You are on page 1of 8

machine design, Vol.6(2014) No.

3, ISSN 1821-1259

pp. 71-78
Original scientific paper

DETERMINATION OF AERODYNAMIC CHARACTERISTICS OF A LIGHT AIRCRAFT


USING VISCOUS CFD MODELING
Zoran STEFANOVI1 - Ivan KOSTI1, * - Olivera KOSTI1
1

University of Belgrade, Faculty of Mechanical Engineering, Belgrade, Serbia

Received (28.04.2014); Revised (16.06.2014); Accepted (19.06.2014)


Abstract: The CFD calculations with included viscous effects represent the latest analyses, aimed for the assessment of
aerodynamic characteristics of a new light airplane. During the evolution of such an aircraft, appropriate calculation
methods should be applied at the different design development stages. The general trend is application of fairly simple,
but reliable analytical and semi-empirical methods at the initial stages, combined with simplified - inviscid CFD
computational models, and complex viscous CFD analyses at its final design levels. At the present stage of light
aviation development, it is assumed that the contemporary design tools for each of those steps should be appropriate
enough, so that they actually verify and additionally fine-tune each other's results. Here presented viscous CFD
calculations have been based on the application of unstructured meshes with reasonably small number of elements,
combined with robust physical model, involving RANS k- SST turbulence treatment, which has enabled analyses at
around-critical and post-stall angles of attack. These calculations have provided very useful, both quantitative and
qualitative aerodynamic data, while the results for aerodynamic coefficients have shown fair agreements with those
obtained by methods at previous design levels
Key words: light aircraft, aerodynamic design, viscous CFD calculations, unstructured mesh, RANS k- SST

1. INTRODUCTION
This paper is primary focused on the latest stage of the
aerodynamic analyses of a new light aircraft (NLA),
designed at the Innovation Centre of the Faculty of
Mechanical Engineering, University of Belgrade. This is a
two-seat, single piston-engine trainer, of metal primary
structure, with plastic composite engine cowling, wing
and tail tips, aimed for primary and advanced flight
training under VFR and IFR conditions, aerobatic
training, sport flying, etc. (see Fig. 1.).
From the historical perspective, in the initial stages of the
aerodynamic computational analyses of NLA, a fairly
simple CFD model based on 3D vortex lattice method
(VLM) was used [1, 2]. Such inviscid calculation model
inherently neglected boundary layer influence and
separation effects, so the effectiveness of the flaps and
control surfaces were overestimated at moderate and
higher deflection angles, and parasite drag components
could not be determined. Also, the lift and moment curves
could have been determined only in their linear domains,
and just induced drag polars could be calculated. In order
to overcome these problems and obtain complete final
diagrams, the authors have applied the hybrid method
approach, developed in two directions.
First, to obtain proper effectiveness of flaps and control
surfaces within the CFD calculations, the non-linear
calibration factors had been successfully derived, from
wind tunnel test data of a domestic aircraft of similar

category. They also included the additional circulation


correction parameters, which took into account the
different lifting characteristics and influences of applied
slotted flaps, instead of the plane flaps simulated by VLM
[3, 4]. Second, the parasite drag components, which can
not be obtained by the applied VLM calculations, were
determined using reliable analytical and semi-empirical
methods. These results have been superimposed with the
induced drag values, which VLM successfully calculates
for many combinations of flaps and elevator deflections,
and angles of attack [5].
On the other hand, these methods also have some
inevitable inherent shorcomings: The VLM (also called
the "inviscid CFD") omits viscous effects and can not
verify analytically determined aerodynamic predictions at
arround-critical angles of attack; obtained parabolic polars
give minimum drag at zero lift, while actual minimums
are at some small positive lift coefficients, etc.
Due to that, at the final stage of aerodynamic design, with
general aircraft configuration finally "frozen" (Fig. 2.),
aerodynamic analysis required the application of the CFD
model with includes viscous effects, taking into account
both micro and macro-vorticity influences on the
flowfield around the airplane. This way, the whole range
of angles of attack could have been investigated,
including deep stall (beyond critical) angles. Considering
control surfaces, in this paper only flaps deflection cases
have been considered, and all results have been compared
with those obtained by previously applied methods.

