Professional Documents
Culture Documents
Roelof Vos
Saeed Farokhi
Introduction
to Transonic
Aerodynamics
Fluid Mechanics and Its Applications
Volume 110
Series editor
André Thess, German Aerospace Center, Institute of Engineering
Thermodynamics, Stuttgart, Germany
Founding Editor
René Moreau, Ecole Nationale Supérieure d’Hydraulique de Grenoble,
Saint Martin d’Hères Cedex, France
Aims and Scope of the Series
The purpose of this series is to focus on subjects in which fluid mechanics plays a
fundamental role.
As well as the more traditional applications of aeronautics, hydraulics, heat and
mass transfer etc., books will be published dealing with topics which are currently
in a state of rapid development, such as turbulence, suspensions and multiphase
fluids, super and hypersonic flows and numerical modeling techniques.
It is a widely held view that it is the interdisciplinary subjects that will receive
intense scientific attention, bringing them to the forefront of technological
advancement. Fluids have the ability to transport matter and its properties as well
as to transmit force, therefore fluid mechanics is a subject that is particularly open
to cross fertilization with other sciences and disciplines of engineering. The
subject of fluid mechanics will be highly relevant in domains such as chemical,
metallurgical, biological and ecological engineering. This series is particularly
open to such new multidisciplinary domains.
The median level of presentation is the first year graduate student. Some texts
are monographs defining the current state of a field; others are accessible to final
year undergraduates; but essentially the emphasis is on readability and clarity.
Introduction to Transonic
Aerodynamics
123
Roelof Vos Saeed Farokhi
Faculty of Aerospace Engineering Department of Aerospace Engineering
Delft University of Technology The University of Kansas
Delft Lawrence, KS
The Netherlands USA
Every effort has been made to contact the copyright holders of the figures and tables which have been
reproduced from other sources. Anyone who has not been properly credited is requested to contact the
publishers, so that due acknowledgment may be made in subsequent editions.
vii
viii Preface
of each subject. The book contains 60 examples and more than 200 practice
problems.
This textbook is intended primarily for senior undergraduate students or graduate
students with prior knowledge in aerodynamics. Although we repeat the funda-
mental equations and flow characteristics in the beginning of the book, we do
assume that the student has had a course on subsonic aerodynamics and is familiar
with its fundamentals. Even though knowledge of transonic aerodynamics is
important for many internal flow applications (e.g., turbo machinery, engine
intakes, exhausts, etc.) the present textbook primarily focuses on external aerody-
namics with limited applications to internal flows. Examples are targeted mainly
toward wings and bodies exposed to a transonic flow field. Many of the examples
reference real aircraft or wing components. Therefore, a strong connection is
present between the content of this textbook and the subject of transonic aircraft
design. To understand why a modern high-subsonic aircraft is designed the way it
is, requires one to understand the subject matter of this textbook.
Acknowledgments
Writing this book has been a privilege. Naturally, we would not have been able to
do so without the support of our respective universities. We therefore would like to
thank the University of Kansas and Delft University of Technology for providing us
with the time and resources to write this book. We would also like to thank the
publisher, who, based on a 135-page summary, trusted that we would extend it to its
current form. We would also like to acknowledge Dr. Luca Guadani and Dr. Ali
Elham who proofread various chapters and performed some of the numerical cal-
culations to support the examples in this book. A special thank you to emeritus
professor Egbert Torenbeek for providing us with meticulous feedback on the
content of various chapters. Finally, we would like to thank the following persons
on both sides of the Atlantic, who over the course of multiple years, aided in the
preparation of this document: Mr. Thomas Statsny, Ms. Lisanne van Veen,
Mr. Maarten Broekhuizen, Mr. Kevin Haagen, Ms. Maaike Weerdesteyn, and
Mr. Amool Raina.
ix
x Contents
Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 551
Chapter 1
Introduction and Historic Perspective
Abstract Airliners and jet fighters are often designed for Mach numbers close to
the speed of sound. At those speeds the flow domain is termed transonic and con-
tains both subsonic and supersonic flow. Typical aerodynamic characteristics in this
domain are shock waves, drag divergence, and transonic buffet. To suppress these
adverse effects or postpone them to higher Mach numbers, multiple design measures
can be taken. A supercritical airfoil can be designed that postpones the onset of
shock waves. In addition, wing sweep can be introduced or the airplane can be area
ruled to further reduce the adverse transonic effects. This chapter presents some of
these practical measures and their effect. It also discusses the definition of transonic
aerodynamics and why it is beneficial to fly at Mach numbers where transonic flow
exits. Furthermore, it demonstrates the challenges associated with simulating tran-
sonic flow and transonic windtunnel experiments. Finally, an outline of the present
textbook is given.
On October 14th, 1947 a new era of flight was entered when Chuck Yeager officially
became the first person to surpass the speed of sound in fully controlled flight. The
rocket-powered Bell X-1 (Fig. 1.1) flew at a Mach number of 1.06, thereby proving
that the sound barrier was in reality not as impenetrable as sceptics had assumed.
This event was the culmination of research efforts over the preceding decades that
attempted to unveil the secrets of flow near Mach one, i.e., transonic flow. Von
Kármán first coined the word transonic to mark this peculiar regime in flight where
a mixture of subsonic and supersonic flow co-exists.
The very first appreciation of transonic flow and its effects on the drag coefficient
dates back to 1910 when the drag coefficient was shown to increase for Mach numbers
close to unity [2]. At that time, transonic and supersonic flow was encountered only by
ballistic projectiles such as ammunition. It is therefore no surprise that the fuselage of
the first supersonic aircraft was shaped like a machine-gun bullet. The first production
aircraft with an air-breathing engine that achieved supersonic speed in a dive was
the XF-86-Sabre, although it was never intended to fly that fast. Only the second
Fig. 1.1 Bell X-1, the first aircraft to officially fly supersonic (Photo USAF)
generation of jet-powered fighter aircraft were designed for (low) supersonic speeds
and featured typical design characteristics (such as an area-ruled fuselage) to reduce
their peak drag coefficient as the airplane surpassed Mach 1.
It took more than 25 years after the first supersonic flight for the first transport
aircraft to fly supersonic. The Russian Tupolev 144 and the Anglo-French Concorde
were the first aircraft that cruised beyond Mach 2 and were thought to initiate a new
era in aviation: supersonic flight [3]. As is generally known now, there never was
a true supersonic era. Concorde flew between 1976 and 2003, but when it retired
there was no other supersonic aircraft to replace it. Environmental issues (supersonic
boom) as well as fuel cost had limited the feasibility of the aircraft. Instead, high-
subsonic jet transport aircraft have dominated the civil aviation market for the past
60 years and their number increased rapidly over the 1970s and 80s. Being relatively
fuel efficient, fast and quiet (no supersonic boom), they have proven to be the best
solution for past and present-day airlines.
Jet transports fly in the lower transonic regime, also known as the high-subsonic
regime. It is therefore of interest to look more closely to the flow phenomena that
are taking place at these speeds and how this effects the design and operation of the
aircraft. To be able to justify a particular aircraft design we need to have a strong
knowledge about the flow phenomena and how we can influence them to meet the
requirements set by the costumers while satisfying the constraints of the aviation
authorities. One of the general goals of aerodynamic design is to maximize the
range parameter (Mach number times lift-to-drag ratio, M × L/D). According to
the Bréguet range equation, the maximum cruise range (Rmax ) for a given thrust-to-
weight ratio is found when the product of Mach number (M) and lift-to-drag ratio
L/D) is maximized [17]:
L a Wbegin cruise
Rmax = M × ln (1.1)
D max SFC Wend cruise
where SFC is the specific fuel consumption, a the ambient speed of sound, and W
the gross weight of the aircraft. Assume that we can find an airplane shape that
1.1 From Subsonic to Supersonic Flight 3
25
Long-Range Jet Aircraft
20
00
00
7-4
00
7-3
-30
74
7-1
74
B-
10
74
B-
7
15
70
DC
B-
B-
0
-30
30
A3
-11
-8
R
P
MD
0
DC
00
IM
7E
00
10
L-1 -10-1
ER
1-5
1-1
76
-30
00
01
B-
01
DC
-10
-6
L-1
00
DC
A3
Fig. 1.2 Aerodynamic efficiency (M × L/D) variation with date of entry into service (after Ref. [5])
4 1 Introduction and Historic Perspective
40
Mdd Douglas definition C-130H
727
20
10
0
0.4 0.5 0.6 0.7 0.8 0.9 1.0
Mach Number, M (-)
Fig. 1.3 Typical compressibility drag behavior and definition of the drag-divergence Mach number
(modified from Ref. [16])
to have a weak shock terminating this supersonic region such that the wave drag is
minimized. It was already shown in the 1960s that by careful design of the airfoil’s
upper surface it is possible to decelerate supersonic flow from as high as M = 1.4
isentropically without having a shock, thereby minimizing the wave drag [14].
shockwave
sonic line
Fig. 1.4 The basic classes of external transonic flows (after Ref. [13])
conditions. When the free-stream Mach number is even further increased, the shock
waves on either side of the airfoil start to move aft as more of the body is immersed
in supersonic flow. As the freestream Mach number exceeds unity, a detached bow
shock forms in front of the body. Only part of the flow behind the bow shock is
decelerated to subsonic conditions. The flow behind the oblique branches of this
bow shock remains supersonic downstream. When the freestream Mach number is
increased even more, the subsonic domain becomes smaller and can be deemed to be
‘sensibly’ empty. The Mach number at which this occurs marks the boundary between
transonic and supersonic flow. Even though there is no physical discontinuity when
the flow transitions between supersonic and transonic, the upper freestream Mach
number for the flow to be termed transonic is generally set at 1.3 [6].
One of the main challenges that was associated with the flight of the Bell X-1 was
the fact that there was no valid wind tunnel data available for speeds between Mach
0.95 and Mach 1.1. The reason for this lack of transonic wind tunnel data stems from
choking of the test section at a freestream Mach number of 0.95. Until the mid 1940s
it was therefore impossible to generate accurate wind tunnel results [1]. The choking
phenomenon had hindered transonic wind tunnel tests at high Mach numbers during
the 1930s and 40s and needed to be resolved if accurate data was to be obtained. On
the other hand, open wind tunnels proved to be unsuitable for transonic flow due to
its high susceptibility to atmospheric disturbances.
6 1 Introduction and Historic Perspective
A break-through came in the late 1940s (a few years after the Bell X-1 had made
its maiden flight). Slotted test sections were introduced, where thin, rectangular slots
were made such that the displaced airflow could move out of the tunnel into a so-
called plenum chamber that surrounded the test section. By doing this, the blockage
problem was essentially solved and the flows could be accelerated through Mach 1
without any problems. These slotted tunnels were essentially the first transonic wind
tunnels and became operational in the early 1950s. Figure 1.5 shows the 16-foot
transonic wind tunnel that is operated by NASA Langley. Later it was found that the
improvement is even greater when the walls are not slotted but perforated [21].
Apart from wind-tunnel blockage, transonic wind tunnel tests faced another prob-
lem: reflected shock waves from the walls. This was particularly an issue for transonic
flows because the shock on the airfoil was close-to normal to the surface of the body.
Therefore, the reflected shock wave interfered with the flow over the body, rather
than with the flow downstream of the body as would be the case for oblique shock
waves [15]. The requirement on the wind tunnel walls was therefore to effectively
cancel the reflected shock waves by producing streamlines downstream of the shock
that corresponded to the flow downstream of the shock. In this respect, porous walls
proved to be superior to slotted walls in which the higher pressure in the viscous
slots extended upstream of the shocks thereby distorting the flow.
In addition to the problem of shock-wave reflection, the growth of the boundary
layer on the walls of the test section proved to be a source of inaccuracy. The thicken-
ing of the boundary layer was caused by small wind-tunnel-wall irregularities. It was
found that a slightly converging wind tunnel wall resulted in a much thinner bound-
ary layer compared to a parallel or slightly diverging wall. The wall inclination was
therefore a parameter that could be changed to reduce the boundary-layer thickness
growth [7]. This problem was targeted in the 1970s. One transonic wind tunnel
incorporated theoretical correction models that used input from a series of test runs.
Fig. 1.5 The present-day slotted-wall test section of the NASA Langley 16-foot transonic tunnel [4]
1.3 Transonic Wind Tunnel Experiments 7
It was then calculated how the walls needed to be oriented such that the wall inter-
ference was minimized and the most accurate results were obtained [22].
Finally, it was found that considerable turbulence formed in the wind tunnel that
modified the boundary layer on the test article in an unpredictable way. Even though
the walls were positioned such that their influence was minimized, this phenomenon
had a considerable influence on the accuracy of the aerodynamic measurements. It
was found that the source of the turbulence came from acoustic disturbances from the
slots or perforations. A practical approach yielded a solution to this problem. Simple
splitter plates in each of the holes diminished the acoustic waves for the perforated
walls. Analogously, a thin wire mesh that covered the slots had the same effect in
case of slotted walls.
Wind tunnel tests are conducted on scale models of aerodynamic bodies that
resemble the true model very accurately. All the conditions in the tunnel are set such
as to match realistic conditions as well as possible. However, the Reynolds number
effect is one characteristic that is hard to simulate when scaling down to wind tunnel
sizes. Since the Reynolds number is dominant in the determination of friction drag,
boundary layer thickness, transition point, and shock position, it is evident that testing
at lower Reynolds number can have severe impact on the predicted aerodynamic
performance of the test object.
For that reason, the idea arose in the early 1970s to modify the test gas such as to
match the Reynolds number encountered by the wind tunnel model to the Reynolds
number of the full-scale body. By increasing the density of the test gas the Reynolds
number increased linearly. To increase the density, three different measures were
taken: the air in the tunnel was replaced with a heavier gas (e.g., carbondioxide), the
gas was cooled by injecting liquid nitrogen into the flow, and the air was pressurized.
Combining these measures resulted in a Reynolds number which could be twenty
times as high as for conventional wind tunnels [8].
Combine all these efforts with the measures that were described in the previous
paragraphs and it becomes clear that transonic cryogenic wind tunnels are among the
most complex aerodynamic facilities to test aerodynamic bodies. Currently, there are
only three large scale facilities in the world: one in the United States (National Tran-
sonic Facility, NTF) and two in Cologne, Germany (European Transonic Windtunnel,
ETW, and Kryo Kanal Köln). The ETW can cool the flow to 110 K in combination
with pressures up to 450 kPa and a Mach number of 1.3 [8]. Advantages of this
tunnel are that compressibility effects, friction effects and deformation effects can
be studied independently by varying Mach number, dynamic pressure and Reynolds
number independently of one another. A photograph taken in the test section of the
ETW is presented in Fig. 1.6.
8 1 Introduction and Historic Perspective
Fig. 1.6 ETW wind tunnel with aircraft model (Photo ETW; printed with permission)
In parallel to the development of the transonic wind tunnel, a series of aircraft design
modifications was investigated to reduce the drag that occurred at high-subsonic
conditions. This included modifications to airfoil shapes, the planform layout of
wings, and the shape of nonlifting bodies such as the fuselage. Both in Europe and
in the United States, great efforts were made to postpone the onset of the sharp drag
rise and to reduce the drag peak that was encountered at high-subsonic conditions.
To appreciate wing aerodynamics in transonic flow, we examine the flow patterns
on airfoils in this regime. In Fig. 1.7 the four stages of transonic flow about an airfoil
are shown. In part (a) the lower subsonic case shows the appearance of a supersonic
bubble that is terminated in a shock on the suction surface of a generic airfoil. The
pressure surface remains entirely subsonic. In part (b) the upper subsonic case shows
the growth of the supersonic bubble on the entire suction surface as well as the
appearance of supersonic flow and a shock on the pressure surface. In part (c) the
freestream is sonic and except for a small subsonic zone near the stagnation point of
the airfoil, the upper and lower surfaces are entirely supersonic with a distinct trailing
edge shock system as shown. The case of slightly supersonic freestream is shown
in part (d). The formation of the bow shock ahead of the airfoil creates a subsonic
region, followed by a supersonic zone on the upper and lower surfaces of the airfoil.
Again the terminal shocks are at the trailing edge, as shown. Note that Fig. 1.7 shows
the Mach distributions (instead of pressure distributions) on the airfoil.
To postpone the onset of a strong shock wave, so-called “supercritical” airfoils
were developed. Supercritical airfoils were developed by Pearcey [14] at the National
1.4 Transonic Aerodynamics of Wings and Bodies 9
M >1 M >1
Upper
Surface
M M M M
1 1 1 1
Lower
Surface
Fig. 1.7 Four flow patterns and corresponding Mach distributions representative of transonic flow
over an airfoil (after Ref. [12]). a M∞ < 1, b M∞ < 1, c M∞ = 1, d M∞ > 1
Physical Laboratory (UK) and Whitcomb [23] at NASA in the early 1960s. The
pressure distributions on a “conventional” (NACA 6-series) airfoil and a supercritical
airfoil are shown in Fig. 1.8. The strong shock that abruptly transitions the supersonic
flow to subsonic on the suction surface of the conventional airfoil (in transonic Mach
number) is replaced by a relatively weak shock that appears further aft on the airfoil.
The corresponding pressure distributions show the difference between a smooth
transition and an abrupt transition to subsonic flow on the two airfoils. We also note
that the supercritical airfoil has a concave curvature near the trailing edge on its
pressure surface. This curvature reversal on the aft pressure surface causes a higher
loading on the airfoil and thus a larger contribution to lift.
We have superimposed a critical pressure coefficient line on the data of Ref.
[11] in Fig. 1.9 to see (or separate) the supersonic zone on the airfoil. The pressure
distribution on a supercritical airfoil at Mach 0.8 with a lift coefficient of 0.54 and a
chord Reynolds number of three million is shown. Noteworthy features are significant
aft loading on the airfoil, as well as a relatively smooth compression on the suction
surface at the end of supersonic bubble. A footprint of flow separation, i.e., a flat C p
profile downstream of the shock, on the suction surface is visible in the data. We also
note that the pressure surface never reaches sonic flow. This pressure distribution is
typical for airfoils that were developed for high subsonic Mach numbers.
Another measure to postpone the drag divergence is the application of swept
wings. The first incarnations of swept-back wings date back to the early 1900s,
when John William Dunne (an Irish aeronautical engineer) designed and built various
tailless biplanes with aft-swept wings. In Fig. 1.10 the Dunne D.8 is shown taking off
from the airfield at the Farnborough air show. Dunne applied wing sweep to shift the
neutral point (i.e., the aerodynamic center of the aircraft) further aft, such that it would
lay behind the airplane’s center of gravity. This ensured a positive static stability
margin, which is an important prerequisite for having acceptable flying qualities for
manually controlled aircraft. Many other tailless aircraft relied on wing sweep to
10 1 Introduction and Historic Perspective
M>1
(b) (d)
Cp
Cp
(+) (+)
Fig. 1.8 Pressure distribution on a conventional (NACA 6-series) airfoil and a supercritical airfoil
(after Ref. [23]). a Supersonic domain over NACA 6-series airfoil. b Pressure distribution over
NACA 6-series airfoil. c Supersonic domain over supercritical airfoil. d Pressure distribution over
supercritical airfoil
Weak shock
Separation (footprint)
-1.0
Supersonic -0.8 Reattachment
-0.6
-0.4 C p, crit = - 0.435
-0.2
Subsonic
Cp 0
0.2
0.4
0.6
0.8
0.0 0.2 0.4 0.6 0.8 1.0
x
c
Fig. 1.9 Pressure distribution on a supercritical airfoil at Mach 0.80, cl = 0.54, Rec = 3.0 × 106
(Flagged symbols are for pressure surface, data after Ref. [11])
1.4 Transonic Aerodynamics of Wings and Bodies 11
mence. After the war, the concept of forward-swept wings was almost completely
abandoned due to the high weight penalty that was introduced due to adverse aeroelas-
tic effects. The aft-swept wing, on the other hand, was embraced by both military
and civil airplane manufacturers in Europe and the United States. It was Boeing that
first introduced aft-swept wings on a large bomber aircraft. This aircraft, the B-47
Stratojet (Fig. 1.11), also featured podded jet engines that were suspended below
the shoulder-mounted wing. This configuration with swept-back wings and engines
mounted in nacelles below the wing is still the dominant aircraft configuration for
high-subsonic airliners.
With the flight of the X-1 in 1947, the supersonic domain had opened up for
military aircraft. However, the jet engines at the time were not yet powerful enough
to overcome the drag rise around Mach one. Therefore, first generation of jet fighters
flew at high subsonic speeds. New specifications from the US Airforce in 1950
dictated an aircraft that could fly with a top speed of Mach 1.2. It was Convair that was
awarded the contract to design and build a prototype aircraft. The resulting aircraft
featured a delta wing and no horizontal tail plane. Test flights showed disappointing
results: the airplane could only achieve M = 0.98, far below the requirement. A major
redesign was initiated to reshape the fuselage such that the peak drag coefficient
around Mach one reduced. This reshaping, commonly referred to as area ruling,
proved to be very effective. The new prototype was able to fly beyond the speed of
sound and met the top-speed requirement of Mach 1.2. This airplane (designated
F-102A) became quite successful with 1,000 copies built. In Fig. 1.12, a comparison
between the first prototype and the second prototype of the aircraft is shown.
The area-rule was initially proposed by Richard Whitcomb as a ‘rule of thumb’
as he called it. This rule of thumb essentially stated that the wave drag of an arbitrary
body at sonic conditions is solely dependent on the cross-sectional area distribution
of that body in the direction of the free stream. He first proposed this rule in 1952
Fig. 1.12 Area ruling applied to Convair F-102 (Photos NASA). a YF-102 prior to area ruling. b
YF-102A with area-ruled fuselage
based on experimental results that he had obtained in the NACA wind tunnels and the
theory of stream pipes that had been introduced the year before by Adolf Büsemann.
However, already during the second world war the effect of cross-sectional area dis-
tribution on transonic drag was acknowledged by Otto Frenzl. Working at Junkers, he
introduced so-called “displacement bodies” that were strategically positioned such
as to reduce the wave drag a high-subsonic conditions. The idea of a favorable cross-
sectional area distribution was to be implemented in 1945 by Waldemar Voigt on the
Messerschmidt P112, which was never completed. In 1946, Dietrich Küchemann
introduced the “pinched fuselage,” a reduction of fuselage diameter at the location
of the wing. This modification was later dubbed the Küchemann Coke-bottle design
and is one of the modifications found in the YF102A. A mathematical description of
the ideal cross-sectional area distribution was presented in two separate articles by
Wolfgang Haack (1941) [9] and William Sears (1947) [19]. For a given combination
of length and volume, the Sears-Haack body possesses the cross-sectional area distri-
bution with the least amount of wave drag at Mach one. For speeds beyond the speed
of sound, a different area distribution should be found according to the supersonic
area rule. This rule was first introduced by Wallace Hayes in his PhD dissertation of
1947 [10].
14 1 Introduction and Historic Perspective
The governing equations of fluid flow are the well-known Navier-Stokes equations.
With present-day numerical techniques, it is possible to solve the Navier-Stokes
equations numerically. However, this technique (termed direct numerical simulation
or DNS) requires a very fine numerical grid and therefore is computationally very
expensive. Therefore, to evaluate the global flow field around arbitrary bodies in
transonic flow, we would rather employ equations that are less expensive to solve
numerically. Examples of a reduced set of equations that are often used are the Euler
equations that hold for inviscid fluid. When neglecting rotation in the flow, also the
full potential equation can be employed. Both neglect the viscous and heat-transfer
characteristics of the fluid. However, to predict the aerodynamic properties of a body
in transonic flow, it is important to at least include viscous effects in the thin layer,
i.e., the boundary layer, close to the body.
In capturing aerodynamics of wings in transonic speeds by computational meth-
ods, the shock-boundary layer interaction plays a big role in accurate flow simulation.
Therefore, inviscid flow solutions, in general, are not expected to be accurate when
boundary layers separate. To test this hypothesis, computational results on a finite
(swept) wing at Mach 0.86 are shown in Fig. 1.13. The pressure distributions in five
semi-spanwise stations are examined. Two computational predictions are also shown
at those stations. The inviscid flow solution is based on the full potential equation
and the results of a coupled inviscid-boundary-layer solution are also depicted. Let
us examine the semispanwise position near the wing tip, i.e., η = 0.89, first. Data
indicates the appearance of a moderately strong shock on the suction surface, around
x/c = 0.25. The full-potential solver over predicts the peak suction and the shock
location by a wide margin. However, the coupled solution does a reasonable job
of predicting the peak suction, shock position and subsequent pressure distribution.
We further note that the pressure surface is also better predicted using the inviscid-
boundary layer coupled scheme. The same trend is also observed at other spanwise
stations. The discrepancy between the inviscid and coupled solution strategy gets
smaller as we approach the wing root, e.g., see η = 0.26.
Another set of data compared with inviscid and the coupled-boundary layer com-
putational approach is shown in Fig. 1.14 for a winglet on a wing-body-winglet con-
figuration. Pressure distributions at four sections at the base and along the winglet
span are shown. In this case the inviscid solver is based on Euler momentum equa-
tions, which means that rotational flows are properly simulated. The coupling of
Euler and a boundary layer solver at least promises to include the viscous effects,
albeit in a sequential and iterative way. We note that the agreement with the measure-
ment is best at the base of the winglet, then it becomes weaker and the predictions
deviate more, as we approach the winglet tip. In particular, Euler solutions always
place a shock further downstream than the data indicate and often the peak suction is
over-predicted. We also note that Euler-coupled boundary-layer method does capture
the viscous dominated events on the winglet and thus better predicts the pressure dis-
tribution. The shortcoming of the coupled inviscid-boundary layer approach stems
1.5 Transonic Flow Calculations 15
-1. -1.
-.8 -.8
-.6 -.6
-.4 Cp -.4
Cp -.2 -.2
0 0
M∞ = 0.86 .2 .4 .6 .8 1.
.2 .4 .6 .8 1.
.2 x/c inviscid .2 x/c
coupled with b.l. .4
.4 η = 0.26 η = 0.89
.6 .6
Fig. 1.13 Impact of viscous coupling with inviscid flow solution on a finite wing in transonic flow
(after Ref. [20])
Euler
-0.8 Euler + Coupled
-0.8
Boundary Layer
Cp -0.4 Cp
-0.4
Test
0.0 0.0
0.4 0.4
M∞ = 0.84
α = 2.08
-0.8 -0.8
Cp -0.4 Cp -0.4
0.0 0.0
0.4 0.4
Fig. 1.14 Impact of viscous coupling on pressure distributions on a winglet in transonic flow (after
Ref. [20])
16 1 Introduction and Historic Perspective
from the shortcomings of the boundary layer simulation of complex viscous flows.
Therefore, such problems demand the use of fully viscous solvers that are based
on the (Reynolds-Averaged) Navier-Stokes equations rather than inviscid solutions
coupled with a boundary layer solver.
In the present text, we introduce the reader to quantitative and qualitative aspects of
transonic aerodynamics. The focus lies on external aerodynamics, rather than flows
through ducts, nozzles or turbomachinery. A brief outline of the each of the chapters
follows below.
In Chap. 2, we (re-)familiarize the reader with concepts from mathematics, ther-
modynamics, and aerodynamics that are fundamental to the methods presented in
the remainder of the book. This includes a short review of vector algebra and partial-
differential equations to provide the mathematical insight into the governing equa-
tions of fluid flow. The Navier Stokes equations are presented and it is shown how
the Reynolds-Averaged Navier-Stokes (RANS) equations, the Euler equations, and
the full-potential equation are derived. Using examples, it is shown how well each
of these models can predict the outcome of aerodynamic experiments in transonic
conditions.
Chapter 3 presents the concept of similarity parameters between “affinely” related
bodies. The theory behind transonic similarity parameters is useful in the prediction of
lift, drag, and moment coefficients of affinely related bodies. In addition, this chapter
presents various compressibility corrections for subsonic and supersonic flow based
on the linearized potential-flow equation. The application of the linearized potential-
flow equation to supersonic flow about two-dimensional and axisymmetric bodies is
demonstrated through a number of examples.
As transonic flow is characterized by the formation of expansion waves, normal
shocks, bow shocks, and/or oblique shock waves, Chap. 4 presents the theoretical
basis for the characterization of these waves. It is shown how the flow characteristics
change over a (normal) shock wave and how this differs from a fan of expansion
waves. It is shown that shock-expansion theory is a powerful tool in analyzing aero-
dynamic forces and moments on pointed bodies in steady supersonic flow where the
oblique shock(s) at the leading edge are attached. This method produces exact results
within the confines of inviscid flow theory. Various examples demonstrate how this
method is used and what its limitations are.
In supersonic flow with large perturbations, a unique set of lines (called character-
istics) can be defined where the hyperbolic partial differential equation that describe
the flow properties reduce to an ordinary differential equation. On these lines, we
can therefore integrate and analytically solve these equations. In Chap. 5 this method
of characteristics is presented. Two approaches to the method-of-characteristics are
demonstrated: the wave-field method and the lattice-point method. Examples of both
1.6 Outline of Present Textbook 17
methods include the supersonic flow through nozzles and the supersonic exhaust
plume of an under-expanded jet.
Chapter 6 presents the sources of drag that are encountered by a nonlifting body
that is moving at transonic Mach numbers. A qualitative characterization of laminar
and turbulent boundary layers is presented along with the concept of transition. In
addition, calculation methods are presented to estimate the friction drag of bodies
subjected to a pressure distribution. Furthermore, a method for the calculation of the
pressure drag over axisymmetric bodies is presented based on the linear potential
flow equation. It is shown how the pressure drag is a function of the cross-sectional
area distribution of the body. The concept of area ruling is presented and examples
are shown of practical implementations of this knowledge.
In Chap. 7 we present the transonic aerodynamics about lift-producing airfoils. It
is shown how various design parameters influence the velocity distribution and the
shock formation over an airfoil. The development of supercritical airfoils is explained
from a historic perspective. The concept of shock-boundary layer interaction, related
shock stall, and shock oscillation are also explained. Furthermore, it is demonstrated
that transonic flow limitations can play a major role in the maximum lift coefficient
of (multi-element) airfoils at Mach numbers as low as 0.25. Examples demonstrate
how low-speed and high-speed stall limit the flight envelope of aircraft that travel at
transonic Mach numbers.
In the final chapter (Chap. 8) we discuss the aerodynamics of swept wings in
transonic flow. It is shown through simple-sweep theory why a swept wing allows for
a higher drag-divergence Mach number and what its effect is on the drag coefficient
in the transonic regime. Apart from these advantages the chapter also presents the
challenges that aerodynamic designers face to reduce the adverse effects of wing
sweep: tip stall, early formation of tip shocks, form drag at the wing root, an early
onset of boundary-layer transition, aerodynamic twist, and aileron reversal. Apart
from aft-swept wings, also the merits and drawbacks of forward-swept wings are
presented.
Each of the chapters includes a set of examples and a set of problems to practice
with the theoretical concepts that are presented. We hope that through the addition
of these examples and problems, the reader can familiarize himself or herself with
important aspects that are related to transonic flow. Answers to selected problems
are given in the back of this book (p. 527 and onwards). A comprehensive list of
symbols and nomenclature is also added (starting at p. 543) such that the reader can
check the meaning and units (all in SI) of the variables that are used throughout the
chapters. Each chapter also lists a variety of references for further reading. Naturally,
if the reader is interested to learn more about particular topics that are presented in
this book we encourage him/her to consult the original sources. They often contain
much more detail about the particular topic at hand and provide a valuable link to
more material on the subject.
18 1 Introduction and Historic Perspective
Problems
1.1 Consider an aircraft with a maximum take-off weight (MTOW) of 560 metric
tons. Assume that at the beginning of the cruise phase this aircraft has burned some
fuel and weighs 96 % of its maximum take-off weight. At the end of the cruise phase
assume that the aircraft weighs only 75 % of its MTOW.
(a) For a cruise altitude of 35,000 ft (10.67 km) calculate the speed of sound, a, using
a temperature lapse rate of −6.5 K/km, a sea-level temperature of 288 K, and a
gas constant of 287 J/kg/K.
(b) For a cruise Mach number of 0.85 and a specific fuel consumption of SFC =
0.000186 [N/N/s] calculate the range, R, when L/D = 16, L/D = 19, and
L/D = 22. Use the Bréguet range formula (1.1).
References
1. Anderson, J.: Modern Compressible Flow With Historic Perspective, 3rd edn. McGraw Hill,
New York (2003)
2. Bensberg, W., Cranz, C.: Uber eine photographische Mehode Zur Messung von
Geschwindigkeiten und Geschwindigskeitverlusten bei Infanteriegeschossen. Artillerische
Monatshefte 41, 333–346 (1910)
3. Burgess, E.H.: Concorde inaugurates the supersonic era. In: Proceedings of the 9th Annual
Meeting and Technical Display. AIAA-1973-16, Washington, DC (1973)
4. Cornelliussen, S.T.: The transonic wind tunnel and the NACA technical culture, Chap. 4. From
Engineering Science to Big Science, pp. 91–133. NASA, Washington (1998)
References 19
5. Denning, R., Armstrong, J.A., Allen, J.E.: The broad delta airliner. Aeronaut. J. 107(1075),
547–558 (2003)
6. Ferrari, C., Tricomi, F.G.: Transonic Aerodynamics. Academic Press, New York (1968)
7. Goethert, B.H.: Transonic Wind Tunnel Testing. No. 49 in AGARDograph. Pergamon Press,
Oxford (1961)
8. Green, J., Quest, J.: A short history of the European Transonic Wind tunnel ETW. Prog. Aerosp.
Sci. 47, 319–368 (2011)
9. Haack, W.: Geschossformen kleinsten Wellenwiderstandes. Bericht 139 der Lilienthal-
Gesellschaft, pp. 14–28 (1941)
10. Hayes, W.D.: Linearized Supersonic Flow. Ph.D. thesis, California Institute of Technology
(1947)
11. Hurley, F.X., Spaid, F.W., Roos, F.W., Stivers, L., Bandettini, A.: Detailed Transonic Flowfield
Measurements about a Supercritical Airfoil Section, NASA TMX-3244 (1975)
12. Kuechemann, D.: The Aerodynamic Design of Aircraft. Pergamon Press, Oxford (1978)
13. Moulden, T.H.: Fundamentals of Transonic Flow. Wiley, New York (1984)
14. Pearcey, H.H.: The aerodynamic design of section shapes for swept wings. Adv. Aeronaut. Sci.
3, 277–322 (1963)
15. Ramm, H.J.: Fluid Dynamics for The Study of Transonic Flow. Oxford University Press,
New York (1990)
16. Roskam, J.: Airplane Design, Part 6: Preliminary Calculation of Aerodynamic. Thrust and
Power Characteristics. DARcorp, Lawrence (2006)
17. Ruijgrok, G.J.J.: Elements of Airplane Performance. Delft University Press, Delft (1989)
18. Scholz, D., Ciornei, S.: Mach number, relative thickness, sweep and lift coefficient of the
wing—an empirical investigation of parameters and equations. DGLR Jahrbuch 2005 (2005)
19. Sears, W.R.: Projectiles of minimum wave drag. Q. Appl. Math. 4(4), 361–366 (1947)
20. Tinoco, E.N.: CFD applications to complex configurations: a survey. In: Henne, P. (ed.) Applied
Computational Aerodynamics, Chap. 15. AIAA, Washington (1990)
21. Tokaty, G.A.: Fluid Dynamics. W & J Mackay & Co Ltd., Chatham (1971)
22. Vidal, R.J.: Wall Interference Effects in Transonic Flows. Calspan Corp., Buffalo (1976)
23. Whitcomb, R.T., Clark, L.R.: An Airfoil Shape for Efficient Flight at Supercritical Mach
Numbers, NASA TM X-1109. Langley (1965)
Chapter 2
Review of Fundamental Equations
2.1 Introduction
Transonic flow conditions are encountered by the majority of today’s jet aircraft.
Almost all major airlines use high-subsonic jet transports on their medium to long
haul flights. In addition, business jets, fighter aircraft, and military UAVs such as the
Northrop X-47 are also confronted with transonic effects. To comprehend the physics
of transonic flow, this chapter presents a review of physical and mathematical topics
that form a basis for the subsequent chapters of the book. For further reading on these
topics, the reader is referred to the reference list at the end of this chapter, that lists a
number of text books that adequately explain the subject matter in more detail. After
this short introductory section, we will present basic reviews on two mathematical
topics: partial differential equations and vector algebra. Since mathematics is often
referred to as “the language of physics,” a good understanding is mandatory for the
As the reader might remember from a course on elementary aerodynamics, the equa-
tions of motion of a particle in a fluid flow can be represented as a set of partial
differential equations (PDEs) called the Navier-Stokes equations. Before we discuss
these relatively complex equations we consider two more elementary equations: the
one-dimensional heat and wave equations. These equations serve as examples of par-
tial differential equations and how they can be solved. At the same time they are also
part of a unique set of PDEs that actually have a closed-form solution. Unfortunately
this is not true for most PDEs that describe physical processes such as the motion of
a fluid. To solve these PDEs we need different solution techniques which often rely
on a (numerical) approximation of the problem. However, to test such a numerical
scheme we can always use the one-dimensional heat or wave equation because we
know their exact solution. They can therefore serve as a test case for our numerical
schemes to gain confidence in the accuracy of the approximation method.
In the subsequent two subsections, we do not attempt to review the broad subject
of PDEs and their solution methods, but merely present a basic review of the topic
within the context of the theory of aerodynamics and computational fluid dynamics.
In this section we derive the governing equation that describes the transverse vibration
of an elastic string, such as a guitar string. When solving this governing equation we
should end up with an expression that describes each individual point on the string in
both space (x, u) and time (t). First, consider a string of length L that we perturb and
at time t = 0 we release the string such that it starts to vibrate. We want to evaluate
its vertical displacement, u(x, t) as shown in Fig. 2.1. To simplify this problem we
assume that the mass per unit length (ρ) of the string is constant along the string and
that it does not have any bending stiffness. In addition, we neglect the gravitational
force and assume that the deflections remain small enough to justify the assumption
that each point on the string only moves in the vertical direction.
To come up with the equation of motion for this string, we consider the force
balance on an infinitesimal part within the string itself (P–Q in Fig. 2.1). We have
α and β denoting the deviation angles from the horizontal axis, and T1 and T2 being
the components of the internal forces at points P and Q, respectively. The horizontal
components of the forces at points P and Q must be balanced:
T1
0 x x+Δx L
Directing our attention to the vertical direction we know that the following force and
acceleration balance should hold over the part Δx:
∂2u
− T1 sin α + T2 sin β = ρΔx (2.2)
∂t 2
Using (2.1) we can rewrite (2.2) as follows:
ρΔx ∂ 2 u
tan β − tan α = (2.3)
T ∂t 2
Since tan α = ∂u/∂x|x and tan β = ∂u/∂x|x+Δx we can now write a partial
differential equation where both derivatives with respect to x and t appear:
1 ∂u ∂u ρ ∂2u
− = (2.4)
Δx ∂x x+Δx ∂x x T ∂t 2
∂2u 2∂ u
2 T
= c with c2 ≡ (2.5)
∂t 2 ∂x 2 ρ
The wave equation is a second order (=the highest power of one of the differential
terms), linear partial differential equation. Not only does this equation describe the
transverse motion of a vibrating string it also describes the wave motion of a plane
wave (e.g. acoustic wave) [16]. In that case we replace the excitation of √ the string by
the sound pressure, p (x, t) and c becomes the speed of sound, c = γ RT , where
γ is the ratio of specific heats, R is the gas constant, and T is the static temperature
in the calorically perfect gas.1
We can now proceed with a solution procedure of the wave equation. There are
several approaches to derive a closed form solution of the problem, one of which
is by using D’Alembert’s approach. This approach relies on a transformation to the
characteristic coordinates ξ and η according to:
ξ = x + ct, η = x − ct (2.6)
∂2u
u ξη = =0 (2.7)
∂η∂ξ
If we assume that two arbitrary functions, f (x) and g(x) describe the initial
position and velocity, respectively, we can express the functions φ(x) and ψ(x) in
terms of the these initial conditions. We have:
If we integrate both sides of (2.11) between x0 and x and divide both sides by c
we obtain:
1x
φ(x) − ψ(x) = g(s)ds with k(x0 ) = φ(x0 ) − ψ(x0 ). (2.12)
c x0
Now, to get φ(x) we simply add (2.12) to (2.9) and divide both sides by two. Similarly,
to obtain ψ(x) we subtract (2.12) from (2.9) and divide both sides by two. The
following results are found:
1 1 x 1
φ(x) = f (x) + g(s)ds + k(x0 ) (2.13)
2 2c 0
x 2
1 1 x 1
ψ(x) = f (x) − g(s)ds − k(x0 ) (2.14)
2 2c x0 2
Substituting φ(x + ct) and ψ(x − ct) in the above expressions and inserting every-
thing back into (2.8) results in the following solution to this initial value problem:
1 1 x+ct
u(x, t) = [ f (x + ct) + f (x − ct)] + g(s)ds. (2.15)
2 2c x−ct
For a brief moment, let us return to interpretation of this result with respect to
the original problem of the vibrating string and assume that the initial velocity,
u t (x, 0) = g(x) = 0. To come up with a valid solution that satisfies the boundary
conditions (u(0) = u(L) = 0), the following should hold:
1
u(0, t) = [ f (ct) + f (−ct)] = 0 (2.16)
2
1
u(L , t) = [ f (L + ct) + f (L − ct)] = 0 (2.17)
2
From (2.16) we learn that the initial shape function f (x) is odd. In other words it
is anti-symmetric with respect to the u-axis. Combining this with (2.17) we obtain
that f (L + ct) = f (−L + ct), which shows that the initial function should have a
period of 2L.
Example 2.1 Calculate the solution to the homogeneous wave equation (2.5) for
u(x, 0) = f (x) = sin(2πx), u t (x, 0) = g(x) = 0, and c = 1.
Solution:
Using D’Alembert’s solution we can directly substitute f (x) = sin(πx) and g(s) =
0 in to (2.15). We obtain:
A graphical representation of this string is sketched over the interval [0, 1] between
time t = 0 and t = 0.25 with time increments Δt = 0.05 (Fig. 2.2).
26 2 Review of Fundamental Equations
u (x,t)
0
−0.5
−1
0 0.2 0.4 0.6 0.8 1
x
where f (x, t) is the disturbance function. We assume that the string is initially
undisturbed, i.e.:
u(x, 0) = u t (x, 0) = 0 (2.19)
Similarly, we can transform the forcing function: f (x, t) = G(ξ, η). Note that when
t = 0 we have ξ = η and ξ = x. Therefore, we can write the first initial condition:
To transform the second initial condition we write the first partial derivative of v with
respect to ξ in terms of u x and u t by employing the chain rule (see Problem 2.5):
1 1
vξ = u x + ut (2.22)
2 2c
Substituting the equalities that hold when t = 0 (same as above) we have:
1 1
vξ (ξ, ξ) = u x (ξ, 0) + u t (ξ, 0) (2.23)
2 2c
2.2 Review of Partial Differential Equations 27
Now, we know that the second part of the right-hand side of the equation is zero
due to the initial condition u t (x, 0) = 0. And, since u(x, 0) = 0 we know that
1/2u(x, 0)|x=ξ must also be zero. Therefore, we now have the second initial condition
in transformed coordinates:
vξ (ξ, ξ) = 0 (2.24)
1
vξη = − G(ξ, η) (2.25)
4c2
We are interested in the excitation at a particular time and location, u(x0 , t0 ), which
corresponds to v(ξ0 , η0 ). Therefore, we need to integrate the left-hand side of (2.25)
twice with appropriate boundary conditions:
ξ0 ξ
vξη (ξ, η)dηdξ
η0 η0
Solving the inner integral by using the fundamental theorem of calculus (FTC)2
results in: ξ
vξη (ξ, η)dη = vξ (ξ, ξ) − vξ (ξ, η0 )
η0
We note that the first term on the right-hand side must equal zero, due to the initial
condition (2.24). The second integration step thus results in:
ξ0
− vξ (ξ, η0 )dξ = v(η0 , η0 ) − v(ξ0 , η0 )
η0
Again, due to the initial condition (2.21) for ξ = η0 , the first term on the right-hand
side equals zero, which leaves us with the negative of what we were initially looking
for: −v(ξ0 , η0 ). Now, by performing the same integration on the right-hand side
of (2.25) and bringing the minus sign to the other side of the equality sign (which
cancels) we have a solution in characteristic coordinates:
1 ξ0 ξ
v(ξ0 , η0 ) = G(ξ, η)dηdξ (2.26)
4c2 η0 η0
To change back to the original coordinate system, the determinant (for explanation
of the determinant see Sect. 2.3) of the Jacobian, det(J ), needs to be calculated, since
dηdξ = det(J )dxdt. It is left to the reader (Problem 2.6) to show that this equals 2c.To
2The fundamental theorem of calculus specifies the relationship between the two central operations
of calculus: differentiation and integration [11].
28 2 Review of Fundamental Equations
(b)
(a)
t (x 0 ,t 0 )
η
ξ0 (ξ 0 ,ξ0)
η0 (ξ 0 ,η0)
(η0,η0)
ξ x
η0 ξ0 x0 - ct0 x0+ct0
Fig. 2.3 Transformation of the area of integration between the characteristic and the original coor-
dinate system. a Region of integration in the (ξ, η)-plane. b Region of integration in the (x, t)-plane
transfer the limits of the integrals it is convenient to look at the region over which we
are integrating G(ξ, η). This region is displayed in Fig. 2.3a. If we look at the vertices
of this triangle, and what they represent in the physical (x, t)-plane, we note that:
(ξ0 , η0 ) → (x0 , t0 )
(ξ0 , ξ0 ) → (x0 + ct0 , 0)
(η0 , η0 ) → (x0 − ct0 , 0)
The region of integration in the (x, t)-plane that corresponds to these vertices is
shown in Fig. 2.3b. We therefore have:
1 t0 x0 +ct0
u(x0 , t0 ) = f (x, t)dxdt (2.27)
2c 0 x0 −ct0
The region of integration in the physical domain can be used to see whether a
point in the domain is influenced by a particular disturbance. When the area under
the graph of the disturbance function and the area of the integration region (partially)
overlap, then the point in the domain is influenced by the disturbance. However,
if there is no overlap at all, the disturbance function does not affect that particular
point in the domain. The characteristic lines (ξ = x + ct and η = x − ct) therefore
mark the boundaries of the so-called domain of dependence. The solution, therefore,
only depends on a disturbance that happens within this domain of dependence. Any
disturbance outside this region does not have any effect on the excitation of this point.
In Example 2.2 this is demonstrated.
2.2 Review of Partial Differential Equations 29
(x,t)
x
-1 0 x-t 1 x+t
Calculate u(x, t) explicitly for (x, t) satisfying: x > t, x < 1, and x + t > 1.
Solution:
To solve this problem we use (2.27) with c = 1. Due to the simple forcing function,
f (x, t), we can directly see that the solution u(x, t) = 1/2 Δ(x,t) f (x, t)dxdt,
where Δ(x, t) represents the region of integration. This region is schematically shown
in Fig. 2.4. Using simple geometry we calculate the area of this region Δ(x, t) =
t2 x2
2 − 2 − xt − x − t − 2 . The solution then becomes:
1
1 2
u(x, t) = (t − x 2 − 2xt + 2x + 2t − 1)
4
In this section we look at the other famous PDE, the heat equation. In the one-
dimensional case we consider a bar with length L (Fig. 2.5). We are interested in
the temperature, T , and its distribution over the bar with time, depending on the
boundary and initial conditions. We assume that the bar is of constant cross section
0 x=L
Fig. 2.5 Model problem for the one-dimensional heat equation: a bar of length L
30 2 Review of Fundamental Equations
and homogeneous material and that it is perfectly insulated, such that heat flows only
along the x-direction.
The derivation of the heat equation starts with Fourier’s law of heat conduction
that states that the heat flux, q, is negatively proportional to the temperature gradient,
∂T /∂x, according to (in one-dimensional space):
where k is the thermal conductivity of the material. The negative sign indicates
the direction of heat flow is from hot to cold, i.e. in the opposite direction to the
temperature gradient. In the absence of work done, a change in internal energy per
unit volume in the material, ΔE, is proportional to the change in temperature, ΔT .
That is,
ΔE = cρΔT (2.29)
where c is the specific heat capacity and ρ is the mass density of the material. Choosing
zero energy at the absolute zero temperature, this can be rewritten as
E = cρT. (2.30)
x − Δx ≤ ξ ≤ x + Δx (2.31)
is given by
x+Δx t+Δt x+Δx ∂T
cρ [T (ξ, t + Δt) − T (ξ, t − Δt)] dξ = cρ dξdτ , (2.33)
x−Δx t−Δt x−Δx ∂τ
where the FTC was used. Additionally, with no work done and the absence of any
heat sources or sinks, the change in internal energy in the interval [x −Δx, x +Δx]
is accounted for entirely by the flux of heat across the boundaries. By Fourier’s law,
this is
t+Δt ∂T ∂T
t+Δt x+Δx ∂ 2 T
k (x + Δx, τ ) − (x − Δx, τ ) dτ = k dξdτ
t−Δt ∂x ∂x t−Δt x−Δx ∂ξ 2
(2.34)
again by the fundamental theorem of calculus. In higher dimensions, the divergence
theorem (Sect. 2.3.3) is used instead. By conservation of energy (Sect. 2.5.3),
t+Δt x+Δx
[cρTτ − kTξξ ] dξdτ = 0. (2.35)
t−Δt x−Δx
2.2 Review of Partial Differential Equations 31
This is true for any rectangle [t − Δt, t + Δt] × [x − Δx, x + Δx]. Consequently,
the integrand must vanish identically:
which is the heat equation. The coefficient k/(cρ) is called thermal diffusivity and
is denoted by α for convenience.
To solve the heat equation (2.38), we use the ‘separation-of-variables’ tech-
nique [18] where we substitute
Denoting ∂/∂t with (. ˙. .) and ∂/∂x with (. . .) , we can write (2.38) as:
Ġ F
= = −β 2 . (2.40)
αG F
F + β 2 F = 0 (2.41)
Ġ + αβ G = 0
2
(2.42)
Equation (2.41) is second order in space and Eq. (2.42) is first order in time.
To find the eigenfunctions of the heat equation we need a set of boundary equations
that give us some details about the heat transfer conditions of the rod at the ends.
In the present case we choose to keep the ends of the rods at zero temperature:
T (0, t) = T (L , t) = 0. A general solution to (2.41) is:
√
Let us now return to (2.42) and substitute the eigenvalue λn = βn α:
Ġ + λ2n G = 0 (2.45)
where Cn are arbitrary constants. This results in the following final solution of this
problem:
nπx
e−λn t for n = 1, 2, . . .
2
Tn (x, t) = Fn (x)G n (t) = Dn sin (2.47)
L
where Dn are arbitrary constants. These are the eigenfunctions of the problem and
λn = cnπ
L are the eigenvalues.
In order to solve the heat equation we need an initial condition, f (x). This initial
condition is prescribing the temperature distribution along the bar at time t = 0
must comply with the boundary conditions. We know that the sum of each of the
solutions in (2.47) is also a solution to (2.38) and we consider the following series
of eigenfunctions:
∞
nπx
e−λn t
2
T (x, t) = Dn sin (2.48)
L
n=1
Since at t = 0 we need to satisfy the initial condition the following relation must be
satisfied:
∞ nπx
T (x, 0) = Dn sin =
ˆ f (x) (2.49)
L
n=1
Multiplying both sides by sin nπxL and integrating both sides over the interval [0, L]
results in the following coefficients of the Fourier sine series:
2 L nπx
Dn = f (x) sin dx (2.50)
L 0 L
Example 2.3 Solve the heat equation, Tt = Tx x on the interval 0 < x < 1 with
boundary conditions T (0, t) = Tx (1, t) = 0 and initial condition T (x, 0) = 1.
Solution:
Using the same methodology as above we separate the temperature function in the
product of F(x) and G(t). We have
Again, we state that due to the first boundary condition A = 0. The second boundary
condition implies: Bβ cos(βx) = 0. This equation is satisfied when
1
β=π n+ for n = 0, 1, 2, . . .
2
∞
2
1 −π 2 n+ 21 t
T (x, t) = Dn sin πx n + e (2.51)
2
n=0
1
Setting T (x, 0) = 1 and knowing that 0 sin π n + 21 x sin π m + 21 x dx = 0
for n = m we multiply both sides of (2.51) with sin π m + 21 x and integrate
between 0 and 1:
1
1 1 2
sin πx m + dx = Dm → Dm =
0 2 2 π m + 21
This gives the following final solution (switching back from m to n for convenience):
∞
2
2 1 −π 2 n+ 21 t
T (x, t) =
sin πx n + e (2.52)
n=0
π n + 21 2
In Fig. 2.6, we see a graphical interpretation of the solution. Note how the initial
condition is satisfied by showing almost a complete horizontal line at t = 0. The
boundary condition at x = 0 is also satisfied, while we can see that the slope of the
line at x = 1 is zero, indicating that we have also satisfied this boundary condition
correctly.
The equations of motion of fluid flow that are presented in this chapter (and through-
out the text) are in conservation form (also referred to as conservative form). This
implies that the partial differential equations (PDEs) that describe the physics of the
flow have coefficients that are either constant or, if variable, their derivatives do not
appear in the equation. We use the following example to illustrate this form.
34 2 Review of Fundamental Equations
0.6 0.2
T (x,t)
0.4 0.4
0.6
0.2 0.8
1.0
0
0 0.2 0.4 0.6 0.8 1
x
Example 2.4 To appreciate the difference between the conservative and nonconser-
vative form of a partial differential equation we consider the one-dimensional heat
equation (2.38) without the assumption that ρ, c, or k are constant with x. In other
words, the thermal diffusivity, α is a function of x: α = α(x). First this equation is
formulated in conservative form:
∂T ∂ ∂T
= α (2.53)
∂t ∂x ∂x
∂T ∂2 T ∂α ∂T
=α 2 + (2.54)
∂t ∂x ∂x ∂x
Note how in (2.53) the coefficient, α, can vary with position (x) but its derivative does
not appear in the equation. Therefore, (2.53) is in conservation form. In (2.54) the
derivative on the RHS of (2.53) has been expanded and now contains the derivative
term ∂α/∂x, which is a non-conservative term in the equation. Therefore we deem
(2.54) to be in non-conservative form.
The same logic that is used for the one-dimensional heat equation can be expanded
to the equations of motion that describe fluid flows (Sect. 2.5). These governing
equations must hold at any distinct point in the flow. To formulate these equations
we take a Eulerian approach where we consider a fixed control volume through which
the fluid passes. The conservation form for each PDE at such a point also allows us
to formulate finite-difference representations that provide a good approximation to
the PDEs. These, in turn, can be used in an iterative numerical code to obtain the
global properties of a flow field. The conservative formulation is therefore often used
to solve partial differential equations by numerical methods.
2.2 Review of Partial Differential Equations 35
A second order partial differential equation (PDE) can be classified based on the
coefficient of its second-order derivative terms. For an arbitrary function, φ, a second-
order differential equation in two dimensions reads:
where a, b, and c are coefficients and H = H (φ, φx , φ y , x, y). Along the so-called
characteristic curves of the PDE the following should hold:
Example 2.6 Find the characteristic curves for the wave equation of Example 2.5.
Solution:
For (2.5) we have dt/dx|1,2 = ± χ1 . Integrating this gives:
1
t1 = x +η
χ
1
t2 = − x + ξ,
χ
where, η and ξ are integration constants. Rearranging these equations results in the
relations of (2.6).
In this arbitrary (orthogonal) coordinate system, the hyperbolic, parabolic, and
elliptic PDEs assume a particular canonical (natural) form. Let us examine these
canonical forms for each class of second order PDEs. We assume that the solution
u(x, t) is written in the transformed coordinates as φ(ξ, η).
Hyperbolic PDE Two characteristic forms exist:
domain of
dependence
domain of
dependence x x x
Fig. 2.7 Domain of dependence and region of influence for the three different classes of second
order PDEs. a Hyperbolic. b Parabolic. c Elliptic
2.2 Review of Partial Differential Equations 37
φξξ = h 3 (φξ , φη , φ, ξ, η) or
φηη = h 4 (φξ , φη , φ, ξ, η)
Similar to hyperbolic PDEs, the parabolic PDEs have a region of influence. How-
ever, this region is now unbounded by characteristics and therefore spans the entire
space (can also be time) beyond the point where the initial disturbance takes place
(see Fig. 2.7b). In addition, the entire region before this point influences this point
in the flow. Parabolic PDEs describe steady boundary layer equations along with
“parabolized” viscous flows.
Elliptic PDE The canonical form of an elliptic PDE can be written in its charac-
teristic coordinates as follows:
Any disturbance in a flow field that is described by elliptic PDEs is felt throughout
the entire flow. In other words, the entire flow domain forms the region of influence.
Conversely, every point in the flow domain influences any other point in the flow
domain. Therefore, the entire flow can also be viewed as the domain of dependence
(see Fig. 2.7c). Elliptic PDEs describe the steady subsonic, inviscid flow along
with an incompressible inviscid flow field.
Consider the two-dimensional potential equation for steady, irrotational, inviscid
flow (we will derive this equation in Sect. 2.7.2):
Φx2 Φ y2 Φx Φ y
1 − 2 Φx x + 1 − 2 Φ yy − 2 2 = 0 (2.58)
c0 c0 c0
The velocity potential function is denoted with Φ(x, y), with the velocities, Φx (x, y)
and Φ y (x, y), as well as the speed of sound, c0 . The three coefficients of interest can
be identified as follows:
Φx2 2Φx Φ y Φ y2
a =1− b=− c =1− (2.59)
c02 c02 c02
Φx2 + Φ y2 − c02
b2 − 4ac = = M 2 − 1, (2.60)
c02
where M is the Mach number of the flow. For subsonic flows M < 1 and the
potential equation is elliptic. When the flow is supersonic M > 1 and the equation
38 2 Review of Fundamental Equations
As demonstrated in Sect. 2.5 the equations of motion of a fluid flow can be conve-
niently represented in vector form. The vector notation is often used to represent
the governing flow equations and their (numerical) approximations. Vector notation
allows us to represent a set of equations in a single line, which make it easier to
perform mathematical operations in order to derive certain numerical approximation
schemes. The following sections give a review of basic vector algebra including a
review of the most common operations and their notation used in this text.
Some properties of a fluid flow are represented as scalar values (such as p, T , and
ρ). Scalars have a magnitude but no direction. Velocity, however, is represented by
a vector, because it has both a magnitude and a direction. In this text a vector is
denoted with a bold font. For example, the velocity vector is denoted with V , while
its magnitude,
V = |V |. In two dimensions, the velocity vector is denoted with
u
V = , where u and v are the magnitude of the velocity vector components in
v
the two orthogonal directions.
Addition and subtraction of vectors (and also matrices) is the same as for scalars.
Each entry of the first vector (matrix) is added or subtracted to the corresponding
entry in the second vector (matrix). This does require the two vectors (matrices) to
be of identical dimension. We distinguish different forms of vector multiplication.
For two vectors, a and b the inner product (also called dot product or scalar product)
is defined as:
2.3 Review of Vector Algebra 39
⎛ ⎞
b1
⎜ .. ⎟
n
a · b = (a1 · · · an ) ⎝ . ⎠ = ai bi (2.61)
bn i=1
a × b = ab sin θn = v (2.62)
where θ is the measure of the smaller angle between a and b (0◦ ≤ θ ≤ 180◦ ), a and
b are the vector magnitudes of vectors a and b, and n is a unit vector perpendicular
to the plane containing a and b in the direction given by the right-hand rule. If the
vectors a and b are parallel (i.e., the angle θ between them is either 0◦ or 180◦ ), by
the above formula, the cross product of a and b is the zero vector 0. In the present
text the cross product in the three-dimensional Euclidian space is considered and
the vectors a and b therefore each have three entries. This allows us to calculate the
entries for v using the third-order determinant:
i j k
v = a1 a2 a3 = (a2 b3 − a3 b2 )i + (a3 b1 − a1 b3 )j + (a1 b2 − a2 b1 )k, (2.63)
b1 b2 b3
where i, j, and k are the unit vectors in the respective directions x, y, and z.
Example 2.7 For the vectors a = (1, 2, 3) and b = (4, 5, 6), calculate the inner
product, the outer product and the vector product of a with b and b with a.
Solution:
The inner product is calculated according to (2.61):
a · b = 1 · 4 + 2 · 5 + 3 · 6 = 32
40 2 Review of Fundamental Equations
a × b = (2 · 6 − 3 · 5, 3 · 4 − 1 · 6, 1 · 5 − 2 · 1) = (−3, 6, −3)
b × a = −(a × b) = (3, −6, 3)
In vector calculus we have two kinds of functions: scalar functions and vector func-
tions. The output of a scalar function at a particular point P is a scalar:
f = f (P)
The output of the vector function is a vector. The domain where a vector function is
defined is a region of space (can also be a surface or a curve in space). Within this
region (or on that surface or line) we say that this vector function defines a vector
field. In Cartesian coordinates a vector field can be denoted by
The following sections present various operations that can be performed on vector
and scalar fields. One of these operations is the gradient of a scalar field. The gradient
of a scalar field is a vector field which points in the direction of the greatest rate of
increase of the scalar field, and whose magnitude is the greatest rate of change.
The gradient (grad) of a scalar field is defined as (in three-dimensional Cartesian
coordinates):
∂f ∂f ∂f
grad f = ∇ f = , , (2.64)
∂x ∂ y ∂z
The ∇ (read: nabla or del) operator is defined as the following differential vector
operator:
2.3 Review of Vector Algebra 41
∂ ∂ ∂
∇≡ i+ j+ k (2.65)
∂x ∂y ∂z
We can define a scalar function whose gradient forms the three components of the
velocity field for irrotational, inviscid flow. This so-called potential-flow equation
reduces the system of equations to a single equation, thereby simplifying the problem
of finding the velocity field considerably. For example, the velocity distribution in a
fluid flow can be represented by the vector field V. At any point, P, throughout the
fluid this vector field can be related to the gradient of a scalar field, Φ, at this point:
V(P) = ∇Φ(P). Since this is true throughout the entire physical domain we define
the velocity potential function as:
V ≡ ∇Φ (2.66)
Example 2.8 Consider a fluid flow where the velocity vector field is given by V =
(2x, yz 2 , zy 2 ). Determine the velocity potential function that describes this flow, if
it exists.
Solution:
To find Φ(x, y, z) find the primitive function of each of the components of the vector
field:
Φ = 2xdx = x 2 + C1 (y, z)
y2 z2
Φ = yz 2 dy = + C2 (x, z)
2
z2 y2
Φ = zy 2 dz = + C3 (x, y)
2
It is possible to combine the above equations into a single expression for the following
potential flow function: Φ = x 2 + 21 y 2 z 2 + C4 , where C4 is a constant. When we
employ (2.66) to Φ, we find V . This confirms that Φ exists and is indeed the potential
function of V .
To assess the magnitude of ‘change’ of a vector field we use the divergence (div) of
a vector field defined as:
∂g1 ∂g2 ∂g3
div g = ∇ · g = + + (2.67)
∂x ∂y ∂z
where g = (g1 , g2 , g3 ).
42 2 Review of Fundamental Equations
Example 2.9 For the vector field of Example 2.8 calculate the divergence.
Solution:
Employing (2.67) we obtain the following:
div V = ∇ · V = 2 + z 2 + y 2
To study the rotation of a vector field we employ the curl. The curl is defined as:
i j k
curl g = ∇ × g = ∂/∂x ∂/∂ y ∂/∂z (2.68)
g1 g2 g3
∂g3 ∂g2 ∂g1 ∂g3 ∂g2 ∂g1
= − i+ − j+ − k
∂y ∂z ∂z ∂x ∂x ∂y
In a velocity field, the curl is called the vorticity vector and is related to the rotation
of the flow. Rotation of the flow is introduced due to the viscosity of the gas and leads
to turbulent conditions. We will see later that the rotation of a flow field is directly
tied to the formation of curved shock waves in supersonic flow via Crocco’s theorem.
Example 2.10 Determine the curl of the vector field of Example 2.8.
Solution:
We can directly apply (2.68):
In Example 2.10 we see that the specified vector field is irrotational. This agrees with
the fact that we can define a potential equation. This confirms that we can define a
potential function only when the vector field is irrotational.
In this section we briefly review the relation between volume, surface, and line
integrals. First, we consider an enclosed volume, V , with an outer surface S. The
divergence theorem (usually attributed to Gauss) states that the flux of a vector field
2.3 Review of Vector Algebra 43
through the surface equals the divergence of that vector field inside the enclosed
volume. For a velocity vector field, V , this theorem can be written as follows3 :
V · dS = ∇ · V dV (2.69)
S V
The gradient theorem states that the gradient of a scalar field, p, integrated over
the control volume, equals the scalar field integrated over the surface vector of the
control volume:
pdS = ∇ pdV (2.70)
S V
where dS is the element of the surface vector in the normal direction to the surface
pointing outward. Finally, we can relate the curl of a vector field over a control surface,
S, to the closed line integral of the vector field in a counter-clockwise direction over
its boundary, C. This is the Kelvin-Stokes theorem (often referred to as Stokes’
theorem):
V · dC = (∇ × V ) · dS (2.71)
C S
Thermodynamics is the science that studies the energy conversion between mechan-
ical work and heat. In fluid flow, thermodynamics plays an important role in trans-
ferring energy from one state to the other. A simple example is a stagnation point at
the leading edge of a wing, where the flow is brought to rest and its kinetic energy is
transferred to heat. Similar effects arise in the boundary layer over the wing where
the friction between the air and the structure also converts kinetic energy into heat.
As we will see in Sect. 2.5 the first law of thermodynamics forms the basis of the
energy equation in fluid flow. This section presents the thermodynamic parameters
that we find in a fluid flow and their interrelation via the state law, the first law of
thermodynamics, and the second law of thermodynamics. It is emphasized that this
section is merely a review of the thermodynamic relations that we will use in sub-
sequent sections and chapters. For a more complete and in-depth discussion of the
topic the reader may consult Sonntag and Van Wylen [21]. For more information on
the application of thermodynamics to aircraft propulsion, the reader is referred to
Farokhi [9].
3 This theorem applies to any vector field, not only V but also mass flow or force fields.
44 2 Review of Fundamental Equations
In fluid flow, the thermodynamic properties of fluid are characterized by scalar fields.
The variables of interest for the present discussion are pressure ( p in Pa or psi), density
(ρ in kg/m3 or lb/ft3 ), temperature (T in K or ◦ F), and the specific internal energy
(e in J/kg or BTU/lb). According to the state principle of thermodynamics, the local
thermodynamic state is determined by any two independent state variables. If e and
ρ are chosen as those variables, both pressure, p, and temperature, T , are dependent
on these two variables. The subsequent paragraphs present the equations that relate
the state variables to each other via a set of laws.
For a calorically perfect gas (intermolecular forces are assumed negligible and the
volume of individual molecules is infinitesimally small) the equation of state reads:
p = ρRT, (2.72)
where R is the gas constant for a specific gas. In terms of specific volume, v = 1/ρ,
(2.72) can also be expressed as pv = RT . A new thermodynamic state variable is
defined as the sum of internal energy, e, and the product of pressure and specific
volume and is called specific enthalpy:
h = e + pv = e + RT (2.73)
Even though specific enthalpy is introduced here without any physical motivation,
we will see later that it is indeed a fundamental state variable for aero-thermodynamic
analysis. Both specific energy and specific enthalpy (which also has unit J/kg) are
both related to the temperature for a perfect gas:
e = e(T ) (2.74)
h = h(T ) (2.75)
In transonic flow we assume that the gas is perfect and we can consequently find
de = cv dT (2.76)
dh = c p dT (2.77)
where cv is the specific heat at constant volume and c p the specific heat at constant
pressure. These coefficients can be assumed to be constant for air up to temperatures
of 1,000 K, which is the basis for the calorically perfect gas assumption. Based on
this assumption we can now define the following relations between the specific heat
coefficients, cv and c p , the specific internal energy and enthalpy, e and h, the gas
constant, R, and the ratio of specific heats, γ:
2.4 Review of Thermodynamics 45
cp R γR
e = cv T, h = c p T, γ = , cv = , cp = , R = c p − cv
cv γ−1 γ−1
(2.78)
Using the above relations, the temperature and pressure can now be defined in terms
of the independent state variables e and ρ:
(γ − 1)e
p = (γ − 1)ρe, T = (2.79)
R
For dry air at temperatures that are normally encountered during flight, the gas
constant equals R = 287.04 J/kg/K.
The viscosity (μ) and thermal conductivity (k) are properties of a gas that are
both dependent on the temperature. The viscosity of the fluid is responsible for the
momentum transport on molecular level. The dynamic viscosity of an ideal gas is:
1
μ= ρλc (2.80)
3
Here λ is the mean free path between molecules and c is the average velocity of the
molecules. Examining this equation we can see that the viscosity increases with the
average velocity and therefore with temperature. Sutherland’s equation states that the
viscosity of air is dependent on its temperature according to the following relation:
3/2
μ T T0 + S
= (2.81)
μ0 T0 T +S
where μ0 and T0 are a reference viscosity and temperature, respectively, and S is the
Sutherland temperature which is S = 110 K for air. For ISA (international standard
atmosphere) conditions these are μ0 = 1.7894 × 10−5 kg/m/s and T0 = 288.16 K.
The thermal heat conductivity was already introduced in Fourier’s law (2.28) and is a
property of the gas. For air k = 0.024 W/m/K at a temperature of 273 K. The Prandtl
number relates the viscosity and the thermal conductivity of a fluid according to:
cp
Pr = μ (2.82)
k
Since k, μ, and c p are dependent on the temperature, the Prandtl number is dependent
on temperature as well. However, for air over a substantial temperature range (up to
600 K) we can assume the Prandtl number remains constant at 0.71 [3]. In Fig. 2.8
the viscosity and conductivity of air are presented as a function of temperature, along
with the assumed values for the constants that have been used. We see that there is
an appreciable change in conductivity and viscosity with temperature, which can be
important if we model high-speed fluid flows.
46 2 Review of Fundamental Equations
40
Viscosity, µ (mg/m/s)
25
35
20 30
25
15
20
10 15
200 300 400 500 600 200 300 400 500 600
Temperature, T (K) Temperature, T (K)
Fig. 2.8 Graphical representation of (2.81) and (2.82) for dry air
The first law of thermodynamics states that energy can only be transformed from one
form into another. In other words, the summation of all energies in a closed system
remains constant. The system has hypothetical boundaries to the surroundings that
we can define based on the problem under consideration. The change in specific
internal energy of the system, de, can be related to the increment in specific work
done by the system, δw, and the specific heat added to the system, δ Q:
de + δw = δ Q (2.83)
Contrary to that, the heat and work depend on the path between the two states.
Therefore, their changes are represented with a δ in (2.83). To demonstrate this
process dependency Fig. 2.9 shows how the total amount of work changes with the
path taken from point 1 to point 2. Note that this figure shows expansion processes
where pV n = constant. This is a polytropic process, with p and V representing
the pressure and volume of the gas, respectively. The exponent, n, depends on the
process. For example if n = 0 we have an isobaric process, when n = 1 we have an
isothermal process (using the perfect gas law). Of course, we can define an infinite
number of ways to get from state 1 to state 2, each described by a different value of n.
If work is done by a system under adiabatic conditions (meaning no heat addition
or extraction, hence δ Q = 0) the internal energy (read: temperature) of the system
2.4 Review of Thermodynamics 47
B
A
2
Area = reversible work
for process A
For a process that is reversible (no dissipative phenomena occur), we can calculate
the specific work (w) according to:
2 p2 v 2 − p1 v 1
for n = 1 (non-isothermal condition)
w= pdv = 1−n (2.86)
1 p1 v1 ln vv21 for n = 1 (isothermal condition)
Substituting the perfect gas law for pv in (2.86), we get the following expression for
non-isothermal processes:
R(T2 − T1 )
w=− (2.87)
n−1
Since we assume that the change in energy equals the work done on the system and
using (2.76), we can write:
Δe = cv (T2 − T1 ) = −w (2.88)
Combining (2.87) and (2.88) leads, after simplification, to the following statement:
cp
n= =γ (2.89)
cv
which demonstrates that for an adiabatic and reversible process the following is true:
Let us recap the result of this process. We have made two assumptions in the
previous derivation: (1) no heat addition or extraction; (2) no dissipative phenomena
occur. Finally, we implicitly assumed that the total amount of mass within the system
48 2 Review of Fundamental Equations
The first law of thermodynamics only states that energy is conserved during a thermo-
dynamic process. However, it does not tell us anything about the direction in which
the energy flow is taking place. This is governed by the second law of thermody-
namics that states that over time, differences in temperature, pressure, and chemical
potential tend to even out in a physical system that is isolated from the outside world.
To quantify this ‘evening-out’ process, a property called entropy, S (denoted by s for
specific entropy), is introduced. Entropy is related to the change in heat in a reversible
system:
δ Q rev
ds = , (2.91)
T
where T is the temperature of the system. Entropy is a state property meaning that
the change in entropy of a system going from one state to another is the same for all
processes. Note that for an irreversible process where dissipative phenomena occur,
we could always assign an effective δ Q rev that relates the initial state and the end state
to each other. However, it is more revealing to say that these dissipative processes
produce their own entropy, dsirrev , i.e.:
δ Q rev
ds = dsrev + dsirrev = + dsirrev (2.92)
T
This statement says that ds has two parts, one is reversible and the other is irreversible
due to dissipation phenomena. The dissipative phenomena within the system always
increase the entropy of the system:
dsirrev ≥ 0 (2.93)
δQ
ds ≥ (2.94)
T
ds ≥ 0 for adiabatic processes (2.95)
When a process is isentropic (literally meaning “equal disorder”) the entropy of the
system and its surroundings remains constant from state 1 to state 2.
Combining the first and second law of thermodynamics, (2.83) and (2.91), respec-
tively, produces Gibbs’ equation:
T ds = de + pdv (2.96)
If we use the chain rule we can find the change in enthalpy (2.73) to be dh =
de + pdv + vd p. Combining this with (2.96) we can now write two alternative forms
for the first law of thermodynamics, one in terms of specific energy and one in terms
of specific enthalpy:
T ds = de + pdv (2.97)
T ds = dh − vd p (2.98)
By inserting the relations between temperature and specific energy (2.76) and specific
enthalpy (2.77), respectively, and by utilizing the state law (2.72) we can find two
expressions for ds: one in terms of specific volume and one in terms of specific
pressure (see Problem 2.22). Integrating both relations between the state properties
at state 1 and state 2 results in the following expressions for the change in entropy
of a calorically perfect gas:
T2 v2
s2 − s1 = cv ln + R ln (2.99)
T1 v1
T2 p2
s2 − s1 = c p ln − R ln (2.100)
T1 p1
Solution:
We calculate the pressure ratio to be p2 / p1 = 1.512 and enter this in (2.100) to find
s2 − s1 = 0.746 J/kg/K. Rewriting (2.99) we get the following:
⎡ ⎤
ρ2 v1 (s1 − s2 ) + cv ln TT21
= = exp ⎣ ⎦
ρ1 v2 R
Based on the relations (2.99) and (2.100) we can investigate the change in state
properties for an isentropic (adiabatic and reversible) process. In this case ds = 0
and we can rewrite (2.99) and (2.100) to show the relation between the pressure,
density and temperature ratios in an isentropic process. We set the LHS of (2.99)
and (2.100) to zero and perform some manipulations to find the following:
−cv /R
v2 T2
= (2.101)
v1 T1
c p /R
p2 T2
= (2.102)
p1 T1
cv 1 cp γ
= and =
R γ−1 R γ−1
With these substitutions and subsequent manipulations (see Problem 2.23) we can
now state the isentropic relations:
γ γ
p2 ρ2 T2 γ−1
= = (2.103)
p1 ρ1 T1
2.4 Review of Thermodynamics 51
Now that we have the isentropic relations we ask ourselves: why are they so
important? We know that in reality a gas flow is viscous and thermally conducting
and therefore dissipative phenomena always occur. However, we also know that these
phenomena are only dominant in small regions in the flow: the boundary layer and
inside the shock waves. Outside of these regions the dissipative phenomena are so
small that they can often be neglected when we want to calculate the local state
properties. However, within these regions the isentropic relations do not hold and
we cannot relate the state properties in this simple way. This becomes evident in the
subsequent section and chapters of this text.
ρV l
Rel = (2.104)
μ
where l is a characteristic length, for example the chord length of an airfoil or the
diameter of a pipe. A second non-dimensional number to characterize a fluid flow is
its Mach number. The Mach number is defined as the ratio of the local flow velocity
to the local speed of sound:
V
M= (2.105)
a
The local speed of sound, a, is a function of the state properties of the fluid. For a
perfect gas the speed of sound is solely dependent on the temperature of the gas:
∂p p
a= = γ = γ RT (2.106)
∂ρ ρ
We consider a fluid flow through a fixed control volume V . We consider the balance
between the mass flowing in and out of the control volume and the time rate of change
52 2 Review of Fundamental Equations
of the mass inside the control volume. We use the empirical law that mass can neither
be destroyed nor created during this process. Mass can only exit the control volume
through its enclosed surface, S. The resulting mass flowing out of the control volume
can be written as follows:
ρV · dS (2.107)
S
where V is the velocity vector. Note that the vector dS has a magnitude of dS and
a direction n, which is perpendicular to the surface of dS. We can express the time
rate of increase in mass inside V using the following volume integral:
∂
ρdV (2.108)
∂t
V
The change in mass within the control volume should be in balance with the net mass
flux through the control surface. This conservation principle results in the following
continuity equation in integral form:
∂
ρV · dS + ρdV = 0 (2.109)
∂t
S V
Equation (2.109) is the conservation of mass in integral form. Applying the diver-
gence theorem to the surface integral combines both terms within one volume integral
(see Problem 2.24). From this integral the conservation form of the continuity equa-
tion can be found. The conservation of mass law in differential form is:
∂ρ
+ ∇ · (ρV ) = 0 (2.110)
∂t
Let us briefly analyze this equation. It is a first order partial differential equation.
At a particular point in space it describes the change in density with time and the
divergence of the product of density and velocity. We can simplify this equation
even further by introducing the substantial derivative operator, D/Dt. The substan-
tial derivative combines the local derivative (∂/∂t) and the so-called convective
derivative (V · ∇), to represent the “total” derivative:
D ∂
= +V ·∇ (2.111)
Dt ∂t
In the case of the density in two Cartesian dimensions it reads:
Dρ ∂ρ ∂ρ ∂ρ ∂ρ
= + (V · ∇)ρ = +u +v (2.112)
Dt ∂t ∂t ∂x ∂y
2.5 Equations of Fluid Motion 53
This equation physically states that the change in density is due to temporal and
spatial variations. Using the definition of the substantial derivative (2.111) we can
rewrite the continuity equation (2.110) in substantial-derivative form:
Dρ
+ ρ(∇ · V ) = 0 (2.113)
Dt
When homogeneous, incompressible flow is considered, the first term on the LHS
can be dropped and only have ∇ · V = 0. However, generally speaking this is only a
valid approximation at Mach numbers below M = 0.3 where density variations are
less than 5 %. In transonic flows with typical Mach numbers in the range of 0.8–1.3
we cannot use the incompressible flow assumption.
d
F= (mV ) (2.114)
dt
Let us again consider a fluid passing through a finite control volume. The forces
on the fluid in this control volume can be divided into forces that are acting on the
fluid (such as gravity) and forces that are acting on the control surface, S (pres-
sure and shear forces). In Fig. 2.10 it is schematically shown how the body force,
pressure and stress vectors act on a control volume in Cartesian coordinates. Note
that τij is the stress on the surface of the control volume with normal vector i in
y y y
p τ22
τ21
τ23 τ12
τ32
τ11
p τ13
x x τ31 x
p τ33
z
f z z
Fig. 2.10 Schematic control volume with body force (left), surface pressure (center) and shear
stresses (right)
54 2 Review of Fundamental Equations
the direction j. The stress components for which i = j are shear stress compo-
nents, while the stress components normal to the surface (i = j) are associated with
the thermodynamic pressure. The thermodynamic pressure can be perceived as the
force exerted on the control volume walls as fluid molecules coincide with it during
their random movement. The latter becomes important only for such effects where
fluid compressibility is essential. Examples would include shock waves and sound
propagation. For incompressible flows the thermodynamic pressure is zero.
If we denote the body forces per unit volume by ρ f we can represent the total
body force by:
body force = ρ f dV (2.115)
V
The most common body force per unit volume is the gravitational force for which
ρ f = ρg. The surface force due to pressure always acts perpendicular to the surface
of the control volume:
surface force due to pressure = − pdS (2.116)
S
Note that the pressure force is negative, since the pressure acts in the opposite direc-
tion to the surface normal vector, n. Finally, the surface force due to friction can be
written as follows:
viscous surface force = τ ij · dS (2.117)
S
where τ ij is the viscous shear stress tensor. In three dimensions this tensor is expressed
as follows: ⎡ ⎤
τ11 τ12 τ13
τ ij = ⎣ τ21 τ22 τ23 ⎦ (2.118)
τ31 τ32 τ33
Now, let us turn our attention to the RHS of (2.114) and look at the time rate of
change of linear momentum. Similar to the components in the continuity equation
we distinguish two contributions to the time rate of change of momentum: first due
to the momentum change with time of the fluid inside the control volume and second
2.5 Equations of Fluid Motion 55
due to momentum entering and leaving the control volume. These components can
be defined as follows:
d ∂
(mV ) = ρV dV + (ρV · dS)V (2.120)
dt ∂t
V S
We can now combine (2.119) and (2.120) to obtain a first version of the fluid-flow
momentum balance in integral form:
∂
ρV dV + (ρV · dS)V = ρ f dV − pdS + τ ij · dS (2.121)
∂t
V S V S S
We can expand the shear stress terms in this equation in terms of state variables
and fluid constants, but let us first evaluate the differential form of the momentum
equation.
By applying the gradient and divergence theorems (Sects. 2.3.2 and 2.3.3, respec-
tively) appropriately to each of the individual terms in (2.121) we can rewrite this
equation only in terms of volume integrals, which allows us to evaluate only the
integrand. This results in the following differential form of the momentum equation:
∂
(ρV ) + ∇ · (ρV V ) = ρ f − ∇ p + ∇ · τ ij (2.122)
∂t
Note that V V is the outer product of V with itself, leading to the second-order
velocity tensor. For Newtonian fluids (stress-rate of strain relation is linear) the
relation between the shear stress tensor and velocity the components can be written
as follows:
∂u i ∂u j ∂u k
τ ij = μ + + δij μ i, j, k = 1, 2, 3 (2.123)
∂x j ∂xi ∂xk
where μ is the dynamic viscosity and μ is second (or bulk) coefficient of viscosity
of the fluid. Furthermore, u i is the velocity component of V in the direction of i and
δij is the Kronecker delta.4 Following Stokes’ hypothesis:
2
μ = − μ (2.124)
3
Therefore, the viscous stress tensor, τ ij , can be written as:
∂u i ∂u j 2 ∂u k
τ ij = μ + − δij (2.125)
∂x j ∂xi 3 ∂xk
1 for i = j
4 The Kronecker delta is defined as follows: δij = .
0 for i = j
56 2 Review of Fundamental Equations
By expanding the LHS of (2.122) using the chain rule and subsequent substitution
of the continuity equation (see Problem 2.25), we can write the momentum equation
in substantial-derivative notation according to:
DV
ρ = ρ f − ∇ p + ∇ · τ ij (2.126)
Dt
Subsequent substitution of (2.123) yields the following set of momentum equations:
DV ∂ ∂u i ∂u j 2 ∂u k
ρ = ρ f − ∇p + μ + − δij μ (2.127)
Dt ∂x j ∂x j ∂xi 3 ∂xk
The first law of thermodynamics (2.83) forms the basis for the energy balance
between a fluid going through a fixed control volume and its surroundings:
de + δw = δ Q (2.83)
Note that (2.83) is the balance between the rate of change in internal (or specific)
energy with the sum of the rate of specific heat added to the fluid and the rate of
specific work done by the fluid. In order to express the first law in energy per unit
volume we need to multiply (2.83) by the density:
Let us expand each of those terms in terms of state variables of the fluid. We start
with the LHS of (2.128), which is the change of internal energy. Remember, that this
law is written for a gas in stationary condition. Since we are considering a moving
fluid, the kinetic energy per unit mass is to be added. If we denote the total energy
per unit volume with E t we have:
V2
Et = ρ e + (2.129)
2
Since the subject of this text is aerodynamics, we have intentionally omitted a poten-
tial energy term in (2.129). The time rate of change of total energy in the control
volume can now be written as:
2.5 Equations of Fluid Motion 57
∂
time rate of change of total energy inside control volume = E t dV
∂t
V
(2.130)
Since fluid is entering and leaving the control volume, the kinetic and internal energy
changes. Each particle that leaves the control volume has a volume-specific energy
density of E t . The total rate of volume-specific total energy leaving the control
volume therefore equals the volumetric flux over the boundary dS times E t : E t V ·dS.
Integrating over the entire control surface yields the following:
time rate of energy transfer across control surface = E t V · dS (2.131)
S
Now, let us turn our attention to the heat term in (2.128). First, we consider the
time rate of change of heat inside the control volume. Similar to the energy flux, we
define two components: the time rate of change of heat inside the control volume
and the heat flow going in and out of the volume via the control surface. We switch
in nomenclature from specific heat (δ Q, unit: J/kg) to heat per unit volume (ρδ Q,
unit: J/m3 ). The time rate of change of heat inside the control surface is given by:
∂
time rate of change of heat inside control volume = ρQdV (2.132)
∂t
V
By using Fourier’s law of heat conduction (see also (2.28) in Sect. 2.2.2), the heat
flow (q, unit: J/s/m2 ) is linearly related to the temperature gradient:
q = −k∇T (2.133)
where k is the coefficient of thermal conductivity and T is the temperature of the gas.
The heat flow into the control volume is in opposite direction to the control surface
vector dS. Therefore, the rate of heat addition is negatively related to the heat flux
over the control surface dS:
heat flow across control surface = − q · dS (2.134)
S
Having examined the energy and heat flux in the control volume, the last part of
(2.128) is the rate of work being done on the fluid in the control volume. We consider
three different contributions that are related to the three different forces we defined
in Sect. 2.5.2. First we have the time rate of change in work due to the body forces.
We know that work can be defined as the force times a displacement. Hence, the
time rate of change of work can be expressed as the force multiplied by the velocity.
Therefore, we have the following:
time rate of change of work due to body forces = ρ f · V dV (2.135)
V
58 2 Review of Fundamental Equations
Secondly, we have the pressure on the control surface that does work on the fluid
inside the control surface. We take the inner product of the RHS of (2.116) with V
to obtain the work flux due to this component:
time rate of work done due to pressure on the control surface = − pdS · V
S
(2.136)
Finally, we look at the rate of work done due to the shear stress on the control surface.
Similar to the pressure-induced work flux, the work flux due to the shear stress can
be found by taking the inner product of the RHS of (2.117) with V :
time rate of work done due to shear stress on the control surface = (τ ij · dS) · V
S
(2.137)
Remember that τ ij is the stress tensor as defined in (2.118).
We have now defined the time derivative of each of the components of (2.128) and
we can formulate the first version of the energy balance between the control volume
and its surroundings:
∂ ∂
E t dV + E t V · dS = ρQdV − q · dS + ρ f · V dV
∂t ∂t
V S V S V
− pdS · V + (τ ij · dS) · V (2.138)
S S
We can cast (2.138) in differential form by applying the gradient and divergence
theorems to the appropriate terms and changing everything to volumetric integral
formulation. The integrand of that equation yields the following energy balance:
∂ Et ∂(ρQ)
+ ∇ · Et V = − ∇ · q + ρ f · V − ∇( p · V ) + ∇ · (τ ij · V ) (2.139)
∂t ∂t
The first term on the LHS of (2.139) represents the rate of increase of E t in the control
volume, while the second term represents the total energy lost due to convection
through the control volume. The first term on the RHS is the amount of heat produced
by external factors. The second term is the rate of heat lost by means of conduction.
The third term represents the work done by body forces while the last two terms
represent the work done by normal and shear stresses on the surface, respectively.
By subsequently employing the continuity and momentum equation (see Problem
2.26) we can write the energy equation in a convenient substantial-derivative form:
De ∂(ρQ)
ρ + p(∇ · V ) = − ∇ · q + ∇ · (τ ij · V ) − (∇ · τ ij ) · V (2.140)
Dt ∂t
The sum of last two terms in this equation is termed the dissipation function, Φ, and
represent the rate at which mechanical energy is expended due to viscosity when
2.5 Equations of Fluid Motion 59
the fluid is deformed. The LHS of (2.140) can be rewritten in terms of enthalpy
by employing (2.73) and the continuity equation (see Problem 2.27). The following
energy balance results:
Dh D p ∂(ρQ)
ρ = + −∇·q+Φ (2.141)
Dt Dt ∂t
where the dissipation function is:
Φ = ∇ · (τ ij · V ) − (∇ · τ ij ) · V (2.142)
D(e + V 2 /2)
ρ = −∇( p · V ) (2.143)
Dt
Here, we have assumed the body forces, f , to be negligible. Using the continuity
equation (2.113), we can rewrite this equation as follows:
D(h + V 2 /2) ∂p
ρ =− (2.144)
Dt ∂t
In steady conditions, we can therefore derive that:
V2
h+ = H = constant (2.145)
2
where H is the definition of the total enthalpy. If we use the relation between enthalpy
and static temperature (2.77), we can write:
V2
cpT + = c p Tt = constant (2.146)
2
where Tt is the total temperature. For a flow where all the streamlines emanate from
the same uniform freestream, the temperature and velocity are therefore uniquely
correlated through (2.146). We can manipulate (2.146) and include the definition of
the speed of sound (2.106) to obtain:
Tt γ−1 2
=1+ M (2.147)
T 2
60 2 Review of Fundamental Equations
Equation (2.147) states that there exists a unique relationship between the Mach
number and the static temperature in the flow. For a given total temperature, Tt , in
the freestream, Eq. (2.147) shows that the static temperature decreases with Mach
number. If we combine (2.147) with the isentropic relations (2.103) we can express
the temperature, density and pressure as a function of the Mach number. We have
tabulated this relation for Mach numbers ranging from 0 through 10 in Appendix A.
We now make two assumptions about the nature of the flow we consider. The first
assumption is that the body forces (such as gravity) are so small compared to the
surface forces that it is unnecessary to account for them during calculations. In
addition, we assume that there is no change in heat per unit volume due to volumetric
heating. Accordingly, we declare that:
d(ρQ)
f =0 =0 (2.148)
dt
Let us rewrite the three equations that represent the conservation of mass (2.110),
momentum (2.122) and energy (2.139), respectively, and apply the assumptions of
(2.148):
∂ρ
+ ∇ · (ρV ) = 0 (2.110)
∂t
∂
(ρV ) + ∇ · (ρV V ) + ∇ p − ∇ · τ ij = 0 (2.149)
∂t
∂ Et
+ ∇ · E t V + ∇ · q + ∇ p · V − ∇ · (τ ij · V ) = 0 (2.150)
∂t
This shortened form permits us to explicitly rewrite all five equations in their con-
servation form:
∂ρ (∂ρu) ∂(ρv) ∂(ρw)
+ + + =0 (2.151a)
∂t ∂x ∂y ∂z
∂(ρu) ∂(ρu 2 ) ∂(ρuv) ∂(ρuw) ∂ p τx x τx y τx z
+ + + + − + + = 0 (2.151b)
∂t ∂x ∂y ∂z ∂x ∂x ∂y ∂z
∂(ρv) ∂(ρuv) ∂(ρv2 ) ∂(ρvw) ∂ p τx y τ yy τ yz
+ + + + − + + = 0 (2.151c)
∂t ∂x ∂y ∂z ∂y ∂x ∂y ∂z
∂(ρw) ∂(ρuw) ∂(ρvw) ∂(ρw2 ) ∂ p τx z τ yz τzz
+ + + + − + + = 0 (2.151d)
∂t ∂x ∂y ∂z ∂z ∂x ∂y ∂z
2.5 Equations of Fluid Motion 61
∂U ∂E ∂F ∂G
+ + + =0 (2.152)
∂t ∂x ∂y ∂z
where:
⎡ ⎤ ⎡ ⎤
ρ ρu
⎢ ρu ⎥ ⎢ ρu 2 + p − τx x ⎥
⎢ ⎥ ⎢ ⎥
U = ⎢ ρv ⎥ E=⎢ ρuv − τx y ⎥
⎣ρw⎦ ⎣ ρuw − τx z ⎦
Et (E t + p)u − uτx x − vτx y − wτx z + qx
⎡ ⎤ ⎡ ⎤
ρv ρw
⎢ ρuv + p − τx y ⎥ ⎢ ρuw − τx z ⎥
⎢ ⎥ ⎢ ⎥
F=⎢ ρv2 + p − τ yy ⎥ G=⎢ ρvw − τ yz ⎥
⎣ ρvw − τ yz ⎦ ⎣ ρw + p − τ yz
2 ⎦
(E t + p)v − uτx y − vτ yy − wτ yz + q y (E t + p)w − uτx z − vτ yz − wτzz + qz
(2.153)
The first and last row in these vectors represent the continuity and energy equation,
respectively. The three middle rows represent the three components of the momentum
equation. The form above is often used because it is easier to code in numerical form.
It still represents the full equations of motion. Note that this set in the subsequent
text is referred to as the Navier-Stokes (NS) equations.
62 2 Review of Fundamental Equations
u = ū + u v = v̄ + v w = w̄ + w ρ = ρ̄ + ρ (2.154)
p = p̄ + p h = h̄ + h T = T̄ + T H = H̄ + H (2.155)
1 Δt
u = u dt = 0 (2.156)
Δt 0
The equations of motion of Sect. 2.5 are modified by substitution of the flow para-
meters that we have defined in (2.155). Subsequently, each of the equations is time
averaged. The result yields a set of equations that have averaged fluctuating terms.
Each of the terms in the equations of motion that have averaged fluctuating terms
2.6 Reynolds-Averaged Navier-Stokes Equations 63
are subsequently set to zero. This yields a more compact version of these equations,
including the fluctuating terms. We briefly discuss each of the governing equations
below.
When substituting the relations of (2.154) in the continuity equation (2.110) the
following averaged equation can be obtained:
∂ ρ̄ ∂
+ ρ̄ū + ρ u j = 0 (2.157)
∂t ∂x j
For incompressible flow, the fluctuating and unsteady density terms can be dropped
from this equation. Furthermore, ρ̄ = ρ = constant. What we are left with is a much
reduced continuity equation:
∂ ū j
=0 (2.158)
∂x j
Starting from the Navier-Stokes equation (2.127) and neglecting body forces,
the Reynolds-averaged momentum equation can be obtained by substitution of the
relations (2.154) and applying subsequent time averaging:
∂ ∂
ρ̄ū i + ρ u j + ρ̄ū i ū j + ū i ρ u j
∂t ∂x j
∂ p̄ ∂
=− + τ̄ij − ū j ρ u i − ρ̄ j u i u j − ρ u i u j (2.159)
∂xi ∂x j
where
∂ ū i ∂ ū j 2 ∂ ū k
τ̄ij = μ + − δij (2.160)
∂x j ∂xi 3 ∂xk
∂ ∂
∂ p̄ ∂
(ρū i ) + ρū i ū j = − + τ̄ij − ρu i u j (2.161)
∂t ∂x j ∂x j ∂x j
This shows how the Reynolds stresses appear as an addition to the mean flow para-
meters that are very much comparable to the original NS equations. The Reynolds-
averaged stress is related to the fluctuating velocities according to:
τ̄ij turb = −ρu i u j (2.163)
Since the last bracketed term within the square brackets on the RHS of (2.165) is often
small compared to the other terms within the square brackets, it is usually neglected.
