You are on page 1of 6

Proceedings of the URECA@NTU 2006-07

Computational Fluid Dynamics for Aircraft Design



Kelvin Kwong Lam Loh
School of Mechanical & Aerospace
Engineering

Asst Prof Kuan Tek Seang
School of Mechanical & Aerospace
Engineering

AbstractThis project is an exploration of the usage
of computational fluid dynamics in aircraft design as
well as the aircraft design process itself. The C-type
grid is explored and used extensively in the NACA
0012 and Clark Y airfoil cases. The software package
used in this project is the commercial CFD code Fluent
and the Gambit pre-processor. As for the aircraft
design process, the design of a light sport aircraft is
used as a case study using the Advanced Aircraft
Analysis (AAA) program.
KeywordsCFD, Fluent, Gambit, NACA 0012,
Clark Y, Aircraft design, Light sport aircraft, AAA

1. INTRODUCTION
Aircraft design involves various phases. Each phase
concerns different fidelity of the aircraft models. For
example, in the conceptual design phase, most aircraft
models are based on estimated sizing methods (i.e.
historical data) and simple physics based equations to
create an aircraft to satisfy the mission requirements.
Numerical methods are used to simulate the models
and help designers in choosing the most suitable design
for the given requirements. Computational fluid
dynamics (CFD) codes help the designers in
determining the aerodynamics of the particular design.
Designers are then able to design better aircrafts with a
better understanding of the flow fields around the
different aircraft designs. In industry, most working
aerodynamic engineers do not involve themselves in
writing CFD codes. Instead, they must learn to use
these codes while being aware of the underlying
assumptions behind the codes.

2. GRID GENERATION
For almost all CFD applications, the uniform,
rectangular grid cannot be used in the physical plane.
Problems such as the flow over an airfoil cannot be
solved using a uniform, rectangular grid because of the
following reasons:
1. Some of the grid points fall inside the airfoil,
and they are completely out of the flow. It
cannot be determined the values of the flow
properties that these points have.
2. The number of grid points that fall on the
surface of the airfoil is too few for a good
determination of the flow. This is because the
airfoil surface is a vital boundary condition and
therefore, the airfoil surface must be clearly
and strongly seen by the numerical solution.
Figure 1 shows the uniform, rectangular grid over an
airfoil. The problems can be seen in the figure. As a
result, a non-uniform, curvilinear grid which is literally
wrapped around the airfoil is appropriate. Figure 2
shows an O-type non-uniform, curvilinear grid over an
airfoil. New coordinate lines, ! and ", are defined such
that the airfoil surface becomes a coordinate line, " =
constant. This is a boundary-fitted coordinate system.
In the physical space shown in Figure 2, the grid is not
rectangular, and is not uniformly spaced. Therefore, a
transformation of the curvilinear grid mesh in physical
space to a rectangular mesh in terms of ! and " must
be done. This is shown in Figure 3. The rectangular
mesh shown in Figure 3 is called the computational
plane.
Grid generation is a very active area in CFD research.
Various grids such as the simple stretched grid,
compressed grid, the boundary-fitted coordinate
systems (e.g., O-type and C-type grids for airfoils), and
adaptive grids are used to effectively capture the flow
for a particular flow problem. Adaptive grids are grid
networks that automatically clusters grid points in
regions of high flow field gradients (i.e., it uses the
solution of the flow field properties to locate the grid
points in the physical plane) [1]. As a side note,
adaptive grids are still very new in CFD, and the
advantages of having an increased accuracy for a fixed
number of grid points are not well established.

Figure 1 A uniform, rectangular grid
For this project, the commercial software, Gambit is
used to mesh the domain of the flow problem. Figure 4
shows a course mesh done using Gambit for a C-type
grid for the NACA 0012 airfoil. This figure is used to
Proceedings of the URECA@NTU 2006-07

show the difference between the C-type and O-type
grid. The C-type grids have the " = constant lines
trailing downstream to the right not totally enclosing
the airfoil (except on the surface of the airfoil). Figure
5 shows a finer mesh used for the transonic flow over a
NACA 0012 airfoil problem and for the lift curve at
near stall angles of attack to fully capture flow
separation.
Figure 6 shows a meshed generic airplane without
empennage using tetrahedral cells and through size
functions attached to the wing root and tip as well as
the exhaust of the engine nacelles. Figure 7 shows a
generic meshed wing produced in Gambit. The size
functions are attached to the symmetry plane with the
root of the wing as the source.