*Correspondence Authors Address: University of Belgrade, Faculty of Mechanical Engineering, Kraljice Marije 16,
11120 Belgrade 35, Serbia, ikostic@mas.bg.ac.rs

Zoran Stefanovi, Ivan Kosti, Olivera Kosti: Determination of Aerodynamic Characteristics of a Light Aircraft Using Viscous CFD Modeling;
Machine Design, Vol.6(2014) No.3, ISSN 1821-1259; pp. 71-78

relevant details have been omitted (like spaces between


shock absorber struts and torque arms, etc.).
In cases with flaps deflected the accurate 3D geometry of
the convergent slot between the single-slotted flaps and
the wing structure had to be modeled (more detailed
information about applied flaps type can also be found in
[3] and [6]). On the NLA, flaps extend from the ailerons
to the trapezoidal centerplane segments of the wing (see
Fig. 1. - (3) and Fig. 2.).

Fig.2. Loft for CFD calculations, flaps deflection = 30o

Fig.1. The NLA development: (1) initial conceptual


design sketch, (2) 3D CAD model, (3) loft for viscous
CFD analyses, (4) airframe used in structural tests, and
(5) completely equipped full size airplane mockup

2. THE CFD CALCULATION MODEL WITH


VISCOUS EFFECTS
To generate tree airplane lofts, necessary for the CFD
analyses, the CAD model (Fig. 1. - (2)) has been used,
with flaps deflections of = 0o, = 20o and = 30o. All
other undeflected control surface gaps on the airplane
have been "sealed", in order to reduce the complexity of
the analyses, without any substantial penalties considering
the overall accuracy of the results. Also, some less
72

The selection of the meshing method and the applied


physical model complexity was based on an optimum
compromise, dictated by the available hardware
resources.
After a certain number of test runs, the optimum choice,
which gave satisfactory outcomes, was based on the
following:
(A) Application of a mesh with reasonably small number
of elements, in this case of the order of 1.000.000
tetrahedral elements for half-model calculations (Fig. 3.),
the exact numbers vary slightly for different flaps
configurations. It should be noted that for very large and
complex configurations, the fine meshes can exceed tens
of millions of elements, but even with very powerful
hardware, the CPU time for a single run might be
measured in hundreds of hours.

Zoran Stefanovi, Ivan Kosti, Olivera Kosti: Determination of Aerodynamic Characteristics of a Light Aircraft Using Viscous CFD Modeling;
Machine Design, Vol.6(2014) No.3, ISSN 1821-1259; pp. 71-78

per one angle of attack had been reduced to only 1 2


hours (longer time required for higher angles of attack).

3. DISCUSSION OF THE RESULTS


The CFD aerodynamic analyses have been performed for
angles of attack [o] ranging from negative, to positive
post-stall values, at Reynolds number MRe 5.3
calculated with respect to the wing chord, for all flaps
positions (to retain full compatibility of the results,
although such MRe value would be a bit too high for
operational flaps applications). The analyses have
provided very useful quantitative and qualitative results.

3.1. Quantitative Assessments

Fig.3. Mesh applied for CFD calculations, in the vicinity


of the model
(B) Application of a very sophisticated calculation
method, capable of dealing with separation effects at high
angles of attack, fully taking into account compressibility
effects (for example, as for transonic 3D flow
calculations, although compressibility influence at the
NLA's cruising Mach number of M = 0.15 is rather
small), etc. Calculation of angles of attack around and
beyond critical has become possible after the RANS
(Reynolds-Averaged Navier-Stokes) k- SST (ShearStress Transport) turbulent model [7], [8], [9] has been
applied, so this model has been adopted as standard for all
calculations.
Since only symmetrical cases have been analyzed, another
reasonable simplification was the analysis of half-models,
instead of full (also see Fig. 3.), while for the visual
presentations, the images were generated using mirroring
option with respect to the plane of symmetry. The
solution convergence had been boosted by the initial
optimum reordering of the mesh domain using the
Reverse Cuthill-McKee method [10], the application of
FMG - the Full Multi-Grid solution initialization at 4
levels [9], [10], and by active solution steering, applying
the automatic optimization of Courant number for the
achieved solution convergence stage. It had been assumed
that the solution for the given angle of attack has
converged when the solution monitors for lift, drag, and
pitching moment coefficients show no change (constant
values) within the last 100 iterations.
By this approach, with the available hardware, the
required CPU time for the convergence of the solutions