What remains is an energy equation comparable to the original energy equation
(2.140) except for the fluctuating term ρu j H . Since this term shares its brackets
with heat flux terms it is often termed the Reynolds heat flux. A similar analysis can
be done for the energy equation where the Reynolds heat flux component is:
∂
− (∇ · q)turb = −ρ̄c p T u j − c p ρ T u j − ū j c p ρ T (2.166)
∂x j
2.6 Reynolds-Averaged Navier-Stokes Equations 65
The Reynolds-averaged equations form a set of five partial differential equations with
more than five unknowns. As was briefly explained, additional fluctuating terms need
to be related to mean flow properties via appropriate turbulence models. We discuss
a typical turbulence model in Sect. 2.6.3.
ρu ρv ρw ρh ρT ρH
ũ = ṽ = w̃ = h̃ = T̃ = H̃ = (2.167)
ρ̄ ρ̄ ρ̄ ρ̄ ρ̄ ρ̄
u = ũ + u v = ṽ + v w = w̃ + w ρ = ρ̄ + ρ
p = p̄ + p h = h̃ + h T = T̃ + T H = H̃ + H (2.168)
where the primed parameters stand for time dependent deviations from the average
Δt
which are zero when integrated over time (i.e. u = Δt
1
0 u dt = 0). The averages
of the doubly primed fluctuating quantities are not zero. Instead the time average
of the doubly primed fluctuations multiplied by the density equals zero (i.e. u =
Δt
Δt 0 ρu dt = 0).
1
∂ ρ̄ ∂
+ ρ̄ũ j = 0 j = 1, 2, 3 (2.169)
∂t ∂x j
∂
∂
∂ p̄ ∂
ρ̄ũ j + ρ̄ũ i ũ j = − + τ̄ij − ρu i u j i, j = 1, 2, 3 (2.170)
∂t ∂x j ∂xi ∂x j
(2.171)
A brief look at the expression reveals a more complex expression than the original,
non-averaged expression of Eq. (2.125). In practice, however, the viscous terms with
doubly primed fluctuations are likely candidates for being neglected based on their
magnitude compared to the mass-averaged variables. By substituting the Reynolds-
averaged continuity equation (2.169) in the Reynolds-averaged momentum equation
(2.170) and employing the substantial derivative yields the following [19]:
D ũ i ∂ p̄ ∂ τ̄ij lam ∂ τ̄ij turb
ρ̄ =− + + (2.172)
Dt ∂xi ∂x j ∂x j
Apparent stress gradients
Mean pressure Laminar-like
Particle acceleration stress gradient due to transport of momentum by
of mean motion gradients turbulent fluctuations and
for the mean motion
deformations attributed to fluctuations
Note that the Reynolds-averaged momentum equation (above) has the same form
as the original Navier-Stokes equation, Eq. (2.127), with the addition of a turbulent
stress term. Explicitly, the laminar and turbulent stress terms are given by:
∂ ũ i ∂ ũ j 2 ∂ ũ k
τ̄ij lam = μ + − δij (2.173)
∂xi ∂x j 3 ∂xk
!
∂u i ∂u j 2 ∂u k
τ̄ij turb = −ρu i u j + μ + − δij (2.174)
∂xi ∂x j 3 ∂xk
To arrive at the Reynolds form of the energy equation, the mass-weighted variables
(2.168) are substituted in (2.140). Subsequent elimination of terms that go to zero
yields the following Reynolds energy equation in mass-weighed variables:
∂ ∂ ∂ T̄ ∂ p̄ ∂
ρ̄ H̃ + ρ̄ũ j H̃ + ρu j H j − k = + ũ i τ̃ij + u i τij
∂t ∂x j ∂x j ∂t ∂x j
(2.175)
A similar analysis can be performed on the energy equation. Apart from the lam-
inar and turbulent stress terms a laminar and turbulent heat flux term can be
defined. The new apparent turbulent stresses and flux terms that appear in both the
Reynolds-averaged momentum and energy equation should be treated as new vari-
ables. Therefore, additional equations are required that make assumptions regarding
the apparent turbulent quantities and the mean flow variables.
2.6 Reynolds-Averaged Navier-Stokes Equations 67
To predict the value of the turbulent kinetic energy, k̄, a transport equation is
developed from the Reynolds-averaged Navier-Stokes equations. Using Boussinesq’s
assumption for eddy viscosity (2.176) this transport equation reads (in substantial-
derivative form):
D k̄ ∂ μT ∂ k̄ 2 ∂u i k̄ 3/2
ρ = μ+ + 2μT Sij − ρk̄δij − CDρ (2.178)
Dt ∂x j Prk ∂x j 3 ∂x j l
A derivation of this equation is beyond the scope of this text but can be found in
texts on turbulence modeling (i.e. [22] or [6]). In (2.178) the Prandtl number for
turbulent kinetic energy (Prk ) appears as a closure constant (Prk = 1.0). C D is the
dissipation coefficient and has been experimentally shown to be C D = 0.164. The
term on the LHS of this equation represents the particle rate of increase of k̄. The
terms on the RHS of Eq. (2.178) represent the diffusion rate, the generation rate
and the dissipation rate of k̄, respectively. Here l is a characteristic length, referred
to as the mixing length. The mixing length can be interpreted as follows: a fluid
parcel will conserve its properties for a characteristic length, l, before mixingwith the
68 2 Review of Fundamental Equations
surrounding fluid. Prandtl [14] described that the mixing length “may be considered
as the diameter of the masses of fluid moving as a whole in each individual case; or
again, as the distance traversed by a mass of this type before it becomes blended in
with neighboring masses…”
The dissipation rate of k̄ which is embedded in the last term in (2.178) is often
3/2
represented with ε: ε = C D k̄ l . Similar to the analysis of k̄, a transport equation
developed from the Reynolds-averaged Navier-Stokes equations can be established
for ε:
Dε ∂ μT ∂ε ε 2 ∂u i ε2
ρ = μ+ + Cε1 2μT Sij − ρk̄δij − Cε2 ρ
Dt ∂x j Prε ∂x j k̄ 3 ∂x j k̄
(2.179)
The term on the LHS of Eq. (2.179) represents the particle rate of increase in dissipa-
tion, while the terms on the RHS represent the diffusion, generation and dissipation
rates of ε. In terms of k̄ and ε the terms in Eqs. (2.178) and (2.179) are as follows:
k̄ 3/2 k̄ 2 4/3
l = CD μT = C μ ρ Cμ = C D (2.180)
ε ε
By appropriate substitution, the k̄-ε model becomes a set of two coupled partial
differential equations with two unknowns: k̄ and ε.
Complementary to closure of the Reynolds-averaged momentum equations, the
k̄-ε closure for the heat flux terms in the energy equation assumes the following
apparent turbulent conductivity:
∂T
ρu j H = ρc p x j T = −k T (2.181)
∂x j
Analogous to the laminar heat conduction coefficient, k, the turbulent heat conductive
heat coefficient is defined according to:
c p μT
kT = (2.182)
PrT
where PrT is the turbulent Prandtl number that is most commonly takes on the value
of PrT = 0.9. The combination of (2.181) and (2.182) form the closure for the
energy equation. That is, if the last bracketed term in (2.175) is neglected because it
is small compared to the other terms within the square brackets. Table 2.1 displays
the constants that are to be used in the k̄-ε model.
The k-epsilon method that has been presented is tailored towards closure of the
incompressible Reynolds-averaged equations of motion. Similar models have been
developed for the closure of the Reynolds-averaged, mass-weighted equations of
motions that were developed in Sect. 2.6.2. The closure of the compressible Reynolds
equations is beyond the scope of this text. Even though in transonic aerodynamics the
flow is compressible, the incompressible turbulence models have been proven to give
good predictions up to Mach 5 [17]. The k̄-ε method for compressible flow is treated
extensively by Launder and Spalding [12] and Mohammadi and Prionneau [13].
Neglecting the viscous and heat-transfer terms in the energy equation (2.140) results
in the following:
De
ρ + p(∇ · V ) = 0 (2.184)
Dt
Alternatively, this equation can be written in terms of enthalpy, h, by modifying
(2.141):
Dh Dp
ρ = (2.185)
Dt Dt
Equations (2.113), (2.183), and (2.184) are generally known as the Euler equations.
The compressible Euler equations can be written in conservation form according
to (2.152). However, the vector representation is now simpler than for the full NS
equations:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
ρ ρu ρv ρw
⎢ ρu ⎥ ⎢ ρu 2 + p ⎥ ⎢ ρuv + p ⎥ ⎢ ρuw ⎥
⎢ ⎥ ⎢ ⎥ ⎢ 2 ⎥ ⎢ ⎥
U=⎢ ⎥
⎢ ρv ⎥ E=⎢
⎢ ρuv ⎥
⎥ F=⎢
⎢ ρv + p ⎥
⎥ G=⎢
⎢ ρvw
⎥
⎥
⎣ρw⎦ ⎣ ρuw ⎦ ⎣ ρvw ⎦ ⎣ ρw2 + p ⎦
Et (E t + p)u (E t + p)v (E t + p)w
(2.186)
The vorticity can be related to the specific entropy, s, according to Crocco’s equation:
∂V V2
− V × ζ = T ∇s − ∇ (2.189)
∂t 2
This equation can be derived from the first and second laws of thermodynamics,
(2.83) and (2.94), respectively. Note that the LHS of (2.189) is also present in the
Lagrange form of the momentum equation (2.188). A physical interpretation of
2.7 Equations of Motion for Inviscid Flows 71
The sharp pressure gradient on the upper surface at approximately 60 % of the chord
length, indicates the presence of a shock wave. If we focus on the airfoil below the
pressure distribution, we see that the streamline that separates the inviscid flow from
the viscous boundary layer has been drawn. Up to the shock wave this line virtually
coincides with the airfoil contour, indicating a very thin boundary layer. However,
due to the sharp adverse pressure gradient, the boundary layer thickness grows and
the streamline is displaced upwards. The flow outside the boundary layer therefore
‘sees’ a different body than the airfoil. This added thickness (known as displacement
thickness) is important to accurately determine the experimentally obtained pressure
distribution. Note that in this figure the pressure distributions are shown for the same
lift coefficient (cl = 0.743), while the angles of attack are slightly different. In
predicting the pressure distribution over an airfoil it has been shown that this often
gives a better correlation to experimental results. Producing these results can be done
in under a minute of computation time, making this a suitable tool for airfoil design.
To stress the importance of including a boundary layer in the calculations we also
show results for flow about a transonic airfoil without the inclusion of the boundary
layer. In Fig. 2.12 we see that the pressure distribution has been predicted using
-0.5
-0.0
0.5
1.0
72 2 Review of Fundamental Equations
-2.0 -2.0
α=3.0° Cl=0.60 M∞=0.68
Expected
-1.6 -1.6 Inviscid
Pressure coefficient, Cp (~)
-0.4 -0.4
0.0 0.0
0.4 0.4
0 .2 .4 .6 .8 1.0 0 .2 .4 .6 .8 1.0
Chordwise position, x/c (~) Chordwise position, x/c (~)
Fig. 2.12 Comparison between predicted and measured pressure distribution at constant angle of
attack (left) and constant lift coefficient (right) (after Ref. [5])
the same angle of attack as in the experiment (left) and at the same lift coefficient
as in the experiment (right). None of these predictions gives an accurate result.
This shows that we must be extremely careful when assessing the results from a
purely inviscid solver. Given the short computation time of viscous-inviscid solvers
such as MSES, it is therefore advised to include the presence of the boundary layer
when predicting the pressure distribution over a body. In particular at low-Reynolds
numbers and transonic Mach numbers this is an important prerequisite for obtaining
reliable results.
Even though the previous analysis showed that the vorticity in transonic flow cannot
be neglected, there is a body of literature that is based on steady, inviscid, irrotational
flow with no body forces [10]. It can be shown that for weak normal shocks (M1 ≈ 1)
the entropy change is relatively small. Based on this finding the approximation can be
made that transonic flow is irrotational and that that therefore the curl of the velocity
field (see Sect. 2.3.4) is zero:
∇×V =0 (2.190)
In potential flow we pose that there exists a scalar potential function, Φ, which’
partial derivatives equal the velocity components:
∇Φ = V (2.191)
2.7 Equations of Motion for Inviscid Flows 73
In Cartesian coordinates, we align the x-axis with the freestream velocity, V ∞ , and
represent the local velocity components, u, v and w by
∂Φ
u= = Φx (2.192a)
∂x
∂Φ
v= = Φy (2.192b)
∂y
∂Φ
w= = Φz (2.192c)
∂z
If we assume the flow to be steady, the continuity equation (2.113) reduces to:
∇ · (ρV ) = 0 (2.193)
∂ ∂ ∂
(ρΦx ) + (ρΦ y ) + (ρΦz ) = 0 (2.194)
∂x ∂y ∂z
Note in this equation that the partials ρx , ρ y , and ρz appear. We will find an expression
for each of these density derivatives by considering the potential form of the steady,
inviscid, irrotational momentum equation.
If we assume that the flow is steady (i.e. ∂/∂t = 0) the Lagrange form of the
momentum equation (2.188) can be reduced to:
V2 1
∇ = − ∇p (2.195)
2 ρ
In isentropic flow the speed of sound (a) is given by (2.106). We can substitute the
expression for the speed of sound (2.106) in (2.196) and obtain:
!
ρ Φx2 + Φ y2 + Φz2
dρ = − 2 d (2.197)
a 2
In the late 1980s and early 1990s the analysis code Matrics-V was developed by
the National Aerospace Laboratory of the Netherlands to estimate the aerodynamic
forces on wing-fuselage combinations (three-dimensional) in high-subsonic condi-
tions. This code relies on a viscous-inviscid formulation, where the full potential
equation (2.199) is numerically solved in the inviscid domain outside the boundary
layer. In the boundary layer the boundary layer equations are numerically solved.
Subsequently, an interaction algorithm is used to match the flow properties at the
1.5 1.5
1.0 1.0
Pressure coefficient, -Cp (~)
0.0 0.0
-0.5 -0.5
2y/b = 0.206000 2y/b = 0.521000
-1.0 -1.0
Matrics-V Matrics-V
Flight test Flight test
-1.5 -1.5
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Chordwise position, x/c (~) Chordwise position, x/c (~)
Fig. 2.13 Comparison of predicted and measured pressure distributions about two sections of the
Fokker 100 wing (from [20]). Note that this is a two-dimensional representation of results that were
obtained from three-dimensional calculations (Matrics-V)
2.7 Equations of Motion for Inviscid Flows 75
edge of the boundary layer to the inviscid flow outside the boundary layer [20]. This
analysis code was specifically developed to investigate the flow about wing-body
combinations. The objective was to develop a code that could produce results in less
than week of computation time. Due to the increase in computational power, this has
now (2015) reduced to less than five minutes on a personal computer, including the
generation of the mesh. In Fig. 2.13 a comparison between the predicted and mea-
sured pressure distribution at two spanwise wing stations of the Fokker 100 can be
seen. We see that the prediction of the pressure distribution is in excellent agreement
with the measured data at the same angle of attack. The accurate prediction in com-
bination with the short computation time make this type of code a viable candidate
in the early stages of preliminary wing design.
2.8 Summary
In the preceding sections we have briefly reviewed the fundamental tools that the
reader needs to be familiar with in order to properly understand the material that
is presented in the subsequent chapters. We have done a very basic review of par-
tial differential equations, their solution methods, and their classification. We have
also re-acquainted ourselves with the mathematical operations in vector algebra. We
have explained the fundamental laws of thermodynamics: the state law that relates
the thermodynamic state variables to each other, the first law that balances work,
heat and internal energy, and the second law that tells us the direction in which a
thermodynamic process takes place. Finally, we have applied all this knowledge in
defining the equations of motion of fluid flow. When expanded we have shown that
the equations of motion are a coupled system of five partial differential equations
[(2.110), (2.127) (3 components), and (2.139)] with five unknowns: ρ, e, and the three
velocity components, u, v, w. These equations are often referred to as the Navier-
Stokes equations and they can be cast in different forms: integral form, substantial-
derivative form, and conservative-derivative form. Because solving these equations
requires vast computational resources we have presented derivatives of these equa-
tions. In the Reynolds-averaged Navier-Stokes equations the flow is decomposed into
a mean flow with fluctuating terms. A turbulence model can then be used to relate
those fluctuating terms to the mean flow parameters. A further simplification of the
Navier-Stokes equations is obtained through the assumption of having an inviscid
and non-conducting flow. The resulting equations are often referred to as the Euler
equations and they can provide an accurate prediction of the flow about a body, in
particular when they are used in parallel with the boundary-layer equations. If also
the assumption is made that the flow is irrotational, the Euler equations reduce to
the full-potential equation. Solving the three-dimensional full-potential equation in
combination with the boundary-layer equations can still produce accurate predictions
of transonic flow about bodies, as long as the shock strength is relatively weak.
76 2 Review of Fundamental Equations
Problems
Review of Partial Differential Equations
2.1 Show that (2.7) results from (2.5) by using the chain rule and substituting ξ =
x + ct and η = x − ct.
2.2 Demonstrate that u(x, t) = sin π(x+ct)
L + sin π(x−ct)
L is a solution to the
following problem:
u tt = c2 u x x
πx
u(x, 0) = 2 sin
L
u(0, t) = u(L , t) = 0
2.3 Consider a string of steel wire, measuring 1 m in length and weighing 0.5 N. It
is stretched by a force of 100 N. What is the corresponding speed c of the transverse
waves?
2.4 Consider the wave equation (2.5) with the following boundary conditions:
u(0, t) = u x (L , t) = 0 .
(a) Use separation of variables technique to calculate the eigenvalues, eigenfunctions
and general solution.
(b) Now, assume L = π and c = 1. With initial conditions u(x, 0) = 0 and
u t (x, 0) = 1, calculate the solution for u(x, t).
(c) With initial conditions u(x, 0) = sin(x/2) and u t (x, 0) = 2 sin(x/2) −
3 sin(5x/2) calculate the solution for u(x, t).
2.5 Demonstrate that (2.22) can be derived from u t = 0 by using the chain rule and
the transformation (2.6).
2.6 Show that the Jacobian for the change in characteristic coordinates (2.6)
equals 2c.
2.7 For the same problem as in Example 2.2, calculate the solution for:
(a) x < t, t > 0, x + t < 0, and check u tt − u x x = 1.
(b) x > 1, t < x < t + 1, and check u tt − u x x = 0.
2.9 Demonstrate that from (2.49) we can obtain (2.50) by using the method outlined
in the text.
u t = ku x x (0 x L)
2.8 Summary 77
u(0, t) = 0 (t > 0)
u x (L , t) = 0 (t > 0)
u(x, 0) = f (x) (0 < x < L)
2.11 For Problem 2.10, provide calculate Dn in closed form for L = π and
f (x) = x.
2.13 Consider the potential equation on p. 74 (2.199). When M > 1, determine the
characteristic coordinates and write the equation in its canonical form.
2.15 For the following velocity vector fields, find the potential function if it exists.
Check that (2.66) holds in each case. φ(x, y, z): (a) (x, 2y, 3z), (b) (yz, x z, x y),
(c) (ye x , e x , 1), (d) (2y, 5x, 0).
2.16 For the vector fields of Problem 2.15, calculate the divergence.
2.17 For the vector fields of Problem 2.15, calculate the curl. Compare your results
to those of Problem 2.15. What do you notice?
Review of Thermodynamics
2.18 With c p as given in Fig. 2.8 and R as given in the text, calculate cv , γ, e, and
h for air at an altitude of 10 km under ISA conditions.
2.19 For the values given in Problem 2.18 also calculate the viscosity of the air, μ
and the thermal conductivity, k.
2.20 Calculate the change in specific internal energy, Δe, for air that is being com-
pressed isentropically. Assume that the initial temperature is 288 K and the final
temperature, due to compression, 340 K.
2.21 Show that (2.97) and (2.98) can be derived by combining the first and second
law of thermodynamics, (2.83) and (2.94), respectively.
78 2 Review of Fundamental Equations
2.22 Derive the following relations by starting from (2.97) and (2.98), respectively.
dT dv
ds = cv +R (2.200)
T v
dT dp
ds = c p −R (2.201)
T p
2.23 Demonstrate that (2.103) can be derived by combining (2.101) and (2.102) and
applying the substitutions as described in the text.
Equations of Fluid Motion
2.24 Use the divergence theorem (2.69) to prove that (2.109) and (2.110) are math-
ematically identical.
2.25 Consider the LHS of (2.122). Show that by using (2.110) this can be simplified
to ρ DV
Dt .
D(E t /ρ) ∂ Et
ρ = + ∇ · Et V
Dt ∂t
(b) Demonstrate that also the following relation is true:
D(E t /ρ) De DV
ρ =ρ +ρ V
Dt Dt Dt
(c) Use (2.122) to show that the second term in the equation above can be written as:
DV
ρ V = ρ f · V − ∇ p · V + (∇ · τ ij ) · V
Dt
(d) Using the three equations above, demonstrate that (2.140) is identical to (2.139).
2.27 Consider the energy equation in substantial-derivative form (2.140).
(a) Show, by employing the continuity equation, that the following identity is true:
De Dh Dp
ρ + p(∇ · V ) = ρ −
Dt Dt Dt
(b) Explicitly write out the dissipation function, Φ, in terms of the velocity com-
ponents in three-dimensional Cartesian coordinates. Use the assumption that
μ = 23 μ.
2.28 Write out (2.110), (2.149), and (2.150) in three-dimensional Cartesian coordi-
nates and demonstrate that you can write this as (2.152).
2.8 Summary 79
References
1. Anderson, J.: Computational Fluid Dynamics: The Basics and Applications, 1st edn. McGraw
Hill, New York (1995)
2. Anderson, J.: Modern Compressible Flow with Historic Perspective, 3rd edn. McGraw Hill,
New York (2003)
3. Anderson, J.: Fundamentals of Aerodynamics, 5th edn. McGraw Hill, New York (2010)
4. Anon.: Direct numerical simulation (DNS) wikipedia. http://en.wikipedia.org/wiki/Direct_
numerical_simulation (2012)
5. Blackwell, J.A.: Scale effects on supercritical airfoils. In: Proceedings of ICAS, pp. 370–283
(1978)
6. Cebeci, T.: Analysis of Turbulent Flows. Elsevier, Amsterdam (2004)
7. Cook, P.H., McDonald, M.A., Firmin, M.C.P.: Aerofoil RAE 2822—pressure distributions,
and boundary layer and wake measurements. In: AGARD-138 (1979)
8. Drela, M., Giles, M.B.: Viscous-inviscid analysis of transonic and low Reynolds number air-
foils. AIAA J. 25(10), 1347–1355 (1987). doi:10.2514/3.9789
9. Farokhi, S.: Aircraft Propulsion, 2nd edn. Wiley, Chichester, UK (2014)
10. Ferrari, C., Tricomi, F.G.: Transonic Aerodynamics. Academic Press, New York (1968)
11. Keener, J.P.: Principles of Applied Mathematics. Westview Press, Boulder (1999)
12. Launder, B.E., Spalding, D.B.: Lectures in Mathematical Models of Turbulence. Academic
Press, London (1972)
13. Mohammadi, B., Pironneau, O.: Analysis of the K-epsilon Turbulence Model. Wiley, Chich-
ester, UK (1994)
14. Prandtl, L.: Über die ausgebildete turbulenz. In: Proceedings of the Second International
Congress of Applied Mechanics, pp. 62–75. Orell Füssli Verlag, Zurich (1927)
15. Reynolds, O.: On the dynamical theory of incompressible viscous fluids and the determination
of the criterion. Philos. Trans. R. Soc. Lond. 186, 123–164 (1895)
16. Ruijgrok, G.J.J.: Elements of Airplane Acoustics. Eburon, Delft (2003)
17. Schlichting, H., Gestern, K.: Boundary Layer Theory, 8th edn. Springer, Berlin (1999)
18. Strauss, W.A.: Partial Differential Equations: An Introduction. Wiley, Chichester (1992)
19. Tannehill, J.C., Anderson, D.A., Pletcher, R.H.: Computational Fluid Mechanics and Heat
Transfer, 2nd edn. Taylor & Francis, Philadelphia (1997)
20. Van Muijden, J., Broekhuizen, A.J., Van der Wees, A.J., Van der Vooren, J.: Flow analysis
and drag prediction for transonic transport wing/body configurations using a viscous-inviscid
interaction type method. In: Proceedings of the 19th ICAS Congress. Anaheim, California
(1994)
21. Van Wylen, G., Sonntag, R.: Fundamentals of Classical Thermodynamics. Wiley, Chichester,
UK (1973)
22. Wilcox, D.: Turbulence Modeling for CFD, 2nd edn. DCW Industries, Canada (1998)
Chapter 3
Transonic Similarity Laws
Abstract This chapter starts the foundation with the small disturbance theory
applied in the subsonic and supersonic flows. Both of these regimes are linear and
thus fail at sonic speed, which is at the heart of transonic flow. Nonlinear small
disturbance theory for the transonic regime is subsequently derived and discussed.
With limited analytical options for the transonic nonlinear differential equation, spe-
cial attention is directed to the development of similarity laws. These laws connect
non-dimensional parameters such as pressure coefficient, lift and drag coefficients on
families of affinely related bodies of different thickness or slenderness ratio and angle
of attack. Brief discussion of hodograph transformation where switching between
the dependent and independent variables causes the governing equation to be lin-
ear is presented. Finally, semi-empirical models on lift curve slope in the transonic
regime and the approximate location of detached shocks in front of blunt bodies are
developed and compared to experimental results. This chapter contains 12 examples
and concludes with 39 practice problems.
3.1 Introduction
similarity parameters, as we shall see. The utility of such concepts is in the power of
generalization inherent in the similar solutions in transonic regime. This allows us to
extrapolate the aerodynamic behavior of affinely related bodies at different transonic
Mach numbers and different thickness, or slenderness ratios and angles of attack. It is
interesting to note that another nonlinear aerodynamic regime in the context of small
disturbance theory, namely hypersonic flow, also lends itself to the development of
similarity parameters on affinely-related bodies. In this chapter we examine similar-
ity solutions from subsonic to supersonic flows including the transonic regime. The
monumental work of Shapiro [9] is a key reference to this classical topic.
To study the aerodynamic characteristics of thin planar and slender bodies, in small
angles of attack (and/or sideslip), the theory of small disturbances is developed. We
propose that for a thin and slender body, with small angle of attack, or sideslip, the
x-component of velocity is but a small perturbation around the freestream speed,
V∞ , and the other two velocity components, Vy and Vz are both small as compared
to freestream velocity. This proposition is the hallmark of small perturbation theory.
It suggests that a thin or slender body can only perturb the flow a small amount, as
compared to freestream. Expressed mathematically, we imply
where the û, v̂ and ŵ are perturbation velocity components in x, y and z directions
respectively and they are small compared to V∞ . Obviously, this approximation
excludes the neighborhood of stagnation point where the change in velocity is large.
Therefore, excluding the neighborhood of stagnation point, we can now introduce a
perturbation potential function that describes the perturbation velocity components
in the flow according to:
û = φx
v̂ = φ y (3.2)
ŵ = φz
Therefore, the full velocity potential, Φ and the perturbation velocity potential, φ are
related by the following set of equations:
Now, for steady, inviscid, irrotational and isentropic flow, we have the full potential
equation, namely
Φx2 Φ y2 Φz2
1 − 2 Φx x + 1 − 2 Φ yy + 1 − 2 Φzz
a a a
Φx Φ y Φx y Φ x Φz Φ x z Φ y Φz Φ yz
−2 2
−2 2
−2 =0 (2.199)
a a a2
This is a second order, nonlinear partial differential equation that is best solved numer-
ically. Also, note that the local speed of sound, a, is related to velocity components
following the energy equation in adiabatic flows according to:
a2 V2 a2 V2
+ = ∞ + ∞ (3.4)
γ−1 2 γ−1 2
γ−1 2
a 2 = a∞
2
− V − V∞
2
(3.5)
2
We then obtain:
γ−1
a 2 = a∞
2
− 2V∞ û + û 2 + v̂2 + ŵ2 (3.7)
2
γ−1 2 (γ − 1) 2 û û 2 v̂2 ŵ2
a =
2 2
a∞ − V − V∞ = a∞ −
2 2
V∞ 2 + 2 + 2 + 2
2 2 V∞ V∞ V∞ V∞
(3.8)
We note that the last two terms in the parenthesis on the right-hand-side of (3.8)
are second-order terms, i.e., O(ε2 ), and may be neglected; therefore the local speed
of sound may be approximated, to first order, by
a2 φx
2
≈ 1 − (γ − 1)M∞
2
+ O(ε2 ) (3.9)
a∞ V∞
84 3 Transonic Similarity Laws
The full potential equation (2.199), in the limit of small perturbations, where we
neglect the squares and products of small parameters, i.e., the second order terms,
as O(ε2 ), and for freestream Mach numbers that exclude transonic and hypersonic
flows, may be written as
(1 − M∞
2
)φx x + φ yy + φzz = 0 (3.10)
This is the classical linear differential equation that governs the dynamics of gas
in steady subsonic and supersonic flows around bodies that are thin (planar), or
slender (3-D) and are at small angles with respect to the flow. The impact of small
perturbations in velocity is also felt at the pressure level. We introduce the pressure
coefficient, C p , which is a non-dimensional coefficient for the pressure differential
at a particular location in the flow:
γ
p − p∞ 2 p 2 T γ−1
Cp ≡ = −1 = −1 (3.11)
q∞ γ M∞2 p∞ γ M∞2 T∞
V2 V2
T+ = T∞ + ∞ (3.12)
2c p 2c p
φx
C p ≈ −2 (3.15)
V∞
The solid surface boundary condition demands that the local flow slope and the local
body slope to be matched at all points along the body surface. This is also known
as the flow tangency condition to the solid surface. An alternative description of
solid surface boundary condition calls for the normal component of the flow to the
solid surface to vanish at all points on the solid surface. For a 2-D body in the x–y
3.2 Linearized Compressibility Corrections 85
θ
V∞
y = f(x)
plane, the body may be mathematically represented by y = f (x) with a local body
slope of dy/dx or d f /dx. This is schematically shown in Fig. 3.1. A representative
flow velocity vector is also shown in Fig. 3.1. The corresponding flow slope may be
expressed in terms of a general angle, θ, which is also valid on the body, i.e.,
v̂ v̂ φy
tan θ = ≈ = (3.16)
V∞ + û V∞ V∞
df φy
≈ (3.17)
dx V∞
In the case of subsonic flow, i.e., M∞ < 0.8 that also excludes transonic regime,
Eq. (3.10) is an elliptic partial differential equation:
βφx x + φ yy = 0 where β ≡ 1 − M∞
2 (3.18)
A transformation of the function φ(x, y), is proposed that converts Eq. (3.18) in
the x–y plane (known as the physical plane) to the Laplace’s equation in φ̄(ξ, η),
which governs the incompressible flow, in the transformed ξ–η plane. Consider the
following transformation:
ξ=x
η = βy (3.19)
φ̄ = βφ
Then, by using the chain rule, we calculate the terms of Eq. (3.18) in the new variables,
for example
86 3 Transonic Similarity Laws
∂φ ∂φ ∂ φ̄ ∂ξ 1
φx = = = φ̄ξ (3.20)
∂x ∂ φ̄ ∂ξ ∂x β
∂2φ ∂ 1 ∂ 1 ∂ξ 1
φx x = = φ̄ ξ = φ̄ ξ = φ̄ξξ (3.21)
∂x 2 ∂x β ∂ξ β ∂x β
∂φ ∂φ ∂ φ̄ ∂η 1
φy = = = φ̄η β = φη (3.22)
∂y ∂ φ̄ ∂η ∂ y β
∂2φ ∂ ∂ ∂η
φ yy = = φ̄η = φ̄η = φ̄ηη β (3.23)
∂ y2 ∂y ∂η ∂y
By substituting Eqs. (3.21) and (3.23) in (3.18), we get the transformed governing
equation in (ξ, η) plane, i.e.,
1
β2 φ̄ξξ + β φ̄ηη = 0 → φ̄ξξ + φ̄ηη = 0 (3.24)
β
dg φ̄η
≈ (3.25)
dξ V∞
But since φ̄η = φ y following (3.22), we note that the transformed and the physical
bodies have the same slope,
dg df
= (3.26)
dξ dy
Note that we have added a subscript “M∞ = 0” to C p in order to signify the Mach-
zero or incompressible condition that prevails in the transformed plane. Replacing
φξ by βφx from (3.20) in (3.27), we relate the compressible and incompressible
pressure coefficients, namely
φ̄ξ βφx φx
Cp M ≈ −2 = −2 = β −2 = β Cp M (3.28)
∞ =0 V∞ V∞ V∞ ∞
3.2 Linearized Compressibility Corrections 87
Or alternatively,
C p M =0
∞
C p M =0
∞
Cp M = =
(3.29)
∞ β 1 − M∞2
1
cl = C p dx (3.30)
c
We conclude that the lift coefficient in subsonic compressible flow also is magnified
by the same factor, (1/β) from the incompressible lift coefficient that appeared in
C p , namely
(cl ) M∞ =0 (cl ) M∞ =0
(cl ) M∞ = =
(3.31)
β 1 − M∞ 2
Similarly, the lift curve slope, in the subsonic compressible plane is related to the
incompressible lift slope, according to
dcl dcl
dcl dα M∞ =0 dα M∞ =0
= =
(3.32)
dα M∞ β 1 − M∞
2
This equation also indicates that the lift curve slope is magnified in the subsonic
compressible flow as compared to the incompressible, by the factor 1/β. To complete
the discussion of two-dimensional lift in the compressible domain, we address the
issue of circulation, Γ, around an airfoil. Since the local strength of the bound vortex
sheet, representing an airfoil, is equal to the tangential velocity jump across the sheet,
which is amplified by 1/β factor in the compressible plane, the vortex sheet strength
is increased by the same factor. Therefore the circulation, which is the integral of the
vortex sheet strength along the mean camber line, is amplified by the same factor.
Another view that leads to the same conclusion is based on static pressure jump and
its proportionality to vortex sheet strength. Therefore, we may conclude that
(Γ ) M∞ =0 (Γ ) M∞ =0
(Γ ) M∞ = =
(3.33)
β 1 − M∞ 2
Finally, we note that the Kutta-Joukowski theorem relating lift to circulation will
remain valid in the compressible subsonic flow as well, i.e.,
L M = ρ∞ V∞ (Γ ) M∞ (3.34)
∞
88 3 Transonic Similarity Laws
x ξ
Fig. 3.2 An airfoil in the physical and the transformed planes following Prandtl-Glauert transfor-
mation
Figure 3.2 shows the same body in two planes, one subsonic compressible and the
other incompressible, following the Prandtl-Glauert transformation that preserves
the shape of the transformed body. Finally, the pitching moment coefficient on an
airfoil, say about the leading edge, is related to the integral of the moment of the
pressure coefficient according to:
1
cm, LE = − xC p dx (3.35)
c2
We conclude that the subsonic compressible pitching moment, is also amplified by
the same factor, 1/β namely
cm, LE M =0
∞
cm, LE M =0
∞
cm, LE M = =
(3.36)
∞ β 1 − M∞2
The hallmark of all of these compressibility correction results that are derived from
the linear theory, is the singularity of the solutions at M∞ = 1. The failure to predict
a finite pressure, lift, drag, pitching moment coefficients in the vicinity of sonic flow
is due to the linearized nature of differential equation, which does not represent the
physics of transonic flow, i.e., dominated by the nonlinearity. We will examine the
suitable transonic small perturbation equation after we review a few examples from
the linear theory.
Example 3.1 The minimum (or peak suction) pressure coefficient on an airfoil at 2◦
angle of attack, in incompressible flow, is C p,min = −0.95. Calculate the pressure
coefficient at the same point on the airfoil, when freestream Mach number is 0.5.
Solution:
Applying the Prandtl-Glauert compressibility correction to this airfoil, we get:
3.2 Linearized Compressibility Corrections 89
(C p, min ) M∞ =0 −0.95
C p, min M =
=
≈ −1.097
∞
1 − M∞ 2 1 − (0.5)2
Example 3.2 The lift curve slope, Clα , of an airfoil in incompressible flow is 6 rad−1 .
Calculate the lift curve slope for the same airfoil at Mach 0.6.
Solution:
The lift curve slope is also amplified by the 1/β factor according to Prandtl-Glauert
compressibility correction formula, namely (3.29), therefore
dcl
dcl dα M∞ =0 6 [rad−1 ]
=
=
≈ 7.5 [rad−1 ]
dα M∞ 1 − M∞2 1 − (0.6) 2
Solution:
According to Prandtl-Glauert compressibility correction, the lift coefficient in sub-
sonic compressible flow is amplified by 1/β factor, therefore,
(cl ) M∞ =0 1.05
(cl ) M∞ =
=
≈ 1.212
1 − M∞ 2 1 − (0.5)2
The circulation around an airfoil is also increased by the same 1/β factor, as Kutta-
Joukowski theorem on airfoil lift is maintained, according to (3.33) and (3.34):
(Γ ) M∞ 1 1
=
=
≈ 1.154
(Γ ) M∞ =0 1 − M∞
2 1 − M∞
2
There are several other compressibility corrections in the subsonic flow. One such
correction is due to Laitone and another one is due to Karman-Tsien. The first, i.e.,
Laitone [4], extends the Prandtl-Glauert correction by replacing the freestream Mach
number with the local Mach number, i.e.,
90 3 Transonic Similarity Laws
C p M =0
∞
Cp M = (3.37)
∞
1 − M L2
The local Mach number on the airfoil in the compressible domain is noted by M L in
(3.37). The effect of this improvement is to create higher suction on airfoils where
Prandtl-Glauert consistently under-predicted the experimental data. Laitone related
M L to the freestream Mach number via an isentropic relation and expressed the
revised Prandtl-Glauert compressibility correction for subsonic flow as:
C p M =0
Cp M =
∞
(3.38)
∞
γ−1 2
1 − M∞ + M∞ 1 + 2 M∞ /2 1 − M∞ C p M
2 2 2
∞ =0
Tsien corrections apply to the local pressure coefficient only and do not extend to the
force coefficients, as in the Prandtl-Glauert correction. However, based on point wise
correction of local pressure coefficient in the subsonic compressible domain, we may
proceed to integrate the pressure jumps and their moment to calculate lift and moment
coefficients on the airfoil using Laitone or Karman-Tsien compressibility corrections.
In subsonic compressible flow over an airfoil, we define a new parameter, called
the critical Mach number, Mcrit . The flight Mach number that corresponds to the
first appearance of a sonic flow on the airfoil is called the critical Mach number. The
first appearance of the sonic point occurs at peak suction, i.e., the point of minimum
pressure, C p, min , or the point of maximum velocity, Vmax on the airfoil, as expected.
Since the pressure coefficient at the sonic point on the airfoil is defined as
2 p∗
C pcrit = −1 (3.40)
γ Mcrit
2 p∞
And we may relate the static pressure ratio, p ∗ / p∞ , to the Mach numbers, Mcrit and
Mach 1 (for sonic point) using an isentropic flow connection (2.147) between the
flight and the sonic point on the airfoil, i.e.,
γ−1 2
γ
p∗ 1+ 2 Mcrit
γ−1
= γ+1
(3.41)
p∞
2
We can express the critical pressure coefficient as a function of the critical Mach
number as:
3.5
3 Supersonic
2.5 Flow
1.5
1 Subsonic
Flow
0.5
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Critical Flight Mach number, Mcrit (~)
92 3 Transonic Similarity Laws
⎡ γ ⎤
γ−1 2 γ−1
2 ⎣ 1+ 2 Mcrit
C pcrit = γ+1
− 1⎦ (3.42)
γ Mcrit
2
The graph of (3.42) for −C pcrit , as is customary, is shown in Fig. 3.4. We note that
C pcrit is the sonic boundary with the region to the right of the curve denoting super-
sonic and the region to the left of C pcrit curve identifying the subsonic regime. We use
the knowledge of critical Mach number on an airfoil at certain angle of attack, in
low speed, to mark the onset of transonic flow phenomenon, or the onset of shock
appearance. We can graphically construct the critical Mach number for an airfoil
where the knowledge of peak suction C pmin exists in the incompressible limit.
Example 3.4 The airfoil of Example 3.1 had a C pmin = −0.95 at the incompressible
limit. First graph the C pmin for the same airfoil, in the subsonic compressible flow,
using the Prandtl-Glauert compressibility correction and second find the intersection
of this curve and the critical pressure coefficient described by (3.42).
Solution:
We first generate
a spreadsheet of the Prandtl-Glauert compressibility correction
applied to C p M=0 = −0.95 for a range of Mach numbers (below 1) and then
graph the two functions, representing C p , as shown in Fig. 3.5.
Example 3.5 Develop a family of compressibility correction curves using Karman-
Tsien approach for the range of incompressible C p s between 1, corresponding to
incompressible stagnation point and −3 corresponding to strong suction at angle of
attack. Also, graph critical pressure coefficient representing the sonic point on the
airfoil.
Critical Pressure Coefficient, -Cp, crit (~)
4.5
-C p,crit
4
3.5
Prandtl - Glauert
3 Compressibility
Correction
2.5
1.5
Incompressible 1
C p ,min = - 0.95
0.5
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Critical Mach number, Mcrit (~)
Solution:
We apply Karman-Tsien compressibility correction to the incompressible pressure
coefficients, described in the problem, and graph them using a spreadsheet program.
The result is shown in Fig. 3.6. The first appearance of sonic flow on the airfoil is
marked by the critical pressure coefficient, C pcrit curve. The Mach numbers beyond
critical, which is also called supercritical Mach numbers, will no longer be reached
isentropically.
Some classical experimental data, shown in Fig. 3.7 reveal the lift coefficient, k L
as a function of angle of attack (here called angle of incidence) for two airfoils.
Since k L in these graphs represents lift coefficient based on double dynamic pressure
(ρ∞ V∞ 2 ), it represents 1 of how we conventionally define lift coefficient, c , i.e., based
2 l
on dynamic pressure (ρ∞ V∞ 2 ). The U in these graphs represents V and a represents
∞ 1
a∞ , therefore a freestream Mach umber range 0.25 < M∞ < 1.70 is investigated.
To relate the lift curve slopes between different subsonic Mach numbers, we may use
5
4.5 -C p,crit
Pressure Coefficient, -Cp (~)
4
3.5
3
2.5
2
1.5
1
Suction
0.5
0
-0.5
Compression -1
-1.5
Incompressible
Stagnation Point -2
(-Cp ) 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Freestream Mach Number, M∞ (~)
Fig. 3.6 Family of Karman-Tsien compressibility correction curves and the critical pressure coef-
ficient
94 3 Transonic Similarity Laws
0.6 0.6
kL kL
0.4 0.4
U=0.7a1 U=0.5a1
U=0.5a1 U=0.25a1
U=0.7a1
U=0.25a1
U=0.8a1
0.2 0.2
U=1.7a1
U=1.7a1
0.0-5 0 5 10 0.0 -5 0 5 10
Angle of Incidence (deg) Angle of Incidence (deg)
-0.2 -0.2
Fig. 3.7 Experimental data on airfoil lift in subsonic compressible flow on two airfoils (Definition
of lift coefficient k L is based on double dynamic pressure, therefore it is 21 cl ), after [11]
dcl
dα 1 − M22
M1 = (3.43)
dcl
dα 1 − M12
M2
The lift curve slopes as a function of Mach number follow a near perfect match
with the Prandtl-Glauert prediction for the circular arc airfoil (i.e., between 0.25 and
0.50). Although, the lift curve slope increase to Mach 0.7 is less accurate, but still
it is in the right direction, i.e., slope increases with compressible subsonic Mach
number. The accuracy is however reduced in the RAF-31a airfoil. In both cases,
the main departure in the subsonic flow occurs when we cross the critical Mach
number boundary and enter the transonic regime on the airfoil. As evidenced by
data, RAF-31a has entered the transonic zone at freestream Mach number of ∼0.5
and the circular arc airfoil enjoys a higher critical Mach number, in excess of 0.7.
The example of subsonic compressible flow over a wavy wall is both instructive
in its own right, as well as serving as a building block for wind tunnel testing in
transonic flow. Here, we use an example to study the flow over a wavy wall.
Example 3.6 Consider a subsonic compressible flow over a wavy wall of infinite
extent, in the ± x-direction, with amplitude, h and the wavelength, . The wall may
be described by a cosine function according to:
2πx
yw = h cos (3.44)
3.2 Linearized Compressibility Corrections 95
Calculate the perturbation potential function, φ(x, y), as well as pressure distribution
on the wall. Also, determine the similarity parameter based on h/.
Solution:
In the limit of small h/, we may use
β 2 φx x + φ yy = 0 (3.45)
As the governing equation for φ, subject to the flow tangency boundary condition on
the wall and total decay of the perturbations at infinity. The separation of variables
suggests that
φ(x, y) = F(x)G(y) (3.46)
Therefore,
φx x = F (x)G(y) (3.47a)
φ yy = F(x)G (y) (3.47b)
F 1 G
+ 2 =0 (3.48)
F β G
We set the pure function of x equal to a constant and the pure function of y equal to
the negative of that constant to satisfy Eq. (2.50), namely
F
= −k 2 (3.49a)
F
1 G
= k2 (3.49b)
β2 G
The periodic function in x matches the periodic wall and boundary condition in x,
i.e., the wavy wall. The solution to (3.49b) is
The coefficients A, B, C and D are as yet arbitrary and need to be determined from
our boundary conditions. The exponentially decaying part of the solution in lateral
direction matches the boundary condition at infinity for subsonic flows and is thus
physical. However, the exponentially growing term is unphysical and we thus need
to set D = 0. The wall boundary condition demands
96 3 Transonic Similarity Laws
φy dyw h
≈ = −2π sin(2πx/) (3.52)
V∞ dx
βk h
− [A cos(kx) + B sin(kx) = −2π sin(2πx/)
V∞
We may satisfy the above equation, by inspection, which sets A = 0, k = 2π/ and
βk h
B = 2π
V∞
hV∞
B=
β
Comparing the pressure coefficient at the wall to the cosine function that described
the wall, we conclude that the pressure is 180◦ out of phase (or −Cp is in-phase)
with the wall and is symmetric with respect to its peaks and valleys; therefore, the
net axial force, i.e., the (inviscid) drag, on the wall is zero. This is D’Alembert
paradox. We also note that the impact of Mach number on the pressure coefficient is
inversely proportional to β, which was predicted by Prandtl-Glauert compressibility
correction. We finally note the attenuation of pressure disturbance away from the wall
(i.e., the exponential decay function) depends on M∞ through the −β term, which
approaches zero as M∞ approaches 1 (see Eq. (3.54)). This is another hallmark of
transonic flow, namely the disturbances that are caused by the body propagate away
from the body without significant decay or attenuation.
3.2 Linearized Compressibility Corrections 97
The general expression for pressure coefficient, invoking the wave number
k = 2π/, is
2hk
Cp ≈ − cos(kx)e−kβ y (3.56)
β
Therefore, the similarity parameter, f , for the wavy wall involving the pressure
coefficient, freestream Mach number and the non-dimensional wall parameter is:
C p 1 − M∞
2
= f (kx, kβ y) (3.57)
hk
Example 3.7 Calculate the perturbation potential for a subsonic compressible flow
over a wavy wall parallel to a straight wall separated by a distance H , as shown
in Fig. 3.8. Also explore the interference effect of the wind tunnel on the wavy
wall pressure coefficient, C pw , as a function of freestream Mach number and non-
dimensional length scales of the model and the test section, e.g., h/ and h/H .
Solution:
For the adaptation of linearized flow, we require the wall amplitude parameter, h/l
1. Consequently, our governing equation, the linearized pressure coefficient and the
wall boundary conditions are:
β 2 φx x + φ yy = 0
φx
C p ≈ −2
V∞
dys
φ y ≈ V∞
dx
98 3 Transonic Similarity Laws
We again approach the problem through separation of variables, as before, and note
that the presence of the upper wall imposes a new boundary condition and thus
remove the necessity to set the coefficient of the exponentially growing term equal
to zero. The new solution, again in terms of the wave number k, is:
V∞ h
φ(x, y) = −2βk
sin(kx)e−βky 1 + e2βk(y−H ) (3.58)
β 1−e H
The pressure coefficient follows the linear theory and at the wall, we approximate
the wall location by y = 0, or the x-axis, to get:
2hk 1 + e−2βk H
C pw ≈− cos(kx) (3.59)
β 1 − e−2βk H
Comparing (3.59) that corresponds to the windtunnel environment to the wavy wall
problem in unbounded space, i.e.,
2hk
C pw ≈ − cos(kx)
β
And the dimensionless length scale is k H . The wind tunnel interference is defined
as the change in C pw (or local speed on the model) caused by the presence of the
windtunnel wall, expressed as a percentage, namely
C pw C pw W T − C pw f light 1 + e−2βk H 2e−2βk H
= = − 1 =
C pw f light C pw f light 1 − e−2βk H 1 − e−2βk H
2
= (3.61)
e+2βk H − 1
The 2-D linearized supersonic flow is governed by the same small perturbation equa-
tion we derived earlier (see 3.10), here cast in a suitable form representing supersonic
freestream Mach number, as
λ2 φx x − φ yy = 0 where λ = 2 −1
M∞ (3.62)
where F is any function of its argument, i.e., (x − λy). Also, any function of (x + λy)
is a solution to the wave equation, namely
M∞
y y
M∞ µ∞ M∞ µ∞
x x
µ∞
µ∞
Fig. 3.10 Wave patterns generated by a thin bi-convex airfoil in supersonic flow as viewed by
linear theory
φx F
C p ≈ −2 = −2 (3.65)
V∞ V∞
Here, F is the derivative of F with respect to its argument (x − λy). Now, we relate
F to φ y via
φ y = −λF (3.66)
Therefore, the pressure coefficient is
F 2 φy
C p = −2 = (3.67)
V∞ λ V∞
The solid surface boundary condition demanded that the flow and the body (locally)
possess the same slope (see 3.17), namely
df φy
≈ (3.68)
dx V∞
In which the body is represented with the function f. The body surface slope is the
tangent of the body surface angle, which for thin bodies may be approximated by
the local body angle, θ, in radians, i.e.,
2θ
Cp ≈ ± (3.69)
λ
Note that a similar result would be produced had we chosen the function G(x +λy)
instead of F(x − λy). The plus and minus signs in (3.69) reflect the possibilities of
compression or expansion on a surface with the local angle θ in radians, in supersonic
flow, respectively. The determination of whether a surface is undergoing compression
or expansion by the flow is obvious as we note whether the flow is locally turning
‘into itself’ or is it locally turning ‘out of itself’ respectively. Here we note that at a
3.2 Linearized Compressibility Corrections 101
point on the body, as the (supersonic) Mach number increases the pressure coefficient
is reduced, by the 1/λ rule in the linear theory. We also note the singularity in C p as
freestream speed approaches sonic.
Now, let us apply the theory of linearized supersonic flow to a few examples.
Example 3.8 A flat plate is at an angle of attack, α, in a supersonic flow, as shown.
Apply the linear theory to this problem to derive an expression for the wave drag and
lift coefficient.
y M∞ >1 Flat Plate
Solution:
From the linear theory, the upper surface experiences a suction pressure and the lower
surface experiences a compression, both proportional to local wall angle, i.e., α,
2α 2α
C pu ≈ − and C pl ≈ + (3.70)
λ λ
Therefore, the flat plate experiences a normal force with the coefficient cn accord-
ing to
4α
Cn ≈ (3.71)
λ
Then, we may resolve the normal force in the lift and wave drag components through
cosine and sine of the plate angle of attack respectively and in the limit of small angles,
the cosine is replaced by 1 and the sine is replaced by α (in radians) to get the final
result, i.e.,
4α
cl ≈ (α in radians) (3.72)
λ
4α2
cd, w ≈ (α in radians) (3.73)
λ
The flat plate, according to linear theory, creates a lift-to-drag ratio in supersonic flow
that is inversely proportional to angle of attack, and is singular at α = 0, namely
L 1
≈ (3.74)
Dw α
102 3 Transonic Similarity Laws
Example 3.9 Apply the supersonic linear theory to the wavy wall problem (of
Example 3.6). Examine the behavior of the solution φ(x, y) and the wall pressure
coefficient, C pw (x). Also, calculate the wave drag coefficient for one wavelength
section of the wall.
Solution:
The equation for the wall is
2π
yw = h cos(kx) where k ≡
The flow on the upper surface of the wavy wall is described by
dyw φy
≈ (3.76)
dx V∞
At y = 0, as before, to get
λ
− hk sin(kx) = − F (x) (3.77)
V∞
hV∞
F(x) = − cos(kx) (3.78)
λ
We can now generate the general solution for F(x − λy) as
hV∞
φ(x, y) = − cos[k(x − λy)] (3.79)
λ
We note that the perturbation potential solution (3.79) is periodic in x, bounded in y
and satisfies the boundary condition at the wall (here approximated at y = 0). The
pressure coefficient is
φx (y = 0) hk
C pw ≈ −2 = −2 sin(kx) (3.80)
V∞ λ
Note that the pressure coefficient on the wall varies as the sine function whereas the
wall is described by the cosine function. Therefore there is a 90◦ phase shift between
the wall and its pressure distribution, as shown below.
3.2 Linearized Compressibility Corrections 103
y w /h (λ /2hk) C p
1
0.5
kx
00 1 2 3 4 5 6 7
-0.5
-1
One of the fruits of linear theory is the principle of superposition. Therefore within
this framework, the aerodynamic forces, i.e., lift and drag, due to a thin symmetrical
body at angle of attack are divided into a separate symmetrical thickness problem and
a separate angle-of-attack problem on a flat plate (i.e., a body with zero thickness).
Since a symmetrical thickness problem at zero angle of attack makes no contribution
to lift, the lift of a thin symmetrical profile is modeled as the lift on a flat plate,
namely
4α
cl ≈ (3.72)
λ
The wave drag is the sum of the wave drag of a symmetrical profile at zero angle of
attack and the wave drag of a flat plate at an angle of attack.
4α2
cd, w = cd, th + cd, AoA = cd, th + (3.83)
λ
In fact, we may even add the friction drag coefficient, c f , to (3.83) to produce the
overall drag coefficient of a thin symmetrical profile in supersonic flow at angle of
attack, namely,
104 3 Transonic Similarity Laws
4α2
cd = c f + cd, th + (3.84)
λ
Example 3.10 Use the principle of superposition in the linear theory to develop
the drag√polar of a 10 % thick, symmetrical double-diamond airfoil, as shown, at
M∞ = 2. Vary the angle of attack between −2◦ and 12◦ and assume the friction
drag coefficient for the profile is 0.005. Also by drawing a tangent from the origin of
c L − c D coordinate system to the drag polar curve, identify the angle of attack that
corresponds to the maximum L /D . Graph L /D as a function of angle of attack
and verify this angle.
θ nose
1 2
α 3 4
M ∞= 2 t/c =10%
Solution:
The lift coefficient is calculated based on a flat plate, i.e., (3.72). The drag coefficient
is the sum of three drag coefficients, as in (3.84). The thickness contribution to drag
is calculated from the pressure coefficients on the four panels of the symmetrical
airfoil, namely
1 t t
cd, th = C p1 + C p3 − C p2 − C p4 = C p1 − C p2 (3.85)
2 c c
Here we used the symmetry of the profile and lumped the contribution of the lower
half of the airfoil with its upper half. The pressure coefficients C p1 and C p2 are
related to the angles in radians that those panels make with respect to x-axis, i.e.,
± 21 vertex angle, respectively. The half-vertex angle, φnose , is
Therefore,
2(0.09976)
C p1 =√ ≈ 0.19934
2−1
−2(0.09976)
C p2 = √ ≈ −0.19934
2−1
t
cd, th = C p1 − C p2 ≈ 0.039868
c
A simple spreadsheet calculation produces the following table (Table 3.1). and
the associated graphs.
3.2 Linearized Compressibility Corrections 105
Table 3.1 Lift and drag characteristics of a diamond airfoil in supersonic flow [linear theory]
α (deg) α (rad) cd cl L/D
−2 −0.034917 0.04974 −0.1396 −2.8070
0 0.000000 0.04487 0.0000 0.0000
2 0.034917 0.04974 0.1396 2.8070
4 0.069813 0.06436 0.2793 4.3387
6 0.104726 0.08873 0.4189 4.7207
8 0.139636 0.12285 0.5585 4.5463
10 0.174533 0.16671 0.6981 4.1876
12 0.209440 0.22033 0.8378 3.8023
The tangent to the drag polar from the origin finds the maximum L/D, on the drag
polar curve, which from our graph is α ∼ 6◦ .
Lift-Drag Polar
10% Thick, Double-Wedge Airfoil
1
0.8
(L/D) max
Lift Coefficient, cl
0.6
0.4
0.2
0
0.00 0.05 0.10 0.15 0.20 0.25
-0.2
Drag Coefficient, cd
The lift-drag ratio is also plotted versus the angle of attack in the following figure,
which confirms the maximum L/D for a 10 % thick symmetrical double wedge
occurs at α ∼ 6◦ .
Lift-to-Drag Ratio
10% Thick, Double-Wedge Airfoil
6
4
L/D
2
Angle of Attack, degrees
0
-2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12
-2
-4
-6
106 3 Transonic Similarity Laws
Assume the axis of a body of revolution lies along the x-axis. The y- and z-coordinates
then describe the cross section of the body. To simulate the flowfield, sources and
sinks may be distributed along the x-axis to create the body of interest, following
the incompressible flow theory. The velocity field is repeated
V = V∞ i + ∇φ (3.86)
where φx /V∞ , φ y /V∞ and φz /V∞ 1 are based on our linearizing assumptions.
The governing perturbation equation for incompressible flow is known as the Laplace
equation, i.e.,
∇ 2 φ = φx x + φ yy + φzz = 0 (3.87)
where the elementary solutions are tabulated and superposed to create the solution
of interest. For example, a 3-D source (or a sink) flow is an elementary solution to
the Laplace equation with the following form (where source is placed at the origin):
−Q −Q
φ(x, y, z) =
= (3.88)
4π x2 + y2 + z2 4πr
where Q is the source strength, i.e., the volume flow rate from the 3-D (incompress-
ible) source. The flowfield associated with a source at the origin is thus purely radial
with
∂φ Q
Vr = = (3.89)
∂r 4πr 2
The denominator of (3.89) is the surface area of a sphere of radius r and the product
of the area and the normal velocity component (Vr ) then constitutes the volume flow
rate, from the source as expected
Figure 3.11 shows a slender body of revolution with a pointed-nose in uniform
flow. We define slenderness and fineness ratios for the body shown in Fig. 3.11.
V∞ x
d
z
These are:
d
τ= slenderness ratio (3.90)
F= fineness ratio (3.91)
d
The governing equation (3.10) for a linearized subsonic flow is:
(1 − M∞
2
)φx x + φ yy + φzz = β 2 φx x + φ yy + φzz = 0 (3.92)
We may divide the governing equation by β 2 (and absorb the β in the y- and z-
derivatives) to get:
φx x + φ(β y)(β y) + φ(β z)(β z) = 0 (3.93)
−Q −Q
φ(x, y, z) =
=
(3.94)
4π x2 + β2 y2 + β2 z2 4π x 2 + β 2 r 2
Thus the constant potential surfaces are evidently ellipsoids of revolution centered
on the source.
In the supersonic linearized case, we have the following governing small pertur-
bation equation, i.e.,
(M∞
2
− 1)φx x − φ yy − φzz = λ2 φx x − φ yy − φzz = 0 (3.96)
3 -D Source Circle,
at origin y
y 2 +z 2 = r 2
r
M∞ x
By analogy to the incompressible source flow solution, we get the so-called super-
sonic source flow potential, which satisfies our governing equation, namely
−Q −Q
φ(x, y, z) =
= √ = φ(x, r ) (3.98)
4π x2 − λ2 y 2 − λ2 z 2 4π x 2 − λ2 r 2
where r = y 2 + x 2
The constant potential surfaces for supersonic source are described by:
x 2 − λ2 r 2 = const. (3.99)
There are two asymptotes x = ±λr where the hyperboloids of revolution tend to as
the solution of Eq. (3.99). The slopes of dr/dx = ±1/λ describe the Mach cones
emanating from the point source both in the upstream and downstream directions, as
shown in Fig. 3.13. However, due to the existence of the zones of action and silence
associated with supersonic flow, the solution bounded by the Mach cone upstream of
the source, i.e., dr/dx = −1/λ, is not acceptable and only the solution downstream
of the point source is acceptable and is considered physical, i.e., within the zone of
action of the source with dr/dx = +1/λ.
To simulate a flow about a body of revolution we need to distribute sources/sinks
along the axis of the body. The potential at point (x, r ) due to a source/sink distrib-
ution of strength f (ξ) spread over a dξ extent along the x-axis is:
f (ξ)dξ
dφ(x, r ) =
(3.100)
4π (x − ξ)2 − λ2 r 2
Fig. 3.13 A Mach cone downstream of a source in a supersonic flow and a constant potential
surface, φ(x, r )
3.2 Linearized Compressibility Corrections 109
dξ
ξ x-ξ
x
Note that a minus sign is now absorbed in the general description of the source/sink
distribution, f (ξ). The parameters in (3.100) are shown in Fig. 3.14.
The overall contribution of the source/sink distribution along the x-axis is obtained
by integrating (3.100) from the leading edge (i.e., x = 0) to the vertex of the Mach
cone on the x-axis that intercepts the point (x, r ), as shown in Fig. 3.15.
Hence the potential at (x, r ) is
x−λr f (ξ)dξ
φ(x, r ) =
(3.101)
0 4π (x − ξ)2 − λ2 r 2
The streamwise perturbation of the flow due to the source distribution along the
x-axis is û(x, r ) = φx (x, r ), or
x−λr f (ξ)dξ
û(x, r ) =
(3.102)
0 4π (x − ξ)2 − λ2 r 2
Fig. 3.15 The segment (x − λr ) contributes to the potential at (x, r ) in a supersonic flow
110 3 Transonic Similarity Laws
û φx
C p ≈ −2 = −2 (3.104)
V∞ V∞
1 x−λr f (ξ)dξ
Cp = −
(3.105)
2πV∞ 0 (x − ξ)2 − λ2 r 2
The flow tangency condition on the body (with body radius R(x)) demands the body
slope and flow inclination to match, namely,
dR v̂ v̂ 1 x−λR x − ξ f (ξ)dξ
= ≈ =−
(3.106)
dx V∞ + û V∞ 4πV∞ 0 R (x − ξ)2 − λ2 r 2
In the limit of slender body, λR is much smaller than x − ξ, and also x, hence the
integrand and the limit of the integral simplify to:
dR 1 x f (x)
≈− f (ξ)dξ = − (3.107)
dx 4π RV∞ 0 4π RV∞
Hence the strength of the source/sink distribution f (x) is related to the slope of
the body, d R/dx, the body radius, R and the freestream velocity, V∞ . We may cast
(3.107) in terms of the body cross sectional area variation, dS/dx, as
dR dS
f (x) ≈ −4πV∞ R = −2V∞ (3.108)
dx dx
Note that the negative sign that we absorbed in the definition of f (x) is now recovered
in (3.107) or Eq. (3.108). Based on the slender body approximation, we may express
the perturbation potential, the streamwise and the lateral velocity perturbations as
well as the pressure coefficient in the flowfield of a slender body in supersonic flow as:
V∞ x−λr dS dξ
φ(x, r ) = −
(3.109)
2π 0 dξ (x − ξ)2 − λ2 r 2
V∞ x−λr d2 S dξ
û(x, r ) = − 2
(3.110)
2π 0 dξ (x − ξ)2 − λ2 r 2
V∞ x−λr d S x − ξ
2 dξ
v̂(x, r ) = − 2
(3.111)
2π 0 dξ r (x − ξ)2 − λ2 r 2
1 x−λr d2 S dξ
C p (x, r ) =
(3.112)
π 0 dξ 2 (x − ξ)2 − λ2 r 2
Now, let us apply these results to a slender cone in supersonic flow. The cone of
semi-vertex angle δ, has a body radius R(x) according to:
3.2 Linearized Compressibility Corrections 111
R(x)
R(x) = x tan δ ∼
= xδ → ≈δ (3.113)
x
Therefore the cross sectional area variation S(x) along the axis of the cone is calcu-
lated from:
S(x) ≈ πx 2 δ 2 (3.114)
This is a quadratic in x; hence the second derivative of cross sectional area of the
cone is constant, namely
d2 S ∼
= 2πδ 2 (3.115)
dξ 2
Now, substitute the second derivative of area in the pressure coefficient integral
(3.112) to get:
x−λR
dξ λR
C p, cone ≈ 2δ 2
= −2δ 2 ln √
0 (x − ξ)2 − λ2 R 2 x+ x 2 − λ2 R 2
(3.116)
Here we conclude that slender cones in supersonic flow obey the following similarity
rule:
C p, cone
= f (λδ) (3.118)
δ2
This is essentially the Göthert rule for axisymmetric bodies, where δ is replaced by
the characteristic thickness ratio of the body, τ. The similarity parameter in (3.118) is
λδ. Therefore, slender cones with different semi-vertex angles, δ1 and δ2 in different
supersonic flows, with M∞1 and M∞2 , will have their cone surface pressures related
by
C p, cone1 δ2
= 12 (3.119)
C p, cone2 δ2
We may also apply the linear supersonic slender body theory to a parabolic body of
revolution, as a second example. The geometry of the parabolic body of revolution,
in the meridian plane, is shown in Fig. 3.16.
The equation for body radius, R, as a function of body axis coordinate, x, is
2 2
R(x) = − (x − 1/4) (3.121a)
F
dR 4x
=− (3.121b)
dx F
d2 R 4
= − = const. (3.121c)
dx 2 F
Here F is defined as the ratio of body length to maximum diameter, or the fineness
ratio, i.e., the inverse of slenderness ratio. Note that the body length is normalized to
one, which serves as a reference length scale. The cross sectional area of the body is
that of a circle or S = π R 2 , which differentiates into
dS dR
= 2π R (3.122)
dx
dx
2
d S dR 2 d2 R
= 2π +R 2 (3.123)
dx 2 dx dx
Now, if we substitute the body area distribution in (3.112) for the pressure coeffi-
cient and integrate the equation we get a closed form solution for the C p distribution.
Note that the new limits of the integral should begin at x = −0.5 instead of zero that
we had in our derivation, which requires x to be replaced by x + 0.5 to represent a
shift to the left on the x-axis, namely
1 x+0.5−λr d2 S dξ
C p (x, r ) =
(3.124)
π −0.5 dξ 2
(x + 0.5 − ξ)2 − λ2 r 2
Again, upon substitution for d2 S/dξ 2 in (3.124) from (3.123), we can evaluate the
integral to get the surface pressure coefficient on a parabolic body of revolution in
supersonic flow, namely
r
Parabola
M∞
x
-0.5 +0.5
=1
A graph of (3.125) for a parabolic body of revolution for a fineness ratio of 10 and
freestream Mach number of M∞ = 1.4 is shown in Fig. 3.17. Also the pressure
distribution on a 2-D parabolic body is also graphed in Fig. 3.17 for comparative
purposes.
The two-dimensional supersonic linear theory predicts:
2θ d R/dx
Cp = ≈2 (3.126)
λ λ
With d R/dx = −4x/F, we get:
8
Cp ∼
=− x (3.127)
Fλ
This equation indicates a linear distribution of C p with x. Also for F = 10, λ = 1,
we have
Cp ∼= −0.8x (3.128)
At the leading edge, we are at x = −0.5, therefore the pressure coefficient at the
nose is 0.4, as graphed in Fig. 3.17. This profile is also symmetrical with respect to
the mid chord. The three dimensional case provides for aft body compression due
to a shrinking streamtube area in 3-D and a weaker nose pressure rise due to the
well-known 3-D relieving effect. The incompressible solution over a symmetrical
body of revolution creates a symmetrical pressure distribution about the mid chord.
Fineness Ratio = 10
0.4 Tw
o Di Three Dimensional
me M ∞ = 1.4
ns
0.2 ion
al,
M
∞
=1
Cp 0 .4
Three Dimensional
-0.2 M∞ = 0
-0.4
-0.5 0 0.5
Fig. 3.17 Pressure distribution on two and three-dimensional bodies with parabolic contours
114 3 Transonic Similarity Laws
Also the bodies sharp nose and the trailing edge in incompressible flow provide for
singularity at the sharp corners, as indicated in the graph of Fig. 3.17 for M∞ = 0.
Note that the term (in 3.129) can be made arbitrarily small when the freestream Mach
number is made arbitrarily close to 1. Consequently, the small perturbation equation
for steady, three-dimensional irrotational flows assumes the following form in the
transonic limit,
φx
(1 − M∞ 2
)φx x + φ yy + φzz = M∞2
(γ + 1) φx x (3.131)
V∞
The non-linear term on the right-hand side of (3.131) speaks to the nature of transonic
flow. Combining the left-hand side term with the right (in φx x ), we get the transonic
small disturbance equation:
2 φx
(1 − M∞ ) − (γ + 1)M∞
2
φx x + φ yy + φzz = 0 (3.132)
V∞
Figure 3.18 shows the pressure coefficient approaching the sonic flow from the
subsonic side and from the supersonic side, where linear theory was deemed valid.
3.3 Transonic Small Disturbance Theory 115
Linearized subsonic
∞ ∞ Nonlinear Transonic flow regime
flow regime -Cp
1
Cp ∼
β
Linearized supersonic flow regime
1
Cp ∼
λ
- Cp, 0
0 M∞
0 1 2 3
Fig. 3.18 Variation of pressure coefficient with Mach number according to linear theory
(1 − M∞
2
)φx x + φ yy + φzz = 0 → φ yy + φzz = 0 as M∞ → 1 (3.133)
Equation (3.133) in the limit of sonic flow shows a lack of dependence on the stream-
wise (i.e., x-) direction. This indicates that the disturbance caused by a body in sonic
flow propagates laterally and stays confined in the streamwise direction. This pic-
ture of near sonic flow helps us understand an important character of transonic flow.
Figure 3.19 shows a cartoon of a body in sonic flow where the disturbance caused
by the body laterally extends to large distances whereas the axial extent is limited to
the body length.
The transonic small disturbance equation (3.132) does not lend itself to analytical
solution; therefore, we seek to establish similarity laws that would hold on bodies that
are affinely-related, but at different transonic Mach numbers and different thickness
or slenderness ratios (and different angles of attack). First, let us define affinely-
related bodies. Figure 3.20a shows a body of maximum thickness-to-chord ratio, τ1
y Stream
surfaces
M∞ =1
Fig. 3.19 A slender body in sonic flow showing extensive lateral and limited axial extent of dis-
turbance
116 3 Transonic Similarity Laws
(a) (b)
τ1 = t1 / t2 τ 2 = t2 /
y y
ys(x) t1 ys(x)
x x
Fig. 3.20 Affinely-related profiles with different thickness/slenderness ratios. a Definition sketch
for a body with thickness ration τ1 . b An affinely-related body to (a) but with thickness ratio τ2
and in Fig. 3.20b we note that the same body is stretched in the y-direction and it
has attained a thickness-to-chord ratio of τ2 .
We define the local body slope as the product of a dimensionless function h(x/)
and the body thickness (or slenderness) ratio, τ, as
dys
= h(x/)τ (3.134)
dx
Although the example profile shown in Fig. 3.20 is symmetrical, i.e., it has zero
camber, affine transformation of cambered profiles will change camber in proportion
to thickness ratio as well. If we define a profiles camber in the x–y plane as yc /
(recall that camber in two-dimensional profiles is customarily defined as z c /c, in x–
z plane, where c is the chord length), then affine transformation adjusts or changes
camber according to (using our notation in x–y plane):
yτ 1/3
y= (3.139)
zτ 1/3
z= (3.140)
φ
φ= (3.141)
V∞ τ 2/3
Note that the dimensionless lateral coordinates, y and z are scaled down by the factor,
τ 1/3 , as compared to the streamwise coordinate, x, with the intention of placing them
all on the same order of magnitude. The choice of dimensionless perturbation poten-
tial φ yields the same governing equation and boundary conditions for affinely-related
bodies, as we shall see. For a rigorous proof of these choices of transformed coordi-
nates, the reader may consult Ref. [9]. The transonic small disturbance equation is
now expressed in terms of the transformed coordinates and perturbation potential as:
(1 − M∞ 2 )
− (γ + 1)M ∞ x φx x + φ y y + φzz = 0
2
φ (3.142)
τ 2/3
The first group in the bracket in (3.142) is identified as a transonic similarity para-
meter, K , i.e.,
(1 − M∞ 2 )
K ≡ (3.143)
τ 2/3
Note that based on this definition, K can be both positive and negative. For example,
it becomes positive when M∞ < 1 and negative when M∞ > 1. The second group
in the bracket may be simplified to
(γ + 1)M∞
2
φx ≈ (γ + 1)φx (3.144)
The boundary condition on the solid body is the flow tangency condition, which
matches the body to flow slope, namely
dys v̂
≈ (3.146)
dx V∞
dys
= τ h(x) (3.147)
dx
Now the right-hand side of (3.146) may be expressed in terms of the transformed
coordinates as
φx
C p ≈ −2 (3.150)
V∞
3.4 Transonic Similarity Parameters 119
This produces the transonic similarity rule for the pressure coefficient, C p , namely
Cp Cp
≈ −2φx = f (K , x, y, z) → = f (K , x, y, z) (3.152)
τ 2/3 τ 2/3
The implication of this similarity rule is that
C p1 C p2
2/3
= 2/3
(3.153)
τ1 τ2
On two bodies that are affinely-related in transonic flow, i.e., at the same x, y and z
locations, when the two transonic flows have K 1 = K 2 and the angle of attacks on
the two bodies follow (3.136), namely
α1 α2
=
τ1 τ2
Now, let us put the transonic similarity rule, (3.153), to use in the following example.
Solution:
First, let us calculate the transonic similarity parameter, K , namely
2 )
(1 − M∞1
K1 ≡ 2/3
≈ 0.88190
τ1
Now, let us consider an affinely-related airfoil with thickness τ2 in Mach 0.95 flow
at an angle of attack, α2 . In order to have the same transonic similarity parameter,
K 2 = K 1 , the freestream Mach number has to be:
2 )
(1 − M∞2
K2 ≡ 2/3
≈ 0.88190
τ2
120 3 Transonic Similarity Laws
Therefore the pressure coefficient at the peak suction point in Mach 0.95 flow is
−0.2822. The following graph shows the schematic drawing of two transonically
similar flows (not to scale).
K=0.88190 y K=0.88190
y τ ≈ 0.03676
τ = 0.10 α ≈ 1.47 o Cp, min ≈ -0.2822
α= 4
o
Cp, min = - 0.550
M∞=0.90 M∞= 0.95
x x
The complete transonic small disturbance equation is re-written here for reference.
(1 − M∞ 2 )
− (γ + 1)M∞ φx φx x + φ y y + φzz = 0
2
τ 2/3
There are different forms of transonic similarity parameter that are possible, which
are all identical at M∞ = 1, but differ at Mach numbers slightly below and above
the sonic point. For example, as we noted earlier, the product of (γ + 1)M∞ 2 in the
3.4 Transonic Similarity Parameters 121
bracket in (3.142) may be replaced with (γ + 1), in the limit of sonic flow and that
led to the definition of the transonic similarity parameter K in (3.143). Here, we first
introduce the transonic similarity parameter that Von Kármán proposed, namely
1 − M∞
2
K ≡ 2/3 (3.154)
γ+1 2
2 2 M∞ τ
What if we take (1 − M∞ 2 ) in the numerator of the Von Kármán parameter and write
It may be shown that the pressure coefficient then takes on the form
C p [(γ + 1)/2]1/3
= f 1 (K , x, y, z) (3.157)
τ 2/3
Although the derivation of the expression (3.157) for pressure coefficient is beyond
the scope of this book, the reader may consult Ref. [9] for detailed derivation. We also
note that (3.157) is more general than the basic expression (3.152), as it accounts for
the effect of different gases through ratio of specific heats, γ. The two expressions for
similarity pressure coefficients share the same functional dependency on slenderness
ratio, τ, namely τ −2/3 .
We are now ready to examine the similarity principles in force and moment coef-
ficients on two dimensional thin/slender bodies in transonic flow. The lift coefficient
is written as
1
cl = C p dx = C p dx (3.158)
cl C
p
= dx = f 1 (K ) (3.159)
τ 2/3 τ 2/3
122 3 Transonic Similarity Laws
From the transonic similarity expression for 2-D wave drag coefficient, we note
that drag coefficient on two-dimensional thin bodies in transonic flow is proportional
to τ 5/3 . The 2-D wave drag then rises as τ 5/3 , which indicates the advantage of thin
profiles in transonic flow. The 2-D drag/lift ratio is proportional to τ as well, which
again points to the advantage of thin bodies in transonic flight.
The pitching moment coefficient about the leading edge is
1
cm, LE = − C p xdx = − C p xdx (3.163)
2
Solution:
First, we establish the transonic similarity parameter, K, according to (3.156), namely
1 − M∞1 1 − 1.175
K = 2/3
= ≈ −0.8123
τ1 0.12/3
2/3
M∞2 = 1 − K τ2 = 1 + 0.81230.152/3 ≈ 1.229
Now, we need to examine the angle of attack on the second airfoil, as it is adjusted
by affine transformation, i.e.,
α2 α1 τ2
= → α2 = α1 = 3◦ (1.5) = 4.5◦
τ2 τ1 τ1
.
τ M∞ α cl
0.1 1.175 3◦ 0.12
0.15 1.229 4.5◦ 0.1572
1 − M∞2
K = (3.165)
(γ + 1)τ 2
124 3 Transonic Similarity Laws
Cp
= f 5 (K , x, y, z) (3.166)
τ2
The wave drag coefficient for a slender axisymmetric body is thus
1 dy 1 dy 1
CD = C p (2π y) dx ∼ 2 2 C p y dx ∼ 2 2 C p τ x[τ h(x)]dx
A 0 dx τ 0 dx τ 0
(3.167)
After cancelling the τ 2 from numerator and denominator of (3.167) and taking the
2 in the integrand to change x to x, we get
1 CD 1 C
h(x)xdx = f (K )
p
CD ∼ C p h(x)xdx → ∼ (3.168)
0 τ2 0 τ2
CD
= f 6 (K ) (3.169)
τ2
C D ∼ d 2 and Dw ∼ d 4 (3.170)
Therefore, the wave drag of a projectile rises as the fourth power of its diameter in
transonic regime, which strongly supports the notion (or necessity) of slender bodies
in transonic flight.
For a planar body, e.g., a wing, the transonic similarity parameter is:
1 − M∞
K = 2/3 (3.171)
γ+1
2 τ
The pressure coefficient is the same as the 2-D flows, but affine transformations
change the wing aspect ratio, A, in addition to the angle of attack and thus the
following three parameters are kept constant on two wings that are affinely-related,
namely:
1 − M∞1 1 − M∞2
= (3.172a)
[(γ1 + 1)τ1 ] 2/3 [(γ2 + 1)τ2 ]2/3
α1 /τ1 = α2 /τ2 (3.172b)
Thus the pressure coefficient for a wing of thickness ratio, τ, aspect ratio, A, span
of b, ratio of specific heats, γ and angle of attack, α is:
3.5 3-D Planar and Axisymmetric Slender Bodies 125
4
Predicted theoretical
critical Mach Number
3 2.5
AR(t/c)1/3= 3.0
(CD p)min 2.0
5/3 1.5
(t/c) 2
1.0
1 .5
.25
0
-2.0 -1.6 -1.2 -.8 -.4 0 .4 .8 1.2
M2- 1
2/3
(t/c)
Fig. 3.21 Curves of the generalized drag coefficient for symmetrical wings for γ = 1.4 (after
Ref. [6])
τ 2/3
C p |α/τ ; A(γ+1)1/3 τ 1/3 = f 7 (K , x, y, z, z/b) (3.173)
(γ + 1)1/3
τ 2/3
C L |α/τ ; A(γ+1)1/3 τ 1/3 = f 8 (K ) (3.174)
(γ + 1)1/3
τ 2/3
C M |α/τ ; A(γ+1)1/3 τ 1/3 = f 10 (K ) (3.175)
(γ + 1)1/3
τ 5/3
C D |α/τ ; A(γ+1)1/3 τ 1/3 = f 9 (K ) (3.176)
(γ + 1)1/3
The transonic zero-lift wave drag coefficient for unswept wings with a symmetrical
NACA 63 profile is presented in Fig. 3.21, as presented by McDevitt [6]. We recognize
the similarity parameters A(t/c)1/3 and (1 − M∞ 2 )/τ 2/3 in the graph as well as the
drag coefficient divided by τ . For a given thickness to chord ratio, aspect ratio, and
5/3
freestream Mach number, this graph allows us to calculate the wave drag coefficient.
A similar graph can be found in the USAF Datcom [3] for general airfoils with
rounded leading edges. We will show in Chap. 8 how this method can be extended
to calculate the wave drag of swept wings at transonic speeds.
126 3 Transonic Similarity Laws
∂ û ∂ v̂ (γ + 1)M∞
2 ∂ û
(1 − M∞
2
) + = û (3.177)
∂x ∂y V∞ ∂x
∂ û ∂ v̂
− =0 (3.178)
∂y ∂x
The system of two equations and two unknowns in û and v̂, as described by
Eqs. (3.177) and (3.178), is still nonlinear. The independent variables are x and
y, which describe the coordinates in the physical plane. However, if we switch the
dependent variables with the independent variables, the resulting set of equations
become linear! This is the essence of Hodograph transformation in transonic small
disturbance theory. From the physical domain we have
û = û(x, y)
v̂ = v̂(x, y) (3.179)
∂ û ∂ û
dû = dx + dy = û x dx + û y dy (3.180)
∂x ∂y
∂ v̂ ∂ v̂
dv̂ = dx + dy = v̂x dx + v̂ y dy
∂x ∂y
v̂ y dû − û y dv̂
dx = (3.181)
û x v̂ y − û y v̂x
−v̂x dû + û x dv̂
dy =
û x v̂ y − û y v̂x
x = x(û, v̂)
y = y(û, v̂) (3.183)
A comparison between the two sets of Eqs. (3.182) and (3.184) shows that
v̂ y −û y
xû = and xv̂ =
Δ Δ
v̂x −v̂x
yû = and yv̂ = (3.185)
Δ Δ
Now, if we substitute these in (3.177) and (3.178), we get the new governing equation
set as:
2 ∂y ∂x (γ + 1)M∞ ∂y
2
(1 − M∞ ) + = û (3.186)
∂ v̂ ∂ û V∞ ∂ v̂
∂x ∂y
− =0 (3.187)
∂ v̂ ∂ û
The new set of governing equations, known as transonic hodograph equations are
now linear in x and y, as independent variables where the new dependent variables
are û and v̂. This clever mathematical exercise has two impediments in real life
applications. First, the physical boundary conditions are always expressed in terms
of x and y as independent variables and not the û and v̂. Second, the solution can only
exist if the determinant of the coefficient matrix Δ
= 0. The determinant may only
be zero in supersonic flows and the loci of all the points with Δ = 0 is called the limit
line. Except for simple boundaries and geometries (as in a wedge), the application
of hodograph transformation to transonic problems is not practical. In addition, with
the advances in Computational Fluid Dynamics (CFD), the nonlinearity of transonic
flow poses no problem to achieving high-fidelity simulation or solution. For more
details on hodograph transformation such classical texts as Shapiro [9] and Liepmann
and Roshko [5] may be consulted.
There are many empirical rules that were developed prior to the advent of supercritical
airfoil sections and area rule in transonic flow in the mid 1950s. Most such rules
involve the old profiles and thus do not translate into a useful practice today in
128 3 Transonic Similarity Laws
10 Typical Unswept
High Aspect Ratio Wings
Lift-curve slope, CL (rad-1)
9
8 Thin Airfoil
Thick Airfoil
α
7
Typical Swept Wing
6
High Aspect Ratio
5 Low Aspect Ratio
4
3
2
1
0
0 .5 1.0 1.5 2.0 2.5 3.0
Mach number, M (~)
3.7 Empirical Rules 129
α
3.
Experiment
2.
Computation
1. Linear Potential
0.
0. 0.5 1.0 1.5 2.0 2.5
Aerodynamic Center
65
aerodynamic center,
60
xac/cr (%)
Computation
55 Linear Potential
Experiment
50
0. 0.5 1.0 1.5 2.0 2.5
Mach number, M (~)
Fig. 3.23 Lift curve slope and aerodynamic center for the Boeing B2707-300 supersonic transport
aircraft (after Ref. [12])
(a) (b)
Detached shock M>1 Detached shock M>1
Sonic line Sonic line
M<1 M<1
M∞ > 1 M∞ > 1
Fig. 3.24 Two examples of a blunt body in supersonic and upper transonic flow. a Blunt body with
a sharp corner (Discontinuous curvature). b Blunt body with continuous curvature
continuous curvature however, as in Fig. 3.24b, there is no simple rule that identifies
the sonic boundary on the body. However, there are two empirically supported ideas
on the location of the sonic boundary on two-dimensional and axisymmetric bodies.
These are (see [7] or [9]):
• The location on the 2-D body where θb = θmax is approximately the sonic boundary
on the body. Here, θb is the body angle and θmax is the maximum turning angle,
or flow deflection angle, corresponding to an attached plane oblique shock at the
freestream Mach number, M∞ .
• The location on the axisymmetric body where θb = θmax is approximately the
sonic boundary on the body. Here, θmax is the maximum turning angle, or flow
deflection angle, corresponding to an attached conical shock at the freestream
Mach number, M∞ .
130 3 Transonic Similarity Laws
(a)
90o
(b)
Approximate
Location of Sonic
Boundary θ b = θ max
σ
M∞> 1
µ
M∞ > 1
θ θ max
Fig. 3.25 Definition sketch of the approximate location of the sonic point on a blunt body [σ is
the shock wave angle, θ is the flow turning angle and θmax is the maximum turning angle]. a Shock
wave angle-deflection diagram boundary. b The approximation location of the sonic boundary
The bow shocks that are created in supersonic flow ahead of blunt bodies vary in
strength from a normal shock to a Mach wave. The normal shock represents the
strongest point on the wave and it is located on the stagnation streamline. The Mach
wave represents the weakest point, which is asymptotically arrived at a large distance
from the body. Therefore, we may assume the basic shape of a detached shock is
approximately a hyperbola, with the Mach wave (of angle μ) running as its asymptote
(see Fig. 3.26).
y
Asymptote y = (tan μ ) x; Equation of where:
tan−1 β asymptote μ ≡ sin−1 (1/ M ∞ )
y =(tan μ ) x 2 −x 02 ; Equation of β ≡ M∞2 − 1
or hyperbola and
x0 is the position of wave vertex
M∞ > 1 µ β y = x 2 − x02
x0 x
Fig. 3.26 Hyperbola as the approximate shape of a detached shock wave in a supersonic blunt-body
flow
3.8 Approximate Location of Detached Shocks 131
η θmax
M∞ > 1
yc= 2y S /3 SB
ySB
yc= yS /2
x
0 x0 xSB
Streamline passing through L
the mass centroid in 2-D
Fig. 3.27 Definition sketch used in the calculation of the location of detached shock waves
132 3 Transonic Similarity Laws
1 1 ys ρ∞ V∞ ys2 /2 ys
yc ≡ y dṁ = y(ρ∞ V∞ ) dy = = (3.188)
ṁ ṁ 0 ρ∞ V∞ ys 2
These positions for the mass centroid are labeled in Fig. 3.27.
By applying the continuity equation to the sonic streamtube upstream of the
detached shock and the sonic flow through the sonic line downstream of the detached
shock, we establish the sonic flow area S − S B (normalizing with respect to the
reference length scale of the problem, namely, y S B in plane 2-D flows and reference
area scale, y S2 B in axisymmetric flows) and thus the length, L, where the vertex of
the detached shock is located upstream of the sonic boundary x S B . Step-by-step
calculation procedure based on the theoretical development of Ref. [7] is outlined in
Shapiro [9], and is not repeated here, for brevity. Once the position of the detached
shock is established, we may perform a momentum balance, in the streamwise, i.e.,
the x-direction, between the sonic streamtube upstream of the shock and the sonic
flow through the sonic line downstream of the shock. This results in an estimation
of nose pressure drag, upstream of the sonic line on the blunt body. Alternatively,
the momentum deficit in the x-direction between upstream and downstream of the
bow shock is equal to the pressure drag of the blunt nose, up to the sonic boundary.
Such calculations were performed by Ferri [2] and Moeckel [7], among others. The
momentum balance method as described here may also be applied to supersonic
inlets that operate in sub-critical mode, i.e., with external spillage (see for example,
Ref. [1]). The sub-critical mode corresponds to the operation of a supersonic inlet
where the normal (or terminal) shock is pushed outside the inlet and thus forms a
bow shock ahead of the inlet cowl lip. The momentum balance method yields inlet
cowl pressure drag in sub-critical operation, or alternatively, the spillage drag of a
supersonic inlet.
Now, we examine some results of the theoretical prediction of the detached shock
position in blunt-body supersonic and upper transonic flows and the experimental
data. Figure 3.28 (from Refs. [7, 9]) shows the detached shock position in supersonic
flow (i.e., Mach 1.7) in front of a cone, sphere and projectile.
First, we note that the shock position is well captured by the theory. Second,
we note that the location of the sonic line is reasonably well predicted, however,
the experimental data point to a curved sonic line, as compared to the straight line
assumption in the theory of Ref. [7]. Finally, we note that the location of the sonic
boundary on the three blunt bodies is nearly independent of the body shape upstream
of the sonic point. Note that the normalizing length scale is yS B , as discussed earlier.
3.8 Approximate Location of Detached Shocks 133
Shock M=1
2.4
Ladenburg
2.0
} (Experimental)
y Moeckel
ySB (Theoretical)
1.6
Projectile
1.2 Sonic
Line Sphere
0.8
Cone ( , )
0.4 Sphere ( , )
Projectile ( )
0
0 0.4 0.8 1.2 1.6 2.0 2.4
x-x0
ySB
(b) 1.5 d
L’ M∞ d (in.)
1/4
d
1/2
1.0 L’
1
(a) 18
Theoretical
16 Experimental 0.6
1.0 1.5 1.9
14 } Axi-Symmetric M∞
0 0.4
1.0 1.4 1.8 2.2 2.6 3.0 3.4 3.8 1.0 1.5 M∞ 1.9
M∞
Fig. 3.29 Detached shock position ahead of a axisymmetric and planar objects, b spheres and c
cones (after [9])
3.9 Summary
Transonic flow is more a physical description of the flow than a mere Mach number
range near sonic. From the lower end, i.e., the subsonic side, transonic flow is entered
at the critical Mach number. On the upper end, i.e., the supersonic side, transonic
flow persists until the profile is submerged entirely in supersonic flow. It is indeed
the mixture of subsonic and supersonic flow with the appearance of shocks on the
body that characterizes and dominates transonic aerodynamics. We learned that this
flow regime is inherently nonlinear and its analysis does not lend itself to full lin-
earization of the perturbation potential equation. Therefore, the solution techniques
are numerical in nature and the classical theoretical developments lead to a group of
transonic similarity parameters and similarity laws. These laws allow for extrapo-
lation of experimental data at different transonic flow conditions on affinely-related
bodies.
Problems
3.1 The peak suction pressure on an airfoil at 5◦ angle of attack is C p, min = −1.25
in low-speed flow. Calculate the pressure coefficient at the same point on the air-
foil, when the freestream Mach number is 0.7 using Prandtl-Glauert compressibility
correction.
3.9 Summary 135
3.2 The lift curve slope, clα , of an airfoil in incompressible flow is 1.95π rad−1 .
Calculate the lift curve slope for the same airfoil at Mach 0.75, using Prandtl-Glauert
correction.
3.4 Calculate the critical pressure coefficient, C p, crit for a range of Mach numbers
starting at 0 and extending to Mach 1. Graph the critical pressure coefficient versus
Mach number.
3.5 An airfoil has a C p, min = −1.01 in low speed. Extend this result to the com-
pressible domain, using three compressibility corrections:
Also, use the graphical technique to determine the critical Mach number for this
profile and flow condition.
M∞ Cp,min
γ =1.4
(a) At what free-stream Mach number, M∞ , will the minimum pressure coefficient
double?
(b) Calculate C p, crit , if the Mcrit = 0.72
(c) Calculate the lift curve slope at Mcrit if the low-speed lift curve slope is a0 =
2π rad−1 .
3.7 A thin airfoil is in a Mach 0.35 flow and develops peak suction, C p, min of −2.4
and a lift coefficient, cl of 0.78.
Calculate:
Cl.q∞ .c
M∞ Cp, min
136 3 Transonic Similarity Laws
3.8 Consider a subsonic compressible flow over a wavy wall of infinite extent. The
wall amplitude is h and its wavelength is . The wavy wall may be described by:
2πx
yw = h cos
Calculate:
(a) the streamwise velocity at (2, h)
(b) the velocity vector at (, 2h)
(c) the minimum and maximum pressure coefficient, C p, min and C p, max on the wall
Calculate:
(a) the Mach number, Mmax , on the airfoil, at C p, min
(b) the critical C p, crit
(c) stagnation point C p , C p, stag
(d) the incompressible value of (C p0 , min )
3.10 A thin airfoil has a minimum pressure coefficient of C p, min = −0.5 at low
speed, i.e., incompressible limit.
y
Cp, min
M∞
γ =1.4 x
Calculate:
(a) C p, min when M∞ = 0.5
(b) C p, stag when M∞ = 0.5
(c) M∞ when C p, min = −1.2
(d) the velocity vector, V/V∞ on the airfoil where C p = −0.60 and the airfoil slope
is θ = −15◦
3.9 Summary 137
Calculate:
(a) Mach number at the point of C p, min
(b) C p, max at M∞ = 0.7
(c) C p, max at M∞ = 0
(d) C p, min at M∞ = 0
V∞
T∞
Calculate:
(a) the critical Mach number
(b) the incompressible C p, min
(c) speed of sound at the critical point
(d) the ratio pcrit / p∞
Assume γ = 1.4 and R = 287 J/kg · K.
3.13 Consider a subsonic flow over a wavy wall as shown. The perturbation velocity
potential is known:
y
v
u y = h cos(2πx)
M ∞= 3 h <<1
2
x
a∞ =300 m/s
0 1/4 1/2 3/4 1
138 3 Transonic Similarity Laws
3.14 For a flat plate in a supersonic flow at an angle of attack, α, apply the linear
theory to calculate the angle of attack that corresponds to L /D = 2.
3.15 For a wavy wall in supersonic flow, the ratio of wall height-to-wave length,
h/ = 0.05 (see Fig. 3.8 for a definition sketch of wavy wall).
Calculate:
(a) the wave drag coefficient (for one wavelength) at Mach 2.0
(b) the wall height-to-wavelength ratio, h/ that gives the same wave drag coefficient
at Mach 3.
3.16 Using the principle of superposition in the linear theory, express the drag of
a 5 % thick, symmetrical double-diamond airfoil as the sum of its friction drag,
thickness drag and wave drag due to angle of attack (on a flat plate), √ similar to
Example 3.10. To develop the drag polar for this airfoil, assume M∞ = 5 and vary
the angle of attack between −2◦ and 12◦ . For a friction drag coefficient of 0.005,
calculate and plot the drag polar for this airfoil.
√
3.17 Consider a slender cone of semi-vertex angle δ = 10◦ in Mach M∞ = 2
flow. Using linear theory, estimate the pressure coefficient on the cone, C p, cone .
√
3.18 Consider a slender cone of semi-vertex angle δ = 5◦ in Mach M∞ = 5
flow. First show that this cone has the same similarity parameter, λδ, as the cone
described in Problem 3.17. Next, use the linear theory to calculate the ratio of pressure
coefficients on the two cones (i.e., the 5◦ and 10◦ cones).