Figure 2 An O-type non-uniform, curvilinear grid

Figure 3 Computational plane

Figure 4 Meshed C-type grid in Gambit

Figure 5 Finer mesh in Gambit

Figure 6 Meshed generic airplane

Figure 7 Meshed wing in Gambit

3. AIRFOIL AERODYNAMIC
CHARACTERISTICS
In aircraft design, the aerodynamic characteristics of
the airfoil selected determine the flying qualities of the
aircraft as well as affects the lift and drag of the
aircraft. Therefore, it is of great importance that the
aerodynamic characteristics of the airfoil be known to
the designer during the conceptual design phase so that
the aircraft do not have to undergo major changes
during the preliminary design phase.
Proceedings of the URECA@NTU 2006-07

3.1 NACA 0012 AIRFOIL
The aerodynamic characteristics for the NACA 0012
airfoil has been tested out and verified by various CFD
codes as well as experimentally [2]. Therefore, this
airfoil is suitable for this project to verify the results
obtained by the different methods and models used.
Figure 8 shows the lift curve for the NACA 0012
airfoil obtained using the most suitable model and
method for this airfoil.
Figure 8 NACA 0012 lift curve
The mesh used is the fine mesh C-type grid from
Figure 5. This is to allow for the mesh size to be small
enough for the wall y+ to be less than 1 so that the
shear stress transport (SST) k-# viscous model used in
the computation is accurate enough. The vertex points
for the airfoil are plotted using the Javafoil program.
The resulting mesh has 196566 cells.
The steady fully implicit pressure based solver is used.
The gradient option for the solver is green-gauss cell
based. The viscous model used is the SST k-# model
without transitional flows. For the determination of the
lift curve of the airfoil, the energy equation is turned
off while the density and viscosity of air are assumed
to be constant. The boundary conditions are set up such
that the freestream velocity is 87.644 ms
-1
and the
turbulence is specified by the turbulent intensity and
length scale of 1% and 0.01 m respectively.
The discretization scheme used is the PRESTO!
scheme for the pressure variable, and the second order
upwind scheme for the momentum, turbulent kinetic
energy, and specific dissipation rate variables. The
semi-implicit method for pressure-linked equations
revised (SIMPLER) algorithm is used as the solution
algorithm for the pressure-velocity coupling in the
steady flow.
The case is calculated repeatedly using different angles
of attack. For the angles of attack that are near the stall
angle of attack, the mesh is adapted after the first
solution has been obtained so that the airfoil wall y+
remains less than 1.
As shown in Figure 8 the lift curve of the airfoil, the
results of this method are accurate as compared to the
experimental data [2].
For the transonic flow case, the k-$ model is used
instead with the energy equation included and the
solver being a coupled solver. The density of air is
based on the ideal gas equation and for the viscosity of
air, Sutherlands formula is used to model the change
in viscosity of the fluid with changing temperature.
The boundary conditions are Mach number = 0.8 and
1.5
o
angle of attack with a temperature of 288.16 K.
Figure 9 shows the pressure coefficient contour plot for
the transonic flow over the airfoil. This case is done
out of curiosity and is not compared with any
experimental data. However, the position for the
normal shocks does seem to closely match the FLO36
code solution [3].

Figure 9 Transonic flow over NACA 0012
3.2 CLARK Y AIRFOIL
This airfoil is done as a familiarization with the Fluent
and Gambit interfaces. However, the results obtained
can still be used to validate the current state-of-the-art
methods.
The mesh used is a course mesh similar to the one in
Figure 4. The resulting mesh has 12150 cells.
The models and solvers are the same as for the NACA
0012 transonic flow case. The boundary conditions are
Mach number = 0.8 and 5
o
angle of attack at 300 K
temperature. Figure 10 shows the pressure coefficient
contour plot for the transonic flow over the Clark Y
airfoil.
Figure 11 shows the pressure coefficient contour plot
for an inviscid flow model over the airfoil at a
freestream velocity of 50 ms
-1
and a 5
o
angle of attack.
Figure 12 shows the same pressure coefficients in an
X-Y plot.