Obtained numerical results can generally be divided into


global and detailed, where the most important global
assessments imply the lift coefficient CL, drag coefficient
CD and pitching moment coefficient CM about the
nominal CG (center of gravity) position, for the entire
airplane configuration, with flaps at three characteristic
positions (retracted, i.e. = 0o, and deflected to = 20o
and = 30o).
The results are graphically shown in Fig.'s 4. and 5, and in
Tables 1. 3.

Fig.4. Diagrams of lift and moment coefficients


for three flaps positions
Presently it has been assumed that these two flaps
positions would satisfy their purpose for take-off and
landing applications, on such a light aircraft (although
with their electric actuator, any angle can be established).
73

Zoran Stefanovi, Ivan Kosti, Olivera Kosti: Determination of Aerodynamic Characteristics of a Light Aircraft Using Viscous CFD Modeling;
Machine Design, Vol.6(2014) No.3, ISSN 1821-1259; pp. 71-78

flaps deflection, CFD calculations indicate rather abrupt


stall, which presents no particular problem at normal
flying altitudes. On the other hand, during the landing
phase, flaps deflections change it to smooth and sustained
stall behavior, specially at = 30o. This is of primary
importance in case of student pilots, whose judgment of
ground proximity before touch down might be poor - even
when at minimum speed, the airplane will smoothly glide
towards the ground.
Table 3. Calculated aerodynamic coefficients for = 30o

[o]

Fig.5. Polar curves for three flaps positions


Table 1. Calculated aerodynamic coefficients for = 0o

[ o]
-4
0
3
6
10
14
15
16
18

CD
0.05003
0.04429
0.05214
0.07032
0.10976
0.16689
0.18233
0.20091
0.27813

CL
-0.184
0.192
0.477
0.760
1.122
1.438
1.468
1.431
1.280

CM
0.1212
0.0446
-0.0088
-0.0571
-0.1184
-0.1863
-0.2106
-0.2621
-0.4380

CL / CD
-3.67
4.34
9.15
10.81
10.22
8.62
8.05
7.12
4.60

Table 2. Calculated aerodynamic coefficients for = 20o

[ o]
-5
-3
0
5
10
12
13
14
16

CD
0.06143
0.05949
0.06516
0.09564
0.15023
0.17666
0.19100
0.20663
0.24574

CL
0.023
0.212
0.495
0.958
1.376
1.496
1.531
1.534
1.481

CM
0.1064
0.0655
0.0066
-0.0823
-0.1603
-0.1980
-0.2253
-0.2540
-0.3434

CL / CD
0.38
3.57
7.59
10.01
9.16
8.47
8.01
7.43
6.03

Presented results, obtained by CFD analyses, show


moderate increase in maximum lift coefficient with flaps
deflection. With the applied flaps design and deflections,
the configuration fully satisfies the EASA CS-23
regulation requirements considering minimum speeds.
The influence of flaps is more important considering the
airplane's stall characteristics (Fig. 4.). Namely, without
74