Calculate:
(a) the pressure coefficient at the stagnation point
(b) the shock Mach number, Ms
(c) the pressure coefficient downstream of the shock
M ∞=2.0
Calculate:
(a) p A / p B
(b) Ratio of freestream dynamic-to-static pressure, q∞ / p∞
Cp, min, 2
t/c=6%
M∞2 α2
3.23 A 2-D body of 5 % thickness ratio, i.e., τ1 = 0.05 is in transonic flow with the
flow Mach number of M1 = 0.90. The pressure coefficient at point A on this body
is C p1= 0.12. Now consider an affinely-related body that has a thickness ratio τ and
is in a Mach 0.95 flow, i.e., M2 = 0.95.
Calculate:
(a) the transonic similarity parameter, K
(b) the thickness ratio of the second (affinely-related) body, τ2
(c) the pressure coefficient, C p2 , on the second body that corresponds to point A on
the first body
A A
M1 M2
140 3 Transonic Similarity Laws
3.24 Use the transonic similarity rule on wave drag coefficients (on two-dimen-
sional bodies) and compare the wave drag coefficients of two affinely-related air-
foils with 5 and 10 % thickness ratios, respectively. Assume the transonic similarity
parameter, K , remains constant between the two airfoils.
3.25 The lift coefficient of a 5 %-thick airfoil at the angle of attack of 2◦ in Mach
M∞ = 1.12 air stream (γ = 1.4) is cl = 0.10. Apply transonic similarity theory to
predict the lift coefficient of an affinely similar airfoil that is 10 % thick in transonic
flight through air (with γ = 1.4)?
θ nose 1 2
α = 6o 3
M∞ = 5 c
3.27 An airfoil in Mach 0.80 flow at 2◦ AoA experiences the pressure distribution,
as shown below (not to scale). Calculate:
(a) Maximum Mach number on the airfoil, Mmax
(b) Pressure coefficient at the stagnation point, C p, stag
(c) Critical pressure coefficient, C p, crit
(d) Shock static pressure ratio on the suction surface
(e) does the shock separate the BL? And if yes what is the extent of the separated
bubble? [Hint: look for the footprint of separation]
Cp, crit
1.0
x/c
0.35 0.70 T.E.
(+)
Cp, stag
M >1
3.9 Summary 141
3.28 For a family of slender axisymmetric bodies (that are affinely-related) in Mach
1 flow, use transonic similarity rule to compare their wave drag coefficients as a
function of their base diameters while keeping their length constant.
3.29 Consider a finite wing of aspect ratio A R1 = 8 and thickness ratio τ1 = 0.10
is in transonic flight at Mach 0.9 (with γ = 1.4) at an angle of attack of 6◦ . First,
calculate the transonic similarity parameter, K , for this wing. Next, consider an
affinely-transformed wing of 5 % thickness, (i.e., τ2 = 0.05), with γ = 1.4 and the
same similarity parameter, K 2 = K 1 . Using transonic similarity laws, calculate:
3.30 A detached shock is formed ahead of a cylinder in a Mach 1.2 flow. Estimate
the location of the sonic boundary on the cylinder. Assume γ = 1.4.
3.31 A detached shock is formed ahead of a sphere in a Mach 1.2 flow. Estimate the
location of the sonic boundary on the sphere. Assume γ = 1.4.
3.33 A detached shock is formed ahead of a 2-D blunt object in a Mach 1.6 flow.
Assuming the shape of the detached shock is described by a hyperbola; calculate
the shock angle, σc on the streamline passing through the mass centroid of the sonic
stream tube. Assume γ = 1.4.
Streamline passing y
through the mass σ
centroid in 2 -D flow
yS θs M=1
S
θmax
SB
yS /2 ySB
x
x0 xSB
3.34 Consider a transonic flow where the freestream Mach number is:
M∞ = 1 + ε where ε 1
142 3 Transonic Similarity Laws
Assuming a detached shock is formed, show that the Mach number downstream
of the normal shock is:
M2 ≈ 1 − ε
M∞ = 1 + ε M2 ≈ 1 - ε
3.35 To demonstrate that weak shocks that appear in transonic flow are nearly isen-
tropic, we choose the freestream Mach number to a normal shock to be:
M∞ = 1 + ε where ε 1
Use the normal shock relations to show that non-dimensional entropy rise across
the normal shock, Δs/R is of order of ε3 , i.e.,
Δs
∼ O(ε3 )
R
3.36 Concept of rotationality of the flow downstream of the curved shocks is derived
from Croccos theorem (for steady, inviscid flows), i.e.,
−V × (∇ × V) = ∇h t − T ∇s
V × (∇ × V) = T ∇s
Use the result of Problem 3.34 to show that the vorticity field downstream of curved
shocks in transonic flow is negligibly weak and thus the flow may be modeled as
irrotational.
3.37 A 2-D wedge of nose angle δ is in a supersonic flow. Apply the linearized
theory to the wedge and show that the expression for pressure coefficient may be put
in the transonic similarity form:
γ+1 2/3
[(γ + 1) /2] C p
1/3 δ
2
= 2
δ 2/3 M∞2 −1
3.9 Summary 143
3.38 Show that from linear theory in supersonic flow, a thin symmetrical bi-convex
airfoil creates wave drag according to:
2
16t
Cd, w =
3 M∞ − 1
2 c
3.39 Prove that for a given thickness-to-chord ratio, t/c, and the location of the
maximum thickness, a symmetrical diamond airfoil creates the lowest wave drag in
supersonic flow, i.e.,
2
t 4
(Cd, w )diamond = (Cd, w )min =
M∞ − 1 c
2
t
for = const.
c
The above expression represents the minimum wave drag. You may use linear theory
in your proof.
Hint: Wave drag coefficient for a general thin airfoil shape using linear theory is
derived to be (see for example Liepmann-Roshko, or Shapiro):
4
Cd, w =
A + B + α02
2 −1
M∞
where
2
1c dh
A≡ dx
c 0 dx
1c 2
B≡ α (x) dx
c 0 c
References
4.1 Introduction
Supersonic flow over bodies involves the creation of waves of compression and
expansion to achieve flow turning. Blunt bodies, for example, create a detached, or
bow shock that forms ahead of the blunt body and the flow downstream of the shock is
a mixed subsonic-supersonic flow. Such mixed flows are not amenable to analytical
solutions. However, there is a class of problems in inviscid supersonic flow over
pointed bodies that lend itself to exact analytical solution (known as shock-expansion
theory). This class of inviscid supersonic flow problems requires the oblique shock
(either plane or conical) to be attached to the body in flight at the leading edge/nose.
This necessitates first the body to be pointed and second the turning imposed by the
body at the leading edge not to exceed the maximum turning angle corresponding
to a plane oblique (or conical) shock at the designated flight Mach number. Given
these stipulations, shock-expansion theory may be applied to any inviscid flow above
sonic, which includes the upper transonic regime.
1
sin μ = (4.1a)
M
1
tan μ = √ (4.1b)
M2 − 1
1 1
μ = sin−1 = tan−1 √ (4.1c)
M M2 − 1
For example in a Mach-2 flow, the Mach angle is sin−1 21 , which is 30◦ . In sonic
flow at Mach 1, the Mach angle is 90◦ , i.e., normal to the flow and for Mach 1.1, the
Mach angle is reduced to 65.38◦ with respect to the local flow direction. However a
single Mach wave due to its infinitesimal strength is unable to provide finite turning
of the supersonic flow that is needed to make room for a body (of finite thickness) in
flight. Consequently, finite compression and expansion waves are created to turn the
supersonic flow. The coalescence of compression Mach waves creates shocks and
M=1
M ² –1
4.1 Introduction 147
(a) Oblique
(b) Streamline
Shock Expansion
Mach waves µ2
Compression M2 > 1
Mach waves 2
M2 > M1
1
2
M2 > 1
1 M1 > 1 µ
µ2 M2 < M 1 1
M1 > 1
µ1
Fig. 4.2 Waves on concave and convex surfaces in supersonic flow. a Waves generated on a concave
surface. b Waves generated on a convex surface
finite expansion waves are created through many expansion Mach waves, known as
the Prandtl-Meyer expansion waves. Figure 4.2a shows the coalescence of compres-
sion Mach waves on a concave surface that culminates in the shock formation and
Fig. 4.2b shows the spreading of the expansion Mach waves on a convex surface that
create finite turning in the flow at supersonic speeds.
We note that the compression Mach waves on a concave surface coalesce, as
shown in Fig. 4.2a, and form an oblique shock wave whereas the expansion Mach
waves on a convex surface (Fig. 4.2b) spread/diverge and thus never form a surface of
discontinuity of rarefaction. The coalescence of compression Mach waves that form
the shock thus creates a region of very steep pressure and temperature gradients in the
gas where normal viscous stress and thermal conductivity of the gas play an important
role in establishing the equilibrium thickness for the shock wave. It is therefore the
dominance of viscous effects within the shock that creates an entropy increase and
thus non-isentropic flow conditions across the shock. We also know from kinetic
theory that the thickness of the shock is proportional to the mean-free path of the gas
at reference sonic temperature condition (e.g., see classical textbooks of Shapiro [6],
Liepmann-Roshko [5], among others). The steep gradients in pressure, temperature,
density, normal component of velocity, entropy and stagnation pressure across the
shock are depicted graphically in Fig. 4.3.
The structure of a normal shock in standard sea level pressure and temperature con-
ditions with upstream Mach number of 1.1 is shown in Fig. 4.4 (from [6]). Subscripts
x and y denote the flow conditions upstream and downstream of the shock, respec-
tively. The streamwise coordinate x shows the region where the steep gradients exist.
Therefore, noting that the x-axis has a 10−5 inch scale, we appreciate the thickness,
or rather thinness of the shock in standard air. To put these numbers in perspective,
we recall that the mean-free path in air at standard conditions is ∼3.7 × 10−5 cm
(or ∼1.46 × 10−5 in.), which makes the thickness of the shock between 3–5 times
the mean-free path in air at standard conditions. The static pressure and temperature
across the normal shock are shown in ratios and the normal velocity component in
units of ft/s. The density ratio is not shown in the graph, but it rises as well as the
gas is compressed across the shock. It is thus the product of density and temperature
148 4 Shock-Expansion Theory
1 2 2
2 1 1
u2 p1 T1
2 2 1
1 1 2
s1 pt
1 2
Fig. 4.3 Schematic drawing of flow gradients across a shock, with shock thickness, δ (not-to-scale).
a Normal component of velocity. b Static pressure. c Static temperature. d Density. e Entropy.
f Total pressure
Fig. 4.4 Variation of flow properties within a normal shock (from [6]). Note that x indicates the
conditions ahead of the shock wave, while y indicates the conditions behind the shock wave
4.1 Introduction 149
ratios following the perfect gas law that produces the static pressure ratio across the
shock, as shown in Fig. 4.4. To demonstrate the variation of normal shock thickness
with Mach number at standard pressure and temperature, we examine Fig. 4.5 (from
[6]). The normal shock thickness on a log-log scale indicates an exponential drop in
shock thickness as shock strength or Mach number increases. As a result of shock
thinning with increasing Mach number (at a given altitude), the gradients increase
and thus total pressure losses mount.
The normal shock relations are derived through the application of the great con-
servation laws of nature, namely conservation of mass, linear momentum and the
energy across the shock. The conservation of mass applied to a normal shock with
stations 1 and 2 identifying up- and downstream of the normal shock, gives birth to
the first constant of motion, related to normal shock flows, namely
ρ1 u 1 = ρ2 u 2 = const. (4.2)
The application of the Newton’s second law of motion to a control volume shown
in Fig. 4.6, gives birth to the second constant of motion for a flow through a normal
shock, namely
p1 + ρ1 u 21 = p2 + ρ2 u 22 = const. (4.3)
Since the flow across a shock is adiabatic, i.e., there is no heat exchange between
an external source and the fluid in the control volume, and since the fluid within the
control volume undergoes no chemical reaction, the law of conservation of energy
for non-chemically reacting, steady flows across a shock produces the third constant
of motion, namely
150 4 Shock-Expansion Theory
h t1 = h t2 = const. (4.4)
Since the sonic state may be reached adiabatically in a gas, the total enthalpy of the
gas at sonic state that is reached adiabatically and the constant stagnation enthalpy
of the flow are equal, namely,
a ∗2 a ∗2 a ∗2 a ∗2 γ+1 a ∗2
h∗ + = cpT ∗ + = + = = h t = const. (4.6)
2 2 γ−1 2 γ−1 2
Any of the constant speeds, at , a ∗ or Vmax can serve as our reference speed in gas
dynamic calculations dealing with normal (or oblique) shocks. For example a ∗ has
served as the reference speed in shock polar and hodograph solutions in gas dynamics
and Vmax is used in Taylor-Maccoll formulation of conical shocks, as we will briefly
review. Also a ∗ is used in the definition of a characteristic Mach number, M ∗ using
the following definition:
4.1 Introduction 151
u
M∗ ≡ (4.9)
a∗
The characteristic Mach number, is related to the Mach number through the law of
conservation of energy equation, namely,
(γ + 1) M 2
M ∗2 = (4.10a)
2 + (γ − 1) M 2
Or inversely,
2
M2 = (4.10b)
(γ + 1) /M ∗2 − (γ − 1)
Note that subsonic flows with M < 1, have M ∗ < 1, sonic flow with M = 1, has
M ∗ = 1 and supersonic flow with M > 1 has M ∗ > 1. The added advantage of M ∗
over M is that M ∗ is bounded whereas M may tend to infinity. This is seen in the
following limit:
∗ γ+1
M → as M →∞ (4.10c)
γ−1
ρu = const.
p + ρu 2 = const.
h t = const.
at = const.
a ∗ = const.
Vmax = const.
The simultaneous solution of the conservation laws yields the celebrated Prandtl’s
relation relating the product of gas speeds across a normal shock to the square of the
(constant) speed of sound at sonic state, a ∗ , following
u 1 u 2 = a ∗2 (4.11)
The immediate result of the Prandtl’s relation is the inverse relationship between the
upstream and downstream characteristic Mach numbers, namely
1 1
M1∗ = ∗ or M2∗ = ∗ (4.12)
M2 M1
upstream Mach number, M, we may express the “jump” conditions across a normal
shock as follows:
ρ2 u1 (γ + 1) M12
= = (4.13)
ρ1 u2 2 + (γ − 1) M12
p2 2γ 2
=1+ M1 − 1 (4.14)
p1 γ+1
T2 2γ 2 2 + (γ − 1) M 2
= 1+ M1 − 1 1
(4.15)
T1 γ+1 (γ + 1) M12
s2 − s1 γ 2γ 2 2 + (γ − 1) M 2
= ln 1 + M1 − 1 1
R γ−1 γ+1 (γ + 1) M12
2γ 2
− ln 1 + M1 − 1 (4.16)
γ+1
⎡ ⎤ γ
γ−1
2+(γ−1)M12
2 + (γ − 1)
pt2 Δs 2γ ⎢ 2γ M 2 −(γ−1) ⎥
= e− R = 1 + M12 − 1 ⎢ ⎥
1
(4.17)
The graphical depiction of the flow properties across a normal shock is very instruc-
tive and the levels and scales of these parameters may be used as an instructional
tool. For example in Fig. 4.7 we have plotted the static pressure, temperature and
density ratios across a normal shock for a calorically-perfect gas with the ratio of
specific heats, γ = 1.4. Note that the upstream Mach number range is intentionally
limited to the supersonic range of up to Mach 5.
We note that the level of pressure jump is higher than both density and temperature
ratios, since in fact it is the product of the two jumps (in density and temperature)
that create the pressure jump. In the transonic range, we note that the behavior of
these jumps in static pressure, temperature and density is nearly linear. Limiting the
Mach number range to 1.3 and re-plotting the ratios of pressure, temperature and
density in Fig. 4.8 we can show this linear behavior.
The Mach number downstream of a normal shock (for γ = 1.4) is shown in
Fig. 4.9. We note that the initial drop in downstream Mach number is rapid and it
gradually plateaus with higher Mach numbers. In fact, there is a limiting downstream
4.1 Introduction 153
31.00
26.00
p2/p1
21.00
16.00 2
/ 1
T2/T1
11.00
6.00
1.00
1 1.5 2 2.5 3 3.5 4 4.5 5
Upstream Mach Number, M1
1.50 2
/ 1
1.40
T2/T1
1.30
1.20
1.10
1.00
1 1.05 1.1 1.15 1.2 1.25 1.3
Upstream Mach Number, M1
Mach number that can be deduced from expression (4.18) in the limit of upstream
Mach number tending to infinity, namely
γ−1
M2 → as M1 → ∞ (4.19)
2γ
0.800
0.700 M2
0.600
0.500
0.400
1 2 3 4 5
Upstream Mach Number, M1 (~)
0.950
range (as compared to a
Downstream
straight line)
0.900
0.850
0.800
0.750
1 1.05 1.1 1.15 1.2 1.25 1.3
Upstream Mach Number, M1 (~)
in Fig. 4.11. The labels for the zones of action and silence signify the space where
disturbance is felt and the space corresponding to “forbidden” signal propagation, or
zone of silence. The Mach cone downstream of the moving source thus represents
the zone of action and the vast zone upstream of the Mach cone represents the zone
of silence.
The motion of finite disturbances at supersonic speeds, as in bodies with finite
thickness moving faster than the speed of sound, involves the creation of oblique
shocks. The analysis of plane oblique shocks is based on the application of conser-
vation principles in the normal and tangential directions to the shock. The schematic
drawing of an oblique shock in a supersonic flow, shown in Fig. 4.12, identifies the
wave angle β, the flow turning angle, θ, and the normal and tangential components
of flow with respect to the oblique shock wave.
4.1 Introduction 155
(a) (b)
Mach Wave µ
Mach Wave a. t
V=a
V>a V. t
Zone of
Silence
Zone of Action Zone of Silence Zone of Action
Fig. 4.11 Delineation of zones of action and silence for a moving source (of small disturbance) in
a compressible medium. a Source moving at the speed of sound. b Source moving at supersonic
speed, V > a
(a) (b)
O.S. O.S.
Mn1 > 1
Mt1 Mn1 w1 u1
M2 V2
M1 V1
Mt2 w2
1 Mn2 1 u2
2 Mn2 < 1 2 w2 = w1
Fig. 4.12 An oblique shock with its normal and tangential flow components. a The components
of flow Mach number across an oblique shock. b The structure of flow velocities across an oblique
shock
w2 = w1 (4.20)
However, as the static temperature and thus the speed of sound are both raised across
the shock, the tangential Mach numbers up- and downstream of the oblique shock
are unequal. Examining part (a) and part (b) of Fig. 4.12 shows the constancy of w,
but a drop in Mt across the oblique shock, while maintaining the same shock wave
angle, β and the deflection angle, θ in the two graphs. Therefore oblique shocks
enjoy an additional constant of motion in comparison to normal shocks and that is
the tangential velocity. In summary the primary constants of motion for an oblique
shock are as follows:
156 4 Shock-Expansion Theory
w = const.
ρu = const.
p + ρu 2 = const.
h t = const.
at = const.
a ∗ = const.
Vmax = const.
Since the normal component of conservation equations are the same for the
oblique and normal shock, all the jump conditions across an oblique shock are thus
solely determined from the normal component of upstream Mach number, namely,
M1 sin β. We will thus replace the upstream Mach number, M1 , in normal shock equa-
tions (4.13)–(4.17) by M1 sin β to establish the jump conditions across an oblique
shock of angle β. These are summarized as follows:
ρ2 u1 (γ + 1) M12 sin2 β
= = (4.21)
ρ1 u2 2 + (γ − 1) M12 sin2 β
p2 2γ 2 2
=1+ M1 sin β − 1 (4.22)
p1 γ+1
T2 2γ 2 2 2 + (γ − 1) M 2 sin2 β
= 1+ M1 sin β − 1 1
(4.23)
T1 γ+1 (γ + 1) M12 sin2 β
s2 − s 1 γ 2γ 2 2 2 + (γ − 1) M 2 sin2 β
= ln 1 + M1 sin β − 1 1
R γ−1 γ+1 (γ + 1) M12 sin2 β
(4.24)
2γ
− ln 1 + M12 sin2 β − 1
γ+1
pt2 − Δs 2γ 2 2
=e R = 1+ M1 sin β − 1
pt1 γ+1
⎡ ⎤ γ
γ−1
2+(γ−1)M12 sin2 β
2 + (γ − 1) 2γ M 2 sin2 β−(γ−1)
⎢ ⎥
⎢ 1
⎥ (4.25)
⎣ 2 + (γ − 1) M1 sin β
2 2 ⎦
From the velocity triangles up- and downstream of the oblique shock, we get the
following two equations:
4.1 Introduction 157
u1
tan β = (4.26a)
w1
u2
tan (β − θ) = (4.26b)
w2
The ratio of these two equations eliminates the constant tangential velocity and
replaces the normal velocity ratio by inverse density ratio, to produce:
The range of the wave angle β lies between the Mach angle, μ and 90◦ that
correspond to a normal shock. The flow turning angle, θ, thus starts from zero when
the oblique shock is an infinitesimal-strength Mach wave to a maximum turning
angle, θmax , which is shown in Fig. 4.13 (from [2]).
Example 4.1 The flow upstream of a normal shock is at Mach M1 = 2.0, in air with
γ = 1.4. The stagnation speed of sound is at = 300 m/s, calculate:
(a) Downstream Mach number, M2
(b) The ratio of speeds of sound, a2 /a1 across the normal shock
(c) The change in gas speed, Δu, across the shock in m/s
Solution:
1+[(γ−1)/2]M12
We may use M22 = γ M12 −(γ−1)/2
to calculate the downstream Mach number, M2 .
We get: M2 = 0.5774 The ratio of speeds √ of sound
√ is equal to square root of the
static temperature ratio, i.e., a2 /a1 = T2 /T1 = 1.6875 ≈ 1.30
From stagnation speed of sound, we get the speed of sound upstream and down-
stream of the shock using a 2 = at2 / 1 + (γ − 1) M 2 /2 , from which we get
a1 ≈ 224 m/s and a2 ≈ 290 m/s, therefore Δu = u 1 −u 2 = M1 a1 −M2 a2 ≈ 280 m/s
There are two reminders in using the oblique shock charts shown in Fig. 4.13.
Namely, for a given turning angle, θ, in a supersonic flow, there are two wave angles
that satisfy the conservation equations. Therefore, β is a double-valued function of
θ (instead of a single-valued function). The smaller wave angle, known as the weak
solution, is the angle to be used in oblique shock calculations. The larger wave angle,
known as the strong solution, does not naturally occur for attached shocks, but may be
created/induced in a supersonic wind tunnel or in ducts, by raising the back pressure.
The reason for the dashed lines in the strong solution portion of the oblique shock
graphs (e.g., Fig. 4.13) is to alert the user to select the weak wave solution (shown as
solid curves). The second reminder concerns θmax . The flow turning angle beyond
θmax may not be supported by an attached oblique shock; rather a detached (bow)
shock is needed to support the flow. These are graphically summarized in Fig. 4.14.
In part (a) of Fig. 4.14, we show the possibility of two oblique wave angles, one
corresponding to a strong solution and one corresponding to a weak solution. The
158 4 Shock-Expansion Theory
90
M1=
8
2010
8
80 6 5
4.5
4.0
3.8
3.6
3.4
3.2
3.0
70 2.8
2.8
60
3.0
β (deg)
3.2
3.4
Weak Shock 3.6
50 3.8
4.0
Shock−wave angle,
4.5
5
6
8
10
20
40
8
30
20
shock wave
M2
M
10 1
β
streamline θ
0
0 4 8 12 16 20 24 28 32
Deflection angle, θ (deg)
Fig. 4.13 Oblique shock charts for air are from Ref. [2]
4.1 Introduction 159
90
80
70
2.2
60
β (deg)
Weak Shock
50
Shock−wave angle,
40
30
shock wave
M2
M
1
β
streamline θ
20
10
0
26 30 34 38 42 46 50 54 58
Deflection angle, θ (deg)
strong solution is not used in the calculations; therefore it is shown as dashed line. In
part (b) of Fig. 4.14 we are showing an attached oblique shock (the weak solution) to
a 2-D wedge or ramp where the nose angle is less than the maximum turning angle
at M1 . We also show an abrupt turning of the supersonic flow across the plane 2-D
oblique shock in Fig. 4.14b. In part (c) of Fig. 4.14 we note the relative position of
the oblique shock and the Mach wave at M1 . It shows that the wave angle for the
oblique shock, β, is greater than the Mach angle, μ, as expected. Finally, in part (d) of
Fig. 4.14, we show a detached shock formation upstream of a blunt body with a nose
angle greater than the maximum turning angle, θmax , allowed for an attached oblique
shock at freestream Mach number, M1 . We may think of bow shock formation as
the wave that is needed to create higher entropy rise across the shock in response to
higher aerodynamic drag of a blunt body.
The conical shocks, which are created on cones in supersonic flow at zero angle
of attack, create a new class of flows known as conical flowfields. Since the flow-
field downstream of an attached conical shock lacks length scale, the flow variables
become a pure function of the (cone) angle θ in spherical coordinate system, as shown
in Fig. 4.15. The flow variables, e.g., p, T , ρ, V , M, are thus constant on constant
θ-cones in between the conical shock and the cone. These surfaces are indeed com-
pression Mach cones that provide isentropic compression and continuous turning of
the flow downstream of a conical shock to accommodate the changing flow area in
the downstream direction. The changing flow area downstream of a conical shock
(a) (b)
Oblique Shock
Oblique Shock
(strong solution) Oblique Shock
(weak solution;
(weak)
acceptable)
strong
weak M1 > 1
M1 > 1
Mach
M1 > 1 wave
M1 > 1
< max
90° nose
> max
Fig. 4.14 Oblique and detached shock formation on pointed bodies of different nose angle in
supersonic flow. a Two theoretically possible oblique shock angles. b Attached oblique shock on
a 2-D compression ramp. c The relative positions of oblique shock and Mach wave in supersonic
flow. d The formation of a bow shock ahead of a body in supersonic flow with θnose > θmax
4.1 Introduction 161
shock
M1 > 1 = =
cone
Cone at
zero angle-
Compression Mach
of-attack
Cones
Fig. 4.15 Formation of the conical shock on a cone at zero-angle-of-attack in supersonic flow
Conical
Mach Shock
Cone >
M1 > 1
Zone of Silence
Zone of Action
Fig. 4.16 Relative position of the conical shock and the Mach cone
is a new feature compared to the constant flow area downstream of a plane oblique
shock (see Fig. 4.14b). It is exactly this feature that leads to 3-D relieving effect, as
we shall see. A graphical depiction of the continuous bending of a streamline is also
shown in Fig. 4.15.
Here again, we emphasize that the angle of the conical shock is greater than the
Mach angle corresponding to M1 , as expected. To emphasize this point and also
show the zones of action and silence on a cone, we graph Fig. 4.16.
162 4 Shock-Expansion Theory
dVr
∇×V =0 → Vθ = (4.28)
dθ
The continuity equation for a steady, axisymmetric flow in spherical coordinates
reduces to:
∂Vθ ∂ρ
2ρVr + ρVθ cot θ + ρ + Vθ =0 (4.29)
∂θ ∂θ
The density is further eliminated from the above equation by the application of Euler
momentum equation and the local speed of sound in the flow downstream of the
conical shock. The result is a second-order non-linear ordinary differential equation
in Vr that is known as the Taylor-Maccoll equation:
γ−1 dVr 2 dVr d2 Vr
Vmax − Vr −
2 2
2Vr + cot θ +
2 dθ dθ dθ2
dVr dVr dVr d2 Vr
− Vr + =0 (4.30)
dθ dθ dθ dθ2
Note that the only independent variable is θ and the only dependent variable is Vr . The
Vmax term is the reference speed that we introduced earlier in this chapter, as a constant
of motion, and is used in conical flowfields, following Taylor-Maccoll’s theoretical
formulations. The Eq. (4.30) is still dimensional and may be non-dimensionalized
by dividing both sides of the equation by Vmax 3 . The resulting non-dimensional equa-
tion is:
⎡ 2 ⎤
γ−1⎣ 2 dVr ⎦ dVr d2 Vr
1 − Vr − 2Vr + cot θ +
2 dθ dθ dθ2
dVr dVr dVr d2 Vr
− Vr + =0 (4.31a)
dθ dθ dθ dθ2
where
Vr
Vr ≡ (4.31b)
Vmax
4.1 Introduction 163
From energy equation, V is related to the local Mach number according to:
−1/2
2
V = + 1 (4.31c)
(γ − 1) M 2
There are many good references on the subject of conical flowfields and conical
shocks and the reader may refer to them to fill in the blanks in the above derivation.
An excellent contemporary reference is by Anderson [1] that is recommended for
further reading. The Taylor-Maccoll equation (4.31a) is solved numerically using the
standard numerical approach of Runge-Kutta fourth order technique starting from the
shock at σ with the known Mach number and flow angle immediately downstream of
the shock. These are used to get Vr and Vθ that serve as the initial conditions for the
Runge-Kutta solution of the Taylor-Maccoll equation. Starting from the immediate
downstream of the conical shock, we march in the negative theta direction by taking
small incremental steps of Δθ. Per step in the numerical procedure, we calculate
the new Vr and Vθ until we capture the cone at Vθ = 0. In this approach, we are
finding the cone that supports the conical shock at σ. The numerical solution of
Taylor-Maccoll equation has been used to develop conical shock charts shown in
Fig. 4.17. In part Fig: 3.17a, only the weak solution is plotted and we clearly note
the maximum cone angles that support attached conical shocks. In part 4.17b, the
ratio of cone surface static pressure to freestream stagnation pressure is plotted.
Part 4.17c is a graph of the surface Mach number on the cone for different cones
in supersonic flow. In part 4.17d, we have the ratio of stagnation pressures across
a conical shock for different cone angles and freestream Mach number. Part 4.17e
highlights the three-dimensional relieving effect by comparing a cone to a wedge and
their corresponding wave angles. The wave angle for the wedge is greater than the
wave angle for the cone and thus the surface of the wedge experiences higher (static)
pressure than the (static) pressure on the corresponding cone. This is the essence
of three-dimensional relieving effect. The cone wave drag coefficient, referenced to
base area, is also shown in part 4.17f.
Example 4.2 Consider a 20◦ (2-D) ramp in a Mach 2 flow of air with γ = 1.4. Also,
there is a cone of semi-vertex angle of 20◦ in the same supersonic flow condition.
Calculate and compare:
(a) the wave angles on the 2-D ramp and the cone
(b) the surface pressure on the ramp and the cone.
Solution:
From the plane oblique shock charts (Fig. 4.13) we read the weak shock wave angle
corresponding to Mach 2 freestream and a turning angle of twenty degrees as β ≈
53.4◦ . From the conical shock charts (Fig. 4.17a) we read the corresponding conical
shock angle of θs ≈ 38◦ . Comparison of these two angles reveals the weaker 3-D
shock as compared to the 2-D shock.
164 4 Shock-Expansion Theory
M1
80 pS
0.4
m
M1 min or
in
p1
or
70 50
60 0.3
, deg.
40°
50 50°
40 30° 40°
20°
0.2 30°
30 20°
10° 10°
20 M1 0°
2 0.1 0°
10
0 0
1.0 2.0 3.0 4.0 1.0 2.0 3.0 4.0
M1 M1
32 ge 30°
2
Co
M1
ed 0.7
W
max
24 0.6
0.5
or
Fig. 4.17 Conical shock charts for γ = 1.4 from Shapiro [6]. a Shock angle versus approach Mach
number, with cone angle as parameter. b Ratio of surface pressure to free stream stagnation pressure
versus free stream Mach number. c Surface Mach number versus free stream Mach number, with
cone angle as parameter. d Ratio of surface stagnation pressure to free stream stagnation pressure,
with cone angle as parameter. e Regions of shock attachment and detachment for cone and wedge.
f Pressure drag coefficient based on projected frontal area
4.1 Introduction 165
The surface pressure on the ramp is calculated based on the normal component of
the flow to the oblique shock, i.e., M1 sin β = 2 sin(53.4◦ ) ≈ 1.6, which produces a
shock (static) pressure ratio of ∼2.82 downstream of the 2-D shock. Using Fig. 4.17b,
we read the ratio of the cone surface static pressure to the freestream stagnation
pressure of ∼0.25. The stagnation and static pressures are related by (the isentropic
relation):
γ
pt γ − 1 2 γ−1
= 1+ M = 7.824 for M = 2
p 2
Therefore the ratio of cone surface pressure to freestream static pressure is: pc / p1 ∼
0.25(7.824) = 1.956 The counterpart of this ratio for the 2-D ramp was calculated to
be 2.82. Here again we reveal the 3-D relieving effect as evidenced by this example,
i.e., lower cone surface pressure.
Example 4.3 Compare the maximum turning angles of Mach-2 flow of air in 2-D
and 3-D through plane oblique and conical shocks, respectively.
Solution:
Referring to the plane oblique shock charts, we note that θmax ≈ 23◦ in 2-D. The
implication of this angle is that in two-dimensional flow a plane oblique shock in
Mach 2 can at most cause ∼23◦ turning.
For the case of 3-D flows, we refer to conical shock charts. From Fig. 4.17e, we
note that maximum turning at Mach 2 is nearly 40◦ . The implication is that conical
shocks can stay attached to cones of up to 40◦ semi-vertex angle in Mach 2 flow,
whereas in 2-D the ramp angle above ∼23◦ causes the shock to be detached. This
too is due to 3-D relieving effect.
We now direct our attention to flow expansion. The counterpart of compression
waves is the expansion waves, which cause flow acceleration and turning in super-
sonic flow. The direction of flow turning (with respect to local upstream flow) deter-
mines whether an expansion or a compression wave is needed to cause the necessary
flow deflection. One description that says “turning the flow into itself” is caused by
compression waves and “turning the flow out of itself” is accomplished by expansion
waves is useful when we consider right- and left-running waves. We identify right-
and left-running waves in supersonic flow through the eyes of an observer who stands
on the wave and is facing downstream. The right-running wave (RRW) appears on
the right side of the observer and the left-running wave (LRW) appears on the left
side of the observer. The expansion waves that are centered at a corner are referred to
as expansion fans and are known as Prandtl-Meyer expansion waves. Let us exam-
ine some compression and expansion corners in supersonic flow and the waves that
cause the flow deflection in Fig. 4.18. The oblique shocks are marked as O.S. and
the Prandtl-Meyer expansion fans are abbreviated to P-M E.F. in Fig. 4.18. We also
have defined the flow angle, θ, with respect to horizontal, or x-axis and have assigned
positive values to θ when the flow direction is above horizontal and have assigned
a negative value to θ when the flow points in the downward (i.e., −y) direction. We
can also discern the “flow turning into itself” and “flow turning out of itself” from
compression and expansion corners that are schematically shown in Fig. 4.18.
166 4 Shock-Expansion Theory
(a) (b)
LRW: Left-Running Waves RRW: Right-Running Waves
LRW
y O.S.
2
> 1 1
X
2 y
M1 > 1 M2
M2 X M1 > 1
1 2
RRW
O.S.
2
< 1
1
M2 1 1 2
2 y
2 RRW
X
M1 > 1 P-M E.F.
2
> 1
Fig. 4.18 Definition sketch of compression and expansion corners and the corresponding waves that
are generated. a Compression corner, oblique shock, LRW. b Compression corner, oblique shock,
RRW. c Expansion corner, Prandtl-Meyer expansion fan, LRW. d Expansion corner, Prandtl-Meyer
expansion fan, RRW
Mach Wave
w µ)
V V
w µ)
d µ-d )
V + dV
Fig. 4.19 An exaggerated view of infinitesimal turning across an expansion Mach wave
4.1 Introduction 167
Expansion waves, as noted earlier are Mach waves, which cause an infinitesimal
turning of the supersonic flow. The analysis of an expansion Mach wave can be
carried out using the definition sketch in Fig. 4.19. Note that we had to exaggerate
the infinitesimal turning in order to “see” the geometrical relations involving flow
angles and velocities across the Mach wave in Fig. 4.19.
By applying the law of sines to the triangle with sides V and V + dV , we may
relate the infinitesimal turning angle, dθ, and incremental speed change, dV , in the
following equation:
V + dV V
= (4.32)
sin π2 + μ sin π2 − μ − dθ
The Eq. (4.32) may be simplified by expanding the sine terms to get:
dV cos μ ∼ cos μ 1
1+ = = = (4.33)
V cos (μ + dθ) cos μ − (dθ) sin μ 1 − (dθ) tan μ
≈ 1 + (dθ) tan μ
By taking isentropic flow conditions across the Mach wave; we may relate incremental
turning angle, dθ, to the incremental Mach number change, dM, according to:
√
M 2 − 1 dM
dθ = (4.34)
1 + γ−1 M 2 M
2
Equation (4.34) may be integrated from zero turning, corresponding to Mach one
(sonic flow), to a general turning angle ν (M) corresponding to supersonic Mach
number, M. The angle ν (M) is called the Prandtl-Meyer angle and has a closed
form solution, i.e.,
γ+1 γ−1 2
ν (M) = tan−1 M − 1 − tan−1 M 2 − 1 (4.35)
γ−1 γ+1
In addition to a table for the Prandtl-Meyer function and Mach angle, we graph these
two functions in terms of Mach number in Fig. 4.20. In Appendix C the relation
between M, μ and ν are tabulated up to a Mach number of 18.
60
40
20
0
1 10 100
Mach Number, M
168 4 Shock-Expansion Theory
Finally, since the expansion Mach waves never coalesce to create a rarefaction
shock, the flow across such waves is isentropic. Consequently, the stagnation pres-
sure remains constant across Prandtl-Meyer expansion waves. The stepwise calcula-
tion involving expanding supersonic flows is dictated by the wall geometry/turning,
which establishes the Mach number and the pressure. Now we solve a supersonic
flow expansion problem using Prandtl-Meyer function, ν (M).
Example 4.4 A supersonic flow approaches an expansion corner, as shown. Assum-
ing the medium is air with γ = 1.4, calculate:
(a) The Mach number downstream of the sharp corner
(b) The wave envelope, i.e., the angle of the head and tail waves
(c) The static pressure on the expansion ramp if p1 = 1 atm.
M1 = 2.0
2
w
= 22°
Expansion
ramp
Head Mach
wave
1
M1 = 2.0
Tail Mach
30° wave
2 20.6°
M2 = 2.836
4.1 Introduction 169
Solution:
The P-M function in region 1 is ν1 ≈ 26.5◦ . The P-M function on the expan-
sion ramp is thus: ν2 = ν1 + Δθ ≈ 26.5◦ + 22◦ = 46.5◦ Therefore the cor-
responding Mach number is M2 = 2.836. The leading Mach wave makes an
angle of μ1 = sin−1 (1/2) = 30◦ and the tail Mach wave makes an angle of
μ2 = sin−1 (1/2.836) ≈ 20.6◦ Finally, the ratio of static pressures on the two ramps
is related to their respective Mach numbers following the isentropic flow condition
up- and downstream of the corner, namely
γ
γ−1
p2 1 + (γ − 1) M12 /2
=
p1 1 + (γ − 1) M22 /2
Upon substitution for the up- and downstream Mach numbers (along with γ = 1.4),
we get: p2 / p1 ∼ 0.2729. Therefore, flow expansion around a 22◦ corner in Mach
2 upstream flow caused a flow acceleration to Mach 2.836 on the downstream ramp
and the static pressure was reduced to ∼27.3 % of upstream value, i.e., p2 = 0.2729
atm., stagnation pressure, pt , remained constant.
A’
D’
N’
R’
L’
M >1 x=c
Fig. 4.21 A pointed 2-D body in supersonic flight with the resultant force (per unit span), R
resolved in normal-axial (N and A ) directions and lift-drag (L and D ) directions
In Fig. 4.21, R is the resultant aerodynamic force, which is then resolved in
different orthogonal directions, as shown. The normal force is simply the integral of
pressure difference across the airfoil along the chord, namely:
c
N = ( pl − pu ) dx (4.38)
0
And the axial force (per unit span) is related to the integral of pressure times the
body slopes (along the chord) according to:
c
dyu
dyl
A = pu − pl dx (4.39)
0 dx dx
In terms of normal and axial force coefficients, cn and ca , we may express Eqs. (4.38)
and (4.39) in terms of the non-dimensional pressure coefficients on the upper and
lower surfaces, namely
1c
cn = C p − C p dx (4.40)
c 0
l u
1c dyu dyl
ca = C pu − C pl dx (4.41)
c 0 dx dx
Now, we are ready to apply these principles to a typical supersonic profile, namely,
a diamond airfoil.
y
1 3
15° 15° x
5° 15° 15°
M = 2.4
2 4
Solution:
Freestream has to turn an additional 10◦ up to reach panel #1 (i.e., from 5◦ to 15◦ ).
This is the case of flow turning into itself. Therefore, the turning has to occur via an
oblique shock wave. From θ − β − M chart for a plane oblique shock, we get:
OS
M∞ = 2.4 and θ = 10◦ −−→ β ≈ 33◦
NS
M∞ sin β ≈ 1.306 −−→ Mn 1 ≈ 0.786 and p1 / p∞ ≈ 1.805
to get:
C p1 ≈ 0.1996
Mn 1 0.786
M1 = ≈ ≈ 2.01
sin (β − θ) sin (33◦ − 10◦ )
The freestream has to turn a total of 20◦ (from +5◦ to −15◦ ), in order to reach ramp
number 2. Reading the plane oblique shock chart, we estimate a wave angle (i.e., the
weak solution):
OS
M∞ = 2.4 and θ = 20◦ −−→ β ≈ 44.2◦
Again the normal component of flow establishes the shock jumps, namely
NS
M∞ sin β ≈ 1.67 −−→ Mn 2 ≈ 0.6458 and p2 / p∞ ≈ 3.126
Mn 2 0.6458
M2 = ≈ ≈ 1.71
sin (β − θ) sin (44.2◦ − 20◦ )
172 4 Shock-Expansion Theory
to get:
C p2 ≈ 0.5273
To march to ramps 3 and 4, we recognize that the net turning angle is twice the
15◦ (i.e., 30◦ ) and since the flow is turning out of itself, we expect an expansion fan
to cause the turn. We start with ν1 which corresponds to M1 and from Prandtl-Meyer
table, we get: ν1 ≈ 26.38◦ and thus
P−M
ν3 = ν1 + 30◦ = 56.38◦ −−−→ M3 ≈ 3.35
The isentropic conditions between panels 1 and 3 can be used to establish static
pressure, p3 ,
γ
p3 1 + (γ − 1) M12 /2 γ−1
= ≈ 0.1272
p1 1 + (γ − 1) M32 /2
or
C p3 ≈ −0.1911
Now, marching onto panel 4 from panel 2, we note that the net turning angle is
again 30◦ . The P-M angle for panel 2 is ν2 ≈ 18◦ , which gives the P-M angle for
panel 4 to be ν4 ≈ 48◦ . From P-M tables, we get M4 ≈ 2.9. The isentropic conditions
between panels 2 and 4 can be used to establish static pressure, p4
γ
γ−1
p4 1 + (γ − 1) M22 /2
= ≈ 0.15861
p2 1 + (γ − 1) M42 /2
or
C p4 ≈ −0.1251
Therefore, we get the normal force coefficient as cn ≈ 0.1968. To calculate the axial
force, we first establish the airfoil thickness-to-chord ratio as t/c = tan 15◦ . This
gives t/c ≈ 0.2678 (a rather thick profile for supersonic flow!). Therefore, the axial
force coefficient may be written as:
1 t
ca = C p1 + C p2 − C p3 − C p4
2 c
Upon substitution, we get ca ≈ 0.1397.
We use Eq. (4.37) to get lift and drag coefficients from the normal and axial force
coefficients and the angle of attack, α, namely cl ≈ 0.1839 and cd ≈ 0.1563. Finally,
lift-to-drag ratio is L /D = 0.1839/0.1563 ∼ 1.18.
Fig. 4.22 Definition sketch of a general (asymmetrical) bi-convex airfoil in supersonic flow at
angle of attack
174 4 Shock-Expansion Theory
yu (x) and yl (x) respectively. The nose angles on the upper and lower surfaces, θnu
and θnl , are also identified. The chordwise position of the maximum thickness, xtmax ,
and the maximum thickness itself, tmax , are shown. The upstream supersonic flow is
at angle α (angle of attack) with respect to chord.
First, we need to identify the waves that are needed to turn the flow on the upper
and lower surfaces at the leading edge (nose). Depending on the relative values of
the upper surface nose angle and the angle of attack, we may have an oblique shock
or a Prandtl-Meyer expansion fan at the leading edge on the upper surface. This is
easily established by the following rules at the leading edge (for the case of positive
angle of attack, i.e., α > 0):
Case 1: α < θnu The wave is an oblique shock, which turns the supersonic flow
by (θnu − α) angle (note that the flow turns into itself ).
Case 2: α = θnu A Mach wave appears at the leading edge on the upper surface
with zero net turning (since α = θnu ) imposed across the Mach wave.
Case 3: α > θnu A Prandtl-Meyer expansion fan will be formed at the leading
edge on the upper surface, with a net turning of (α − θnu ). Note that the flow turns
out of itself.
(a)
(b)
(c)
Fig. 4.23 The wave pattern on a bi-convex airfoil at positive angle of attack, α. a Case 1: with
α < θnu (an oblique shock will form at the leading edge on the upper surface). b Case 2: with
α = θnu (upstream flow is aligned with the upper surface at the leading edge). c Case 3: with
α > θnu (Prandtl-Meyer expansion fan is formed at the leading edge on the upper surface)
4.3 Bi-Convex Airfoil 175
The leading edge of the lower surface will encounter an oblique shock, which
turns the flow by (θnl + α) angle. The identification of the leading-edge waves will
allow us to march across the waves and onto the upper and lower surfaces of the
bi-convex airfoil. The waves that are formed on the airfoil beyond the leading edge
are all expansion Mach waves that cause the flow to accelerate and the static pressure
to drop. In the vicinity of the airfoil, we will have the wave patterns corresponding
to Case-1, Case-2 and Case-3, shown in Fig. 4.23 where solid lines show oblique
shocks and the dashed lines are expansion Mach waves.
The waves at the trailing edge (on the upper and lower surfaces) are formed to
equalize the static pressure and flow direction downstream of the airfoil. In supersonic
airfoil theory, we define the slipstream as a (free) vortex sheet that is formed at the
trailing edge of the airfoil that separates the upper and lower flows with flow directions
on two sides coinciding with the slipstream and the static pressure being continuous
across the slipstream. Although the flows on two sides of the slipstream are parallel,
they are not equal in magnitude, i.e., V4 = V5 . It is indeed the jump in tangential
velocity across the slipstream that is called the vortex sheet strength. Also, since the
airfoil is in the zone of silence of the waves at the trailing edge, the flow on the airfoil
is not affected by the waves at the trailing edge. However, these waves impact the
formation of slipstream (i.e., vortex sheet), which may impact aircraft components
downstream of the wing (e.g., engine inlet). We show the slipstream downstream
of a supersonic airfoil at high angle of attack and its two boundary conditions in a
definition sketch in Fig. 4.24. We have (arbitrarily) chosen to show an oblique shock
wave as well as a Prandtl-Meyer expansion fan at the trailing edge of the airfoil in
Fig. 4.24. However, it is possible to have two oblique shock waves at the trailing
edge, if the flow environment downstream of the airfoil (i.e., static pressure and
flow direction) demands it. Indeed, the nature of these waves at the trailing edge is
established by the two boundary conditions on the slipstream, namely the equality
of static pressure ( p4 = p5 ), which is known as the pressure equilibrium condition,
and the flow angles (θ4 = θ5 ).
Once, we established the nature of the waves at the leading edge, we may proceed
to calculate the pressure and Mach number immediately downstream of the leading
Boundary Conditions
p4 = p5
P-M E.F. = 5=
O.S. 4
M4
2
M1 > 1 Slipstream
(Vortex sheet)
4 4 p4
1
M4
M5
M5 5
p5
3
5
O.S. P-M E.F.
Fig. 4.24 The slipstream downstream of a supersonic airfoil at angle of attack with possible trailing
edge waves
176 4 Shock-Expansion Theory
edge on the upper and lower surfaces of the airfoil. The flow downstream of the
leading edge on a bi-convex airfoil undergoes a continuous acceleration on both
upper and lower surfaces (as shown in Figs. 4.23 and 4.24). Thus, the flow expands
isentropically on both sides of the airfoil downstream of the leading edge where
oblique shocks may appear. The difference between the upper and lower surfaces is
therefore in the initial conditions that are set at the nose by the type and the strength of
the waves that are formed at the leading edge. Otherwise, the solution methodology
is identical on the upper and lower surfaces of the bi-convex airfoil.
Solution Methodology
We start at the leading edge, calculating the nose angle on the upper surface, accord-
ing to:
θnu = tan−1 (dyu /dx) (4.42)
x=0
Depending on the angle of attack, α, we may have one of the three cases that we
described earlier as determining the wave at the leading edge on the upper surface.
We then march across the wave at the leading edge and establish:
1. Mnu , Mach number of the flow immediately downstream of the nose on the upper
surface
2. pnu , static pressure immediately downstream of the nose on the upper surface
We then propose to march downstream on the upper surface by creating a table of
values for the surface pressure (and Mach number). For this purpose, we may divide
the airfoil into (n + 1)-chordwise calculation stations, from leading edge to trailing
edge, similar to the graph shown in Fig. 4.25. Here, we used the chord length, c,
as the normalizing length scale in the problem and graphed y/c versus x/c, as the
non-dimensional coordinates. Counting the leading and trailing edges as calculation
stations, we have (arbitrarily) chosen 11 positions along the airfoil upper surface
in Fig. 4.25, which divides the airfoil into ten segments/panels along the chord, as
shown.
Next, we calculate the surface (inclination) angle at every calculation station,
namely at n = 2 to 11 [note that we have already calculated the leading edge angle,
i.e., n = 1 in Eq. (4.42)]. For the ith station, we use:
x/c
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Fig. 4.25 The upper surface of a general bi-convex airfoil divided into 11 equally-spaced calculation
stations and 10 panels
4.3 Bi-Convex Airfoil 177
θu(i) = tan−1 (dyu /dx) (4.43)
x=xi
The Mach number immediately downstream of the nose that we calculated earlier
is now used to determine the Prandtl-Meyer angle at the nose, i.e., νn or ν1 . Since
the flow is continuously expanding on the bi-convex airfoil, the Prandtl-Meyer angle
continually increases on the upper surface, from one station to the next, by the net
flow turning angle between those stations, namely
νu(i+1) = νu(i) + θu(i+1) − θu(i) (4.44)
The absolute value in Eq. (4.44) shows that the Prandtl-Meyer angle increases as we
march towards the trailing edge on the bi-convex airfoil. Now, we can determine the
local Mach numbers, Mu(i) based on the Prandtl-Meyer angles, νu(i) . The isentropic
flow condition on the upper (and lower) surface can be used to calculate the static
pressure according to isentropic rule:
⎡ ⎤ γ−1
γ
pu(i+1) 1 + γ−12 M 2
u(i)
=⎣ ⎦ (4.45)
pu(i) γ−1
1 + 2 Mu(i+1) 2
We have now created the pressure distribution on the upper surface of the airfoil at
discrete points along the airfoil, from the leading edge to the trailing edge. By defining
the effective panel areas for each calculated pressure as the half-way points between
the adjacent calculation stations (or panels), we can proceed with the calculation
of the incremental force, per effective panel area. A definition sketch is shown in
Fig. 4.26 as an aid to understand the effective panel area and the calculation of the
incremental force and pitching moment based on pu(i) . The increment of normal
components of force (per unit span) on the upper and lower surfaces due to the static
pressures pu(i) and pl(i) are:
ΔNu(i) = − pu(i) xi+1/2 − xi−1/2 (4.46a)
ΔNl(i) = + pl(i) xi+1/2 − xi−1/2 (4.46b)
x/c
i-1/2 i+1/2
178 4 Shock-Expansion Theory
The increments of axial component of force (per unit span) on the upper and lower
surface due to static pressures pu(i) and pl(i) are:
ΔAu(i) = + pu(i) yu(i+1/2) − yu(i−1/2) (4.47a)
ΔAl(i) = − pl(i) yl(i+1/2) − yl(i−1/2) (4.47b)
The increments of pitching moment about the leading edge (per unit span) on the
upper and lower surfaces are:
ΔMu(i) = pu(i) xi+1/2 − xi−1/2 xi + pu(i) yu(i+1/2) − yu(i−1/2) yu(i) (4.48a)
ΔMl(i) = − pl(i) xi+1/2 − xi−1/2 xi − pl(i) yl(i+1/2) − yl(i−1/2) yl(i) (4.48b)
Finally, we add all the increments to get the normal and axial force components
(per unit span) on the upper surface and proceed similarly on the lower surface.
Equation (4.37) will then be used to express lift and wave drag (per unit span) in terms
of the normal and axial force and the angle of attack, α. We will now demonstrate
the shock-expansion theory applied to an asymmetric, parabolic profile bi-convex
airfoil with negative camber in the following example.
Example 4.6 Consider a 5 % thick bi-convex airfoil of parabolic arc upper and lower
profiles with negative camber, as shown in the schematic drawing. The upper surface
contributes 2 % to thickness at 50 % chord and is thus defined by:
yu (x) = 0.08x (1 − x)
The lower surface contributes 3 % thickness to the airfoil at mid chord and is thus
defined by:
yl (x) = −0.12x (1 − x)
By virtue of negative camber, a stronger oblique shock appears at the leading edge
on the lower surface than the oblique shock on the upper surface. Therefore, we
expect this airfoil to produce lift (albeit small) even at zero angle of attack. Use
shock-expansion theory to first calculate the pressure coefficient on the upper and
lower surfaces, then calculate the lift, wave drag and pitching moment (about leading
edge) coefficients for this airfoil at the freestream Mach number of M∞ = 1.4 and
at zero angle of attack. Graph the airfoil and its upper and lower surface pressure
distributions along the chord.
y nu 0.02c
M = 1.4
x
= 1.4 0.03c
c
nl
Not-to-Scale
4.3 Bi-Convex Airfoil 179
Solution:
First, we calculate the leading edge angles for the upper and lower surfaces.
dyu
= (0.08 − 0.16x)|x=0 = 0.08
dx x=0
Therefore, the nose angle on the upper surface is θnu ≈ 4.57◦ . Similarly, we calculate
the leading edge angle for the lower surface,
dyl
= (−0.12 + 0.24x)|x=0 = −0.12 → θnl ≈ −6.84◦
dx x=0
In both cases, the supersonic flow turns into itself and thus requires an oblique shock
to form at the leading edge on both upper and lower surfaces. Using oblique shock
charts (Fig. 4.13), we get:
OS
M∞ = 1.4 and θnu ≈ 4.6◦ −→ βu ≈ 52.2◦
NS
M∞ sin β ≈ 1.11 −→ Mnunose ≈ 0.907 and pn,u / p∞ ≈ 1.261
Munose = Mnunose / sin (β − θ) ≈ 1.228
The Prandtl-Meyer angle immediately downstream of the leading edge on the upper
surface corresponds to Munose = 1.228. We can use Eq. (4.35) to get:
νunose ≈ 4.25◦
Now, from the equation of the slope of the upper surface, dy dx = 0.08 − 0.16x,
u
OS
M∞ = 1.4 and θnl ≈ −6.84◦ −−→ βl ≈ 56.2◦
NS
M∞ sin β ≈ 1.16 −−→ Mnlnose ≈ 0.866 and pl / p∞ ≈ 1.4124
Mlnose = Mnlnose / sin (β − θ) ≈ 1.141
The Prandtl-Meyer angle immediately downstream of the leading edge on the lower
surface corresponds to Mlnose = 1.141. We can use Eq. (4.35) to get:
θlnose ≈ 2.185◦
180
Table 4.1 Geometric and flow calculations on a bi-convex airfoil using Shock-expansion theory
M∞ τu τl x yu yl ycamber θ u (◦ ) θ l (◦ ) ν u (◦ ) ν l (◦ ) Mu Ml
1.4 0.02 0.03 0 0 0 0 4.57 −6.84 4.2525 2.1848 1.227 1.141
1.4 0.02 0.03 0.05 0.004 −0.006 −0.001 4.12 −6.16 4.7082 2.8635 1.245 1.171
1.4 0.02 0.03 0.10 0.007 −0.011 −0.002 3.66 −5.48 5.1645 3.5440 1.262 1.199
1.4 0.02 0.03 0.15 0.010 −0.015 −0.003 3.21 −4.80 5.6212 4.2260 1.279 1.226
1.4 0.02 0.03 0.20 0.013 −0.019 −0.003 2.75 −4.12 6.0783 4.9094 1.296 1.253
1.4 0.02 0.03 0.25 0.015 −0.023 −0.004 2.29 −3.43 6.5358 5.5939 1.313 1.278
1.4 0.02 0.03 0.30 0.017 −0.025 −0.004 1.83 −2.75 6.9936 6.2795 1.329 1.303
1.4 0.02 0.03 0.35 0.018 −0.027 −0.005 1.37 −2.06 7.4516 6.9658 1.346 1.328
1.4 0.02 0.03 0.40 0.019 −0.029 −0.005 0.92 −1.37 7.9098 7.6527 1.362 1.353
1.4 0.02 0.03 0.45 0.020 −0.030 −0.005 0.46 −0.69 8.3681 8.3401 1.378 1.377
1.4 0.02 0.03 0.50 0.020 −0.030 −0.005 0.00 0.00 8.8264 9.0276 1.394 1.401
1.4 0.02 0.03 0.55 0.020 −0.030 −0.005 −0.46 0.69 9.2848 9.7151 1.41 1.425
1.4 0.02 0.03 0.60 0.019 −0.029 −0.005 −0.92 1.37 9.7431 10.4024 1.426 1.448
1.4 0.02 0.03 0.65 0.018 −0.027 −0.005 −1.37 2.06 10.2013 11.0893 1.441 1.472
1.4 0.02 0.03 0.70 0.017 −0.025 −0.004 −1.83 2.75 10.6593 11.7757 1.457 1.495
1.4 0.02 0.03 0.75 0.015 −0.023 −0.004 −2.29 3.43 11.1170 12.4612 1.473 1.518
1.4 0.02 0.03 0.80 0.013 −0.019 −0.003 −2.75 4.12 11.5745 13.1458 1.488 1.542
1.4 0.02 0.03 0.85 0.010 −0.015 −0.003 −3.21 4.80 12.0316 13.8291 1.504 1.565
1.4 0.02 0.03 0.90 0.007 −0.011 −0.002 −3.66 5.48 12.4884 14.5112 1.519 1.588
1.4 0.02 0.03 0.95 0.004 −0.006 −0.001 −4.12 6.16 12.9446 15.1916 1.535 1.611
1.4 0.02 0.03 1 0 0 0 −4.57 6.84 13.4003 15.8703 1.55 1.634
4 Shock-Expansion Theory
Table 4.2 Geometric and flow calculations on a bi-convex airfoil using Shock-expansion theory
pu / pnose pl / pnose pu / p∞ pl / p∞ C p, u / p∞ C p, u / p∞ dcn dca,u dca,l dcm,u dcm,l
1 1 1.2610 1.4124 0.190 0.301 0.0055 0.0007 0.0017 0 0
0.9773 0.9626 1.2324 1.3596 0.169 0.262
0.9554 0.9283 1.2048 1.3111 0.149 0.227 0.0078 0.0010 0.0022 0.0015 −0.0023
4.3 Bi-Convex Airfoil
Now, from the equation of the slope of the lower surface, dy dx = −0.12 + 0.24x,
l
(a)
0.03
0.01
y/c
-0.01
-0.03
-0.05
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
x/c
(b) -0.3
-0.2
-0.1
Cp
0 Cp, upper
0.1
0.2
Cp, lower
0.3
0 0.2 0.4 0.6 0.8 1
x/c
Fig. 4.27 Application of shock-expansion theory to a 5 % thick bi-convex airfoil with negative
camber. a The 5 % thick parabolic bi-convex airfoil with negative camber (2 % upper thickness and
3 % lower thickness). b Pressure distribution on the upper and lower surfaces
4.3 Bi-Convex Airfoil 183
The normal force coefficient is the sum of dcn column (in Table 4.2): cn ≈
0.00233. The axial force coefficient is the sum of dca, u and dca, l columns (in
Table 4.2): ca ≈ 0.01453. The pitching moment coefficient about the leading edge
is the sum of dcm, u and dcm, l (in Table 4.2): cm, LE ≈ 0.0132.
Since the airfoil is at zero angle of attack, the normal and axial force directions
coincide with the lift and drag directions respectively. The shape of the airfoil, includ-
ing the mean camber line is shown in Fig. 4.27a and the pressure distribution is shown
in Fig. 4.27b, from the Excel files.
Flow around axisymmetric bodies in supersonic flow involves conical shocks and
expansion Mach cones. In the absence of angle of attack, a pointed axisymmetric
body that supports a conical shock may be analyzed using the conical shock charts
in Fig. 4.17. However when the body is in an angle of attack, the axis of the conical
shock will no longer coincide with the body axis. Figure 4.28 shows an example of a
cone at an angle of attack and zero yaw. The axis of the cone is along x-direction, as
shown. The cone cross section is therefore in the y-z plane. We define the meridian
angle, ϕ in the y-z plane with Φ = 0 at the top and 180◦ at the bottom of the cone,
as shown. Therefore the windward side of the cone in positive angle of attack is at
Φ = 180◦ and the leeward side of the cone is at Φ = 0◦ , as indicated. The meridian
angle becomes important in asymmetrical flows as it describes the rays along the
cone surface where flow attains unique properties.
We expect the windward side of the cone to experience higher static pressures and
the leeward side of the cone to experience lower static pressures. The same arguments
apply to a cone in sideslip or yaw. Therefore, in general we expect the static pressure
gradient that is set up (in the meridian plane) around a cone with angle of attack and
yaw to create a spiraling flow around the cone, as shown in Fig. 4.29.
Fig. 4.28 Cone at an angle of attack (and zero yaw) in a supersonic flow showing an asymmetrical
disposition of a conical shock attached to the cone at the vertex. a An axisymmetric cone at an angle
of attack. b Meridian angle, Φ at the base of the cone
184 4 Shock-Expansion Theory
z
Spiraling streamlines
on a cone with angle
Conical Shock
of attack and yaw
=0º
z Leeward side -
- -
-
- -
- x = =270º y
+ 90º
M >1 +
+
+ +
M >1 + =180º
Windward side
Cone
Fig. 4.29 Spiraling streamlines on a cone in supersonic flow at angle of attack and yaw
Fig. 4.30 A (pointed) slender body of revolution in supersonic flow at zero angle of attack. a Slender
body of revolution at zero angle of attack. b Local cone (tangent) to point A has semi-vertex angle, δ
Example 4.7 Consider a slender body of revolution at zero angle of attack in super-
sonic flow with M∞ = 2.0. The semi-vertex angle is δ0 = 30◦ . Calculate the pressure
coefficient at the vertex and estimate the pressure coefficient at point A, where the
body angle is δ = 10◦ using the method of local cones.
A
M = 2.0
= 10º
0
0
= 30º
x
Solution:
From conical shock charts (Fig. 4.17), for Mach 2.0 and a cone of 30◦ semi-vertex
angle, we get the conical shock angle at the vertex to be σ0 ≈ 48◦ . The normal
component of flow to the shock is therefore:
M∞ sin σ0 ≈ 2 sin 48◦ = 1.486
pN 2γ 2
=1+ M∞ sin2 σ0 − 1
p∞ γ+1
We note the static pressure at point A is significantly less than the pressure at the
nose, as expected. The static pressure at point A is ∼55 % of the static pressure at
the nose.
The same approximation applies to a slender axisymmetric body in angle of attack,
α. For a general axisymmetric body at an angle of attack, we define an equivalent
cone angle, δc that is tangent to point A and the freestream direction is the axis of the
equivalent cone, as shown in Fig. 4.31. The body angle at A is δ, the nose angle is δ0
and the equivalent cone angle is δc . The meridian angle is also shown in Fig. 4.31,
which affects the equivalent cone angle according to (4.49).
186 4 Shock-Expansion Theory
c 270º
x
0 90º
M >1 A
180º
Fig. 4.31 Definition sketch used in the method of local cones for a slender body in angle of attack
α2
δc = δ − α cos Φ + cot δ · sin2 Φ (4.49)
2
Example 4.8 Consider a slender axisymmetric body, similar to the body shown in
Fig. 4.31, with the semi-vertex angle, δ0 = 20◦ , in a supersonic flow with M∞ = 2
and at 10 degrees angle of attack, i.e., α = 10◦ . Apply the method of local cones to
point A on the body where the local body angle is 10◦ ; and A is on the windward
side of the body with Φ = 180◦ . Estimate the pressure coefficient at A using the
method of local cones.
Solution:
The nose angle with respect to the flow direction on the windward side is the sum of
20◦ for the nose and 10◦ for the angle of attack (this may also be deduced from
Eq. (4.49)). Therefore, the conical shock angle corresponding to M∞ = 2 and
equivalent cone angle of 30◦ is (the same as Example 4.7): σ0 ≈ 48◦ on the wind-
ward side. The rest of the calculations immediately downstream of the nose on the
windward side follow the Example 4.7, namely
M∞ sin σ0 ≈ 2 sin 48◦ = 1.486
and the static pressure ratio at the nose (on the windward side) is calculated to be:
p N 2γ 2
= 1 + M sin 2
σ − 1 ≈ 2.411
p∞ ϕ=180◦ ∞ 0
γ+1
To get the equivalent cone angle for point A, we substitute for the body angle at A,
namely δ = 10◦ and the meridian angle, Φ = 180◦ as well as the angle of attack,
α = 10◦ in Eq. (4.49). We get:
4.4 Axisymmetric and Slender Bodies 187
α2
δc = δ − α cos Φ + cot δ · sin2 Φ → δc = 20◦
2
Therefore the corresponding (equivalent) conical shock produces the ratio of the cone
surface pressure to freestream stagnation pressure of pc / pt∞ ≈ 0.25 at M∞ = 2
(from Fig. 4.17b). From isentropic table, we get pt∞ / p∞ = 7.824 for Mach 2, and
therefore we estimate p A / p∞ to be:
pA p A pt∞
= ≈ (0.25) (7.824) ≈ 1.956
p∞ pt∞ p∞
In Example 4.8, we assumed that the body surface pressure was the same as
the equivalent cone surface pressure. However, within the framework of method of
local cones; there are two hypotheses that are advanced on the flow conditions, i.e.,
pressure and velocity, at point A as compared to an equivalent cone that is tangent
to A, namely:
Hypothesis 1: pressure at A is identical as on the equivalent cone with δc , good
for high velocities
Hypothesis 2: velocity at A is identical as on the equivalent cone, which is better
for low velocities
2.0
= 6º
No. 1 1.0
1.6 No. 1 5
= 0º
R6.5
1.2
No. 2 1.0
= 6º
CD
1.5 3.5
0.8 No. 2
5
= 0º
= 6º R6.5
0.4
No. 3 1.0
No. 3
= 0º 2.5
0
1 2 3 4 5
M
Fig. 4.32 The effect of angle of attack and body shape on nose wave drag (after Ref. [3])
188 4 Shock-Expansion Theory
We will examine these hypotheses more closely in the next section (Examples and
Applications). In addition, the base pressure is assumed to be equal to the freestream
pressure, i.e., pb = p∞ , which is a reasonably accurate assumption for most appli-
cations of truncated bodies in supersonic flow. Once the pressure distribution on the
slender axisymmetric body is obtained [using (4.49) for an equivalent cone] and
further assuming that the base pressure is equal to the ambient pressure, we may
approximate the aerodynamic forces and moments on the body, namely, lift, wave
drag, side force and three moment coefficients for the pointed slender body in super-
sonic flow. The effect of angle of attack and body shape (i.e., nose bluntness) on
the nose wave drag coefficient is shown through experimental data in Fig. 4.32 (from
Krasnov 1970). Body number 1 is a flat-head cylinder, whereas the bodies designated
as number 2 and 3 represent a streamlined nose of 1.5 and 2.5 (nose) fineness ratios,
respectively. Note that the lower-end range of freestream Mach numbers dips into the
transonic zone. Here, the effect of angle of attack is seen as an upward shift in C D .
The nose fineness ratio may be related to the non-dimensional radius of curvature
via:
FN = R − 1/4 (4.53)
The local body slope is obtained by differentiating r (x) with respect to x, i.e.,
−1/2
dr 1 dr FN FN2
tan δ = = = (1 − x) 1 − 2 (1 − x) 2
(4.54)
dx 2FN dx R R
FN FN
sin δ0 = = (4.55)
R FN2 + 1/4
Based on the method of local cones described in Sect. 4.4, we propose to examine
the two hypotheses as they apply to a tangent ogive, namely:
1. the pressure on the body is equal to the pressure on the equivalent cone, of the
same local angle δ, and the stagnation pressure corresponding to δ0 cone (at the
nose)
2. the velocity on the body is equal to the velocity on an equivalent cone, of the same
local angle, δ and the stagnation pressure corresponding to a cone of δ, semi-vertex
angle (i.e., for higher total pressure recovery than the first hypothesis)
r (x)
dr
dx
r(x)
0
x
d
x
lN
The first hypothesis assumes the same pressure results in a pressure coefficient, C pp
and the second method assumes the same surface velocity produces C pv . These
pressure coefficients are related to each other via [3]:
2 (ν − 1)
C pv = C pp ν + (4.56)
γ M∞2
where,
ν0
ν≡ (4.57)
ν
pt |δ0
ν0 ≡ ; Stagnation pressure ratio on a cone with the nose angle δ0 (4.58)
pt∞
pt |δ
ν≡ ; Stagnation pressure ratio on a cone with the nose angle δ (4.59)
pt∞
The second hypothesis, i.e., C pv proves to be a more accurate estimate of the local
pressure coefficient than the C pp . The C pv leads to a negative pressure coefficient
on the aft sections of the ogive, i.e., near the shoulder and the shadow region that is
borne by both the Method of Characteristics as well as the experimental data. The
graphs of C pp and C pv on tangent ogives at Mach numbers 2 and 5 are shown in
Figs. 4.34 and 4.35 respectively (from Ref. [3]). Note that the pressure coefficient
C pv near the shoulder region of the tangent ogive where δ ≈ 0, dips into the negative
territory for tan δ0 > 0.35, or semi-apex angle of >19.3◦ .
0.50
0.5
0.45
0.4
0.40
Cp
Cpv
ta
0.35
0.3
0.30
0.2 0.25
0.20
0.1 0.15
0.10
0
0.05
0
-0.1
0 0.2 0.4 0.6 0.8
tan 0
4.5 Examples and Applications 191
0.5 0.55
0.50
n
ta
0.4
Cp
0.45
0.3 0.40
Cpv
0.35
0.2 0.30
0.25
0.1
0.15
0
0
-0.1
0 0.2 0.4 0.6 0.8
tan 0
Example 4.9 A tangent ogive has a nose fineness ratio of FN = 2.825. Calculate:
(a) the non-dimensional radius of curvature, R for the nose
(b) the semi-vertex angle, δ0 in degrees
(c) the pressure coefficients C pp and C pv at M∞ = 2 where the body angle is 11.2◦
(d) the wave drag coefficient of this tangent ogive (nose) at M∞ = 2
Solution:
From Eq. (4.53), we establish the non-dimensional radius of curvature of the nose
to be:
R = FN2 + 1/4 = 8.23
For the body angle of 11.2◦ and the semi-apex angle of 20◦ we calculate the tangents
to be: tan δ ≈ 0.20 and tan δ0 ≈ 0.365. We use these values in Fig. 4.34 to get the
two pressure coefficients: C pp ≈ 0.13 and C pv ≈ 0.12 Finally, the pressure drag
coefficient of the tangent ogive is estimated from the semi-empirical correlation
(4.60). But first, we need the pressure coefficient on the cone of the semi-apex angle
20◦ , in Mach-2 flow (from Fig. 4.17f, noting that the cone pressure coefficient and
the wave drag coefficients are the same since the base pressure is assumed to be equal
to p∞ ), C pc ≈ 0.325
196FN2 − 16
C D = C pc 1 − ≈ 0.0998 ≈ 0.1
14 (M∞ + 18) FN2
A slender body in a slightly supersonic flow, i.e., for M∞ = 1+ε creates a normal
shock that stands ahead of the body. The flow downstream of the normal shock is
subsonic with M ≈ 1 − ε. This is the result of the linear behavior of normal shocks
in slightly supersonic flows, as we demonstrated earlier in this chapter. Therefore
the local Mach number on a slender body, M, does not change appreciably when the
freestream flow condition is near sonic. This argument, first forwarded by Liepmann
and Bryson [4] can be expressed mathematically as:
∂ M
=0 (4.61)
∂ M∞ M∞ =1
This is the principle of stationary local Mach numbers near sonic freestream flow
conditions on slender bodies. This condition can be used as an extrapolation tool to
0
0.6
=1
0.20 M =1
(CDp )M
(CDp)M
0.4
c
0.10
M =1
0.2
0 0
0 30 60 90 0 30 60 90
c
(deg) 0
(deg)
Fig. 4.36 Wave drag coefficient, C D p , at sonic flow (after Ref. [3]). a Cone. b Ogive
4.5 Examples and Applications 193
extend the results of force and moment coefficients from subsonic and supersonic
sides through the transonic region. We will apply this principle to a slender cone and
an ogive in transonic flow. First, the experimental results for the wave drag coefficient
for cones and ogives are shown in Fig. 4.36a, b respectively at sonic freestream Mach
number [3]. Note that the cone semi-apex angle of 90◦ is the case of a flat-head
cylinder and the ogive with the semi-vertex angle of 90◦ is the case of hemisphere.
Applying the principle of stationary Mach numbers to slender axisymmetric bodies,
e.g., cone or ogive, where transonic similarity applies, produces:
2
2 +1
γ M∞ 2 M∞ −1
CD = (C D ) M∞ =1 + (4.62)
(γ + 1) M∞
2 (γ + 1) M∞
2
This expression is valid for axisymmetric slender bodies that include both cones
and ogives. Therefore the experimental data on C D -cone at M∞ = 1, as shown in
Fig. 4.36a substituted in Eq. (4.62) allows for the extension of wave drag in transonic
range, i.e., from subsonic to supersonic. In particular the slope of the drag coefficient
near sonic flow is determined. The proof of the match between the theory and the
experimental results on a cone in the transonic range is shown in Fig. 4.37a [3]. The
rising dashed lines through sonic Mach number are produced from the theoretical
model, i.e., Eq. (4.62). Similarly, the wave drag coefficient on an ogive in the vicinity
of sonic flow may be explored using Eq. (4.62) in conjunction with the ogive wave
drag coefficient for sonic flow, i.e., Fig. 4.36b. Comparison with experimental results
shown in Fig. 4.37b is in good agreement with the theoretical model.
dmid
(a) M c
(b) M
lN
N
= lN /dmid
1.2 = 45º
c 0.12 N
3
35º
0.8 0.08
p
4
p
CD
CD
25º
0.4 20º 5
0.04
6
10º
0 0
0.8 1.2 1.6 2.0 0.8 1.2 1.6 2.0
M M
Fig. 4.37 Experimental data on wave drag coefficient in transonic flow (after Ref. [3]). Note that
λ N is the nose fineness ratio, FN in our notation. a Cone. b Ogive
194 4 Shock-Expansion Theory
4.6 Summary
Problems
p1
M = 3.0 5°
10°
p p
p2
c
4.2 Calculate the stagnation pressure measured by a Pitot tube on an inclined ramp
in a supersonic flow, as shown.
4.6 Summary 195
10°
M = 2.0
Pitot
p = 100 kPa
5º p3
MT.S. = 2.0
5º
y p2
p1
x
4.4 A supersonic flow expands around a sharp corner, as shown. Calculate the fol-
lowing parameters:
(a) Downstream Mach number, M2
(b) Flow area ratio, A2 /A1
(c) The angle of Prandtl-Meyer fan envelope, δ.
1
Streamline
M1 = 2.0 2
A1
= 30º A
2
196 4 Shock-Expansion Theory
4.5 A symmetrical half-diamond airfoil has a nose angle θnose = 30◦ and is exposed
to a supersonic flow. A Pitot tube is installed on each of the three surfaces, as shown.
Calculate the Pitot tube readings on the airfoil surfaces at zero angle of attack.
Pitot tubes
2 3
M1 = 3.0
pt1 = 100 kPa 30º 30º
4 Pitot tube
4.6 Calculate the wave drag coefficient of a 2-D sharp-nosed projectile, as shown,
assuming the base static pressure is equal to ambient, i.e., pbase = p∞ . The 2-D
wave drag coefficient here is defined based on b, i.e., Cd ≡ D /q∞ · b. Assume the
angle of attack is zero.
20º
M = 2.0
b
4.7 A Prandtl-Meyer centered expansion wave is visualized with the wave angles
as shown. Calculate:
(a) the flow turning angle, i.e., wall angle, θw
(b) velocity ratio across the expansion waves V2 /V1 .
V1
30º
16.6º
w
V2
4.8 A half-diamond airfoil with a nose angle of 15◦ and a thickness-to-chord ratio
of 10 % is in a supersonic flow, as shown. Calculate:
(a) the chordwise location of the maximum thickness point (in % c)
(b) the trailing-edge angle
(c) the non-dimensional pressure on the three surfaces, p1 / p∞ , p2 / p∞ , p3 / p∞ .
4.6 Summary 197
1 tmax 2
M = 2.4
3
15º c
4.9 Calculate the 2-D lift and drag coefficients of a two-dimensional wave rider, as
shown. Assume the base pressure is equal to the ambient, p4 = p1 . Shock wave
angle is β1 = 45◦ . The upper surface is aligned with the flight direction (surface 2).
Surface 3 is the lower surface and surface 4 is the base.
4
2
M1 = 3.0 p4 = p1
1
45º 3
Oblique
shock surface
c
5º 1
t
2 4
10º 10º
10º 10º
M1 = 5.0
= 1.4 3 5
4.12 Consider a 10 % thick bi-convex airfoil of parabolic arc upper and lower profiles
with negative camber. The upper surface contributes 3 % to thickness at 50 % chord
and the lower surface contributes 7 % thickness to the airfoil at mid chord. First,
write the equations yu (x) and yl (x) for the upper and lower profiles. Next, use
shock-expansion theory to calculate:
(a) the pressure coefficient at the leading edge, and increments of 10 % chord until
trailing edge
(b) the lift coefficient
(c) the (wave) drag coefficient
(d) the pitching moment coefficient about the leading edge.
y yu(x) 0.03c
M =2
x
= 1.4 0.07c
c
yl(x)
4.13 A hexagonal airfoil is a suitable shape for supersonic flow. Assuming the airfoil
shown is 10 % thick and it is at zero angle of attack in Mach 2.4 flow, use shock-
expansion theory to calculate:
(a) pressure coefficients on all six panels
(b) the airfoil wave drag coefficient.
y/c
0.05 x/c
1.0
0.25 0.75
4.14 Identify the waves and draw the wave pattern around the two-dimensional
symmetrical body at points A, B and C, as shown. For expansion waves, calculate
the wave angle of the head and tail waves and with the oblique shocks calculate the
wave angle. Note that the body is symmetrical and is set at zero angle-of-attack.
4.6 Summary 199
M1 = 3.0
B C
30º
A
10º
4.15 A supersonic flow approaches an expansion corner, as shown. The static pres-
sure ratio p2 / p1 is 0.25. Calculate:
(a) Mach number on the expansion ramp, M2
(b) Wall angle, θw (in degrees).
p1 1
M1 = 2.5 2
p2
w
4.16 Two compression ramps in supersonic flow create two plane oblique shocks,
as shown. Calculate:
(a) shock wave angle, β1 (in degrees)
(b) ramp angle, θ1 (in degrees)
(c) shock wave angle, β2 (in degrees)
(d) estimate the second ramp angle, θ2 (in degrees).
M1n = 1.2 2 2
1
M2n = 1.2
M1 = 2.4
2
1 3
OS1 M2
OS2
200 4 Shock-Expansion Theory
4.17 A sharp hexagonal airfoil is placed in a supersonic flow. The Mach numbers
on its upper three surfaces are as indicated. Calculate:
(a) The semi-nose angle, θLE
(b) Static pressure ratio, p2 / p1
(c) Static pressure ratio, p3 / p1
(d) The semi trailing-edge angle, θTE .
1 2
M M2 = 3 3
M1 = 2 M3 = 4
L.E. T.E.
y
t/c = 0.10
M = 1.8
= 1.4 x
x=0 x=1
4.19 Apply shock-expansion theory to the diamond airfoil in supersonic flow (as
shown), to calculate:
(a) upper and lower thickness-to-chord ratios, t1 /c and t2 /c
(b) the pressure coefficients, C p1 , C p2 , C p3 , C p4
(c) the normal force coefficient, cn
(d) the axial force coefficient, ca
(e) lift-to-drag ratio, L /D
(f) the pitching moment coefficient about the leading edge, cm,L E .
3
1
t1
12º 12º
2º 14º 14º
M = 2.0 t2
2 4
c
4.6 Summary 201
3
2
4.21 An oblique shock is described by its upstream flow components (u and w are
normal and parallel to the shock, respectively) as shown. Calculate:
(a) Upstream speed of sound, a1
(b) Characteristic Mach number, M2∗ .
O.S.
1 u2
w2
u1 = 360 m/s
V1 V2
w1 = 410 m/s
*
= 280 m/s w1
= 1.4 u1 2
R = 287 J/kg.K
4.22 A flat plate is in Mach 1.2 flow at 2◦ angle of attack. Use shock-expansion
theory to predict its lift-to-wave drag ratio. The ratio of specific heats is γ = 1.4.
4.23 A bi-convex airfoil has a nose angle, θnose = 10◦ . For an angle of attack of 5◦
and freestream Mach number, M∞ = 2.4, calculate:
202 4 Shock-Expansion Theory
10º A
5º
y
= 2º
x
M = 2.8 p
= 1.4
= 1.4
M = 3.0 A
0 = 20º
= 40º
0 x
4.6 Summary 203
4.26 Consider a slender axisymmetric body, as shown, with the semi-vertex angle,
δ0 = 30◦ , in a supersonic flow with M∞ = 2.4 and at 5 degrees angle of attack, i.e.,
α = 5◦ . Apply the method of local cones to point A on the body where the local body
angle is 10◦ ; and A is on the meridian plane with Φ = 120◦ . Estimate the pressure
coefficient at A.
= 5º
0
= 30º
x
A
M = 2.4
= 1.4 = 10º
4.28 Use the wave drag coefficient for a cone of semi-vertex angle of 45◦ at M∞ =
1.0, from Fig. 4.36a to create the pressure drag behavior of the cone in the transonic
range from M∞ = 0.8 to M∞ = 1.5. Graph your results and compare it to the
experimental data in Fig. 4.37a.
4.29 Consider four cones with semi-vertex angles of 0◦ , 15◦ , 30◦ and 40◦ , as shown.
First for a flow Mach number of M∞ = 3.0, calculate the cones’ wave drag coeffi-
cients and graph them on a chart, (as shown). You may use conical shock charts of
Shapiro (Fig. 4.17) for γ = 1.4.
204 4 Shock-Expansion Theory
M
=0º
M
=15º
=30º
=40º
1.0
0.8
0.6
0.4
0.2
0.0
0 10 20 30 40
(deg)
Finally, if we install a Pitot tube on the 30◦ cone in the Mach 3 flow, calculate the
stagnation pressure that the Pitot tube reads, as a fraction of flight stagnation pressure.
4.30 Use the wave drag coefficient for an ogive of fineness ratio 3.0 at M∞ = 1.0,
from Fig. 4.36b. You first need to calculate the semi-vertex angle for a tangent ogive
that has a fineness ratio of 3.0 before you can use Fig. 4.36b. Next, calculate the
pressure drag behavior of the ogive in the transonic range from M∞ = 1.0 to M∞ =
1.4. Graph your results and compare it to the experimental data in Fig. 4.37b for the
ogive with fineness ratio 3.0.
4.31 A cone of 30◦ semi-vertex angle is in supersonic flow at M∞ = 2.0 and zero
angle of attack. Use conical shock charts to find:
4.6 Summary 205
Mc
M = 2.0
4.32 Consider a 30◦ (2-D) ramp in a Mach 3 flow of air with γ = 1.4. Also, there is
a cone of semi-vertex angle of 30◦ in the same supersonic flow condition. Calculate
and compare:
(a) The wave angles on the 2-D ramp and the cone
(b) The surface pressure on the ramp and the cone
(c) Explain the 3-D relieving effect as evidenced by this problem.
30º
1 b
M = 3.0
3
l
2
30º
M1 = 2.4
p1 = 30 kPa
1
T1 = -40 ºC
R = 287 J/kg.K
= 1.4
4.35 An oblique shock intersects a wall at a corner, as shown (not to scale). What
kind of wave(s) will form at the corner to accommodate the turning of the flow?
Calculate the Mach number in region 3.
1 M1 = 2.4 2 10º
30º
3
P-M E.F.
O.S.
M = 1.6 10º
= 1.4 y
x
O.S.
P-M E.F.
4.6 Summary 207
4.37 In the method of local cones, we introduced a function for the cone semi-apex
angle, δc that was related to the angle of attack (in radians), local body angle and the
meridian angle following:
α2
δc = δ − α cos Φ + cot δ · sin2 Φ
2
We wish to graph δc as a function of x for a tangent ogive, for α = 10◦ and the local
ogive angle δ varying between the nose angle, δ0 = 45◦ to the shoulder value, where
δ = 0 around the circumference of the ogive, i.e., with Φ varying between 0◦ and
360◦ . First make a table for δ versus x for a tangent ogive in increments of Δx = 0.1
along the axis of the ogive, i.e.,
Then, calculate (e.g., using a spreadsheet) and graph δc versus Φ at each x. Note
that there is a singularity at the shoulder where δ = 0 [due to cot (δ) term], that
corresponds to x = 1. To avoid the singularity at the shoulder; we limit x to 0.99.
Identify the shadow region on the ogive at angle of attack, where δc is negative.
4.38 Consider a flat plate at angle of attack in supersonic flow. Use shock-expansion
theory to calculate and graph the drag polar (i.e., graph of Cl vs. Cd ) for the airfoil
in the range of α from 0 to 12◦ for constant freestream Mach numbers of M∞ = 2,
3 and 5.
4.39 A symmetrical diamond airfoil has a semi-vertex angle of 20◦ in a Mach 2 flow
in air (γ = 1.4).
(a) At what angle of attack, will the leading-edge shock detach?
(b) At what angle of attack will the flow downstream of the leading-edge shock be
sonic?.
4.40 Consider a 15◦ (2-D) ramp in a Mach-3 flow of air with γ = 1.4. Also, there is
a cone of semi-vertex angle of 15◦ in the same supersonic flow condition. Calculate
and compare:
(a) surface Mach number on the ramp and the cone
(b) the wave angles on the 2-D ramp and the cone
(c) the surface pressure coefficient on the ramp and the cone
(d) wave drag coefficient, referenced to base area for the ramp and the cone.
208 4 Shock-Expansion Theory
A H pa
pa pa pa
D
1 2 4
60º
y 3
M1 = 2.6 F Centerline
x
G
C
E
B
Centered expansion fan
Cp, min
Cp, crit
Cp, stag
References 209
References
1. Anderson, J.: Modern Compressible Flow with Historic Perspective, 3rd edn. McGraw Hill,
New York (2003)
2. Anon.: Equations, Tables, and Charts for Compressible Flow. NACA TM 1135, Moffett Field
(1953)
3. Krasnov, N.F.: Aerodynamics of Bodies of Revolution. Elsevier Publication, New York (1970)
4. Liepmann, H.W., Bryson, A.E.: Transonic flow past wedge sections. J. Aeronaut. Sci. 17(12),
745 (1950)
5. Liepmann, H.W., Roshko, A.: Elements of Gas Dynamics. Wiley, New York (1957)
6. Shapiro, A.H.: The Dynamics and Thermodynamics of Compressible Fluid Flow. Ronald Press,
New York (1953)
7. Taylor, G.I., Maccoll, J.W.: The air pressure on a cone moving at high speed. In: Proceedings
of Royal Society, vol. 139. London (1933)
Chapter 5
Method of Characteristics
Abstract The governing equations of motion for two and three-dimensional inviscid
irrotational supersonic flows are derived. Although these equations are nonlinear,
there are unique curves in the physical space, known as characteristics that turn
the governing partial differential equations into ordinary differential equations that
may be integrated along the characteristics. Therefore, the method of characteris-
tics (MOC) is an exact solution technique that is graphical in nature and since it is
nonlinear, it applies to upper transonic flow when the local Mach number is sonic or
supersonic. The theory of MOC is developed and applied to 2-D irrotational flows.
Within MOC, two competing techniques known as wave-field method and lattice-
point approach are presented. The first example that is detailed is the minimum-length
supersonic nozzle design. The principle of wave cancellation at a solid boundary is
used in supersonic nozzle design and it is well described. The next application is
the wave pattern in supersonic exhaust plume. Here, wave reflection from a free
boundary, i.e., the free shear layer, is introduced and analyzed. The cases of under-
and over-expanded exhaust plumes are considered and MOC is applied to capture
the spatial evolution of the jet. MOC is extended to axisymmetric irrotational flows.
Family of supersonic nozzles, deflecting jets, non-uniform inlet condition, stream-
lines and ducts and curved shocks are the applications of MOC that are treated in
this chapter. This chapter contains 9 MOC application examples and 21 practice
problems at the end of the chapter.
5.1 Introduction
The governing partial differential equations in fluid mechanics are nonlinear, even in
the case of 2-D inviscid, irrotational flows. Although, linearization as an approximate
method is successfully developed for thin, slender bodies at small angle of attack
in subsonic and supersonic flows, the accuracy of solution is diminished when per-
turbations are no longer small. In addition, we learned that some flow regimes, e.g.
transonic, may not be linearized even for slender bodies. Interestingly, in supersonic
flow where the governing partial differential equation is hyperbolic, there are unique
curves in physical space called characteristics that turn the governing nonlinear
V = ui + vj (5.1)
For irrotational flows where the curl of the velocity field vanishes, i.e.,
∂u ∂v
∇×V =0 → = (5.2)
∂y ∂x
The mathematical implication of the irrotationality condition (i.e., Eq. (5.2)) is that
a scalar potential function, Φ exists such that:
V = ∇Φ (5.3)
∂Φ
u≡ (5.4a)
∂x
∂Φ
v≡ (5.4b)
∂y
∂u ∂v u ∂ρ v ∂ρ
+ + + =0 (5.5)
∂x ∂y ρ ∂x ρ ∂y
∂u ∂u 1 ∂p
u +v =− (5.6a)
∂x ∂y ρ ∂x
∂v ∂v 1 ∂p
u +v =− (5.6b)
∂x ∂y ρ ∂y
Since the flow is isentropic, we may relate the pressure gradient to the density gradient
and the local speed of sound according to:
∂p ∂ρ
= a2 (5.7a)
∂x ∂x
∂p ∂ρ
= a2 (5.7b)
∂y ∂y
Combining Eqs. (5.6) and (5.7) substituting them for the density terms in Eq. (5.5)
we get the following equation in velocity components u and v and the local speed of
sound:
u 2 ∂u 2uv ∂u v2 ∂v
1− 2 − 2 + 1− 2 =0 (5.8)
a ∂x a ∂y a ∂y
Replacing the velocity components in Eq. (5.8) with partial derivatives of the scalar
potential function, Φ, we get the full potential equation in steady two-dimensional,
irrotational flows:
Φx2 2Φx Φy Φy2
1 − 2 Φxx − Φxy + 1 − 2 Φyy = 0 (5.9)
a a2 a
∂u ∂u ∂2Φ ∂2Φ
du = dx + dy = dx + dy (5.10a)
∂x ∂y ∂x 2 ∂x∂ y
∂v ∂v ∂2Φ ∂2Φ
dv = dx + dy = dx + 2 dy (5.10b)
∂x ∂y ∂x∂ y ∂ y
Therefore, the system of three equations for the three unknowns, Φx x , Φx y and Φ yy
are written as:
u2 2uv v2
1 − 2 Φxx − 2 Φxy + 1 − 2 Φyy (5.11)
a a a
∂u ∂v
The solution for Φx y , which is our ∂y or ∂x , is expressed in terms of Cramer’s rule
as the ratio of two determinants:
(1 − u 2 ) 0 (1 − v2 )
a2 a2
dx 0
du
0 dv dy
Φxy = (5.14)
(1 − u 2 ) −2uv (1 − v2 )
a2 a2 a2
dx 0
dy
0 dx dy
The condition for the indeterminate Φx y is satisfied by demanding the numerator and
denominator of Eq. (5.14) to be simultaneously zero. First, by setting the determinant
in the denominator equal to zero, we get the equation for the characteristics, namely:
u2 −2uv v2
1 − 2 (dy)2 − dx dy − 1 − dx = 0 (5.15)
a a2 a2
Dividing both sides of Eq. (5.14) by (dx)2 and rearranging, we get a quadratic equa-
tion for the slope of the characteristic curve, dy/dx:
2
u2 dy 2uv dy v2
1− 2 + 2 + 1− 2 =0 (5.16)
a dx a dx a
5.2 2-D Irrotational Flows 215
We note that the square root of (M 2 − 1) written at the end of Eq. (5.17) holds
the key to the nature of the characteristics, namely for subsonic flows the square
root becomes imaginary and thus we conclude that no (real) characteristics exist in
subsonic flows. In supersonic flows however, the square root yields real numbers and
thus we conclude that two (real) characteristics exist. At sonic flow, the square root
identically vanishes and thus one real characteristic exists. These in the language
of classifications of the second order partial differential equations (see Sect. 2.2.4)
are called elliptic, hyperbolic and parabolic respectively, corresponding to subsonic,
supersonic and sonic flows. Figure 5.1 shows the definition sketch for the velocity
vector in x − y plane, its components and the flow angle, θ that it makes with respect
to x-axis. Based on this, we express the velocity components in terms of the velocity
magnitude and the angle, θ, according to:
u = V cos θ (5.18)
v = V sin θ (5.19)
√
dy M 2 sin θ cos θ ± M 2 − 1
= (5.20)
dx characteristic 1 − M 2 cos2 θ
We express Mach number in terms of the Mach angle, μ in Eq. (5.20), to get
sin θ cos θ
cot μ
dy sin2 μ sin θ cos θ ± sin μ cos μ
= 2θ
= (5.21)
dx characteristic 1 − cos sin2 μ − cos2 θ
sin2 μ
u
x
216 5 Method of Characteristics
Equation (5.22) reveals that the local characteristics about a (local) velocity vector
with flow angle, θ, are indeed Mach waves above and below the velocity vector. The
characteristic that is above the velocity vector is called C+ and the characteristic
that is below the velocity vector is called C− . The local flow at point A and its two
characteristics are shown in Fig. 5.2, as definition sketch.
Now, by setting the determinant in the numerator of Eq. (5.14) equal to zero, we
establish the compatibility condition along the two characteristics, namely
2 2
1 − ua 2 0 1 − av 2
dx du 0 =0 (5.23)
0 dv dy
Using the slope of the characteristics dy/dx (from Eq. (5.17)) in Eq. (5.24) and the
substitutions (5.18) and (5.19), we simplify the compatibility equation to the follow-
ing form:
u2 u2
√
dv 1− a2 dy d (V sin θ) 1− a2
− auv2 ± M2 − 1
= = =−
du 1− v2 dx characteristic d (V cos θ) 1− v2
1− u2
a2 a2 a2
(5.25)
C+ Characteristic with
y / dx)A = tan( θA + µA )
( dy Tangent to C+ characteristic at A
(Left-Running Mach Wave)
VA (> a)
µA
θA µA
A Tangent to C- characteristic at A
(Right - Running Mach Wave)
C- Characteristic with
(dy / dx)A = tan( θA − µA )
Fig. 5.2 Local characteristics are right- and left-running Mach waves in two-dimensional super-
sonic flow
5.2 2-D Irrotational Flows 217
dθ ∓ dν = 0 (5.28)
Hence from an initial data line in 2-D supersonic flows, we establish both local char-
acteristic directions as well as the constants (K + and K − ) along those characteristics.