Figure 10 Transonic flow over Clark Y
Proceedings of the URECA@NTU 2006-07


Figure 11 Inviscid flow over Clark Y

Figure 12 Pressure coefficient plot for inviscid flow
4. AIRCRAFT DESIGN
In this project, the aircraft design part is covered by the
practical application of the design process as outlined
in most textbooks. A case study is done by designing a
special light sport (S-LSA) aircraft using the methods
outlined by design texts as well as the aid of software
programs such as Advanced Aircraft Analysis (AAA)
and Solidworks. Also, in this aircraft design section
of the paper, the units are primarily in Imperial units.

4.1 THE S-LSA AIRCRAFT
The S-LSA aircrafts are primarily used for
recreational purposes and a sport pilot may only
operate the aircraft during daylight hours [4]. The
requirements for this aircraft are as follows:
1. Maximum take-off weight of not more than
1,320 pounds (600 kilograms) for aircraft
not intended for operation on water.
2. Maximum airspeed (full-power, max
continuous) VH of not more than 120 knots
CAS under standard atmospheric conditions
at sea level.
3. Maximum flaps-up (no lift enhancing
devices) stalling speed of not more than 45
knots CAS at the aircraft maximum
certificated take-off weight and most
critical center of gravity.
4. Maximum of two seats.
5. Single, reciprocating engine (includes rotary
or diesel engines), non-turbine.
6. Fixed or ground-adjustable propeller.
7. Non-pressurized cabin.
8. Fixed landing gear.
9. Fixed wing.
10. Can be manufactured, ready to fly under
the Special Light Sport (S-LSA) aircraft
certification without FAR 23 compliance.
11. Performance requirements from ASTM F-
2245 04 section 4.4.
Based on these requirements, a tandem seater design is
selected with the passenger seated at the back and
higher than the pilot. A single tailboom is used instead
of a conventional fuselage tail. Figure 13 shows the
mission profile for this aircraft. Table 1 shows the
summary of the weights obtained from the fuel
fractions method.

Figure 13 Mission profile
The aircraft is designed to meet performance and
airworthiness objectives in the following categories:
1. Stall speed
2. Take-off field length
3. Climb requirements
4. Maximum speed
5. Landing field length

Table I. Weights summary
Payload weight 240.0 lbs
Crew weight 200.0 lbs
Mission fuel weight 121.4 lbs
Trapped fuel and oil
weight
13.0 lbs
Useful load 561.4 lbs
Empty weight 727.0 lbs
Maximum take-off weight 1301.4 lbs



Proceedings of the URECA@NTU 2006-07

The aircraft parameters that drive the performance are:
1. Wing area, S
w

2. Take-off power, P
TO

3. Maximum required lift coefficient with flaps
up, C
Lmax_cln

4. Maximum required take-off lift coefficient,
C
Lmax_TO

5. Maximum required landing lift coefficient,
C
Lmax_L

Figure 14 shows the complete performance matching
sizing plot. To satisfy the performance requirements,
the design point must be located below the FAR 23.65
rate of climb with all engines operational curve and
less than the stall speed sizing line. The design point
chosen for this aircraft is the wing loading at take-off
having a value of 9.32 lb/ft
2
and a power loading at
take-off of 16.0 lb/hp. This corresponds to a wing area
of 138.85 ft
2
and a required power of 81 hp. The Rotax
912UL fits the power requirement for this design point
and is chosen as the power plant for this aircraft.

Figure 14 Performance sizing matching plot
The airfoil selection for the wing is limited to only
NACA airfoils. In particular, the chosen airfoil for the
wing is the NACA 23012 airfoil as it has relatively
good lift characteristics and approximately produces no
moment at the aerodynamic centre although the airfoil
does suffer in poor stall characteristics. The aspect
ratio of the wing is determined to be 10.0, with no
sweep angle. Taper ratio of the wing is decided to be
one to have a constant chord to minimize the cost. The
wing area is determined to be 138.85 ft
2
. This gives a
chord length of 3.73 ft and a span of 37.26 ft. The
trimmed lift coefficient of the whole aircraft is
determined to be 0.5585.
The sizing of the empennage is done via the volume
coefficient method. The volume coefficient for the
horizontal tail is assumed to be 0.6128 and for the
vertical tail to be 0.0435. This corresponded to a
horizontal and vertical tail area of 18.0 ft
2
and 13 ft
2
.
Just like the wing, the horizontal and vertical tails both
have no sweep angle and both have a constant chord.
Figure 15 shows the trim diagram for the aircraft. The
trim triangle represents the trim envelope of the
airplane, bounded by the forward and aft center of
gravity lines, and the airplane stall angle of attack line.
To ensure that control power exists throughout the
center of gravity envelope, the elevator deflection lines
should cover the entire trim triangle, as is the case in
Figure 15. This shows that there is enough longitudinal
control such that the aircraft can be controlled up to the
stall angle of attack. For stability, the horizontal tail
area of 18.0 ft2 meets the minimum static margin
requirement of higher than 10%.