-7
-4
0
4
8
11
12
13
14
15
16

CD
0.07960
0.07476
0.08373
0.10905
0.14968
0.18903
0.20273
0.21559
0.23141
0.24912
0.29063

CL
-0.049
0.237
0.611
0.979
1.326
1.556
1.616
1.619
1.618
1.599
1.383

CM
0.1299
0.0688
-0.0091
-0.0816
-0.1459
-0.1907
-0.2065
-0.2368
-0.2636
-0.3050
-0.3785

CL / CD
-0.61
3.17
7.29
8.98
8.86
8.23
7.97
7.51
6.99
6.42
4.76

The assessed maximum value of (CL / CD) max 11 is


quite realistic and satisfactory for light aircraft of metal
design (values in the range of 10 12 are most commonly
reported by pilots in operational use, for such airplane
category).
The detailed numerical results imply that each of the
coefficients shown in Tables 1. 3. can be divided into
separate contributions of the airplane loft members, also
called the zones, or "named selections". Table 4. provides
an example of the detailed contributions to the drag
coefficient CD = 0.20273, obtained for = 30o and angle
of attack = 12o (see also Table 3.).
Table 4. Detailed contributions to total drag coefficient,
for = 12o and = 30o (number of decimal digits is
shown as provided by the software)
Zone
flap
wing
nose leg
main leg
fuselage
horiz. tail
vert. tail
Total

Pressure

Viscous

Total

0.061359989

0.000577269

0.061937259

0.091292484

0.009768767

0.101061250

0.001511985

5.11388e-05

0.001563124

0.005003766

0.000163604

0.005167371

0.017627113

0.006439333

0.024066446

0.005470524

0.001742480

0.007213005

0.000730859

0.000993846

0.001724705

0.18299672

0.01973644

0.20273316

3.2. Qualitative assessments


The most relevant qualitative characteristics of the
airplane design have been obtained using flow pattern
(represented by eddy viscosity) and pressure distribution
visualization tools (Fig. 6.).

Zoran Stefanovi, Ivan Kosti, Olivera Kosti: Determination of Aerodynamic Characteristics of a Light Aircraft Using Viscous CFD Modeling;
Machine Design, Vol.6(2014) No.3, ISSN 1821-1259; pp. 71-78

Fig.6. Examples of flow pattern (eddy viscosity - left), and


pressure (right) visualizations, = 6o and = 0o
In this sense, the CFD post-processing tools provide
information which are, at least, of the same relevance as
those obtained by the wind tunnel visualization
techniques. It must be kept in mind that the applied CFD
calculation method had previously been thoroughly tested
and verified by comparisons with the relevant
experimental data, used for the so called software
parameter calibration purposes.
Figure 7. shows the flow field pattern development for
flaps-retracted configuration, and angles of attack 10o.
The eddy, or vortex generated viscosity visualization
gives the most relevant insight in the flow characteristics
around, and behind the airplane. Small red colored
domains in Fig. 7. (3) indicate initial separation, which
occurs close to the wing roots (also see point 3 on CL
diagram), at the critical angle of attack of cr = 15o,
where maximum lift coefficient is achieved. One degree
more, at = 16o, the separated flow intensity obviously
drastically increases, indicating the airplanes tendency of
rather abrupt stall with flaps retracted.
Only three degrees beyond critical angle of attack, at =
18o, the airplane is in a deep stall condition, with very
massive separation, which is clearly identified by dark-red
and yellow domains, looking like fire coming out of the
wings (actually, these are the zones of very intensive air
turbulence). Very important is the fact that, at this angle
of attack, the bottom side of horizontal tail is out of the
turbulent domain (see Fig. 7. (5) right). This contributes
to the airplanes natural tendency for stall recovery, by
pitching the nose down towards smaller, sub-critical
angles of attack, which is very important for training
airplanes. Also, up to about = 14o, the horizontal tail is
also out of the turbulent flow generated by the wing (see
Fig. 7. (2) left). This way, flow visualization has
confirmed the appropriate positioning of the tail.

Fig.7. Quite abrupt stall, occurring with flaps retracted


75

Zoran Stefanovi, Ivan Kosti, Olivera Kosti: Determination of Aerodynamic Characteristics of a Light Aircraft Using Viscous CFD Modeling;
Machine Design, Vol.6(2014) No.3, ISSN 1821-1259; pp. 71-78

Figure 8. shows quite different behavior of the airplane in


landing configuration, with flaps deflected to = 30o. The
maximum lift coefficient is reached at about = 12o, and
this value remains practically constant until = 15o (Fig.
8. - (2) (5)), and after that the airplane drops into a deep
stall (Fig. 8. - (6)). During such sustained stall behavior,
the flow separation, which laterally occurs at the wingcenterplane junction (flaps root section, see Fig. 8.),
longitudinally slowly moves from the trailing edge, for
some 40% chord forward, within 3o of domain. That
keeps most of the wing area still "flyable" in this
range, and as already mentioned, helps student pilots to
perform smoother landings, considering their limited
flying experience during the initial flight training stages.