Although characteristic directions change as we march in the flow direction, the con-
stants that we established from the initial data line, (K + and K − )-, remain the same.
We had introduced the zone of action and zone of silence in Chap. 4 to lie within the
Mach cone downstream of a point in supersonic flow and the entire space upstream of
the Mach cone, respectively. In this chapter, the two characteristics that emanate from
a point in supersonic flow in the downstream direction are indeed consistent with the
boundaries of the zone of action. We can now extend this concept to upstream and
define a domain of dependence that impacts the flow at a point in supersonic flow that
lies within the upstream characteristics passing through the point. In this context, we
may regard the domain of dependence upstream of a point as the “past” events that
influenced the state of flow at point A and the region of influence in the downstream
direction to represent the “future” of the flow that is impacted by the “presence” at
point A. These physical arguments impact the finite differencing strategy (of informa-
tion flow) in computational fluid dynamics. Figure 5.3 is a definition sketch that shows
the concepts of domain of dependence and region of influence in supersonic flow.
The construction of the characteristic network relies on three unit processes. These
unit processes help calculate flow properties at internal points, wall points and at
shock points. First, the internal point is created at the intersection of two character-
istics that still lies within the flowfield. Consider points 1 and 2 in the x − y plane
218 5 Method of Characteristics
y
C+ Characteristic
Domain of Dependence VA
of point A A Region of Influence
of point A
C- Characteristic
Fig. 5.3 Definition sketch of intersecting C+ and C− characteristics and their physical significance
in supersonic flow
C-3
C+2
2
y2
M2
C-2
x
x2 x1 x3
(see Fig. 5.4) for which we know the flow inclination/direction, θ1 and θ2 as well as
the Mach numbers, M1 and M2 . From local Mach numbers, we calculate the Mach
angles, μ1 and μ2 as well as the Prandtl-Meyer (P-M) angles, ν1 and ν2 . These angles
are sufficient to establish the characteristic directions for C+1 , C−1 , C+2 and C−2
as well as the compatibility constants K +1 , K −1 , K +2 and K −2 along those charac-
teristics respectively. The intersection of C−1 and C+2 creates point 3, which shares
the same compatibility constant with K −1 and K +2 that form two equations with
two unknowns, i.e.,
θ3 + ν3 = θ1 + ν1 = K −1 (5.30a)
θ3 − ν3 = θ2 − ν2 = K +2 (5.30b)
The solution to Eqs. (5.30a) and (5.30b) yields the flow angle θ3 and (indirectly,
through P-M angle, ν3 ) Mach number at point 3:
5.2 2-D Irrotational Flows 219
K −1 + K +2
θ3 = (5.31a)
2
K −1 − K +2
ν3 = (5.31b)
2
To graphically construct point 3, from the known points 1 and 2, we graph the
C+2 characteristic at the average angle corresponding to points 2 and 3 (along a C+
characteristic), i.e., C+2 is graphed at the angle of
θ2 + θ3 μ2 + μ3
+ (5.32)
2 2
Similarly, the C−1 will be graphed at the average angles corresponding to points 1
and 3 on the C− characteristic, namely,
θ1 + θ3 μ1 + μ3
− (5.33)
2 2
Note that we are averaging the angles of the characteristics between points 1 and
3 and 2 and 3 and not the slopes (i.e., the tangent of the angles). Also, since we
are drawing straight lines for the characteristic network between the adjacent grid
points, in general the accuracy will improve if the points are closer to each other.
Figure 5.5 shows the average angles used in connecting C− and C+ characteristics
to create point 3.
We next treat the unit process that involves the intersection of a characteristic and
an adjacent wall. Consider a point in the flowfield (e.g., point 1 in Fig. 5.6) that has
its local flow angle, θ1 and the local Mach number, M1 known. Based on these, we
can establish the angles μ1 and ν1 from the Mach number M1 and thus calculate the
K +1 and K −1 corresponding to point 1 in the flow. Now, consider an adjacent wall
to point 1, as shown in Fig. 5.6.
y3 3
(C+ ) 2−3
⎛ θ 2 + θ3 ⎞ ⎛ μ2 + μ3 ⎞
⎜ ⎟+⎜ ⎟
⎝ 2 ⎠ ⎝ 2 ⎠
2
y2
x
x2 x1 x3
220 5 Method of Characteristics
x1 x2 x
The C−1 characteristic intersects the wall at point 2 with coordinates (x2 , y2 ), as
shown in Fig. 5.6. We thus conclude that K −2 and K −1 are equal, which means that
K −2 = θ2 + ν2 = θ1 + ν1 = K −1 (5.34)
The flow angle at point 2 coincides with the wall angle at point 2 (from flow tan-
gency condition), which means that we have an estimate for θ2 , based on the C−1
characteristic line that was drawn at the angle of (θ1 − μ1 ) from point 1 towards
point 2. The estimate of θ2 , from the known wall angle θw at (x2 , y2 ), substituted in
Eq. (5.34) yields an estimate for P-M angle ν2 , according to:
ν2 = K −1 − θw2 (5.35)
The P-M angle ν2 from Eq. (5.35) then yields M2 and μ2 . Now based on the new
information at the wall, namely μ2 , we can redraw the C−1 characteristic for the
average angle of
θ1 + θ2 μ1 + μ2
− (5.36)
2 2
The intersection of the new (C− )1−2 characteristic and the wall will result at a new
point 2, with its corresponding θw2 . Depending on the correction of the wall angle
in our calculations, i.e., θw2 − θw2 , we may have to iterate to reach the desired level
of accuracy. The factors that influence the correction angle of the wall are the wall
curvature and the proximity point 1 to the wall.
The next unit process involves the intersection of a characteristic, which is a
pressure wave, and an oblique shock. The drawing in Fig. 5.7 serves as a definition
sketch for this interaction.
At point 1 in the flowfield, we know the local Mach number, M1 and the flow
angle, θ1 . Therefore, we know K +1 , which remains constant on C+1 characteristic.
The point 2 is at the intersection of the C+1 characteristic and the oblique shock
wave, as shown in Fig. 5.7. The interaction of a characteristic, i.e., a Mach wave
or pressure wave, with an oblique shock causes the shock strength to change, and
a characteristic, i.e., a pressure wave, to be reflected from the shock wave. These
are schematically shown in Fig. 5.7 where the angle of the oblique shock, β∞ has
changed to β2 after the C+1 characteristic has interacted with the wave. Also note
that a C−2 wave is reflected from the oblique shock. Since point 2 is on the same C+
5.2 2-D Irrotational Flows 221
Fig. 5.7 Definition sketch of a C+ characteristic (pressure) wave interaction with an oblique shock
wave
K +2 = θ2 − ν2 = θ1 − ν1 = K +1 (5.37)
The flow angle at 2 is the turning or deflection angle across the oblique shock at 2,
which is related to M∞ and β2 via oblique shock relations. The strategy is to guess a
flow deflection angle, θ2 and then find the corresponding wave angle β2 from oblique
shock relations establish M2 and thus ν2 . Then check Eq. (5.36) to see whether θ2 −ν2
is indeed equal to K +1 . We then need to iterate on our choice of flow angle at 2 to
converge on a consistent solution. Depending on the application, we may set the level
of acceptable accuracy at for example 0.01◦ . The classical textbooks of Liepmann
and Roshko [5] and Shapiro [7] as well as the contemporary book by Anderson on
compressible flow [1] should be consulted.
Example 5.1 We have experimentally measured the flow directions and Mach num-
bers at two points (1 and 2) in a supersonic flow, as shown. Calculate:
(a) The flow angle and Mach number at the intersection of the C− 1 and C+2 , i.e.,
point 3.
(b) The angle of the C− characteristic between the points 1 and 3.
(c) The angle of the C+ characteristic between the points 2 and 3.
(d) The coordinates of point 3.
y
M1
y1 =1
1 (C− )1− 3
M3
3
y3 M1 1 (deg ) M2 2 (deg )
(C+ )2−3 2 +5 1.8 0
y2 =0 M2
2
x
x1 = x 2 =0 x3
222 5 Method of Characteristics
Solution:
Based on the Mach number and the flow angle specified for points 1 and 2, we
complete the table:
Note that the fineness of the characteristic net is set by the closeness of the points
along the initial data line.
The method of characteristics can be used to design the contour of a supersonic nozzle
or wind tunnel. A convergent-divergent nozzle accelerates the gas from subsonic flow
in the converging section, through sonic flow at the throat to supersonic flow in the
divergent section. The method of characteristics may be initiated at the sonic throat
and be extended to the diverging section with the help of the two characteristic
families, C+ and C− . The function of these characteristics, or Mach waves, is to
change the local flow direction by a small amount and thus accelerate or decelerate
the flow. Associating the turning angle with the strength of these waves, we may
assign a prescribed strength; say 1◦ or 2◦ , to each wave. We may then construct
a network of C+ and C− characteristics (i.e., of expansion Mach waves for nozzle
design) of known strength that cause the flow to change direction and thus accelerate
in supersonic flow. The local flow angle, θ, decreases when it encounters a Left-
Running Wave (LRW), which we called a C+ characteristic and the local flow angle
5.3 Design of a 2-D Supersonic Minimum-Length Nozzle (MLN) 223
x x
Fig. 5.8 Basic expansion corners that create two families of characteristics that change the flow
angle to ±θ a Encountering Right-Running expansion Mach waves increasing the flow angle, θ.
b Encountering Left-Running expansion Mach waves decreasing the flow angle, θ
y e
θ w, max
th
C+ C+
C+
Mth =1.0 C- Uniform flow in the
θ th =0 Nozzle exit plane
γ, R
Centerline x
Fig. 5.9 Centered expansion waves at the throat of a choked nozzle with uniform flow that form a
network of C+ and C− characteristics
sonic line is straight and the flow at the throat is uniform, as shown in Fig. 5.9.
The condition of uniform flow is reasonable for nozzles with slow area-variation
convergent sections. However, if the area variation in the converging section prior to
the throat is large, it will affect the throat curvature and thus the sonic line is curved,
which causes flow non-uniformity. For additional discussion on this topic the reader
may consult Shapiro [7] and Ferri [3].
We take advantage of flow symmetry with respect to the x-axis and use the MOC
to design only the upper contour (or upper half) of the nozzle. The lower contour is the
reflection of the upper contour through the mirror of the x-axis. In essence since the
x-axis is a streamline, we may treat it as a solid wall. The reflected C+ characteristics
are then treated as reflections from a solid wall. Note that the symmetry argument
allowed the number of the nodes in the characteristic network to be reduced by
half. Figure 5.9 shows a sonic throat with uniform flow and four C− characteristics
when we include the first, i.e., the μ = 90◦ , characteristic. The reflection of the
90◦ C− characteristic from the x-axis is upon itself, as expected. However, it is
the next C− characteristic with a prescribed incremental flow turning angle, θinitial
that creates a reflection as C+ characteristic, as shown in Fig. 5.9. The subsequent
C− characteristics emanating from the sharp corner cause additional turning and
reflections that create the network of C− and C+ characteristics. Returning to the
four characteristics shown in Fig. 5.9, we conclude that the characteristic that sets off
the MOC is the one with a non-zero θinitial . Let us use an expanded view of Fig. 5.9
and label the characteristic nodes, for further analysis (see Fig. 5.10). Note that the
exit Mach number, Me , is considered to be a nozzle design parameter, i.e., known.
Consequently, the Prandtl-Meyer angle at the nozzle exit, νe , is also a known design
parameter. The ratio of specific heats of the gas is γ and is treated as constant in our
application of MOC. In reality, the expansion process in the nozzle causes a reduction
in the gas temperature and thus chemical composition of the gas and γ change. The
same arguments apply to the gas constant, R, which changes with gas composition.
However, we treat both γ and R as constant, which means that we assume frozen
chemistry in the nozzle.
5.3 Design of a 2-D Supersonic Minimum-Length Nozzle (MLN) 225
Fig. 5.10 The grid points in a characteristic network in a 2-D supersonic nozzle
According to Fig. 5.10, the flow angle at node 8 is zero, θ8 = 0, since it lies on
the x-axis (i.e., plane of symmetry). The C− characteristic that passes through node
8 and the sharp corner at point “A” has to maintain the same K − value, namely
θ 8 + ν8 = θ A + ν A (5.38)
Node 8 is the last node on the x-axis; therefore its Prandtl-Meyer angle is the same
as the exit value, νe . Also, since the flow turning angle started from zero at the sonic
condition the value of the wall angle is the same as the corresponding Prandtl-Meyer
angle, as discussed in Chap. 4. Therefore, Eq. (5.38) may be re-written as
We have now established a geometrical relation between the maximum wall angle,
θw, max and the Prandtl-Meyer angle at the nozzle exit, i.e., a nozzle design parameter
corresponding to the exit Mach number, Me . We have shown:
νe
θw, max = (5.40)
2
Since we started the MOC with the selection of a small θinitial angle, node 1 attains
that angle. However node 1 is on the x-axis and its flow angle should be identically
zero. This indeed is the small discrepancy that our MOC introduces in the design
of the supersonic nozzle. The characteristic that connects points A and 1 is a C−
characteristic that is also a Mach wave. We first determine the Mach number and the
corresponding Mach and Prandtl-Meyer angles at node 1. The flow angle was the
θinitial that we had chosen to begin the MOC procedure, i.e.,
θ1 = θinitial (5.41)
226 5 Method of Characteristics
Since the flow expanded from the sonic throat condition and turned by the θinitial
angle, the Prandtl-Meyer angle attains the same value, namely
ν1 = θinitial (5.42)
K −1 = θ1 + ν1 = 2θinitial (5.43a)
K +1 = θ1 − ν1 = 0 (5.43b)
νA = θA (5.44b)
Equation (5.45) is one equation with two unknowns, θ2 and ν2 . To establish a second
equation involving the same unknowns, we note that node 2 is on the C+ characteristic
that passes through node 1, therefore, it shares the same K + value with node 1, i.e.,
K +2 = θ2 − ν2 = K +1 = 0 (5.46)
θ2 = ν2 = θinitial + Δθ (5.47)
From Prandtl-Meyer angle ν2 we get the flow Mach number M2 and subsequently
the Mach angle, μ2 . Now, we have two straight characteristics to graph, one from A
to node 2 and one from node 1 to node 2. But this process is identical to the worked
Example 5.1.
The third C− characteristic takes the flow turning at “a” through an additional
turning angle, Δθ. Following the same arguments as node 2, we get the angles θ3
and ν3 and the corresponding K values as
5.3 Design of a 2-D Supersonic Minimum-Length Nozzle (MLN) 227
K +3 = θ2 − ν3 = 0 (5.48c)
The connection between the design exit P-M angle and the steps in our MOC is
revealed in Eq. (5.49a). Namely, for a give design νe , we divide the P-M angle by
two and select the initial turning angle and the subsequent steps in Δθ. In the simple
example of Fig. 5.10, we have taken two steps beyond θinitial .
Continuing with our calculations, we note that the flow properties at node 4 and
3 are the same, i.e.,
We graph the wall angle between the two nodes on the wall at the average flow angles
at the two nodes, which in the case of A and 4, is θw, max . Now, let us proceed to
node 5 on the x-axis. We conclude that the flow angle at 5 has to be zero and it lies
on the same C− characteristic as node 2, therefore
θ5 = 0 (5.51a)
K −5 = K −2 (5.51b)
We get the Prandtl-Meyer angle at node 5 from Eqs. (5.51a) and (5.51b) to be:
The C+ characteristic that is reflected form the plane of symmetry has the value
K −3 + K +5
θ6 = = Δθ (5.54a)
2
228 5 Method of Characteristics
K −3 − K +5
ν6 = = νe = Δθ (5.54b)
2
K −6 = νe (5.54c)
ν7 = ν6 = νe − Δθ (5.55b)
The wall that connects nodes 4 and 7 is drawn at the average of flow angles at 4 and
7, namely
θ4 + θ7 θw, max + Δθ
θw |4−7 = = (5.56)
2 2
Proceeding to node 8, we note that it lies on the x-axis and it is the last node, as we
had discussed earlier. Therefore,
θ8 = 0 (5.57a)
ν8 = ν e (5.57b)
K −8 = νe (5.57c)
K +8 = −νe (5.57d)
The flow properties at node 9 on the wall are the same as those at node 8. Therefore
the wall angle connecting nodes 7 and 9 is the average of the two flow angles at 7
and 9, namely
θ7 + θ9 Δθ
θw |7−9 = = (5.58)
2 2
Equation (5.58) that states that the angle of the last wall segment is one half of our
step size, Δθ, is a general result. This means that regardless of the number of steps
in the Δθ selection in MOC, the last wall segment makes an angle of Δθ/2 with
respect to the x-axis.
Based on our graphical construction, the ratio y9 /ya in the x − y plane represents
the ratio of exit-to-throat height, which for two-dimensional nozzles, it is the ratio of
exit-to-throat area. This should theoretically be equal to the Ae /A∗ corresponding
to Me and γ. However, there will be a discrepancy that is attributed to our non-zero
θinitial angle as well as the finite number of steps, i.e., Δθ’s, which we took to reach
νe /2. A reduction in θinitial has the largest impact on reducing the discrepancy in
area ratios followed by the size of Δθ steps. Now, we have completed the design of
a 2-D, minimum-length isentropic nozzle with uniform inlet condition using MOC,
albeit with only three C− characteristics to demonstrate the principle.
5.3 Design of a 2-D Supersonic Minimum-Length Nozzle (MLN) 229
Solution:
The Prandtl-Meyer angle corresponding to Me = 2.2 is νe = 31.73◦ . The maximum
wall angle is one half of νe , therefore the maximum wall angle is
θw, max = 15.865◦
A suitable choice may be θinitial = 0.865◦ and the remaining 15◦ divided into five
steps of 3◦ each, i.e., let Δθ = 3◦ .
There is a rule on the number of characteristic nodes in a Minimum-Length-Nozzle
(with uniform inlet condition), which states that:
n
No. of nodes = [2 + (n + 1)] (5.59)
2
where n is the number of C− characteristics in the problem, i.e., the number of θ
sub-divisions at the throat. In this example, we chose a total of six subdivisions (i.e.,
θinitial and five Δθ’s). Therefore, we expect (9) · (3) = 27 grid points. Had we chosen
7 subdivisions, we would have produced (10) · (7/2) = 35 characteristic nodes.
Tables 5.1, 5.2 and 5.3 show the flow parameters at the characteristic nodes, the
characteristic slopes and the wall slopes, respectively. In addition to the flow vari-
ables, the Cartesian coordinates of the grid and wall points are also tabulated. The
graph of the upper half of the nozzle contour is shown in Fig. 5.11. The lower contour
is symmetrical with respect to the centerline.
The nozzle area ratio based on the MOC and our graphical construction is thus
2.0117, since the 2-D nozzle has a unit depth. The isentropic area ratio is based on
continuity equation:
Table 5.2 Slopes of C− and Nodes (dy/dx) char Nodes (dy/dx) char
C+ characteristics
A to 1 −2.51235 1 to 2 2.126231
A to 2 −1.27408 2 to 3 1.576156
A to 3 −0.897 3 to 4 1.424069
A to 4 −0.68756 4 to 5 1.373924
A to 5 −0.5307 5 to 6 1.360623
A to 6 −0.40374 6 to 7 1.361322
2 to 8 −1.17509 8 to 9 1.064162
3 to 9 −0.87118 9 to 10 1.037232
4 to 10 −0.67937 10 to 11 1.033785
5 to 11 −0.53173 11 to 12 1.04338
6 to 12 −0.40801 12 to 13 1.049892
9 to 14 −0.84059 14 to 15 0.837755
10 to 15 −0.67344 15 to 16 0.845641
11 to 16 −0.53676 16 to 17 0.865011
12 to 17 −0.42178 17 to 18 0.879234
15 to 19 −0.68017 19 to 20 0.696655
16 to 20 −0.55126 20 to 21 0.721989
17 to 21 −0.44178 21 to 22 0.735424
20 to 23 −0.57345 23 to 24 0.601679
21 to 24 −0.4671 24 to 25 0.618436
24 to 26 −0.49611 26 to 27 0.51031
5.3 Design of a 2-D Supersonic Minimum-Length Nozzle (MLN) 231
Table 5.3 Slopes of the Nodes (dy/dx) char Nodes (dy/dx) char
nozzle wall segments
A to 7 0.284045 18 to 22 0.131585
7 to 13 0.247946 22 to 25 0.078662
13 to 18 0.185243 25 to 27 0.026173
γ−1 2
γ+1
A 1 1+ 2 M
2(γ−1)
= γ+1
(5.60)
A∗ M
2
The area ratio for an exit Mach number of 2.2 and a gas with γ = 1.4 is
(A/A∗ )isentropic = 2.004975. We note that the area ratio based on the MOC (2.0117)
is within 0.33 % of the theoretical value. This level of accuracy is remarkable when
we consider that we only did a 27-point calculation.
In Fig. 5.11, the wall point at the throat is at (1, 1) whereas the nozzle exit coor-
dinate at the wall, i.e., grid point 27, is at:
Since xthroat is at 1, the physical distance of the last C− characteristic that intersects
the axis of symmetry is at 2.3346 (units of throat half-width) downstream of the throat.
The implication here is that the flow accomplishes its exit condition on the axis of
symmetry at only 2.3346/6.2537 or 37.3 % of the nozzle length. The flow continues
to evolve away from the axis up to the wall. Therefore the condition of uniform flow
is accomplished after the entire flow has accelerated to the exit condition. Figure 5.12
shows the evolution of Mach number along the centerline of the supersonic nozzle.
2.5
25 27
2 18
22
13
1.5 7
A
1 6
12
17
0.5 21
24
26
1 8
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8
Fig. 5.11 The outer contour of a minimum-length 2-D isentropic nozzle with uniform inlet condi-
tion designed for exit Mach number of 2.2 using MOC (γ = 1.4)
232 5 Method of Characteristics
2
M 1.5
1
1 2 3 4 5 6 7 8
x
Fig. 5.12 The evolution of Mach number along the nozzle centerline (γ = 1.4)
Note that the waviness of the Mach distribution curve along the nozzle axis is caused
by sparse (only 6 in this example) data points along the centerline (remember that
we used 6 C− characteristics to accelerate the flow at the throat). Increasing the
number of the characteristics, or equivalently, reducing the turning angle, Δθ, of
each characteristic will smooth the Mach distribution curve along the nozzle axis.
In supersonic flow the turning is accomplished by two kinds of waves, the right-
running and left-running waves that may be of expansion or of compression type.
Therefore there are two families of waves that dictate the flow direction and mag-
nitude in supersonic flow. There are regions in the flow where only one type of
waves uniquely determines the flow variables and they are called simple-wave flow.
Figure 5.13 shows a convex wall where the flow upstream of the shoulder (point 1)
has been uniform and parallel. The subsequent turns in the wall are accomplished by
left-running Mach waves, as shown. Here we note that the Prandtl-Meyer expansion
fans at points 1, 2 and 3 are replaced by an “average” single Mach wave that is
based on the average Mach numbers on the two sides of the respective expansion
corners. The Mach angles that are shown in Fig. 5.13 are based on the average Mach
numbers in regions 1 and 2, 2 and 3, and 3 and 4 respectively. Therefore crossing of
each wave implies a net turning that is dictated at the wall. The left-running waves turn
the flow in clockwise, i.e., in negative θ, direction whereas right-running
waves turn the flow in the counter-clockwise, i.e., the positive, direction. For example,
LRW
V Streamline LRW
1
M1 2 LRW
=0 M2
V 1 µ1 µ2 M3
y 2 µ3 3
1
x 2 3
1 3
M4
2 4
3 4
Fig. 5.13 Simple-wave problem in supersonic expansion flow over a convex surface with LRWs
5.4 Wave-Field Method Versus Lattice-Point Approach 233
when the flow crosses from field 1 to field 2 in Fig. 5.13, the flow angle is reduced
by the turning at the wall at point 1. The Prandtl-Meyer angle, ν, also jumps across
the waves. In an expanding flow, such as the convex wall in Fig. 5.13, the Prandtl-
Meyer angle jumps up by the amount of wall turning angle, since the Mach number
in flow increases along a streamline that crosses the waves emanating from corners
1, 2 and 3.
For the general case where the flow upstream of corner 1 makes an angle, θ1 , with
respect to the x-axis, the flow angles in regions 2, 3 and 4 are:
θ2 = θ1 − Δθ1 (5.61a)
The Prandtl-Meyer angles in regions 2, 3 and 4 are related to the P-M angle in
region 1 and the wall turning angles, following:
ν2 = ν1 + Δθ1 (5.62a)
ν3 = ν2 + Δθ2 (5.62b)
ν4 = ν3 + Δθ3 (5.62c)
From Eqs (5.62a), (5.62b), and (5.62c) we get the Mach numbers in regions 2, 3 and
4 respectively. The waves at the corners, as noted earlier, are C+ characteristics that
are drawn at the average Mach and flow angles according to definition:
dy
= tan(θ̄ + μ̄) (5.63)
dx C+
θ1 + θ2
θ¯1 = (5.64a)
2
θ2 + θ3
θ¯2 = (5.64b)
2
θ2 + θ3
θ¯3 = (5.64c)
2
2
μ¯1 = sin−1 (5.64d)
M1 + M2
2
μ¯2 = sin−1 (5.64e)
M2 + M3
234 5 Method of Characteristics
2
μ¯3 = sin−1 (5.64f)
M3 + M4
Figure 5.14 shows the definition sketch of the physical Mach waves and the “mean”
Mach waves in an expansive supersonic flow over a convex surface.
It is important to note that the concept of “mean” Mach wave, as described by
Eqs. (5.64d)–(5.64f), is mainly used as a tool that facilitates the numerical-graphical
solution to supersonic flows using wave-field method. It is used to replace the head
and tail Mach waves that envelope the P-M expansion fans. However, these are all
physical waves, including the mean Mach wave that exist in nature and are funda-
mental to supersonic flows (see Fig. 5.14).
The counterpart of a convex wall is a concave surface, which in supersonic flow
creates compression Mach waves to turn the flow. The P-M angle across a (left-
running) compression Mach wave is thus reduced by the turning angle of the wall. The
reduced P-M angle across compression Mach waves creates a reduction in flow Mach
number and an increase in Mach angle. Consequently, the compression Mach waves
on a concave wall converge, in contrast to the expansion Mach waves that diverge.
Figure 5.15 shows the simple wave problem in supersonic flow over a concave wall
and a representative streamline.
The flow deceleration across (the left-running) compression Mach waves as shown
in Fig. 5.15 satisfy the P-M relations:
ν2 = ν1 − (θ2 − θ1 ) (5.65a)
ν3 = ν2 − (θ3 − θ2 ) (5.65b)
ν4 = ν3 − (θ4 − θ3 ) (5.65c)
Note that the left-running compression Mach waves cause an increase in flow
angle, i.e., θ4 > θ3 > θ2 > θ1 , therefore P-M relations (5.65a)–(5.65c) result in a
____
Physical Mach Waves
Mean Mach wave ___
“Mean” Mach Waves
Head Mach wave
Tail Mach wave
M1 µ2 M
2 µ3 Physical Mach Angle:
µ1 M3
µ2 µ = sin-1(1/M)
y 1 2 µ3 µ 4
θ1 3
x θ2 M4
θ3 4
Convex wall θ4
Fig. 5.14 Definition sketch for the physical and mean Mach waves in supersonic flow over expan-
sion corners
5.4 Wave-Field Method Versus Lattice-Point Approach 235
4
M1 4 θ4
y 3
x Streamline 1 2
θ3
3
2 θ2
1
Concave wall
Fig. 5.15 Supersonic flow over a 2-D concave surface—A simple wave problem with compression
Mach waves
reduction in P-M angle, i.e., ν4 < ν3 < ν2 < ν1 , which in turn cause a reduction
in Mach number, M4 < M3 < M2 < M1 . To generalize the principle, we now use
right-running compression Mach waves on a concave surface a shown in Fig. 5.16.
The flow angle is reduced across the right-running compression Mach waves,
which is the opposite of the left-running waves of the same type. However, the flow
deceleration still applies, namely M4 < M3 < M2 < M1 . Therefore to generalize
the P-M relations that we used in the case of left-running compression Mach waves,
Eq. (5.65a)–(5.65c), may be cast in terms of the absolute value of flow turning angle
according to:
ν2 = ν1 − |θ2 − θ1 | Compression Mach wave (5.66a)
1
2 Concave wall
3
Streamline
1 2
3
y 4
x 4 4
Fig. 5.16 Turning of supersonic flow over a 2-D concave surface by right-running compression
Mach waves
236 5 Method of Characteristics
The generalized Eq. (5.65a)–(5.65c) apply to both types of compression Mach waves,
i.e., RRW as well as LRW. By using the same principle, the generalized case of a
convex surface in supersonic flow where expansion Mach waves cause a flow turning
and acceleration follow the P-M relations:
Figure 5.17 is a definition sketch for an expansive flow over a convex surface with
right-running (Mach) waves.
The non-simple wave problem in supersonic flow involves the creation and inter-
action of both types of waves, i.e., left-running and right-running waves. We observed
an example of non-simple wave in the design of minimum-length supersonic nozzle
where the throat created C− characteristics (as in Fig. 5.10) and the C+ character-
istics were created by the reflection of C− characteristics from the x-axis, i.e., the
nozzle centerline. Now, we are ready to apply the wave-field method to isentropic
supersonic flows with waves of both families.
The unit processes in wave-field method are: wave generation, wave reflection
from a solid boundary, wave reflection from a free boundary and wave cancelation.
The method is best learned through an example. Wave-field approach is applied to
design of a minimum-length 2-D nozzle as shown in Fig. 5.18. The exit Mach number
is 2.06, which requires 14◦ of flow turning to accelerate a sonic flow to Mach 2.06
flow. Subsequently, seven waves of each 2◦ strength are created at the sharp corner,
A (see Fig. 5.18). The numbers in cells are characteristic numbers. Once we learn the
rules for assigning the characteristic numbers, the process becomes easy. The first
rule is that the sum of the two numbers, 600 and 400, at the sonic cell adds up to
1,000. In subsequent cells the sum is less than 1,000. In the first cell after the sonic
throat, where the flow has crossed a 2◦ turn (Left-Running) Mach wave, the sum of
the two numbers is 998 (i.e., 598 and 400). Therefore the difference between 1,000
4
Convex wall
3 RRW
2 4
1
Streamline RRW
3
y
x 2 RRW
1
RRW
Fig. 5.17 A convex wall in supersonic flow with expansion Mach waves (RRW)
5.4 Wave-Field Method Versus Lattice-Point Approach 237
Minimum-Length Nozzle
B
Wave Reflection
598 598
400 398
600
400 586
386 586 Exit Mach number
586
400 388 Me = 2.06
A
586
398
Wave Generation
Wave Cancellation Region
Fig. 5.18 A 2-D minimum-length nozzle (MLN) designed using wave-field method (adapted from
[3])
and the sum of the numbers in the cell constitute the P-M angle, ν, which is equal to
θ in magnitude. Now we present the rule about the difference of the two numbers.
The difference between the numbers in the cell minus 200 is the flow angle in the
cell. Note that at the sonic throat the difference between 600 and 400 is exactly 200,
which means that the flow angle at the throat is zero. In the next cell, the difference
between the numbers 598 and 400 is 198, which taking away 200 leaves a −2◦ of
flow angle in that cell. The next cell where the numbers 598, 398 are written mean
the following: the sum is 996, which means the P-M angle in that cell is 4◦ . The
difference between the characteristic numbers is 200, which means that the flow
angle is 0. Note that the cell in discussion is adjacent to the centerline, where flow
has to attain zero-angle due to flow symmetry. In addition the flow has crossed a
LRW and its reflection, i.e., a RRW; therefore the flow angles of −2◦ encountered
by a LRW is canceled by the +2◦ induced by the RRW that was reflected from the
centerline. The reason for the P-M angle to be 4◦ is due to crossing of two 2◦ Mach
waves, which add up to 4◦ . The cell that has the maximum wall angle is designated
with two numbers, 586 and 400. The sum of the numbers is 986, which means that
the P-M angle in that cell is 14◦ . The difference between the two numbers is 186,
which implies the flow angle in the cell is −14◦ . Note that the flow angle crossing
the LRW is reduced; therefore we expect the flow angle to increase as we cross
the RRW, as in the reflection of the LRW from the centerline. The last cell that we
examine is characterized by two numbers, 586 and 386. Their sum is 972, which
implies that the P-M angle is 28◦ . This angle corresponds to the exit Mach number
of 2.06. So far we have experienced the wave generation at the throat where seven
LRWs created the initial expansion. We also experienced the wave reflection from a
solid boundary, namely the centerline behaves as solid boundary. The rule was that
the refection from a solid boundary is in like-manner where expanding Mach waves
reflect as expanding Mach waves, except the reflections are of the opposite family,
namely the RRWs in this case.
238 5 Method of Characteristics
Table 5.4 Characteristic parameters that are used in wave-field approach (γ = 1.4)
N ν (deg) M μ (deg) p/ p ∗
1,000 0 1.000 90.000 1.000000
999 1 1.081 67.640 0.907248
998 2 1.133 61.988 0.851025
997 3 1.176 58.217 0.804867
996 4 1.219 55.153 0.761912
995 5 1.258 52.665 0.723304
994 6 1.294 50.617 0.688964
993 7 1.330 48.763 0.655897
992 8 1.365 47.114 0.624887
991 9 1.400 45.585 0.594836
990 10 1.435 44.176 0.566083
989 11 1.469 42.901 0.539221
988 12 1.503 41.701 0.513255
987 13 1.537 40.595 0.488759
986 14 1.571 39.546 0.465116
985 15 1.605 38.551 0.442326
984 16 1.638 37.620 0.420759
983 17 1.672 36.730 0.399939
982 18 1.706 35.876 0.379809
981 19 1.741 35.061 0.360539
980 20 1.775 34.283 0.342107
979 21 1.810 33.542 0.324543
978 22 1.845 32.828 0.307674
977 23 1.880 32.143 0.291536
976 24 1.915 31.483 0.276111
975 25 1.950 30.847 0.261340
974 26 1.986 30.233 0.247249
973 27 2.024 29.617 0.233236
972 28 2.060 29.041 0.220336
971 29 2.097 28.484 0.208035
970 30 2.132 27.972 0.196900
969 31 2.173 27.405 0.184785
968 32 2.211 26.897 0.174144
967 33 2.249 26.407 0.164089
966 34 2.288 25.918 0.154276
965 35 2.329 25.428 0.144668
964 36 2.369 24.965 0.135835
963 37 2.411 24.508 0.127330
962 38 2.453 24.063 0.119294
(continued)
5.4 Wave-Field Method Versus Lattice-Point Approach 239
The reflected waves from the centerline are still responsible for more flow accel-
eration in subsequent cells, as their expansive nature is preserved in a like-manner
reflection. We learn that the opposite is true for the reflection of waves from free
boundaries when we discuss under-expanded nozzle flows. The wave cancellation is
accomplished at the wall where we turn the wall to meet the flow angle upstream of
the wave, i.e., before it meets the wall. To demonstrate this point let us examine the
cell next to the maximum wall angle cell where the characteristic numbers are 586
and 398. The sum is 984 which means that the flow P-M angle is now 16◦ , which
is 2◦ more than the previous adjacent cell (with 586 and 400 numbers). The dif-
ference between the numbers is 188, which means that the flow angle in the cell is
5.4 Wave-Field Method Versus Lattice-Point Approach 241
Intermediate
Streamlines
Intermediate
Streamline
Fig. 5.19 Wave-field method is used to design a 2-D supersonic nozzle with an extended initial
expansion region that produces longer nozzle than MLN (adapted from [3])
−12◦ . Since the flow angle in the wall cell and the wall angle have to match, the
wall angle is −12◦ . The cell where the maximum wall angle is a boundary had a
flow and wall angle of −14◦ ; therefore the wall has turned in the positive θ direc-
tion, i.e., in counter-clockwise direction by 2◦ . This is the wave cancellation step in
wave-field method. We note that the wave strength does not change in reflections,
i.e., the 2◦ that we started the initial expansion persists through reflections. In this
example, we learned about the characteristic numbers, the wave generation step, the
wave reflection from a solid boundary step and the wave cancellation step at the wall.
To facilitate the numerical-graphical calculation, a suitable table that has a column
for characteristic number, N , which is the sum of the two numbers in a cell, is added
to a P-M table that includes Mach angle and p/ p ∗ as additional columns. Table 5.4
shows the characteristic parameters that are used in the wave-field approach in MOC
for a gas with γ = 1.4. These parameters and more detail may be found in NACA
Report 1135 [2].
The wave generation step in Fig. 5.18 was concentrated at a point that led to the
design of minimum-length nozzle (MLN). However the concentrated expansion at the
throat is very abrupt which may affect the flow quality at the nozzle exit. The MLN
is suitable for flight due to its reduced weight/cost, but in stationary applications,
such as supersonic wind tunnels, the weight or cost advantage is over-shadowed by
the flow quality advantage of accomplishing uniform, wave-free flow condition in
the test section. As a result, supersonic nozzles with an arbitrarily stretched wave-
generation region are designed for stationary applications. The example shown in
Fig. 5.19, which is adapted from Ferri [3], shows a spread of the wave generation near
the throat, followed by wave reflections from the centerline and wave cancellations
at the wall similar to the previous MLN case. In this example Ferri addresses the
case of a nozzle with fixed length and then works in the reverse order to establish
the waves that are necessary in its design. Our purpose was however to show that the
wave generation region may be spread to accomplish higher quality flow in the exit
plane/test section of a supersonic wind tunnel.
242 5 Method of Characteristics
Fig. 5.20 Incident and reflected pressure waves (of expansion and compression types) from a solid
wall
Example 5.3 Show that the strength of the reflected (isentropic) pressure wave from
a solid wall is the same as the strength of the incident wave, i.e., show:
(Δθ)reflected = −(Δθ)incident
(Δν)reflected = −(Δν)incident
Solution:
We consider two cases of expansion and compression waves, as shown in Fig. 5.20.
The proof of the first part, namely flow turning of the incident wave from a solid wall
is canceled by the reflected wave is imposed at the solid surface. Mathematically, we
state that:
θ1 = θ3 = θw (5.68)
The ± sign in Eq. (5.69) depends on whether the incident wave was of compressive or
expansive nature as well as whether the incident wave was of the left- or right-running
type. For example, the two RRWs that are the graphed as incident waves in Fig. 5.20,
cause the flow turning to be in the positive direction for the expansion wave and
in the negative direction for the compression wave. Also note that by choosing the
straight wall to lie along the x-axis, the flow angles in regions 1 and 3 as well as the
wall angle are all zero. Combining Eqs. (5.68) and (5.69) proves the first assertion:
(Δθ)reflected = −(Δθ)incident
Since the origin of this condition is set at the solid wall (and not by the nature of
the wave, i.e., whether the wave was isentropic or non-isentropic), the incident wave
may even be non-isentropic, e.g., an oblique shock wave. The flow turning induced
by an oblique shock is reversed upon reflection at a plane solid wall to satisfy the
flow tangency condition at the solid surface.
5.4 Wave-Field Method Versus Lattice-Point Approach 243
In Prandtl-Meyer flow theory, we have shown that the change in P-M angle is
equal to the magnitude of the flow turning angle. The expansion waves will cause
flow acceleration and thus P-M angle increases by the absolute value of the turning
angle. In this context, the LRW or RRW both cause an increase in P-M angle for an
expansion wave. The compression waves will cause flow deceleration and thus P-M
angle decreases upon encountering of such waves. This may be stated mathematically
for expansion waves as:
Therefore, the change in P-M angle across the incident expansion wave and its
reflection is:
ν2 − ν1 == (Δν)incident = |Δθ| (5.72a)
The change in P-M angle across the incident and reflected compression waves is:
The Eqs. (5.72) and (5.73) prove the second assertion that (Δν)reflected = (Δν)incident .
The case of wave reflection from a free boundary is of interest when we encounter
under-expanded (or over-expanded) nozzles. The problem statement is that a super-
sonic stream at Mach, Me , and static pressure, pe , that is discharged in an atmosphere
with static pressure, pa , where pe = pa has to adjust its pressure to match the ambi-
ent pressure at the free boundary separating the jet from the ambient. We know that
the interface between the two fluids, i.e., the free shear layer, cannot sustain a static
pressure jump. The case of pe > pa is called under-expanded flow and the oppo-
site is true for over-expanded flows. In under-expanded flows, there is an excess of
static pressure at the nozzle exit that needs to be (abruptly) relieved when it meets
the ambient fluid. The abrupt static pressure relief mechanism in supersonic flow is
through centered expansion fans, which are known as P-M expansion Mach waves.
The case of over-expanded flows where the nozzle exit pressure is lower than the
ambient static pressure, i.e., sub-atmospheric, the mechanism for abrupt rise in static
244 5 Method of Characteristics
pressure in supersonic flow is through shock waves. Figure 5.21 shows a definition
sketch of an under-expanded jet and its wave train in the exhaust plume. Representing
the centered expansion fan at the nozzle lip by only two (finite) expansion waves,
we arrive at field numbers 1–12. The jet centerline is the plane of symmetry, which
is a streamline and may thus be replaced by a solid wall. The reflection of expansion
waves from the centerline (i.e., a solid surface) is in like-manner; therefore further
expansions below atmospheric occur upon reflection. The jet in field number 3 is
perfectly expanded, which means that p3 = pa thereby maintaining the constant
pressure condition at the free boundary. To compensate for over expansion of the
flow in fields number 4 and 5, the expansion waves reflect as pressure waves from
the free boundary to create fields’ number 7 and 8. This is the principle of unlike
reflection at a free boundary. Compression waves that are generated at the boundary
reflect from the centerline (that behaves like a solid wall) as compression waves and
thus create fields with static pressure that are in excess to ambient pressure (in fields
number 10 and 11). The process of unlike wave reflections from the free boundary
and like reflections from the centerline repeats to form a periodic exhaust plume
shape with a spatial wavelength of λ, as depicted in Fig. 5.21.
The periodicity of the wave structures is noted in Fig. 5.21 by repeating cell
numbers 2, 3, 4 and 5 towards the end of the plume, where in theory the flowfield
is identical to their counterparts near the nozzle lip. The case of wave pattern in
an over-expanded nozzle is shown in Fig. 5.22. The waves that are formed at the
nozzle lip are shock waves (here are drawn as oblique shocks) and the subsequent jet
___
Free Shear layer
Streamline Expansion Waves
Compression Waves
Constant-Pressure pa Streamline
Free Boundary
pa 7 pa pa
pa
3
3 9
Me 5 5
pe > p a 8
γ, R
1 2 11 2
y 6 12
4 10 4
x
Wavelength, λ Centerline
pe = p1 p12 = pe
p2 p11
p 3 = pa p7 = pa p9 = pa
p5 < pa p8 < pa
Fig. 5.21 The wave pattern in the exhaust plume of an under-expanded jet and the pressure distri-
bution along a representative streamline (not to scale)
5.4 Wave-Field Method Versus Lattice-Point Approach 245
Oblique Shock
___
Free Shear layer
Streamline Expansion Waves
Constant-Pressure Compression Waves
pa Free Boundary pa Streamline
pa pa
pa pa pa
pa
Me pa pa
pe < p a pa
γ, R
y
x Centerline
Wavelength, λ
Incident and Reflected
Shock waves
pa pa pa pa
pe < pa
Fig. 5.22 The exhaust plume of an over-expanded nozzle with a representative Streamline and its
pressure distribution along the jet (not to scale)
contraction. It is through the process of like and unlike reflections from the centerline
and the free boundary respectively that gives the exhaust plume a periodic structure.
We recall that the strength of the shock wave at the lip depends on the severity of over-
expansion. It is even possible to have a normal shock inside the nozzle of severely
over-expanded flows. Figure 5.22 shows the case of an over-expanded nozzle where
the pressure rise across the oblique shock at lip is sufficient to reach the ambient
static pressure, pa .
Example 5.4 Calculate the initial exhaust plume angle for an under-expanded jet
with Me = 2.2, pe / pa = 2.0 with the ratio of specific heats, γ = 1.4.
Solution:
Since the centered expansion waves are isentropic, stagnation pressure remains con-
stant, i.e., pt2 = pt1 . From isentropic tables we get:
From the ratio of pt2 / p2 and isentropic table we get M2 ≈ 2.64. The flow turning
angle across P-M waves is Δθ = Δν, therefore from ν1 = 31.73 and ν2 = 42.31,
we get θ2 ≈ 42.31◦ − 31.73◦ ≈ 10.58◦ .
246 5 Method of Characteristics
2
2
M2
1
1
Me = M 1
Solution:
The wave-field approach in MOC is best suited for supersonic problems involving a
free boundary. The flow parameters at the exit are:
pte
Me = 1.435 → νe = 10◦ , μe = 44.176◦ and ≈ 3.344
pe
Since the expansion process at the nozzle lip is isentropic, the total pressure at the
nozzle exit remains constant along the jet, therefore
pt3 pte pe pa
= = 3.344 · 1.347 · 1 = 4.504 → M3 = 1.639 → ν3 = 16◦
p3 pe pa p3
The flow turning angle across a centered expansion fan is equal to the change of P-M
angle, Δν across the fan, namely
θ3 − θ1 = ν3 − ν1 = 6◦
Assuming that the jet axis coincides with the x-axis, we use the flow angle at the
nozzle exit equal to zero (i.e., as reference), and thus the flow angle in field number
5.4 Wave-Field Method Versus Lattice-Point Approach 247
θ1 = 0 and θ3 = +6◦
Therefore, the initial expansion at point A makes an angle of +6◦ with respect to
the x-axis. The equation for the free boundary AD is therefore:
y AD = 1 + tan(6◦ )x ≈ 1 + 0.105x
we attribute an average θ and μ that correspond to the two neighboring fields to the
wave. We use these average angles in the construction of the characteristic network.
With the known slopes and the starting coordinate, e.g., A, we start our march
in the streamwise direction by writing the equations of straight lines (to represent
waves). For example, we start with the wave AB.
With the coordinate of point B known in the previous step, we write the equations
for the waves AC and BC.
yAC = 1 − 0.69005x
y BC = 0.857326 · (x − 1.155)
From the intersection of AC and BC, we get the coordinates of point C, namely
xC = 1.2862 and yC = 0.11246. We continue with the remaining waves and their
intersections to establish the coordinates of the nodes.
From the intersection of AD and CD, we get the coordinates of point D, namely,
x D = 2.6327 and y D = 1.2764.
yDG = 1.27644
From yIK and the x-axis, we get the coordinates of point K , as: x K = 4.7255 and
y K = 0.
yIJ = 0.11246 + 0.69 · (x − 4.5943)
From the intersection of yIJ and yKJ , we get the coordinates of point J as x J = 5.8805
and y J = 1.0.
Table 5.7 is a summary of the coordinates of the characteristic nodes. To graph
the lower half of the jet, we have created a (−y) column that is the mirror image of
the upper half of the jet. Figure 5.23 shows the 2-D under-expanded exhaust plume,
the incident expansion waves, the reflected compression waves and the free (jet)
boundary. We also note that the waves and their reflections form a periodic plume
structure with a wavelength of 5.881 unit of jet half-width. Since we had neglected
fluid viscosity, the supersonic shear layer does not spread due to the actions of
fluid viscosity, turbulent stresses and wave interactions at the jet boundary. The
periodic plume structure is also subject to the dissipative actions of fluid viscosity
and turbulence.
250 5 Method of Characteristics
B
E
-0.5
-1.0
-1.5
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
x
Fig. 5.23 The exhaust plume of an under-expanded supersonic jet discharging into a stationary
ambient using wave-field approach in MOC [Uniform exit flow with Me = 1.435, pe = 1.347 pa
and γ = 1.4 = constant]
Wave-field approach is more suitable for plane flows where wave strength remains
constant, even after interaction with other waves. Also the wave method is advanta-
geous when there is a free boundary to be determined, as in under-expanded nozzle
flows. But in 3-D or axisymmetric flows, wave strength varies along the wave and
the method is less useful. The lattice-point MOC is readily applied to 2-D as well as
3-D flowfields.
5.5 Axisymmetric Irrotational Flows 251
The last term in Eq. (5.80) is identically zero, since w = 0 as well as ∂/∂ϕ = 0.
The first two terms in the continuity equation may be expanded to:
∂u ∂v ∂ρ ∂ρ ρv
ρ + +u +v + =0 (5.81)
∂x ∂r ∂x ∂r r
Axisymmetric
Shock x = constant planes
y
r = y2 + z 2 tan = dr/dx
r
r
M >1 nose
x
Axisymmetric
z Body
Fig. 5.24 A pointed axisymmetric body at zero-angle-of-attack and yaw in supersonic flow
252 5 Method of Characteristics
ρV dV = −d p (5.82)
The pressure term in Eq. (5.82) may be related to the square of the speed of sound
and density perturbation since the flowfield is isentropic, namely
d p = −a 2 dρ (5.84)
In partial differential form, Eq. (5.85) may be written as two equations, namely
1 ∂ρ 1 ∂u ∂v
=− 2 u +v (5.86a)
ρ ∂x a ∂x ∂x
1 ∂ρ 1 ∂u ∂v
=− 2 u +v (5.86b)
ρ ∂r a ∂r ∂r
We may now eliminate the density form the continuity Eq. (5.81), to get:
∂u ∂v v u ∂u ∂v v ∂u ∂v
+ + − 2 u +v − u +v =0 (5.87)
∂x dr r a ∂x ∂x a2 ∂r ∂r
Irrotational flow demands that the curl of the velocity field to vanish, namely:
e re e
1 ∂r ∂ϕ ∂x
∇ × V = ∂r ∂ϕ ∂x = 0 (5.89)
r
v rw u
The remaining two complimentary equations in du and dv are written similar to the
2-D case as:
∂u ∂u ∂u ∂v
du = dx + dr = dx + dr (5.92)
∂x ∂r ∂x ∂x
∂v ∂v
dv = dx + dr (5.93)
∂x ∂r
Now, we have formed the system of three Eqs. (5.91)–(5.93) and three unknowns
(∂u/∂x, ∂v/∂x and ∂v/∂r ) in cylindrical polar coordinates. We solve for ∂v/∂x
using Cramers rule:
(1 − u 2 /a 2 ) −v/r (1 − v2 /a 2 )
dx du 0
∂v 0 dv drN
= = (5.94)
∂x
(1 − u /a ) −2uv/a (1 − v /a )
2 2 2 2 2 D
dx dr 0
0 dx dr
By setting the denominator equal to zero, we get the equations for the characteristics,
namely:
(1 − u 2 /a 2 )dr 2 − dx −2uv/a 2 dr − 1 − v2 /a 2 dx = 0 (5.95)
Note that Eq. (5.95) is identical to (5.15) for 2-D flow except coordinate “y” is
replaced with coordinate “r ”, therefore the slope of the characteristics in 3-D axisym-
metric flow which is dr/dx follows (Eq. (5.22)):
dr
= tan(θ ∓ μ) (5.96)
dx characteristic
As expected, the characteristics are Mach waves that lie above and below the local
velocity vector. Since we are dealing with 3-D flows, the two characteristics are
indeed the projections of the Mach cone in the meridian plane. The C− characteristic
makes an angle θ − μ and the C+ characteristic makes an angle θ + μ with respect
to the x-axis in the meridian plane.
254 5 Method of Characteristics
By setting the numerator, N , in Eq. (5.94), equal to zero, we get the compatibility
conditions along C− and C+ characteristics in 3-D axisymmetric flows.
u2 v dr
dv 1− a2 dr
=
−
r du 2 (5.97)
du 1− v 2 dx characteristic 1 − av 2
a2
By substituting for the characteristics slope, dr/dx, in Eq. (5.97), and replacing veloc-
ity components u and v by the velocity magnitude and flow angle according to:
Since the first term in Eq. (5.99) is the differential of the Prandtl-Meyer angle, dν, we
can combine it with dθ on the LHS to arrive at the simplified compatibility equation
along C− and C+ characteristics in axisymmetric irrotational flows as:
1 dr
d(θ + ν) = √ Along C− characteristic (5.100)
M 2 − 1 − cot θ r
1 dr
d(θ − ν) = − √ Along C+ characteristic (5.101)
M 2 − 1 + cot θ r
Note that the compatibility conditions in 3-D axisymmetric flows is no longer the
algebraic equations θ ± ν = constant that we derived in 2-D irrotational flows.
The compatibility equations are now differential equations that have to be solved
numerically. By taking small steps in r in an x = constant plane, we replace the
differential of r by Δr and the corresponding differential of θ ± ν, by Δ(θ ± ν)
and create finite differences connecting two points along the characteristics network.
Here we show an example of axisymmetric irrotational flow that is solved by the
MOC.
Example 5.6 Apply the MOC to a cone with 30◦ half-vertex angle that is matched
with a circular arc ogive (from [3]) at point 1 (i.e., x = 6), as shown in Fig. 5.25.
The freestream Mach number is 3.07.
Solution:
This graphical-numerical solution is worked out by Ferri [3]. The conical shock angle
at the nose is established by the Taylor-Maccoll equation [9]. It makes an angle of
5.5 Axisymmetric Irrotational Flows 255
y
45 46
47
37 48
36 38
30 39
8 17 29 31
40
32
16a 22 24 25
7 16 18 19 26 33
12 20 27
6 11 13 14 21
7 15
5 4 8 9
2 10
4 6
3 5
1
3
2
1 = 30o
s
s
0 x
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Fig. 5.25 MOC applied to a sharp-nosed axisymmetric body of revolution in Mach 3.07 (Adapted
from [3])
39.42◦ with respect to the x-axis. The C+ characteristics at the nose form a simple
wave region and are Mach cones with different vertex angles η. Corresponding to
each η is a flow angle, θ that is established by Taylor-Maccoll equation. The angular
space between the cone and the conical shock is divided into 5 incremental steps. The
start of the calculation is at point 1 (on the cone-ogive) where the C+ characteristic
from the cone intersects the 6 characteristics (compression Mach waves) emanating
from the nose at points 2, 4, 7, 11 and 16.
The table of values produced by Ferri is shown in its original form in Table 5.8.
The fractional notation for angle is expressed in minutes (where there are 60 min in
1◦ ). Also note that the y-coordinate is the same as r in our notation and W is the
non-dimensional velocity, which is the ratio of local velocity V , to the maximum
velocity, Vmax (that we had defined in Chap. 3). Krasnov [4] is recommended for
further reading on aerodynamics of bodies of revolution.
M=2.91
M=2.37
592
382
(b)
600 598 596 594 590
382 382 382 382 382 588
596 382 586
398 594 592 584 382 580
598 394 392 384 584 380
398 596 592 590 588 586 382 578 576
396 394 390 388 386 582 378
600 382 376
400 576
598 576 378
400 576 380
578 382
582 580 382
382
582 384
576 576 584 582 386 576 576
594 594 400 398 390 582 388 576 386 384
400 398 390 388
594 576 576 576
396 396 576 392 390
394
(a)
Fig. 5.26 Wave-field approach is used in the design of a Mach 2.37 and Mach 2.91 isentropic
nozzles (from [3])
Design of supersonic nozzles and wind tunnels are the best candidates for MOC.
With the assumption of uniform flow in the throat section, we may apply the wave-
field MOC to design a family of supersonic nozzles. Figure 5.26 shows such a family
for two exit Mach numbers, 2.37 and 2.91. Note that the waves in Fig. 5.26 are all
of equal strength with Δθ = 2◦ . By choosing 9 steps of 2◦ each, we arrive at the
Mach 2.37 supersonic nozzle design (see Fig. 5.26). By taking an additional 3 steps
of 2◦ each at the throat (i.e., the wave generation section), the nozzle becomes a
Mach-2.91 design. Taking these additional 3 steps in designing the Mach 2.91 nozzle
design did not change the first 9 steps in the wave field method (see Fig. 5.26).
582 582
(a) 590 382 390
390 582 582
382 390
Horizontal Axis
Streamline
582
390
590
390 582
390
598
(b) 598 390
390
Streamline
598 Horizontal Axis
390
598
598 390
390
Fig. 5.27 Wave-field method applied to a nozzle with its exit cross section inclined with respect to
the jet axis (from [3]) a Expansion b Compression
from the horizontal axis as the exhaust plume evolves. We note that the asymmetry
in the exit plane of supersonic nozzle, as shown in Fig. 5.27, presents a case for thrust
vectoring, albeit of fixed and weak form.
C+
r = const. + w
C+
C+
M>1 9
C+
4 8
C+ =0 x
+ M=const. 3 7
2 6 C-
1 5
Virtual Source C-
C-
C-
- w
Initial Data Line r1 C-
r2
r3
r4
Fig. 5.28 Supersonic flow in a straight 2-D diverging duct with a few representative C± charac-
teristics
258 5 Method of Characteristics
MOC may also be applied to supersonic flows with non-uniform inlet flow condition.
The starting point is flow description on an initial data line, which then creates a
two-family network of characteristics. As an example, consider a 2-D supersonic
flow with radial streamlines, which resembles a source flow. Figure 5.28 shows the
definition sketch of a supersonic flow in a straight diverging duct with non-uniform
inlet condition with a few representative C± characteristics. The initial data line is
shown on a circular arc at r = r1 . The Mach number is constant on the initial data line
(of circular arc shape), but the flow angle changes from −θw to +θw corresponding to
the symmetrical diverging duct. The flow angle θ on the initial data line corresponds to
the polar coordinate θ shown in the definition sketch of Fig. 5.28. The unit processes
in this calculation involve interior and wall points that we have presented in Sect. 5.2.
γ−1 2
γ+1
h2 M1 1+ 2 M2
2(γ−1)
= γ−1 2
C+ characteristics or (left-running) Mach waves.
h1 M2 1+ 2 M1
(5.102)
For a given wall angle in region 2, θ2 , we calculate the flow Mach number in
region 2 by using the P-M expansion theory (presented in Chap. 4, shock-expansion
theory). The lengths of the C+ characteristics along the head and tail Mach waves
that are cut by the streamline are l1 and l2 respectively. Since these lengths are the
hypotenuse of a right triangle with respective heights h 1 and h 2 , they are related
according to:
l1 = h 1 / sin μ1 (5.103a)
l2 = h 2 / sin μ2 (5.103b)
5.6 Examples and Applications 259
γ−1 2
γ+1 γ−1 2
γ+1
l2 sin μ1 M1 1+ 2 M2
2(γ−1)
1+ 2 M2
2(γ−1)
= γ−1 2
= γ−1 2
(5.104)
l1 sin μ2 M2 1+ 1+
2 M1 2 M1
We have established the orientation of line BF, i.e., the tail characteristic, as well as its
length, l2 . For intermediate C+ characteristics, e.g., BE, we choose an intermediate
Mach number between M1 and M2 , which immediately produces its correspond-
ing Mach and P-M angles. The corresponding flow turning angle is the difference
between the two P-M angles (i.e., the difference between the intermediate PM angle
and the upstream value). The length BE is related to the length BD according to
Eq. (5.104) where M2 is replaced by the Mach number that we chose for the interme-
diate C+ characteristic. In the Cartesian coordinates attached to point B in Fig. 5.29,
we establish the coordinates of the bending streamline, DEF by noting that
x F = 2 cos(μ2 − θ2 ) (5.105c)
Example 5.7 A parallel uniform flow at Mach 1.6 approaches a sharp corner that
turns the flow in the clockwise direction by 21.89◦ , similar to the schematic drawing
of Fig. 5.29. Calculate the coordinates of the bending streamline with h 1 = 1 and
graph the streamline. Assume that the flow is planar 2-D with γ = 1.4.
Streamline M1 D
E
µ1 µ2
F
1 y l1 M2
h1 l2 Streamline
1 =0
(µ2 – 2)
2
x h2
2
A B
M2 = 1.8
M1 = 1.6
M3 = 2.0
M=1.6 C+1 C+2 M4 = 2.2
C+3
θ =0 C+4
y M5 = 2.4
C+5
x
M=2.4
o
θ =21.89
Fig. 5.30 Definition sketch for the five C+ characteristics in the expansion corner
Solution:
The first C+ characteristic is the head Mach wave that makes Mach angle with respect
to local upstream flow:
−1 1 −1 1
μ1 = sin = sin = 38.68◦
M1 1.6
The P-M angle corresponding to the incoming flow is ν1 = 14.86◦ . The P-M angle
after 21.89◦ of wall turning is the sum of the upstream P-M angle and the turning
angle, i.e., ν = 36.75◦ . The corresponding Mach number is M = 2.4. We divide
the centered expansion fan into finite steps in Mach number, e.g., steps of 0.2. This
choice produces 5 C+ characteristics as shown in Fig. 5.30.
Since the 5 C+ characteristics are chosen based on Mach number, we immediately
associate a Mach angle and P-M angle with each characteristic. The flow angle, θ,
starts from zero (based on our choice of the coordinate system) and the steps in θ
are the difference between the local P-M angle and the incoming P-M angle. The
length ratio along different characteristics associated with an intersecting streamline
is calculated from Eq. (5.104). The coordinates for the bending streamline follow
Eq. (5.105). These are summarized in table form from Excel calculations (Tables 5.9,
5.10 and 5.11).
The coordinates of the wall are summarized in table form:
The coordinates of the streamline are also summarized in table form:
The graph of the expansion corner and the corresponding streamline is shown in
Fig. 5.31.
We may now select a second streamline with a different h 1 value than the
Example 5.7, say h 1 = 2.0 The second streamline bends similar to the first and
the two streamlines, as an example, form the boundaries of a bending duct that
accelerate a Mach 1.6 flow to the exit Mach number of 2.4.
5.6 Examples and Applications 261
y
Streamline 1.5
C+1 C+2 Streamline
1 C+3
C+4
Wall 0.5
C+5
0 x
-2 -1 -0.5 0 1 2 3 4 5
-1
Wall
-1.5
-2
Fig. 5.31 Expansion corner of Example 5.7 and the graph of a representative streamline (γ = 1.4)
y
2.5
2
1.5
1
0.5
0 x
-2 -0.5 0 2 4 6 8 10
-1
-1.5
-2
Fig. 5.32 The expansion corner of Example 5.7 and the graph of the second streamline (γ = 1.4)
Example 5.8 Calculate and graph a second streamline for the expansion corner of
Example 5.7 with h 1 = 2.0.
Solution:
We follow the same procedure that we outlined in Example 5.7 and produce the
following tables in a spreadsheet calculation. Note that the flow characteristics, M,
ν, μ and θ remain unchanged on the second streamline (Tables 5.12 and 5.13).
Figure 5.32 shows the second streamline, with h 1 = 2. We may use the two
streamlines as the boundaries of a curved duct and is shown in Fig. 5.33.
The curved duct of Example 5.8 is a supersonic nozzle with an exit flow angle
of 21.89◦ with respect to the entrance flow direction. By incorporating a second
expansion fan of opposite family, it is possible to expand the flow and accomplish
Streamline 2 y
2.5
2 Curved Duct
Streamline 1 1.5
1
0.5
0
-2 -0.5 2 4 6 8 10
-1
-1.5
-2
Fig. 5.33 The shape of a curved duct that accelerates a uniform, parallel Mach 1.6 flow to a uniform
and parallel Mach 2.4 flow (γ = 1.4)
an axial exit flow. Figure 5.34 shows a schematic drawing of an S-duct supersonic
nozzle. The inlet flow is uniform and parallel, in field #1 and is in the x-direction.
Field #2 is a continuous (expansive) turning of the flow that is associated with a
centered expansion fan, which creates a non-uniform and non-parallel flowfield. The
flow in field #3 is again uniform and parallel, i.e., non-accelerating or decelerating.
Field #4 is a continuous expansion process associated with a centered expansion fan
of opposite family. Therefore, the turning of the flow is in the counter-clockwise
direction. Field #5 is the uniform parallel exit flow, which is in the axial direction.
A slender pointed body in supersonic flow creates an oblique shock at its leading edge,
as shown in Fig. 5.35. The convex curvature of the body on the upper surface creates
expansive Mach waves, or C+ characteristics, that interact with the oblique shock
and weaken it. We may use MOC to study the Mach wave—shock wave interaction.
According to Crocco’s theorem, the flow downstream of a curved shock is rotational
due to entropy gradient (see Eq. (2.189)). Therefore, to be concise, we have to use the
MOC suitable for rotational flows. However, in case of thin bodies at small angle of
attack, the shock curvature becomes slight and thus entropy gradient downstream of
the shock becomes negligibly small. Under these conditions, the MOC for potential
flows may be applied. Another acceptable simplification that is often used neglects
the reflected Mach waves, or C− characteristics, from the shock in establishing the
shape of the curved shock.
We may use the lattice-point method (or the wave-field approach) in calculating
the shock wave angle, βC upon interaction with the C+ characteristic that emanates
from point B on the body (see Fig. 5.35b). Since we neglected the reflected C−
characteristic from the shock at point C, the problem only involves the left-running
C+ characteristics and thus it becomes a simple-wave problem. Since the flow turning
is imposed at the wave, all streamlines turn the same amount. Therefore, the flow that
emerges from point C on the shock has to have the same angle as the wall at point
B. The reason for the nearly equal sign used in θc ≈ θ B in Fig. 5.35b is due to our
approximation of no reflected C− characteristic at point C on the shock, otherwise the
flow angles are identical in simple-wave problems. Now, with flow angle downstream
of the shock at point C known, we may use an oblique shock chart to establish the
shock wave angle βC (using the weak solution). At any point on the body, say B, we
know the body angle, θ B and using shock-expansion theory (Chap. 4), we calculate
the local Mach number, M B . There is a Mach angle μ B that corresponds to M B and
the sum of the two angles; θ B + μ B is the angle of the local C+ characteristic that
emanates from point B (see Fig. 5.35b). Subsequently, we move to other points on
the body, e.g., D, and repeat the process.
(a) (b)
G MC
C
E C
c B
C- Reflected C-
C C+ characteristic
1 A (neglected)
B
M1 > 1 H B
F
D B
A B A
Fig. 5.35 The waves on a slender body in supersonic flow a Interaction of Mach waves and an
oblique shock (simplified view), b expanded view of the flow and shock angles
Example 5.9 A 2-D pointed body at zero angle of attack is in a Mach 2 flow, as
shown. Its leading edge angle on its upper surface is θ A = 10◦ . Point B on the body
has an angle of θ B = 8◦ . Calculate:
(a) The oblique shock angle, β A in degrees
(b) The oblique shock angle, βC in degrees
(c) Mach number at the nose (downstream of the O.S.)
(d) Mach number at point B on the body
(e) The angle of C+ characteristic at point B
C o
c B =8
C
M1 = 2.0
C+
A
B o
B B =8
o
A A = 10 x
Solution:
With M1 = 2 and the nose angle of 10◦ , we use the oblique shock charts (Fig. 4.13)
to get the wave angle, β A ≈ 39.5◦ . Since the flow angle downstream of the shock at
C is estimated to be 8◦ , i.e., the same as the body at B, we use the oblique Shock
charts to get βc ≈ 37.2◦ .
We march across the O.S. at the nose, by using the following relations:
M2n
M2 = M A = ≈ 1.63
sin(β A − θ A )
5.7 Summary
Problems
2 w= +15o
C+ C-
y
M1 =2.0 x
o
1 =+10
1
5.2 In a steady, 2-D, irrotational, isentropic supersonic nozzle flow, we have the
flow Mach number and direction at point 1, as shown. The nozzle wall is straight and
makes an angle, θw = −9.5◦ with respect to x-axis. Calculate the flow Mach number
on the nozzle wall that corresponds to the C− characteristic that passes through point
1 (i.e., point 2, M2 ). Also calculate the slope of the C− characteristic, dy/dx, between
points 1 and 2.
M1 =2.2 o
1 1 =- 4
C- y
x
2
5.3 Based on the initial data line in a diverging section of a 2-D supersonic nozzle,
as shown, M1 = M3 = 1.6, θ1 = −θ3 = −15◦ , M2 = 1.8, θ2 = 0◦ . Use Method
of Characteristics (MOC) for 2-D irrotational flow to calculate:
(a) M5 and θ5
(b) M7 and θ7
8
y 3
x2 5
7
1 4
6
M ∞ = 3.2
β2
M2
2 θ 2 = 12o
M∞ = 3.2
C+ characteristic
y
x
1 M1 = 2.2
θ 1 = 9o
w, max
C+ C+
C+ C+
C- Me=2.8
y C-
x
A
Me=1.2
2
x
1 4
Centerline
Calculate:
(a) K −1 in degrees
(b) K +3 in degrees
268 5 Method of Characteristics
y 12 13
x 10
8
3
2 5 7
4
9
1
6 11
Centerline
Calculate:
(a) Flow angle θ and Mach number at node 5, i.e., θ5 and M5
(b) Wall angle connecting nodes 3 and 8 on the wall, θw |3−8
(c) Mach number, M11
5.9 A 2-D minimum-length nozzle is designed using MOC. There are 5 C− char-
acteristics at the throat and each turn the flow equally by 2◦ (Fig. 5.37). Calculate:
(a) The exit Mach number, Me
(b) The slope dy/dx of the first C− characteristic, i.e., AB
(c) The flow angle in region 2, i.e., θ2
(d) The flow angle in region 3, i.e., θ3
5.7 Summary 269
22
18
13
th 7
M e=1.8
A w,max
6
12 e
y Centerline
1 8 14 19 23 26 x
3
e
A
2
Me
1
y
B x
e y Centerline
x
Me = 2.06
pe=1.621 p
a
pa
5.10 Calculate the initial exhaust plume angle for an under-expanded jet with Me =
2.6, pe / pa = 3.0 with the ratio of specific heats, γ = 1.4. Assuming the flow in
the nozzle was isentropic; calculate the nozzle pressure ratio, NPR that is defined as
pt0 / pa where pt0 is the total pressure at the nozzle entrance.
5.13 The initial data line in a 2-D straight diverging duct is a circular arc with
radius of curvature R, where R = 1 m. The flow Mach number is M = 2.0 =
constant in magnitude on the initial data line. The straight, radial diverging duct has
its two sidewalls symmetrical with respect to the axis of the duct and their angles are
θw = ±15◦ , as shown. There are five points on the initial data line. Point number 1
is on the lower wall. Point number 3 is on the centerline. Point number 5 is on the
upper wall. Points 2 and 4 are half-way between 1 and 3 and 3 and 5 respectively.
Fill out the table below for the coordinates and flow parameters at the 5 initial data
points.
R =1m
y o
w= +15
5
x 3
Initial Data Line
1 o
w= -15
5.14 A parallel uniform flow at Mach 1.4 approaches a sharp corner that turns the
flow in the clockwise direction. The flow downstream of the corner reaches Mach
number of 2.8. Calculate the wall turning angle and use the wave-field method to
5.7 Summary 271
calculate the coordinates of the bending streamline with h 1 = 1 (see Fig. 5.29 for
definition sketch) and graph the streamline. Assume that the flow is planar 2-D with
γ = 1.4.
M1=2.6 C+ C+
o
1 =0
=1.4
y 3
1 2
x C
B
A
5.17 A parallel uniform flow at Mach 2.0 approaches a sharp corner that turns
the flow in the clockwise direction by 10.37◦ . First calculate the flow Mach number
downstream of the expansion corner. Second, calculate the coordinates of the bending
streamline with h 1 = 3.0 and graph the streamline. Assume that the flow is planar
2-D with γ = 1.4.
5.18 To accelerate a Mach 2.0 stream to Mach 2.4 isentropically, an expansion fan
that turns the flow by 10.37◦ will be formed, as shown schematically by lines AC and
272 5 Method of Characteristics
AE. This is the same problem as 5.16 with a single streamline that passes through
h 1 = 3.0. Now, construct a second streamline with h 1 = 4.0, which along with
the first streamline may form the walls of a supersonic nozzle. Calculate and graph
a second streamline with h 1 = 4.0. What is the area ratio, A2 /A1 , of this nozzle?
Assume flow is planar 2-D and γ = 1.4.
h
4.0 M =2.0 1 C
1 E
3.0 B 2
D
M 2=2.4
A
x
5.19 A 2-D S-shaped supersonic nozzle is designed based on two centered expansion
fans of opposite family, ABC and DEF, following the schematic drawing shown. The
flow at the entrance to the S-shaped nozzle is parallel and uniform in the x-direction
with M1 = 2.0. The nozzle is divided into three fields. The field #2 represents the
flow that has emerged from the first bend. It has attained a flow angle of θ2 = −20◦ .
The field #3 represents the flow that has emerged from the second bend and is purely
axial, i.e., θ3 = 0◦ . Calculate:
(a) The Mach number in field #2, M2
(b) The Mach number in field #3, M3
(c) The nozzle exit-to-entrance area ratio, A3 /A1
D
B
C
y
x 1
2
3
A
E
F
5.20 We apply wave-field theory to the half diamond airfoil in supersonic flow, as
shown. Assuming that the flow downstream of the shock is irrotational, we may use
the irrotational MOC at the shoulder with two C+ characteristics, AB and AC. Mach
number in region 2 is M2 = 2.0 and the flow angle is θ2 = +4◦ . The flow angles in
regions 3 and 4 are θ3 = 0◦ and θ4 = −4◦ respectively. Calculate:
(a) Mach numbers in regions 3 and 4
(b) The angles that C+ characteristics make with respect to the x-axis
(c) Flight Mach number, M1
5.7 Summary 273
C
B
3
y
C+ C+
1 2 4
A
x
5.21 Apply the lattice-point method to a curved (convex) body ABDF to estimate
the shock angles for the segments, AC, CE and EG. The body angles are known to be:
Body Angle (deg) A B D F
12 8 6 4
G
E
C
1
M1=2.4 H
= 1.4 F
D
B x
A
References
1. Anderson, J.: Modern Compressible Flow with Historic Perspective, 3rd edn. McGraw Hill,
New York (2003)
2. Anon: Equations, Tables, and Charts for Compressible Flow. NACA TM 1135, Moffett Field,
CA (1953)
3. Ferri, A.: Elements of Aerodynamics of Supersonic Flows. Macmillan Company, New York
(1949)
4. Krasnov, N.F.: Aerodynamics of Bodies of Revolution. Elsevier Publication, New York (1970)
5. Liepmann, H.W., Roshko, A.: Elements of Gas Dynamics. Wiley, New York (1957)
6. Prandtl, L., Busemann, A.: Näherungsverfahren zur Zeichnerischen Ermittlung von Ebenen
Strömungen mit Überscallgeschwindigkeit. Stodola Festschrift p. 499 (1929)
7. Shapiro, A.H.: The Dynamics and Thermodynamics of Compressible Fluid Flow. Ronald Press,
New York (1953)
8. Shapiro, A.H., Edelman, G.M.: Method of characteristics for two-dimensional supersonic flow
graphical and numerical procedure. J. Appl. Mech. 14(2), A-154 (1947)
9. Taylor, G.I., Maccoll, J.W.: The air pressure on a cone moving at high speed. Proc. R. Soc. Lond.
139, 298–311 (1933)
Chapter 6
Aerodynamics of Non-lifting Bodies
Abstract This chapter presents the sources of drag that are encountered by a
nonlifting body that is moving at transonic Mach numbers. In inviscid conditions
drag can be produced if the integrated pressure over a body has a component in the
opposite direction to the flow (pressure drag). First, the reader is introduced to the
relation between the geometry of a body and the pressure distribution over the body.
Subsequently, a method for the calculation of the pressure drag over axisymmetric
bodies is presented based on the linear potential flow equation. It is shown how the
pressure drag is a function of the cross-sectional area distribution of the body. The
concept of area ruling is presented and examples are shown of practical implementa-
tions of the area rule. If the flow is assumed viscous, other drag sources arise: friction
drag and drag due to boundary-layer separation. A qualitative characterization of lam-
inar and turbulent boundary layers is presented along with the concepts of transition
and separation. Also, the interaction between shock and boundary layer (both weak
and strong) is further detailed and it is shown how this influences drag divergence in
transonic conditions. In addition, calculation methods are presented to estimate the
boundary-layer properties of laminar and turbulent boundary layers along with their
transition region for bodies subjected to an external pressure and velocity distribution.
This chapter contains 8 examples and concludes with 29 practice problems.
6.1 Introduction
There exist two types of drag: friction drag and pressure drag. Friction drag is mostly
caused by the boundary layer between the body and the outer flow and is a function of
the viscosity of the fluid. A nonlifting body in transonic flow can experience pressure
drag through two mechanisms. One is due to a wake that is formed when the flow
streamlines separate from the body surface (drag due to flow stagnation and separated
flow). The second cause is through the formation of shock waves (wave drag). Wave
drag is an aerodynamic phenomenon which is unique to supersonic flow. This type
of drag is associated with the energy which is radiated away from the vehicle in
the form of pressure waves. In the present chapter we focus on these three causes
for drag. The aim of the chapter is to explain to the reader how a nonlifting body
produces drag and what parameters are of influence.
© Springer Science+Business Media Dordrecht 2015 275
R. Vos and S. Farokhi, Introduction to Transonic Aerodynamics,
Fluid Mechanics and Its Applications 110, DOI 10.1007/978-94-017-9747-4_6
276 6 Aerodynamics of Non-lifting Bodies
control surface, S u2
u1 C
p F
body
-F
body
A
D
2D control volume
control surface, σ
Fig. 6.1 Body force generated by pressure and shear stresses on the body and their relation to the
velocity deficit in the wake
In Fig. 6.1 a body is shown in a control volume. It is assumed that the upper and
lower boundary of the volume are formed by streamlines of the flow. The pressure
and shear forces that act on the body result in a total body force, F body . By Newton’s
third law, the body exerts an equal and opposite force on the fluid, −F body . Let us
investigate the effect of this body force on the momentum in the flow. We start from
Newton’s second law:
d
F = (mV ) (2.114)
dt
Remember that the right-hand side (RHS) of (2.114) physically represents the time
rate of change in momentum that is moving through the control volume. Its repre-
sentation has been given in the LHS of (2.121):
∂
F= ρV dV + (ρV · dS)V (6.1)
∂t
V S+σ
The integral over the second control surface, σ is always zero (regardless of control
volume shape) due to the fact that the inner product of velocity and surface area
vanishes on the body surface. If we neglect the small body forces on the fluid ( f = 0),
the external force F can be broken down into two parts: the reaction force from the
body, −F body , and the force exerted across the outer boundary:
F=− pdS − F body (6.2)
S
In (6.3), the force exerted on the body is coupled to properties of the flow inside
the control volume. We would like to determine the x-component of the body force
from the parallel velocity components at the beginning of the control volume, u 1 , and
at the end of the control volume, u 2 . We write the x-component of (6.3) as follows:
Fbodyx = − (ρV · dS)u − pdSx (6.4)
S S
First we notice that when the boundaries of the control volume are chosen at an
infinite distance from the body the pressure, p, should be constant on this boundary.
For a constant pressure, the integral yields:
pdSx = 0 (6.5)
S
Since the upper and lower surface of the control volume are streamlines of the flow,
the velocity vector and the control surface are locally aligned. This means that the
inner produce V · dS = 0 along those boundaries. Therefore, we can write for Fbodyx :
B C
Fbodyx = ρ1 u 21 dy − ρ2 u 22 dy (6.6)
A D
By employing the continuity equation in integral form (2.109) over the entrance
and exit of this control volume and subsequent substitution yields the following
expression:
C
Fbodyx = ρ2 u 2 (u 1 − u 2 )dy (6.7)
D
Loss in total
48 pressure due to
boundary layer
47
46
In this section we investigate the effect of the geometry and Mach number on the
supervelocities over a nonlifting body. In high subsonic conditions the flow over
curved surfaces can surpass the speed of sound creating expansion waves or shock
waves. The effect of shock waves is wave drag and possible separation of the boundary
layer. It is therefore instrumental to know how the geometric shape of a body affects
the increase in local Mach number.
This section partially follows the introduction to pressure fields and fluid acceleration
as presented by Shapiro [51]. In Chap. 2 we derived the Euler equations in Cartesian
coordinates. Along a streamline we can reduce the system of equations according to:
d p = −ρV dV (6.8)
We can derive a relation between the local pressure in the fluid and the local velocity.
We can integrate both sides of (6.8) between two arbitrary points along the streamline.
If we assume that the flow is incompressible (i.e. ρ1 = ρ2 = ρ) we obtain Bernoulli’s
equation (see Problem 6.3):
1 1
p1 + ρV12 = p2 + ρV22 (6.9)
2 2
For incompressible flow we can therefore conclude that:
1
ˆ p + ρV 2 = constant along a streamline
pt = (6.10)
2
The second term in (6.10) is the dynamic pressure, often denoted with q. Similar to the
static pressure, p, the dynamic pressure is a scalar. Bernoulli’s equation literally states
that the local velocity is inversely proportional to the square root of the local pressure.
The pressure coefficient may be written as
p − p∞
Cp =
ˆ (3.11)
q∞
Note that the subscript ∞ implies that we are dealing with a point in the flow at
an infinite distance from the measuring point. This is referred to as the freestream
condition. Substituting (6.10) and assuming incompressible conditions yields the
following expression for the pressure coefficient:
2
V
Cp = 1 − (6.11)
V∞
280 6 Aerodynamics of Non-lifting Bodies
This relation directly ties the pressure coefficient to the local velocity. We can
approximate this equation by assuming that the local velocity is a relatively small
deviation from the freestream velocity: V = V∞ + ΔV and ΔV /V∞ ≪ 1. In that
case, (3.11) reduces to:
ΔV
Cp = −2 (6.12)
V∞
We now have a very simple linear relationship between the supervelocity, ΔV , and
the pressure coefficient, Cp . Next, we have to correlate the pressure coefficient to the
local curvature of the streamlines.
Perpendicular to the streamline we can derive a relation between the pressure on
a fluid particle and its velocity along a curved streamline. A schematic drawing of
such a streamline is shown in Fig. 6.3. We define the direction perpendicular to the
streamline as n, and along the streamline with s. The free-body diagram (FBD) shows
the pressure forces that are acting on this fluid particle in the n-direction, in inviscid,
irrotational flow, while the kinetic diagram (KD) shows its centripetal acceleration.
If we assume the dimensions of the fluid particle to be Δs in length and Δr in width,
then Newton’s second law reduces to:
mV 2
= ( p + Δp − p)Δs (6.13)
r
We know that Δp = (d p/dr )Δr and that m = ρΔsΔr . Combining this with (6.13)
yields the following relation:
ρV 2 dp
= (6.14)
r dr
This relation shows that the pressure always increases outward from the center of
curvature.
We find curved streamlines over curved bodies in subsonic, transonic and super-
sonic flows. These bodies can be concave or convex (see Fig. 6.4). If the body is
convex the normal vector to the body wall, n is defined in the same direction as r
in Fig. 6.3 and the pressure gradient, d p/dn is therefore positive. In other words,
the pressure is increasing with the distance from the wall. Because we know that
p+Δp
n
m s m V V
m
p mV 2
r
r
FBD KD
Fig. 6.3 Forces and accelerations on a fluid particle perpendicular to a curved streamline
6.2 Pressure Distribution Over Nonlifting Bodies 281
the pressure at an infinite distance normal to the body should equal the freestream
pressure, p∞ , we can deduce that the pressure at the wall, p, should be lower than
p∞ . In other words, a convex shape leads to a reduction in pressure and consequently
a negative pressure coefficient, Cp . A similar argument can be made for the concave
shape. However, in this case the direction normal to the body wall, is in opposite
direction to the radius vector, r . Therefore, the pressure gradient in this direction is
negative and the pressure continuously decreases when we move further away from
the wall. Consequently, we can deduce that the pressure at the wall should therefore
be higher than the pressure infinity: p > p∞ . The pressure coefficient at the body
for a concave shape is therefore positive.
We have now qualitatively determined that streamlines along convex curves have
decrease Cp and streamlines along concave curves increase Cp . In addition, we can
see from (6.14) that the there is an inverse relation between the radius of curvature,
r , and the pressure gradient, d p/dn. From this relation we can deduce that when
r → 0, the pressure gradient must tend to infinity, i.e. d p/dn → ∞. We also see that
when limr →∞ d p/dn → 0 and that V → V∞ . The effect of the radius of curvature
of the body, r = R, can be evaluated on a qualitative basis. We can plot the pressure
coefficient Cp as a function of the distance from the center of curvature r , using the
aforementioned relations at the respective boundaries at r = 0 and r → ∞. This is
shown in Fig. 6.5. We can observe that the position of the wall has an effect on the
local pressure, p, at r = R. Whenever the wall is positioned close to the center of
dp/dn < 0
dp/dn > 0
p R p
8
convex wall
p<p Cp< 0 p>p Cp > 0
8
0 R
0 r
8
- convex shape
282 6 Aerodynamics of Non-lifting Bodies
curvature (i.e. it has a large curvature itself) the pressure difference between pr = R
and p∞ is large and we have a high Cp . Conversely, when R is large (i.e. the wall
has little curvature) the pressure difference is much smaller and Cp tends to zero.
Hence, on a purely qualitative basis we can state that the radius of curvature of the
body is negatively related to the pressure coefficient. In other words, the larger the
curvature, the higher the absolute value of the pressure coefficient.
We can combine the relation between curvature and Cp and between Cp and
ΔV in a straightforward manner. We know that when a geometry displays a convex
curvature the pressure coefficient, Cp < 0. Consequently, employing (6.12) we note
that the supervelocity, ΔV > 0. A convex curvature, therefore, leads to local speeds
that surpass the freestream velocity. Similarly, when the geometry has a concave
curvature, the pressure coefficient is positive and ΔV < 0. Concave shapes therefore
reduce the speed of the fluid near the body. Finally, we can state that the absolute
curvature (whether convex or concave) has a magnification effect on ΔV . A body
with sharp corners, for example, yields a large speed fluctuations at these corners,
while a body with little curvature only induces mild speed variations.
Now that we have established the effect of local curvature on the pressure coeffi-
cient, we take a look at the distribution of the pressure coefficient over a body. The
value of the pressure coefficient at various locations on the body is often termed the
pressure distribution. Strictly speaking this is incorrect. However, we have used the
term “pressure distribution” in relation to the Cp distribution and we will continue to
do so in this chapter. In subsonic flow, the pressure at any point A has an effect on the
pressure at an arbitrary point B (point B is in point A’s “region of influence”). Every
change in pressure coefficient due to the presence of curvature at point A therefore
also affects the pressure coefficient at any other arbitrary point B on that body. This
makes the intuitive prediction of the pressure coefficient somewhat more difficult.
However, we can still relate the geometric features of a given body to its pressure
distribution by relating the change in Cp to the local curvature. For example, if the
flow encounters a convex shape with a small radius of curvature we expect a sharp
increase in local suction. Conversely, if the flow encounters a concave shape with a
large radius of curvature, we expect a gradual increase in local pressure.
This is shown in Fig. 6.6 where two axis-symmetric bodies of identical thickness-
to-length ratio are placed in an axial flow field. The ellipsoid has a blunt leading
edge, which leads to a high local curvature. The Cp therefore increases rapidly.
Because the curvature over the ellipsoid is low over the center section of the body,
the pressure coefficient shows little variance. The absence of the blunt leading edge
on the paraboloid makes for a more gradual increase of the pressure coefficient.
However, due to the higher curvature in the center section, the pressure coefficient
reaches a higher peak.
In Fig. 6.7 we show the pressure distribution over an important part of the fuse-
lage: the cockpit. In this case, the aircraft in question has been modified with the
addition of a large nose fairing (radome) to house a radar. The radome has a signifi-
cant impact on the pressure distribution over cockpit. The pressure distribution over
the top of the fuselage shows the most erratic behavior. Initially the flow is acceler-
ated from near stagnation pressure (Cp = 1) at the nose towards a minimum of about
6.2 Pressure Distribution Over Nonlifting Bodies 283
-0.08
Paraboloid
-0.06
II
-0.04
I
CP -0.02
Ellipsoid
0.02
0.04
0 0.2 0.4 0.6 0.8 1.0
x
X= l
R
II
Fig. 6.6 Pressure distribution over an ellipsoid and paraboloid of identical maximum thickness
and length at incompressible axial flow (after Ref. [49])
-0.6
PRESENT METHOD EXPERIMENTAL DATA
TOP
-0.4 BOTTOM
-0.2
0
50 100 150 200 300 350 400
FUSELAGE STATION ~INS
CP
0.2
0.4
0.6
0.8
1.0
BASIS FUSELAGE
Fig. 6.7 Pressure distributions on a C-135 fuselage with a large radome at zero angle of attack
(after Ref. [6])
284 6 Aerodynamics of Non-lifting Bodies
Cp = −0.24. Then, the flow decelerates once more to a value close to stagnation at the
kink between the radome and the cockpit window. Due to the high convex curvature
over the window, the flow accelerates rapidly beyond the kink. At the bottom curve
we also see the presence of the concave kink between fuselage and radome in the
pressure distribution. Before the kink the velocity on the uncurved part of the radome
is close to its freestream value (Cp ≈ 0). At the kink the flow decelerates (Cp > 0)
after which it accelerates once more over the convex part of the lower fuselage.
Example 6.1 Below an axis-symmetric body is shown. Assuming inviscid, incom-
pressible flow, sketch the notional pressure distribution about this body.
Solution:
We first identify some of the key characteristics of the body under consideration.
Since we consider inviscid conditions we have a stagnation point at the leading edge
and at the trailing edge where the pressure coefficient is 1. This body has a blunt nose,
so we expect a sharp increase in suction over the first few percent of the body. Then the
curvature decreases, and so does the flow velocity. Hence the slope dCp /dx reduces.
Once the flow reaches the cylinder it does not see any change in curvature. Therefore,
the flow decelerates towards freestream velocity. The Cp will therefore increase to
a value close to 0. Even further downstream a convex shape is present again which
accelerates the flow and creates a second suction peak. Finally, the flow decelerates
to stagnation over the last few percent of the body. Due to the symmetric shape of
the body we also expect a symmetric shape in pressure distribution. Applying this
line of reasoning results in the notional pressure distribution drawn in Fig. 6.8.
1
0 1
x/l
6.2 Pressure Distribution Over Nonlifting Bodies 285
compressible pressure coefficient, Cp M is related to the incompressible pressure
∞
coefficient according to:
Cp M = 0
∞
Cp M = (6.15)
∞
1 − M∞ 2
In other words, the pressure coefficient increases with the freestream Mach number
up to infinity at M∞ = 1. We know from practice that in reality this theoretical value
is not reached. However, it does show qualitatively that the pressure coefficient and
thus the supervelocities over a given body in high-subsonic flow are affected by the
compressibility of the air.
Let us clarify this qualitative statement through the relationship between the pres-
sure coefficient and Mach number in steady isentropic flow. If we employ the relation
between Mach number and static temperature in isentropic conditions (2.147) and
the relation between pressure and temperature (2.103), we have:
γ γ
γ − 1 2 γ−1 γ − 1 2 γ−1
pt = p 1 + M = p∞ 1 + M∞ (6.16)
2 2
This relation shows that under the assumption of isentropic compression there is a
relation between the local Mach number, M, and the Mach number at infinity, M∞ ,
via the static pressure. We use the following definition of the pressure coefficient:
2 p
Cp = −1 (3.11)
γ M∞2 p∞
Combining (6.16) and (3.11) results in the following relation between pressure coef-
ficient, local Mach number, and freestream Mach number:
⎡ ⎤
γ−1 2
2 ⎣ 1 + M ∞
M2 = 2
− 1⎦ (6.17)
γ−1 γ−1
1 + 2 γ M∞ Cp
1 2 γ
For air flow about curved bodies this equation can be used up to a local Mach number
of 1.58. This is the (theoretical) maximum supersonic speed that can be achieved over
a body in subsonic flow. It will be derived in Sect. 7.5. A graphical overview relating
the local Mach number to the pressure coefficient is shown in Fig. 6.9. It can be seen
from this graph that high negative pressure coefficients (suction) in combination
with moderate to high subsonic Mach numbers lead to a local Mach number well
beyond 1.
If we define the increase in local Mach number as ΔM = M − M∞ then we can
also plot the relative increase in Mach number ΔM/M∞ . This is shown in Fig. 6.10.
If we take this graph and we compare the relative increase in Mach number for
Cp = −1 then we can see that for M∞ = 0.1 the increase is around 42 %, while
for M∞ = 0.8 the increase is around 62 %. In other words, there is a progressive
increase in local Mach number with increasing freestream Mach number.
286 6 Aerodynamics of Non-lifting Bodies
fre
es
0.5
tr e
1
am
0.4
M
ac
0.8
hn
0.3
um
be
0.6
r, M
0.2
8
0.4
0.1
0.2
0
0 -5 −10 -15
Pressure Coefficient, C p
0.5
0.4
0.3
0.2
0.1
0
0 −0.5 −1 −1.5 −2
Pressure Coefficient, C p
We have now established that the curvature in a body is directly responsible for
the supervelocities about that body and the related pressure coefficient. In addition,
we have shown that the local Mach number is strongly affected by the pressure
coefficient and value of the freestream Mach number. This should give the reader
some indication about the relation between the geometry of the body, the freestream
Mach number and the expected pressure coefficient. In addition to the effect close to
the body, the body has a larger effect on the streamlines further away from the body
when the Mach number approaches 1. As can be seen in Fig. 6.11, at a Mach number
of 0.8, region of influence is much larger that at a Mach number that is well-below 1.
This is a second effect of compressibility that is important for transonic aircraft.
6.2 Pressure Distribution Over Nonlifting Bodies 287
M ∞ << 1 M ∞ ≈ 0.8
As was established in Chap. 3, this effect plays an important role when wind tunnel
tests are carried out to predict the aerodynamic coefficients of the airplane.
The displacement effect caused by the body on the streamlines far away from
it continues to grow with Mach number. In supersonic conditions, the linearized
potential theory shows that the streamlines even at an infinite distance from the body
are affected. In fact, they follow exactly the same path as the wall itself. In supersonic
conditions, it is the slope of the surface with respect to the freestream velocity vector
that determines the local pressure coefficient [see (3.69)].
The aerodynamic shape of high-subsonic aircraft is to be designed such that the Mach-
dependent drag penalty in its cruise configuration is minimized. One instrument to
do this is by creating a favorable cross-sectional area distribution of the airplane. In
1951 it was Adolf Büsemann who first described that a streamtube of air at near-sonic
speeds no longer varied in diameter with speed but remained constant. Conversely,
when a streamtube of air would flow over a curved surface, the velocity inside the
tube would increase but its diameter would stay the same.
Let us investigate this claim by looking at the steady inviscid flow equations for a
one-dimensional streamtube with a variable cross section A. The continuity equation
(2.110) states that the mass flow entering a section of the streamtube equals the mass
flow that exits the streamtube. In other words:
ρu A = constant (6.18)
d(ρu A) = 0 (6.19)
288 6 Aerodynamics of Non-lifting Bodies
dρ du dA
+ + =0 (6.20)
ρ u A
d p = −ρudu (6.21)
Furthermore, we have already seen that the speed of sound can be written as:
dp
a= (2.106)
dρ
If we combine (6.19), (6.21), and (2.106), we can show that (see Problem 6.8):
dA 2
du
= M −1 (6.22)
A u
We observe that when M = 1, dA/A = 0 even though a finite du exists. This
shows that Büsemann’s claim is indeed true for flows near the speed of sound. It
also corresponds to our earlier observation that the region of influence in transonic
conditions is much larger than in subsonic conditions. Streamtubes around a body
(or wavy wall, as in Fig. 6.11) are simply displaced rather than reduced in area. At
M = 1 this effect extends, in theory, to infinity.
Richard Whitcomb realized that the individual streamtubes over a wing-body
combination would interfere with each other due to dramatic increases in cross-
sectional area where the wing joined the fuselage. Rather than looking at the shape
of the wing and fuselage separately he postulated that one needed to look at the cross-
sectional area of the entire airplane configuration and keep it as smooth a curve as
possible as it increased and decreased around the fuselage, wing, and empennage.