Figure 15 Trim diagram
The fixed landing gear for the aircraft is chosen to have
the conventional tricycle configuration. It is initially
determined that the load will purely be absorbed by the
tires. The static load for the nose gear strut is
calculated to be 208.384 lbf. The static load imposed
on the main gear strut is determined to be 546.5 lbf.
The static loads are calculated from moment equations.
The coordinates for the nose wheel contact point on the
ground with respect to the nose of the aircraft as origin
are (1, 0, -2.62) where the units are in ft. Similarly, one
side of the main wheel contact point coordinates are
(9.0488, 5, -2.62). The longitudinal tip-over criterion
for tricycle gears are fulfilled with the angle between
the main gear contact point and the most aft center of
gravity (CG) at 15
o
. As for the lateral tip-over criterion,
it is found that the minimum Y displacement from the
aircraft nose is 5 ft for a tip-over angle of less than 55
o
.
The ratios of static load and maximum take-off weight
give an estimate on the tire dimensions and the internal
pressure [5]. For the aircraft, the ratio for the nose gear
is 0.1601 and the corresponding tire size is determined
to have a diameter of 9 inches and a tire width of 3.4
inches at 45 psi. As for the main gear, the ratio is
calculated to be 0.8399 and the corresponding tire size
is determined to have a diameter of 15 inches and a tire
width of 5 inches at 45 psi.
Figure 16, Figure 17, and Figure 18 shows the three
views of the aircraft in Solidworks. Figure 19 shows
the rendered (dimetric) view of the aircraft in
Solidworks.
Proceedings of the URECA@NTU 2006-07


Figure 16 Front view

Figure 17 Top view

Figure 18 Right view

Figure 19 Rendered (dimetric) view

5. CONCLUSION
Commercial CFD codes such as Fluent helps designers
determine the aerodynamic characteristics of
conceptual aircrafts in a virtual environment. Currently
there are no standards in which a designer can simply
analyze the aircraft without going through the
complicated process of grid generation, model, and
solvers selection for each flight condition that the
designer wants to analyze. However, given the
increased computing power over time and more
extensive studies on grids, models, and solvers, CFD
codes can help designers in validating their conceptual
designs quicker and with much greater accuracy than
current conceptual design practices.

ACKNOWLEDGMENT
I would like to acknowledge the help and support
provided by my advisor, Dr. Kuan Tek Seang
throughout this project and to Dr. Yongki Go for
providing the AAA software as well as to the lab
technicians over at the main hangar for giving me the
space to complete this project. I also wish to
acknowledge the funding support for this project from
Nanyang Technological University under the
Undergraduate Research Experience on Campus
(URECA) programme.

REFERENCES
[1] J.D. Anderson, G. Degrez, E. Dick, and R.
Grundmann, Computational Fluid Dynamics: An
Introduction, A Von Karman Institute Book series,
John F. Wendt (Editor), Springer-Verlag, NY,
USA, 1992.
[2] I.H. Abbott, A.E. von Doenhoff, and L.S. Stivers,
Jr., Summary of Airfoil Data, Report No. 824,
NACA, Langley Memorial Aeronautical
Laboratory, Langley Field, VA, USA, 1945.
[3] W.H. Mason, Transonic Aerodynamics of
Airfoils and Wings, Chapter 7, AOE 4124
Configuration Aerodynamics Class Text, Virginia
Polytechnic Institute and State University,
Blacksburg, VA, USA, 2007.
[4] W. Anemaat, B. Kaushik, and
K.Wiedenmann,Preliminary design of a tandem
seater light sport airplane, Paper 2006-01-2394,
General Aviation Technology Conference and
Exhibition, SAE International, Kansas, USA,
August 29-31, 2006.
[5] J. Roskam, Airplane Design Parts I toVII, 2
nd

Edition, Roskam Aviation and Engineering
Corporation, Kansas, USA, 1989.

You might also like