This design rule to minimize wave drag was termed the Whitcomb transonic area
rule [63]. The area rule states that the wave drag of an airplane configuration depends
primarily on the longitudinal distribution of the total cross-sectional area. This tells
us that a simple body of revolution that has the same distribution of cross sectional
area as a complex airplane produces the same wave drag. To evaluate the wave drag
of such a complex aircraft we can therefore reduce the complexity of the problem
by evaluating its equivalent axis-symmetric body. The equivalent body can be con-
structed by measuring the cross-sectional area of a fuselage station perpendicular to
the longitudinal axis of the airplane (see Fig. 6.12).
To demonstrate the applicability and limitations of the area rule, Fig. 6.13 shows a
comparison study carried out in the wind tunnel by Chan [8, 9]. Four different wind
tunnel models are subjected to a Mach number ranging from 0.8 to 1.1. We see a
1 See a text book on calculus such as Ref. [2] for details on logarithmic differentiation.
6.3 Wave Drag 289
swept-wing configuration with its equivalent body (1) and a delta-wing configuration
with its equivalent body (2). The equivalent bodies are axisymmetric and have the
same cross-sectional area distribution as the swept-wing configuration and delta-
wing configuration, respectively. The graph shows us the variation in zero-lift drag
coefficient of the four wind tunnel models over the Mach range. We see that when the
Mach number is increased beyond M = 0.9 all models show a significant increase
in drag. This drag is attributed to the formation of shock waves and is therefore
0.25
0.20
0
Zero-lift drag coefficient, CD
equivalent body 1
0.15
0.10
equivalent body 2
0.05
Fig. 6.13 Zero-lift drag coefficient of wing-body combinations and their equivalent axisymmetric
bodies (data from Refs. [8, 9])
290 6 Aerodynamics of Non-lifting Bodies
termed wave drag. This wave drag can be quantified by the momentum deficiency
over the shock wave as we have seen in Sect. 6.4.3. If we compare the curves for
the delta-wing configuration and its equivalent body, we see that their wave drag
in the transonic domain is very similar. This is in line with Whitcomb’s area rule.
However, when we look at the comparison between the swept-wing configuration
and its equivalent body, we notice that there is a significant difference between the
wave drag of the two bodies.
6.3 Wave Drag 291
Let us investigate why there is such a strong difference between the swept-wing
configuration and its equivalent body. It is expected that the formation of shock waves
is an important reason for discrepancy. Since pressure waves, such as shock waves or
Mach waves, induce density gradients in the flow, schlieren optics is an often used tool
to visualize these waves and the flow field. Schlieren (German, meaning “streaks”)
optics captures density gradients in compressible flows. The physical principle behind
such visualization techniques as schlieren or shadowgraph is based on light beam
deflection as it propagates through a medium with density gradients. For details
of these flow visualization techniques, their distinguishing features, limitations and
applications, the reader may refer to classical gas dynamic books such as Shapiro [50],
Liepmann and Roshko [37], Ferri [20] or books on wind tunnel testing principles.
Figure 6.14 shows schlieren images of the two wind tunnel models at M = 0.98. We
see that both models show shock waves in the same longitudinal position. However,
the shock wave over the wing of the swept-wing model is much stronger than the
corresponding shock wave over the equivalent model. The shock wave over the
swept-wing model causes boundary-layer separation at the outer portion of the wing
which causes additional drag with respect to the equivalent body. This explains the
larger drag over the swept-wing configuration compared to its equivalent body.
Based on the observations above, we conclude that the area rule works well for
delta-wing aircraft of low aspect ratio but works less well for swept-back wing-bodies
with moderate aspect ratio. The area rule is therefore limited to low-aspect-ratio
thin-wing body combinations. Viscous effects, such as shock-wave-boundary-layer
interaction, prevent a correct application of the area rule to wing-body combinations
with higher aspect ratio or thicker wings.
There is a large body of literature that supports the concept of the transonic area rule
in order to minimize the wave drag of a nonlifting body in transonic conditions, e.g.
Ref. [65]. We would like to know what cross-sectional area distribution would give
us the lowest wave drag. In the analysis that develops in this section we follow the
classic text of Ashley and Landahl for the computation of wave drag in inviscid flow
[4].
We start the derivation of wave drag in supersonic flow by looking at the axis-
symmetric body in uniform flow conditions. A two-dimensional representation of
such a body is presented in Fig. 6.1. The force acting on this body is governed by
(6.3). We introduce unit normal vector, n, such that we can bring both terms on the
RHS of (6.3) within a single integral:
F body = − [ρV (V · n) + pn]dS (6.23)
S
292 6 Aerodynamics of Non-lifting Bodies
In this equation ndS = dS. Under the same assumptions, the continuity equation in
integral form (2.110) can be written as follows:
ρ(V · n)dS = 0 (6.24)
S
We can employ the continuity (6.27) to show that (6.26) can be simplified. If we
assume that S is an enclosed surface we can introduce the freestream static pressure,
p∞ , and reduce the momentum equation to (see Problem 6.9):
Fbody = − ( p − p∞ ) n + ρV∞ 2
v (v · n + i · n)dS (6.28)
S
Equation (6.28) is a compact equation that relates the total body force to the properties
of the freestream flow ( p∞ , V∞ ) and the local perturbation velocities (v), density (ρ)
and pressure p under the assumption of an inviscid and steady state flow.
We are interested in calculating the wave drag of a three-dimensional body in
supersonic flow. An example of such a body in its cylindrical control volume is
shown in Fig. 6.15. The surface area, S of the control volume is divided into three
parts marked S1 –S3 . Let us first extract the drag force, D, from (6.28):
D = Fbodyx = − ( p − p∞ ) i + ρV∞ 2
u (v · n + i · n) dS (6.29)
S
S1 S3
We can simplify this expression by acknowledging that in supersonic flow the pres-
sure and velocity at S1 must remain unperturbed by the body (i.e. p = p∞ and
n φ̂x = 0). Similarly, we assume that at S3 is sufficiently far behind the body such
that the flow has become essentially two-dimensional in the y–z plane (“Trefftz
plane”). Therefore, φ̂x = 0 such that the last integral in (6.32) vanishes. In addition,
we employ Bernoulli (6.9) to relate the pressure to the velocity disturbance velocity
potential in cylindrical coordinates:
1 2 2
p − p∞ = − ρV∞ φ̂r + 2φ̂x + φ̂2x (6.33)
2
We can use this relation to write p − p∞ only in terms of φ̂r since φ̂x = 0 at S3 .
Finally, we note that φ̂r2 = φ̂2y + φ̂2z . Combining all of the information above we can
now rewrite (6.32) according to:
ρV∞2
D = −ρV∞
2
φ̂x φ̂r dS2 + (φ̂2y + φ̂2z )dS3 (6.34)
2
S2 S3
294 6 Aerodynamics of Non-lifting Bodies
This equation expresses the total pressure drag experienced by the body. In supersonic
flow, the pressure drag stems from two sources: the wave drag and the vortex drag.
The second integral in (6.34) gives the vortex drag, which is identical to the induced
drag for subsonic flow. The first integral represents the wave drag:
Dwave = −ρV∞
2
φ̂x φ̂r dS2 (6.35)
S2
This is a remarkably simple expression for the wave drag in terms of the derivatives
of the perturbation velocity potential. We must note that the present result is only
valid for bodies that are pointed in the rear.
It can be shown that flow around a body can be constructed by adding elementary
source solutions. Examples of this technique can be found in text books on the
fundamentals of aerodynamics such as Ref. [3]. In spherical coordinates a point
source should satisfy Laplace’s equation for incompressible conditions: ∇ 2 Φ = 0. In
spherical coordinates it can be shown that a source of strength Q located at r = 0 has
the potential function: Φ = −Q/4πr [see also (3.88)]. Since the disturbance velocity
potential can be expressed as a function of the velocity potential [Φ = V∞ (x + φ̂)]
we write for the disturbance velocity potential in spherical coordinates:
Q
φ̂ = − (6.36)
4πr V∞
where r is the distance measured at any point to the origin of the source.
Since we know that the superposition of elementary sources can represent the
shape of any axis-symmetric body we would like to know what the wave drag would
be of a line of sources. Therefore, we introduce a so-called lineal source distribution:
a straight line of sources of strength Q = V∞ f (x). Here, f (x) represents the vari-
able source strength along the line. Without derivation we show that the associated
perturbation velocity potential in supersonic flow is [4]:
1
x−βr
f (x1 )dx1
φ̂ = − (6.37)
2π
0 (x − x1 )2 − β 2 r 2
√
where β = M 2 − 1 and r is the distance from any point in space to the origin of
the y–z plane. We assume that f is continuous and that f (0) = f (l) = 0. We can
substitute (6.37) in (6.35) to find a closed-form expression of the wave drag. The
derivation involves complex algebraic operations that are beyond the scope of this
6.3 Wave Drag 295
book. Reference [4] can be consulted for details on the exact derivation. We merely
present the reader with the result:
2 l l
ρ∞ V∞
Dwave = − f (x1 ) f (x2 ) ln |x2 − x1 | dx1 dx2 (6.38)
4π
0 0
2 l
ρ∞ V∞ x
Dwave =− f (x)dx f (x1 ) ln(x − x1 )dx1
(6.39)
2π
0 0
If we compare (6.37) with (6.39) we notice that the drag coefficient has become
independent of β provided that the body ends in a point (as in Fig. 6.15) or ends in
a cylindrical portion. Finally, the tangentiality condition at the surface of the bodies
relates the source strength to the first derivative of the cross-sectional area, S:
Example 6.2 We consider a body of revolution with the following radius distribution:
1 1 2
R(x) = − x − for 0 ≤ x ≤ 1
4 2
Solution:
(a) We first calculate the cross-sectional area distribution that corresponds to this
radius distribution:
2 2
1 1
S(x) = π R 2 (x) = π − x−
4 2
(b) To calculate the wave drag coefficient with respect to the frontal area (π/16) we
rewrite (6.39) accordingly:
16 x
l
C Dwave = − f (x)dx f (x1 ) ln(x − x1 )dx1
π2
0 0
When examining (6.39) we notice that for x = x1 the integrand of the second
integral tends to negative infinity. This means that we cannot evaluate this inte-
gral using standard quadrature methods. We therefore replace the integral by a
double series:
16
I i−1
C Dwave ≈ − π2 f (xi )Δx f (x j ) ln(xi − x j )Δx
i=2 j=1
We evaluate this double summation in Matlab and find for Δx = 0.001 and
I = 1,000:
C Dwave ≈ 2.63
Note that this drag coefficient is calculated with respect to the frontal area of the
body.
For a slender body of revolution, we are interested in the cross-sectional area distrib-
ution that yields the least amount of wave drag. We start from the the expression for
6.3 Wave Drag 297
l2
π ∞
S(θ) = − An sin (nθ1 ) sin θ1 dθ1
2
θ n=1
∞
l 2 sin (2θ) sin (n + 1) θ sin (n − 1) θ
= A1 π − θ + + An −
4 2 n+1 n−1
n=2
(6.45)
If we integrate the area distribution between 0 and π we obtain an expression for the
volume of the body in terms of the first two Fourier coefficients:
l
π
πl 3 1
V = S(θ) sin θdθ = A1 − A2 (6.46)
2 8 2
0
Finally, we can obtain an expression for the wave drag in terms of the Fourier coef-
ficients by substituting (6.42) in (6.38):
2 l 2 0 0
ρ∞ V∞
∞ ∞
Dw = − An n cos(nθ1 ) Am m cos(mθ2 )
16π π π
n=1 m=1
298 6 Aerodynamics of Non-lifting Bodies
l
× ln (cos θ − cos θ1 ) sin θ1 sin θ2 dθ1 dθ2
2
∞
2 l2
πρ∞ V∞
= n A2n (6.47)
8
n=1
We consider a body which is pointed at both ends and has a given volume, V . This
implies that the cross-sectional area is zero at θ = 0 and θ = π. Via (6.45) and
S(0) = S(π) = 0 we obtain that A1 = 0. For a given volume and length we employ
(6.46) to obtain a value for A2 :
16V
A2 = − (6.48)
πl 3
If we scrutinize (6.47) we can immediately see that in order to minimize Dwave we
would have to ensure that as many Fourier coefficients as possible are equal to zero.
In practice this means that the drag is smallest when only A2 is non-zero while all
the others are zero. Substituting (6.48) into the equation for the wave drag, (6.47),
results in the equation for wave drag for this body:
64V 2
Dwave = ρ∞ V∞
2
(6.49)
πl 4
In order to find the ideal area distribution for this body one can substitute A1 = 0,
A2 = −16V /πl 3 into (6.45) and let all remaining An = 0. We then obtain the
following equation:
4V 1
S(θ) = sin θ − sin 3θ (6.50)
πl 3
This is is the cross-sectional area distribution that yields the lowest wave drag. The
axis-symmetric body that it represents is termed the Sears-Haack body after the
English scientist William Sears and the German scientist Wolfgang Haack, who
independently derived (6.50). In Fig. 6.17 one can see a Sears-Haack body for a
fineness ratio of approximately 6.8.
To minimize the zero-lift wave drag on any body in transonic flow, the cross-
sectional area distribution should be as close as possible to the Sears-Haack body.
6.3 Wave Drag 299
R(x), S(x)
0.04
0.02
0
0 0.2 0.4 0.6 0.8 1
x/l
The actual shape of the aircraft is fairly independent of its wave drag according to the
theory of Büsemann and Whitcomb. It is therefore important that large volumetric
additions to the body (e.g. wings, vertical and horizontal tail planes) are positioned
such that a cross-sectional area distribution results that matches as closely as possible
to the Sears-Haack body. Small deviations from the ideal Sears-Haack area distribu-
tion do not lead to excessive amounts of wave drag. However, discontinuities in area
distribution can cause a large increase in wave drag and should therefore be avoided.
The Sears-Haack body gives minimum drag for a body with a pointed nose and rear.
A second shape of interest is the optimum nose cone shape for a straight body of
revolution (e.g. a rocket, or bullet). If one applies different boundary conditions to
(6.45) one can find the area distribution for which the nose cone produces minimum
drag. For a nose cone with a given base area, S(l), at x = l we have θ = 0. Substituting
θ = 0 in (6.45) we obtain:
4S(l)
A1 = (6.51)
πl 2
Since A1 is nonzero, the wave drag is minimized when An = 0 for n ≥ 2. Substituting
(6.51) into (6.47) one obtains the expression for the minimum wave drag of the nose
cone: 2 S(l)2
2ρ∞ V∞
Dw = (6.52)
π l2
If we substitute the expression for A1 into (6.45) we obtain the area distribution for
the minimum-drag nose cone:
S(l) 1
S= π − θ + sin 2θ (6.53)
π 2
This resulting body is called the Von Kármán ogive and gives the minimum drag for
a given base area and length. Figure 6.18 shows an example of a Von Kármán ogive.
300 6 Aerodynamics of Non-lifting Bodies
R(x), S(x)
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
x/l
Application of the transonic area rule had the same objective as the supercritical airfoil
and the swept wing concept: reduce wave drag in the transonic regime. Fighter aircraft
at the time (e.g. Convair F-102) could not penetrate the sound barrier because of the
exponential drag increase that they experienced near Mach 1. Their jet engines could
not produce enough thrust to overcome the exponential increase in drag. The cross-
sectional area distribution of the original YF-102a prototype (Fig. 1.12a) had such a
high wave drag penalty near M = 1 that the available thrust could not overcome the
drag. Flight tests showed that this aircraft could not exceed M = 0.98. This is shown
in Fig. 6.19a where the thrust and the drag of the prototype aircraft are plotted at an
altitude of 35,000 ft. Because the US Air Force was not satisfied with the maximum
speed of this interceptor airplane a redesign was commissioned that had to result in a
top speed of M = 1.2. This redesign included a narrower canopy, revised intakes, and
a “pinched” fuselage (Fig. 1.12b). This dramatically altered the cross-sectional area
distribution of the airplane resulting in a smaller increase in drag coefficient around
M = 1 (see Fig. 6.19a) and a top speed of M = 1.22 as can be seen in Fig. 6.19b.
10,000
min
Minimum drag
.03
8,000 thrust available
revised
.02 improved
6,000
revised improved nose
nose 4,000
.01
2,000
0 0
.6 .7 .8 .9 1.0 1.1 1.2 .6 .7 .8 .9 1.0 1.1 1.2
Mach number, M (~) Mach number, M (~)
Fig. 6.19 Effect of cross-sectional area distribution on performance of the F-102 prototype at
35,000 ft (after Ref. [17]). a Drag coefficient and area distribution. b Drag an thrust versus Mach
number
6.3 Wave Drag 301
In the late 1960s when commercial transports started to emerge, it became evident
that aircraft needed to be designed with the area rule and Sears-Haack body in mind.
This resulted in some particular features of commercial transports that are still very
obvious this day. An example of how the cross-sectional area distribution was altered
on the Boeing 747 is shown in Fig. 6.20. It can be seen that by extending the upper
deck the area distribution became more smooth, which resulted in a higher drag-
divergence Mach number.
Another famous example of transonic area ruling is found in the anti-shock bodies
that were developed by Whitcomb and Küchemann. The anti-shock bodies were
positioned near the trailing edge of the wing to smoothen the cross-sectional area
distribution of the airplane. The effect of these bodies was a reduction in shock
strength over the upper surface of the wing, resulting in a lower wave drag. This was
demonstrated by experiments at NASA, which are shown in Fig. 6.21. Notice the
more aft position of the shock wave due to the application of the anti-shock bodies.
Even though these bodies increased the wetted area of the airplane (i.e. more friction
drag), their benefit in terms of wave drag reduction outweighed this penalty. The
socalled Küchemann carrots found their way onto several aircraft such as the Convair
990 (fastest high-subsonic transport airplane to date) and the Handley Page Victor.
Yet another aircraft that benefited from the research of Whitcomb and Küchemann
was the Soviet Union’s revolutionary Tu-95 bomber. The Tu-95 demonstrated the
ability to cruise at jet-like speeds of Mach 0.85 while using propeller-driven engines.
One of the reasons for this aircraft’s exceptional performance was the use of large
engine fairings aft of the inboard engines. These pods brought the external shape
of the Tu-95 closer to the Sears-Haack body to reduce drag and increase maximum
speed.
Basic cab
Extended cab
Extended 0.032
Cross-section aera
W/B fairing
Body 0.024 CL = 0.4
CL = 0.3
0.016
Body station 0.80 0.90 1.00
Mach number, M (~)
Fig. 6.20 Example of transonic area ruling on the Boeing 747 [24]. a Extended cab resulted in a
smoother area distribution. b Drag-divergence Mach number increased substantially
302 6 Aerodynamics of Non-lifting Bodies
Friction between a body and the surrounding fluid is one of the main causes of aircraft
drag. On typical high-subsonic transports the friction drag is responsible for 40–60 %
of the total drag of the airplane [57]. Every component (lifting or nonlifting) of the
aircraft that is wetted by the flow contributes to the friction drag of the airplane.
An obvious way to reduce friction drag is therefore to reduce the total wetted area.
6.4 Fundamentals of Boundary-Layer Flow 303
In steady potential flow the inertia forces of the fluid are balanced by pressure forces.
Viscosity does not play any role. If we consider the potential flow about a thin
plate attached to a cylinder, Fig. 6.22a, we see that near the body the streamlines
follow the contours of the body perfectly. Two stagnation points are present, one on
the leading edge of the plate and one on the opposite end of the cylinder. In high-
Reynolds-number flow, however, the effect of viscosity dominates the flow around
304 6 Aerodynamics of Non-lifting Bodies
(a) (b)
Fig. 6.22 Flow visualization of flow around a thin plate attached to a cylinder (after Ref. [1]).
a Potential flow streamlines. b Streamlines of water flowing past the object
the boundaries of the body Fig. 6.22b. We see that the streamlines do not follow the
curvature of the body anymore but separate from the surface near the top and bottom
of the cylinder.
If we consider the flow past a simple symmetric airfoil and we assume that the
Reynolds number is relatively large and the airfoil is aligned with the direction of the
freestream, then the effect of viscosity is confined to a very narrow region close to
the body. The streamlines about the the airfoil resemble very much the potential flow
streamlines (see Fig. 6.23a). Now, when we place this same airfoil at a significant
incidence angle with respect to the freestream flow direction, the viscous effects
become very pronounced and the flow field about the airfoil is completely changed
(Fig. 6.23b). The airfoil is said to be in a stalled condition. The boundary layer has
separated from the surface and a large wake is present between the streamlines and
the body. Separation is caused by a large adverse pressure gradient on the body
surface which causes separation of the boundary layer.
We first look at the development of a boundary layer in the absence of any pressure
gradient. In Fig. 6.24 the velocity profile of the flow about a flat plat is represented by
white areas. We clearly see that there is a no-slip condition at the wall: the fluid is at
(a) (b)
Fig. 6.23 Flow visualization of flow past an airfoil (after Ref. [1]). a At small angle of attack. b At
high angle of attack
6.4 Fundamentals of Boundary-Layer Flow 305
a standstill there. With increasing distance from the wall, y, the velocity, u, increases
progressively until a point where it matches the freestream value, U . We typically
define the thickness of the boundary layer, δ, as the y coordinate where the velocity
vector reaches 99 % of the freestream value. If we look progressively downstream
we see that the thickness of the boundary layer increases, while the velocity gradient,
du/dy, decreases. The velocity gradient at the wall is directly related to the shear
stress, τw , the fluid is exerting on the wall:
du
τw = μ (6.54)
dy
where μ is the dynamic viscosity of the fluid. This relation was originally pro-
posed by Isaac Newton and confirmed by experiments carried out by Poiseuille
in 1849 using liquid flow in tubes. The local friction coefficient is the wall-stress
non-dimensionalized by the freestream dynamic pressure according to:
τw
cf = (6.55)
2 ρ∞ V∞
1 2
Based on our observations of the velocity profiles in Fig. 6.24 and Eqs. (6.54) and
(6.55), we can deduce that the local friction coefficient is highest near the leading
edge and that it reduces further downstream.
To understand why the boundary layer is growing in thickness downstream we
look at the time history of vorticity within the boundary layer. We know that for a
given control surface, the vorticity vector (∇ × V ) is related to the velocity around
the contour of the control surface:
(∇ × V ) · dS = V · dS (2.71)
S C
The vorticity enclosed within the contour of the control surface is called the circula-
tion, Γ :
Γ = V · dS (6.56)
C
Let us try to estimate the circulation in the boundary layer. In Fig. 6.25 we show a
contour drawn about the velocity profile in the boundary layer. Due to the no-slip
306 6 Aerodynamics of Non-lifting Bodies
condition, the velocity at the wall is zero. Furthermore, we assume that locally the flow
is aligned with the wall and that the perpendicular velocity component is therefore
zero. Evaluating the circulation per unit length about this contour therefore results in:
Γ = V · dS = 0 + 0 − U0 + 0 = −U0
C
We can conclude that at any station, the circulation per unit length equals −U0 . The
vorticity is constant at any streamwise station along the plate: −U0 times a unit
length. The total amount of vorticity within each contour remains the same. Because
the flow upstream of the plate has no vorticity, we can conclude that the vorticity is
introduced at the leading edge of the plate due to the no-slip boundary condition.
Even though the circulation per unit length is constant at any streamwise station
on the plate, the distribution of vorticity normal to the plate does change. Due to the
viscosity of the fluid the vorticity is spread transversely as it travels downstream. In a
laminar boundary layer, this goes according to the principle of molecular diffusion.
The local boundary layer thickness represents the distance the vorticity has diffused
away from the plate. This growing process is a function of time, t, and the diffu-
sivity of momentum. The latter quantity is more commonly known as the kinematic
viscosity:
μ
ν= (6.57)
ρ
√
The transverse diffusion length, δ, can be shown to correlate with νt (a derivation
of this quantity is beyond the scope of this text2 ). At any distance from the leading
edge, x, the time that has passed can be calculated according to t ≈ x/U0 . Therefore,
the relative boundary layer thickness can be expressed as:
δ μ 1
∝ =√ (6.58)
x ρU0 x Rex
This relationship is only valid at high Reynolds numbers where δ/x 1 and for
laminar boundary layers. We can note from (6.58) that with higher freestream velocity
the boundary layer becomes thinner at any station along the plate. This is simply
because the boundary layer has had less time to grow. Bodies exposed to high-
Reynolds number flows therefore have a thinner boundary layer with respect to their
length.
To characterize the boundary layer, we frequently use the terms displacement thick-
ness, momentum thickness, and shape factor. In the subsequent paragraphs we show
how these parameters relate to the velocity distribution in the boundary layer. These
properties are applicable to both laminar and turbulent boundary layers.
The displacement thickness, δ ∗ , is a measure of the missing mass flow due to the
presence of the boundary layer.
δ ρu
δ∗ = 1− dy (6.59)
0 ρe u e
If we assume that the flow in the boundary layer is incompressible its displacement
thickness can be reduced to:
δ u
δi∗ = 1− dy (6.60)
0 ue
The incompressibility assumption is often used in practice, since it has shown to give
good correlation to experimental results.
Similar to the displacement thickness we can also define a thickness that accounts
for the missing momentum in the boundary layer. This thickness is termed the
momentum thickness, θ, and can be found from:
δ ρu ρu
θ= 1− dy (6.61)
0 ρe u e ρu e
Again, if we assume that in the boundary layer the flow is incompressible we have:
δ u u
θi = 1− dy (6.62)
0 ue ue
δ∗
H= (6.63)
θ
The shape factor is a function of the shape of the velocity distribution within the
boundary layer, hence its name. Notional velocity profiles for both high and low
shape factor are shown in Fig. 6.26. Notice that for a small H the velocity gradient at
the wall is higher compared to a high H . The shape factor will prove its importance
in predicting, for example, transition and separation (Sect. 6.4).
308 6 Aerodynamics of Non-lifting Bodies
low H
u/u e
Let us expand on the effect of the Reynolds number on the friction coefficient. As
is evident from (6.58), the Reynolds number is responsible for the relative thickness
of the boundary layer. Using Blasius’ equation (which is presented in Sect. 6.5.2)
we can deduce a velocity profile inside the incompressible laminar boundary layer.
We can relate the velocity gradient at the wall to the kinematic viscosity and the
freestream velocity according to:
∂u U0
= 0.332 U0 (6.64)
∂ y y=0 νx
Using this result and (6.55) we have a very straightforward quantification of the
friction drag:
0.664μ U0 0.664
cf = U0 =√ (6.65)
ρU02 νx Rex
This shows us that the friction coefficient indeed decreases with increasing Reynolds
number. The average friction coefficient of a complete plate can be found by inte-
grating the local friction coefficient over the complete length of the plate:
1 l
Cf = cf dx (6.66)
l 0
Substituting (6.65) we find that:
1.328
Cf = √ (6.67)
Rel
This equation demonstrates once more that an increase in the Reynolds number
decreases the friction drag coefficient of a surface wetted by an airflow. If we relate
that to airplanes, we can conclude from this simple derivation that both an increase in
size as well as an increase in cruise velocity decreases the average friction coefficient
of the aircraft. However, we must note that the effect of surface roughness has not
yet been taken into account and that we solely speak of laminar boundary layers. It
should also be noted that absolute drag does increase with velocity and size.
6.4 Fundamentals of Boundary-Layer Flow 309
(a) (b)
40 40
Normal Distance Parameter, x Re∞ (~)
M∞ = 12
32 32
M=1
y
M∞
=
12
24 M∞ = 8 24
M∞ = 6
M∞
16 16
=
8
M ∞ M ∞= 4
M∞ = 4
=6
8 M∞ = 2 8
M∞ = 0 M∞ = 2
0 0
0 0.2 0.4 0.6 0.8 1.0 0 10 20 30
Mach number ratio, M/M∞ (~) Temperature ratio, T/T∞ (K/K)
Fig. 6.27 Mach number and temperature distribution across a laminar boundary layer on an insu-
lated flat plate for various Mach numbers. Calculations carried out by Van Driest [59] assuming
Pr = 0.75 and S/Tw = 0.505. a Mach number distribution. b Temperature distribution
For compressible flows the density is not constant and the assumption of a con-
stant viscosity and heat transfer coefficient through the boundary-layer thickness are
no longer valid. The friction between the flow and the wall generates a temperature
difference between the wall and the free-stream. The resulting temperature distrib-
ution in the boundary layer has an effect on the velocity distribution and therefore
on the velocity gradient at the wall. At the same time, the temperature has an effect
on the viscosity through Sutherland’s formula (2.81). Both effects contribute to a
change in friction coefficient.
For a laminar boundary layer and adiabatic wall conditions, Van Driest [59] cal-
culated the change in velocity profile with free-stream Mach number using Crocco’s
method.3 Assuming adiabatic wall conditions and a constant heat transfer coefficient,
the temperature and Mach number profiles within the boundary layer could be deter-
mined for various free-stream Mach numbers. The Mach number and temperature
distribution is shown in Fig. 6.27.
If we examine Fig. 6.27a, we can observe that an increase in Mach number has
several effects. First of all, it increases the boundary layer thickness. Secondly, the
Mach number gradient at the wall decreases with Mach number. It can be shown that
also the velocity profile follows this trend [59]. This results in a lower shear stress
at the wall, which causes a reduction in friction coefficient with Mach number. The
sonic line is also shown in Fig. 6.27a. With increasing Mach number, the sonic line
moves away from the wall. However, if we compare the location of the sonic line to
the thickness of the boundary layer, we conclude that the position of the sonic line
only changes slightly. For example, for M∞ = 2 the sonic line lies at approximately
0.40δ, where δ has been defined as the distance where u/V∞ = 0.995. For M∞ = 8
the sonic line is located at approximately 0.35δ.
The temperature distribution in Fig. 6.27b shows the effect of friction on the wall
temperature. Note that at M = 2 the wall temperature is 40 % higher than the free-
stream temperature. At M = 4 there is more than a factor of 3 between the wall
temperature and the free-stream temperature. We can therefore deduce that for adi-
abatic conditions, the temperature increase over at high supersonic Mach numbers
results in temperatures that are higher than conventional aluminum can sustain with-
out degrading material properties. Therefore, supersonic aircraft that fly faster than
M ≈ 2.5 have to rely on more heat resistant materials, such as titanium. In the
hypersonic realm (M∞ > 5) the wall temperatures become extremely high, which
requires aerospace vehicles that operate in this regime to have an ablative heat shield
or a thermal-soak heat shield. In the transonic domain, aerodynamic heating is com-
paratively small and has little effect on the velocity gradient and friction coefficient.
Therefore, the incompressible boundary-layer equations can often be successfully
used to predict the boundary layer properties in the transonic flow domain.
Using the method of Van Driest, the compressible friction coefficient can be
correlated to the incompressible friction coefficient. An approximation of the rela-
tion between compressible and incompressible friction coefficient was derived by
Johnson and Rubesin [27]:
Cf 1
= 0.12 (6.68)
C fi 1 + 0.1305M 2
Both the approximation of Johnson and Rubesin as well as the calculation of Van
Driest is shown in Fig. 6.28. We see a gradual decay of the friction coefficient with
increasing Mach number. We also note that in the transonic flow domain the decrease
in friction coefficient is relatively small. Using (6.68), a reduction of 1.5 % at M = 1
can be computed. The change in turbulent friction coefficient is also presented in
Fig. 6.28 and is discussed in more detail in Sect. 6.4.2.
Coles 1
Chapman and Kester 2
Monaghan and Cooke 1
Rubesi, Maydew and Varga 1
(~)
1
Brinich and Diaconis 1
M=0 Korkegi 1
Lobb, Winkler and Persh 1
Friction coefficient ratio, Cf /Cf
0.8
{ Pr = 0.75
Crocco S/T = 0.505
method T∞ =∞ 217.8 K
LAMINAR
0.6
(1+0.1305M 2 ) -0.12
(1+0.2M 2 )-0.467
0.4
Sommer & Short method TURBULENT
0.2
ic nic
son transo
sub supersonic hypersonic
0
0 1 2 3 4 5 6 7 8 9 10
Mach number, M (~)
Fig. 6.28 Variation of skin friction coefficient with Mach number. Data from Refs. [25, 27, 37, 39,
46, 53, 58, 59]. 1: Mean skin friction values. 2: Local skin friction values, for Reynolds number
Reθ = 8,000 based on momentum thickness
(a) (b)
δ =.66 δ
h h =.53
h
δ h δ
UPSTREAM DOWNSTREAM
Fig. 6.29 Flow visualization of velocity profiles in a converging channel (after Ref. [1]). a Velocity
profile in a converging channel. b Comparison between upstream and downstream velocity profiles
When looking at the velocity profile in the boundary layer we see that the boundary
layer thickness reduces considerably. Most of this reduction can be explained by
the two-to-one decrease in flow area. However, using the distance to the nearby
streamline, h, as a reference dimension, we can see that also the relative thickness
δ/ h has decreased. If we consider the vorticity within the layer up to h we can deduce
that the vorticity increases linearly with the increase in freestream velocity. This new
vorticity is added to the vorticity already present in the boundary layer at the wall.
It is as if a new boundary layer is being formed at every increment in the converging
duct. The combined profile at the exit of the contraction is relatively thinner because
the vorticity has had little time to diffuse. In the narrow part of the duct we now have
a boundary layer with a larger percentage of vorticity close to the wall. This results
in a higher velocity gradient, shear stress, and friction coefficient.
312 6 Aerodynamics of Non-lifting Bodies
In a divergent channel, the opposite effect occurs. When the diffuser angle is
relatively low, the flow velocity in the channel will decelerate and an unfavorable (or
adverse) pressure distribution is present: ∂ p/∂x > 0. The pressure downstream is
higher than the pressure upstream. The result is a thickening of the boundary layer
both in the absolute and in the relative sense. In other words, ∂ hδ /∂x > 0. The reduced
velocity slope at the wall reduces the shear stress and the friction coefficient. The
velocity profile becomes less full: less vorticity is concentrated near the wall and more
is concentrated near the edge of the boundary layer. This can be seen by comparing
the velocity profiles at the first two stations of Fig. 6.30b.
If the deceleration of the flow continues to take place the velocity gradient (∂u/∂ y)
near the wall becomes smaller and smaller. Eventually the shear stress goes to zero,
followed by a reversal of the flow direction near the wall. This is visualized in the
3rd station of Fig. 6.30b. We see that the flow at the second station is still directed
to the right. While the flow at the third station is directed to the left. Somewhere in
between those two stations the flow at the wall has come to a complete stand still.
The corresponding shear stress at this point is zero. We call this point the separation
point. Beyond this point, the fluid, which was in contact with the wall in the upstream
boundary layer, has separated from the wall by a region of reversed (or recirculating)
flow. This is visualized in Fig. 6.31.
We have now established that the development of the boundary is influenced by
the pressure gradient. Given the fact that any body in a flow field generates a pressure
distribution, we can deduce that the boundary layer development is influenced by
the shape of the pressure distribution. In addition, the separation point (if there is
any) is affected by the pressure distribution. The shape of the velocity profile at any
point in the boundary layer is also dependent on the conditions of the boundary layer
downstream. For the simple case of a flat plate we know that the boundary layer
transitions from laminar to turbulent and subsequently grows in thickness the further
we move downstream. A thicker boundary layer has a lower velocity gradient at the
wall and is therefore more prone to separate whenever an adverse pressure gradient
is present. If we recall the lessons we learned from Sect. 6.2 we can state that in order
to prevent separation we need to modify the outer shape of the body such that the we
obtain relatively small pressure (adverse) gradients in regions where the boundary
layer is relatively thick.
(a) (b)
Fig. 6.30 Flow visualization of velocity profiles in a diverging channels of different diffuser angle
(after Ref. [1]). a Diverging channel with small diffuser angle. b Diverging channel with larger
diffuser angle
6.4 Fundamentals of Boundary-Layer Flow 313
Point of
separation
∂u ∂u ∂u
>0 =0 <0
∂y ∂y ∂y
As was demonstrated by Osbourne Reynolds in the late 19th century, fluid flow in a
pipe transitions from laminar to turbulent beyond a certain critical Reynolds number
(Fig. 6.32).4 The same happens in boundary layers. The steps in the transition from
laminar to turbulent are complicated and interdependent. First there is the growth of
the nearly two-dimensional Tollmien-Schlichting waves. These waves are the first
(infinitesimal) indicators of laminar-flow instabilities. Whether these instabilities
grow depends on the amplification factor of the instability at a particular position in
the boundary layer.
Experiments of boundary layer flow over a flat plate show that the unstable
Tollmien-Schlichting waves grow downstream and become three-dimensional. At
regions of high localized shear stress, several types of unstable vortices occur. These
vortices are longitudinally stretched and begin a cascading breakdown into smaller
units. In this fluctuating state, intense local changes occur at random locations and
at random times in the shear layer near the wall. Turbulence occurs in these spots
which further coalesce downstream into fully turbulent flow. This three-dimensional
transition process is shown in Fig. 6.33 after Kegelman and Mueller [28]. Note that
the turbulent spots mark wedges of turbulent flow. This results in the typical zig-zag
pattern of the transition line.
Since it is generally known that laminar boundary layers have a lower friction drag
than turbulent boundary layers, it is of interest to know where transition originates.
As we will see more explicitly in Sect. 6.5.4, amplification or damping of Tollmien-
Schlichting waves largely depends on the local pressure gradient. A negative pressure
gradient reduces the amplification of disturbances, while a positive (adverse) pressure
gradient increases the amplification rates. An adverse pressure gradient therefore
hastens the onset of transition compared to boundary layer flow in the absence of a
pressure gradient.
If the critical Reynolds number is defined as the Reynolds number for which a
disturbance with a particular wave length causes an unstable growth of that partic-
ular disturbance, the minimum critical Reynolds number is the Reynolds number
4 In this context, the word critical is unrelated to the flow properties at sonic conditions.
314 6 Aerodynamics of Non-lifting Bodies
(c)
U Stable
laminar
flow
x Edge
Recrit contamination
0
δ (x)
U
for which any disturbance of arbitrary wave length causes the boundary layer to
become unstable. Based on a theoretical investigation of compressible boundary
layer stability over a flat plate, Lees [35] concludes that for subsonic conditions any
laminar boundary-layer flow is unstable at sufficiently high Reynolds numbers. The
minimum critical Reynolds number is largely determined by the product of mean
density and the mean vorticity across the boundary layer. The addition of heat to the
fluid through the solid surface reduces the minimum critical Reynolds number and
therefore hastens the onset of transition. Cooling the wall has the opposite effect.
For adiabatic conditions, Laufer and Vrebalovich [34] concluded that the Mach
number has no effect on the stability of the laminar boundary layer but is solely
dependent on the displacement-thickness Reynolds number, Reδ . In addition, the
amplification factor is also shown to be independent of the Mach number. However,
it is shown that the disturbances propagate faster with increasing Mach number. Con-
sequently, a disturbance propagating to a certain x-position will not be amplified as
much in a higher Mach number flow because its time available to grow is less, even
6.4 Fundamentals of Boundary-Layer Flow 315
though its amplification coefficient is the same. Therefore, the adiabatic compress-
ible boundary layer is shown to be more stable with respect to small disturbances
than the incompressible one.
Once the boundary layer has transitioned from laminar to turbulent, its velocity
profile changes significantly. Whereas the laminar boundary layer showed a smoothly
changing velocity profile, the large amount of vorticity in the turbulent boundary layer
creates a more rugged velocity profile. In Fig. 6.34 the instantaneous velocity of a
laminar and turbulent boundary layer are shown on opposite sides of a flat plate. The
ruggedness of the velocity profile in the turbulent boundary layer is evident. When we
superimpose a large number of these instantaneous velocity profiles over time, we can
distinguish a mean velocity profile for both the turbulent and the laminar boundary
layer. It can be seen from Fig. 6.34 that the deviation from the mean velocity profile
is much larger in the turbulent boundary layer than it is in the laminar boundary layer,
due to the randomness of the vortices that are created in the turbulent boundary layer.
In Fig. 6.35 the two velocity profiles are compared. When superimposing them
we see that the turbulent boundary layer shows a larger velocity gradient at the wall.
The turbulent boundary layer therefore generates more shear stress as per (6.54).The
incompressible friction coefficient at a particular station x on a flat plate can be
approximated by the “power law” [66]:
0.027
cf = 1/7
(6.69)
Rex
Note that this relation only holds for low subsonic conditions. For turbulent boundary
layers the incompressible friction coefficient therefore decays much slower with
−1/7 −1/2
Reynolds number: cf ∝ Rex while cf ∝ Rex for laminar boundary layers. In
Fig. 6.36 the local friction coefficient is plotted for both a turbulent and a laminar
boundary layer. In addition, a typical transition line is shown which shows the rapid
increase in local friction coefficient when the boundary layer transitions from laminar
to turbulent.
The circulation is the same in both the laminar and turbulent boundary layers.
However, the distribution of circulation is different. For the turbulent boundary layer
(a) (b)
Fig. 6.34 Boundary layer around a flat plat: flow is turbulent on upper surface and laminar
on lower surface (after Ref. [1]). a Instantaneous velocity profiles. b Superposition of many
instantaneous velocity profiles
316 6 Aerodynamics of Non-lifting Bodies
(a) (b)
LAMINAR TURBULENT
Fig. 6.35 Comparison between velocity profiles in laminar and turbulent boundary layers (after Ref.
[1]). a Comparison of mean laminar (solid) and turbulent (dashed) velocity profiles. b Comparison
of velocity gradients at the wall for laminar and turbulent boundary layers
0.003
0.002 Typical
transition
curve
Laminar flow:
0.001
0.664
(Blasius)
Re
0
105 106 107 108 109 1010
Local Reynold number, Rex (~)
more vorticity is concentrated near the plate even though some vorticity has also
spread farther from the plate. The thickness of the turbulent boundary layer on a flat
plate at point x can be approximated according to [66]:
δ 0.16
≈ 1/7
(6.70)
x Rex
Comparing this to the laminar boundary layer we see that the laminar boundary layer
grows according to δ ∝ x 1/2 , while the turbulent boundary layer grows much faster
according to δ ∝ x 6/7 .
The distribution of momentum is also different between the two boundary layers.
The unsteady random rotary motions associated with vorticity that is aligned with the
flow transports high-momentum fluid towards the plate while low-momentum fluid
is transported away from the plate. This extra momentum close to the wall enables
the turbulent boundary layer to withstand much larger adverse pressure gradients
6.4 Fundamentals of Boundary-Layer Flow 317
Fig. 6.37 Gloster Javelin FAW.9 with three rows of vortex generators on the wing to delay the
onset of boundary layer separation. Photo Adrian Pingstone
before separating from the wall. This is the reason why vortex generators are often
seen on airplane wings. They generate additional vorticity in the direction of the
flow that transports momentum towards the wing surface and hence delays the onset
of separation. Rows of vortex generators can be seen on various modern transonic
airplanes. A historic example is the Gloster Javelin having three rows of vortex
generators on its outer wing (Fig. 6.37).
We have already noted that the effect of the Reynolds number on the friction coef-
ficient is less strong in turbulent boundary layers than in laminar boundary layers.
There exist many empirically derived relations that correlate the turbulent friction
coefficient, Cf , of a hydraulically smooth plate to the freestream Reynolds number.
Schlichting [48] proposed the following interpolation formula:
0.472
Cf = (6.71)
(log10 Re)2.58
Note that the two coefficients in (6.71) are empirically determined. Von Kármán and
Schoenherr used the following equation to fit the empirical data:
0.242
= log10 Cf (6.72)
Cf Re
Equation (6.72) needs to be solved iteratively to find a solution for Cf . Note that the
Von Kármán-Schoenherr relation relies on a single empirical coefficient. In Fig. 6.38
the two relations are shown together with various experimental results. Some of the
experiments date back to Froude in the late 19th century. There exists quite some
scatter between the various experimental data. According to Ref. [40] the exper-
imental data from Smith and Walker (1959) is most reliable due to the accurate
measurement techniques. It can be seen from Fig. 6.38 that both theoretical lines fol-
318 6 Aerodynamics of Non-lifting Bodies
−3
x 10
4.0
Froude, 1872 -- 16, 28, and 50 foot varnished planks
Average Friction Coefficient, Cf (~)
Fig. 6.38 Variation of turbulent skin friction coefficient with Reynolds number. Data selected
from [40]
low approximately the same trend. It depends on the experimental data set to judge
which theoretical line gives a better prediction.
It is of practical interest to investigate the effect of the Mach number on the
friction coefficient of a turbulent boundary layer. We would like to know if the friction
coefficient in transonic domain is influenced by the Mach number. Figure 6.28 on
p. 309 shows experimental results for the compressible, turbulent friction coefficient
up to a Mach number of 10. In addition, two theoretical lines are plotted, which
are shortly discussed hereafter. It can be observed that the friction coefficient of the
turbulent boundary layer has a much stronger dependency on the Mach number than
the friction coefficient of a laminar boundary layer.
Rubesin et al. [46] base their prediction of the compressible, turbulent friction
coefficient on the interpolation curve of Schlichting (6.71):
0.472
Cf =
0.467 (6.73)
(log10 Re)2.58 1 + γ−1
2 M 2
∞
where the viscosity ratio is given by the Sutherland equation (2.81). The relationship
between Cf and Re is given by the Von Kármán-Schoenherr relation (6.72):
0.242
= log10 Cf (6.77)
Cf Re
Finally, the compressible skin friction coefficient is obtained from Cf according to:
T∞
Cf = Cf (6.78)
T
If we compare the two theoretical lines we can conclude that the method of Sommer
and Short is most consistent with experimental results up to M = 10. However, in the
transonic regime equation (6.73) gives an approximation that is sufficiently accurate
given the scatter in experimental results. For M = 1, the compressible friction
coefficient is about 8 % less than its incompressible counterpart. If we compare that
to the laminar boundary layer (−1.5 % at M = 1), we can conclude that the turbulent
boundary layer is more sensitive to changes in Mach number and that the change in
friction coefficient due to compressibility effects is significant in the transonic regime.
In a mixed subsonic and supersonic domain often normal shock waves form. On the
wing surface of many high-subsonic aircraft a weak normal shock (recompression
shock) terminating a supersonic region is present in the design condition. The
associated pressure distribution allows for a high cruise lift coefficient and a rela-
tively high lift-to-drag ratio. On non-lifting bodies the presence of a shock should be
avoided. It produces wave drag and it can also cause separation of the boundary layer
at the foot of the shock wave. Shock waves on nonlifting bodies often occur when two
320 6 Aerodynamics of Non-lifting Bodies
Flow
(c)
Flow
6.4 Fundamentals of Boundary-Layer Flow 321
(d)
Flow
Fig. 6.39a, b). This is referred to as weak interaction of the shock with the boundary
layer. For a slightly higher Mach number, namely 0.895, the shocks are strong enough
to cause boundary layer separation (see Fig. 6.39c). We call this strong interaction.
The turbulent eddies are visible at the interface of the separated shear layer and the
subsequent pressure waves that emanate from the wake. The effect of boundary layer
type, i.e., laminar versus turbulent, on shock-boundary layer interaction is qualita-
tively shown in Fig. 6.39c, d. Note that the flow Mach number is maintained between
the two cases (both are at 0.895) but the Reynolds number is doubled. The laminar
boundary layer shows earlier separation on the airfoil than the turbulent boundary
layer case, as expected. Consequently, the wake downstream of the airfoil in the
laminar boundary layer case is thicker than in the turbulent boundary layer case. We
also note the lambda shock formation at the trailing edge of the airfoil in Fig. 6.39d.
Consider Fig. 6.40, which gives a general sketch of the flow field within the turbulent
boundary layer and the interaction with a normal shock wave. In front of the shock
wave the outer flow is supersonic. This implies that the edge velocity at the boundary
layer edge is also larger than the speed of sound. Descending into the boundary layer
the velocity progressively decreases up to the point where u = a (i.e. M = 1). If we
connect the points in the streamwise direction where this condition holds we form
the sonic line. The location of the sonic line is determined by the boundary layer
velocity profile and the edge Mach number, Me . The velocity profile, in turn, is a
function of the pressure gradient under which the boundary layer developed before
encountering the interaction. Below the sonic line information can travel upstream.
The pressure rise that results from the presence of the shock is spread ahead of the
shock location resulting in the boundary layer thickening.
The flow field shown in Fig. 6.40 forms a complex equilibrium of flow parameters.
We see that at the interaction the boundary layer becomes thicker. This implies that
322 6 Aerodynamics of Non-lifting Bodies
M e <1
M e >1
M>1
large
M<1 eddy
M>1 region
viscous sublayer
Fig. 6.40 Detail of weak shock-wave boundary-layer interaction for a turbulent boundary layer
(after Ref. [41])
the streamlines move away from the wall. The resulting curvature in the streamlines
causes the formation of (isentropic) compression waves. These compression waves
start at the sonic line and allow the supersonic flow to change the flow direction
(remember that in supersonic conditions the direction of the flow can only be altered
through the formation of waves as we explained in Chap. 4). The fan of compression
waves merge into a shock wave outside of the boundary layer. Close to the boundary
layer edge the shock wave is still slightly oblique which reduces the flow velocity
behind the shock to low supersonic. Within the boundary layer this is also the case:
the flow behind the compressions waves can still be (slightly) supersonic. This gives
rise to a pocket of supersonic flow behind the shock wave that is termed the super-
sonic tongue. Within the supersonic tongue the flow decelerates isentropically to
subsonic speeds. The size of the supersonic tongue is case-dependent and is merely
notional in Fig. 6.40.
The profile of the velocity distribution in the boundary layer is altered in the short
interaction region. The outer velocity decreases substantially and the boundary layer
itself thickens. This results in a higher value for the shape factor (H ) and a lower
velocity gradient at the wall.
The outer flow through the shock wave experiences an almost instantaneous rise in
static pressure. In other words: dCp /dx → ∞. Based on our previous discussion on
the effect of the pressure distribution on boundary layer separation we might therefore
wrongfully conclude that the boundary layer would invariably separate from the sur-
face. However, a closer look into the shock-wave boundary-layer interaction reveals
that the pressure increase is smeared out over a larger streamwise region. This smear-
ing property is due to the formation of the compression waves within the boundary
layer. The effective pressure gradient within the boundary layer is therefore much
smaller and merely results in an increase in shape factor and boundary layer thickness.
Finally, we note that even in the case of a weak interaction the boundary layer
properties are adversely affected. Beyond the shock wave the boundary layer can
6.4 Fundamentals of Boundary-Layer Flow 323
When the shock wave strength increases, the interaction between shock wave and
boundary layer becomes stronger. The sonic line descends deeper into the boundary
layer and a different equilibrium results. Let us try to explain this equilibrium by
considering Fig. 6.41. Here we have a pocket of reversed flow (separation bubble).
A separation line is present between the reversed flow and the remaining boundary
layer flow. The streamlines outside of the separation bubble need to bend away from
the wall. In the supersonic domain (above the sonic line) this strong curvature can
only be supported by the formation of compression waves. Outside the boundary
layer these compression waves merge into an oblique shock. The bending of the
streamlines above the sonic line cause the boundary layer to rapidly grow in thick-
ness. To align the streamlines with the wall a second oblique shock is present. The
two oblique shocks meet at the triple point above the boundary layer to form a normal
shock wave. This is the typical lambda shock pattern that is often seen on the upper
surface of transonic wings.
Flow outside the boundary layer that goes through the two oblique shocks expe-
riences a different deceleration than flow going through the single normal shock.
A velocity difference is therefore present resulting in two velocity domains which
are separated by a shear layer (known as slip line). Further downstream the velocity
differences over the slip line are reduced to zero. The flow that goes through the two
oblique shocks experiences less total pressure loss than the flow going through the
normal shock wave. Typically the region between the triple point and the boundary
layer edge is relatively small. To reduce the loss in total pressure (and the corre-
sponding wave drag) it is therefore beneficial to smear the separation region over a
oblique shocks
M e <1
M>1 M>1
M e >1
M>1
separation line M<1
separation
S bubble R
Fig. 6.41 Detail of strong shock-wave boundary-layer interaction for a turbulent boundary layer
(after: [16])
324 6 Aerodynamics of Non-lifting Bodies
larger distance such that the triple point is moved upwards and the integrated total
pressure loss over the shock system decreases.
The velocity profile of the boundary layer changes substantially over this interac-
tion. Initially, the profile shows a high pressure gradient near the wall. However, due
to the large adverse pressure gradient under the oblique shock, the velocity at the wall
quickly reduces to zero at point S. The flow separates and a bubble of recirculating
flow forms. The boundary layer shows a negative velocity gradient at the wall. The
reverse flow aids significantly in the upstream transmission of the post-shock pres-
sure rise that causes the separation of the boundary layer in the first place. Beyond
the second oblique shock the deceleration of the subsonic flow in the boundary layer
causes the streamtube to expand. This aids in the reattachment of the boundary layer.
At the reattachment point (R) the velocity gradient at the wall is again zero. Over
the distance of the separation bubble the sonic line gradually moves away from the
wall towards the boundary layer edge. Further downstream of the bubble the velocity
profile has significantly thinned compared to the initial velocity distribution due to
the lower edge velocity and thicker boundary layer.
The onset of separation at the foot of the shock is dependent on the development of
the boundary layer in front of the shock and the surface curvature at the shock. How-
ever, for general purposes we can assume that on a flat plate separation starts at shock
Mach numbers (Me ) in the range of 1.3–1.35 [5]. In many practical cases the flow
behind the shock wave is further decelerated towards the trailing edge of the body.
Under influence of such an adverse pressure gradient the flow might separate close to
the trailing edge (trailing edge separation). Increasing the shock Mach number pushes
the shock further aft. In addition, the separation bubble grows, which also moves the
re-attachment point further aft. Finally, the reattached boundary layer is now more
likely to separate under the adverse pressure gradient. This moves the trailing-edge
separation point more forward. At some Mach number the reattachment point gets
close enough to the point of trailing edge separation such that the flow does not reat-
tach anymore. This generates a large wake that originates at the foot of the shock.
This shock-induced separation is an important contributor to the exponential increase
in drag (drag divergence) that occurs at the drag divergence flight Mach number.
For more information on the interaction between the normal shock and the bound-
ary layer the reader is referred to Refs. [5, 16].
Now that we have familiarized ourselves with the qualitative aspects of boundary-
layer flow, we turn to the mathematical models that describe it. In this section we
present the boundary layer equations and show how they can be reduced to model
the flow in laminar and turbulent boundary layers. In addition, we also present
the e N method to predict where the boundary layer transitions from laminar to
turbulent.
6.5 Boundary-Layer Computations 325
For a body in a high Reynolds number flow, viscous forces are dominant in essen-
tially three domains: the boundary layer, the wake of the body, and the shock wave
(if present). If the boundary layers stays attached to the airfoil, inviscid models that
permit the existence of strong shocks and rotational flow (e.g. Euler equations, see
Sect. 2.7.1) can be used in the region outside of the boundary layer. Prandtl’s equa-
tions can be used in the laminar boundary-layer region [33]. The governing continuity,
momentum and energy equations for a general unsteady, compressible, viscous flow
are a set of coupled nonlinear partial differential equations, which have been pre-
sented in Sect. 2.5. If we neglect the normal stress (τx x = τ yy = 0), and assume
steady, incompressible, two-dimensional flow, these equations can be simplified for
two-dimensional flow as follows:
continuity equation
∂u ∂v
+ =0 (6.79)
∂x ∂y
x-momentum equation
∂u ∂u ∂p ∂ ∂v ∂u
ρu + ρv =− + μ + (6.80a)
∂x ∂y ∂x ∂y ∂x ∂y
y-momentum equation
∂v ∂v ∂p ∂ ∂v ∂u
ρu + ρv =− + μ + (6.80b)
∂x ∂y ∂y ∂x ∂x ∂y
energy equation
V2 ∂2 T ∂2 T ∂(uτx y ) ∂(vτx y )
ρV · ∇ e + = k 2 + k 2 − ∇ · pV + + (6.81)
2 ∂x ∂y ∂y ∂x
∂u ∂v
+ =0 (6.82)
∂x ∂y
x-momentum equation
∂u ∂u dpe ∂ ∂u
ρu + ρv =− + μ (6.83a)
∂x ∂y dx ∂y ∂y
y-momentum equation
∂p
=0 (6.83b)
∂y
energy equation
2
∂h ∂h ∂2 T d pe ∂u
ρu + ρv =k 2 +u =μ (6.84)
∂x ∂y ∂y dx ∂y
Note that subscript “e” represents the conditions at the boundary-layer edge. Also
note that the y-momentum equation basically states that the pressure stays con-
stant throughout the boundary layer. The pressure is therefore solely dependent on
x. Hence, from now on we replace its partial derivative operator (∂) with a total
derivative operator (d).
By defining the boundary values at a solid surface, this set of equations can be
seen as a boundary value problem (BVP). To assess this BVP, consider a coordinate
system which is bound to a solid object, with the y-axis perpendicular to its surface
and the x-axis aligned with the flow. The wall boundary values usually imply the
no-slip condition:
u(x, 0) = 0, v(x, 0) = v0 (6.85)
In the general case, Tw is a function of time, i.e. Tw = Tw (x, t). In the case of
adiabatic wall conditions the heat flow to the wall, qw is zero and the following
boundary condition results:
∂T
=0 (6.87)
∂ y y=0
If adiabatic wall conditions exist, the adiabatic wall temperature (Tw = Taw ) is a
part of the solution rather than a set boundary condition. The steady-state adiabatic
wall temperature can, however, be computed by using (6.74). At the outer edge
(subscript e) of the boundary layer, the following boundary conditions exist:
An estimate of the values of u e (x) and Te (x) can be obtained by calculating the
flow around an object using for example the Euler equations. The five boundary
conditions together with the coupled equations of motion [(6.82)–(6.84)] form the
basis for the analysis of laminar boundary layer flow. A significant simplification with
respect to the original equations of motion can be observed. It is stressed that the
above equations are derived for laminar boundary layers, assuming incompressible
flow and a thin boundary layer. A more complex order-of-magnitude analysis can
be carried out for boundary layers which exhibit a significant thickness [48, 55].
However, this is beyond the scope of the present text.
When we integrate the boundary-layer momentum Eq. (6.83a) from y = 0 to
y = δ and also employ (6.82) we obtain the Von Kármán momentum integral relation,
(also known as the momentum integral equation):
dθ θ du e τw cf
+ (2 + H ) = 2 = (6.89)
dx u e dx ρu e 2
The Blasius equation can be derived from the boundary layer equations. It reduces
the system of equations [(6.82)–(6.84)] to a single ordinary differential equation (see
Ref. [3] for complete derivation):
1
f + ff =0 (6.90a)
2
subject to
f = f = 0 on η = 0 (6.90b)
f → 1 as η → ∞ (6.90c)
328 6 Aerodynamics of Non-lifting Bodies
where
V∞
η=y (6.90d)
νx
u
f (η) = (6.90e)
V∞
The resulting velocity gradient at the wall has been quoted in (6.64). The displacement
thickness of a typical Blasius velocity profile (incompressible conditions) can be
shown to be a function of the Reynolds number according to:
1.72x
δ∗ = √ (6.91)
Rex
0.664x
θ= √ (6.92)
Rex
We note that both √the momentum thickness and the displacement thickness increase
proportional to x. At the trailing edge of the plate we substitute x = c. If we
compare the result to (6.67) we see that the momentum thickness at the trailing edge
is directly related to the friction drag over the plate:
2θx=c
Cf = (6.93)
c
This is an important result because it is not limited to laminar boundary layers or
the Blasius velocity profile. It is a universal observation that holds for both turbulent
and laminar boundary layers: the reduced momentum in the boundary layer at the
trailing edge is a direct measure for the amount of friction drag of the profile. More
general, any loss of momentum can directly be correlated to the drag of a body (see
also Fig. 6.2). In this case, the momentum deficiency is caused by the friction of the
body.
The Blasius solution only holds for a flat plate where the external velocity dis-
tribution is constant. To predict the boundary-layer properties in the presence of a
pressure distribution, an engineering approach can be used. For example, the method
of Thwaites [56] predicts these properties regardless of the exact shape of the veloc-
ity profile inside the (laminar) boundary layer. We will derive this method and show
some of its properties.
Thwaites’ method starts from the Von Kármán integral relation. We first multiply
the momentum integral equation (6.89) by θu e /ν:
τw θ u e θ dθ θ2 du e
= + (2 + H ) (6.94)
μu e ν dx ν dx
6.5 Boundary-Layer Computations 329
θ2 du e
λ= (6.95)
ν dx
Subsequently, we correlate the shear stress parameter [(S(λ)] and the shape factor
[H (λ)] to λ using Thwaites empirical relations [66]:
τw θ
S(λ) = ≈(λ + 0.09)0.62 (6.96)
μu e
H (λ) ≈2.0 + 4.14z − 83.5z 2 + 854z 3 − 3337z 4 + 4576z 5 (6.97)
with z = (0.25 − λ). Note that when S → 0 the shear stress goes to zero. In other
words, the boundary layer separates when λ < −0.09. Substituting (6.95) and (6.96)
in (6.94) reduces it to:
d dx
ue λ = 2 [S − (2 + H )λ] = F(λ) (6.98)
dx du e
Thwaites shows that F(λ) = 0.45 − 6λ. We can therefore rewrite (6.98) to the
following differential equation [employing (6.95)]:
d θ2 θ2 du e
ue +6 = 0.45 (6.99)
dx ν ν dx
Integrating this differential equation and assuming that the integration constant is
zero results in the following relation between boundary-layer edge velocity and
momentum thickness:
0.45ν x
θ= 6 u 5e dx (6.101)
ue
0
used to obtain estimates of the displacement thickness and shear stress at any point
on the body. By doing a number of iterations of inviscid and displacement-thickness
calculations, a converged value of the displacement thickness results. It can be shown
that the displacement thickness is about 1/3 of the boundary layer thickness [48].
The following example illustrates this process.
Solution:
(a) The analytic nature of the velocity distribution allows us to compute θ from
(6.101):
0.45ν x x
5 νc x
−6
θ= 6
5 1−
V∞ dx = 0.075 1− −1
6 1− x
V∞ c V∞ c
c 0
The turbulent boundary layer differs substantially from the laminar boundary layer.
Let us first examine which layers we can distinguish within the turbulent boundary
layer. Four layers can be identified: a viscous sublayer that touches the wall (denoted
with subscript w), a merging layer, a large-eddy region, and a viscous superlayer. To
represent the velocity distribution in each of these layers we need to introduce a new
variable: u τ , the wall friction velocity (see Problem 6.19):
τw
uτ = (6.103)
ρ
We nondimensionalize the velocity in the boundary layer and the boundary layer
thickness by introducing two additional new variables:
u
u+ = (6.104)
uτ
√
+ |τw |/ρw yu τ
y =y = (6.105)
νw ν
u+ = y+ y+ < 5 (6.106)
The velocity in the merging layer (sometimes called “overlap” layer or “blending”
layer) between the laminar sublayer and the large-eddy region can be predicted using
a formula provided by Spalding [54]:
+ + −κB κu + + (κu + )2 (κu + )3
y =u +e e − 1 − κu − − y + < 100 (6.107)
2 6
From experiments, the Von Kármán constant is found to be κ = 0.41 and B = 5.0
for a smooth wall. This equation is termed the law of the wall.
In the large-eddy region the velocity changes logarithmically with distance from
the wall. The following approximation by Coles and Hirst [12] is often used:
1
u+ = ln y + + B with κ = 0.41 and B = 5.0 y + > 30 (6.108)
κ
The logarithmic region in the boundary layer extends to 10–20 % of the total boundary
layer thickness, depending on the value of Reδ . In the “wake” component of the
boundary layer, the local pressure distribution has a significant effect on the velocity
distribution. Therefore, Coles [11] proposed to add an additional quantity, Π , to
account for the local pressure distribution. The resulting law of the wake is written
as follows:
1 2Π y
u+ ≈ ln y + + B + f y + > 30 (6.109)
κ κ δ
It was shown that Coles’ wake parameter, Π , could be related to the displace-
ment thickness of the boundary layer by integrating (6.109) over the boundary layer
thickness:
δ∗ 1+Π 2
≈ with λ = (6.110)
δ κλ cf
The value of λ is a measure for the local skin friction coefficient and is useful in
the analyses which follow. Note that this λ is not the same as the λ we used in the
analysis of the laminar boundary layer with Thwaites method (Sect. 6.5.2). Finally,
the wake function f (y/δ) is of the form [11]:
y
πy
f = sin2 (6.111)
δ 2δ
This implies that the wake function is zero at the wall and unity at the edge of the
boundary layer. This function allows for the S-shape we expect to see in the velocity
profile in the wake of the boundary layer whenever there is a pressure gradient present
(see Sect. 6.4). At the edge of the boundary layer, we can now relate the edge velocity
334 6 Aerodynamics of Non-lifting Bodies
50
Viscous sublayer y + < 5
45 Coles and Hirst: Large−eddy layer y > 30
+
+
Spalding: Merging layer y < 100
40
35 Π=6
Π=4
30 Π=2
u = u/u τ
viscous merging
sublayer layer
25 Π=0
+
20
15
logarithmic outer
10 region region
large-eddy
5 layer
0
0 1 2 3 4
10 10 10 10 10
+
y = y u τ /ν
Fig. 6.42 Velocity distribution in a turbulent boundary layer for various values of Π , assuming
δ + = 10,000
ue 1 Reδ 2Π
u + (y = δ) = = ln +B+ (6.112)
uτ κ λ κ
The velocity distributions as presented in this section are summarized in Fig. 6.42
for various values of Π . If we connect the prediction of Spalding to that of Coles and
Hirst, we observe the typical S-shape in the velocity profile of the boundary layer flow.
The effect of the pressure distribution on the development of the turbulent boundary
layer has been demonstrated qualitatively in Fig. 6.42. We see that for various values
of Π the velocity profile in the outer region can change significantly. We would like
to know how the value of Π is related to the pressure distribution outside of the
boundary layer. Therefore we return to the momentum integral equation as derived
by Von Kármán [62]:
dθ θ du e τw cf
+ (2 + H ) = 2 = (6.89)
dx u e dx ρu e 2
In order to use this equation we assume that an outer velocity distribution is known
and that we need to obtain values for cf , θ, and H . In order to do so we require at
least two additional equations. One is given by Coles’ wall-wake law: (6.112). Based
6.5 Boundary-Layer Computations 335
on this relation the following algebraic equation due to Felsch et al. [19] describes
the relation between H , Reθ and cf :
The third relation stems from a large body of experimental research where Π is
correlated to Clauser’s equilibrium parameter β which is defined as follows [10]:
δ ∗ d pe θ du e
β= = −λ2 H (6.114)
τw dx u e dx
where d pe /dx is the pressure gradient at the edge of the boundary layer and λ is as in
(6.110). Clauser’s equilibrium parameter is directly related to the pressure gradient.
Whenever β < 0 we have a favorable pressure gradient (d pe /dx < 0). Vice versa,
when β > 0 we have an adverse pressure distribution (d pe /dx > 0). The proposed
correlation between Π and β stems from Ref. [15]:
H 2 + 3.2Π + 1.5Π 2
λ= (6.116)
H −1 κ(1 + Π )
Let us review this method and put this in a practical perspective. We would like to
calculate the properties of the boundary layer at various stations along a plate under
the presence of a pressure distribution. To that extent we intend to solve the ordinary
differential equation (6.89) with known boundary conditions. These boundary con-
ditions could for example stem from an assumed value of the momentum thickness
or Reynolds number at x = 0. Subsequently, we use the algebraic equations (6.113)–
(6.116) to close the problem.
Example 6.4 Assume we have a pressure distribution over a curved plate of unit
length that can be represented by the following polynomial:
Cp = bx(ax − 1)
where a and b are coefficients. Let us further assume that the boundary layer over the
curved plate is made fully turbulent at x = 0 and that the initial Reynolds number
based on momentum thickness, Reθ = 100. Furthermore, the freestream velocity is
V∞ = 50 m/s and 0 m ISA conditions apply. Consider the following cases:
336 6 Aerodynamics of Non-lifting Bodies
1. a = 21 and b = 1
2. a = 1 and b = 21
3. a = 2 and b = 2
For these three cases perform the following tasks:
(a) Plot the pressure distribution, velocity distribution and distribution of the velocity
gradient for each of the aforementioned cases.
(b) Plot the distribution of the following boundary layer parameters for the three
aforementioned cases: H , β, Π , cf , θ, and δ ∗ .
(c) Briefly reflect on the results that have been plotted.
Solution:
du e V∞ b(1 − 2ax)
= √
dx 2 1 + bx(1 − ax)
We substitute each combination of a and b in the above equations and plot the
results in Fig. 6.43. A quick evaluation of these three graphs shows us that for
case 1 we have a completely favorable velocity gradient. For cases 2 and 3 we
have a velocity gradient that is favorable until x = 0.5. Beyond this point the
pressure gradient is negative, which implies that the outer flow is decelerated.
This effect is four times larger for case 3 than four case 2.
(b) To answer the second part of the question we must solve the first order differential
equation (6.89). We rewrite (6.89) according to:
dθ cf θ du e cf θb 1 − 2ax
F= = − (2 + H ) = − (2 + H ) (6.117)
dx 2 u e dx 2 2 1 + bx(1 − ax)
−0.5 65 50
−0.4
du /dx (1/s)
60
ue (m/s)
−0.3
Cp (~)
0
−0.2
e
55 a=0.5, b=1
−0.1 a=1, b=0.5
a=1, b=2
0 50 −50
0 0.5 1 0 0.5 1 0 0.5 1
k1 = F(xn , θn )Δx
1 k1
k2 = F(xn + Δx, θn + )Δx
2 2
1 k2
k3 = F(xn + Δx, θn + )Δx
2 2
k4 = F(xn + Δx, θn + k3 )Δx
1
θn+1 = θn + (k1 + 2k2 + 2k3 + k4 )
6
In this method we march from position xn to position xn+1 with steps of Δx. At
each position we calculate the value of θ. To start this iteration we need a starting
value for θ. We compute this through the value of Reθ in combination with the
0 m ISA properties and the value of u e at x = 0. For all three cases u e = V∞ at
x = 0. We have:
Reθ μ
θ1 = = 29 µm
ρV∞
In order to do determine the values of H at each position we need to solve the four
algebraic equations (6.113)–(6.116) simultaneously. If we substitute (6.114) in
(6.115) we are left with three equations that are written in the following func-
tional format:
−0.268
ueθ
G(1) = 0.058 (0.93 − 1.95 log10 H )1.705 − cf = 0
ν
2 θ du e
G(2) = −0.4 + 0.76Π + 0.42Π 2 + H =0
cf u e dx
H 2 + 3.179Π + 1.5Π 2 2
G(3) = − =0
H −1 κ(1 + Π ) cf
Notice that we have three equations with three unknowns: H , cf , and Π . The
Matlab routine “fsolve” is used to compute the roots of this system of equations.
Having computed H and cf we can now evaluate the RHS of (6.117) and com-
pute θn+1 . Finally, we use this new value of θ in combination with H to compute
5For details on the Runge-Kutta method the reader is directed to an introductory text on numerical
methods for differential equations, which is also covered in Ref. [30].
338 6 Aerodynamics of Non-lifting Bodies
−3 −3
x 10 x 10
5 5
4 4
δ* (m)
θ (m)
3 3
2 2
1 1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
0.01
3
a = 0.5, b =1
2 a = 1, b = 0.5
cf (~)
H (~)
0.005 a = 2, b = 2
1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x (m) x (m)
Fig. 6.44 Distribution of boundary layer parameters according to three different pressure
distributions
δ ∗ via (6.63). We repeat this procedure at every xn until xn = 1. In Fig. 6.44 the
results are plotted for the three cases.
(c) If we look at case 1, a favorable pressure gradient is present. This favorable
pressure gradient results in an almost constant value of the shape factor. If we
compare this to the second case we see the effect of the adverse pressure gradient
on the momentum thickness and displacement thickness. We also see that the
friction coefficient becomes smaller due to a slightly higher value of the shape
factor. Remember that this corresponds to a lower velocity gradient at the wall
(see also Fig. 6.26). Finally, we have case 3 where the shape factor rises much
quicker due to the larger adverse pressure gradient. If we examine the friction
coefficient we see that it tends towards zero around x = 0.62. We know that this
is the condition of separation. In this example the flow separates from the surface.
We see an exponential increase in θ and δ ∗ and we notice that H goes to a value
close to 3.0. Beyond x = 0.62 the calculation of the roots of the functions G gives
a complex solution and is therefore neglected in the generation of these plots.
If we consider real airplanes, we know that the surface of the airplane is covered with
various imperfections ranging from small rivet heads to larger antennas. To represent
these imperfections, for every airplane an equivalent sand grain roughness can be
defined. To predict the total friction drag of the airplane, the skin is assumed to be
covered with this uniform sand grain roughness instead of the real excrescences. The
6.5 Boundary-Layer Computations 339
parameter that determines the significance of the roughness is the ratio between the
sand grain height (k) and the boundary-layer thickness. We distinguish two cases:
1. Hydraulically smooth surface. If k is smaller than the height of the viscous sub-
layer, i.e. k < δv , k does not affect the boundary layer significantly. In other
words: the friction coefficient is unaffected by the presence of roughness. We
can imagine each sand grain to be a bluff body that is immersed in a very low
Reynolds-number flow. In such flow (see Fig. 6.22a) the streamlines are attached
over the complete body. The drag that results in this flow (Stokes flow) is due to
friction and therefore has the same dependency on Reynolds number as a perfectly
smooth surface.
2. Hydraulically rough surface. If k is larger than the height of the viscous sub-
layer, i.e. k > δv , the sand grains create small wakes behind them due to flow
‘separation.’ The pressure drag that is generated at each sand grain results in
the approximate friction drag. Because the pressure drag is less dependent on
the Reynolds number than the friction drag, the friction coefficient over a rough
surface also has much lower dependency on the Reynolds number. The value of
the friction coefficient is dominated by the relative size of the sand grains with
respect to the boundary layer thickness: k/δ.
If we assume the sand grains to have an average height of y = k, we can introduce
a nondimensional sand grain height according to (6.105):
ku τ
k+ = (6.118)
ν
From experiments (e.g. Clauser [10]) it is known that the layer where velocity changes
logarithmically with wall distance still exists when roughness is present. However,
the intercept of the logarithmic curve moves downward with progressively larger
values of k + . This alters (6.108) as follows:
1
u+ = ln y + + B − ΔB(k + ) (6.119)
κ
1
ΔBsand grains ≈ ln(1 + 0.3k + ) for k + > 60 (6.120)
κ
The result is the following expression for the nondimensional velocity profile in a
boundary layer that is disturbed by roughness:
1 1
u+ = ln y + + B − ln(1 + 0.3k + ) (6.121)
κ κ
340 6 Aerodynamics of Non-lifting Bodies
From the experiments it is know that for values below k + < 4 roughness has no
effect on the velocity profile in the boundary layer. Between 4 < k + < 60 there
is a transitional roughness regime. Finally, above k + > 60 the flow has become
hydraulically rough. Based on (6.121) it can be shown that as the sand grains become
larger, the effect of viscosity becomes negligible and the friction coefficient therefore
becomes independent of the Reynolds number (see Problem 6.23):
1 y
u+ = ln + 7.9 (6.122)
κ k
We can insert (6.121) into (6.83a) under the assumption of an absent pressure
gradient. Integrating the result over the thickness of the boundary layer and the length,
x, results in a relation between the friction coefficient, cf , the Reynolds number, Rex
and the grain size, k [66]:
0.3k +
Rex = 1.73(1 + 0.3k + )e Z Z 2 − 4Z + 6 − (Z − 1) (6.123)
1 + 0.3k +
where Z = κλ. This relation, even though it is implicit, is valid for uniform sand-
grain roughness over the complete range of hydraulically smooth, transitional, and
fully rough walls in turbulent flat plate flow. If we write k + = ku τ ν = Rex (k/x)
λ , we
can plot the local friction coefficient as a function of the local Reynolds number for
various values of x/k. This is shown in Fig. 6.45. The effect of the roughness is evident
from this figure. The larger the grains compared to the length x the higher the friction
coefficient. In addition, we see that the friction coefficient of a smooth plate decays
with Reynolds number (as expected). However, we also see that the decaying effect
of the Reynolds number diminishes for roughened plates above a certain Reynolds
number. The smaller the sand grains compared to the size of the distance x, the larger
the Reynolds number at which the Reynolds number ceases to decrease the friction
coefficient. This Reynolds number is called the cut-off Reynolds number. This is in
x
=10
k
−2 100
10
1,000
104
105
106
−3
10
5 6 7 8 9
10 10 10 10 10
Reynolds number, Re x (~)
6.5 Boundary-Layer Computations 341
.0015
.0010
20 30 40 50 60 80 100 200
Reynolds number, Rel x 10-6
good agreement with classical experiments carried out by Nikuradse on pipe flow
with roughened walls [42].
The results plotted in Fig. 6.45 can be directly translated to the friction drag coeffi-
cient of full scale aircraft and wind tunnel models. In Fig. 6.46 the variation in friction
coefficient with Reynolds number is shown for three airplanes of the same generation
and built by the same company (Lockheed). It could be assumed that the manufac-
turing techniques for all three airplanes was of the same level leading to identical
values for the equivalent sand grain roughness of the skin of the airplane. What is
shown is the relative sand grain height, scaled with respect to a characteristic length,
l, of the airplane. Similar to the theoretical curves in Fig. 6.45, the largest airplane
(C-5) demonstrates to have the largest cut-off Reynolds number (≈100 million). This
advocates the statement that large aircraft benefit from their size when it comes to
the friction drag coefficient.
Example 6.5 Assume we have a flat plate with an equivalent sand grain roughness of
0.010 mm. Calculate using (6.123) the friction coefficient (cf ) at x = 2.5 m. Assume
that we have 0 m ISA conditions and a freestream velocity of 100 m/s. Also calculate
the wall stress, τw and the velocity gradient at the wall, (∂u/∂ y)w .
Solution:
We first calculate the Reynolds number at x = 2.5 m:
k 0.010 × 10−3
= = 2.5 × 105
x 2.5
342 6 Aerodynamics of Non-lifting Bodies
Because (6.123) cannot be solved directly, we use a “for” loop in Matlab according to
the following routine. We first establish a set of possible values for λ. Subsequently,
for each value of λ we calculate the value for k + , Z , and Rex :
for
λ = [1, 1.001, 1.002, . . . , 50]
do
Rex (k/x)
k+ =
λ
Z = κλ
+ 0.3k +
Rex = 1.73(1 + 0.3k )e Z
Z − 4Z + 6 −
2
(Z − 1)
1 + 0.3k +
end
Finally, we compare the outcome of Rex to the value that we computed above:
We subsequently find one value of λ for which ΔRex is closest to zero. This results
in λ = 26.3. Substituting this in (6.110) results in:
2
cf = = 0.00290
λ2
From the friction coefficient we can calculate the local shear stress according to
(6.55):
1 1
τw = cf ρU02 = 0.00290 · · 1.225 · 1002 = 17.7 N/m2
2 2
The local velocity gradient can be found by employing (6.54):
∂u τw m/s
= = 0.994
∂y w μ µm
Example 6.6 Considering the same roughened plate as in Example 6.5, calculate the
value of the total friction coefficient over the entire plate. Assume the plate has a
length of l = 3 m and that the boundary becomes turbulent at x = 0.10 m. Ignore
the contribution of the laminar boundary layer over the first 10 cm to the friction
coefficient.
Solution:
The friction coefficient of the plate is given by (6.66) with the upper bound x = 3 and
the lower bound x = 0.10. We evaluate this integral numerically by dividing the plate
6.5 Boundary-Layer Computations 343
1
29
Cf = C f j = 0.0030
l
j=1
From an operational perspective, there has been a long desire to minimize the drag
of an airplane to a minimum. In the context of boundary-layer theory, this desire has
translated to the objective of having as much laminar flow over a surface as possi-
ble. In a two-dimensional boundary layer, transition is governed by the formation
of Tollmien-Schlichting waves. As stated before, the amplification of these waves
depends on many factors. Experience has shown that a “fuller” boundary layer usu-
ally results in lower disturbance growth [47]. Let us rewrite (6.83a) for y ≈ 0 and
add the body force term that we have left out before:
∂2u ∂u dp dμ ∂T ∂u
μ = ρv0 + − − fx (6.124)
∂y 2 ∂y dx dT ∂ y ∂ y
Note that we have split the partial derivative ∂/∂ y(μ∂u/∂ y) and have used ∂μ/∂ y =
(∂μ/∂T )·(∂T /∂ y). The left-hand side of (6.124) is a measure for the curvature of the
boundary layer profile near the wall. Generally speaking, a “fuller” boundary layer
corresponds to a convex velocity profile, i.e. a more negative value of ∂ 2 u/∂ y 2 . A neg-
ative pressure gradient (d p/dx < 0) therefore results in lower disturbance growth,
while an adverse pressure gradient (d p/dx > 0) increases disturbance growth. One
can also heat or cool the wall. Wall cooling in air (∂T /∂ y > 0, ∂μ/∂T > 0) stabi-
lizes the boundary layer, while heating it (∂T /∂ y < 0, ∂μ/∂T > 0) destabilizes the
boundary layer. Alternatively, one can apply suction perpendicular to the boundary
layer (v0 < 0) in order to reduce the growth of disturbances. This has been success-
fully demonstrated to postpone transition in the wind tunnel and in flight [7, 44].
Finally, one can apply a body force to the flow in the direction of the velocity vector.
This can be done by locally ionizing the air using an electric field. By employing
an asymmetric electrode arrangement, a body force can be exerted on the resulting
plasma [13]. It has been shown that a periodic body force is effective in attenuating
the Tollmien-Schlichting waves [29]. Both suction and plasma control are examples
of active laminar flow control (LFC).
344 6 Aerodynamics of Non-lifting Bodies
In the early 20th century William Orr and Arnold Sommerfield independently
derived the linear homogeneous ODE describing the infinitesimal laminar flow
instability based on the steady continuity equation (2.113) and the unsteady linear
momentum equation (2.126). In contrast to our earlier assumption, Orr and Sommer-
field allowed infinitesimal velocity perturbations perpendicular to the flow direction
within the boundary layer. They assumed that these perturbations changed harmon-
ically with time and could grow exponentially in time and space depending on the
frequency of the wave, the wave propagation speed, the kinematic viscosity, and
the shape of the velocity profile. For negligible viscosity, the solution of the Orr-
Sommerfield equation showed that only velocity profiles with an inflection point in
the velocity profile would amplify the instabilities. Additionally, the second deriva-
tive of the velocity with respect to y (distance perpendicular to the surface) needed
to be smaller than zero. This led to the mistaken assumption that a favorable pres-
sure gradient could stabilize the boundary layer up to an infinite Reynolds number.
Prandtl demonstrated, however, that viscosity can indeed be destabilizing for certain
wave numbers at a finite Reynolds number.
A more elaborate discussion of the the stability of laminar boundary layers can
be found in the text of White [66]. A selection of the results for boundary layers over
smooth flat plates is listed below:
1. The minimum or critical Reynolds number for initial instability is Rex crit ≈
91,000.
2. The smallest unstable wavelength is λmin ≈ 6δ. Thus, unstable Tollmien-
Schlichting waves are long compared to boundary layer thickness.
3. Compared to Rex crit ≈ 91,000, the point of final transition to turbulence is at
about Rex tr ≈ 3 × 106 , or about 30 times further downstream.
In practice, the transition process is influenced by many factors: surface roughness,
freestream disturbances, pressure gradients, vibration, sound, etc. Therefore, the
exact Reynolds number (or position, for a given fluid and speed) can vary with time
in a random way. Placing objects in the boundary layer results in a swift amplification
of the instabilities in the boundary layer and hastens the onset of transition. Such trip
strips are often used in wind tunnel experiments to fix the point of transition at a fixed
percentage of the wing chord. Other sources of instabilities stem from the attachment
line of the boundary layer on the leading edge of objects that are placed oblique to
the flow (such as the leading edge of a swept wing). In addition, instabilities arising
from three-dimensional cross flow (also found on swept wings) hasten the onset of
laminar to turbulent transition. This will be further discussed in Chap. 8 when we
treat the aerodynamics of swept wings.
In the derivation of the critical Reynolds number a Blasius6 profile was assumed
for the velocity distribution, u(y), in the boundary layer. However, it was shown that
the critical Reynolds number can be 90 times higher than this value when the velocity
6 The Blasius velocity profile satisfies the nonlinear ordinary differential equation that can be derived
from the boundary layer equations under specific assumptions, see (6.90a).
6.5 Boundary-Layer Computations 345
Point of dp
inflection =0
dx
dp
< 0 (favorable)
dx
profile in the boundary layer satisfies the condition: ∂ 2 u/∂ y 2 < 0. Such velocity
profiles are very stable and postpone transition to a much higher Reynolds number.
Such a stable velocity profile can be formed when a favorable pressure gradient
is present as is shown in Fig. 6.47. The boundary layer profile that is subjected to
an adverse pressure gradient shows an inflection point, which means that ∂ 2 u/∂ y 2
changes sign in this point. In this case the critical Reynolds number (and the transition
Reynolds number) is decreased.
The critical Reynolds number can be shown to correlate to the shape factor of the
boundary layer at the point where instabilities start to occur. The shape factor can,
in turn, be calculated by the velocity distribution in the x-direction outside of the
boundary layer (as was shown through Thwaites method on pp. 305–306). In Fig. 6.48
this correlation is experimentally demonstrated through a range of experiments. Both
the critical Reynolds number as well as the transition Reynolds number are shown.
10 7
Re x, tr
10 6
Re x, crit
10 5
10 4
10 3
2.0 2.2 2.4 2.6 2.8 3.0
Shape factor, H = δ
*
(~)
θ
346 6 Aerodynamics of Non-lifting Bodies
The curve that is fitted through transition data points corresponds to the following
curve-fitted equation [64]:
By using Thwaites method for the calculation of the local shape factor and combin-
ing that with (6.125), a swift prediction of the transition Reynolds number can be
obtained.
Example 6.7 Consider the flow problem of Example 6.3. Use the criterion of (6.125)
to calculate where transition is expected to occur.
Solution:
We need to calculate the local Reynolds number at each x-position, Rex . Subse-
quently, we need to evaluate whether the following inequality holds:
factor at the beginning of transition. For turbulence levels below Tu < 0.1 % the
experimental data shows too much scatter to warrant a simple relation as (6.126).
The turbulence intensity (in percentages) is defined as follows:
RMS(u )
Tu = 100 · (6.127)
ū
dN
N= (Hk ) Reθ − Reθ0 (Hk ) (6.128)
dReθ
In the above equations the parameter Hk is the kinematic shape factor, which is
calculated using the following empirical relation assuming a constant density across
the boundary layer [67]:
H − 0.290Me2
Hk = (6.132)
1 + 0.113Me2
with Me being the Mach number at the boundary-layer edge and H being the shape
factor from (6.63).
The method described above predicts the onset of transition. The laminar
boundary-layer variables can be calculated using (6.59), (6.61), and (6.63). If the
outcomes are combined with the properties at the boundary-layer edge (u e , pe , Te ),
the above equations can be used to calculate the value of N . If N surpasses a prede-
fined threshold (e.g. N = 9), the boundary layer transitions from laminar to turbulent.
Example 6.8 Assume we have curved plate of length c = 1m, that supports a pres-
sure distribution according to:
348 6 Aerodynamics of Non-lifting Bodies
x
x
!
Cp = 8 −1 +1
c c
The flow over this plate can be assumed incompressible and adiabatic. The freestream
velocity of this flow is V∞ = 50 m/s and the plate is at mean sea-level conditions
of the international standard atmosphere (i.e. 0 m ISA). Carry out the following
assignments:
(a) Use Thwaites’ method to calculate the position of laminar separation and the
momentum thickness distribution.
(b) For Tu = 0.1 % and Tu = 2 % compute the logarithm of the amplification factor,
N , at which transition takes place.
(c) Use the method presented prior to this example to compute the distribution of
the logarithm of the amplification factor, N .
(d) For each turbulence level, compute the position of transition point and mark
these points in the plot of the reduced amplification factor.
Solution:
(a) We follow Example 6.3 to answer the first question. The most convenient way
to do this is to implement all the relevant equations in a spreadsheet program
or other computational software such as Matlab. We first convert the pressure
distribution to a velocity distribution by employing (6.11):
u e = V∞ 1 − Cp
Let us briefly reflect on the result of Example 6.8. First of all, note that the transition
point is ahead of the calculated laminar separation point. The predicted onset of
transition will therefore cause a turbulent boundary layer before the boundary layer
can separate. The separation point therefore has lost its meaning in this example.
Secondly, observe the arrows in Fig. 6.49 indicating the stabilizing region and the
destabilizing region. When N < 0 the amplification factor is smaller than 1, meaning
6.5 Boundary-Layer Computations 349
Logartithm of amplification
15
e9
10
factor, N (~) Tu = 0.1%
5
Tu = 3.0% destabilizing
0
stabilizing
−5
xtr
−10
−15
0 0.2 0.4 0.6 0.8 1
x (m)
Fig. 6.49 Distribution of logarithm of amplification factor: N = ln e N belonging to Example 6.8
that small disturbances in the boundary layer are damped. This corresponds to the
pressure distribution, which features a favorable pressure gradient up to x = 0.5. We
see that when the pressure gradient becomes less negative, the value of N starts to rise
quickly. Moreover, when the pressure gradient is positive (beyond x = 0.5) a very
rapid increase in N can be seen. This implies that initial disturbances in the boundary
layer are rapidly amplified under an adverse pressure gradient and therefore cause a
quick onset of transition. When we discuss natural-laminar flow airfoils in Sect. 7.3,
we will see how an airfoil can be designed in order to sustain laminar flow over a
large percentage of the chord.
We have seen in the preceding sections that any body of a particular curvature causes
the flow over the body to accelerate or to decelerate. In addition we have seen that
the supervelocities directly (via friction) and indirectly (via shocks) affect the drag
of a body in transonic conditions. If we have two bodies that each have their own
drag (e.g. D1 and D2 ) and we would join these bodies together we mostly find that
the cumulative drag of these two bodies is larger than the sum of the individual com-
ponents. If the difference, ΔD = D1+2 − (D1 + D2 ) > 0, we speak of interference
drag. This effect is not only present when two bodies intersect each other but also
when two bodies are in each others vicinity. At this point we would like to empha-
size that the term ‘interference drag’ is merely a term that describes the effect on
drag when two components intersect. The physical cause for drag can, in turn, be
attributed to an increase in pressure drag or an increase in friction drag.
In the classic book of Hoerner [26] various types of interference drag are dis-
tinguished. In the context of the present chapter we focus on the interference drag
caused by the intersection of two bodies. In practice, at least one of the two bodies
is often a lifting body. For example, the intersection of the horizontal tail plane with
350 6 Aerodynamics of Non-lifting Bodies
the vertical tail can cause interference drag. For the explanation of the concept of
interference drag it is not important whether the interfering bodies are designed to
generate lift or not.
0.25
y
0.20 ReC= 4 .105
interference
t
0.15
0
CD
0.10 c = 3t
0.05
0
0 1 2 3 4 5 6
lateral spacing, y/t (~)
Fig. 6.50 Interference drag between two streamlined bodies (after Ref. [26])
6.6 Interference Drag 351
The example above illustrates that two bodies do not need to intersect in order
for their mutual interference to cause separation of the boundary layer. This type
of interference is mainly based on the flow outside the boundary layer, which is
essentially inviscid. However, the magnified acceleration and deceleration of the flow
outside the boundary layer also has an effect on the development of the boundary
layer, as we have seen in Sect. 6.4.2. Therefore, the local friction coefficient and
shape factor of the boundary layer are directly affected by the interference of the two
bodies. The increase in adverse pressure gradient for the bodies in Fig. 6.50 increased
the shape factor of the boundary layer which made the boundary layer more prone
to separation than without the interference of the two bodies.
The effect of interference on the boundary layer is even greater when two bod-
ies intersect. A relevant example is shown in Fig. 6.51a where a streamlined strut
intersects with a flat plate. Both the strut and the flat plat generate a boundary layer.
The boundary layer on the plate is, generally speaking, a lot thicker than the one on
the strut due to the longer distance it has traveled and therefore the time it has had
to develop. This results in a relatively small du/dy|wall . At the intersection of the
two components the boundary layers join. At the leading edge (stagnation) and over
(a)
0.15
increment
CD 0.10
0
l
t
0.05 interference on two ends ( = 0)
Separation
Separation
Fig. 6.51 Interference drag due to the intersection of a curved body with a flat wall. a Interference
between strut and wall (after Ref. [26]). b Interference causing separation between cascades and
wall (after Ref. [22])
352 6 Aerodynamics of Non-lifting Bodies
the rear of the airfoil, the boundary layer is invariably retarded, which often leads to
separation (du/dy|wall < 0). This is illustrated in Fig. 6.51b where the interference
between airfoil cascades and the wall is shown. Near the wall the separated region
on the cascades expand more upstream. In addition, a separated region starts to form
near the leading edge of the cascades. The separated flow, in turn, causes a wake
which results in additional pressure drag. As can be seen in Fig. 6.51a the angle
between the strut and the wall is an important factor for the value of the interference
drag coefficient. When the angle between the strut and the wall becomes smaller
than 90◦ the interference drag also increases. This is due to increased separation in
the narrower one of the two corners.
In order to minimize interference drag between components we need to make
sure that the boundary layer remains attached at all times. The unfavorable addition
of pressure distributions should be avoided. One way to achieve this is to apply
fillets between the two surfaces. This fillet locally reduces the supervelocities at
the intersection of the two bodies. Thereby it reduces the magnitude of the adverse
pressure gradient, which shifts the point of separation further downstream or prevents
the formation of separation all together.
Fig. 6.52 Leading-edge fairing at the wing root of an A340-300 (photo A Pingstone)
-0.1
-0.3
-0.4
-0.5
Cp = 0.0
-0.2
-0.6 Cp = 0.0
Fig. 6.53 Theoretical isobar patterns on area ruled empennage for M = 0.80 and C L h = −0.117
The first example is the wing-body fairing which smoothes the angle between the
wing and the fuselage. Without the presence of such a fairing the boundary layer
that has developed over the fuselage side is likely to separate. An example of such a
fairing is presented in Fig. 6.52.
Another example is the junction between the horizontal and vertical tail plain in
a T-tail (see Fig. 6.53). Often, an additional fairing (acorn) is positioned to reduce
the added supervelocities that are caused by interference. Adding this body allows
the designer to locally change the area distribution of the intersection such that
interference is minimized.
The position of the nacelle with respect to the wing is a third example of how
interference-drag considerations dominate this geometric feature. The positioning
of the nacelle under the wing implies a strong interference between the pressure
distribution over the nacelle, pylon, and lower wing surface. An interesting example
in this respect is the development of the Convair 990. The specification for this
airplane included a very high cruise Mach number (Mcruise = 0.89). During early
experimentation it surfaced that supersonic flow accompanied by strong shock waves
354 6 Aerodynamics of Non-lifting Bodies
were present on the inboard sides of the pylons. The drag rise7 of the airplane
coincided with the drag rise of the nacelles.
To remedy this problem and postpone the onset of drag divergence to a higher
Mach number local area ruling was applied. A representative streamtube was cho-
sen by including (part of) the cross-sectional area of nacelle, pylon and wing. This
streamtube is shown in Fig. 6.54a. A target area distribution was defined that had
A B C
A B C
1.5 Rnacelle
max 1.5 Rnacelle
max
nacelle exit
600
added
area
400
original area distribution
200
0
40 80 120 160 200 240 280 320 360 400
Longitudinal station (in)
Fig. 6.54 Local area ruling between nacelle, pylon and wing of Convair 990 (after Ref. [31]).
a Streamtube boundaries. b Area distribution prior and post modifications. c Drag curve prior and
post modification. d Proposed modifications to nacelle and pylon
7 “Drag rise” is a term that describes the exponential increase in drag coefficient with Mach number.
6.6 Interference Drag 355
coefficient, D (cts)
.0040
.0020
0
.80 .82 .84 .86 .88 .90 .92
Mach number, M (~)
(d) WING
LE
A B terminal
Inboard fairing
nacelle
B
A
forward pylon Sect. A-A Sect. B-B
fairings aft pylon
Wing
fairings
A LE
B
Outboard
nacelle
A B
Sect. A-A Sect. B-B
is even more likely to occur. The integration of winglets to the tip of the wing is an
example where the interference of two force-producing bodies is likely to produce
interference drag. Creating a fillet with a large radius can prevent the onset of strong
shocks in the corner of the two lifting surfaces. For more examples on interference
drag created by lifting and non-lifting components the reader is referred to the text
of Obert [43].
6.7 Summary
There are several parts of a high-subsonic airplane that do not contribute to the
production of lift during the cruise phase of the mission (e.g. fuselage, fin, nacelle).
These surfaces do provide a significant contribution to the total drag that is produced
by the airplane. We have investigated three important sources of drag: friction drag,
drag due to separation, and wave drag.
Wave drag that is associated with the formation of shock waves. A simple method
has been presented to calculate the wave drag of an arbitrary nonlifting body. This
method, which is based on linearized supersonic potential theory, uses an equivalent
axis-symmetric body with the same cross-sectional area distribution as the body
under investigation (transonic area rule). It has been shown how the cross-sectional
area should be distributed along the length of the body in order to minimize the wave
drag. Several examples of the application of area ruling have been presented for both
civil and military aircraft.
Friction drag is associated with shear forces that are generated in the boundary
layer between the outer flow and the airplane surface. It has been shown that the shape
of the velocity distribution in the boundary layer has an effect on the local friction
coefficient of both a turbulent and laminar boundary layer. In addition, it affects the
streamwise position at which the laminar boundary layer transitions to a turbulent
boundary layer. A turbulent boundary layer has a higher velocity gradient at the wall
and produces more friction drag than a laminar boundary layer. In addition, the turbu-
lent boundary layer is much thicker than the laminar boundary layer which affects the
outer pressure distribution through the displacement effect. The shape of the velocity
profile in the boundary layer, in turn, is affected by the outer pressure distribution.
The shape of the body (which is responsible for the pressure distribution) is therefore
of pivotal importance for the total amount of friction drag that is generated. While
the turbulent boundary layer is less sensitive to changes in Reynolds number than the
laminar boundary layer, it is much more sensitive to changes in Mach number. The
presence of surface roughness has an effect on the transition point (shifts forward)
as well as on the local value of the friction drag (becomes larger). Additionally,
it is shown that the friction coefficient becomes smaller with increasing Reynolds
and Mach number. However, for roughened surfaces there exists a cut-off Reynolds
number above which the friction coefficient remains constant with Reynolds number.
6.7 Summary 357
6.1 Consider the flow about the body in the control volume of Fig. 6.1.
(a) Show that the
continuity equation (2.109) can be rewritten according to: intake
ρ1 u 1 dy = exhaust ρ2 u 2 dy.
(b) Show that (6.7) can be derived from (6.6) and the relation shown under (a).
6.2 Consider the flow about the body in the control volume of Fig. 6.1.
(a) If we assume that the velocity defect behind the wing (u 1 −u 2 ) is small compared
C
to the free stream velocity, V∞ , show that D ≈ ρV ∞ (V∞ − u 2 ) dy.
D
(b) Show that, under the same assumptions, the drag can also be derived from the
C
total-pressure defect in the wake of the wing: D ≈ pt∞ − pt2 dy.
D
(c) Based on the items above, how would you measure the drag of a body at section
CD?
358 6 Aerodynamics of Non-lifting Bodies
6.4 Using the same approach as in Example 6.1, draw the notional pressure distrib-
utions about the following axis-symmetric bodies:
6.5 In the figure below you find a side-view of a British experimental aircraft for
radar surveillance (Early Warning), the British Aerospace Nimrod AEW. Sketch the
relation of the static pressure coefficient at the lower and upper crown line of the
forward fuselage in the plane of symmetry.
-1
Cp 0
x
6.6 (a) Derive (6.17) based on the expression for the pressure coefficient and the
isentropic relation between total pressure and static pressure.
(b) If M∞ = 0.8 and Cp = −2, calculate the local Mach number.
(c) If M∞ = 0.8 and Cp = −3, calculate the local Mach number.
(d) Calculate for what value of the Cp the theoretical local Mach number tends to
infinity if M∞ = 0.8.
6.7 Summary 359
6.7 For the free stream Mach numbers ranging from 0.1 to 0.8, graph the increase in
Mach number, ΔM, as a function of pressure coefficient. Use a range of Cp between
−2 and 0.
Wave Drag
6.8 Show that (6.22) can be derived from (6.19), (6.21), and (2.106).
6.9 Consider the perturbed momentum equation in integral form (6.26).
(a) Show that, by employing the continuity equation (6.27), this equation can be
simplified according to:
!
F body = − pn + ρV∞
2
v (v · n + i · n) dS.
S
(b) Argue that the expression above is identical to (6.28) because we integrate over
an enclosed surface, S.
6.10 Consider (6.50) in terms of the Glauert variable θ.
(a) Rewrite (6.50) in terms of x.
(b) If R(x) is the distribution of the radius of the Sears-Haack body, express R(x)
in terms of the maximum radius, Rmax , and x.
6.11 Consider the body described in Example 6.2.
(a) Calculate the volume, V , and maximum radius, Rmax corresponding to the afore-
mentioned body.
(b) Calculate the wave drag of a Sears-Haack body with the same length and volume
as the aforementioned body.
(c) Calculate the relative decrease in wave drag coefficient (in percentages) that can
be attained when modifying this body to the Sears-Haack body with identical
volume and length.
6.12 The aft fuselage of the Airbus A320 exhibits a tailored waist at the location
of the horizontal tail. In addition, the vertical tail is moved forward relative to the
horizontal tail. What are the explanations for their relative positioning?
horizontal tail
vertical tail
360 6 Aerodynamics of Non-lifting Bodies
6.13 The figure below reveals the large wing-fuselage fairing of the Airbus A380.
This fairing is so large that it received the nickname “the bathtub.” Associated with it
are a large wetted area and a weight penalty. Carefully explain the reason of existence
of a wing-fuselage fairing in general, despite the drawbacks mentioned.
6.14 The Convair 990 (see below) was designed to be a jet transport with a cruise
Mach number of M = 0.89. In 1958 this was quite challenging due to the limited
computational resources. Wind tunnel tests on the initial design demonstrated that the
drag coefficient was too large to attain the desired cruise Mach number. A somewhat
unconventional solution was proposed and eventually adopted in the final design.
The solution was to position large bodies near the trailing edge of the wing on the
top surface.
(a) Carefully explain how the addition of these bodies can result in a lower drag
coefficient at the design Mach number.
(b) Name two disadvantages of these bodies.
(c) Why do modern transport aircraft lack these additional bodies?
6.7 Summary 361
Photo NASA
6.15 Consider Fig. 6.24. If we assume that the freestream velocity value of this flow
of water equals 1 m/s and that the streamtube in this figure measures 1 cm.
(a) Calculate the velocity gradients at the three different positions.
(b) Calculate the local value of the wall stress at these three positions.
(c) Calculate the local value of the friction coefficient at these three positions.
Boundary-Layer Computations
6.17 Consider a plate with chord length of c = 10 m, which is exposed to a flow with
V∞ = 100 m/s at 0 m ISA conditions. Consider the following velocity distribution
x
over the plate at the edge of the boundary layer: u e = V∞ 1 + b c 1 − a xc .
For a = 1 and b = 2, perform the following tasks:
(a) Plot the pressure coefficient, Cp , over the plate.
(b) Calculate the position of separation, (x/c)separation .
(c) Plot the momentum thickness over the plate between x/c = 0 and (x/c)separation .
(d) Plot the displacement thickness over the plate between x/c = 0 and (x/c)separation .
(e) Plot the shear stress over the plate between x/c = 0 and (x/c)separation .
6.19 Derive (6.103) by using the definition of u + and the definition of τw in (6.54).
6.20 Plot the velocity profile according to the law of the wall. On the vertical axis
show u + , ranging between 0 and 25. On the horizontal axis, show y + , ranging
between 1 and 1,000.
6.21 To investigate the effect of the Coles wake paramter on the velocity profile,
plot (6.109) for Π = 0, 1, 2.5, 5, 10, 15. On the vertical axis show u + , ranging from
0 to 100. On the horizontal axis, show y/δ between 0 and 1.0.
6.22 Show that (6.112) can be derived from (6.109) by substituting y = δ in (6.109).
6.24 Assume we have a flat plate with an equivalent sand grain roughness of
0.020 mm. Assume that we have 0 m ISA conditions and a freestream velocity of
100 m/s.
(a) Calculate using (6.123) the friction coefficient (cf ) at x = 2 m.
(b) Calculate the wall stress, τw .
(c) Calculate the velocity gradient at the wall, (∂u/∂ y)w .
6.25 Consider the plate and flow conditions of Problem 6.24 and the method of
Example 6.6. Imagine we have 5 plates with lengths: 1, 3, 10, 20, and 50 m. Assume
that on each plate the turbulent boundary layer starts at x = 0.10 m. Calculate
(a) The Reynolds number, Rel , for each plate
(b) The friction coefficient, Cf , for each plate.
(c) Make a plot of the Reynolds number versus the friction coefficient.
6.26 Consider a flat plate of 10 m long. Assume that we have 0 m ISA conditions,
a freestream velocity of 100 m/s, and that the turbulent boundary layer starts at
x = 0.10 m. We would like to evaluate the effect of various sand grain sizes: k = 0,
0.025, 0.050, 0.075, 0.10 mm.
(a) Plot the value of the local friction coefficient, cf as a function of the position x
along the length of the plate for the five roughness heights.
(b) Plot the value of the total friction coefficient, Cf , as a function of the roughness
height, k.
Interference Drag
6.27 In the figures below you see pictures of two versions of a British Navy-fighter,
the Hawker, later Armstrong-Whitworth “Sea Hawk.” The later version, powered
by an engine with 5400 lb static thrust at sea level, has a fairing at the junction of
the horizontal and vertical tailplane. The earlier version, with a 5000 lb static thrust,
lacked this fairing. Explain the (aerodynamic) cause for the presence of the fairing
on the later version.
6.7 Summary 363
6.28 The picture below shows the Tupolev, Tu-154 in landing configuration. Visible
is the pointed cone at the intersection of the horizontal and vertical tailplane. Explain
the aerodynamic function of that cone using sketches of pressure distributions.
Photo A. Pingstone
6.29 The Lockheed P-38 Lightning (below) was in many ways innovative. An unde-
sirable characteristic of this aircraft was heavy horizontal tail buffeting at high speed.
Initially it was expected that it was classical flutter of the horizontal tail itself and
could be cured using mass balances. However, that did not solve the problem at all.
Only after a fairing was added at the root of the wing’s leading edge, where it meets
the fuselage, did the buffet disappear. What can be concluded about the precise cause
of this phenomenon?
Photo USAF
364 6 Aerodynamics of Non-lifting Bodies
References
26. Hoerner, S.F.: Fluid Dynamic Drag. Hoerner Fluid Dynamics (1965)
27. Johnson, H.A., Rubesin, M.W.: Aerodynamic heating and convective heat transfer—summary
of literature survey. Trans. ASME 71, 447–456 (1949)
28. Kegelman, J.T., Mueller, T.J.: Experimental studies of spontaneous and forced transition on an
axisymmetric body. AIAA J. 24(3), 397–403 (1986)
29. Kotsonis, M., Giepman, R., Hulshoff, S., Veldhuis, L.: Numerical study of the control of
Tollmien-Schlichting waves using plasma actuators. AIAA J. 51(10), 2353–2364 (2013).
doi:10.2514/1.J051766
30. Kreyszig, E.: Advanced Engineering Mathematics. Wiley, New York (1988)
31. Kutney, J.T., Piszkin, S.P.: Reduction of drag rise on the Convair 990 airplane. J. Aircr. 1(1),
8–12 (1964)
32. Kutney Sr, J.: The inside story of the Convair 990 the fastest subsonic airliner in the world. In:
Proceedings of AIAA Joint Propulsion Conference and Exhibit, July, pp. 1–18. AIAA-2007-
5338 (2007)
33. Kuz’min, A.G.: Boundary-Value Problems for Transonic Flow. Wiley, Chichester (2002)
34. Laufer, J., Vrebalovich, T.: Stability and transition of a supersonic laminar boundary layer on
an insulated flat plate. J. Fluid Mech. 9(2), 257–299 (1960). doi:10.1017/S0022112060001092
35. Lees, L.: The Stability of the Laminar Boundary Layer in a Compressible Fluid. NACA TN
1360. Langley Field, Virginia (1947)
36. Liepmann, H.W.: The interaction between boundary layer and shock waves in transonic flow.
J. Aeronaut. Sci. 13(12), 103–138 (1946)
37. Liepmann, H.W., Roshko, A.: Elements of Gas Dynamics. Wiley, New York (1957)
38. Little Jr., B.H.: Advantages and problems of large subsonic aircraft. In: AGARD-LS-37-High
Reynolds Number Subsonic Aerodynamics. AGARD (1970)
39. Lobb, R.K., Winkler, E.M., Persh, J.: Experimental investigation of turbulent boundary layers
in hypersonic flow. J. Aeronaut. Sci. 22(1), 1–9 (1955)
40. MacWilkinson, D.G., Blackerby, W.T., Paterson, J.H.: February correlation of full-scale drag
predictions with flight measurements of the C141A aircraft—phase II. Wind Tunnel Test,
Analysis, and Prediction Techniques, NASA CR-2333, Marietta (1974)
41. Moulden, T.H.: Fundamentals of Transonic Flow. Wiley, New York (1984)
42. Nikuradse, J.: Stromungsgesetze in rauhen Rohren. Forsch. Arb. Ing.-Wes, p. 631 (1933)
43. Obert, E.: Aerodynamic Design of Transport Aircraft. IOS Press, Delft (2009)
44. Pfenniger, W.: Laminar flow control—laminarization. In: Special Course on Concepts for Drag
Reduction, pp. 3.1–3.75. AGARD R-654 (1977)
45. Reynolds, O.: On the experimental investigation of the circumstances which determine whether
the motion of water shall be direct or sinuous, and the law of resistance in parallel channels.
Philos. Trans. R. Soc. Lond. Ser. A 174, 935–982 (1883)
46. Rubesin, M.W., Maydew, R.C., Varga, S.A.: An Analytical and Experimental Investigation of
the Skin Friction of the Turbulent Boundary Layer on a Flat Plate at Supersonic Speeds. NACA
TN 2305. Moffett Field, California (1951)
47. Saric, W.S., Reed, H.L., White, E.B.: Stability and transition of three-dimensional boundary
layers. Annu. Rev. Fluid Mech. 35(1), 413–440 (2003). doi:10.1146/annurev.fluid.35.101101.
161045
48. Schlichting, H., Gestern, K.: Boundary Layer Theory, 8th edn. Springer, Berlin (1999)
49. Schlichting, H., Truckenbrot, E.: Aerodynamik des Flugzeuges. Springer, Heidelberg (1960)
50. Shapiro, A.H.: The Dynamics and Thermodynamics of Compressible Fluid Flow. Ronald Press,
New York (1953)
51. Shapiro, A.H.: Pressure fields and fluid acceleration. In: Illustrated Experiments in Fluid
Mechanics. National Committee for Fluid Mechanics Films. Cambridge (1962)
52. Smith, A., Gamberoni, N.: Transition, Pressure Gradient, and Stability Theory. Douglas Aircraft
Company, Rept. ES 26388 (1956)
53. Sommer, S.C., Short, B.J.: Free-Flight Measurements of Turbulent-Boundary-Layer Skin Fric-
tion in the Presence of Severe Aerodynamic Heating at Mach Numbers from 2.8 to 7.0. NACA
TN 3391, Moffett Field (1955)
366 6 Aerodynamics of Non-lifting Bodies
54. Spalding, D.B.: A single formula for the law of the wall. J. Appl. Mech. 28, 455–457 (1961)
55. Tannehill, J.C., Anderson, D.A., Pletcher, R.H.: Computational Fluid Mechanics and Heat
Transfer, 2nd edn. Taylor & Francis, Philadelphia (1997)
56. Thwaites, B.: Approximate calculation of the laminar boundary layer. Aeronaut. Q. 1, 245–280
(1949)
57. Torenbeek, E., Wittenberg, H.: Generalized maximum specific range performance. J. Aircr.
20(7), 617–622 (1983)
58. Van Driest, E.R.: Turbulent boundary layer in compressible fluids. J. Aeronaut. Sci. 18(3),
145–160 (1951)
59. Van Driest, E.R.: Investigation of Laminar Boundary Layer in Compressible Fluids Using the
Crocco Method. NACA TN 2597, Washington (1952)
60. van Ingen, J.L.: A Suggested Semi-Empirical Method for the Calculation of the Boundary-
Layer Transition Region. TU Delft, Report VTH-74, Delft (1956)
61. van Ingen, J.L.: The eN method for transition prediction. Historical review of work at TU Delft.
In: 38th Fluid Dynamics Conference and Exhibit, pp. 1–49. Seattle (2008)
62. von Kármán, T.: Uber laminare und turbulent Reibung. Zeitschrift fur Angewandte Mathematik
und Mechanik 1(4), 233–252 (1921)
63. Wallace, L.E.: The Whitcomb area rule: NACA aerodynamics research and innovation. In:
Mack, P.E. (ed.) From Engineering Science to Big Science, Chap. 5. National Aeronautics and
Space Administration, History Office, Washington (1998)
64. Wazzan, A.R., Gazley Jr. C., Smith, A.: H-Rx method for predicting transition. J. Aircr. 19(6),
810–812 (1981)
65. Whitcomb, R.T.: A study of the zero-lift drag-rise characteristics of wing-body combinations
near the speed of sound. NACA TR 1273 (1956)
66. White, F.M.: The stability of laminar flows. In: Viscous Fluid Flow, 2 edn. New York (1991)
67. Whitfield, D.: Integral Solution of Compressible Turbulent Boundary Layers Using Improved
Velocity Profiles, Arnold Engineering Development Center AEDC-TR-78-42. Arnold Air Force
Station, Tennessee (1978)
68. Whitford, R.: Chapter 3: fuselage design. In: Design for Air Combat, pp. 148–160. Jane’s,
London (1987)
69. Wickens, R.H.: Aerodynamic design of low-drag fuselages. Can. Aeronaut. Space J. 36(4),
189–201 (1990)
Chapter 7
Airfoil Aerodynamics
7.1 Introduction
Experimental research into transonic aerodynamics was initiated before the first
world war. Although initially tailored towards munition, at the end of WWI transonic
effects were showing up in the increased tip speeds of propellers. It was found that
thicker sections at the tip yielded much less thrust and a much higher drag coefficient.
Soon it was found that when these sections were made thinner, these adverse effects
were not encountered [2]. Even though the physical understanding of transonic aero-
dynamics was still non-existing at the time, practical solutions to the adverse effects
were already conceived. Experimental research in transonic aerodynamics always
prevailed above the theoretical research because of the many challenges associated
with modeling of transonic flow.
With the introduction of the jet engine on transport aircraft in 1949 on the de
Havilland Comet, the operating Mach numbers for passenger aircraft increased
substantially. It was soon found that the profile shapes of the popular natural-laminar
flow airfoils produced significant wave drag and were therefore not suitable for this
new type of aircraft. The application of such airfoils required relatively large sweep
angles in order to increase the Mach number at which strong shock waves started to
cause excessive wave drag. Even though the swept wing solution worked well from
a cruise-performance point of view, it posed drawbacks in the low-speed regime as
well as in weight and aeroelastic performance. Better airfoil sections were therefore
pivotal in the development of contemporary wings on high-subsonic aircraft.
We have seen in Chap. 6 that any body submersed in a flow field produces a
pressure distribution. When the integral of the pressure distribution over the entire
body is nonzero, a resultant force is produced. We showed that for nonlifting bod-
ies this resulting force is comprised only of the drag force. Such conditions only
apply for (axis) symmetric bodies at zero angle of attack. Whenever either of these
requirements is violated a resultant force component perpendicular to the freestream
direction results. This discrete force (often referred to as the lift) has a point of appli-
cation somewhere on the body. The resultant force is now a combination of the lift
force and the drag force, as shown in Fig. 6.1. When this body is two-dimensional
it reduces to an airfoil. For many high-aspect ratio wings, the flow over the center
section of the wing can be treated as two-dimensional. Airfoil shape therefore plays
an important role in the lift and drag characteristics of the wing.
This chapter explains the typical transonic flow characteristics about two-
dimensional airfoil sections (commonly referred to as airfoils). We use a histor-
ical approach to take the reader on a quest to discover the relevant aerodynamic
effects related to transonic flow about airfoils. We demonstrate how experimental
researchers such as Pearcey and Whitcomb in the late 1940s and early 1950s came
up with airfoils designed for supercritical conditions. They showed how sections
could be shaped to operate at high Mach numbers and at high lift coefficients with-
out a large (wave) drag penalty. We present the key differences between laminar-flow
airfoils and supercritical airfoils. We also discover that transonic flow about airfoils
is not limited to high-subsonic freestream Mach numbers. On the contrary, the maxi-
mum lift coefficient is shown to be highly dependent on the freestream Mach number.
Especially multi-element airfoils are shown to yield lower maximum lift coefficients
due to transonic effects at Mach numbers as low as 0.2. Finally, we present the con-
cept of transonic buffeting: a high-frequency, self-sustained oscillation of the shock
wave that severely limits the maximum lift coefficient of an airplane at its operating
Mach number and beyond.
The pressure distribution about airfoils can be understood by following the same
rationale as in Sect. 6.2. Concave curvatures on a body cause a deceleration of the
flow, which are accompanied by an increase in static pressure. Convex curvatures
accelerate the flow and result in a decrease in static pressure (suction). The magnitude
7.2 Pressure Distribution About Airfoils 369
1 The program that is used for this evaluation is Xfoil version 6.94. XFOIL carries out a vortex-
panel analysis of subsonic isolated airfoils and has the option to include a boundary layer. More
information can be found on http://web.mit.edu/drela/Public/web/xfoil/.
370 7 Airfoil Aerodynamics
0.0 0.0
0.5 0.5
1.0 1.0
Fig. 7.2 Comparison of pressure distribution between two airfoils of different thickness as simu-
lated by Xfoil version 6.94. a NACA 0006. b NACA 0018
have been plotted in Fig. 7.2. The aerodynamic coefficients of each of the airfoils are
displayed next to the pressure distribution.
Let us start with the similarities between the two airfoils. First of all, they are
both symmetrical, which means their camber is identical. Classical airfoil theory
tells us that the lift curve slope for a symmetrical airfoil should equal dcl /dα = 2π.
Since we know that for this airfoil cl = 0 at α = 0 we expect cl to be close to 0.66
for α = 6◦ . Comparing that to the value shown in Fig. 7.2a, b we see that the lift
coefficient for both airfoils are close to this value, meaning that both airfoils generate
almost the same amount of lift. Secondly, we look at the moment coefficient about
the quarter-chord point. Assuming classical airfoil theory this value should be zero
for both airfoils. We see that this is indeed the case for the NACA 0018, but that the
NACA 0006 displays a small nose-up pitching moment about the quarter chord.
We have established that both profiles have a similar overall effect on the lift
and moment coefficient of the airfoil. Yet their pressure distributions are somewhat
different. The thinner profile of the NACA 0006 causes a much higher suction peak
over the leading edge of the airfoil. This large suction peak is followed by a rela-
tively sharp adverse pressure gradient that gradually lessens in magnitude further
downstream. The thick airfoil is much more blunt. The radius of curvature at the
nose of this airfoil is therefore larger resulting in a lower suction peak. The lower
suction peak is followed by a shallower adverse pressure gradient. The magnitude of
the suction peak can directly be translated to the magnitude of the supervelocities.
A high suction peak yields high supervelocities that can become supersonic at rela-
tively low freestream Mach numbers. At M = 0.2 the critical pressure coefficient is
−16.3 according to Eq. (3.42).
In the previous chapter we have seen that the value of the adverse pressure gradi-
ent in combination with Reynolds number determines whether or not the boundary
layer separates. Even though that information cannot be extracted from the analysis
carried out above, the shape of the pressure distribution can tell us qualitatively where
we should expect flow separation to start. For the thinner airfoil the large adverse
pressure gradient near the leading edge will become so large with increasing angle of
7.2 Pressure Distribution About Airfoils 371
attack that the boundary layer is likely to separate from the surface close to the lead-
ing edge (even though the boundary layer itself is relatively fresh). Leading edge stall
is therefore to be expected on this profile. The thicker airfoil shows a more shallow
adverse pressure gradient which is likely to separate the flow starting at the trailing
edge, where the boundary layer shape factor (H ) has its highest value [see (6.63)
on p. 305 for the definition of the shape factor]. Typically when H is between 2.4
and 2.6 boundary layer separation occurs. With increased angle of attack this point
is likely to move forward leading to a relatively gentle onset of stall. These deduc-
tions on the stall behavior of the airfoil are purely conjectural based on the pressure
distribution. However, experimental results from Ref. [1] support these conclusions.
We will elaborate more on the separation of the boundary layer in low-speed and
high-speed conditions in Sects. 7.5 and 7.6, respectively.
The effect of compressibility on the local pressure coefficient has been discussed
qualitatively in Sect. 3.2. The Prandtl-Glauert compressibility correction magnifies
the value of the pressure coefficient. It has been shown that this correction (or deriv-
atives thereof) can accurately predict the value of the pressure coefficient by sim-
ply multiplying this correction factor to the pressure coefficient obtained from an
incompressible-flow calculation (see Chap. 3). However, when Mach numbers are
increased beyond Mcrit the compressibility correction loses its accuracy. In Fig. 7.3
it is shown why this correction cannot be applied in transonic conditions. The same
airfoil (NACA 0012) is evaluated with two different prediction tools for the same
freestream conditions (a constant angle of attack and varying Mach number). The
prediction in Fig. 7.3a is based on a vortex panel method (Xfoil) with an embedded
compressibility correction. We see that the pressure distributions at low Mach num-
bers are merely amplified. The graphs in Fig. 7.3b are generated using an Euler solver
M = 0.74 M = 0.74
pressure coefficient, C p (~ )
M = 0.68
pressure coefficient, C p (~)
−1 −1 M = 0.68
M = 0.2 M = 0.2
−0.5 −0.5
0 0
0.5 0.5
NACA 0012 NACA 0012
1 1
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
position, x/c (~) position, x/c (~)
Fig. 7.3 Comparison between results from two prediction tools for the flow over a NACA 0012
airfoil with α = 2.1 and Rec = 10 million. a Vortex panel method + compressibility correction
(Xfoil v. 6.94). b Euler approximation (MES v. 3.04)
372 7 Airfoil Aerodynamics
(MSES2 ). The graphs of Fig. 7.3b show a distinct change in pressure distribution
when the Mach number is M = 0.68 or higher. We see how the adverse pressure gra-
dient of Fig. 7.3a is replaced by a more gradual increase in negative C p in Fig. 7.3b.
The steep drop in (negative) C p is indicative of a shock wave, which is also absent in
Fig. 7.3a. This comparison demonstrates the limitation of compressibility corrections
in the transonic regime. We will discuss this change in pressure distribution in more
detail in Sect. 7.4.
Having looked at the basic characteristics of airfoils we now wonder: what airfoil
performs best? Even though this answer might seem simple enough, there exists
no one-size-fits all airfoil that performs best for all applications. First of all, what
do we quantify as airfoil performance? This question cannot be answered quite
straightforwardly. There are a few key performance parameters of an airfoil: its
maximum lift coefficient, its lift-curve slope, its lift-do-drag ratio at the design lift
coefficient(s), and its moment coefficient at the design lift coefficient(s). The plural
form of the lift coefficients already indicates that airfoils are often designed for a range
of operating conditions (read: angle of attack and Mach number). Therefore, finding
the best airfoil might eventually result in an airfoil that finds the best compromise in
performance parameters over the specified range of operating conditions [45].
To maximize airplane range and endurance the lift-to-drag ratio of the airfoil is
of relatively high importance. The first efforts to maximize this ratio during the pre-
WWII period concentrated on creating large domains of laminar boundary-layer flow
over the upper and lower side of the airfoil. At the National Advisory Committee for
Aeronautics (NACA) a series was developed for precisely this reason: minimize the
drag by generating a favorable pressure coefficient over the upper and lower surface
such that the point of transition would be shifted more and more aft. This resulted
in the NACA 6-series airfoils that are described in detail in Ref. [1]. These airfoils
were designed with a specific lift coefficient in mind. In addition, the position of
minimum pressure was specified, indicating the chordwise location of the onset of
the adverse pressure gradient. We compare a NACA 6-series airfoil to a NACA 4-
series airfoil in Fig. 7.4. Two airfoils of identical thickness are evaluated3 at identical
lift coefficients, Mach numbers, and Reynolds number. Their pressure coefficient
and friction coefficient distribution are shown in Fig. 7.4a, b, respectively.
Let us evaluate the results presented in Fig. 7.4. First of all, we would like to
emphasize the similarity between the two airfoils: they have the same thickness and
2 MSES is a viscous-inviscid analysis program that couples the numerical solution of the Von Kár-
mán equation (6.89) in the boundary layer to the numerical solution of the steady state conservative
Euler equations (2.183) outside the boundary layer. More information can be found in Refs. [10, 15].
3 This prediction is carried out by Xfoil 6.94 using an e N method for transition prediction with
Ncrit = 10.
7.3 Laminar-Flow Airfoils 373
(a)
Pressure Coefficient, C p (~)
(b)
0 0
−3 −3
x 10 x 10
Friction Coefficient, c (~)
f
upper surface
4 4
upper surface
lower surface
2 2
lower surface
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
position, x/c (~) position, x/c (~)
Fig. 7.4 Predicted pressure and shape factor distribution for cl = 0.3, M = 0.2, and Rec = 6×106 .
a NACA 2412 at α = 0.4◦ . b NACA 661 212 at α = 0.9◦
both of them produce the same section lift coefficient. If each airfoil would be used
in a separate wing, the structural weight of each wing for a given planform shape
would be virtually identical. Even though the integrated difference of the pressure
coefficient between upper and lower surface is identical for both wings, their pressure
distributions are quite different. The thickest point of the 2412 airfoil is located
(by default) at 30 % chord. On the upper surface we see the start of the adverse
pressure gradient at approximately x/c = 0.2. Transition to a turbulent boundary
layer occurs around x/c = 0.45. On the lower surface the adverse pressure gradient
is slightly steeper and transition occurs closer to the leading edge than on the upper
surface. Notice that in this simulation transition can be seen in the jump in the friction
coefficient. This indicates where the boundary layer has transitioned from laminar to
turbulent. In this particular simulation the step change in displacement thickness that
is caused at the transition point can also be seen in the pressure distribution curve by
means of a little bump.
The thickest point of the 661 212 has been shifted beyond the 50 % chord. This
creates an almost flat plateau in the pressure distribution over the suction side of the
airfoil up the 60 % chord. On the lower surface there is a favorable pressure gradient
pressure up to x/c = 0.6. Transition occurs on both sides beyond the 60 % chord.
Notice the effect of the favorable pressure gradient on the lower side: transition is
shifted more aft. We know that the average friction coefficient, C f is correlated to
the integral of the local friction coefficient, c f , over the chord length. The total area
under the two lines in the friction-coefficient diagram are therefore representative
for the total amount of friction drag of the profile. Based on that knowledge we can
directly see that the total friction drag produced by the 661 212 is considerably lower
374 7 Airfoil Aerodynamics
.012
.008
smooth
.004
drag bucket
0
-1.6 -1.2 -0.8 -0.4 0 0.4 0.8 1.2 1.6
Section lift coefficient, cl (~)
than the friction drag produced by the 2412. The predicted drag coefficient is 0.0031
or 31 drag counts4 for the 661 212 against 51 counts for the 2412.
Comparison with experimental data of Ref. [1] shows that drag coefficient is
somewhat underestimated. In Fig. 7.5 the experimental wind tunnel measurements
are presented for a Reynolds number of 6 million. It can be seen that at cl = 0.3 the
drag coefficients is around 34 counts for the 661 212 versus 63 counts for the 2412.
We notice a 45 % decrement in drag when switching from a 2412 to a 661 212. The
lift-to-drag ratio of the airfoil is increased with more than 80 % from 48 to 88.
As can be seen from Fig. 7.5, the natural-laminar-flow (NLF) airfoil has a limited
range of lift coefficients where the friction drag is low. As a matter of fact, the
designation of the airfoil already shows this:
The design lift coefficient is 0.2 and we can expect a low drag coefficient (also known
as the drag bucket) between cl = 0.1 and cl = 0.3. Outside of the drag bucket the drag
of the NLF airfoil is either close to, or higher than the drag coefficient of the NACA
2412. In addition, when the skin is roughened (using NACA standard roughness) the
drag coefficient at cl = 0.3 increases to 95 cts while the drag of the 2412 increases to
100 cts, making the difference between the two airfoils very small. This illustrates the
sensitivity of both profiles to skin roughness. In practice, roughness can stem from
irregularities on the skin, rivets, or waviness in the skin surface. In addition, steps
When the thrust of engines increased with the commercial introduction of the jet
engine in the 1950s, it quickly became clear that close to the speed of sound the
drag of the airplane increased exponentially (drag divergence). The Mach number
at which drag divergence starts is often defined as the Mach number at which the
numerical value of the slope of the curve of cd versus M is 0.10 [19]:
∂cd
Mdd = = 0.10 (7.1)
∂M cl =constant
This exponential increase in drag was attributed to the formation of strong shock
waves on the wing surface in combination with separation at the shock foot. It
was found that the natural laminar flow airfoils produced a significant amount of
wave drag whenever they were exposed to Mach numbers beyond the critical Mach
number (see Fig. 7.6a). This limited the maximum Mach number of early jet fighters
0 0
c l = 0.7
c l = 0.5
c l = 0.3
0.5 0.5
NACA 661 212 NASA SC(2) 0412
1 1
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
position, x/c (~) position, x/c (~)
Fig. 7.6 Predicted pressure distribution for cl = 0.3, cl = 0.5, and cl = 0.7 at M = 0.75,
Rec = 20 × 106 . a NACA 661 212. b NASA SC(2) 0412
376 7 Airfoil Aerodynamics
that used NACA 6-series airfoil sections. One effort to tackle this problem was
to investigate whether airfoil shapes could be developed that would have a higher
critical Mach number in combination with lower wave drag in supercritical (i.e.
Mlocal > 1) conditions. In the United Kingdom it was H.H. Pearcey that investigated
new airfoil shapes that would allow for supersonic flow on the upper surface of the
airfoil without a strong terminating shock wave [35]. In the Unites States under the
supervision of R.T. Whitcomb a new family of supercritical airfoils was developed
specifically designed to produce low wave drag at high Mach numbers [17]. The
pressure distribution of one of these airfoils is shown in Fig. 7.6b.
If we compare the pressure distributions between the NLF airfoil and the super-
critical (SC) airfoil of identical thickness, we see a large difference on both the upper
and lower surface. In the graph for the NLF airfoil we already recognize the presence
of a shock wave at cl = 0.3 (remember that this cl is still the low-speed design range
for this airfoil). We recognize the shock wave by the sharp increase in C p close to
x/c = 0.6. If we look at the SC airfoil we see that the suction over the upper surface
is much lower at cl = 0.3. Most of the lift for this airfoil stems from the difference in
pressure on the aft part of the airfoil. The concave shape of the lower surface causes
an increase in pressure on the lower surface which makes a large contribution to the
total lift coefficient. We call this aft loading and it is one of the means to lower the
required supervelocities on the upper surface of the airfoil for a given lift coefficient
(see Problem 7.2). When the lift coefficient is increased to 0.5 the shock wave on
the NLF airfoil increases in strength (larger pressure rise). On the SC airfoil we also
see the formation of a shock wave around x/c = 0.35. However, the latter shock
wave is much weaker than the one over the NLF airfoil as can be observed from the
smaller pressure increase. Finally, at cl = 0.7 the shock over the NLF airfoil has
shifted forward, while its strength has remained approximately the same. However,
the trailing-edge C p has now become substantially more negative, indicating that
the boundary layer has separated from the trailing edge. On the SC airfoil we see
that the shock has moved aft and has increased in strength. There is no indication of
boundary layer separation.
In Fig. 7.7 the corresponding curves for the drag coefficients are shown. Notice
that the drag coefficient (on the vertical axis) has been expressed in terms of drag
counts. The drag coefficient consists of two fundamental components: the friction
drag coefficient, cd f , and the pressure drag coefficient, cd p . In this analysis the pres-
sure drag coefficient has been found by integrating the pressure over the section and
projecting the resulting force vector onto the direction of the flow.5 If we compare
the drag coefficient development between the two airfoils, we immediately see that
the NLF airfoil shows a sharp increase in drag coefficient between cl = 0.3 and
cl = 0.5. Since the friction drag is almost constant, this increase in drag is almost
solely due to the formation of a strong shock wave on the upper surface. The SC
airfoil shows almost a constant value for the total drag coefficient over this range. At
cl = 0.5 the drag of the SC airfoil is therefore 37 % lower than for the NLF airfoil.
At cl = 0.7 the shock has become significantly stronger, which results in an increase
5 This calculation procedure for obtaining the pressure drag is often called a “near-field analysis”.
7.4 Supercritical Airfoils 377
cd c
p dp
120 120
d
d
100 100
80 80
60 60
40 40
20 20
0 0
0.3 0.4 0.5 0.6 0.7 0.3 0.4 0.5 0.6 0.7
lift coefficient, c (~) lift coefficient, c (~)
l l
Fig. 7.7 Predicted drag coefficients at M = 0.75, Rec = 20 × 106 . Note cd p indicates pressure
drag, while cd f indicates friction drag. a NACA 661 212. b NASA SC(2) 0412
in pressure drag. However, the flow remains attached and the pressure drag does not
rise as fast as for the NLF airfoil.
For a given airfoil shape, the thickness ratio (t/c) and lift coefficient are often
the most important parameters that influence the drag divergence Mach number.
Figure 7.8 shows how for supercritical and NACA airfoils the thickness ratio influ-
ences the drag-divergence Mach number. According to this graph, for supercritical
airfoils, the drag-divergence Mach number decreases linearly according to the fol-
lowing statistical relation:
t
Mdd = 0.92 − 1.16 for cl = 0.5 (7.2)
c
0.76
Conventional 66-210
Airfoils 64A211
0.72
NA Rockwell 64A412
NASA
0.68 NACA
65-215
Northrop
0.64
The decrease in drag divergence Mach number can be explained by the fact that the
thicker airfoils create higher supervelocities over the upper and lower surface than
a thin airfoil. A thicker airfoil will therefore encounter critical conditions at a lower
free stream Mach number than a thin airfoil. In general, this also results in an earlier
onset of shock-induced boundary-layer separation and therefore drag divergence.
In Fig. 7.9 we show an example of the effect of lift coefficient on the drag diver-
gence Mach number. The drag divergence Mach number in this graph is defined as
the point on the line where ∂cd /∂ M = 0.10. We can observe that with increasing lift
coefficient the drag divergence Mach number decreases. This can be explained by
the fact that the increase in lift coefficient is a result of higher supervelocities on the
suction side of the airfoil, which causes an earlier formation of the shock wave and
associated shock-induced separation. A rudimentary equation to compute the drag
divergence Mach number that includes both the thickness-to-chord ratio and the lift
coefficient is given by Korn:
where κ is a technology factor that amounts to 0.95 for supercritical airfoil sections.
Torenbeek [44] derived a modified version of Korn’s equation based on empirical
data of second generation supercritical airfoils [17]:
It should be emphasized that a different airfoil shape can highly influence the depen-
dency of the drag divergence Mach number to the lift coefficient. It was already
shown by Göthert in 1944 that careful modification of a NACA 0012 could yield a
constant drag divergence Mach number of 0.78 for a lift coefficient ranging from 0
to 0.4 [16]. Blackerby and Johnson show in Ref. [7] that changing the forward 12 %
of the airfoil can have a profound effect on the drag divergence Mach number.
.032
cl = 0.50
.030
.028 cl = 0.45
.026
.70 .72 .74 .76 .78 .80
Mach number, M (~)
7.4 Supercritical Airfoils 379
What we now know as a supercritical airfoil was still to be discovered in the early
1960s. In a 1963 paper on the development of low-drag SC airfoils, Pearcey states
the questions that needed to be answered at the time [35]: “Just how much can the
velocity be allowed to exceed sonic locally without incurring strong, drag-producing
shock waves, and how does this in turn depend on the velocity distribution and
section shape?” Keeping in mind that numerical tools such as those employed in
this chapter were unavailable at this time and that the analytical methods to predict
subsonic pressure distributions were inadequate for transonic flow, these questions
could only be answered through a thorough experimental investigation. The goal of
this investigation was to find a section shape that would allow for supersonic flow to
develop over the top surface but without the drag-producing shock wave. One of the
key findings of this investigation was a clear elaboration on the effect of the expansion
waves generated at the leading edge of the airfoil. We will follow this elaboration
(Ref. [35]) to demonstrate the complex interaction of expansion and compression
waves in the supersonic domain of the flow.
We first consider a generic airfoil in high subsonic conditions (Fig. 7.10). The
flow over the airfoil accelerates to beyond M = 1, which introduces a supersonic
flow domain on the upper side of the airfoil. The line separating this supersonic
bubble from the subsonic domain is called the sonic line. This sonic boundary forms
a constant pressure surface and attaches to the airfoil close to the leading edge (at
the point where the flow accelerates locally beyond M = 1). We assume that right
behind this line an expansion wave (left-running Mach wave) is generated by a small
disturbance. This wave travels through the supersonic flow field until it reaches
the sonic line. Since expansion waves cannot travel in the subsonic flow domain
outside the sonic line, it is reflected towards the surface. Because the sonic line is
a constant-pressure surface the expansion wave is reflected as a compression wave
(reflection in unlike sense). The supersonic flow in between the expansion wave
and the compression wave experiences an increment in Mach number (ΔM) due to
Compression
Shock Wave
Expansion
Sonic
boundary
M - M M
Fig. 7.10 Notional reflection of a single expansion wave emanating close to the leading edge (after
Ref. [35])
380 7 Airfoil Aerodynamics
the expansion wave. The flow behind the reflected compression wave experiences a
decrement in Mach number of equal strength. The resulting Mach number behind
this compression wave would therefore be equal to the Mach number in front of the
first expansion wave.
When the compression wave reaches the airfoil surface it is being reflected. If the
surface is flat the compression wave is reflected as a compression wave (reflection in
like sense). However, if the surface has sufficient convex curvature, the flow tangency
condition requires the streamline to bend towards the surface and the incident com-
pression wave is reflected as an expansion wave. Alternatively, the surface can have
exactly the right amount of convex curvature such that there is no reflection from the
wall. To decelerate the flow it is beneficial to reduce the convex curvature such that
the reflected wave is also a compression wave. This has been implicitly assumed in
the notional sketch of Fig. 7.10. The flow behind this second compression wave has
a lower Mach number than the flow in front of the first expansion wave (−ΔM in
Fig. 7.10). The compression waves therefore aid in the reduction of the Mach number
further downstream over the airfoil. Because this compression process is isentropic
it is termed isentropic recompression.
In reality the single expansion wave should be replaced by an infinite number of
expansion waves that emanate from the convex surface close to the leading edge and
possibly also further downstream on the airfoil. This expansion fan is reflected from
the sonic boundary as a compression fan. The expansion waves and compression
waves are intersecting. This is visualized in Fig. 7.11 where the expansion waves are
shown as characteristics. We see the expansion characteristics as dashed lines and
the compression characteristics as solid lines. In the lower graph the Mach number
distribution is shown, where ω is the Prandtl-Meyer angle corresponding to the local
Mach number of the flow. In this graph the upper dashed line shows the Mach
number distribution if only the expansion waves would be taken into account. Due to
the reflection of the compression waves from the sonic line and subsequently from
the crest, it can be seen how the Mach number is lowered.
In Ref. [39] it is described that the formation of a shock is the result of a coalescence
of the recompression characteristics. If the reflected compression waves coalesce,
they merge to form a shock wave. If even a short segment of the profile is straight-
sided (instead of convex) a shock will be formed as the result of a coalescence of
the compression waves reflected from this segment. In Fig. 7.12 we see two airfoils
and their pattern of characteristics. The airfoil in Fig. 7.12a has been designed to be
shock free. In other words, the recompression waves emanating from the sonic line
do not coalesce in the physical plane. This can only happen if the shape of the sonic
line and the flow deflection are carefully tailored to each other [39]. If only a small
disturbance travels downstream along the reflected Mach waves it will result in their
coalescence: a shock (Fig. 7.12b). This shows how carefully the airfoil geometry
must be arranged in order to obtain shock-free supersonic flow.
Pearcey demonstrated experimentally that even subsonic flows that locally accel-
erate up to M = 1.4 can be decelerated isentropically without the formation of a
shock. To achieve these properties the airfoil had to have the following properties [35]:
7.4 Supercritical Airfoils 381
Sonic
boundary
} Subsequent reflections
} Limiting characteristic
and reflection
Free-stream
direction P
Crest
Surface distribution
(Mcrest -1)
1.0
Fig. 7.11 Sketch of Mach-wave pattern in a region of supersonic flow over an airfoil and the Mach
number distribution resulting from the supersonic position of the first simple wave, ω1 , and the
compressive effect, ω2 (after Ref. [33])
(a) (b)
Fig. 7.12 Pattern of characteristic waves over two airfoils (after Ref. [39]). a Shock-free. b Coa-
lescence causes shock
(a) a sharp suction peak needs to be present at the leading edge during subsonic flow
conditions at a particular angle of attack
(b) behind the suction peak rapid deceleration has to occur in subsonic flow, and
(c) behind the leading edge the curvature distribution of the upper surface has to be
such that the expansion waves in the supersonic region, the sonic line, and the
reflected compression waves are formed behind the suction peak.
382 7 Airfoil Aerodynamics
(a) (b)
CP CP
-1.0 -1.0
CP*
CP*
-0.5 -0.5
inviscid theory
0 0
0.5 0.5
1.0 1.0
x/c x/c
Fig. 7.13 Pressure distribution about a two different shock-free airfoils. a M = 0.709, α = 3.6
(modified from Ref. [8]). b M = 0.765, cl = 0.58, Rec = 21 × 106 (after Ref. [23])
Due to the sharp suction peak and high adverse pressure gradient in the subsonic
pressure distribution these airfoils are often termed ‘peaky’ airfoils. Pearcey showed
that the significant amount of isentropic recompression could result in shock-free air-
foils or airfoils with a weak normal shock wave terminating the supersonic domain.
Other researchers (e.g. [8, 23, 39]) demonstrated the potential and limitations of
shock-free supercritical airfoils. Figure 7.13 shows the pressure distribution for two
of such airfoils. In both graphs the critical C p has been indicated with a dashed line.
We see that both airfoils show a smaller adverse pressure gradient allowing for a
larger supersonic bubble. According to Ref. [32] one can expect that for given values
of lift coefficient and thickness ratio, the design Mach number for shock-free airfoils
will be generally somewhat lower for the peaky designs than for the “nonpeaky”
designs of Fig. 7.13. However, this does not necessarily mean that these nonpeaky
airfoils also have a higher drag divergence Mach number. On the contrary, they have
been shown to be more sensitive to off-design conditions than peaky airfoils.
We have now shown that shock-free airfoils have the possibility of producing
relatively high lift coefficients (around cl = 0.6) without producing any wave drag.
In theory this is an ideal solution. However, in practice these airfoils are hardly used
on high-subsonic transport aircraft. One of the main reasons for this is their sensitivity
to changes in the flow condition. These could stem from a change in Mach number,
angle of attack, or even Reynolds number. This is illustrated for the Garabedian-Korn
airfoil of Fig. 7.13b in Ref. [21] where the airfoil is predicted6 to be shock free for
M = 0.751 and cl = 0.625 while for M = 0.750 and cl = 0.629 it is predicted to
have two shocks. In reality the displacement effect of the boundary layer also plays
6 The prediction was made by assuming the flow to be inviscid (Euler solution).
7.4 Supercritical Airfoils 383
an important role in whether the shock-free condition can be attained. Given the fact
that the displacement thickness is dependent on the Reynolds number it is very well
possible that an airfoil shows to be shock-free in the wind tunnel while having a
weak shock during flight for the same Mach number and lift coefficient.
P
sonic line
Q
G
rounded nose
relatively flat central upper surface boundary layer thickening
B F
A
NASA SC(2) 0412
the (flap) structure. Figure 7.14 shows the geometric characteristics of a 12 % thick
supercritical airfoil.
In addition to the geometric characteristics, Fig. 7.14 also shows key aerodynamic
properties of a supercritical airfoil. A notional boundary layer thickness has been
drawn for reference. We see that the sonic line, which separates the two flow domains,
forms a continuous boundary around the supersonic bubble. We see that the shock
wave is slightly curved. The tip of the shock (point P) is in the supersonic domain.
The curved part of the shock (oblique with respect to the local flow direction) causes
the flow decelerate to a Mach number higher than unity, causing a small supersonic
region behind this part of the shock. The flow decelerates isentropically to subsonic
conditions in this region. The sonic line therefore attaches to the shock at point Q.
Below point Q the shock is normal to the local flow direction and the flow decelerates
to subsonic conditions. If we move towards the foot of the shock (point F) we see
that the shock is formed by the coalescence of compression waves that smear the
pressure increase over a finite interval. Due to this pressure increase, the boundary
layer thickens (as is explained in Sect. 6.4.3).
Most modern jet transports employ supercritical airfoils or use some form of super-
critical geometry in their airfoil design. For example, all the Airbus aircraft employ
supercritical airfoils, as do the latest Boeing aircraft (777, 787). Other Boeing air-
craft (e.g. 737-700 and 767) also include airfoils that are modified using supercritical
airfoil technology [4].
Example 7.1 Consider an aircraft that performs an accelerated climb from M = 0.50
to M = 0.77 while the lift coefficient of the aircraft remains constant. At 70 % semi-
span the wing has a DSMA-523a supercritical airfoil with a thickness-to-chord ratio
7.4 Supercritical Airfoils 385
DSMA−523a
Solution:
First we need to chose a suitable solver for this question. Since we know that the
flow about a supercritical airfoil in high-subsonic conditions will be mixed (both
subsonic and supersonic domains) we choose a numerical approximation of the Euler
equations. In this case we use a finite-volume approximation of the Euler equations
on an unstructured grid. A detailed description of this approximation can be found
in Ref. [9]. Since these calculations are based on the assumption that the flow is
inviscid, we will not be able to deduce any viscous effects such as separation. In
addition, the exact shock position is most likely to be somewhat inaccurate due to
the lacking shock wave-boundary layer interaction (SWBLI). However, it does give
a good insight into the development of the shock wave and its position and gives
a reasonable prediction for high Reynolds-number flows, such as found over high-
subsonic transport aircraft. It is therefore found to be a suitable tool to determine
the pressure distributions about this airfoil. The results for the prescribed conditions
are found in Fig. 7.15. Note that these plots show the input values above the graph:
Mach number and angle of attack. The output in terms of aerodynamic coefficients is
displayed within the plot. To keep the lift coefficient constant at each Mach number
we performed several iterations where we changed the angle of attack, until the result
showed a lift coefficient close to the target of cl = 0.85.
−1 −1 −1
CP*
−0.5 −0.5 −0.5 CP *
0 0 0
c = 0.854 c = 0.852 c = 0.857
l l l
0.5
c = 0.00336 0.5
c = 0.00638 0.5
c = 0.00384
d d d
cm, c/4 = −0.131 cm, c/4 = −0.163 cm, c/4 = −0.203
1 1 1
0 0.5 1 0 0.5 1 0 0.5 1
chordwise position, x/c (~) chordwise position, x/c (~) chordwise position, x/c (~)
Fig. 7.15 Development of pressure distribution with constant lift coefficient and increasing Mach
number over a DSMA-523a airfoil. Note this is the result of an inviscid calculation, hence cd = cd p
386 7 Airfoil Aerodynamics
Let us analyze the changing pressure distribution of Example 7.1. First we consider
the effect of the increasing Mach number on the pressure distribution over the airfoil
while the lift coefficient remains constant (see Fig. 7.15). In the left graph we see
the expected peaky behavior of the supercritical airfoil at low-speed conditions. The
suction peak at the leading edge is followed by a relatively sharp adverse pressure
gradient, due to the lack of curvature on the top side of the airfoil. We also identify
the aft loading, which is caused by the concave shape of the lower side of the airfoil
over the aft 40 %c. When the freestream Mach number goes beyond the critical
Mach number, a plateau is formed in the pressure distribution. This represents the
supersonic domain of the flow and is terminated by a sharp pressure increase: the
shock wave. As we increase the freestream Mach number even more the shock
wave starts to move backwards towards the trailing edge, while the crest of the
pressure distribution lowers. Notice that by only increasing the Mach number by
0.05, we now have more than 85 % of the top surface emersed in supersonic flow. Also
note that the pressure distribution over the bottom side of the airfoil hardly changes.
The flow is fully subsonic there and we only distinguish a magnification of the
pressure coefficient between M = 0.5 and M = 0.73 due to compressibility effects.
Furthermore, we see that the drag coefficient decreases when the Mach number
increases from 0.73 to 0.77. This large decrease is rather unrealistic and is attributed
to the absence of viscosity in the flow solver. Finally, we would like to point out
the rapid change in moment coefficient about the quarter chord point. In subsonic
conditions we would expect the moment coefficient to stay constant about this point.
It is clear from the values for cm, c/4 that the nose-down pitching moment increases
substantially. This phenomenon is often referred to as Mach tuck or tuck under. To
compensate for this and balance the aircraft around its center-of-gravity, we need to
apply a larger nose-up pitching moment. For a conventionally configured airplane
this is achieved by providing more downforce on the horizontal stabilizer, which
increases the trim drag.
Solution:
Having done the previous example (Example 7.1), we use the same numerical approx-
imation of the steady Euler equations, keeping in mind that we assume the flow to be
inviscid. We evaluate the airfoil at the prescribed lift coefficients. However, since the
lift coefficient is an output of the calculation (by summing the pressure difference
7.4 Supercritical Airfoils 387
−1 −1 −1
*
Cp
−0.5 −0.5 −0.5
0 0 0
cl = 0.498 cl = 0.889 cl = 1.08
0.5
c = 0.0018 c = 0.00754 c = 0.0163
d 0.5 d 0.5 d
cm, c/4 = −0.178 cm, c/4 = −0.161 cm, c/4 = −0.163
1 1 1
0 0.5 1 0 0.5 1 0 0.5 1
chordwise position, x/c (~) chordwise position, x/c (~) chordwise position, x/c (~)
Fig. 7.16 Development of pressure distribution with increasing lift coefficient and constant Mach
number over a DSMA-523a airfoil. Note this is the result of an inviscid calculation, hence cd = cd p
between the suction and pressure surface) we need to do this iteratively by varying
the angle of attack in the input file. The result is displayed in Fig. 7.16.
In the example above we can distinguish the effect that changing lift coefficient
has on the pressure distribution over the wing. We see that for a relatively low lift
coefficient, the main contribution to the lift coefficient comes from the aft loading of
the airfoil, which, in turn, produces quite a large nose-down pitching moment about
the quarter-chord point. When the lift coefficient increases (through the change in
angle of attack), we see that a plateau is formed in the pressure distribution. This
plateau rises and becomes longer when the lift coefficient is even further increased. At
the same time we see that the shock wave becomes stronger, yielding more wave drag
with higher lift coefficient. A quick analysis of the results shows that the average
lift curve slope of this airfoil at M = 0.73 is 14.1(1/rad), which is more than
twice the value of what we expect from this airfoil at low speeds, which is around
2π = 6.28 (1/rad). This demonstrates that in the present conditions this airfoil is
much more responsive to changes in angle of attack, than at low speeds. In practice
this means that atmospheric gusts can have a relatively large impact on the pressure
distribution over the wing and can hence introduce large fluctuations in wing lift. In
the cabin we feel this as if the aircraft is driving on a road dotted with potholes. We
will see in Chap. 8 that the introduction of sweep reduces the lift-curve slope of the
aircraft and, therefore, partially compensates for this large sensitivity.
The results shown in Examples 7.1 and 7.2 are predictions of an inviscid Euler
code. In Fig. 7.17 a Mach and cl sweep is shown about a supercritical airfoil. Both
experimental results obtained in a wind tunnel (Rec = 14.5 × 106 ) and numeri-
cal results obtained by an Euler code with boundary layer approximation (viscous-
inviscid prediction) are shown. If we look at the wind tunnel test results we see
a similar behavior as in the results of the aforementioned examples. With increas-
ing Mach number and constant lift coefficient the shock wave moves aft while the
minimum pressure coefficient decreases somewhat. When the Mach number is kept
388 7 Airfoil Aerodynamics
inviscid calculation:
-2.0 -2.0 -2.0 const.
Cp Cp Cp const. CL
-1.5 M =0.499 -1.5 M =0.690 -1.5 M =0.732
CL =0.770 CL =0.927 CL =0.898
-1.0 -1.0 -1.0
0 0 0
20 40 60 80 20 40 60 80 20 40 60 80
x/c (%) x/c (%)
0.5 0.5 0.5
-1.0
1.0 1.0 1.0
M =0.732
-0.5 CL =0.645
-6 0.5
Sym Ref. RC x 10
Lift -1.0
Windtunnel test data 14.5 coefficient 1.0
Douglas / Garabedian 14.5 sweep -0.5
Cp 0
20 40 60 80
DSMA 671 x/c (%)
0.5
M =0.732
1.0 CL=0.499
Fig. 7.17 Mach number and cl sweep for a 14 % thick supercritical airfoil (after Ref. [22])
constant while the lift coefficient is increased, a shock wave starts to form. The
plateau in the lift distribution rises with increasing lift coefficient, while the shock
position remains almost constant.
Finally, we note that the prediction of the pressure distribution of the viscous-
inviscid flow solver is quite accurate. Both position of the shock and the supercritical
C p is accurately predicted. We see in the top right graph of Fig. 7.17 an Euler pre-
diction (inviscid) for two conditions: constant α and constant cl . This means that the
equations are being solved respectively for the same α and cl as in the wind tunnel
experiment. We see that the Euler prediction is unable to properly capture the position
of the shock wave. Due to the absence of the displacement thickness of the boundary
layer, the higher effective curvature of the upper side of the airfoil generates a higher
supercritical Mach number which pushes the shock wave more aft. For the same
lift coefficient (constant cl ) and higher Mach number the airfoil is positioned under
a lower angle of attack, resulting in an underprediction of the supercritical Mach
number. For the evaluation at constant angle of attack, the prediction is even more
off: the supercritical speeds are over predicted and the shock wave is too far aft. This
combination results in an over estimation of the lift coefficient. The large discrepancy
between the Euler prediction and the measured data is evidence that the inclusion
of the boundary layer is important to accurately capture the pressure distribution
7.4 Supercritical Airfoils 389
(a) .8
0
.6 -2°
-1°
0°
1°
Section lift coefficient, cl (~)
.4 2°
4°
6°
8°
.2 10°
12°
-.2
-.4
NACA
-.6
.2 .3 .4 .5 .6 .7 .8 .9 1.0
.06 0
-2°
-1°
0°
.04 1°
2°
4°
6°
.02
NACA
0
.2 .3 .4 .5 .6 .7 .8 .9 1.0
Mach number, M (~)
Fig. 7.18 The measured variation of section properties with Mach number of a NACA 64A010
airfoil for various angles of attack [18]. a Lift coefficient. b Drag coefficient
about supercritical airfoils. The results of Examples 7.1 and 7.2 can therefore not be
perceived as an accurate prediction, but merely serve as examples.
The change in lift and drag coefficient with Mach number is shown in Fig. 7.18
for a modified NACA 6-series airfoil of 10 % thickness. In Fig. 7.18a we observe
that initially the low-subsonic lift coefficient is amplified by the compressibility
effect, in accordance with the compressibility corrections that were introduced in
Chap. 3. However, at higher Mach numbers, the formation of shock waves break the
amplification trend and with ever higher Mach number, shock-induced separation
(see also Sect. 7.6) causes the lift coefficient to drop. At the same time, the drag
coefficient due to the presence of the shock wave and shock-induced separation starts
to rise. With increased angle of attack, the onset of the drag rise occurs at ever lower
Mach numbers, showing that the drag-divergence Mach number is a function of both
390 7 Airfoil Aerodynamics
Mach number and angle of attack. This behavior is not unique to this specific airfoil
but applies to most supercritical airfoils. The higher the required lift coefficient, the
higher the angle of attack needs to be to achieve that. With higher angle of attack,
higher supervelocities on the upper surface of the airfoil are encountered, resulting
in an earlier onset of shock waves and shock-induced separation.
Whereas on supercritical airfoils the supersonic region on the upper surface of the
airfoil is terminated with a weak shock, so-called ‘sonic rooftop’ airfoils lack a
supersonic region all together. The ‘sonic rooftop’ refers to the sonic flow over the
first part of the airfoil, similar to the supersonic region on a supercritical airfoil. A
sonic rooftop airfoil is designed for transonic speeds, but such that the flow at the
design condition just reaches sonic speeds over the flat region of the upper surface.
This implies that a terminating shock wave is absent and wave drag amounts to zero
in the design condition. The drag divergence Mach number for sonic rooftop airfoils
is higher than for supercritical airfoils. However, this does come at an expense.
Sonic rooftop wings are thinner and, consequently, hold less fuel. They also require
a heavier structure to transfer the bending moment in the wing. Furthermore, the
nose radius is smaller which decreases their maximum lift coefficient at low speeds.
In Fig. 7.19 we compare two airfoils of similar but different geometry. One is
designed as an SC airfoil with a design lift coefficient of 0.635, while the second
airfoil is a sonic rooftop airfoil with a design lift coefficient of 0.500. Both airfoils
have a design Mach number of 0.72. Their respective pressure distributions are shown
in Fig. 7.19a, b. The sonic rooftop profile of the pressure distribution is clearly visible
in Fig. 7.19b. If we compare the two drag components of each airfoil we see that the
friction drag is predicted to be the same, while pressure drag of the SC airfoil is
10 drag counts higher than for the sonic rooftop airfoil. The lift to drag ratio of the
SC airfoil in the design condition is cl /cd = 78, while for the sonic rooftop airfoil
this value is estimated to be cl /cd = 70. Looking at Fig. 7.19c we see that the drag
divergence Mach number is higher for the sonic rooftop airfoil than for the SC airfoil.
Furthermore, we distinguish in this figure the relative geometry of both airfoils which
corresponds to the description of the previous paragraph.
One important characteristic of the pressure distribution over the upper surface of
a supercritical airfoil is the strong adverse pressure gradient that exists over the aft
part of the airfoil. We know from Chap. 6 that the boundary layer is more susceptible
to separate under a strong adverse pressure gradient. Therefore transonic airfoils
are often designed to reduce this adverse pressure gradient by incorporating two
7.4 Supercritical Airfoils 391
-1 -1
CP, crit Cp, crit
CP Cp
0 0
1 1
(c) c sonic rooftop: Mdesign = .72 cl, design = 0.5 t/c = 0.110
d
.012 supercritical: Mdesign = .72 cl, design = .635 t/c = 0.126
.010
Mdd
.008 supercritical
sonic rooftop
.6 .7 .75 M
Fig. 7.19 Comparison between a supercritical and a sonic rooftop airfoil of similar but different
geometry (after Ref. [33]). Note cd f indicates friction drag while cd p indicates pressure drag. a
Supercritical airfoil. b Sonic rooftop airfoil. c Drag coefficient versus Mach number
Fig. 7.20 Trailing edge shapes relating to the Kutta condition (after Ref. [3])
geometric measures: a cusp trailing edge, and a finite thickness of the trailing edge.
We will discuss these two measures subsequently.
In Fig. 7.20 we show two generic trailing edges. The first trailing edge has a finite
angle, while the upper and lower surface of the second trailing edge are parallel
at point a. In the theoretical analysis of incompressible, inviscid flow the Kutta
condition is applicable (see for example Ref. [3]). This simply implies that the flow
smoothly leaves the trailing edge. If we apply this condition to the trailing edge with
finite angle, we see that it can only be satisfied for the upper and lower surface when
the velocities there are zero. This implies a stagnation point (C p = 1) located at
point a. If we consider the cusped trailing edge, we notice that the Kutta condition
can be satisfied while the upper and lower velocity are larger than zero C p < 1.
392 7 Airfoil Aerodynamics
The only condition that applies is that upper and lower velocity are equal. This
implies that C p < 1 for a cusp trailing edge.
In practice we do not have inviscid conditions, and the presence of the boundary
layer allows trailing-edge pressure coefficients to be lower than 1 (as is shown in
many of the figures in this chapter). However, we still see that airfoils with cusped
trailing edges allow for lower trailing edge pressure coefficients without separation
than airfoils with a finite trailing-edge angle. Therefore, cusped trailing edges have
a smaller adverse pressure gradient over the aft part of the airfoil, which postpones
separation. Many supercritical airfoils therefore employ cusped trailing edges (e.g.
SC(2)-0412 and DSMA-523a that have featured in this chapter).
One of the major disadvantages of having a cusped trailing edge is the lack of
physical space in the trailing edge to make a structure. In this region of the wing
often flaps or ailerons are positioned. Due to the diminishing trailing edge thick-
ness, the second moment of area is relatively small. To give the trailing edge its
required stiffness (to prevent aeroelastic deformation) a relatively heavy structure
is required. Obviously, this introduces weight penalties, which eventually translate
into an increase in induced drag (see Problem 7.6). To remedy this problem, aero-
dynamicists investigated the effect of cusped trailing edges with a finite thickness.
It was shown that blunt trailing edges reduce the adverse pressure gradient on the
upper surface by utilizing off-surface pressure recovery [41]. In practice this means
that the pressure coefficient of the lower surface could be lower than the pressure
coefficient on the upper surface.
Let us try to imagine what the effect of such a blunt trailing edge would be. First
of all, we foresee an increase in base drag: the low pressure a the trailing edge acts
on an aft-facing surface creating additional pressure drag. In subsonic conditions we
therefore expect an increase in drag coefficient for a given lift coefficient (hence,
a reduction of the lift-to-drag ratio). However, with increasing Mach number and
the formation of a shock wave, the boundary layer is exposed to an ever increasing
adverse pressure gradient. Due to the lower adverse pressure gradient of the airfoil
with blunt trailing edge, the momentum thickness will be smaller (see also Exam-
ple 6.4) reducing the total momentum loss in the boundary layer. So we have two
effects at transonic conditions: an increase in pressure drag due to the blunt base of
the airfoil, and a decrease in pressure drag due to the reduced momentum thickness
at the trailing edge. Which of these two phenomena dominates depends on the rel-
ative thickness of the trailing edge compared to the airfoil chord. Investigations at
NASA showed that the optimum trailing-edge thickness varied with the maximum
thickness of the airfoil and was somewhat below 0.7 %c [17]. Figure 7.21 shows the
effect of a blunt trailing edge on the drag coefficient development of an 11 % thick
supercritical airfoil at cl = 0.7. It becomes clear from this figure that at subsonic
conditions the drag coefficient due to the blunt trailing edge is higher, while at tran-
sonic conditions (between M = 0.68 and M = 0.79) the blunt trailing edge reduces
the drag coefficient.
7.5 Low-Speed Stall 393
.018 10 sharp
y/c (%)
.016 blunt
-10
0 10 20 30 40 50 60 70 80 90 100
.014 x/c (%)
drag coefficient, cd (~)
.012
.010
(t/c)TE in percent:
.008 0 (sharp)
1.0 (blunt)
.006
.60 .64 .68 .72 .76 .80 .84
Mach number, M (~)
Fig. 7.21 Effect of blunt trailing edge on drag coefficient for an 11 % thick airfoil at cl = 0.7 (after
Ref. [17])
When an aircraft goes into a stall, the flow over the wing starts to separate from the
upper wing surface and creates a wake of turbulent air. With increasing angle of attack
this separated region becomes larger and larger, which reduces the rate at which lift
increases. At a particular angle of attack, the rate at which the lift rises with angle of
attack (dL/dα) reduces to zero. At this point the aircraft experiences its maximum
lift coefficient (C L max ). Increasing the angle of attack even further reduces the total
lift of the airplane. At the same time, the wake becomes larger, which increases
the drag. As the lift over the wing decreases, the moment balance on the aircraft
is disturbed. Many aircraft are designed such that the horizontal tailplane provides
a large enough nose down pitching moment to reduce the angle of attack once the
wing stalls. During this motion the airplane starts to lose altitude and increases speed.
While the angle-of-attack decreases, the flow remains separated. At an angle of attack
well below the stall onset angle-of-attack the flow finally re-attaches to the wing and
the airplane recovers from the stall. The latter effect is called stall hysteresis. This
behavior is graphically shown in Fig. 7.22. We see the hysteresis effect on both the
lift coefficient (Fig. 7.22a) and pitching moment coefficient (Fig. 7.22b).
The maximum (trimmed) lift coefficient is an important parameter for airplane
performance. It is directly related to the required take-off field length and the landing
394 7 Airfoil Aerodynamics
(a) (b)
0 25 0 25
Anlge of attack (deg) Anlge of attack (deg)
Fig. 7.22 Typical stall behavior of a passenger aircraft during flight. Data obtained from Ref. [12]
for a Dornier 328. a Lift coefficient. b Moment coefficient
length of the aircraft. Therefore, many aircraft employ high-lift devices to increase
the lift coefficient of the wing such that they can have a lower stall speed and therefore
meet field length requirements. Typical high-lift devices are slotted flaps at the trailing
edge and flaps, slats, or drooped noses at the leading edge. Fighter aircraft also employ
their high-lift devices during combat maneuvers, which often occur in the transonic
speed realm. At any lift coefficient the high-lift devices are automatically positioned
such that maximum lift-to-drag ratio is achieved. This is an important parameter for
the minimum turn radius or maximum turn rate that can be achieved.
Back to the maximum lift coefficient at low speeds and why it is important to treat
this in a text on transonic aerodynamics. In the following subsections we discuss the
various types of stall and the effect of Reynolds number and Mach number on the
maximum lift coefficient of an airfoil. Furthermore, we present the aerodynamics of
high-lift devices. We also show that transonic effects can often be the limiting factor
for the maximum lift coefficient that can be achieved, even at low subsonic speeds.
pulled inwards, the jet moves to the body instead. Gasses experience the same effect
around an airfoil. The flow is accelerated due to the presence of the body. Its pressure
reduces (Bernoulli) and the flow ‘sticks’ to the body. The further away from the body,
the lower the velocity increase. This principle has been illustrated in Sect. 6.2.1. The
fact that the flow attaches to the body can therefore be attributed to the lower pressure
of the fluid around it. When this pressure rises due to an adverse pressure gradient in
the streamwise direction, the velocity of the air around the airfoil decreases and the
Coandă effect diminishes. At the same time, the boundary-layer momentum reduces
and reverse flow within the boundary layer marks the beginning of boundary layer
separation (Chap. 6). When the angle of attack increases further, the boundary layer
separates from the surface and becomes a shear layer between a turbulent wake and
the laminar outer flow: the wing has stalled.
We distinguish three types of stall: leading-edge stall, trailing-edge stall, and thin-
airfoil stall. The latter one is merely an artifact of low-Reynolds number flows and,
contrary to what its name suggests, not only pertains to thin airfoils. At a certain
angle of attack the laminar boundary layer separates due to the high adverse pressure
gradient right after the nose of the airfoil. Further downstream the laminar boundary
layer transitions to a turbulent boundary layer and re-attaches to the surface. The
bubble of reverse flow that is formed between the separation and re-attachment point
is called the laminar separation bubble. When increasing the angle of attack, this
bubble expands downstream. Beyond a certain angle of attack the boundary layer
does not reattach and a turbulent wake is present over the entire aircraft. This means
an instant loss in lift and a large increase in drag. Most aircraft that are designed for
high-subsonic speeds actually fly at higher Reynolds numbers, making thin-airfoil
stall almost never seen in practice. However, scaled versions of these aircraft which
are tested in at a lower Reynolds number in the wind tunnel could develop thin-airfoil
stall. Thin-airfoil stall can be prevented by forcing transition to occur prior to the
point where the boundary-layer separates from the surface.
Leading-edge stall occurs when the boundary layer is not capable of negotiating
the large adverse pressure gradient that follows the suction peak on the nose. Before
the stall angle of attack is reached as small laminar separation bubble forms just aft
of the suction peak. This bubble may span as little as 1 % of the chord length and
forms behind the suction peak under the steep adverse pressure gradient. The sepa-
rated laminar boundary layer transitions over the bubble to turbulent and reattaches.
Increasing the angle of attack moves this bubble forward and shortens it until the
laminar boundary layer is incapable of reattachment. The result is that flow separates
and that a wake is created starting at the nose of the airfoil (see also Fig. 6.23b).
Alternatively, when a turbulent boundary layer is capable of reattachment it could
still separate at a short distance behind the bubble due to the adverse pressure gradi-
ent. When either of these two events happens the suction peak over the leading edge
collapses and a relatively constant pressure coefficient over the airfoil surface results.
In the lift curve of the airfoil we see a step decrease in lift over a very small increment
in angle of attack. A wake over the entire airfoil is created which creates a sudden
increase in drag. In Fig. 7.23 leading-edge stall is shown for a 10 %-thick airfoil.
396 7 Airfoil Aerodynamics
1.6
1.2
0.8
0.4
-12
-8
Cp
-4
Fig. 7.23 Leading edge stall: notional lift and drag curves (top), typical pressure distributions
(middle), and upper-surface stream line (bottom). Data based on Ref. [28] for Re = 4.0 × 106 ,
M = 0.17
Trailing-edge stall, on the other hand, is more gradual. Once the boundary layer
has survived the steep adverse pressure gradient behind the leading edge, it progres-
sively grows thicker and gets a higher shape factor. This means that the momentum
of the boundary layer close to the wall is diminishing. At a certain angle of attack
the boundary layer starts to separate at the trailing edge (usually when H is between
2.4 and 2.6), creating a small wake. Increasing the angle of attack further increases
the adverse pressure gradient over the aft part of the airfoil and moves the separation
point upstream. At the same time a laminar separation bubble could occur closer to
the leading edge of the airfoil. With increasing angle of attack the size of the wake
from the rear separation point increases and the pressure coefficient behind the sep-
aration point remains relatively constant. Increasing the angle of attack even further
could merge the two separated regions creating a complete wake over the airfoil.
Figure 7.24 visualizes the interaction between separation, pressure distribution and
lift coefficient.
In the previous discussion we have treated leading-edge and trailing edge stall as if
they were two independent types of stall. In practice often mixed types of stall occur.
In such cases a short laminar separation bubble occurs after which the boundary layer
reattaches. Further downstream the turbulent boundary layer separates indicating
trailing-edge stall. The laminar separation near the leading edge can have a profound
effect on the thickness of the boundary layer as is noted in [30]. The increase in the
7.5 Low-Speed Stall 397
1.6
1.2
0.8
0.4
-8
Cp -4
Fig. 7.24 Trailing edge stall: notional lift and drag curves (top), typical pressure distributions
(middle), and upper-surface stream line (bottom). Data based on Ref. [31] for Re = 6.3 × 106 ,
M = 0.15
extent of an existing bubble increases the initial thickness of the turbulent boundary
layer and, therefore, increases the tendency for the latter to separate. In this way the
laminar separation bubble promotes the onset of turbulent stall near the trailing edge.
If we compare Figs. 7.23 and 7.24 we see that the difference in stall behavior is
a function of the airfoil shape. Important parameters that determine whether leading
or trailing-edge stall occurs include the thickness-to-chord ratio, the leading-edge
radius, and the leading-edge camber. Typically, blunt and thick airfoils are more
likely to suffer from trailing-edge stall, while sharp and thin airfoils suffer from
leading-edge stall. In addition, the Reynolds number plays an important role whether
the leading-edge or trailing-edge stall prevails.
With increasing wing size and flight speed, the Reynolds number grows. With increas-
ing Reynolds number the (turbulent) boundary layer at a given location on the airfoil
reduces in thickness according to (6.58). In general, increasing the Reynolds num-
ber increases the maximum lift coefficient for a given type of stall (leading-edge or
trailing-edge). The thinner boundary layer is capable of negotiating a larger adverse
pressure gradient without separating. There is, however, a more important Reynolds
398 7 Airfoil Aerodynamics
number effect that is instrumental for the maximum lift coefficient: the position of
the transition region. An increase in Reynolds number could also imply a switch
from leading-edge stall to trailing-edge stall. In Fig. 7.25 this is demonstrated for a
moderately thick airfoil, which’ stall behavior is evaluated at three different Reynolds
numbers. Note that this airfoil shows thin-airfoil stall at low Reynolds numbers even
though it is not considered a thin airfoil. This confirms that thin-airfoil stall is not
limited to thin airfoils. We explain the various Reynolds-number-related effects on
stall in the subsequent paragraphs.
As the Reynolds number increases the position where the boundary layer tran-
sitions from laminar to turbulent shifts forward (see Sect. 6.5.4). As the region of
transition has a large effect on the boundary layer development over the airfoil it has
an impact on when (at what angle of attack) and where (right after re-attachment
or at the trailing edge) the turbulent boundary layer separates. For a low Reynolds
number the transition point lies downstream of the suction peak. If a small laminar
separation bubble is present the boundary layer transitions from laminar to turbu-
lent over the bubble and reattaches. The reattached boundary layer at the end of the
separation bubble has a significantly higher thickness, which still grows in thickness
downstream of the reattachment point. Increasing the Reynolds number shifts the
transition region forward, which reduces the size of the separation bubble and conse-
quently reduces the thickness of the boundary layer downstream of the bubble. The
downstream boundary layer therefore separates at a higher angle of attack, which
increases the maximum lift coefficient. When the transition region moves in front of
the separation point, the effect of a higher Reynolds number is limited to the thick-
ness decrease of the turbulent boundary layer, which could increase the maximum
lift coefficient even further.
7.5 Low-Speed Stall 399
(~)
(after Ref. [40]). Note data
1.2
not at constant Reynolds
max
number
1.0
P-51B
0.9 YP-80A
P-39N
0.8
F8F-3
0.7
P-38F
0.6
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Mach number, M (~)
The Mach number can have a significant effect on the maximum lift coefficient
that can be attained. Even at low-subsonic conditions this statement holds up. In
Ref. [40] it is shown through a series of flight tests how the airplane maximum
lift coefficient is affected both by Reynolds and by Mach number. In Fig. 7.26 the
maximum lift coefficients of six different propeller-powered aircraft are graphed as
a function of the Mach number. Three aircraft used modified airfoils designed to
promote a laminar boundary layer over a large part of the upper surface during cruise
conditions (the ‘low-drag airfoils’). The other three aircraft relied on NACA 4 and 5
series airfoils (the ‘conventional airfoils’). We see a distinctly different maximum lift
behavior between the airplanes having low-drag airfoils and the airplanes having the
conventional airfoils. In the subcritical domain the uniform lowering of the maximum
lift coefficient is attributed to the steepened adverse pressure gradient resulting from
compressibility effects.
At Mach numbers beyond the critical Mach number the maximum lift coefficient
is governed by the suction peak. In Fig. 7.27 it can be seen that increasing the Mach
number reduces the suction peak for both types of airfoils. In both cases we see
that the suction peak is widening with increasing Mach number. Furthermore, the
low-drag airfoil develops a shock wave beyond M = 0.63. The pressure coefficient
in front of the shock remains below the dash-dotted line, which corresponds to a
maximum local Mach number. The minimum pressure coefficient always appears
to stay below the dashed line, which also corresponds to a maximum local Mach
number. We shall investigate why this happens.
400 7 Airfoil Aerodynamics
-2 -2
Cp Cp
0 M = 0.40 0
CL = 1.23 CL = 0.96
max max
-4 Cp * -4
M = 2
-2 -2
Cp Cp
0 M = 0.50 0
CL = 1.06 CL = 0.93
max max
Cp γ+3
-2 M = -2
2
Cp Cp Cp
crit
0 M = 0.63 0
CL = 0.77 CL = 1.08
max max
-2 -2
Cp Cp
0 M = 0.68 0
CL = 0.68 CL = 1.03
max max
Fig. 7.27 Typical pressure distributions about NACA 4-series airfoils (left) and NACA 6-series
airfoils (right) at maximum lift coefficient (after Ref. [40])
The same theoretical boundary has been shown in Fig. 6.9 where the dashed
line shows a theoretical limit in local Mach number for supersonic flow over a
convex surface embedded within
√ subsonic flow. It has been shown that a characteristic
Mach number of M ∗ = 2 is the highest Mach number for which Mach waves
can exist for which the curvature is monotonic.7 This corresponds to a situation
where the streamline curvature is still continuous and the flow is isentropic and
7 A monotonic curvature means that the Mach wave is either entirely convex or entirely concave.
If Mach wave curvature is not monotonic it means an inflection point is present somewhere on the
Mach wave.
7.5 Low-Speed Stall 401
uniform [27]. Via Eq. (4.10b) we calculate that this corresponds to a maximum local
Mach number of:
2
M=√ ≈ 1.58 for γ = 1.4 (7.5)
3−γ
We can compare this maximum Mach number to the maximum Mach number that
can be achieved in the throat of a supersonic convergent-divergent channel. In that
case the flow is said to be choked. Whatever we do upstream or downstream of the
throat, the Mach number can never exceed M = 1. In our case we have a similar
condition only this time we consider external flow and limit the freestream Mach
number to values below unity. The theoretical value of 1.58 is limited to flow over
convex surfaces that provide monotonic curvature to the Mach waves. The absolute
maximum is attained when discontinuities in streamline curvature are allowed. √ In
this case the first discontinuity in streamline curvature starts at M ∗ = 2. This
corresponds to theoretical accelerations of infinity at the point of discontinuity. The
theoretical maximum that can be attained is referred to as the Ringleb solution [36]:
4
M= ≈ 2.50 (7.6)
3−γ
For practical cases (7.5) is “the most likely maximum value to be approached in
most cases,” according to Laitone [24]. Substituting (7.5) in (6.17) and rewriting this
for C p yields a theoretical minimum local pressure coefficient as a function of the
freestream Mach number:
γ
2 3−γ γ−1 2 γ−1
C p M ∗ =√2 = 1+ M∞ −1 (7.7)
γ M∞ γ+1 2
This minimum has been indicated in Fig. 7.27 with the dashed line. Note that the
minimum pressure comes very close to this value but never exceeds it. This confirms
the statement that (7.7) is a practical limit on the minimum pressure coefficient
and that (7.5) is a limiting Mach number. This analysis also demonstrates that even
at speeds well below the speed of sound, transonic effects dictate maximum lift
coefficient.
The theoretically obtained minimum pressure coefficient has been compared to
experimental data obtained from a series of wind tunnel tests by Axelson [5]. As
can be seen in Fig. 7.28 at Mach numbers beyond 0.5, considerably lower pressure
coefficients can be obtained than those predicted by (7.7). An alternative prediction
by Mayer comes closer to the measured values. Mayer based his predictions on a
series of wind tunnel tests between M = 0.4 and M = 2.2 [29]. He proposed a
statistical relation between the minimum value of the pressure coefficient and the
freestream Mach number:
−1
C pmin = 2 (7.8)
M∞
402 7 Airfoil Aerodynamics
(~)
ahead of shock
min
−3
Pressure coefficient, Cp
−2.5
−2
-1
Cp = 2
min M
−1.5
8
2
Mlim=
−1 3−γ
γ+3
−0.5 Mlim=
2
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9
Free−stream Mach number, M (~)
8
It can be shown (see Problem 7.9) that this results in a static pressure ratio of approx-
imately 0.30 between the suction peak and the free stream. In other words, it is
predicted to achieve 70 % of pure vacuum under the suction peak. At Mach numbers
higher than 0.5 it can be seen from Fig. 7.28 that Mayer’s prediction comes closer to
the measured values published in [5], but still underestimates in certain cases.
If we return to Fig. 7.27 we observe that at M = 0.63 and M = 0.68, a shock
wave is present for the low-drag airfoil. In addition, the dash-dotted line indicates a
theoretical maximum local Mach number that cannot be exceeded in front of a shock
wave without creating separated flow [24]:
γ+3
M= ≈ 1.48 for γ = 1.4 (7.9)
2
In Ref. [24] it is demonstrated that for constant stagnation conditions ahead of the
shock and a Mach number of 1.48 the maximum absolute value of p2 (static pressure
behind the shock) is achieved. In other words, for this limiting Mach number the
maximum magnitude of static pressure recovery is achieved (although this is usually
less than the free stream static pressure). If we imagine the local velocity to increase
beyond this Mach number and form a shock wave further downstream, the static
pressure recovery would be less ( p2 would be lower). Therefore, the higher subsonic
free stream static pressure would push the normal shock wave back upstream until
the maximum absolute static pressure would be produced. Substituting (7.9) in (6.17)
and rewriting this for C p yields a theoretical limiting local pressure coefficient in
front of the normal shock wave:
⎡ ⎤
2 γ−1
γ
2 ⎣ 2 γ−1 2
C p γ+3 = 1+ M∞ − 1⎦ (7.10)
M= 2 γ M∞
2 γ+1 2
7.5 Low-Speed Stall 403
This is confirmed by the graphs for M = 0.63 and M = 0.68 for the NACA
6-series airfoil in Fig. 7.27. We observe that the pressure coefficient is increasing
between the suction peak and the shock. This indicates that the local Mach number
is reducing from a value close to 1.58 to a√lower value in front of the shock.
It is demonstrated in [11] that M = (γ + 3)/2 is the only Mach number for
which a curved streamline can become straight when passing through a normal shock
wave. Consequently, it would be the most likely Mach number in front of a shock
wave whenever boundary layer separation occurs at the foot of the shock. If we
combine these observations we can conclude that on convex profiles the maximum
Mach number in front of a normal shock wave never exceeds 1.48 and that it is
likely that this shock causes boundary layer separation. Evidence of the accuracy of
this predicted pressure coefficient is presented in Fig. 7.28. The experimental data
consistently shows that the value of M = 1.48 ahead of the shock is never exceeded
in any of the experimental tests.
Most high-subsonic aircraft have leading-edge and trailing-edge devices to allow for
a higher lift coefficient and hence a lower stall speed. Many contemporary aircraft
use slats or leading-edge flaps to increase the stall angle of attack, while trailing-
edge flaps are used to increase the camber of the wing and extend the local chord.
The combined effect can increase the maximum lift coefficient by more than 100 %.
The reader is referred to [33] or [38] for a general discussion on the aerodynamics
of high-lift devices. In this section we limit ourselves to the transonic effects that
accompany the flow over multi-element airfoils.
The idea behind high-lift devices is to increase the maximum lift coefficient.
This can only be achieved if the suction over the upper surface is magnified without
boundary layer separation. The slots between the individual components play a piv-
otal role to achieve this [38]. However, the same aerodynamic limitations in terms of
Reynolds and Mach number effects, as discussed previously, exist for multi-element
airfoils. Increasing the Reynolds number has a beneficial effect on the maximum lift
coefficient, while the maximum local Mach number limits the minimum pressure
coefficient and therefore the lift coefficient. The main difference with single-element
airfoils is the magnitude of the suction peak. For well-designed multi-element air-
foils, the highest suction peak occurs on the slat and can easily exceed C p = −20 at
a theoretical Mach number of zero. As we now know, the transonic effects therefore
occur at much lower freestream conditions sometimes as low as M = 0.2. This
means that the stall speed of an aircraft in landing configuration at sea level, Vs0 ,
could in fact be limited by the maximum supersonic Mach number over the leading
edge of the slat (see Problem 7.10).
Figure 7.29 shows the airfoil of the A320 wing in landing configuration with slat
and flap deployed. The dash-dotted lines indicate the edge of the various boundary
layers that are formed on this multi-element airfoil. We can see how the wake of
404 7 Airfoil Aerodynamics
slat wake
slat wake
wing wake
wing boundary layer
flap boundary-layer separation
shock-wave-boundary-
layer interaction
transition
transition
supersonic bubble cove separation cove separation
SLAT WING FLAP
Fig. 7.29 A320 airfoil in landing configuration (modified from Ref. [13])
the slat merges with the boundary layer of the wing. This is termed boundary layer
confluence. This happens again over the upper surface of the flap where the wake of
the wing surface merges with the flap boundary layer. In this example, there exists
reversed flow over the flap, which indicates the onset of stall. Over the slat we also
observe a local bubble of supersonic flow that is terminated with a shock wave.
Under the shock wave the flow transitions from laminar to turbulent resulting in a
thickening of the boundary layer. We can imagine how, with increasing Mach number,
the strength of the shock wave increases, which eventually results in boundary-
layer separation at the shock foot (see also Sect. 7.6.1). It is this phenomenon that
effectively limits the maximum lift coefficient as a function of the freestream Mach
number.
To substantiate these claims, consider Fig. 7.30. Here we have a typical multi-
element airfoil consisting of a slat and a double-slotted flap system. The graph shows
the minimum pressure coefficients measured at the suction peak on the leading edge
of the slat. In addition, the statistical limit by Mayer (7.8) as well as the theoretical
limit of (7.7) are shown. Note that the theoretical limit predicts a lower pressure
coefficient than the statistical prediction by Mayer on this low-subsonic domain.
The symbols indicate various combinations of flap and slat deflection. We√clearly
see that the boundary is formed by the limiting Mach number of M = 2/ 3 − γ,
regardless of the flap or slat setting. We also observe a decrement in minimum
pressure coefficient before the local Mach number reaches the limiting value. In the
cases where the slat is deployed the decrement in lift coefficient between M∞ = 0.19
and M∞ = 0.28 for this particular airfoil is between 0.3 and 0.4 [46]. This causes
a 10 % decrease in maximum lift coefficient due to the transonic “choking” of the
flow over the slat leading edge.
7.6 High-Speed Stall 405
δf
δs 1
−26 δf
2
2
(~)
−24 Mlim=
3−γ
min
−22 -1
Cp =
Pressure coefficient, Cp
δf
2
min M δs δf
8
−20 1 2
−18 10 35 10
−16 10 35 -
- 35 10
−14
- 35 -
−12
- - -
−10
0.18 0.2 0.22 0.24 0.26 0.28 0.3
Free−stream Mach number, M (~)
8
Fig. 7.30 Minimum pressure coefficients at the leading-edge suction peak for Re = 6.9 × 106
(after Ref. [46])
CL, max
buffet boundary
boundary for
critical freestream
ΔCL
conditions
CL = const. CL, max
ΔM∞
design point
1.0
Freesetream Mach number, M∞ (~)
Fig. 7.31 Separation boundary at various Mach numbers and lift coefficients (after Ref. [42])
In addition, this combination of Mach number and lift coefficient effectively limits
the altitude an airplane can reach at a given Mach number when flying in steady,
symmetric, horizontal flight (see Problem 7.11).
Fig. 7.32 Schlieren images of the flow around a 6 % thick RAE 104 airfoil at 2◦ angle of attack
(from Ref. [34]). a M = 0.84. b M = 0.86. c M = 0.88. d M = 0.90
0.2
0.1
0
0.5 0.6 0.7 0.8 0.9 1.0
Freestream Mach number, M (~)
Fig. 7.33 Development of the section lift coefficient of the 6 % thick RAE 104 airfoil at α = 2
(after Ref. [34]). Note p2 indicates the pressure downstream of the shock, PTE is the trailing edge
pressure, while psonic corresponds to the pressure for sonic flow at points downstream of the shock,
i.e. psonic = 0.528 pt
on the upper surface creates an inclined ramp for the flow outside the boundary layer
resulting in a more oblique shock wave terminating the supersonic domain.
The effect of shock-induced separation on the lift coefficient is shown in Fig. 7.33.
The letters a through d refer to the respective images in Fig. 7.32. We see that the
maximum lift coefficient at a constant angle of attack is reached around M = 0.87.
When the boundary layer fully separates from the surface the lift coefficient starts to
decrease rapidly. Most of this decrease is caused by the increase in suction over the
lower surface due to the decrease in trailing-edge pressure when the boundary layer
has separated. With increasing Mach number, the shock wave on the lower surface
moves faster towards the trailing edge than the upper shock wave creating more
suction over the lower surface and hence reducing the lift coefficient. A minimum
is obtained when the lower shock reaches the trailing edge. Increasing the Mach
number even further results in a higher value of the lift coefficient that corresponds
with both shocks having reached the trailing edge.
In Fig. 7.34 we schematically repeat this development of the shock pattern over
a generic airfoil of low thickness. The same effects as in the wind tunnel are seen
as can be verified by comparison to Fig. 7.32. In Fig. 7.34 the strength of the shock
has been indicated by the thickness of the shock line. Note how the shock strength
is highest at the surface, while its strength decrease further away from the body.
Intuitively, we can verify this because the supervelocities are highest close to the
surface. Furthermore, the relative size of the shock wave increases with higher Mach
number. At M = 1.4 the shocks at leading and trailing edge even exceed the bounds
of this figure. Finally, the notional point of separation is shown along with the growth
of the wake over the transonic regime. Note how the wake grows up to M = 0.9 and
decreases again above that value.
7.6 High-Speed Stall 409
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.6
Freestream Mach number, M (~)
Fig. 7.34 Notional development of section lift coefficient over a symmetric airfoil at α = 2
Fig. 7.35 Schlieren images of the flow around a 6 % thick RAE 104 airfoil at M = 0.75 (from Ref.
[34]). a α = 2.7◦ . b α = 3.7◦ . c α = 4.7◦ . d α = 5.7◦
410 7 Airfoil Aerodynamics
Let us follow Pearcey [34] in his investigation into the behavior of shock-induced
separation at constant Mach number and increasing angle of attack. In Fig. 7.35 we
see schlieren images of the same airfoil as in Fig. 7.32. The Mach number has been
fixed at M = 0.75 while the angle of attack increases progressively from α = 2.7◦
through α = 5.7◦ . It is important to note that the transition has been artificially
induced at the nose and that the shock wave interacts with a turbulent boundary
layer at any of the aforementioned angles of attack. We see that somewhere between
α = 2.7 and α = 3.7 the boundary starts to separate at the shock foot. The shock is
still quite close to the leading edge and the boundary layer reattaches downstream.
With increasing angle of attack the shock moves aft and the separation bubble expands
rapidly towards the trailing edge. This causes the trailing edge pressure to decrease
and the shock to halt before it starts to move forward. At even higher angles of attack
the shock moves further forward together with the separation point.
In Fig. 7.36 the position of the shock foot is plotted for a range of angles of attack
and Mach numbers. We distinguish uniform behavior at the various Mach numbers:
initially the shock moves aft up to the point that the separation bubble reaches the
trailing edge. Beyond that angle of attack the shock wave and the separation point
move forward. With decreasing Mach number the angle of attack at which shock-
induced separation starts becomes progressively higher. Simultaneously, the shock
position at which separation starts becomes closer to the leading edge. It becomes
difficult to judge at what Mach number the separation ceases to be induced by the
shock instead of being induced by the adverse pressure gradient behind the suction
peak, as seen at low speeds.
Let us take a look at the effect of shock-induced separation on the lift coefficient.
In Fig. 7.37 we see the lift coefficient development over the RAE 104 airfoil at two
Mach numbers. In 7.37a we see that lift coefficient remains close-to constant beyond
the point where the separation bubble reaches the trailing edge. For M = 0.80 we
expect the separation to start at a lower lift coefficient. This is indeed the case as
can be seen in Fig. 7.37b. However, here we have somewhat of a different case.
0.8 M
Upper-surface shock position,
0.95
0.6
0.90
0.4
0.85
0.2
0.80
0.70 0.75
0
0 2 4 6 8 10
Angle of attack, (deg)
7.6 High-Speed Stall 411
0.8 0.8
0.6
0.6
0.4
0.4
0.2
0.2
0
-0.2
0 2 4 6 8 10 0 2 4 6 8 10
Fig. 7.37 Lift versus angle of attack for a 6 % thick RAE 104 airfoil at two transonic Mach numbers
(after Ref. [34]). a M = 0.70. b M = 0.80
The lift coefficient increases even when the shock-induced separation covers the
upper surface from shock foot to trailing edge. Such behavior is attributed to off-
surface pressure recovery and occurs beyond the angle of attack where the pressure
coefficient at the trailing edge falls below the critical pressure coefficient. When
this is the case, the pressure coefficient at the trailing edge has no effect anymore
on the development of the flow over the lower surface. With increasing angles of
attack, the pressure coefficient on the lower surface begins to rise again. At the same
time, the pressure coefficient on the upper surface decreases even further below the
critical pressure coefficient allowing the shock wave to move backwards resulting in
an increase in suction over the upper surface. The combined effect is an increase in
lift coefficient beyond the angle of attack of shock-induced separation. This effect
is dependent on the presence of off-surface pressure recovery and therefore partially
on the shape of the airfoil trailing edge.
In the previous section we have examined the effect of separation at the shock foot
at transonic speeds. We know from experience with low speed stall that separation
and the associated wake cause pressure fluctuations that are transmitted to the airfoil.
These fluctuations in pressure can be felt as structural vibration. Apart from these
fluctuations the separated flow is considered to be steady (e.g. invariant with time). In
transonic conditions, however, the separated boundary layer can also cause a periodic
motion of the shock wave over the surface of the airfoil. Such fluctuations usually
have a frequency on order of 102 Hz and generate large fluctuations in pressure (see
Fig. 7.38). These large fluctuations can be felt as a strong vibration in the airframe
412 7 Airfoil Aerodynamics
(a) (b)
2.0 -Cp 2.0 -Cp
1.5 1.5
1.0 1.0
.5 .5
0 = 3.5o 0 = 3.5o
M∞ = 0.723 M∞ = 0.723
-.6 -.6
R = 4.2·106 x/c
R = 4.2·106 x/c
-1.0 -1.0
0 .2 .4 .6 .8 1.0 0 .2 .4 .6 .8 1.0
Fig. 7.38 RA16SC1 airfoil. Envelope of the pressure distributions at buffet onset (after Ref. [25]).
a Calculation. b Experiment
M Wake deflected
M1 M Vs upwards
Weak shock
(attached flow)
Weak shock
Vs
M M
Wake
M1>1.3
Vs
Fig. 7.39 Shock wave development over a 14 % thick circular arc airfoil with increasing subsonic
freestream Mach number (after Ref. [14]). a M1 < 1.2 (no separation). b M1 > 1.2 start periodic
flow. c 1/2 cycle after (b). d M1 > 1.3 shock too strong for reattachment
structure. This phenomenon is termed transonic buffet and its mechanics is further
explained in this section.
Before we look at a realistic case of transonic buffet on a supercritical airfoil
it is instructive to consider the shock motion over a circular arc airfoil. The shock
development over a 14 % thick circular arc airfoil at symmetric flow conditions
is presented in Fig. 7.39. In Fig. 7.39a we see shocks of equal strength appearing at
both the upper and lower surface. In Fig. 7.39b we assume that a pressure disturbance
upstream of the shock on the upper surface causes the shock to move more forward.
The instantaneous shock Mach number (Ms ) is given by:
1 dxs
Ms = M1 + (7.11)
a dt
7.6 High-Speed Stall 413
where xs is the location of the shock and dxs /dt = Vs is the velocity of the shock
wave with respect to the airfoil. As the shock moves forward Ms increases beyond the
Mach number at which separation is triggered. So as the shock moves forward it also
shifts the separation point forward creating an asymmetric wake. At the same time, the
trailing edge pressure decreases which moves the shock on the lower surface closer
to the trailing edge. In moving backward, Ms decreases which prevents the boundary
layer from separating. As the upper shock moves forward over the convex surface it
encounters a slower supersonic region. Therefore, its strength weakens as it moves
upstream. The flow attaches at a point where the shock becomes too weak to initiate
separation. On the lower side the reverse is happening. As the shock moves closer
to the trailing edge its strength is increasing and at some point it initiates separation.
This causes the wake to move to deflect downward as can be seen in Fig. 7.39c. This
process is now repeating itself on the lower surface. The deflected wake essentially
causes the shock on the lower surface to move forward, while the shock on the
upper surface starts to move backward. Even though there is a time lag between the
downward deflection of the wake and the motion of the shock wave, the flow can
maintain a self-sustained periodic motion. Finally, when M1 > 1.3 (Fig. 7.39d) the
shock is strong enough at all times to cause separation. No reattachment can take
place and the periodic motion of the shock wave ceases.
Close to the surface of the airfoil we can approximate the flow as one dimensional.
We follow Ref. [26] to relate the oscillating shock position to the oscillating strength
of the shock wave. The latter one is given as the ratio of static pressure over the
shock wave. We denote the properties of the flow ahead of the shock wave with a
subscript 1 and behind the shock with subscript 2. The shock strength would then
be quantified as p2 / p1 . We assume that the flow has a varying velocity, u, a varying
static pressure, p, and a varying speed of sound, a, in the streamwise (x) direction.
The displacement of the shock in time is given by Δx. The pressure and velocity in
front and behind the shock can be written as follows:
dP1
p1 = P1 + Δx + p̃1 (7.12)
dx
p2 = P2 + p̃2 (7.13)
dU1
u 1 = U1 − u s + ũ 1 + Δx (7.14)
dx
u 2 = U2 − u s + ũ 2 (7.15)
where p̃ and ũ are perturbation pressures and velocities, respectively, and P and U
are the steady-state pressures and velocities, respectively. We know that the pressure
ratio over the shock wave is given by the Rankine-Hugoniot relation:
p2 2γ u 21
=1+ −1 (4.14)
p1 γ+1 a12
where ū 1 is the time average of the perturbation velocities ahead of the shock. Sub-
stituting (7.18) in (7.17) yields an expression for the time dependency of the shock
strength:
p2 P2 4γ dM1 iω ū 1
= + M1 − + Δx0 eiωt (7.19)
p1 P1 γ+1 dx a1 a1
If we assume that the disturbances in the supersonic region upstream of the shock
are negligible (i.e. ũ 1 ≈ 0) the relation reduces to:
p2 P2 4γ dM1 iω
= + M1 − Δx0 eiωt (7.20)
p1 P1 γ+1 dx a1
Let us try to interpret the contents of (7.20). On the RHS it requires the input of the
shock oscillation: amplitude (Δx0 ) and frequency (ω in rad/s). In addition, it requires
an estimate for the change in upstream Mach number with displacement (dM1 /dx)
at the neutral position of the shock. Finally, it requires an estimate of the steady state
properties of the shock: P2 /P1 and M1 .
Example 7.3 Assume we have a NACA 64A006 airfoil at M = 0.90 which has an
induced shock oscillation (this experiment is described in Ref. [43]). The steady-state
Mach number in front of the shock has been determined from quasi-steady conditions
and is M1 = 1.18 and dM1 /dx = 1.7. A periodic shock motion is induced with a
frequency of 120 Hz and an amplitude of Δx0 /c = 0.05. The static temperature in
front of the shock wave is 290 K. Calculate the following:
(a) The steady-state pressure ratio, P2 /P1
(b) The amplitude of the pressure ratio
(c) The phase shift in radians between pressure jump and shock position.
7.6 High-Speed Stall 415
Solution:
To calculate the steady shock jump in static pressure we employ (4.14):
P2 2γ 2
=1+ M1 − 1 = 1.46
P1 γ+1
This means that the time-averaged pressure ratio is 1.46 over this shock. To find the
amplitude of the pressure ratio we employ part of (7.20):
2 2
4γ dM1 iω
amplitude = M1 − Δx0 = 0.380
γ+1 dx a1
Finally, we determine the face shift between the shock position and the pressure
jump. We plot one period for the shock position and pressure jump, respectively, in
Fig. 7.40. We see that the pressure jump is lagging the shock position. We calculate
the phase shift, ϕ, as follows:
−1 ω dx
ϕ = tan = 0.91 (rad)
a1 dM1
xs/c (~)
0.05
Shock displacement,
−0.05
0 1 2 3 4 5 6 7 8
Time, t (ms)
Shock pressure jump, p2/p1 (~)
1.8
1.6
1.4
1.2
1
0 1 2 3 4 5 6 7 8
Time, t (ms)
Fig. 7.40 Oscillating shock position and corresponding pressure jump according to (7.18) and
(7.20), respectively
416 7 Airfoil Aerodynamics
The phase shift between shock position and pressure jump amounts to 1.21 ms.
For clarity, the oscillating shock position and oscillating shock jump are shown in
Fig. 7.40.
In the previous example it becomes clear that the shock pressure jump periodically
alters between p2 / p1 = 1.83 and p2 / p1 = 1.09. With the knowledge in mind that
the pressure magnitude of the pressure jump is dominant for boundary layer separa-
tion, we can deduce from this simple example that when the shock is weakest it is
conceivable that the boundary layer remains attached. Vice versa, it also conceivable
that when the shock strength passes a certain threshold the boundary layer fully sep-
arates. Pearcey suggested separation to occur when p2 / p1 > 1.4. The limiting Mach
number of (7.9) predicts separation to occur at p2 / p1 = γ + 1 (see Problem 7.13).
The shock motion in the previous example was induced by deflecting a control
surface (c f = 0.25c) with an amplitude of 1◦ . The deflection of the control surface
caused an oscillation in the wake and an oscillation in the back pressure of the shock
wave. When separated and attached flow follow each other rapidly we see similar
behavior: an oscillating wake causing an oscillating shock wave (see also Fig. 7.39).
The amplitude and phase shift of the pressure jump are largely dependent on the
sensitivity of the upstream Mach number (Ms ) to the shock position as well as on the
frequency, f . It can be shown (see Problem 7.14) that when dM1 /dx → 0 the phase
shift tends to π/2 or 90◦ . Furthermore, when the frequency increases the amplitude
of the shock decreases as is demonstrated in Ref. [43]. This results in a frequency at
which the amplitude of the static pressure jump is maximum (see Problem 7.15).
Example 7.4 One of the prime parameters responsible for the phase shift is the
value of dM1 /dx. Calculate this value for an SC(2) 0412 supercritical airfoil around
M = 0.775 for cl = 0.5 at a Reynolds number of 20 million.
Solution:
A reasonable quick way to investigate the change of Mach number ahead of the shock
with the position of the shock is to evaluate the pressure distribution. This has been
plotted for three values of the Mach number (M = 0.77, M = 0.775, and M = 0.78)
in Fig. 7.41. We see how the small changes in Mach number have a relatively large
effect on the shock position. From the jump in pressure coefficient we can calculate
the Mach number ahead of the shock wave. First we rewrite the pressure jump in
terms of the pressure coefficient using (3.11):
p2 γ M∞
2 C +2
p2
=
p1 γ M∞
2 C +2
p1
p
−0.5
0.5
1
0 0.2 0.4 0.6 0.8 1
position, x/c (~ )
We tabulate the values for the pressure coefficient, pressure jump, M1 , and the shock
position at each Mach number (Table 7.1).
Based on the values of M1 and xs /c at the three conditions we make an estimate
for dM1 /dx according to the following formula:
dM1 1 1.197 − 1.169 1.169 − 1.140 0.747 + 0.916
≈ + = = 0.831
d(x/c) 2 0.627 − 0.596 0.596 − 0.558 2
In the previous example the Mach number in front of the shock wave was still
relatively low (≈1.17). Separation would therefore be unlikely. However, the example
does demonstrate how one could calculate the change in shock strength with its
position. With modern CFD tools unsteady separation from the shock foot can be
predicted. In Fig. 7.42 the estimated flow field about a supercritical airfoil is shown
at four instances within one oscillation. A numerical approximation of the Unsteady
Reynolds-Averaged Navier-Stokes equations is used for these predictions. Errors on
the frequency prediction are between 9 and 17 %, which demonstrates that there is
still room for improvement. However, the numerical simulations do give us insight
into the mechanism of the shock-boundary layer interaction.
The shock motion on the upper surface of the supercritical airfoil of Fig. 7.42 is
slightly different from the case of the biconvex circular arc airfoil of Fig. 7.39. In this
case, we only have a single shock wave on the upper surface and no shock is formed
418 7 Airfoil Aerodynamics
(a)
Mach number, M 0.1 0.2 0.4 0.5 0.7 0.8 1.0 1.1 1.3 1.4
(b)
Mach number, M 0.1 0.2 0.4 0.5 0.7 0.8 1.0 1.1 1.3 1.4
(c)
Mach number, M 0.1 0.2 0.4 0.5 0.7 0.8 1.0 1.1 1.3 1.4
(d)
Mach number, M 0.1 0.2 0.4 0.5 0.7 0.8 1.0 1.1 1.3 1.4
Fig. 7.42 Mach number and flow streamline predictions for the RA16SC1 airfoil at developed
buffet conditions, M∞ = 0.732, α = 4◦ , and Re = 107 (from Ref. [20]; printed with permission
of the authors). a Most downstream position of the shock. b Middle position as the shock is
moving upstream. c Most upstream position of the shock. d Middle position as the shock is moving
downstream
on the lower surface during the cycle. The onset of buffet is shown to start after the
onset of flow separation. In other words, the flow has already fully separated from the
shock foot and oscillation only starts when the angle of attack is further increased.
7.6 High-Speed Stall 419
As the separated region grows, this high-pressure bubble pushes the shock upstream.
When the shock moves upstream it becomes stronger due to a combination of effects.
We have discussed the effect of a moving shock on the incident Mach number above.
Iovnovich and Raveh [20] call this the dynamic effect. In addition, they mention
the wedge effect, which alludes to the shock becoming more oblique due to the
formation of a wedge-like separation region at its foot. The shock strengthens as
it moves forward as is evidenced by the increase in inclination angle (Fig. 7.42b).
Finally, as the shock moves upstream and encounters more curvature (closer to the
leading edge) the region of expansion reduces and the shock strength weakens. In [20]
this is referred to as the curvature effect. In this example, the boundary layer does
not fully reattach at its most forward position (see Fig. 7.42c). Because the Kutta
condition is not satisfied, the circulation around the airfoil decreases substantially
leading to lower accelerations ahead of the shock. This is another contribution to
further weakening of the shock wave. When the flow reattaches the shock moves
downstream until separation starts again and the cycle is repeated.
7.7 Summary
In the quest for higher speeds in the late 1940s and early 1950s the wing aerodynamics
proved to be the limiting factor. The increase in wave drag that they produced at
Mach numbers close to the speed of sound proved to be prohibitive to achieve the
desired top speeds. This sparked research into new airfoil shapes that would yield a
comparatively low drag coefficient at the specified lift coefficient and Mach number.
Originally, laminar-flow airfoils were proposed for this purpose. Their favorable
pressure gradient over the first half of the airfoil allowed for a laminar boundary
layer up to 50 or 60 % of the airfoil upper surface. Even though these airfoils proved
to be suited for subsonic application, they developed strong shock waves at relatively
low Mach numbers resulting in excessive wave drag. New airfoils were developed
that could decelerate the flow over the upper surface without producing a strong
shock. The supercritical airfoil was born.
Shock-free supercritical airfoils are designed to isentropically decelerate the
supersonic flow on the upper surface of the airfoil to subsonic conditions. The local
Mach number on the upper surface is dominated by the interaction of compression
and expansion waves. This interaction is highly dependent on the shape of the sonic
line separating the supersonic bubble from the subsonic outer flow as well as on the
leading edge radius and the upper surface curvature of the airfoil. A slight change
in either of these parameters can cause the compression waves to coalesce and form
a shock wave. To avoid this sensitivity, many supercritical airfoils are designed to
allow for a (weak) shock wave to terminate the supersonic bubble. The pressure
distribution about these airfoils are characterized by a flat plateau over the first half
of the airfoil, followed by a sharp drop (normal shock) and a steep adverse pressure
gradient towards the trailing edge. The lower surface often has increased camber
420 7 Airfoil Aerodynamics
near the trailing edge. This increases the pressure on the aft part of the lower surface
contributing to a higher lift coefficient for a given angle of attack. A cusped trailing
edge is often used to allow for off-surface pressure recovery and a reduction of the
adverse pressure gradient preventing trailing-edge stall. In addition, a finite trail-
ing edge thickness has been demonstrated to have a similar effect. Both measures
contribute to lower profile drag at the design condition of the airfoil.
The stall of airfoils starts with the separation of the boundary layer. In low speed
conditions it has been shown that a limiting Mach number of 1.58 exists on the top
surface of the airfoil that effectively “chokes” the flow over continuously convex
profiles in conditions where M∞ < 1. This effectively limits the maximum lift
coefficient that can be achieved by profiles in subsonic conditions. This Mach number
limit is extended to flows over multi-element airfoils (employing flaps and slats). The
limiting Mach number can reduce the achievable lift coefficient at Mach numbers as
low as 0.25. In addition, a theoretical maximum Mach number of 1.48 exists in front
of a normal shock wave. At this Mach number the highest static pressure behind the
shock can be achieved for given stagnation conditions. Boundary layer separation is
likely to occur at values somewhat below this Mach number. The relatively strong
shock causes the boundary layer to separate at its foot. If the resulting separation
bubble merges with the separated boundary layer from the trailing edge we have
high-speed stall. The interaction between separated boundary layer and shock wave
can cause shock oscillation. The resulting high-frequency fluctuation in pressure
distribution causes vibrations. These vibrations are referred to as transonic buffet.
The buffet boundary forms a limiting line in the Mach-C L space of a wing that
effectively limits the achievable lift coefficient at a specified Mach number.
Problems
Supercritical Airfoils
7.1 Calculate the increase in drag divergence Mach number when we reduce a 10 %
thick supercritical airfoil down to 8 % for the same lift coefficient of cl = 0.5.
7.2 Explain why introducing aft-loading on an airfoil can reduce the required super-
velocities on the upper surface of the airfoil for a given lift coefficient.
7.3 In graph below the relationship between the drag coefficient and the Mach num-
ber of two aircraft are shown: the DC-10-30, and the MD-11, which was the successor
of the DC-10. For each lift coefficient the graphs for both aircraft display an expo-
nential increase at the so-called drag-divergence Mach number. Although the drag
divergence Mach numbers are similar for both aircraft, the DC-10 suffers from drag
creep (a steady increase of the drag coefficient with Mach number) while the MD-11
does not.
7.7 Summary 421
CD x 104
DC-10-30
MD-11
300
CL = 0.5
250
CL = 0.4
CL = 0.3
200
7.4 (a) What are the differences between a modern supercritical airfoil and a
so-called “sonic rooftop” profile in an aerodynamic sense? Refer in your answer
to the critical pressure coefficient.
(b) Indicate the differences in a sketch, drawing the two profiles on top of each other.
7.5 Below the first supercritical airfoil with slot is shown (from Ref. [47]). Can
you explain why the addition of a slot resulted in an increase in the drag-rise Mach
number?
7.6 Thin trailing edges require a heavier rib structure to be properly supported.
Argue why such an increase in structural weight can eventually lead to an increment
in (induced) drag.
7.7 The lower rear surface of the wing of the Airbus A340 exhibits a strong camber
(see figure below). What are its specific advantages and disadvantages?
422 7 Airfoil Aerodynamics
Photo R.Slingerland
7.8 The Lockheed P-38 Lightning (below) was in many ways innovative. An unfa-
vorable characteristic was that in straight high-speed dives the aircraft experienced
a nose-down pitching tendency (termed tuck-under or Mach tuck) above M = 0.65.
This tendency increased the dive angle and thereby the speed even more, leading to
structural break-up of the aircraft. Only after the addition of dive recovery flaps this
problem was solved. Explain why the nose had the tendency to drop at these Mach
numbers.
Photo USAF
Low-Speed Stall
7.9 Show that the Mayer limitation on the minimum pressure distribution corre-
sponds to a static pressure ratio of pmin / p∞ ≈ 0.30.
7.7 Summary 423
7.10 Consider an airplane with a 1-g stall speed of 66 m/s at 0 m ISA conditions.
This stall speed is achieved by deploying slats and flaps. Assume that the suction
peak over the slat amounts to C p = −24.
(a) Calculate the stall Mach number at 0 m ISA conditions.
(b) Calculate the minimum allowed pressure coefficient at 0 m ISA using (7.7).
(c) Calculate the stall speed at 1,655 m and Mach number at 1,655 m altitude.
Assume the lift coefficient and wing loading are the same at both altitudes.
(d) Calculate the minimum allowed pressure coefficient at 1,655 m altitude.
(e) Based on your calculations, do you think the maximum lift coefficient at 1,655 m
is still the same as at sea level?
High-Speed Stall
7.11 Prove that for an airplane in steady, symmetric, and horizontal flight the buffet-
onset boundary essentially limits its maximum operational altitude.
7.12 An airplane is cruising at its maximum cruise lift coefficient in steady horizontal
flight at a speed of 250 m/s.
(a) Calculate the smallest turn radius it should be able to attain without encountering
buffet at n = 1.3.
(b) Calculate the associated bank angle to reach n = 1.3
7.13 Derive from (7.9) and (4.14) that the pressure jump at which the boundary
layer separates at the shock foot amounts to p2 / p1 = 1 + γ.
Lift coefficient,
CL (~) CL (~) CL (~) CL (~) CL (~) CL (~)
0.9 0.8 0.7 0.6 0.5 0.4
0 1 2 3 4 5
M=0.70 (°)
0.2 0.1 0 R
0 1 2 3 4 5
M=0.60 (°)
0.1 0 R
0 1 2 3 4 5
M=0.50 (°)
0 R
0 1 2 3 4 5 6
R
(°)
Pressure coefficient trailing edge,
-0.2 C (~) M=0.50 -0.2 C (~) -0.2 C (~)
P,TE P,TE M=0.60 P,TE M=0.70
R
(°) R
(°) R
(°)
0 0 0
1 2 3 4 5 6 1 2 3 4 5 1 2 3 4 5
+0.2 +0.2 +0.2
-0.2 C (~) M=0.75 -0.2 C (~) -0.2
P,TE P,TE
M=0.80 CP,TE (~) M=0.85
R
(°) R
(°) R
(°)
0 0 0
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
+0.2 +0.2 +0.2
is measured for four frequencies: f = 10, 30, 60, and 80 Hz. The corresponding
amplitudes measure Δx0 = 0.105, 0.102, 0.063, and 0.056, respectively.8
(a) Calculate the phase shift in degrees between the shock position, xs and the
pressure jump p2 / p1
(b) Calculate the corresponding amplitude of the static pressure jump in each of the
above cases.
(c) Plot the phase shift in radians and the amplitude as a function of the frequency, f .
7.16 In the graphs of Fig. 7.43 the lift curves and trailing edge pressure coefficients
are drawn for the outboard wing of a given swept wing. Determine the buffet onset
boundary and plot this in an M versus C L graph.
8 The data presented in this problem is based on an experiment described in [43]. In this experiment
the shock movement was introduced trough a periodic change in angle of attack around α = 3.0 of
a NLR 7301 airfoil at M = 0.70.
References 425
References
1. Abbott, I.H., von Doenhoff, A.E.: Theory of Wing Sections. Dover Publications, New York
(1959)
2. Anderson, J.: Modern Compressible Flow with Historic Perspective, 3rd edn. McGraw Hill,
New York (2003)
3. Anderson, J.: Fundamentals of Aerodynamics, 5th edn. McGraw Hill, New York (2010)
4. Anon.: The supercritical airfoil. NASA TF-2004-13 DFRC (2004)
5. Axelson, J.A.: Estimation of transonic aircraft aerodynamics to high angles of attack. AIAA
paper 75-996, Los Angeles (1975)
6. van den Berg, B.: Reynolds number and Mach number effects on the maximum lift and the
stalling characteristics of wings at low speed. NLR TR 69025 U, Amsterdam (1969)
7. Blackerby, W., Johnson, J.: Application of advanced technologies to improve C-141 cruise
performance. In: 17th Aerospace Sciences Meeting. AIAA 79-0066, New Orleans (1979)
8. Boerstoel, J.W., Uijlenhoet, R.: Lifting aerofoils with supercritical shockless flow. In: Proceed-
ings of ICAS (1970)
9. Carpentieri, G.: An adjoint-based shape-optimization method for aerodynamic design. Ph.D.
thesis, Delft University of Technology (2009)
10. Drela, M., Giles, M.B.: Viscous-inviscid analysis of transonic and low Reynolds number air-
foils. AIAA J. 25(10), 1347–1355 (1987). doi:10.2514/3.9789
11. Emmons, H.W.: The theoretical flow of a frictionless, adiabatic, perfect gas inside of a two-
dimensional hyperbolic nozzle. NACA TN 1003 (1946)
12. Fischenberg, D., Jategaonkar, R.V.: Identification of aircraft stall behavior from flight test data.
In: RTO SCI Symposium on “System Identification for Integrated Aircraft Development and
Flight Testing”. RTO MP-11, Madrid, Spain (1998)
13. Flaig, A., Hilbig, R.: High-lift design for large civil aircraft. In: AGARD Conference Proceed-
ings CP-515, pp. 31.1–31.12. AGARD (1993)
14. Gibb, J.: The cause and cure of periodic flows at transonic speeds. In: Proceedings of ICAS,
pp. 1522–1530. ICAS-88-3.10.1 (1988)
15. Giles, M.B., Drela, M.: Two-dimensional transonic aerodynamic design method. AIAA J.
25(9), 1199–1206 (1987). doi:10.2514/3.9768
16. Göthert, B.: Widerstandsanstieg bei Profilen im Bereich hoher Unterschallgeschwindigkeiten,
Deutche Versuchsanstalt fur Luftfahrt, E.V. Untersuchungen und Mitteilungen Nr. 1167 (1944)
17. Harris, C.D.: NASA supercritical airfoils; a matrix of family related airfoils. NASA TP2969,
Langley (1990)
18. Hemenover, A.D.: Tests of the NACA 64-010 and 64A010 airfoil sections at high subsonic
Mach numbers. NACA-RM-A9E31. Moffett Field (1949)
19. Hoak, D.E., Anderson, R., Goss, C.R.: USAF Stability and Control Datcom. Air Force Wright
Aeronautical Laboratories, Wright Patterson Air Force Base, Ohio (1978)
20. Iovnovich, M., Raveh, D.E.: Reynolds-averaged Navier-Stokes study of the shock-buffet insta-
bility mechanism. AIAA J. 50(4), 880–890 (2012). doi:10.2514/1.J051329
21. Jameson, A., Ou, K.: 50 years of transonic aircraft design. Prog. Aerosp. Sci. 47(5), 308–318
(2011). doi:10.1016/j.paerosci.2011.01.001
22. Johnson, D.A., Spaid, F.W.: Supercritical airfoil boundary-layer and near-wake measurements.
J. Aircr. 20(4), 298–305 (1983)
23. Kacprznski, J., Ohman, L.H., Garabedian, P.R., Korn, D.G.: Analysis of the Flow Past a Shock-
less Lifting Airfoil in Design and Off-Design Conditions. NRCC LR-554, Ottawa (1971)
24. Laitone, E.: Local supersonic region on a body moving at subsonic speeds. In: Oswatitish,
K. (ed.) Symposium Transsonicum, vol. I, pp. 57–70. Springer, Heidelberg (1964)
25. LeBalleur, J.C., Girodroux-Lavigne, P.: Viscous-inviscid strategy and computation of transonic
buffet. In: Zierep, J., Oertel, H. (eds.) Symposium Transsonicum III. IUTAM, pp. 49–63.
Springer, Berlin (1988)
26. Lee, B.: Self-sustained shock oscillations on airfoils at transonic speeds. Prog. Aerosp. Sci.
37(2), 147–196 (2001). doi:10.1016/S0376-0421(01)00003-3
426 7 Airfoil Aerodynamics
27. Lin, C.C., Rubinov, S.I.: On the flow behind curved shocks. J. Math. Phys. 27(2), 105–129
(1948)
28. Maki, R., Hunton, L.: Investigation at subsonic speeds of several modifications to the leading-
edge region of the NACA 64A010 airfoil section designed to increase maximum lift. NACA
TN 3871. Moffett Field (1956)
29. Mayer, J.P.: A limit pressure coefficient and an estimation of limit forces on airfoils at supersonic
speeds. NACA RM L8F23, Langley Field (1948)
30. McCullough, G.B., Gault, D.E.: Examples of three representative types of airfoil-section stall
at low speed. NACA TN 2502. Moffett Field (1951)
31. McGhee, R., Beasley, W.: Low-speed aerodynamic characteristics of a 17%-thick-airfoil
section for general aviation applications. NASA TN-D-7428. Hampton (1973)
32. Nieuwland, G.Y., Spee, B.M.: Transonic airfoils: recent developments in theory, experiment,
and design. Annu. Rev. Fluid Mech. 5, 119–150 (1973)
33. Obert, E.: Aerodynamic Design of Transport Aircraft. IOS Press, Delft (2009)
34. Pearcey, H.H.: Some Effects of Shock-induced Separation of Turbulent Boundary Layers in
Transonic Flow Past Aerofoils. National Physics Laboratory R&M 3108 (1955)
35. Pearcey, H.H.: The aerodynamic design of section shapes for swept wings. Adv. Aeronaut. Sci.
3, 277–322 (1963)
36. Ringleb, F.: Exakte Loesungen der Differntialgleichungen einer adiabatischen Gasstroemung.
Zeitschrift fur angewandte Mathematik und Mechanik 20(4), 185–198 (1940)
37. Roskam, J.: Airplane Design, Part 6: Preliminary Calculation of Aerodynamic, Thrust and
Power Characteristics. DARcorp, Lawrence (2006)
38. Smith, A.M.O.: High-lift aerodynamics. J. Aircr. 12(6), 501–530 (1975)
39. Sobieczky, H., Seebass, A.R.: Supercritical airfoil and wing design. Annu. Rev. Fluid Mech.
16, 337–363 (1984)
40. Spreiter, J.R., Steffen, P.J.: Effect of Mach and Reynolds numbers on maximum lift coefficient.
NACA TN 1044 (1946)
41. Standish, K.J., van Dam, C.P.: Aerodynamic analysis of blunt trailing edge airfoils. J. Sol.
Energy Eng. 125(4), 479 (2003). doi:10.1115/1.1629103
42. Stanewsky, E., Basler, D.: Mechanism and Reynolds Number Dependence of Shock-Induced
Buffet on Transonic Airfoils. In: Zierep, J., Oertel, H. (eds.) Symposium Transsonicum III.
IUTAM. Springer, Berlin (1988)
43. Tijdeman, H., Seebass, A.R.: Transonic flow past oscillating airfoils. Annu. Rev. Fluid Mech.
12, 181–222 (1980)
44. Torenbeek, E.: Elements of aerodynamic wing design. Advanced Aircraft Design, Chap. 10.
Wiley, Chichester (2013)
45. Viken, J., Wagner, R.D.: Design limits of compressible NLF airfoils. In: 29th Aerospace Sci-
ences Meeting. AIAA-91-0067, Reno (1991). doi:10.2514/6.1991-67
46. Vogelaar, H.: Wind-tunnel investigation on the two-dimensional F-29 model 12-1 airfoil section
with various high lift devices at various Mach and Reynolds numbers. NLR TR 83059C (1983)
47. Whitcomb, R.T., Clark, L.R.: An Airfoil Shape for Efficient Flight at Supercritical Mach
Numbers. NASA TM X-1109. Langley (1965)
Chapter 8
Aerodynamics of Swept Wings
8.1 Introduction
In the previous chapter we have seen transonic flow about airfoils. We now add a
third dimension to the airfoil and investigate the flow over finite wings in transonic
conditions. If we look at airplanes that fly in the transonic regime, there is one
distinct feature that separates these aircraft from their subsonic counterparts: they
exhibit wing sweep. On a first glance the use of wing sweep is not very obvious. For
a given aspect ratio, it lengthens the structural span of the wing, which means that
a heavier structure is required. In addition, wing weight is even more increased to
resist the aerodynamic twist that is incurred during bending (see Sect. 8.5). Moreover,
the control surfaces and high-lift devices on the trailing edge of a swept wing are
less effective. In view of these drawbacks one might wonder why virtually all high-
subsonic aircraft still exhibit a swept-back wing.
The answer to this question lies in the higher drag divergence Mach number.
The swept wing simply allows an aircraft to fly at a higher Mach number without
the exponential increase in wave drag. For supersonic aircraft, a higher wing sweep
reduces the maximum value of the drag coefficient when traversing from the subsonic
to the supersonic domain. For high-subsonic aircraft wing sweep is a complimentary
measure to the supercritical airfoil. Both are designed to push the boundary of drag
divergence to a higher Mach number. However, we shall discover in this chapter that
the introduction of wing sweep triggered a string of changes to the wing design to
counter the adverse effects that were alluded to in the first paragraph.
In this chapter we mainly focus on the effect of sweep on the aerodynamic prop-
erties of the wing. Alternative texts, such as Anderson [4], present the effects and
associated aerodynamic theory of finite wings in subsonic flow. We presume that
the reader is familiar with this theory and therefore try to extend his/her knowl-
edge to wings that are swept. Section 8.2 presents the aerodynamic rationale behind
swept wings and discusses the inviscid theory of infinite swept wings. In Sect. 8.3
we present the effect of wing sweep on the streamline curvature over infinite wings
as well as over finite wings. In Sect. 8.4 we revisit the flow in the boundary layer over
a swept wing. We shall discover that different transition mechanisms exist and that
the spanwise flow of the boundary layer can be a dangerous source of separation.
Finally, in Sect. 8.4 we address the adverse aeroelastic effects that are encountered
on swept-wing aircraft.
The effect of sweep on drag reduction has long been known in transonic/supersonic
aircraft, well before the invention of the supercritical airfoil and development of
the area rule. It was in fact in the Fifth Volta Conference in 1935 in Rome, Italy
where Adolf Büsemann outlined the theory of swept wings and their aerodynamic
characteristics in supersonic flow. Büsemann argued that it was the normal component
of flow to the wing leading edge that dictated its high-speed performance. He thus
argued that swept wings experience a delayed compressibility effect and as a result
postpone drag-divergence behavior observed in unswept wings in transonic flow.
Schlichting [57] shows the drag polar of a wing with zero-sweep and another wing
with 45◦ of sweep at transonic Mach numbers in Fig. 8.1. Side-by-side comparison
of drag at constant lift in Fig. 8.1a, b shows the dramatic reduction in drag as a result
of wing sweep.
Note that the effect of sweep, as depicted by Schlichting, also introduces the effect
of aspect ratio since both wings have equal structural span. Therefore, we may ask for
8.2 The Advantages of Wing Sweep 429
(a) (b)
1.0
M∞=0.7
0.8 M∞=0.7
0.6
M =0.9
0.4
CL M ∞=0.9 M∞
Λ = 45
0.2
0.0
-0.2
-0.4
0.0 0.1 0.2 0.3 0.0 0.1 0.2 0.3
CD CD
Fig. 8.1 The effect of sweep on drag polar in transonic speeds (data from [57]). a Λ = 0◦ .
b Λ = 45◦
0
-0.2 0 0.2 0.4 0.6
CL
8
V
8
v
8
8
w Voblique w’ V =0
u
u V u’
8
8
Fig. 8.3 Oblique flow and straight flow over a wing section with constant airfoil
Now that we have seen the favorable effect of wing sweep in transonic conditions,
let us investigate the effect of wing sweep on the flow over a wing following Ringleb
[51]. In Fig. 8.3 we see two straight wings of infinite span. The left wing is swept
such that the airfoil is exposed to oblique flow. In other words, the direction of the
freestream velocity vector is under an angle Λ with respect to the chordwise direction
of the airfoil. The airfoil on the right is in straight flow. Both wings experience the
same freestream velocity, V∞ .
We investigate the velocity at a random point somewhere in the flow field over
both airfoils. The velocity over the oblique wing is influenced by the curvature of
the airfoil, resulting in velocities u and w. In addition, the spanwise component of
the freestream velocity, v∞ remains constant. This component does not encounter
any curvature and therefore remains constant over the wing. The total velocity can
therefore be written as:
2
Voblique = u 2 + w2 + v∞ 2
(8.1)
For the wing in straight flow we have a similar expression without the cross-flow
component:
2
VΛ=0 = u 2 + w2 (8.2)
u∞
Because the velocity, V∞ is related to u ∞ according to V∞ = cos Λ the velocity
components scale with the same factor:
u
u = (8.3a)
cos Λ
w
w = (8.3b)
cos Λ
8.2 The Advantages of Wing Sweep 431
The pressure coefficient in incompressible conditions at any point in the flow can be
calculated according to:
V 2
Cp = 1 − (6.11)
V∞
Substitution of (8.5) in (6.11) results in the following relation between straight wing
pressure coefficient, C p, Λ=0 , and the oblique wing pressure coefficient, C p, oblique
(see Problem 8.1):
C p, oblique = C p, Λ=0 cos2 Λ (8.6)
We can extend the result of (8.6) to the lift coefficient and pitching-moment coeffi-
cient of the airfoil by employing (3.30) and (3.35), respectively. This results in the
following relations:
The derivation shown above is known as the simple sweep theory. Let us briefly
reflect on the results that have been obtained. We have now shown that the super-
velocity as well as the absolute value of the pressure coefficient at any point in the
flow about the oblique airfoil is lower than at the same point in the flow about the
straight airfoil. If we remember our discussion on airfoil aerodynamics in Chap. 7,
we can directly deduce that the formation of supersonic flow is delayed to higher
freestream Mach numbers for the airfoil in oblique flow. The critical Mach num-
ber therefore increases. The same argument can be applied to the drag-divergence
Mach number. Due to the lower supervelocities over the airfoil in oblique flow, a
higher freestream Mach number is required to yield strong shock waves that induce
separation at the shock foot. Therefore, drag divergence is postponed to a higher
freestream Mach number (this will be elaborated in Sect. 8.2.2). If we go back to
our discussion on the relation between shape and pressure distribution (Sect. 6.2) we
understand why the result of our analysis makes sense. If we assume that the flow
over the airfoil continues along the direction of the freestream velocity vector, we an
imagine that the flow ‘sees’ an airfoil with a smaller thickness-to-chord ratio. The
smaller thickness-to-chord ratio induces lower supervelocities, reducing the absolute
value of the pressure coefficient, which is in line with the findings of (8.6). A simple
432 8 Aerodynamics of Swept Wings
alternative for the swept wing would be to simply reduce the thickness of a straight
(λ = 0) wing. However, this would result in a heavier structure to carry aerodynamic
loads on the wing (see Problem 8.4).
To demonstrate that simple sweep theory can be used to estimate the pressure
distribution over swept wings once the pressure distribution over a straight wing is
known, we consider Fig. 8.4 from Abbott and von Doenhoff [1]. This figure shows
an experimental setup of a straight wing which is positioned at three different sweep
angles in a wind tunnel: Λ = 0◦ , Λ = 20◦ and Λ = 40◦ . The pressure is measured
over the center cross section at each sweep angle. The resulting pressure coefficient
is scaled back to the equivalent pressure coefficient of the straight wing by using
(8.6). The result is plotted on the right of Fig. 8.4. This shows that the simple sweep
theory indeed gives a good relation between the pressure distributions of the swept
and straight wing. The discrepancies between the distributions could be explained by
the effect of the wind tunnel walls, which force the stream lines at the wall to remain
parallel to the wall. We will see in Sect. 8.3 that in reality the flow over a swept wing
causes the streamlines to curve over the planform. The wind tunnel walls prevent
this curving, which influences the pressure distribution.
So far we have only considered the effect of sweep on infinite wing sections. In
Fig. 8.4 we already see that effects at the boundary have an effect on the pressure
distribution. In practice, swept wings have a finite span and are often tapered (meaning
the tip chord is smaller than the root chord). A tapered wing does not have a uniform
sweep angle. The local sweep angle depends on the chordwise position and varies
between the leading edge and the trailing edge. It is therefore impossible to use the
simple sweep theory directly on a swept wing that is also tapered. Therefore, we need
to choose a reference sweep line if we want to predict the pressure distribution about
an airfoil in a swept wing. From a practical point of view it has been demonstrated
that using the quarter-chord sweep line (Λc/4 ) gives good results in predicting the
Cp
80 mm
-0.5
+1.0
80
Fig. 8.4 Comparison between pressure distribution over a straight wing and a swept wing at
Λ = 20◦ and Λ = 40◦ for constant normal Mach number (after Ref. [1])
8.2 The Advantages of Wing Sweep 433
Λ=45°
-2.0
-2.0
-1.6 η=0.50
-1.6
η=0.815 -1.2
-1.2 Experiment, η=0.815 Experiment, η=0.50
Theory - chordwise (64A010) Theory - chordwise (64A010)
Theory - streamwise (64A007) -0.8 Theory - streamwise (64A005)
-0.8
cl=0.4 cl=0.26
-0.4
-0.4
0 0
0.4 0.4
0.8 0.8
0 .2 .4 .6 .8 1.0 0 .2 .4 .6 .8 1.0
Chordwise position, x/c (~) Chordwise position, x/c (~)
Fig. 8.5 Comparison of theoretical section pressure distributions with experimental loadings on
finite wing panels (after Ref. [30]). a 45◦ swept wing with NACA 64A010 sections normal to the
quarter chord line. b 60◦ swept wing with NACA 64A010 sections normal to the quarter chord line
drag-divergence Mach number. This is a line that connects the quarter-chord points on
the chord line of each individual airfoil in the wing. Figure 8.5 shows the comparison
between the theoretical prediction and the experimentally obtained results for the
same lift coefficient. The solid theoretical line has been obtained by calculating the
chordwise pressure distribution about the airfoil (in this case NACA 64A010) and
applying (8.6). We see that it matches closely to the measured pressure distribution.
The dashed line is a prediction based on an equivalent “sheared” wing. The difference
between these two approaches becomes clear in the next section.
In the previous discussion we have assumed that airfoils are defined perpendicular
to the wing LE. Hence we have assumed an unswept wing that is rotated about a
pivoting axis in order to orient its airfoils at an angle with respect to the freestream
flow (hence have them experience a lower velocity component than the freestream
434 8 Aerodynamics of Swept Wings
Λ Λ
Airfoil perpendicular to LE
Airfoil parallel to freestream
Fig. 8.6 Generation of a swept wing from a straight wing: rotating or shearing (sketch: G. La
Rocca)
by a factor cos ΛLE ). However, it is often industry practice to design a swept wing by
shearing the airfoil in parallel planes to the freestream rather than pivoting the wing.
The difference between the two cases is demonstrated in Fig. 8.6. In the case of a
sheared wing, the effective flow component (u ∞ ) does not see the selected airfoil, but
a similar one with a higher thickness-to-chord ratio (actually with the same thickness
but shorter chord). The theory that is presented in the previous section then applies
to the airfoil perpendicular to the leading edge. By shearing the wing forward or
backwards we always have the same airfoil in streamwise direction. This allows us
to investigate the effect of wing sweep on the drag divergence Mach number for an
airfoil at constant angle of attack.
Let us first compare the sheared wing to the oblique (or “yawed”) wing in terms
of the lift coefficient. If we consider the airfoil to be thin we can replace it by an
infinite vortex sheet [4]. For an unswept wing, the vortex vectors are perpendicular
to the oncoming flow resulting in the following value of the two-dimensional lift
coefficient: c
L Λ=0 = ρ∞ V∞ γ(x)dx (8.8)
0
For a swept wing the vorticity vector, γ is inclined at an angle π/2 − Λ. Therefore,
the two-dimensional lift vector, L , can be written as:
c
L sheared = ρ∞ V∞ γ(x) cos Λdx (8.9)
0
We therefore conclude that the lift coefficient of the sheared airfoil relates to the lift
coefficient of the unswept airfoil as:
Note that this result is different from what we found for the oblique wing, i.e. (8.7a).
This makes sense because for the oblique wing we evaluated the lift coefficient of the
reference airfoil of the unswept wing. This airfoil is not being deformed in streamwise
direction when the wing is being sheared, while it is stretched when the wing is being
yawed. Depending on the way a wing is designed, one of the two prediction methods
should be used to evaluate the effect of wing sweep on the aerodynamic coefficients.
Now that we have defined this alternative form of the sweep theory, let us return
to the drag-divergence Mach number. The Mach number at which drag divergence
starts is strongly dependent on the Mach number in front of the shock wave. As we
know from Chap. 7, the wave drag is directly proportional to the momentum loss
over the shock wave, which, in turn, depends on the static pressure increase over
the shock. To make a first estimate of the drag divergence Mach number, Mdd , we
can rely on a crude but simple empirical relationship between the drag divergence
Mach number (defined here as the Mach number for which ∂C D /∂ M = 0.05) and
the critical Mach number of a finite swept wing [59]:
Mdd = Mcrit 1.02 + 0.08(1 − cos Λc/4 ) (8.11)
To find the critical Mach number we employ a modified procedure to the one
outlined in Chap. 3 (Fig. 3.4). We consider a nonlifting swept wing at zero angle
of attack. First we modify the relation between the critical pressure coefficient and
the critical Mach number, (3.42). We replace the critical freestream Mach number,
Mcrit , by Mcrit cos Λ in accordance with simple sweep theory. We obtain the following
relation: ⎡ γ ⎤
γ−1 2 2 Λ γ−1
2 ⎣ 1 + M cos
C pcrit = 2
γ+1
crit
− 1⎦ (8.12)
γ Mcrit
2
2
ξ
c
η
x
We can subsequently write the full potential equation (3.2) in the rotated coordi-
nate system for two-dimensional flow:
Φξ2 Φξ Φ z Φz2
1− Φξξ − 2 2 Φξz + 1 − 2 Φzz = 0 (8.14)
a2 a a
Φξ2 − Φz2
1− =0 (8.16)
a2
Φξ = Vξ = V∞ cos Λ + û (8.17a)
Φz = Vz (8.17b)
Note that we have used Vξ and Vz as the velocity components parallel to the ξ and
z-axis, respectively. Substituting this in (8.16) results in the following expression for
the critical condition:
Note that in this equation it is implied that v (spanwise velocity component tan-
gent to the leading edge, V∞ sin Λ) does not change over the wing; an assumption
of simple-sweep theory. When examining the critical condition (8.18) we observe
that only the velocity component normal to the direction of sweep ( Vξ2 + Vz2 )
8.2 The Advantages of Wing Sweep 437
must be sonic. The total velocity, V , may exceed sonic speed at critical conditions:
V 2 = a 2 + V∞ 2 sin2 Λ ≥ a 2 . Pure subsonic flow over a wing exists until the veloc-
ity along the peak suction line becomes sonic. For unswept wings this is also the
end of the subcritical regime. However, when the wing is swept, subcritical flow
exists until the component perpendicular to the peak-suction line becomes sonic.
For swept wings this marks the end of the subcritical regime. In other words, a swept
wing can be in subcritical conditions even when local supersonic flow exists. While
being in the subcritical regime, the flow cannot have any shock waves. This intu-
itively demonstrates the effect of wing sweep on the drag divergence boundary: the
subcritical domain is extended, such that (strong) shocks occur at higher freestream
Mach numbers.
To investigate the effect of the sweep angle on the compressibility factor, β (3.18),
let us return to (8.14). In the limit of small perturbations, where we neglect the squares
and products of small parameters, i.e. the second order terms, and for freestream Mach
numbers that exclude transonic and hypersonic flows (8.14) may be written as:
2
Vξ
β Φξξ + Φzz = 0 with β = 1 −
2 2
(8.19)
a
V − V∞2 sin2 Λ
β2 = 1 − (8.20)
a2
Remember that for isentropic flow the following holds:
γ−1 2
a 2 = a∞
2
− (V − V∞
2
) (3.8)
2
Substitution of (3.8) in (8.20) and subsequent manipulation yields the following
expression:
2
cos2 Λ + VV∞ − 1
β 2 = 1 − M∞ 2 (8.21)
2
γ−1
1 − 2 M∞ V∞ − 1V
many cases the velocity will be known for incompressible flow. The velocity term
can then be replaced by the incompressible pressure coefficient, −(C p ) M∞ =0 =
(V /V∞ )2 − 1. The final expression for β 2 then becomes:
β 2 = 1 − M∞
2
cos2 Λ − (C p ) M∞ =0 (8.22)
438 8 Aerodynamics of Swept Wings
−1
Pressure Coefficient, C p (~)
−0.8
Λ=0
30
45
−0.6 60 60
45
30
Λ=0
Cp
−0.4
(C p ) = -0.5 Cp
M =0 crit
8
−0.2
0
0.5 0.6 0.7 0.8 0.9 1
Free−Stream Mach Number, M (~)
8
curves that show the effect of the freestream Mach number on the pressure coefficient.
With increasing sweep angle we observe that the compressibility correction becomes
weaker. In other words, the change in pressure coefficient with Mach number is
lessened. Finally, the intersection between the two curves of the same sweep angle
marks the critical Mach number. We see that the critical Mach number increases
from M∞ = 0.68 when Λ = 0 to M∞ = 0.87 when Λ = 60◦ .
We have now established the relation between the critical Mach number, the sweep
angle, and the incompressible pressure coefficient. If we set the right-hand sides of
(8.12) and (8.23) equal to each other we can deduce the critical Mach number for a
given incompressible pressure coefficient and sweep angle. If we do this for several
values of the incompressible pressure coefficient, we can make a chart where we relate
the critical Mach number to the sweep angle for a given value of the incompressible
pressure coefficient. Such a chart is presented in Fig. 8.9. We can use this chart in
combination with (8.11) if we would like to determine how much sweep is required
for a given drag-divergence Mach number. The following example demonstrates the
use of the chart.
Example 8.1 Many transonic aircraft cruise at or slightly beyond their drag-
divergence Mach number. Assume we desire a cruise Mach number of 0.85 and
have an airfoil which has a peak suction pressure coefficient of C p = −0.35 at low
Mach number and zero angle of attack. Calculate the sweep angle of the airfoil such
that drag divergence occurs at M = 0.85.
(C p ) -1.0 -.90 -.80 -.70 -.60 -.55 -.50 -.45 -.40 -.35
M =0
60
8
50 -.25
-.20
Sweep Angle, Λ (deg)
40 -.15
-.10
30
-.05
20
10
0
0.5 0.6 0.7 0.8 0.9 1
Critical Mach Number, M (~)
crit
Fig. 8.9 Critical Mach number for two-dimensional flow based on Weber’s compressibility
correction (8.22)
440 8 Aerodynamics of Swept Wings
Solution:
To solve this problem, we need to combine the chart of Fig. 8.9 and Eq. (8.11). We
use Λ = Λc/4 . Because both the chart and the formula contain critical Mach number
and the sweep angle, the equation cannot be solved in a closed expression. Therefore,
we need to solve this problem iteratively. We start with a first guess for the critical
Mach number, making sure it is somewhat below the drag-divergence Mach number:
Mcrit = 0.8. We find a corresponding sweep angle of Λ = 32◦ (see Fig. 8.10). We
then employ (8.11):
Our first guess yields a drag-divergence Mach number that is a little too low. We try
Mcrit = 0.82, which gives a corresponding sweep angle of Λ = 36◦ (see Fig. 8.1).
With (8.11) we find:
We have now found a sweep angle that yields a drag divergence Mach number that
is very close to the desired cruise Mach number. The iterative solution is illustrated
in Fig. 8.10.
60
(C p ) = -0.35
M =0
8
50
Sweep Angle, Λ (deg)
40
30
first iteration
20
initial guess
10
0
0.5 0.6 0.7 0.8 0.9 1
Critical Mach Number, Mcrit (~)
Fig. 8.10 Graph of sweep angle versus critical Mach number belonging to Example 8.1
8.2 The Advantages of Wing Sweep 441
At speeds close to the drag-divergence Mach number we would like to calculate the
change in drag coefficient with Mach number. We therefore use a modified version of
the transonic similarity laws of Chap. 3. First step is to construct the wave-drag versus
Mach number graph for the unswept wing. We can do this by employing Fig. 3.21
for a given value of thickness-to-chord ratio, τ , and aspect ratio, A. Based on this
graph, we read a value for the maximum value of the wave drag coefficient, C Dw, max ,
and the corresponding Mach number (MC Dw, max , Λ=0 ) as well as the drag-divergence
Mach number. The latter one can be obtained by using the definition of Mdd from
(7.1): ∂C D /∂ M = 0.10.
Now, for a wing with sweep these values need to be converted. We consider a
wing that is pivoted about a hinge point. In the theoretical case of such an oblique
wing, the drag divergence Mach number would scale with cos Λc/4 , similar to the
velocity. In real life, as we will see in the subsequent sections, the flow over the wing
is fairly three-dimensional. Therefore, the assumption that the airfoil only sees the
normal component with respect to the quarter chord line is an oversimplification.
The actual drag divergence Mach number should be somewhere between the one for
the airfoil and the one for the oblique wing [5]:
Following the same reasoning we can state that the effective Mach number, Meff , can
be related to the freestream Mach number according to:
For the case of ideal conditions n = 1 [33]. Torenbeek [66] uses n = 1/2. We will use
the latter approximation to estimate the drag divergence Mach number of the swept
wing and the Mach number at which the maximum drag coefficient is obtained:
(Mdd )Λ=0
(Mdd )oblique = (8.26a)
cos Λc/4
(MC Dw, max )Λ=0
(MC Dw, max )oblique = (8.26b)
cos Λc/4
The drag wave-drag coefficient of the oblique wing can be correlated to the unswept
wing according to [31]:
(C Dw, max )oblique = (C Dw, max )Λ=0 (cos Λc/4 )2.5 (8.27)
We can now construct the line of M versus C Dw . We have the combination of Mach
number and wave-drag coefficient at the peak value. In addition, we have the position
442 8 Aerodynamics of Swept Wings
Determine the variation of the wing’s wave drag coefficient with Mach number
between M = 0.80 and M = 1.4.
Solution:
To solve this problem, we first determine the variation of the wave drag coefficient
with Mach number for an unswept wing. We rely on the relation shown in Fig. 3.21
where the zero-lift drag coefficient for round-nose airfoils is related to Mach number,
thickness ratio, and Mach number. We need to calculate the following quantity:
Aτ 1/3 = 1.568
0.03
unswept
0
0.8 0.9 1 1.1 1.2 1.3 1.4
Mach number, M (~)
Fig. 8.11 Graph of drag coefficient versus Mach number belonging to Example 8.2
(Mdd )Λ=0
(Mdd )oblique ≈ = 1.05
cos Λc/4
(C Dw, max )oblique = (C Dw, max )Λ=0 (cos Λc/4 )2.5 = 1.34
(MC D )Λ=0
(MC Dw, max )oblique = w, max = 0.0121
cos Λc/4
To construct the curve that relates the wave drag coefficient of the swept wing to
the Mach number we now have the following information: (1) at M = 1.05 the
slope of the line is ∂C Dw /∂ M = 0.10; (2) at M = 1.34 we know the maximum
drag coefficient is C Dw = 0.0121 and that the slope ∂C Dw /∂ M = 0. With this
information we can construct the C Dw curve for the swept wing. The curve for the
unswept wing can be used as a guide for constructing the curves between the two
points. The result of this exercise is shown by the solid curve in Fig. 8.11.
The approximation that is demonstrated in the previous example is close to the
experimentally obtained prediction indicated by the dash-dotted grey line in the graph
of Fig. 8.11. We notice that the prediction of the drag divergence Mach number is too
high, and that the drag coefficient beyond M = 1 is underestimated by this method.
444 8 Aerodynamics of Swept Wings
On the other hand, this analytic treatment can yield a quick estimation without the
requirement of using computational fluid dynamics.
The minimum drag (drag at zero-lift) of a finite wing is dependent on many
factors: aspect ratio, sweepback angle, taper ratio, airfoil section family, thickness-
to-chord ratio, twist distribution, dihedral, Reynolds number, and Mach number. We
would like to know what the effect is of each of these factors on the transonic drag
performance of the wing. To reduce the complexity of this problem, we follow Ref.
[35] and experimentally investigate the effect of sweep, thickness-to-chord ratio, and
Mach number on the minimum drag coefficient, C Dmin . We do this for one family of
airfoils: NACA 65A0XX, where XX represents the value of the thickness-to-chord
ratio. Furthermore, we set the aspect ratio to 4 and the taper ratio to 0.6 and we
consider wings without dihedral or twist. Finally, we evaluate the drag properties
at Reynolds numbers higher that 10 million. From the experimental results we can
now plot the minimum drag as a function of the three independent variables. This is
shown in Fig. 8.12. Note that this graph is constructed based on experimental results
from various wind tunnel tests as described in Ref. [35].
The graph in Fig. 8.12 shows both the effect of sweep and thickness-to-chord
ratio on the minimum drag coefficient in the transonic realm. If we consider the
unswept wing (Λc/4 = 0◦ ) we see an increase in drag coefficient by a factor of about
20 between M = 0.7 and 1.0 for the airfoil of 12 % thickness. By decreasing the
thickness-to-chord ratio we observe three effects: the drag-divergence Mach number
increases, the drag peak decreases, and the absolute value of the slope dC Dmin /dM
decreases. This can all be explained based on the two-dimensional aerodynamics:
lower thickness leads to lower supervelocities and therefore weaker shock waves.
A lower thickness-to-chord ratio is therefore a powerful means to reduce excessive
drag in the transonic domain.
We now turn to the effect of the sweep angle. Let us consider the wings with the
thickest profiles (t/c = 12 %). By comparing the lines at Λc/4 = 0◦ and Λc/4 = 60◦ ,
we see a reduction of the peak drag coefficient of 80 % when sweeping the wing
backwards over 60◦ . The effect of sweepback is lessened for the thinner wings. For
a wing with 9 % thickness the reduction in peak drag coefficient is around 60 %.
Apart from the absolute value of the drag peak, we also distinguish a backwards
shift in drag peak and the drag divergence Mach number. Both can be explained if
we remind ourselves that the critical conditions on the wing are determined by the
velocity component perpendicular to the local isobars on the wing. On a swept wing
of infinite span these isobars run parallel to the local sweep angle of constant-chord
points. Due to the increase in sweep angle we therefore postpone the Mach number
at which critical conditions occur and simultaneously push back the drag-divergence
Mach number as well as the peak in drag coefficient.
Figure 8.12 gives a quantitative view on how the transonic drag behavior can
be altered by modifying either the thickness-to-chord ratio or the sweep angle. In
practice, considerations on wing weight, fuel tank volume, aeroelastic character-
istics, control characteristics, and low-speed performance dictate which combina-
tion of these two parameters yields the best compromise. Depending on the type of
8.2 The Advantages of Wing Sweep 445
.09
.08
.07
.06
0
.05
Sw
15
eep
.04
an
gle
.03
at
30
qu
art
.02
e rc
ho
45
rd,
.01
Λ c/4
(°)
60 .2 0
.3 .4 .5 .6 .7 .8 .9 1.0 1.1 1.2 1.3 1.4 1.5 1.6
Mach number, M (-)
Fig. 8.12 Effect of sweep and thickness ratio on minimum drag coefficient (after Ref. [35])
aircraft this could yield thick wings with relatively much sweep (e.g. high-subsonic
transports) or thin wings with zero sweep (e.g. supersonic interceptors).
If one knows the drag-coefficient behavior of a given airfoil section (2-D) as a
function of Mach number and lift coefficient, Ref. [32] gives an alternative method
to calculate the drag-coefficient versus Mach number for a 3-D wing. An important
446 8 Aerodynamics of Swept Wings
assumption for this method is that the spanwise loading of the wing is neglected
and the airfoil in the wing experiences an average lift coefficient. In this method
it is assumed that the change in airfoil section wave drag coefficient due to Mach
number and lift coefficient is known, i.e. cdw = f (MΛ=0 , clΛ=0 ). Such data could,
for example, be obtained from experimental tests on wing profiles or from two-
dimensional airfoil analysis. In addition, one should also know how the section
minimum drag coefficient, cd0 , is affected by the sweep angle. Reference [75] shows
that the profile drag coefficient of a swept NACA 64-012 airfoil section can be
approximated as follows:
√
cd0 (Λ) = cd0 , Λ=0 cos Λ for 106 < Re < 108 (8.28)
If cd0 (Λ) and cdw (M, cl ) quantities are known, the wing drag can be estimated
according to [32]:
C D = cd0 cos Λ + cdw cos3 Λ + C Di (8.29)
The induced drag, C Di , is a function of the lift coefficient and spanwise loading of
the wing and is independent of the Mach number in subsonic conditions. The vortex
induced drag coefficient of a finite wing is inversely proportional to wing aspect
ratio, A, in low subsonic Mach numbers and directly proportional to the square of
lift coefficient:
C L2
C Di = (8.30)
πAφ
The symbol φ is the span efficiency factor due to non-elliptic wing loading.1 For
elliptical lift distribution φ = 1 (minimum induced drag coefficient). In all other
cases φ is less than one. Plotting (8.29) constitutes the wing drag polar. The data
of Ref. [60] shown in Fig. 8.13a supports the dominance of parabolic behavior of
C D versus C L and the inverse relationship with aspect ratio at low subsonic Mach
numbers (e.g., M∞ = 0.5). The lowest wing drag coefficient is seen to occur at the
highest aspect ratio. Now, examine the drag polar of the same wing, but at a high
subsonic Mach number of M∞ = 0.875 in Fig. 8.13b. Due to the dominant wave
drag, the drag polar behavior seems to be turned upside down. The high aspect ratio
wing (e.g., A = 9) shows the highest drag and vice versa. In this high-subsonic case
we see that for high-aspect ratio wings a small variation of lift coefficient results in a
large variation of C D (note the higher scale of drag coefficient in 8.13b versus 8.13a).
This behavior is due to the appearance of shocks on the wing and subsequent flow
separations causing a loss of lift and rise in drag. Here the wing drag is not dominated
1 The Oswald factor, e, of a full airplane is not the same as the span efficiency factor. However, the
two are related according to 1e = X π A + φ1 . X is the coefficient that relates the profile and parasite
drag to C L2 . For more information on this relation, the reader is referred to Ref. [53].
8.2 The Advantages of Wing Sweep 447
(b) 0.09
A=9
0.08
5
0.07
0.06
0.05
CD
3
0.04
(a)
0.03 0.03 2
A= 2 3 5 9
0.02 0.02
CD ∞
0.01 0.01
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 0.1 0.2 0.3
CL CL
Fig. 8.13 The effect of wing aspect ratio on drag polar of an unswept wing with NACA 0012 airfoil
at freestream Mach numbers of 0.5 and 0.875 (after Ref. [60]). a M∞ = 0.5. b M∞ = 0.875
by the contribution of vortex lift to drag, but rather by the shock stall behavior of the
airfoil section and the associated compressibility drag. The low aspect-ratio wings
(2 and 3) have reduced drag compared to the higher aspect ratio wings (5 and 9).
This change in characteristics is associated with the delayed and less rapid rise of
drag as the aspect ratio is decreased [60].
The wing drag coefficient of a 6.4 aspect ratio wing composed of NACA airfoil
section [65(112) −213] for three constant-lift coefficient cases in the range of transonic
Mach numbers is shown in Fig. 8.14. The figure shows the critical Mach number for
the three constant C L cases. A seemingly constant plateau before Mcrit is followed by
a rapid and dramatic (nearly 7-fold) rise in drag up to sonic flight, where seemingly
another plateau is reached. The reader should note that the classical data shown in
Figs. 8.13 and 8.14 pre-date the supercritical airfoil era.
Implicitly, C D is dependent on the lift coefficient, C L , and Mach number, M
through the section wave drag coefficient, cdw . To correlate these wing properties to
the airfoil properties we use simple sweep theory and the assumption that the lift
coefficient of the airfoil section can be correlated to the average wing lift coeffi-
cient [32]:
448 8 Aerodynamics of Swept Wings
0.08
0.06
0.04
Mcrit
0.02
0
0.6 0.7 0.8 0.9 1.0 1.1
Mach number, M∞ (∼)
CL
clΛ=0 = (8.31a)
cos2 Λ
MΛ=0 = M cos Λ (8.31b)
For each combination of C L and M, one can now find a value for the section drag
coefficient, cd (M, C L ) through the drag data of the section. We can subsequently
find cdw as follows:
cdw = cd − cd0 (8.32)
In Fig. 8.15 the comparison between measured drag data and the estimation of (8.28)
through (8.32) is shown for three wings with moderate sweep angle (30◦ and 35◦ ).
The two-dimensional data on which the estimation is based, stems from experimental
tests on airfoils and is therefore considered to be reliable. We see that the prediction
of the drag-divergence Mach number is within 10 % of the measured drag-divergence
Mach number. Even though the correlation is not very good, we do see the effect of
the wing’s lift coefficient on the drag-divergence Mach number and also the shape
of the estimated drag curve matches well with the experimentally obtained results.
In Ref. [32] it can be seen that for wings of higher sweep angle the correlation
between estimation and experimental results deteriorates (see Problem 8.6). With the
estimation of (8.29) it is possible to find a drag-divergence Mach number well beyond
1. In practice, it is shown in [32] that there exists a limit in drag-divergence Mach
8.2 The Advantages of Wing Sweep 449
(a) (b)
.08
Drag coefficient, CD (~)
CL=.6
.06
.04
.4 CL=.6
.02 .4
0
0
0
0 .2 .4 .6 .8 1.0 1.2 0 .2 .4 .6 .8 1.0 1.2
Mach number, M∞ (~) Mach number, M∞ (~)
(c) (d)
.08
Drag coefficient, CD (~)
.06
.04
CL=.6
.02 .4 CL=.4
0 0
0
0 .2 .4 .6 .8 1.0 1.2 0 .2 .4 .6 .8 1.0 1.2
Mach number, M∞ (~) Mach number, M∞ (~)
Fig. 8.15 Drag coefficient versus Mach number for three wings based on Eq. (8.29) (after Ref.
[32]). Solid line represents experimental data while dashed line represents the prediction according
to (8.29). Note λ is the taper ratio of the wing. a Λc/4 = 35, A = 3.5, λ = 0.2. b Λc/4 = 35,
A = 5.1, λ = 0.71. c Λc/4 = 30, A = 7.4, λ = 0.38. d Λc/4 = 35, A = 10, λ = 0.5
number of about 0.95. Clearly, more advanced methods (e.g. CFD) are required to
accurately predict the drag behavior of arbitrary wing shapes at high Mach numbers.
It has now been established that wing sweep has a favorable effect on the high-
subsonic drag characteristics of wings. For the remainder of this chapter we will
illuminate some of the challenges associated with swept-wing design. In this section
we will start by examining the inviscid flow over a swept wing. Remember that
the flow over a body can essentially be divided in two categories: inviscid flow and
viscous flow. As stated before, viscous forces are only important in the boundary
layer and in the thin region around the shock wave. Outside these regions, the flow
can essentially be treated as being inviscid. We investigate how sweep influences
the inviscid flow field. We first present the effect of wing sweep on the streamline
curvature over infinitely swept wings. Subsequently, we show how the presence of a
450 8 Aerodynamics of Swept Wings
wing tip and a wing center section affects the streamline curvature. Furthermore, we
will demonstrate how the three-dimensional wing can be adapted in order to reduce
the adverse root and tip effects that are caused by wing sweep.
If we assume that the inviscid crossflow velocity over an oblique wing is constant and
amounts to v∞ , then we can calculate the flow direction over the wing by comparing
the local flow component in chordwise direction to the flow component in spanwise
direction (crossflow component). The angle between these two values indicates the
direction of the flow: v
∞
Λlocal = tan−1 (8.33)
u
If we know the local sweep angle of the streamline in the plan view of the wing, we
can also determine the path of the streamline over the wing. The following example
shows how this path can be determined.
Solution:
To determine the pressure distribution over the airfoil we choose to use a two-
dimensional Euler solver with an uncoupled boundary-layer solver.2 The pressure
distribution is shown in Fig. 8.16a.
Based on the C p distribution we calculate the velocity distribution about the airfoil
by using (3.11) and assuming that the flow is incompressible:
Voblique
= 1 − Cp
u∞
2 The solver that is used is MSES. See Refs. [21, 26] for details on this viscous-inviscid solver.
8.3 Inviscid Flow over Swept Wings 451
M = 0.2 1.3
−1 c l = 0.5
pressure coefficient, C (~)
−0.5 1.1
1
0
0.9
0.5 0.8
NASA SC(2) 412 0.7
1
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
chordwise position, x/c (~) streamwise position, x/c’ (~)
Fig. 8.16 Predicted pressure and velocity distributions about NASA supercritical airfoil. a Pressure
distribution. b Velocity distribution
The resulting velocity distribution is shown in Fig. 8.16b. Note that we denote the
streamwise chord length with c , where c = c/ cos Λ.
We employ (8.33) to calculate the local sweep angle of the stream line. Since both
u and v∞ are known, this is a straight-forward operation. To calculate the path of a
particle, we subsequently assume that a particle on the stream line follows a straight
line between two grid points. This line has an angle Λlocal, i between xi and xi+1 .
We calculate the displacement in y direction as follows:
yi+1 yi xi+1 − xi
= + tan Λlocal, i
c c c
In addition, we assume that y1 = 0. We now have a displacement yi at every chord-
wise airfoil point, xi . We can now plot the stream line path in the coordinate system
of the airfoil. This has been shown in Fig. 8.17. However, we have rotated the coor-
dinate system of the airfoil over an angle Λ, such that the streamlines can easily be
compared to the freestream direction of the flow.
If we examine the streamline pattern over the swept wing of Fig. 8.17, we observe
that the upper and lower streamlines do not follow the same path. While the lower-
surface streamline stays close to the streamline direction of the free stream, the
upper-surface streamline clearly bends in the negative y-direction. If this wing were
aft-swept, the upper-surface streamline would therefore bend towards the wing root.
Even though we are aware of the fact that simple-sweep theory is not a very good
representation of the real flow conditions, the example above intuitively shows why
inviscid streamlines over a swept wing curve inboard or outboard depending on the
local pressure distribution.
452 8 Aerodynamics of Swept Wings
8
x V
c
V
8
V
8
V
8
c’
Fig. 8.17 Plan view of inviscid stream lines on upper and lower surface of oblique wing according
to simple sweep theory
Tip region of
sweptback wing
Centre region of
sweptback wing
In Fig. 8.18 a sketch of the streamline pattern over a finite swept wing is presented.
This wing has a symmetric airfoil profile, no twist, and is positioned in the flow at
zero angle of attack. In other words, the streamline pattern on the upper and lower
surface is purely caused by the thickness distribution of the airfoil and is identical
on the lower and upper side of the foil. If we zoom in on the “infinite” portion of the
wing (far away from root and tip), we indeed see the same streamline paths as for
8.3 Inviscid Flow over Swept Wings 453
le
ad
in
g
ed
ge
isobars
fluid particle
direction of motion
direciton of steepest
pressure gradient
Cp
=
0. .0
5
0 .5
-0 0 (su
-1
.
tra
-0
ili
.5
ng
ct
io
ed
n
ge
0.
pe
0
ak
)
Fig. 8.19 Notional isobar pattern over an infinitely long swept wing
the upper surface of the airfoil in Example 8.3. We see that the streamlines ahead of
the wing curve outboard. Over the leading edge of the airfoil the flow experiences
the most inboard curvature, while further downstream this is reduced.
The curvature of the streamlines over infinitely swept wings can be explained by
the three-dimensional pressure distribution over the symmetric wing. If we would
map the pressure distribution onto the wing’s surface, we would have a contour plot
of the isobars. The isobars would all be aligned with the sweep angle of the wing
(see Fig. 8.19). Any particle flowing over the leading edge is now being subjected
to a pressure gradient that is highest in the direction perpendicular to the isobars.
Therefore, the path of the fluid particle ahead of the suction peak is curved in the
direction of the strongest pressure gradient: inboard on an aft-swept wing. Behind
the suction peak there exists an adverse pressure gradient that is again highest per-
pendicular to the isobars. This gradient has a component which is in the outboard
direction. Therefore, the particle that is experiencing this adverse pressure gradient
is pulled more outboard on an aft-swept wing. This results in the s-shaped path that
a fluid particle describes over a swept wing under inviscid conditions.
In Fig. 8.20 we show the spanwise lift distribution over two plain semi-wings of finite
span (aspect ratio 8) without any taper, twist or camber. These results have been
computed by Matrics V, a viscous-inviscid flow solver based on the full potential
equation (2.199) for the inviscid solution [67, 68]. The first wing has a sweep angle
of 35◦ (aft-swept) while the second wing has a sweep angle of −35◦ (forward-
swept). Both wings are at a lift coefficient of C L = 0.5. As a reference, we have
454 8 Aerodynamics of Swept Wings
0
0 0.2 0.4 0.6 0.8 1
Spanwise location, y/(b/2) (~)
also plotted the elliptical lift distribution for a wing of C L = 0.5 We see that the lift
distribution of the aft-swept wing is quite different from the forward swept wing. The
center of lift on the aft-swept wing is shifted more outboard. We also notice that the
lift distribution over the forward-swept wing is much closer to the ideal elliptical lift
distribution. This results in a higher span efficiency factor for the forward-swept wing
(φ = 0.97) than for the aft-swept wing (φ = 0.86). The change in lift distribution
can be explained by the wing sweep. As the center of the aft-swept wing is ahead
of the outboard wing, it effectively generates upwash over the outboard wing that
increases the effective angle of attack of the outboard wing. Vice versa, the tip of
the forward-swept wing is ahead of the inboard wing creating more upwash over
the inboard wing. The effective angle of attack over the inboard wing is therefore
increased leading to a higher local lift coefficient. The sweep angle of the wing
therefore has an impact on the lift-induced drag of the wing. It should be noted that
for both wings a change in taper, twist distribution or camber distribution can result
in lift distribution very close to the elliptical lift distribution.
Let us return to Fig. 8.18 where at the junction of two wing halves (termed center
section, also referred to as root), the requirement of symmetry forces the streamline
to be straight. In other words, there exists no net pressure gradient perpendicular to
the freestream direction at the root. For this statement to be true, the isobars should
be exactly perpendicular to the freestream direction at the root. At the root, the
isobars are therefore completely unswept. This means that the swept isobaric pattern
of Fig. 8.19 should somehow blend with the unswept isobars at the root. In Fig. 8.21
we see that the isobars indeed curve backwards at the root and become completely
orthogonal to the flow direction over the center section. At the tip, there is no strict
flow symmetry requirement. Therefore, the isobars do curve forward but show some
variation. This variation is caused by the three-dimensionality of the flow around
the tip. We see that the isobaric pattern on the wing with aspect ratio 1.05 is in fact
dominated by the presence of root and tip. The wing with aspect ratio 2 has a larger
portion where the isobars are aligned with the sweep angle of the wing. The effect
of root and tip on the pressure distribution over a swept wing are often termed the
root effect and tip effect, respectively.
8.3 Inviscid Flow over Swept Wings 455
A=2
-0 in.
A=1.05
.
0
M5
-0 20
M5
0
5
0
0
-0 5
-0 15
in
.2
.2
.2
.2
.1
.1
.0
.2
.2
.2
.1
.1
.0
1
.
0.
0.
-0
-0
-0
-0
-0
-0
-0
-0
-0
-0
=-
=-
C
C
p
Fig. 8.21 Isobar pattern over two 53◦ swept wings of different aspect ratio and zero lift (after
Ref. [39])
The fact that the isobars are curved at root and tip has a profound effect on the
chordwise pressure distribution at these two stations. The pressure distribution on
two spanwise locations along a swept wing at transonic Mach number is shown in
Fig. 8.22. The nomenclature on the figure indicates that a 9 % thick (symmetrical)
A=6
t / c = 0.09 cC
30
cT / cC = 1/3
M0 = 0.9
α= 0 cT
-0.6
experiment y=0 y/b/2 = 0.5
CFD
-0.4
0
0.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 1.0
x/c x/c
0.2
Fig. 8.22 Comparison between CFD and experiment on pressure distributions at two spanwise
locations on a swept wing at Mach 0.9 (after Ref. [29])
456 8 Aerodynamics of Swept Wings
(a) 1.15
Aspect ratio 1.05 Exp.
Aspect ratio 2
Centre
1.05
1.0
0 0.2 0.4 0.6 1.0
Chordwise station, x/c (~)
t/c=0.12 at 0.3c
1.05
1.00
0 0.2 0.4 0.6 0.8 1.0
Chordwise station, x/c (~)
Centre Mid-semispan Tip
Fig. 8.23 Velocity distributions over swept wing sections (modified from Ref. [39]). a Velocity
distributions at center and tip sections of two 53◦ swept-back wings of constant chord. b Velocity
distributions at three spanwise stations on a 50◦ swept-back wing of constant chord with modified
section shapes
profile is used with a taper ratio of 1/3rd, the wing sweep angle at the mid-chord
is 30◦ , the aspect ratio of the wing is 6 and the (geometric) angle of attack is zero.
The wing is in Mach 0.9 freestream. The symbol C ∗p indicates the critical pressure
coefficient and thus the demarcation between locally subsonic and locally supersonic
flows on the wing at two spanwise locations (note that y = 0 is the wing root and
8.3 Inviscid Flow over Swept Wings 457
y/b/2 = 0.5 is at the mid point of the semispan). The appearance of shocks in both
spanwise locations where the flow transitions from supersonic to subsonic is evident
in the data. The computational method is based on the Transonic Small Disturbance
(TSD) equation. Although in general the trend in pressure distribution is captured, the
CFD prediction misses the peak suction and the chordwise location of the shock foot.
By comparing the two pressure distributions in Fig. 8.22 we see that the shock
wave near the root of the wing is positioned much farther aft than at 50 % of the semi
span. The root effect causes a backward shift in suction peak, a more aft position of the
shock wave, and a steeper adverse pressure gradient behind the shock. Furthermore,
the magnitude of the suction peak is also reduced. As is evident in Fig. 8.23a, the
curved isobars reduce the suction over the front part of the airfoil and increase the
suction over the aft part of the airfoil. This results in a net drag force that can
be attributed to the root effect. The tip experiences the opposite effect. The forward
sweeping isobars induce additional leading-edge suction over the tip of the wing. This
results in a force in opposite direction to the freestream (a thrust force). In addition,
we see in Fig. 8.23a that the suction peak is higher. In high-subsonic conditions, this
means higher supercritical velocities near the leading-edge of the tip and therefore
an early onset of shock formation and shock-induced separation.
The adverse effects at the root and tip can be remedied by modifying the wing
shape such that the chordwise pressure distribution at these locations is almost iden-
tical to the one at the mid point of the semispan. This can be achieved by modifying
planform and/or airfoil parameters. In Fig. 8.23b it can be seen how the thickness
distribution of the airfoils can be modified in order to achieve the same shape of
the pressure distribution at root, mid section, and tip. We see that the thickest point
on the root airfoil has been moved forward to move the suction peak more forward.
Analogously, the thickest point on the tip airfoil has been moved rearwards. We now
have three different airfoils at the three aforementioned spanwise stations. Because
they are part of a three-dimensional swept wing, they can be designed such that the
shape of the pressure distribution in the design condition is the same. By increasing
the thickness of the root airfoil and the tip airfoil, the pressure distribution can be
exactly matched to the pressure distribution of the section at the mid semispan. A
more forward position of the thickest point and an increase of the thickness-to-chord
ratio are indeed two characteristics of root airfoils of modern high-subsonic wings.
A more aft-position of the thickest point on the tip airfoil has a negative effect
on the maximum lift coefficient due to the sharper leading edge that results. To
remedy the tip effect we can therefore also choose to modify the tip planform shape
of the wing. Küchemann and Weber [39] introduced a parabolically curved leading
edge leaving the streamwise section shape unaltered. The curved part of the leading
edge extends inboard by 25 % of the tip chord. The idea behind such a shape is to
increase the leading-edge thrust not by increasing the suction peak, but by reducing
the positive pressure near the leading edge. Such a Küchemann tip does not straighten
the isobars completely but increases their spacing near the leading edge significantly
(see Fig. 8.24). A Küchemann tip can easily be designed by following the guidelines
presented on the right-hand side of Fig. 8.24.
458 8 Aerodynamics of Swept Wings
-0.3
min.
53°
Pe
-0.2
ak
-su
P1 P1
cti
-0.1
on
A
lin
e
B
Cp=0
C
53°
P2
A’ A simple geometric
A=2 construction of the
ϕ =53° curved tip as the
envelope of the
Aerofoil section
lines AA’, BB’, etc
with t/c=0.12 B’ where A, B, and A’, B’
at 30% c.
are equal intercepts
Conventional on P1P 2 and P2P3
rounded tip respectively
Tip with curved C’
leading edge
P3 P3
Fig. 8.24 Effect of planform modifications to isobar pattern at the wing tip (after Ref. [39])
So far, we have discussed modifications to root and tip such that the velocity
distribution due to thickness is constant over the span of the wing. All the pre-
vious examples concern untwisted wings without camber and in symmetric flow
conditions (zero angle of attack). To produce lift, the wing is usually exposed to
an angle of attack (unsymmetric flow conditions). Figure 8.25 shows the pressure
distribution at 5 different spanwise stations of a 45◦ swept wing of aspect ratio 3.
This wing has a symmetric airfoil section and exhibits no twist. We observe that
the pressure distribution around the midspan of the semi-wing remains constant
(Fig. 8.25d). However, if we move inboard, we see a gradual change in pressure dis-
tribution. The suction peak reduces, and close to the center section the suction peak
has shifted back to about 20 % of the local chord length. Furthermore, the upper
surface experiences more suction over the aft part of the airfoil. If we focus on the
lower surface, we observe that ahead of the airfoil’s thickest point, a higher pressure
is present near the center section than further outboard. These three changes in pres-
sure distribution have the combined effect that the pressure-induced drag (form drag)
around the center airfoil increases compared to the airfoils positioned more outboard.
In other words, the application of sweep on a lifting wing introduces additional form
drag near the midsection.
8.3 Inviscid Flow over Swept Wings 459
(a) -0.6
(b) -0.6
η=0.07
-0.4 -0.4
-0.2 -0.2
0 0
0.2 0.2
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Chordwise position, x/c (~) Chordwise position, x/c (~)
(c) -0.6 (d) -0.6
η=0.41
η=0.61
-0.4 -0.4
-0.2 -0.2
0 0
0.2 0.2
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Chordwise position, x/c (~) Chordwise position, x/c (~)
Fig. 8.25 Pressure distribution at various spanwise stations over a 45◦ swept wing of aspect ratio
y y y y y
3 (after Ref. [70]). a b/2 = 0.07. b b/2 = 0.14. c b/2 = 0.27. d b/2 = 0.41 and b/2 = 0.61
We have already seen that a modification of the thickness distribution of the root
can have an effect on the pressure distribution over the airfoil at zero lift. It was shown
by Weber et al. that this modification only works for wings at low angles of attack. At
angles of attack of 4◦ and upwards the isobars are distorted near the center because of
the change in chordwise lift in that region [70]. To bring the isobars near the root more
forward at elevated angles of attack, Küchemann [37] proposes a different solution:
increase the incidence and add negative camber to the wing root. These combined
measures ensure the desired value of the lift coefficient at the center section and a
chordwise lift distribution that matches to the one on the outboard wing. Küchemann
proposes an elegant method based on a vortex representation of the wing to predict
the required twist and camber distribution that matches a predefined chordwise and
spanwise lift distribution. The interested reader is referred to Refs. [37, 38] for further
reading. In Fig. 8.26 we show the result of wind tunnel tests on a 45◦ swept wing of
aspect ratio 3. As can be seen in the figure, the center airfoil features a positive twist
angle of 2.8◦ , which gradually reduces to zero at the mid-semispan. The graphs in
Fig. 8.26a, b show the chordwise lift distribution at various spanwise stations for an
angle of attack of 2.3◦ and 4.4◦ , respectively. For comparison, the dashed line shows
the lift distribution for the center airfoil without any twist or camber. The effect of
460 8 Aerodynamics of Swept Wings
basic wing
basic symmetrical section, t/c = 0.12
1.0 1.0
basic wing
η
0 wing with
0.8 0.8 0.05 camber and
0.24 twist
η 0.51
0 wing with
0.6 0.05 camber and 0.6
0.24 twist
0.51
0.4 0.4
α =2.3°
0.2 0.2
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Chordwise station, x/c (~) Chordwise station, x/c (~)
Fig. 8.26 Chordwise lift distributions between on a symmetrical wing and on a wing with camber
and twist. Both wings are untapered, have a 45◦ sweep angle, and an aspect ratio of 3 (after Ref.
[70]). a α = 2.3. b α = 4.4
the modification is striking: the lift distribution at the center section matches closely
to the more outboard sections. This indicates that the suction peak of the pressure
distribution has shifted forward considerably. These conclusions hold for both angles
of attack that are shown, indicating that the result of this modification is independent
of the overall lift coefficient of the wing.
In summary, we have seen two modifications to the wing root to ensure straight
and parallel isobars up to the center section of a swept wing. The first modification
requires the thickest point of the root airfoil to move forward and the thickness-to-
chord ratio to grow. This ensures that supervelocities due to thickness are constant
over the span. The second modification is an increase in local incidence angle of
the root airfoil in combination with a reduction in camber. This results in a constant
chordwise lift distribution with span and ensures that the spanwise lift distribution
meets a designated shape. Both modifications gradually merge with the baseline
airfoil within one chord length of the wing. Each of these modifications can be seen
in the wing of the Airbus A310 aircraft, which is depicted in Fig. 8.27. We can see
that the airfoil in the outboard wing has its thickest point at 28 % of the local chord.
In addition, the thickness to chord ratio in the wing outboard of the kink gradually
decreases from 11.8 % at the kink to 10.8 % at the tip. The incidence angle of the
airfoil at the kink is +1.5◦ while at the tip the incidence angle is reduced to −3◦
8.3 Inviscid Flow over Swept Wings 461
28%
28% 10,8%
22%
28%
28% -3° 2017
11,8/6%
2 +1°5
2
11,6% 3400
+1°5
+3° 4096
1 11,8% 7006 1
15% +5°
8380
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Meters
Fig. 8.27 Example of a wing for a high-subsonic transport aircraft (after Ref. [7])
to postpone stall at the tip3 (see Sect. 8.4.3). If we turn our attention to the root airfoil,
we see its thickness has grown to 15 %, its thickest point has moved forward to 22 %,
the incidence angle is raised to +5◦ , and clearly some negative camber has been
applied. All of these modifications are there to account for the three-dimensional flow
effects of swept wings and to make sure that the form drag of the wing is minimized.
Additional benefits include a reduction in wing weight and a larger internal wing
volume.
In the early 1950s when wind tunnel tests on swept wings commenced, it was
observed that the transition process in boundary layers over swept wings differed
from that over unswept wings. Where transition over unswept wings is dominated
by the Tollmien-Schlichting waves (see Sect. 6.4.2), it was noted that the transition
3 A negative incidence angle of the wing is often termed “washout”. This wing would therefore
ut
inflection
resulting
point
component
crossflow
tangential component
component
zt wall shear
xt
mechanism over swept wings was remarkably different. The combination of pres-
sure gradient and wing sweep deflects the streamlines inboard (see Fig. 8.18). In the
pressure-recovery region, closer to the trailing edge, the streamlines are bent outboard
again under influence of the local pressure gradient. Outside the boundary layer, the
high momentum of the flow causes a relatively small deflection of the streamline
due to the presence of the pressure gradient. However, inside the boundary layer,
the momentum of the flow is considerably reduced due to friction with the surface.
Therefore, the adverse pressure gradient causes a larger deflection of the streamlines
inside the boundary layer resulting in boundary-layer crossflow.
In Fig. 8.28 a notional boundary-layer profile is shown over a swept wing. The
axes system is oriented such that the xt -axis is aligned with the inviscid streamline
outside of the boundary layer. The boundary-layer profile spirals upwards creating
an angle between the wall shear vector and the inviscid stream line. The projections
of the boundary-layer profile on the xt yt -plane and z t yt -plane are also shown. The
shape of the boundary layer on the xt yt -plane looks familiar to us. However, in
the z t yt -plane the cross-flow profile looks entirely different. At the wall and at the
boundary-layer edge, the crossflow component is zero. Somewhere in the middle,
the crossflow component reaches a maximum. The notional sketch of Fig. 8.28 is
somewhat misleading as quantitative results show that the value of maximum cross-
flow velocity is typical around 5–10 % of the boundary-layer-edge velocity [9]. The
crossflow profile itself shows an inflection point (where the curvature changes from
concave to convex).
8.4 Viscous Flow over Swept Wings 463
∂p suction peak
<0
inviscid streamline ∂x
∂p
>0
xM ∂x
u
w
β0
Fig. 8.29 Laminar boundary layer development on a swept wing. xM is the location of the inviscid
streamline inflection point. β0 is the angle between the wall shear stress vector and the inviscid
streamline (after Ref. [9])
464 8 Aerodynamics of Swept Wings
Fig. 8.30 Boundary layer cross flow visualized on a swept wing with fluorescent oil on the right
wing and mini-tufts on the left wing (Photo Douglas Aircraft Company)
8.4.2 Transition
In Sect. 6.4.2 the transition of a laminar boundary layer to a turbulent boundary layer
was presented. We showed that flow over a flat plate transitions from laminar to
turbulent due to the amplification of Tollmien Schlichting (TS) waves. In addition, in
Sect. 7.3 we demonstrated that laminar-flow airfoils benefit from a favorable pressure
distribution to limit this amplification and postpone transition to more than 50 %c
over the suction-side of an airfoil. On swept wings it is more difficult to define a
8.4 Viscous Flow over Swept Wings 465
TS instability
Crossflow instability
V∞
Λ
Leading-edge contamination
Attachment-line instability Centrifugal instability
TS instability
Crossflow instability
Fig. 8.31 Notional sketch of instability mechanisms that might dominate in different parts of
a three-dimensional swept wing boundary layer (courtesy of Tempelmann [63], modified from
Ref. [19])
wing geometry that supports natural laminar flow over a substantial part of the wing
surface. Numerous efforts have been undertaken to achieve laminar flow over swept
wings and other lifting surfaces such as the horizontal or vertical tail. However, the
vast majority of these endeavors have not been implemented in production aircraft.
So, why is postponing transition so difficult on swept-wing aircraft? To answer this
question, we need to investigate the fundamental instability mechanism in three-
dimensional boundary layers. Apart from the TS waves, three additional instability
mechanisms can be identified: attachment-line instability, cross-flow instability, and
centrifugal instability. Each of these mechanisms can cause the boundary layer to
transition from laminar to turbulent. The region in which each of the mechanisms is
important is shown in Fig. 8.31.
The attachment-line instability refers to the instability that occurs at the leading-
edge attachment line. The leading-edge attachment line is an imaginary line that
connects the maximum pressure points on the leading-edge of the wing. For an
airfoil, this would be a stagnation point but when sweep is added, the flow does
not fully come to a standstill but starts moving outboard along the leading edge.
Laminar-flow flight experiments with swept-wing aircraft in the 1950s and 1960s
had demonstrated unexpected early transition near the leading edge. Pfenniger [47]
demonstrated that the stability of the attachment line boundary layer can be corre-
lated to the momentum-thickness Reynolds number at the attachment line, Reθal .
He showed that when Reθal > 240 the attachment line is inherently unstable and
causes transition. When 90 < Reθal < 240, the boundary layer is stable but sus-
ceptible to exterior disturbances that can cause transition. For example, disturbances
can be caused by local surface waves or a turbulent boundary layer stemming from
the wing/fuselage junction (referred to as leading-edge contamination). When Reθal
is less than 90–100, the boundary layer is stable, resistant to exterior disturbances,
and does not propagate turbulence along the wing span. To calculate the momen-
tum thickness at the attachment line, Saric and Reed [54] propose the following
approximate relation (for incompressible flow) based on [47]:
466 8 Aerodynamics of Swept Wings
r/c = 1.8%
NASA SC(2) 0412
r/c = 0.8%
NACA 661 212
Fig. 8.32 Estimated airfoil radii of two airfoils (belonging to Example 8.4)
0.404 V∞r sin2 Λle
Reθal =√ (8.34)
1+ε v cos Λle
where r is the leading-edge radius of the airfoil, v is the dynamic viscosity of the fluid,
and ε is the thickness-to-chord ratio of an equivalent ellipse, matching the leading
edge geometry. The value of ε can be computed as a function of the leading-edge
radius and the maximum thickness of the airfoil, tmax [12]:
2r/c
ε= (8.35)
tmax /c
A quick assessment of (8.34) reveals that in order to reduce Reθal one needs to
have a small leading-edge radius and/or a low sweep angle. The following example
illustrates this.
Example 8.4 Consider a sheared wing without taper and twist and with a chord of
6.0 m. We consider two airfoils for this wing that are both oriented in the streamwise
direction: a NACA 661 212 and a NASA SC(2) 0412 (see Fig. 8.32). The first airfoil
has a leading-edge radius of r/c = 0.008, while the second airfoil has r/c = 0.018.
Compute for each airfoil, the maximum sweep angle for which attachment-line dis-
turbances are damped, i.e. Reθal < 100. Assume that V∞ = 100 m/s and that sea-level
conditions apply.
Solution:
We first calculate the value for the dynamic viscosity at sealevel:
μ 18.1 × 10−6 [kg/(sm)]
ν0m ISA = = = 14.91 × 10−6 [kg/m2 ]
ρ 0m ISA 1.225 [kg/m3 ]
Subsequently, we employ (8.35) to compute that ε is 0.13 and 0.30 for the NACA
661 212 and a NASA SC(2) 0412, respectively. Using these values, we calculate
the value of the attachment-line momentum thickness using (8.34) for a range of
leading-edge sweep angles. If we plot this relation for both of the airfoils and add
a line for the transition boundary we can find the leading-edge sweep angles for
which attachment-line transition occurs. This is shown in Fig. 8.33. We can therefore
8.4 Viscous Flow over Swept Wings 467
conclude that for the given conditions, the maximum leading-edge sweep angle for
the SC(2) 0412 amounts to only 18◦ , while the 661 212 allows a sweep angle of 26◦ .
A different criterion is proposed by Poll [49]. For compressible flows, he defines
a Reynolds number based on the length scale ζ, and the boundary-layer-edge cross-
flow velocity along the attachment line, ve :
we ζ
Reζal = (8.36)
ν
where ν is the kinematic viscosity (μ/ρ) and ζ is defined as follows:
ν
ζal = du (8.37)
dx x=0
Poll demonstrates that when Reζal < 245 attachment line turbulence decays. This cor-
responds to an attachment-line momentum-thickness Reynolds number of 100. For
compressible flows, the dynamic and kinematic viscosity are dependent on the flow
temperature through Sutherland’s equation (2.81). Since the temperature changes in
the boundary layer, Sturdza [62] proposes to use an intermediate temperature, T̄ :
where Tr is the recovery temperature, which in adiabatic conditions equals the wall
temperature, Tw . Te is the inviscid edge temperature of the boundary layer. If we
write the cross-flow component as we = V∞ sin Λ, the Reynolds number based on
¯ ζal , can then be expressed as follows:
this intermediate temperature, Re
50 212
A 66 1
NAC
0
0 5 10 15 20 25 30
Leading−edge sweep angle, Λle (deg)
468 8 Aerodynamics of Swept Wings
Assuming the boundary-layer flow remains laminar along the attachment line,
crossflow instabilities are a second cause for the early onset of transition. The type
of crossflow instability that occurs within the boundary layer depends on the recep-
tivity of the boundary layer to external disturbances such as freestream turbulence,
acoustic waves, or roughness. Receptivity is defined as the mechanism by which
freestream disturbances enter the boundary layer and create the initial conditions for
unstable waves [55]. For low levels of turbulence (Tu < 0.2 %), crossflow transition
is dominated by stationary waves, while for higher turbulence levels (Tu > 0.2 %)
transition is caused by traveling waves [19]. The inflection point in the crossflow
boundary-layer profile (see Fig. 8.28) causes crossflow vortex structures, which have
their rotational axis within a few degrees aligned with the inviscid streamline. These
crossflow vortices all rotate in the same direction and create a “cat’s eye” structure
when viewed in the streamwise direction. In Fig. 8.34 the notional vortex structure
of these vortices is shown along with a photo showing the cross flow vortices. The
y
(a)
z
(b)
Fig. 8.34 Cross flow vortices over a swept wing. a Notional sketch of “cat’s eye” structure of cross-
flow vortices (after Ref. [50]). b Naphthalene surface patterns showing crossflow vortices (Photo
from Ref. [54]; reprinted by permission of the American Institute of Aeronautics and Astronautics,
Inc.). Flow is from left to right
8.4 Viscous Flow over Swept Wings 469
breakdown of these vortices results in a serrated transition pattern over the span of
the wing as is clear from Fig. 8.34b.
Experiments show that surface roughness represents the key parameter respon-
sible for the initiation of stationary crossflow vortices. Increased surface roughness
hastens the onset of crossflow vortices and moves the transition region upstream [19].
The receptivity to a two-dimensional surface protuberance and especially acoustic
disturbances is very weak.
The effect of turbulence levels on crossflow transition is more complicated than
for Tollmien-Schlichting waves in a two-dimensional boundary layer. Remember
that for TS waves there exists a logarithmic relation between the turbulence level and
the critical amplification factor, N (see Sect. 6.5.4). In contrast to two-dimensional
boundary layers, up to moderate values of turbulence intensity, freestream turbulence
affects transition only indirectly. It attenuates the growth of the stationary vortices and
retards transition. Only at values of Tu > 0.2 %, depending on the roughness height,
do traveling modes become dominating. The stationary vortex structure is washed
away by traveling crossflow waves, which results in an earlier onset of transition and
a more diffuse transition pattern [54]. A further increase of freestream turbulence
then accelerates transition. An airplane in flight is likely to see turbulence levels
below 0.1 %, while turbulence levels in the wind tunnel can be higher than 0.2 %.
Since the crossflow transition mechanism is different for both situations, it is difficult
to use wind-tunnel tests for the prediction of crossflow transition on full-scale wings.
Computationally expensive simulation methods such as direct numerical simulation
(DNS) can be used to include the effect of receptivity on the crossflow instability
mechanisms and transition.
The final transition mechanism in three-dimensional boundary layers is the cen-
trifugal instability. Centrifugal flow occurs when streamlines are curved by the pres-
ence of convex or concave surfaces, as was shown in Fig. 6.4. If, in an inviscid circular
flow, |r V | decreases with an increase in r the Rayleigh circulation criterion is sat-
isfied and the flow is unstable. In this case the instability is in the form of steady,
streamwise-oriented, counter-rotating vortices as shown in Fig. 8.35. These vortices
are commonly called Görtler vortices after H. Görtler who first described the forma-
tion of these vortices over concave walls in his doctoral dissertation in 1940 [27].
The Görtler instability is an important boundary-layer instability that, under some
conditions, leads the flow through a transition to turbulence. It is known that a Görtler
instability can cause transition on the wall of a supersonic nozzle in a boundary layer
that would otherwise be laminar. Moreover, the Görtler vortex structure exists in a
laminar boundary layer over a concave surface such as turbine-compressor blades.
If we return to Fig. 8.31, we see that Görtler instabilities are indicated as a possible
cause for transition over the concave region on the lower surface of the wing. How-
ever, it has been shown that beyond a very small sweep angle, the principal instability
on a swept wing with concave curvature is a crossflow instability and not a Görtler
instability. This reduces the concern about Görtler vortices in swept-wing flows [55]
but makes them important for unswept wings such as found on sail planes.
470 8 Aerodynamics of Swept Wings
8.4.3 Separation
The oil flow pattern in Fig. 8.30 is aligned with the local shear vector on the wing’s
surface. This image shows how the flow inside the boundary layer moves outboard.
We should recall that the boundary layer thickness at any point on the surface is a
function of the distance that the flow inside the boundary layer has traveled from
the leading edge (see the discussion on p. 305). With that notion, we can easily
deduce that the boundary layer over the outboard wing is thicker than what would
be expected based on two-dimensional boundary-layer properties. Conversely, the
crossflow in the boundary layer causes the thickness of the boundary layer at the
root to decrease compared to its two-dimensional counter part. Also recall that a
thicker boundary layer has less momentum close to the wall and therefore tends to
separate more easily than a thin boundary layer when an adverse pressure gradient
is present. We can therefore deduce that the root of a swept wing is relatively stall
resistant, while the outboard wing is more susceptible to stall. Since most aft-swept
wings are also tapered, this effect is amplified due to the lower Reynolds number
at the tip compared to the root. In Fig. 8.36 we show that for a 45◦ aft-swept wing
of aspect ratio 3 this stall behavior can indeed be observed. The airfoil close to the
y
tip ( b/2 = 0.91) stalls when the wing reaches an angle of attack of 17◦ . The more
inboard profiles stall at a progressively higher angle of attack. Even at an angle of
attack of 27.5◦ the two most inboard profiles show no sign of stall yet.
Due to the addition of wing sweep to postpone adverse compressibility effects,
we have now encountered a serious problem at low subsonic Mach numbers. The
early onset of tip stall could cause the outboard wing to lose lift. As the tips of an
aft-swept wing are located behind the airplane’s center of gravity, this effectively
creates a nose-up pitching moment. This pitching moment, in turn, causes the angle
of attack to increase, aggravating the stall even further! This unstable stall behavior
was a serious threat to the first generation of swept-wing aircraft such as the F-86
Sabre and MiG-15. If the onset of separation starts asymmetrically (one tip stalling
before the other), a large rolling moment results. This phenomenon is called roll off
and was another threat to early jet fighters with swept wings.
To prevent tip stall on swept wings, various measures can be taken. The MiG-15
featured full-chord fences to stop the cross flow of the boundary layer and start a fresh
8.4 Viscous Flow over Swept Wings 471
1.4
45° 0.14
20’’
η=0 0.27 0.41
1.2 BE 0.61
0.83
1.0 38’’ Span
0.91
Aspect ratio 2
Lift coefficient, CL (~)
59” Span
Aspect ratio 3.04
0.8
0.6
0.4
0.2
Fig. 8.36 Local lift coefficients over a 45◦ swept wing of aspect ratio 3 (after Ref. [71])
boundary layer outboard of each of the fences (see Fig. 8.37). We can easily imagine
that such fences cause interference drag at high subsonic conditions and are therefore
not ideal. Therefore, over the course of several decades several more subtle add-ons
were invented to prevent tip stall while keeping the drag penalty at high-subsonic
speeds acceptable. Examples include small fences, leading-edge snags, vortilons,
and vortex generators. The interested reader is referred to [46] for examples of these
devices and their effect on high-lift performance of various transport and military
aircraft. In addition, Ref. [46] also presents modifications to the airfoil shape of the
outboard wing to postpone tip stall.
Whether wing sweep causes a higher or lower maximum lift coefficient depends
on multiple factors. The aspect ratio and the thickness-to-chord ratio are among them.
An important parameter is the sharpness of the airfoil. The leading-edge sharpness
can be represented by the value of Δy (see Fig. 7.1) and expressed as a fraction of
the chord length. For untwisted wings of a constant airfoil section and aspect ratio
larger than 4, it is shown in Ref. [31] that when Δy > 2 %c, the maximum lift
coefficient decreases with increasing sweep angle. In other words, the maximum
lift coefficient of wings with blunt leading edges deteriorates with increasing sweep
472 8 Aerodynamics of Swept Wings
Fig. 8.37 Example of a swept-wing aircraft featuring wing fences: MiG-15 (Photo USAF)
angle. However, when Δy < 2 %c the maximum lift coefficient tends to increase
with increasing sweep angle. This is attributed to the formation of a leading-edge
vortex which forms when the flow separates from the leading edge. Leading edge
separation can lead to the formation of a separation bubble and the spanwise pressure
gradient converts this bubble into a stable leading-edge vortex. The high velocity in
the leading edge vortex causes a low static pressure, which acts on (part of the) wing’s
upper surface. The leading-edge vortex is therefore contributing to the wing’s lift and
can therefore increase the maximum lift coefficient of the wing. At the same time,
since the suction peak at the leading edge has diminished, the leading-edge vortex
also increases the drag of the wing (see Problem 8.16).
It is shown in Ref. [48] that the formation of a leading-edge vortex is dependent on
a combination of leading-edge sweep angle, airfoil sharpness, angle-of-attack, and
Reynolds number. In Fig. 8.38, the oil-flow pattern of two wings is shown. For the
sharp-nosed wing (Fig. 8.38a) we observe a streamline pattern that is indicative of
the presence of a leading edge vortex. In this particular example, a continuous flow
structure exists running from the leading edge at the root to the trailing edge at the
tip. The lines of tip separation and leading-edge separation have joined in this case.
If we compare this to the blunt-nosed wing of Fig. 8.38b, we see a different pattern.
There is no indication of the presence of a vortex but the flow is starting to separate
near the wing tip. However, as is shown in Ref. [48], this wing produced a part-span
vortex at angles of attack beyond 20◦ , demonstrating that the critical angle-of-attack
to produce a leading-edge vortex increases with increasing nose radius.
8.4 Viscous Flow over Swept Wings 473
(a)
Secondary
seperation
line
Short seperation
bubble
Seperation
line forming
Fig. 8.38 Surface oil-flow about two wings of A = 3.39, λ = 1, Λ = 45◦ , α = 14◦ , and
Re = 2.1 · 106 . Both wings feature a RAE 101 airfoil (t/c = 10 % at 30 %c) with modified leading
edge (after Ref. [48]). a Leading edge radius = 0.03 %c. b Leading edge radius = 3.0 %c
Fig. 8.39 Examples of highly-swept surfaces to create a leading-edge vortex. a Northrop YF-17
featuring leading-edge root extensions ahead of the main wing (Photo USAF). b Fokker 50 featuring
a dorsal fin ahead of the vertical tail (Photo Aleksandrs Samuilovs)
fighter aircraft that feature leading-edge root extensions (also known as “strakes”).
These are essentially highly swept wings without a distinct profile but with a sharp
leading edge (see Fig. 8.39a). These surfaces are added to create a stable vortex and
increase the maximum lift coefficient of the airplane and the corresponding stall
angle-of-attack. This is also true for the highly-swept dorsal fins that are sometimes
added to vertical tails of low-subsonic and high-subsonic aircraft (see for example
Fig. 8.39b). They allow for a higher sideslip angle, increase the maximum side-force
coefficient of the fin, and avoid the “rudder lock” problem.
In this final section of this chapter (and of this textbook!), we are turning our attention
to another Achilles heel of aircraft with swept wings: aeroelasticity. Aeroelasticity
is the common term that describes the interaction between aerodynamic, elastic and
inertial forces as described by the Collar triangle [18] in Fig. 8.40. Static aeroelas-
ticity is the interaction between aerodynamic and elastic forces. Examples include
wing bending due to aerodynamic loading or wing twisting due to control-surface
deflection. Flutter is an example of dynamic aeroelasticity. The inertial forces play
an important role in dynamic aeroelasticity as they affect the natural frequency of
the structure. Since the early days of flying, flutter, wing divergence, and control
effectiveness have been primary problems for airplane designers [22]. It can be eas-
ily imagined that with the increase in aircraft size, their flexibility has increased
considerably. Combine this with high dynamic pressures and compressibility effects
in the transonic realm and aeroelasticity becomes a very important and complex
phenomenon.
As many passengers of commercial aircraft have experienced, atmospheric turbu-
lence can lead to uncomfortable vibrations inside the aircraft. These vibrations are
often caused by atmospheric instabilities and are usually a combination of a rigid-
body oscillation and elastic vibrations. Atmospheric turbulence takes the shape of
8.5 Aeroelasticity of Swept Wings 475
vibration
stability and control
Dynamic Aeroelasticity
static aeroelasticity
discrete gusts that can come from any direction. As the reader knows, the loads on the
aircraft are directly coupled to the local velocity. Any change in local velocity (both
in magnitude and direction) immediately changes the loads on the aircraft. Aero-
dynamically induced vibrations can induce a cyclic loading on the structure which
can accelerate the growth of micro-cracks in metal parts. This is known as metal
fatigue and is related to the operational life span of the aircraft. Fatigue occurs due
to cyclic stresses on the structure. Reducing these stresses can be achieved by adding
more material to the structure. If a structure is reinforced to reduce the aeroelastic
effects on the structure, this is referred to as an aeroelastic weight penalty (or simply
aeroelastic penalty) [73].
In this section we discuss four different aeroelastic effects that swept wings in
high-subsonic conditions (might) experience: static deformation, reduced control-
surface effectiveness, structural divergence, and flutter. In Fig. 8.41 we show a qual-
itative relationship between the wing sweep and the three critical aeroelastic speeds.
We see that aft-swept wings are most sensitive to a change in aileron reversal speed
and we therefore place some more emphasis on this topic than on divergence and
flutter. No attempt is made to present an in-depth physical model for aeroelastic
behavior. For an extensive treatment of this subject, the reader is referred to texts
on aeroelasticity (e.g. [14, 72, 73]). The intention is to present the reader with the
aeroelastic implications of having a swept-wing.
Static aeroelastic effects occur when elastic and aerodynamic forces result in a
strained aircraft structure. This can result in quite large displacements. Inertia plays
no roll in static deformation. One can imagine that for the same surface area, a high
aspect-ratio wing experiences a larger static deformations than a low-aspect-ratio
wing. There exists a profound difference between the shape of the wing when it is
manufactured (often referred to as the jig shape or 0-g shape) and the shape it takes
during flight (1-g shape). Due to the pressure distribution over the wing, the structure
deforms in two ways: it twists and it bends.
If we consider a typical monocoque primary wing structure, the two spars and
stiffened upper and lower skins form a box. This is the load-carrying structure of the
wing that transfers the aerodynamic loads to the fuselage. When the box experiences
a torque it will twist about its elastic axis. This is an imaginary axis, with position
dependent on the material choice and material distribution in the wing box. If the
resultant force on a wing section does not coincide with the position of the elastic
axis, a torque is introduced and the section will twist. The amount of twist depends
on the torsional stiffness of the wing and the distance between the resultant aerody-
namic force and the elastic axis. To avoid twist in the structure, one can design the
box structure such that (in the design condition) the distance between the resultant
aerodynamic force and the elastic axis is zero. Such a measure is an example of
aeroelastic tailoring.
Even though the structural twist (or geometric twist) in a wing can be limited
by modifying the torsional stiffness distribution over the wing section, the wing will
still bend under aerodynamic loading. For an unswept wing, bending only affects the
effective dihedral angle4 but has a negligible effect on the effective angle of attack
of the wing. The effective angle of attack is the local angle of attack that the wing
or wing section experiences. When a wing deforms, its orientation with respect to
the flow changes resulting in a change in effective angle of attack. On swept wings,
pure bending (no torsion) leads to aerodynamic twist. The aerodynamic twist angle
at an arbitrary spanwise location is defined as the zero-lift angle of attack of the local
airfoil section. An aerodynamic twist distribution can thus be created by changing
the camber of the wing along the span. In the present context, we look at aerodynamic
twist that is caused by bending of an aft-swept wing. This is schematically shown
in Fig. 8.42 for a wing of infinite torsional stiffness about its elastic axis but a finite
bending stiffness. When the wing bends due to aerodynamic loading, it bends about
an imaginary bending axis. As we can see in Fig. 8.42, the bending axis is oriented at
approximately a right angle with respect to the elastic axis. This implies that when
the wing bends, the rear spar and front spar displace exactly the same amount along
the local bend axis. If we project these displacements on the rear view of the wing,
we can observe that at a given spanwise station the rear spar is elevated more than the
front spar. This results in a negative incidence angle of the wing sections outboard
of the root reducing their effective angle of attack.
4 The dihedral angle is the upward angle of the wing measured with respect to the horizontal.
8.5 Aeroelasticity of Swept Wings 477
lea
din
ge
fuselage side
dg
fro e Top view
nt
spa
r
ela
stic
axi ben
s din
ga
rea xes
r sp
ar
trai
ling
edg
e
Side view
Rear view
1-g shape
r
rear spa
fuselage side
ar
front sp 0-g (jig) shape
Fig. 8.42 Aerodynamic twist due to upward wing bending under 1-g loading assuming infinite
torsional stiffness and finite bending stiffness of the wing box (after [46])
Solution:
The geometric twist (θ) at the tip can be calculated by comparing the displacement,
w, of the leading edge (LE) and trailing edge (TE), respectively:
−1 wTE − wLE
α = sin (8.40)
c
To calculate these displacements at the tip, we first need to calculate the displacement,
w and rotation, ϕ at the point where the lift force attaches. We use the following
478 8 Aerodynamics of Swept Wings
b/2
Λ b/4
fro
nt sp
ar
c
rea
r sp L
ar
l1
l2
l3
Ll13
w(l1 ) = (8.41a)
3EI
Ll 2
ϕ(l1 ) = 1 (8.41b)
EI
where l1 = b/(4 cos Λ). In this example we assume the total lift to be concentrated in
a single attachment point. Beyond this discrete attachment point, the wing therefore
remains straight (no bending). The displacement at the leading and trailing edge can
be calculated as follows:
Flexible wing
tan Λ Lb2
α = − sin−1 (8.44)
1 + cos(2Λ) 8EI
We can now substitute the values of each of the wing properties and find a tip twist
of αtip = −2.76◦ .
Airplanes with high aspect-ratio wings and high sweep angles can have a con-
siderable amount of aerodynamic twist under 1-g loading. The Boeing B-47 (1947)
was one of the first airplanes that featured such a high-aspect-ratio swept-back wing.
In addition, its engines were podded in nacelles attached to the lower surface of the
wing by means of pylons (see Fig. 1.11). Since it had not been designed with the
aerodynamic twist in mind, its spanwise lift distribution differed considerably from
the lift distribution produced by the undeformed wing (Fig. 8.44). Because we know
that the induced drag of a wing is dependent on its lift distribution (elliptical lift
distribution yields the lowest induced drag), we can easily deduce that the deformed
swept wing experienced a higher drag and a lower bending moment at the root than
the undeformed wing.
The discrepancy between the jig-shape and the in-flight shape of the swept wing
should be considered when the aircraft is designed. The desired in-flight shape of
the wing does not correspond to the jig-shape of the wing. The wing should be
designed such that under 1-g loading, it takes on the desired shape. This requires
a more in-depth analysis of the elastic and geometric coupling under 1-g loading.
Moreover, bending due to lift has a significant effect on the bending moment due
to maneuvering and gusts when the load factor, n > 1. To arrive at the optimal
design under 1-g loading a structural design loop can be employed where first the
desired wing shape is determined and then the wing structure is designed that would
deform to this shape under 1-g loading [3]. The B-52 is an early example of a high-
subsonic aircraft (maximum operating Mach number, MMO = 0.93) that utilized the
difference between jig shape and the 1-g shape. Because the B-52 had a wing sweep
angle of 35◦ and an aspect ratio of 8.5, the bending of the wing resulted in a change
in angle of attack of −4.7◦ at the tip of the wing [15]. However, the designers at
480 8 Aerodynamics of Swept Wings
(a) (b)
-1.5
Pressure coefficient, Cp (~)
-1.5
0.0 0.0
η=0.1 η =0.1
η=0.3 η =0.3
0.5 η=0.5 η =0.5
η=0.7 0.5
η =0.7
η=0.9 η =0.9
1.0 1.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Chordwise position, x/c (~) Chordwise position, x/c (~)
Fig. 8.45 Difference in section pressure distribution on B-52 wing between flexible and rigid
conditions (after Ref. [45]). a Flexible wing. b Rigid wing (jig shape)
Limit up
nlimit, up=2.0g
Jig
14.16 ft
position
4.52 ft
Limit down
Fig. 8.46 Bending of the wing of a B-52 under aerodynamic loading (modified from Ref. [15])
Boeing had designed for such an aerodynamic twist and the wing assumed a shape
that resulted in a more favorable pressure distribution than would be expected from
the rigid wing design (Fig. 8.45). Figure 8.46 shows the difference between the jig
position of the wing and the shape under aerodynamic loading. The deformed wing
features reduced outboard loading as can be deduced from Fig. 8.45.
Apart from the wing, also the empennage suffers from static aeroelastic effects.
Due to the flexible attachment of the tailplane to the fuselage and due to the flexibility
of the fuselage itself, the effectiveness of the tailplane is reduced. As an example,
Fig. 8.47 shows how the lift-curve slope (C L αh ) of the horizontal tail plane of an
Airbus A300 depends on the Mach number at various altitudes. We see that for a
rigid connection, the effectiveness increases with Mach number due to the com-
pressibility effect. However, when the flexibility of the tailplane and the fuselage are
taken into account, we see that particularly at low altitudes and high Mach numbers
(i.e. high dynamic pressure, 21 γ pM 2 ) the effectiveness of the horizontal tailplane
decreases. This reduced effectiveness directly influences the longitudinal stability of
the aircraft. If C L αh reduces more than the lift-curve slope of the wing-fuselage com-
bination, C L αwf , the longitudinal stability of the aircraft reduces. Aeroelastic effects
are therefore not limited to just the change in wing shape; they also influence the
stability and controllability of the aircraft.
8.5 Aeroelasticity of Swept Wings 481
3.5
∂CLT Rigid structure
∂αT
3.0
H= 9000 m
2.5
6000 m
3000 m
2.0
0m
1.5
Tailplane, attachment
and fuselage empennage
1.0
0.5
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Mach number, M (~)
Fig. 8.47 Influence of fuselage flexibility on tailplane lift-curve slope (after Ref. [58])
When a wing is loaded during flight, the lift applies close to the elastic axis and
therefore does not typically cause twist in the wing. However, when an aileron is
deflected, say downward, the pressure distribution over the airfoil is altered, such
that the section’s center of pressure shifts substantially backwards. This creates a
significant torque about the elastic axis and can result in aileron-induced (geometric)
twist. When the aileron is deflected downward, the outboard wing will twist leading-
edge down. This results in a lower effective angle of attack of the section and therefore
reduces the increment in lift that was initially created by the aileron deflection. When
the wing has a relatively low torsional stiffness, the lift decrease due the aileron-
induced twist can be higher than the increase in lift due the aileron deflection itself.
When this is the case, roll control on the aircraft is reversed. In other words, when
the pilot puts in right aileron, the airplane rolls to the left. When this happens we
speak of aileron reversal.
Reduced aileron effectiveness is typically found on high-aspect-ratio swept-back
wings. Reduced aileron effectiveness is more profound in swept wings because the
aileron-induced bending of the wing also results in aerodynamic twist. A high-aspect
ratio, aft-swept wing therefore experiences both geometric and aerodynamic twist
upon aileron deflection. The speed at which the aileron effectiveness reduces to zero
is called the aileron reversal speed. Introducing 45◦ sweep in the wing requires the
torsional stiffness to increase with 40 % with respect to the unswept wing in order to
maintain the same aileron-reversal speed [16]. The aileron reversal speed is dependent
on the altitude and Mach number of the airplane. The following derivation is presented
482 8 Aerodynamics of Swept Wings
b
M∞
Top View Λ
k b
flexible T s /2
attachement c/4
ec A
k
aerodynamic center
of semi wing
ela
cf stic L m
axi
2s 2
fla
p
Section AA
L m
s
c
elastic axis 2 2
A c
α
ξ ec 0.25c
cf c
Fig. 8.48 Flexibly attached non-tapered rigid wing with full-span plain flap
to determine the effect of Mach number and sweep angle on the effectiveness of the
control surface.
Consider an airplane with a swept wing of aspect ratio A, surface area S, and
sweep angle, Λ (see Fig. 8.48). Assume the wing has a symmetric section and a
trailing-edge control surface over its entire span. When this control surface deflects
downwards, the camber of the wing is increased and a lift force and aerodynamic
moment is generated. The lift force and aerodynamic moment apply at the aerody-
namic center of each wing half, which is assumed to be located at the quarter-chord
line of the wing at 50 % of the structural semi-span, i.e. at bs /4. Since we consider
a semi-wing, the lift force and aerodynamic moment are also halved. If we focus
on section AA in Fig. 8.48 we have defined an aerodynamic moment vector, m/2,
perpendicular to the freestream direction. The semi-wing is modeled as a rigid com-
ponent that is constrained in twist (θ) and dihedral (Γ ) by two torsional springs
of stiffness k T and k, respectively. These springs represent the torsional and bend-
ing stiffness, respectively. We express the effectiveness of the control surface as the
change in lift coefficient with control surface deflection, C L δ , where δ is defined
perpendicular to the hinge line.
To start our calculations, we first need to determine the span and aerodynamic
chord of the wing. They are defined follows:
8.5 Aeroelasticity of Swept Wings 483
√
b= AS (8.45a)
c = S/b (8.45b)
The structural span and structural chord of a single wing can then be calculated,
respectively:
b
bs = (8.46a)
cos Λ
cs = c cos Λ (8.46b)
When the semi-wing bends or twists, its angle of attack α, changes. If we assume
that the effective angle of attack, dihedral angle, Γ , and the twist angle, θ, are small,
we can establish the following relation:
The twist angle and dihedral angle can be related to the lift (L/2) and pitching moment
(m/2) about each semi wing. As can be seen from Fig. 8.48, the pitching moment can
be divided into a component that exerts torsion on the wing (m T = m cos Λ) and a
component that induces bending of the wing (m b = m sin Λ). These two components
of the pitching moment in combination with the lift force can be related to the twist
angle and dihedral angle as follows:
mT L
+ ecs = k T θ (8.48a)
2 2
mb bs L
− + = kΓ (8.48b)
2 4 2
Traditional wing structures consist of a box structure with stringers and ribs. The box
consists of an upper and lower skin panel together with a front and an aft spar. When
loaded in bending, the upper skin is being compressed, while the lower skin is put
in tension. The flanges of the spars transfer shear stresses. When loaded in torsion,
the upper and lower skin together with the two spar flanges transfer shear stresses.
By placing ribs in the box structure, the torsional stiffness of the wing box can
be significantly increased. This reduces the amount of twist that is produced under
loading. Let us consider a semi-wing box that is rectangular shaped and which has
a height (h) that is equal to the maximum thickness of the airfoil, a width, w, and
a length, bs /2. The upper and lower parts of the box are formed by the wing skins
with thickness tskin . The front and aft part of the box are formed by the spar flanges
of thickness tflange . Furthermore, we assume that the elastic axis of the box is located
in the geometric center of the box. Intuitively, we know that when the box grows in
size, or when the thicknesses increases, the box becomes stiffer. The bending and
torsional stiffness of the box are related to the second moments of area of the box’
cross section (I x , I y , and J , respectively). For the box dimensions presented above,
484 8 Aerodynamics of Swept Wings
wh 2 tskin
Ix = (8.49a)
3
w2 htflange
Iz = (8.49b)
3
J = I x + Iz (8.49c)
The second factor that determines the box’ stiffness is the Young’s modulus (E) of
the material that is used. The shear modulus of the material (G) is related to the
Young’s modulus via the Poisson ratio of the material (ν):
E
G= (8.50)
2(1 + ν)
If we assume a box without any ribs, we can calculate the flexural rigidity as the
product of the Young’s modulus and the moment of inertia about the x-axis. The
torsional rigidity is the product of the polar moment of inertia (J ) and the shear
modulus:
flexural rigidity = EI
torsional rigidity = GJ
If we model the wing box as a rigid component that attaches to a set of flexible springs,
the flexural and torsional rigidity need to be mapped onto the spring stiffnesses. For
the flexural rigidity, we do this as follows. We assume a flexible beam of length bs /2
with flexural rigidity EI that is subjected to a uniformly distributed positive load
L/bs [N/m]. Under this load, the beam bends upward over a distance Δ at the tip.
Secondly, we consider a rigid bar of length bs /2 that is connected to a torsion spring
of stiffness k [Nm/rad], which is subjected to the same uniformly distributed load
L/bs [N/m]. This bar rotates such that the tip displacement in vertical direction is
also Δ. With this comparison we can show that the spring stiffness is related to the
flexural rigidity as follows (Problem 8.20):
8EI x
k= (8.51)
bs
We do the same for the torsional spring that represents the torsional rigidity of the box.
The flexible beam with torsional rigidity GJ is subjected to a uniformly distributed
moment m/bs [Nm/m]. It can be shown that under such a moment distribution the
twist distribution is quadratic with a maximum twist at the tip of θtip = m T bs /8GJ.
The average twist amounts to θaverage = m T bs /24GJ. We set the average rotation to
the rotation of the rigid bar that is connected to the spring and deduce that the spring
stiffness is as follows (Problem 8.21):
8.5 Aeroelasticity of Swept Wings 485
12GJ
kT = (8.52)
bs
The force and moment can be expressed as a function of the Mach number per-
pendicular to the leading edge and the freestream static pressure ( p∞ ) as follows:
L clα clξ p∞ γ M ∞
2 S
= f (A) α+ ξ (8.53a)
2 2 2 2 1 − M∞ 2
m cm ξ p∞ γ M∞
2 Sc
= f (A) ξ (8.53b)
2 2 2 1 − M∞ 2
Here, we have implicitly assumed that the moment coefficient due to angle of attack
about the quarter-chord point equals zero. We recognize in (8.53) the lift coefficient
due to angle of attack (clα ) and the change in lift and moment coefficient due to
flap deflection (clξ and cm ξ , resp.). These three derivatives pertain to the airfoil, not
the wing. The lift coefficient on a sheared airfoil is related to the lift coefficient of
the unswept airfoil according to (8.10). The angle of attack and the deflection angle
of the flap in spanwise direction are assumed to be affected by the sweep angle as
follows:
In the above equations we define δ as the deflection angle of the flap perpendicular
to the hinge line. The angles ξ and δ are interrelated according to:
A p∞ γ M ∞
2
K = √ S (8.57)
2 + A2 + 4 4 1 − M∞ 2
486 8 Aerodynamics of Swept Wings
We can now combine the previous equations into a set of three linear equations
in α, θ and Γ :
kT
clα ecs α − θ = − cm ξ + clξ e cs ξ (8.58a)
K
bs k bs
clα α − Γ = cm ξ c sin Λ − clξ ξ (8.58b)
4 K 4
α − θ cos Λ + Γ sin Λ = 0 (8.58c)
This system of equations can be solved for α, θ and Γ . The wing lift (L) due
to flap deflection can subsequently be calculated by employing (8.53a). The flap
effectiveness can hence be calculated according to:
L(δ) − L(0)/q S
C L δ = lim (8.60)
δ→0 δ
The following example illustrates the effect of wing sweep on the flap effectiveness.
Example 8.6 Consider the wing of Fig. 8.48. Assume this wing has a planform area
of 93.5 m2 and an aspect ratio of 8.4. Furthermore, assume it features a box structure
of width w = 0.5cs and height h = 0.12cs . The geometric center of this box
is located at 40 % of the structural chord (i.e. e = 0.15). The thicknesses of the
box are tskin = 1 cm and tflange = 5 mm. The Young’s modulus of the material is
70 × 109 N/m2 and the Poisson ratio is 0.31. The flap measures 20 % of the chord, i.e.
c f = 0.20c. Assume the wing is sheared with the following sweep angles: Λ = 0,
Λ = 15◦ , Λ = 30◦ , and Λ = 45◦ .
(a) Make a topview of the four wings.
(b) Plot the flap effectiveness (C L δ ) as a function of the Mach number ranging from
M = 0 to M = 0.9. Assume a static pressure of 26.4 kN/m2 (which is the
atmospheric pressure at an altitude of 10 km).
Solution:
First, we draw the top views of the wing to get a feeling for the relative dimensions of
the wing. This is done on the left-hand side of Fig. 8.49. Subsequently, we estimate the
following values for the sectional aerodynamic derivatives from handbook methods
[1, 52] for an airfoil with c f = 0.20c:
5
Sweep angle, Λ (deg):
4.5
Sweep angle, Λ (deg):
δ
15 3
15
2.5
2
30
1.5 30
0.5 45
45
0
0 0.2 0.4 0.6 0.8
Mach number, M∞ (~)
Fig. 8.49 Aileron effectiveness of four different wings at 10 km altitude (belonging to Example 8.6)
For calculation purposes, we assume δ = 1◦ and calculate ξ for every sweep angle
according to (8.55). We subsequently solve (8.59) for various combinations of M and
Λ to find the values for Γ , θ and α. For each combination of M and Λ we substitute
the value of α along with the value of ξ into (8.53a) and using (8.60) to obtain a
value for C L δ . The result is plotted in the graph on the right-hand side of Fig. 8.49.
Let us briefly reflect on the result of Example 8.6. What we have simulated is the
effect of wing sweep on the full-span flap effectiveness of a wing. In shearing the
wing backwards, both the structural and aerodynamic properties of the wing change.
With increasing sweep angle the structural chord decreases and the structural length
increases. With decreasing structural chord comes a decrease in flexural and torsional
rigidity. Aerodynamically, the wing becomes less sensitive to changes in angle of
attack or changes in flap angle. For the unswept wing (Λ = 0) we observe that
the flap effectiveness is continuously increasing with Mach number. This trend is
attributed to the compressibility effect. We will see in Problem 8.22 that the torsional
flexibility reduces the aileron effectiveness of the unswept wing and that this effect
is magnified with sweep angle. A nose-down pitching moment is generated due to
the flap deflection, which results in a nose-down twist of the wing.
When the sweep angle increases the flap effectiveness decreases. Even at low
speeds, one can observe a decrease in flap effectiveness due to wing sweep. This
aerodynamic effect is independent of the flexibility of the wing. It demonstrates that
flaps become less efficient on a swept wing, compared to a straight wing. At increased
Mach number, the bending of the flexible wing starts to affect the angle of attack
of the section. As soon as the wing bends upwards, the angle of attack decreases
and the amount of lift is decreased. This results in a decrease in flap effectiveness
488 8 Aerodynamics of Swept Wings
with Mach number. For a 30◦ swept wing, the flap effectiveness has halved at a
Mach number of 0.85. This demonstrates the detrimental effect of wing sweep on
the effectiveness of a flap on a flexible wing. In practice, swept wings therefore
have reduced aileron effectiveness or even aileron reversal at combinations of speed
and altitude that result in a high dynamic pressure. To remedy this problem and
allow for roll control during the entire flight envelope, many high-subsonic aircraft
either employ outboard spoilers or inboard ailerons close to the wing root (known
as high-speed ailerons).
In our example, we have made a gross simplification of a wing to demonstrate
the effect of wing sweep on flap effectiveness on a flexible wing. We have assumed
an untapered wing of constant cross section. In practice, most wings are tapered,
resulting in an increase in both flexural and torsional stiffness towards the root. In
addition, the control surface usually spans only a part of the trailing edge where we
have assumed a full-span flap. Also, we implicitly assumed a constant lift distrib-
ution over the wing, where in reality the lift distribution would be closer to being
elliptical. Furthermore, the Prandtl-Glauert compressibility correction is known to
be inaccurate at Mach numbers approaching unity, which overestimates the forces
and resulting displacements. Finally, we made a gross simplification of the structure
(e.g. no ribs) and subsequently mapped the flexural and torsional stiffness onto the
spring stiffness of two torsional springs. Any of the aforementioned assumptions
has an effect on the results. However, with this gross simplification of an airplane
semi wing we have demonstrated in a very simple manner the adverse effect of wing
sweep and Mach number on flap effectiveness. In the following example we will
demonstrate how Mach number and altitude dictate the flap effectiveness for a wing
of given geometry.
Example 8.7 We would like to investigate the effect of altitude on the full-span flap
effectiveness of the wing of Example 8.6 with Λ = 30◦ . For this wing, plot the
flap effectiveness as a function of Mach number for Mach numbers between 0 and
0.9 and for altitudes of 0, 3, 6, and 9 km. Assume international standard atmosphere
conditions. In addition, plot the relation between the flap altitude and Mach number
at which the flap effectiveness is C L δ = 1.5, 1.0, 0.5 and 0 per radian. In the latter
plot, also indicate where control reversal occurs.
Solution:
To make these two plots, we follow the same approach as in Example 8.6. However,
this time we keep the sweep angle constant at 30◦ . To estimate the pressure ( p) at
each altitude (h) we employ the following equation for the troposphere [53]:
−g
λh Rλ
p = p0m ISA 1 + (8.62)
T0m ISA
The following constants are used: p0m ISA = 101325 N/m2 , T0m ISA = 288.15 K,
g = 9.81 m/s2 , λ = −0.0065 K/m, and R = 287 J/kg/K. For a given value of the
8.5 Aeroelasticity of Swept Wings 489
2.5 9000
Sweep angle, deg 8000
Aileron effectivenss, C (1/rad)
2 7000
1.15
1.75
1.45
0.85
L
6000
Altitude, h (m)
(1/rad) =
h
1.5
(k
0.55
5000
m
)=
9
L
4000
C
6
1
3
3000
0.25
0
0.5 2000
1000
0 0
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
Mach number, M (~) Mach number, M (~)
altitude (up to 11 km), we can now find the corresponding pressure. Using the same
wing properties as in Example 8.6, we can hence make the two required graphs. They
are shown in Fig. 8.50.
Again, we briefly reflect on the result of Example 8.7. It can be seen in the
graph on the left-hand side of Fig. 8.50 that the flap effectiveness increases with
increasing altitude and decreases with increasing Mach number. The lower pressure
at higher elevations results in a lower dynamic pressure, which results in a smaller
loss in control-surface effectiveness. Vice versa, at lower altitudes the effectiveness
is decreased. If we shift our attention to the second graph in Fig. 8.50, we could
imagine that this graph could be used to define an envelope of altitude and Mach
number for which a predefined value of control power is guaranteed. For example,
if C L δ > 1.15, we could use the corresponding line in the graph to bound the flight
envelope of the airplane in terms of Mach number and altitude. In that case, the
airplane would be restricted to fly below M = 0.35 at sea level (ISA). However, at
9 km (ISA) it would be allowed to fly M = 0.61.
This notional example demonstrates the practical implications of reduced control-
surface effectiveness. The addition of wing sweep makes it more difficult to fulfill
control power requirements. If for a given wing geometry the control power is insuf-
ficient due to aeroelastic effects, the structure needs to be made stiffer. This can be
done by adding more material to the structure, which increases its mass. This increase
in mass to reduce adverse aeroelastic effects is an example of the aeroelastic penalty.
We have qualitatively shown that the stability and control characteristics of large
transonic transport aircraft can be significantly altered due to aeroelastic effects.
Simple control laws corrected for small aeroelastic effects obtained from handbooks
or wind tunnel tests do not suffice [22]. To determine the stability and control deriva-
tives (e.g. C L αh and C L δ ) and how they change with Mach number, the fluid-structure
490 8 Aerodynamics of Swept Wings
interaction needs to be taken into account. This can, for example, be done by cou-
pling a finite-element (FE) model of the aircraft structure to a numerical fluid model
(CFD).
Fig. 8.51 X-29 with forward swept wing and supercritical airfoils [8]
8.5 Aeroelasticity of Swept Wings 491
Dynamic aeroelastic effects occur when elastic, inertial and aerodynamic forces inter-
act in such a manner that an oscillation is produced [73]. Dynamic aeroelastic effects
are not exclusive to transonic flow. However, the increasing influence of compress-
ibility results in particularly strong pressure and density changes in transonic flow
when instabilities occur. This results in severe loads on the boundaries between fluid
and structure (e.g. wing) and strong pressure waves in the flow [42]. Damped oscilla-
tions are forced vibrations that are caused by instabilities in the flow that feed energy
to the structure. These forced vibrations can have different causes, e.g. cyclic sepa-
ration of the boundary layer due to shockwave impingement (buffet, see Sect. 7.6) or
atmospheric turbulence. The result of this oscillatory loading is a forced vibration in
the structure. If the amplitude of the vibration grows in time towards an asymptote,
we speak of a limit cycle oscillation (LCO). If the amplitude grows unbounded, we
speak of flutter. Contrary to damped oscillations, flutter is an unbounded dynamic
form of aeroelasticity. At the flutter speed (V f ), any infinitesimally small disturbance
results in an unbounded vibration of the wing that eventually leads to a failure of
the structure. Flutter can allude to a pure bending motion of the wing, a pure torsion
motion, or a coupling of both motions. Probably the most dangerous type of aircraft
flutter results from coupling between the bending and torsional motions of a relatively
large aspect ratio wing. In this section we discuss flutter and limit-cycle oscillations
of swept wings. The reader is referred to other texts that explain the mechanisms of
flutter in more detail [14, 73].
We further explain the mechanism of bend-torsion flutter by considering the forces
acting on a two-dimensional (rigid) wing section of mass (m) and inertia (I y ) as
shown in Fig. 8.52. In this example, sc is the distance between the elastic axis and
the center of mass referred to as the static unbalance of the section. The structural
stiffness is represented by a linear spring system consisting of a torsion spring and
a tension/compression spring. The section is able to plunge (h) and pitch (α) about
the elastic axis. The aerodynamic moment (M y ) is considered about the elastic axis,
while the inertial force (m ḧ) and moment (I α̈) act at the center of mass of the section.
L
sc
c
My undeflected airfoil center line
α elastic x
axis
h
kT k
center of
mass
Fig. 8.52 Notional sketch of aerodynamic, structural, and inertial forces acting on a two-
dimensional wing section
492 8 Aerodynamics of Swept Wings
We can imagine that when this system is in oscillation, each of the forces and moments
shown in Fig. 8.52 are somehow dependent on the position or acceleration of the
section. The only sources of damping or instability in the system are the lift and the
aerodynamic moment.
If we only consider the plunging motion, we notice that it has an effect on the
angle of attack of the wing section. For example, when the outboard wing heaves
downwards, it experiences an increase in local angle of attack, which induces an
upward lift force. This lift force, in turn works opposite to the direction of motion and
therefore decelerates the downward motion. This is what is what we call aerodynamic
damping. The damping force depends on the rate of change in position. If we again
consider a downward motion, there is no aerodynamic force that depends on the
displacement (heave) of the outboard section, the damping is entirely due to the
motion of the wing. In other words, the aerodynamic stiffness due to bending is (close
to) zero. Whether the aerodynamic stiffness due to torsion is positive or negative
depends on the value of e, the nondimensional distance between the aerodynamic
center and the elastic axis. If the aerodynamic center of the wing lies ahead of
the elastic axis, the torsional stiffens is negative. This implies that the increase in
twist causes an aerodynamic moment that increases the twist even further. Vice versa,
when the aerodynamic center is behind the elastic axis, there is positive aerodynamic
torsional stiffness. It can be shown that the quasi-steady aerodynamic moment is
also dependent on α̇, resulting in damping or amplification of the torsional motion
depending on the speed.
The flutter speed is defined as the speed at which this system becomes critically
damped, i.e. when the aerodynamic damping goes to zero. The flutter speed therefore
depends on a variety of factors including the static unbalance, the location of the
elastic axis, the location of theaerodynamic center, the relative mass ratio (m/ρc2 ),
the radius of gyration (r = I y /m) about the elastic axis, and the ratio of the
uncoupled natural frequencies of the bending and torsional motion (ωh /ωα ).
Typical wings of high-performance aircraft have large density ratios and posi-
tive static unbalances. The frequency of the bending-torsion flutter lies somewhere
between ωh and ωα , with ωh being the smaller of the two. Increasing the density
ratio increases the flutter speed. The most critical condition is therefore encountered
at sea-level rather than at high altitude. Increasing the static unbalance or decreasing
the radius of gyration both reduce the flutter speed. This has led to the concept of
mass balancing. Placing additional mass ahead of the elastic axis reduces the static
unbalance. The torsional inertial loads become smaller and therefore the torsional
natural frequency increases resulting in an increase in flutter speed. However, even
when the mass and elastic axis are aligned with the aerodynamic center on the quarter
chord of the wing, flutter can still occur due to negative damping terms [73]. It can
be shown that for a given position of the aerodynamic center, a displacement of the
elastic axis closer to the leading edge also results in a an exponential increase in the
flutter speed.
The Mach number has an effect on the flutter speed. In the lower transonic
regime, the flutter speed reduces when the Mach number increases towards unity.
This results in a dip in the flutter speed versus Mach number graph, which we term
8.5 Aeroelasticity of Swept Wings 493
3.2
“transonic dip” 5 a
bωα = 1
2.4
1.6 1/2
0.8
c = 2b
0
0 0.8 1.6 2.4 3.2
Mach number, M∞ (~)
the transonic dip. When the Mach number increases towards unity the flutter speed
also increases again and continues to increase in the supersonic domain. References
[24, 43] were among the first to publish analytically obtained results based on poten-
tial flow calculations that showed the transonic dip for high relative mass ratios.
The relation between flutter speed (V f ) and Mach number is shown in Fig. 8.53. We
clearly observe the transonic dip in flutter speed for high relative mass ratios. The
quantitative accuracy of these curves can be questioned due to the nonpotential flow
below M∞ = 1 and the erratic behavior of the aerodynamic coefficients near the
speed of sound. However, the curves give a good indication of the compressibility
effect on the flutter speed.
Let us further discuss the effect of Mach number on dynamic response of the
wing. With increasing Mach number the section becomes more sensitive to changes
in angle of attack (remember the Prandtl-Glauert compressibility correction that was
discussed in Chap. 3). This amplifies the aerodynamic forces when there is a change in
angle of attack. Furthermore, the formation of a shock wave and the associated mixed
subsonic-supersonic flow have a profound effect on the energy exchange between
494 8 Aerodynamics of Swept Wings
airstream and wing and hence on the flutter speed. This is caused by the strong
dependence of the shock location on the angle of attack. During the oscillatory motion
of the wing, the shock oscillates over the profile or even temporarily vanishes [64].
We also saw this effect when we discussed the periodic shock-induced separation
that causes transonic buffet (Sect. 7.6). Generally speaking, the flutter frequencies
of the coupled bending-torsion mode are one to two orders of magnitude smaller
than for buffet of full-scale high-aspect ratio wings. Therefore, there exists a smaller
phase shift in the shock location with respect to the angle of attack. A representative
relation between shock position and angle of attack has previously been shown in
Figs. 7.16 and 7.17.
Bendiksen [11] performed a numerical investigation into the (de)stabilizing effect
of shocks waves on wings in the transonic domain. It was shown that the destabilizing
effect of shocks becomes noticeable when the shocks reach a position between 40
and 50 % chord. The most destabilizing shock location is 75 %c, which generally
coincides with the Mach number at the bottom of the transonic dip. With further
Mach number increases, the shock approaches the trailing edge and the influence of
the shock changes from destabilizing to strongly stabilizing, resulting in the post-dip
rise in the flutter speed. The mean position of the shock also relates to the location
of the aerodynamic center. The aerodynamic center shifts progressively rearwards
from the 25 % chord below the critical Mach number towards the 50 % chord in
supersonic conditions. At supersonic Mach numbers no flutter exists when the center
of mass is ahead of the midchord. Finally, the shock-boundary layer interaction can
play an important role as the boundary layer might separate at the shock foot when
the section’s angle of attack becomes large enough. This can effectively change a
bending-torsion instability into a limit-cycle oscillation. In Fig. 8.54 the effect of
Mach number on the flutter speed of a swept-wing wind tunnel model is shown
s
rd
0.15
xi
as height
ho
b
ca
r-c
sti
te
Λ=45º
ela
ar
0.1
qu
0.05
cr
0
0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2
Mach numer, M (~)
8.5 Aeroelasticity of Swept Wings 495
along with aerodynamic predictions. Clearly, a flutter dip can be observed in the
experimental results around the speed of sound.
Aircraft with high aspect ratio wings are more prone to flutter than low-aspect
ratio wings due to the larger absolute displacement and rotation of the wing for a
given aerodynamic load [10]. However, due to the coupling between wing bending
and aerodynamic twist, aft-swept wings generally have a higher flutter speed than
their unswept counterparts (as was shown in the notional graph of Fig. 8.41). The
interaction between aerodynamic, elastic and inertial forces can still result in struc-
tural oscillation. Since the natural frequencies (in torsion and bending) are a direct
function of the stiffness of the fluid-structure system, they change with the velocity
of the aircraft. On an aft-swept wing, the aerodynamic bending stiffness is introduced
due to the coupling of bending and aerodynamic twist. Whenever the wing heaves
downwards, the effective angle of attack is increased to provide an upward lift force
trying to restore the wing to its neutral position. An aft-swept wing therefore has
a positive aerodynamic stiffness, which contributes to a higher overall stiffness of
the system. Vice versa, a forward-swept wing typically reduces the overall stiffness
of the system. Experimental and theoretical investigations have demonstrated that
aft-swept wings have a tendency to reduce the amplitude of the vibration due to
the coupling of wing bending and aerodynamic twist. Due to this coupled effect the
vibration beyond the flutter speed does not grow unbounded immediately, but results
in a limit cycle oscillation [10].
Flutter is not only limited to wings. Tail flutter can also be a potential risk in
transonic flow. Moreover, a coupling between fuselage, wing, engines, and empen-
nage can even flutter. For example, the location of a wing-mounted engine has a
major effect on the flutter behavior of the aircraft. Flutter grows unbounded when
the energy that is extracted from the airstream is larger than the energy that is dissi-
pated by damping in the fluid-structure system (whether this is just the wing or the
full aircraft). It occurs when the aerodynamic forces couple with a natural vibration
mode of the structure to produce a periodic motion in the structure that leads to
failure if the damping goes to zero. As mentioned previously, the speed at which
this occurs is referred to as the flutter speed and forms an important boundary of
the flight envelope. The flutter motion of wings is usually a coupling between the
bending and torsion modes. If an aircraft reaches its flutter speed, there can be a
coupling between wing bending, wing torsion, rigid-body translation of the fuse-
lage, and rigid-body rotations of the fuselage. The interaction between the various
elastic and rigid-body motions demonstrates the complexity of the flutter motion.
Adding the fact that transonic aerodynamics is challenging from a modeling point of
view makes predicting flutter speeds and flutter behavior one of the most challenging
problems in the aeronautical sciences.
It was already acknowledged in the early 1950s that flutter performance can be
improved significantly if a rapidly responding closed-loop control system is inte-
grated in the aircraft structure (this is often referred to as aeroservoelasticity). The
flutter frequencies are generally too high for pilots to respond to and should there-
fore work autonomously [14]. Commercial aircraft can therefore be equipped with
a flight control system that has a gust load alleviation system [44]. For example,
496 8 Aerodynamics of Swept Wings
the Boeing 787 uses advanced in-flight controls to allow the aircraft to sense and
respond to atmospheric turbulence (gusts), creating a smoother ride [34]. The allevi-
ation systems rely on sensors (usually accelerometers and rate gyros) that are placed
close to the center of gravity of the aircraft. Air data sensors are used to inform the
control system about the aircraft’s speed and attitude. Control input is then given to
the control surfaces (e.g. ailerons and spoilers) such as to balance any gust-induced
motion of the aircraft. Such a vibration-suppression system could potentially also be
used to fly beyond the flutter speed of the aircraft. This has only been successfully
demonstrated on wind tunnel models [73].
can be used for the equations that describe the fluid flow (i.e. unsteady Euler or
unsteady RANS equations). The equations of motion are then changed from the time
domain (where all variables are a function of time, t) to the frequency domain by
means of a Fourier transformation.5 By doing this, the aerodynamic properties can
be transformed to a truncated Fourier series that can generally be reduced to contain
just three, two, or even one Fourier term. This essentially reduces the system to
a multi-DoF damped spring-mass model with a forcing function [13]. Using such
an approximation of the fluid-structure system, we can relatively quickly determine
whether a structure will be prone to flutter. However, due to the many simplifications
the resulting prediction is not as accurate as we would like it to be. An overview of
various numerical methods to analyze the static and dynamic aeroelasticity is given
by Livne in Ref. [41].
Flutter predictions can also be done by wind tunnel experiments or by flight test-
ing full scale aircraft. Which of these measures is most effective depends on a variety
of factors including the anticipated margin of safety from flutter, the Mach number
range, and the number of different mass and structural configurations to be analyzed
[14]. The last argument is particularly applicable in the case of military aircraft
that can carry a variety of external weapons or stores. Testing each configuration
during flight tests is expensive and time-costly. Even though present-day analytic
predictions are still underdeveloped, they can be used to indicate which combina-
tion of stores/weapons can potentially be flutter sensitive. By test-flying only these
configurations the number of flight tests can be reduced significantly [20].
To determine the dynamic aeroelastic characteristics of a transonic aircraft, wind
tunnel tests can be carried out on a dynamically scaled model of the aircraft. A
dynamically scaled model of the aircraft is constructed such that the (dynamic) elas-
tic deformations under aerodynamic load are proportional to that of the full-scale
aircraft, although the time required for a certain motion is different and requires
mathematical factors for interpretation of the results. Even though these tests are
not sufficiently reliable to determine the absolute flutter speed of the full-scale air-
craft, it helps the main certification calculations by validating unsteady aerodynamic
methodologies. In addition, it is used for performing parametric studies, studying
new configurations, investigating interference and compressibility effects, or study-
ing the effect of advanced control algorithms on the aeroservoelasticity (e.g. gust
response) of the aircraft [17]. Figure 8.55 shows two wind tunnel experiments in a
transonic wind tunnel. In Fig. 8.55a a semispan model of a wing is mounted to the
wind tunnel wall. This model has been dynamically scaled such that the relative dis-
placements of the wing under aerodynamic loading are close to the full-scale wing.
With this model the amount of bend-twist coupling under aerodynamic loading can
be determined and the dynamic aeroelastic response can be measured. In Fig. 8.55b
a model of the Boeing 747 is shown. Similar to the wing of Fig. 8.55a, this model is
also dynamically scaled. However, it is also strung to the side and upwards to mimic
free-flight conditions. These strings allow the aircraft to perform rigid-body motions,
which often occur when the aircraft experiences limit-cycle oscillations.
Fig. 8.55 Flutter tests in the NASA Langley transonic wind tunnel facility [17]. a Semi-span model
of Cessna citation (Photo NASA). b Scale model of Boeing 747 (Photo NASA)
Fig. 8.56 Ground vibration test of a Lockheed Jetstar at NASA Dryden Flight Research Laboratory.
Photo from Ref. [36], courtesy of NASA
8.6 Summary
the angle between the shear vector at the wall and the inviscid streamline can be
as high as 90◦ . Due to the inflection point in the crossflow boundary-layer profile,
crossflow instabilities occur in the form of (stationary) crossflow vortices. These
vortices are responsible for the early onset of boundary-layer transition on airfoils that
would otherwise have remained laminar. In addition, on an aft-swept wing the mean
boundary-layer flow tends to move outboard under an adverse pressure gradient. This
results in a thinner boundary layer near the root, making it more resistant to stall,
while a thicker boundary-layer is formed over the outboard wing and at the wing
tip. The tip therefore has a lower stall angle-of-attack, which can lead to premature
tip stall, one of the most dangerous forms of stall on swept-wing aircraft due to the
resulting destabilizing pitching moment.
Finally, we presented the effect of wing sweep on some fundamental aeroelas-
tic properties of the wing. We showed that swept wings exhibit aerodynamic twist
whenever they bend under aerodynamic loading. A swept wing should therefore
be designed for 1-g loading. Furthermore, we showed that aft-swept wings reduce
the effectiveness of trailing-edge control surfaces. Due to the coupling between wing
bending and aerodynamic twist on an aft-swept wing, the effectiveness of control sur-
faces deteriorates. At the reversal speed, the control-surface effectiveness has reduced
to zero and a further increase in velocity causes control reversal. For high-aspect ratio
wing with ailerons this is known as aileron reversal, a potentially dangerous charac-
teristic when flying at high dynamic pressures. On the other hand, aft-swept wings
generally have a higher flutter and divergence speed as compared to their unswept
counterparts. In low transonic conditions the flutter speed decreases rapidly with
Mach number due to destabilizing oscillating shock waves. Once the shocks move
closer to the trailing edge, they become stabilizing and the flutter speed increases
rapidly with Mach number in the upper-transonic and supersonic domain.
Problems
The Advantages of Wing Sweep
8.2 The R-4009 profile of Fig. 8.4 has a peak pressure coefficient of −0.60. Calculate
the drag divergence Mach number of an infinite-span wing consisting of this airfoil…
(a) …when it is swept over 20◦ .
(b) …when it is swept over 40◦ .
8.6 Summary 501
8.3 Write a computer program to generate a graph similar to the one in Fig. 8.9
with on the x-ordinate the drag divergence Mach number instead of the critical
Mach number. You need to use (8.12), (8.23), and (8.11). Use this graph to answer
Problem 8.2. Do you get the same answers?
8.4 Consider a typical lift distribution over a straight, unswept wing. Argue why the
weight of this wing increases when the thickness-to-chord ratio reduces.
8.5 In Fig. 7.18 the lift and drag coefficient of a NACA 64A010 are shown as a
function of Mach number. We would like to estimate the drag coefficient as a function
of lift coefficient and Mach number of a swept, untwisted wing of Λ = 25◦ . We
assume that this wing has an aspect ratio of 6 and a span efficiency factor of φ = 0.85.
Assume cd0 is independent of the sweep angle and that we may use (8.29) to predict
the wing’s drag coefficient as a function of Mach number and lift coefficient.
(a) Combine Fig. 7.18a, b and tabulate the values of the lift coefficient and drag
coefficient at each Mach number in a spreadsheet.
(b) Use this spreadsheet to draw the drag coefficient (cd ) of the airfoil as a function
of the Mach number for the following lift coefficients: cl = 0, 0.2, 0.4, and 0.6.
(c) Use your spreadsheet and (8.29) to draw the drag coefficient (C D ) of the wing
as a function of the Mach number for the following lift coefficients: C L = 0,
0.2, and 0.4.
(d) Repeat (c) for a wing of the same aspect ratio and span loading but with Λ = 15◦
and Λ = 35◦ , respectively. Can you discover a trend in drag-divergence Mach
number at each of the lift coefficients?
8.6 In the figure below you can see the drag-curve for a 45◦ swept wing of aspect
ratio 5.5 and taper ratio of 0.6. The wing features a NACA 64A010 airfoil. The
estimation of the drag coefficient is based on (8.29). Why is the drag-divergence
Mach number for the wing so much lower than would be expected from elementary
swept-wing theory?
Drag coefficient, CD (~)
Measured
Estimated
CL=.3
8.7 (a) Explain what is meant by the ‘root effect’ of swept wings.
(b) List four modifications that should be made to the geometry of the wing root in
order to reduce the root effect.
8.8 Consider the planform of the A321 wing in the picture below. Explain why the
tip has been rounded.
Photo A.Pingstone
8.9 In the pictures below an early design of the Boeing 7E7 is shown (top) as well
as the production version (bottom). One of the striking features is the highly-swept
“shark-fin” vertical tail. The tip has been rounded-off and there is a curved contour
at the base where the trailing edge blends in with the fuselage, both at the leading
and trailing edge.
(a) Do the tip and base fin shapes have any aerodynamic significance and why (not)?
(b) Explain why you think the production version of the 787 had a more conventional
(straight) fin leading edge?
8.6 Summary 503
7E7
787
8.10 In the figure below the airfoils of the Airbus A310 are shown at three spanwise
locations. The airfoil at the wing root does not look like a supercritical airfoil. Explain
why its shape is so different from the airfoils in the outboard sections.
Wing tip
Parallel
to Datum
Planform crank
Datum
Wing root
Parallel
to Datum
After Ref. [23]
8.11 In the sketches below the pressure distribution is shown for M = 0.78 and C L =
0.27 for the wing profile as applied to the North American Sabreliner, a business jet
of the 1950s. During the 1970s another company came up with a modification of the
inboard wing that eliminated the inboard shock which reduced high speed drag. In the
bottom left figures you may notice the difference in pressure distribution before and
after the modification (indicated by “Mark Five”). Sketch what profile modification
yields the aforementioned change in pressure distribution at cross section A (data
from Ref. [65]).
504 8 Aerodynamics of Swept Wings
-.1
0
.1
A
-.5
M=0.78
-.4
-.3 CL=0.27
-.2
-.1
0
Shock wave
-.1
A
Pressure distribution at AA
Pressure coefficient upper surface, Cp,upper (~)
-1.0
Modified airfoil
-0.8 Original airfoil
-0.6
M = 0.78
-0.4 CL= constant
-0.2
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
0.2 Chordwise position, x/c (~)
0.4
8.6 Summary 505
8.12 In Sect. 6.6.2 modifications to the wing-pylon-nacelle area of the Convair 990
have been presented. It was stated that the strong drag increase merely stemmed from
the inboard side of the pylons. Why would the inboard of the pylon produce more
interference drag than the outboard side?
8.13 Let us consider a starboard wing with a mean aerodynamic chord of 6.0 m,
a taper ratio of 0.25, an aspect ratio of 5, and a quarter-chord sweep angle of 30◦
(akin to the Airbus A330). Furthermore, assume the wing has a SC(2) 0412 airfoil
(see Example 8.4), oriented in streamwise direction. Assume this wing is subjected
to air at a velocity of 0.8 times the local speed of sound at an altitude of 10 km and
that (8.34) may be used to compute the attachment-line momentum thickness.
(a) Calculate the span, b of the wing in meters.
(b) Calculate the tip-chord length (ctip ) and the root-chord length (croot ) in meters.
(c) Calculate the sweep angle of the leading edge (Λle ).
(d) Calculate the density, the viscosity, and the local speed of sound for air at an
altitude of 10 km ISA.
(e) Define the relation between nose radius (r ) and spanwise position (y).
(f) Plot the relation between attachment-line momentum-thickness Reynolds num-
ber versus spanwise position. Also mark the border where the attachment line
switches from being stable to being susceptible to disturbances and the border
where the attachment line switches from being susceptible to disturbances to
being inherently unstable.
(g) Based on these findings, what conclusions can you draw?
8.14 Near the end of WWII the Germans applied the swept-wing concept to attain
higher speeds. An example is the Junkers, Ju-287, which was a jet-powered bomber.
This aircraft had the peculiar feature that the wings were swept forward.
(a) Explain the general principle of a swept wing.
(b) Why do you think a forward swept wing was chosen, rather than an aft swept
wing?
(c) What is an important drawback of a forward swept wing compared to and aft
swept wing?
506 8 Aerodynamics of Swept Wings
8.15 A plain wing (without camber and twist) with Λc/4 = 45◦ having constant
profile has a calculated spanwise distribution of the local lift coefficient as indicated
in the figure below. The profile, considered perpendicular to the quarter-chord line,
has a maximum lift coefficient clmax of 1.5. At what wing lift coefficient do you expect
the initiation of tip stall and unstable behavior?
Local lift coefficient, cl, local (~)
1.4
CL,wing
1.2
1.0
0.90
0.8 0.75
0.6 0.50
0.45
0.4
0.30
0.2 0.15
0
0 0.5 1
Spanwise position, η (~)
8.16 When the suction peak of a swept wing diminishes due to the formation of a
stable leading-edge vortex, explain why/how this increases the profile drag of the
wing.
8.17 Consider the wing of Example 8.5. We want to investigate the effect of the
geometric wing parameters on the aerodynamic twist.
(a) Calculate the aerodynamic twist when the wing span increases with 10 %.
(b) Calculate the aerodynamic twist when the flexural rigidity increases with 10 %.
8.6 Summary 507
(c) Calculate the aerodynamic twist when the sweep angle increases with 3◦
(d) To which of the aforementioned parameters is the aerodynamic twist most
sensitive?
8.18 Consider the wing of Example 8.5 but with Λ = 45◦ . We want to know the
sensitivity of the aerodynamic twist to the wing parameters.
(a) Calculate the sensitivity in [degree/m] of the aerodynamic twist w.r.t. the span
of the wing, i.e. dθ/db.
(b) Calculate the sensitivity in [degree/(Nm2 )] of the aerodynamic twist w.r.t. the
flexural stiffness of the wing, i.e. dθ/d(E I ).
(c) Calculate the sensitivity in [degree/degree] of the aerodynamic twist w.r.t. the
sweep angle of the wing, i.e. dθ/dΛ.
(d) To which of the aforementioned parameters is the aerodynamic twist most sensi-
tive? In other words, if each of the aforementioned parameters is increased with
1 %, which has the largest effect on the aerodynamic twist of the wing?
8.19 Below a generic cross section of a wing box is shown. Calculate the second
moment of area about the x and y axis, respectively. Do they match with (8.49)?
What assumptions were made?
z
tskin h
x
tflange
8.20 Below, two beams of the same dimensions are shown that are both subjected
to the same uniform lift distribution. The upper beam is completely rigid and rotates
about a hinge point. The lower beam is clamped and bends under the uniform loading.
Both beams have an identical tip deflection of Δ, where Δ bs /2. Show that the
spring stiffness, k can be expressed as in (8.51).
Δ
Γ
k
l
Δ
EI
b /2
s
8.21 Below, two beams of the same dimensions are shown that are both subjected
to the same uniform moment distribution. The upper beam is completely rigid and
rotates about a hinge point. The lower beam is clamped and twists under the uniform
loading. The average twist of the flexible beam equals the rotation of the rigid beam.
508 8 Aerodynamics of Swept Wings
(a) plot the twist distribution in the flexible beam as a function of the spanwise
position, y. Assume m̄/GJ = 1 and bs /2 = 1.
(b) calculate the average rotation in the flexible beam and indicate this in the plot
of (a).
(c) Show that the spring stiffness, k T can be expressed as in (8.52).
kT mT
GJ
bs /2
8.23 Use the computer program developed in Problem 8.22 to make the following
assignments.
(a) Change the aspect ratio to 2 and 10, respectively. Keep all other parameters as in
Example 8.6. Plot the relation between Mach number and C L δ for sweep angles
of 0, 15, 30, and 45◦ . Also plot the change in α, θ, and Γ . Based on your results,
describe the effect of aspect ratio on the flap effectiveness.
(b) Change the Young’s modulus of the material to 35 and 140 GPa, respectively.
Keep all other parameters as in Example 8.6. Plot the relation between Mach
number and C L δ for the aforementioned sweep angles. Also plot the change in
α, θ, and Γ . Based on your results, describe the effect of material stiffness on
the flap effectiveness.
(c) Keep all parameters as in Example 8.6. Plot the relation between Mach number
and C L δ for sweep angles of 0, −15, −30, and −45◦ . Also plot the change in
α, θ, and Γ . Based on your results, describe the effect of forward sweep on flap
effectiveness. What adverse aeroelastic effect do you observe?
References
1. Abbott, I.H., von Doenhoff, A.E.: Theory of Wing Sections. Dover Publications, New York
(1959)
2. Adler, A.: Effects of Combinations of Aspect Ratio and Sweepback at High Subsonic Mach
Numbers. NACA RM L7C24, Langley Field (1947)
3. Aly, S., Ogot, M., Pelz, R., Siclari, M.: Jig-shape static aeroelastic wing design problem: a
decoupled approach. J. Aircr. 39(6), 1061–1066 (2002). doi:10.2514/2.3035
4. Anderson, J.: Fundamentals of Aerodynamics, 5th edn. McGraw Hill, New York (2010)
5. Anderson, J.D.: Introduction to Flight. McGraw Hill, New York (2000)
6. Anon.: Transonic data memorandum; method for predicting the pressure distribution on swept
wings with subsonic attached flow. ESDU TM 73012, London (1973)
7. Anon.: Airbus Industrie A 310. L’Aeronautique et L’Astronautique 6(91), 15 (1981)
8. Anon.: Fact Sheets: X29. www.nasa.gov/centers/dryden (2008)
9. Arnal, D., Casalis, G.: Laminar-turbulent transition prediction in three-dimensional flows. Prog.
Aerosp. Sci. 36, 173–191 (2000)
10. Bendiksen, O.O.: Transonic limit cycle flutter of high-aspect-ratio swept wings. J. Aircr. 45(5),
1522–1533 (2008). doi:10.2514/1.29547
11. Bendiksen, O.O.: Influence of shocks on transonic flutter of flexible wings. In: Proceedings of
the 50th AIAA/ASME/AHS/ACS Structures, Structural Dynamics, and Materials Conference,
AIAA 2009-2313, pp. 1–29. Palm Springs, CA, May 2009. doi:10.2514/6.2009-2313
12. Bergrun, N.R.: An Empirically Derived Basis for Calculating the Area, Rate, and Distribution
of Water-Drop Impingement on Airfoils. NACA TR 1107 (1952)
13. Biao, Z., Zhide, Q., Chao, G.: Transonic flutter analysis of an airfoil with approximate boundary
method. In: Grant, I. (ed.) Proceedings of the 26th Congress of the International Council of the
Aeronautical Sciences. ICAS 2008-7.10.5 (2008)
14. Bisplinghof, R.L., Ashley, H., Halfman, R.: Aeroelasticity. Dover Publications, Mineola, New
York (1996)
15. Boyne, W.J.: Boeing B-52: A Documentary History. Smithsonian Institution Press, Washington
(1981)
16. Broadbent, E.G., Mansfield, O.: Aileron Reversal and Wing Divergence of Swept Wings, ARC
Technical Report R&M 2817, London (1954)
17. Chambers, J.R.: Concept to Reality: Contributions of the NASA Langley Research Center to
U.S. Civil Aircraft of the 1990s. SP 2003-4529, NASA (2003)
18. Collar, A.R.: The first fifty years of aeroelasticity. Aerospace 5(2), 12–20 (1978)
19. Deyhle, H., Bippes, H.: Disturbance growth in an unstable three-dimensional boundary layer
and its dependence on environmental conditions. J. Fluid Mech. 316(1), 73–113 (1996). doi:10.
1017/S0022112096000456
20. Dowell, E.H., Hall, K.C., Thomas, J.P., Kielb, R.E., Spiker, M.A., Li, A., Charles, M., Dene-
gri, J.: Reduced order models in unsteady aerodynamic models, aeroelasticity and molecular
dynamics. In: Grant, I. (ed.) Proceedings of the 26th International Congress of the Aeronautical
Sciences. ICAS 2008-0.1 (2008)
21. Drela, M., Giles, M.B.: Viscous-inviscid analysis of transonic and low Reynolds number air-
foils. AIAA J. 25(10), 1347–1355 (1987). doi:10.2514/3.9789
22. Dusto, A.R.: An analytical method for predicting the stability and control characteristics of
large elastic airplanes at subsonic and supersonic speeds, part I—analysis. In: Aeroelastic
Effects from a Flight Mechanics Standpoint, vol. 46. AGARD (1970)
23. Dykins, D.H., Jupp, J.A., McRae, D.M.: Esso energy award lecture, 1987. Application of
aerodynamic research and development to civil aircraft wing design. Proc. R. Soc. Ser. A.
Math. Phys. Sci. 416(1850), 43–62 (1988)
24. Garrick, I.E., Rubinov, S.I.: Flutter and Oscillating Air-Force Calculations for an Airfoil in a
Two-Dimensional Supersonic Flow. NACA Report No. 846, Langley Research Center, Hamp-
ton Virginia (1946)
510 8 Aerodynamics of Swept Wings
25. Gere, J.M., Timoshenko, S.P.: Mechanics of Materials, 4th edn. Stanley Thornes Ltd, Chel-
tenham (1972)
26. Giles, M.B., Drela, M.: Two-dimensional transonic aerodynamic design method. AIAA J.
25(9), 1199–1206 (1987). doi:10.2514/3.9768
27. Görtler, H.: Über eine dreidimensionale Instabilität laminarer Grenzschichten an konkaven
Wänden. Ges. d. Wiss. Göttingen, Nachr. a. d. Math. 2(1) (1940)
28. Hadcock, R.: X-29 composite wing. In: Presented at the American Institute of Aeronautics and
Astronautics Symposium on the Evaluation of Aircraft/Aerospace Structures and Materials.
Air Force Museum, Dayton, Ohio (1985)
29. Hall, M.G.: Transonic flows. In: Proceedings of IMA Conference on Numerical Methods and
Problems in Aeronautical Fluid Mechanics. Academic Press (1975)
30. Harper, C.W., Maki, R.L.: A Review of the Stall Characteristics of Swept Wings. NACA TN
D-2373 (1964)
31. Hoak, D.E., Anderson, R., Goss, C.R.: USAF Stability and Control Datcom. Air Force Wright
Aeronautical Laboratories, Wright Patterson Air Force Base, Ohio (1978)
32. Hunton, L.W.: A study of the application of airfoil section data to the estimation of the high-
subsonic-speed characteristics of swept wings. NACA-RM-A55C23 (1955)
33. Jenkinson, L.R., Simpkin, P., Rhodes, D.: Civil Jet Aircraft Design. American Institute of
Aeronautics and Astronautics, Reston, VA (1999)
34. Jenks, M.: The Boeing 787 dreamliner programme. In: Grant, I. (ed.) Proceedings of the 26th
Congress of the International Council of the Aeronautical Sciences. ICAS 2008-0.8 (2008)
35. Johnson, R.: Minimum Drag Coefficient of Wings. RAND RM-604, Santa Monica (1951)
36. Kehoe, M.W., Freudinger, L.C.: Aircraft Ground Vibration Testing at the NASA Dryden Flight
Research Facility. Technical Memorandum 104275, NASA (1993)
37. Kuechemann, D.: The distribution of lift over the surface of swept wings. Aeronaut. Q. 4,
261–278 (1953)
38. Kuechemann, D.: The Aerodynamic Design of Aircraft. Pergamon Press, Oxford (1978)
39. Kuechemann, D., Weber, J.: The Subsonic Flow Past Swept Wings at Zero Lift Without and
With Body. Aeronautical Research Council Reports and Memoranda 2908, London (1956)
40. Kuzmina, S., Karas, O., Ishmuratov, F., Zichenkov, M., Chedrik, V.: Analysis of static and
dynamic aeroelastic characteristics of airplane in transonic flow. In: Grant, I. (ed.) Proceedings
of ICAS 2012, pp. 1–10. Brisbane, Australia (2012)
41. Livne, E.: Future of airplane aeroelasticity. J. Aircr. 40(6), 1066–1092 (2003). doi:10.2514/2.
7218
42. Meier, G.E.A.: Unsteady phenomena. In: Zierep, J., Oertel, H. (eds.) Symposium Transsonicum
III, IUTAM, pp. 441–464. Springer, Berlin (1988)
43. Nelson, H.C., Berman, J.H.: Calculations on the Forces and Moments for an Oscillating Wing-
Aileron combination in Two-Dimensional Potential Flow at Sonic Speed. NACA TN 2590.
Langley Research Center, Hampton (1952)
44. Newberry, C.F.: Consideration of stability augmentation systems for large elastic aircraft. In:
Aeroelastic Effects from a Flight Mechanics Standpoint, vol. 46. AGARD (1970)
45. Obert, E.: Aerodynamic Design and Aircraft Operation, Part III. Aerodynamics of Combat
Aircraft. Delft University of Technology, Delft (1996)
46. Obert, E.: Aerodynamic Design of Transport Aircraft. IOS Press, Delft (2009)
47. Pfenniger, W.: Laminar flow control - laminarization. In: Special Course on Concepts for Drag
Reduction, AGARD R-654, pp. 3.1–3.75 (1977)
48. Poll, D.I.A.: On the generation and subsequent development of spiral vortex flow over a swept-
back wing. In: AGARD-CP-342, pp. 6.1–6.14 (1983)
49. Poll, D.I.A.: Some observations of the transition process on the windward face of a long yawed
cylinder. J. Fluid Mech. 150, 329–356 (1985). doi:10.1017/S0022112085000155
50. Reed, H., Saric, W.: Stability of three-dimensional boundary layers. Annu. Rev. Fluid Mech.
21, 235–284 (1989)
51. Ringleb, F.: Some Aerodynamic Relations for an Airfoil in Oblique Flow. NACA TM 1158
(1947)
References 511
52. Roskam, J.: Airplane Design, Part 6: Preliminary Calculation of Aerodynamic, Thrust and
Power Characteristics. DARcorporation, Lawrence, KS (2006)
53. Ruijgrok, G.J.J.: Elements of Airplane Performance. Delft University Press, Delft, the
Netherlands (1989)
54. Saric, W., Reed, H.: Crossflow instabilities—theory & technology. In: 41st Aerospace Sciences
Meeting & Exhibit. AIAA 2003-0771, Reno, NV (2003)
55. Saric, W.S.: Görtler vortices. Annu. Rev. Fluid Mech. 26, 379–409 (1994)
56. Saric, W.S., Reed, H.L., White, E.B.: Stability and transition of three-dimensional boundary
layers. Annu. Rev. Fluid Mech. 35(1), 413–440 (2003). doi:10.1146/annurev.fluid.35.101101.
161045
57. Schlichting, H., Truckenbrot, E.: Aerodynamik des Flugzeuges. Springer, Heidelberg (1960)
58. Schoernack, W., Hässler, E.: Theoretical investigation of aeroelastic influences on the lift
distribution and the aerodynamic derivatives of swept wings at symmetric and antisymmetric
stationary flight conditions. In: Aeroelastic Effects from a Flight Mechanics Standpoint, vol.
46. AGARD (1970)
59. Shevell, R.S., Kroo, I.: Compressibility drag: 3D effects and sweep. In: Aircraft Design: Syn-
thesis and Analysis. Department of Aeronautics and Astronautics, Stanford University, CA
(1992)
60. Stack, J., Lindsey, W.: Characteristics of Low Aspect Ratio Wings at Supercritical Mach Num-
bers. NACA TN 1665. Langley Air Force Base (1948)
61. Strauss, W.A.: Partial Differential Equations, An Introduction. Wiley, Chichester, UK (1992)
62. Sturdza, P.: An Aerodynamic Design Method for Supersonic Natural Laminar Flow Aircraft.
Ph.D. thesis, Stanford University (2003)
63. Tempelmann, D.: Stability and Receptivity of Three-Dimensional Boundary Layers. Disserta-
tion, Royal Institute of Technology, Sweden (2009)
64. Tijdeman, H.: Investigations of the Transonic Flow Around Oscillating Airfoils. Ph.D. thesis,
Delft University of Technology, the Netherlands (1977)
65. Timmons, L.M.: Improving Business Jet Performance: The Mark Five Sabreliner. SAE Paper
790582 (1979)
66. Torenbeek, E.: Synthesis of Subsonic Airplane Design. Delft University Press, Delft (1976)
67. Van der Wees, A.: A Nonlinear Multigrid Method for Three-Dimensional Transonic Potential
Flow. Ph.D. thesis, Delft University of Technology, the Netherlands (1988)
68. Van Muijden, J., Broekhuizen, A.J., Van der Wees, A.J., Van der Vooren, J.: Flow analysis
and drag prediction for transonic transport wing/body configurations using a viscous-inviscid
interaction type method. In: Proceedings of the 19th ICAS Congress. Anaheim, California
(1994)
69. Weaver, J.H.: A method of wind tunnel testing through the transonic range. J. Aeronaut. Sci.
15(1) (1948)
70. Weber, J.: The Calculation of the Pressure Distribution over the Surface of Two-Dimensional
and Swept Wings with Symmetrical Aerofoil Sections. Aeronautical Research Council Reports
and Memoranda No. 2918 (1953)
71. Weber, J., Kuechemann, D., Brebner, G.: 2882 Low-Speed Tests on 45-deg Swept-Back Wings,
Aeronautical Research Council Reports and Memoranda No. 2882 (1958)
72. Weisshaar, T.A.: Aircraft Aeroelastic Design and Analysis. Purdue University, West Lafayette
(2009)
73. Wright, J., Cooper, J.: Introduction to Aeroelasticity and Loads. Wiley, Chichester, UK (2007)
74. Yates, E.C.: 765 AGARD Standard Aeroelastic Configurations for Dynamic Response I- Wing
445.6, AGARD Report 765 (1985)
75. Young, A.D., Booth, T.B.: The Profile Drag of Yawed Wings of Inifinite Span, Cranfield College
of Aeronautics Report No. 38 (1950)
Appendix A
Isentropic Flow Table
Table C.1 shows the Prandtl-Meyer fucntion and Mach angle for γ = 1.4.
Chapter 1
1.1
(a) a = 296 m/s
(b) R| L/D=16 = 5350 km, R| L/D=19 = 6353 km, R| L/D=22 = 7356 km.
1.2
(c) L/D|max = 17.7 at C L = 0.74
(f) V × L/D|max = 3.3 km/s at C L = 0.43 and ρ = 0.3.
Chapter 2
2.7
(a) u = 1/2t 2
(b) 1/2(x 2 + t 2 + 1) − xt − x − t
2 (−1)n
2.11 Dn = π ( 1 +n)2 .
2
2.15
(a) φ = x 2 /2 + y 2 + 3z 2 /2
(b) φ = x yz
(c) φ = ye x + z
(d) does not exist.
2.16
(a) 6
(b) 0
(c) ye x
(d) 0.
2.17
(a) 0
(b) 0
(c) 0
(d) (0, 0, 3).
2.18
γ = 1.4, cv = 716 J/kg/K, e = 160 kJ/k, h = 224 kJ/kg, and T = 223 K.
2.20
37.3 kJ/kg.
Partial Answers to Selected Problems 531
2.27
(b)
2 2
∂u 2 ∂v ∂w ∂v ∂u 2 ∂w ∂v 2
Φ= μ 2 + + + + + +
∂x ∂y ∂z ∂x ∂y ∂y ∂z
∂u ∂w 2 2 ∂u ∂v ∂w 2
+ + − + +
∂z ∂x 3 ∂x ∂y ∂z
2.29
(a) p = (γ − 1)ρe
(b) q = − ckv ∇e
3/2
c T0 + 110
(c) q = − Prp μ0 cveT0 e + 110cv ∇e; not conservative.
Chapter 3
3.1 −1.704
3.3 cl = 0.756; The circulation is 51.2 % higher than the circulation in the incom-
pressible limit
3.7
(a) cl2 = 0.9615
(b) Cp = −2.248
min
(c) clα M =0.65 / clα M = 1.32
∞ ∞ =0
3.9
(a) Mmax = 1.372
(b) C p,crit = −2.133
(c) C p,stag = 1.064
(d) C p0 ,min = −3.031
3.11
(a) M A = 1.0909
3.13
(b) u = 178.18 m/s
(c) C p = 0.53
3.15
(a) cd,w = 0.0569
(b) h/ = 0.0256
532 Partial Answers to Selected Problems
3.21
(a) p A / p∞ = 1.56
(b) q∞ / p∞ = 2.8
3.23
(a) K = 07368
(b) τ2 = 0.0177
(c) C pmin ,2 = 0.06
3.25 cl2 = 0.1587
3.27
(a) Mmax = 1.43
(b) C p,stag = 1.117
3.29
(a) M∞2 = 0.937
(b) α2 = 3◦
(c) = 10.079
(d) C L2 /C L1 = 0.6299
(e) C D2 /C D1 = 0.3149
(f) Cm2 /Cm1 = 0.6299
3.31 δmax = 20◦
3.33 σc = 78.11◦
Chapter 4
4.1
(a) p1 / p∞ = 2.971
(b) p2 / p∞ = 1.991
(c) cd = 0.02755
4.3
(a) p1 = 20.793 kPa
(b) p1 = 11.923 kPa
(c) p1 = 9.953 kPa
Partial Answers to Selected Problems 533
4.5 pt,2 = 53.38 kPa; pt,3 = 6.02 kPa; pt,4 = 32.83 kPa
4.7
(a) θw = 32.15◦
(b) V2 /V1 = 1.264
4.9 cl = 0.65, cd = 0.31
4.11
(a) t/c = 0.08816
(b) C p2 = 0.1061, C p3 = 0.4236, C p4 = −0.1531, C p5 = −0.1537
(c) ca = 0.03687
4.13
(a) C p1 = C p4 = 0.433, C p2 = C p5 = −0.1099, C p3 = C p6 = 0.1124
(b) cd = 0.0545
4.15
(a) M2 = 3.4
(b) θw = 17.79
4.17
(a) θLE = 23.38
(b) p2 / p1 = 0.213
(c) p3 / p1 = 0.2419
(d) θTE = 16.02
4.19
(a) t1 /c = 0.1063, t2 /c = 0.1246
(b) C p1 = 0.2239, C p2 = 1.3896, C p3 = −0.2151, C p4 = −0.1322
(c) cn = 0.1652
(d) ca = 0.0609
(e) L /D = 2.46
(f) cm,LE = −0.0995
4.21
(a) a1 = 260 m/s
(b) M2∗ = 0.78
4.23
(a) pnose / p∞ = 1.350
(b) p A / p∞ ≈ 0.735
4.25 C p,N = 0.821, C p,A = 0.250
534 Partial Answers to Selected Problems
4.27
(a) R̄ = 6.5
(b) δ0 = 22.6
(c) C pp = 0.05, C pv = 0.20
(d) C D = 0.118
4.31
(a) σ = 48◦
(b) Mc = 0.8
(c) C D p = 0.65
4.33
(a) C p1 = 0, C p2 = 0.837
(b) cd = 0.4835
(c) cl = 0.8374
4.35 M3 = 2.85
4.39
(a) |α| > 3◦
(b) α = ±2.5◦
Chapter 5
5.5
(a) θw,max = 22.875◦
(c) dy/dx = −2.37
(d) dy/dx = 0.3823
Partial Answers to Selected Problems 535
5.7
5.13
5.15
(a) M2 = 2.1
(b) M3 = 2
(c) (dy/dx)RRW = −0.536, (dy/dx)LRW = 0.582
(d) p3 / p1 = 1
5.17
3.2
3
2.8
y 2.6
2.4
2.2
2
0 5 10 15
x
536 Partial Answers to Selected Problems
5.19
(a) M2 = 2.85
(b) M3 = 4.05
(c) A3 /A1 = 6.645
Chapter 6
6.2
C C
(a) see text on p. 275: D = ρ D u 2 (u 1 − u 2 )dy ≈ ρV∞ D (V∞ − u 2 )dy
(b) pt∞ − pt2 = 21 ρ(V∞ 2 − u 2 ) ≈ ρV (V − u )
2 ∞ ∞ 2
(c) Using a Pitot rake the total pressure deficiency over the wake can be measured.
6.4 Notional pressure distributions:
- c
Cp
a
b
0
0 1
x/l
6.6
(b) 2.40
(c) 1.69 + 1.76i
(d) 2.23
6.8
(c)
6
Relative stream tube area, S/Smin (~)
1
0 0.5 1 1.5 2
Mach number, M (~)
Partial Answers to Selected Problems 537
6.10
x
2 3
(a) S(x) = 16V
− 4 xl ) 2
(4
3lπ
2 3
l
(b) R(x) = Rmax (4 xl − 4 xl ) 4
6.11
1 π π
(a) V = π 0 R 2 (x)dx = 30 and Rmax = 16 .
(b) C Dwave = 0.461
(c) C Dwave = −11 %
6.15
(a) 5.12, 3.47, 2.60 [m/s/m]
(b) 5.13, 3.47, 2.60 [mN/m]
(c) 10.3 ·10−6 , 6.94 ·10−6 , 5.20 ·10−6
6.16
−3
x 10
3
Friction coefficient, C (~)
2.9
f
2.8
2.7
2.6
2.5
2.4
0 0.5 1 1.5
Mach number, M (~)
6.17
(a)
−0.4
Cp (~)
−0.2
0
0 2 4 6 8 10
postion, x (m)
538 Partial Answers to Selected Problems
θ, δ* (m)
1 δ*
0
0 2 4 6
postion, x (m)
(e)
10
τ (N/m )
2
5
w
0
0 2 4 6
postion, x (m)
6.18
(a) (x/c)separation = 0.72
(b) (x/c)separation = 0.54
(c) (x/c)separation = 0.30
6.23
Rewrite (6.121) to u + = 1/κ ln(y/k) + 1/κ ln 1
1/k + 0.3 + B. Substitute B = 5.0
and evaluate for k+ → ∞.
6.24
(a) c f = 0.00320
(b) τw = 19.6 N/m
(c) (∂u/∂ y)w = 1.07 μm
m/s
6.25
(a) Rel = 6.70, 20.1, 67.0, 134, 335 ·106
(b) C f = 0.0145, 0.0136, 0.0129, 0.0126, 0.0123
Partial Answers to Selected Problems 539
(c)
−3
x 10
3.5
f
C
2.5
2
0 1 2 3 4
Re 8
l x 10
6.26
(a)
−3
x 10
7
Local friction coefficient, cf (−)
2
0 1 2 3
x−position (m)
(b)
−3
x 10
4
Friction coefficient, C (~)
f
3.5
2.5
2
0 0.2 0.4 0.6 0.8 1
sand grain height, k (m) x 10−4
540 Partial Answers to Selected Problems
Chapter 7
7.1
Mdd increases from 0.80 to 0.83.
7.10
(a) M0m ISA = 0.197
(b) C p M ∗ =√2 = −27.7
(c) Vstall = 72.4 m/s. Mstall = 0.217
(d) C p M ∗ =√2 = −22.8
(e) Since C p M ∗ =√2 > C pmin it is likely that the suction peak is lower in reality than
we have anticipated here and that the maximum lift coefficient is therefore also
lower.
7.12
(a) 7.7 km
(b) 40◦
7.14
(a) ϕ = 0.5433, 1.096, 1.783, 2.083 ms.
(b) amplitude = 0.7505, 0.4064, 0.3068, 0.2990
(c)
1
Shock jump amplitude (~)
0.8
0.6
0.4
0.2
5 4 3 2 1 0
2
Phase shift (rad)
1.5
0.5
0
5 4 3 2 1 0
dM1/dx (1/m)
7.15
(a) ϕ = 7, 20, 40, 44 (deg).
(b) amplitude = 0.44, 0.45, 0.32, 0.32
Partial Answers to Selected Problems 541
(c)
0.5
0.4
0.35
0 20 40 60 80
25
20
Phase shift (rad)
15
10
0
0 20 40 60 80
frequency (Hz)
Chapter 8
8.2
(a) Mdd = 0.70
(b) Mdd = 0.77
8.6 As long as the drag divergence of the wing is governed by section characteristics
(such as for wings with 35◦ sweep or less) a rather rough estimate of Mdd is possible
with the method of (8.29), but that an upper limit exists for Mdd in the neighborhood
of 0.95 caused by three-dimensional sonic-flow conditions (onset of wave drag).
This boundary cannot be altered by a further increase in sweep or changes in section
other than by their contribution to a change in the longitudinal distribution of cross-
sectional area.
8.13
(a) b = 30 m
(b) croot = 9.6 m, ctip = 2.4 m
(c) le = 32.5◦
(d) ρ = 0.412 kg/m3 , μ = 14.7 · 10−6 kg/(ms)
c −c
(e) c = tip b root y + croot
542 Partial Answers to Selected Problems
(f)
250
200
150
Re θ (~)
al
100
50
0
0 10 20 30
y (m)
8.17
(a) θ = −3.33◦
(b) θ = −2.5◦
(c) θ = −3.31◦
(d) Most sensitive to changes in span (b).
8.18
(a) dθ/db = −1.45 deg/m
(b) dθ/d(E I ) = 7.15 · 10−6 deg/(Nm2 )
(c) dθ/d = −0.509 deg/deg
(d) Most sensitive to changes in sweep angle ().
l(bs /2)4
8.20 Use = 8E I , L/2 b4s = k, = (bs /2), and L = lbs .
8.21
y m̄η
(a) Use θ (y) = 0 G J dη
-0.5
-0.4
-0.3
θ (rad)
-0.2
average
-0.1
0
0 0.5 1
y (m)
Partial Answers to Selected Problems 543
b /2
(b) Use θaverage = bs1/2 0 s θ (y)dy to find θaverage = 1/24 rad
(c) Use k T θaverage = m̄bs /2
8.22
(b)
Λ=
0.6 0
15
30
0.4 45
0.2
α, θ, Γ (deg)
Λ= Λ=
0
15 0
0
15 45 30
−0.2 30
45
−0.4 α
θ
Γ
0 0.2 0.4 0.6 0.8
Mach number, M (~)
8.24
(b)
0.6
h=
0
0.4 3
6
9
α, θ, Γ (deg)
0.2
h=
0 9
6
3
h= 0
−0.2 9
6
3
0
−0.4 α
θ
Γ
0 0.2 0.4 0.6 0.8
Mach number, M (~)
Glossary
Latin symbols
Greek Symbols
Vectors
Subscripts
0 Stagnation
∞ Free stream conditions
α Angle of attack
θ Polar angle
Φ Azimuth angle
al Attachment line
aw Adiabatic wall
b Bending
C Center
crit Critical condition, i.e. ML = 1
d Drag
dd Drag divergence
e Boundary layer edge
f Friction or flutter
h Horizontal tailplane or bending
i, j Indices: 1,2,3
i Incomnpressible or induced
k Kinematic
L local
l Lower or lift
LE Leading edge
m Moment
min Minimum
N Nose
n Normal
p Pressure
r Radius
s Surface, static or structural
t Total
T Turbulent, thermal, tip, or torsion
th Thickness or throat
Glossary 549
tr Transition
turb Turbulent
u Upper
w Wall, wave, or wing
Wet Wetted
Supscripts
* Critical condition, i.e. ML = 1
+ Normalized distance or speed within the boundary layer
Acronyms
AGARD Advisory Group for Aerospace Research and Development
AIAA American Institute of Aeronautics and Astronautics
CFD Computational Fluid Dynamics
CS Certification Specification
DC Douglas Company
DNS Direct Numerical Simulations
DoF Degrees of Freedom
DSMA Douglas Santa Monica Airfoil
ETW European Transonic Windtunnel
FAR Federal Aviation Regulation
FTC Fundamental Theorem of Calculus
ISA International Standard Atmosphere
KD Kinetic Diagram
LCO Limit Cycle Oscillation
LHS Left Hand Side
LRW Left Running Wave
MD McDonnell Douglas
MiG Mikoyan-and-Gurevich
MLN Minimum Length Nozzle
MOC Method of Characteristics
NACA National Advisory Committee for Aeronautics
NASA National Aeronautics and Space Administration
NLF Natural Laminar Flow
NLR Nationaal Lucht-en Ruimtevaartlaboratorium
NS Navier-Stokes
NTF National Transonic Facility
ODE Ordinary Differential Equation
P-M Prandtl-Meyer
PDE Partial Differential Equation
RAE Royal Aircraft Establishment
RAF Royal Air Force
RANS Reynolds Averaged Navier Stokes Equations
RHS Right Hand Side
550 Glossary
A B
Adiabatic, 46 Blasius equation, 327
Aerodynamic Boundary layer, 303
center, 482 blending layer, 333
damping, 492 confluence, 404
stiffness, 492 crossflow, 461
twist, 476 laminar, 303
Aeroelastic merging layer, 333
penalty, 489 overlap layer, 333
tailoring, 476 separation, 302, 304, 470
Aeroelastic penalty, 475 shock wave interaction, 319
Aeroelasticity, 474 thickness, 305, 306
dynamic, 474 transition, 343, 372, 464
static, 474 turbulent, 313
Aeroservoelasticity, 495 Boussinesq assumption, 67
Aft loading, 376 Bréguet range formula, 2
Aileron Buffet, 405
high-speed, 488 onset boundary, 405
reversal, 481 transonic, 405, 412
Airfoil, 368
bi-convex, 173, 198
bi-convex parabolic, 200 C
circular-arc, 173 Calorically-prefect gas, 150
hexagonal, 198, 200 Camber, 369
laminar flow, 372 line, 369
natural-laminar-flow, 374 Canonical form, 35
‘peaky’, 382 Center of mass, 491
shock-Free, 379 Characteristic
sonic rooftop, 390 coordinates, 24
supercritical, 375 curves, 35
Angle-of-attack Characteristic Mach number, 150, 151
effective, 476 Characteristics, 211
Area rule, 12 Chord
supersonic, 13 length, 369
transonic, 288 Clauser’s equilibrium parameter, 335
Attachment line, 465 Coles, 333
instability, 465 Coles’ wake parameter, 333, 362
© Springer Science+Business Media Dordrecht 2015 551
R. Vos and S. Farokhi, Introduction to Transonic Aerodynamics,
Fluid Mechanics and Its Applications 110, DOI 10.1007/978-94-017-9747-4
552 Index