Fig.8. Sustained stall with flaps deflected to = 30o


76

Fig.9. Pressure distribution, flaps at = 30o , = 12o

Zoran Stefanovi, Ivan Kosti, Olivera Kosti: Determination of Aerodynamic Characteristics of a Light Aircraft Using Viscous CFD Modeling;
Machine Design, Vol.6(2014) No.3, ISSN 1821-1259; pp. 71-78

This phenomenon can directly be attributed to the applied


slotted flaps type. Through the convergent gap, generated by
their deflection (again see Fig. 2.), a certain amount of air is
accelerated towards the upper flaps surfaces, where the initial
stall begins. This air energizes the flow in the domains where
it tends to separate, suppresses the abrupt stall tendency, and
prolongs the generation of maximum lift.
A benefit of the presented visualization procedure, which
should be additionally emphasized, is the detection of
lateral spanwise position of initial flow separation. It is
obvious that in both cases, without and with flaps
deflected, it occurs close to the wing root, at the junction
of the outer rectangular wing and the trapezoidal
centerplane sections. Another requirement for primary
flight training airplanes is that the initial stall should
appear as close to the wing root as possible. In case of
asymmetric stalling conditions, where flow over one side
of the wing separates before the other (in sideslip, with
sideways wind gusts, etc.), such an airplane will show
little tendency to roll-over and fall into a spin, after stall.
Figure 9. shows the pressure distribution over the loft
with flaps fully deflected, calculated with respect to the
ambient atmospheric pressure at the cruising altitude,
taken as zero reference. By this approach, the underpressure is quantified by negative, and over-pressure by
positive values. The under-pressure, seen as bright yellow
domain over the upper wing surface overflows the
fuselage in the domain of the canopy, by which this part
of the fuselage additionally contributes to a certain extent
to the overall lift. It is known that in the presence of wing,
the fuselage generates more lift than the same fuselage
shape as an isolated body. In the domain of profiled wing
tips, which could be treated as very small winglets, the
same low pressure extends all the way to the trailing edge,
also improving lift in that part of the wing, etc.

calculations. On the other hand, the order of drag values


at lift coefficients used in standard operational flying
conditions has been mutually confirmed by all methods.

3.3. Comparisons with previous results


The results calculated by here presented CFD model are
compared in Fig.'s 10. and 11. with those obtained by
methods applied in previous design steps [3, 4, 5].
Keeping in mind the fact that several principally different
calculation methods have been applied during three
consecutive design development stages, the obtained
results for the NLAs aerodynamic characteristics show
quite fair agreements.
Lift coefficient curves give very close values for lift curve
slopes for all three configurations. Viscous CFD
calculations provide more optimistic assessments of
maximum lift coefficient values (higher in all three cases
than obtained by Datcom). The only principal difference
is in stall characteristics unlike Datcom, the viscous
CFD has predicted abrupt stall with flaps retracted, and
slow and sustained stall with full flaps extended.
Considering pitching moment characteristics, the VLM
and viscous CFD results coincide well for all three
configurations, where slopes of moment curves define the
measure of airplanes longitudinal static stability (pitching
moment had not been analyzed by Datcom).
Finally, certain differences in polar curves are readily
expected. Namely, in hybrid method calculations, polars
are obtained in simplified parabolic forms of the type
CD = A + B CL2, suitable for performance calculations,
which give minimum drag at zero lift coefficient. In
reality, minimum drag is achieved at certain small
positive lift coefficients, as predicted by viscous CFD

Fig.10. Comparisons of lift and moment curves


77

Zoran Stefanovi, Ivan Kosti, Olivera Kosti: Determination of Aerodynamic Characteristics of a Light Aircraft Using Viscous CFD Modeling;
Machine Design, Vol.6(2014) No.3, ISSN 1821-1259; pp. 71-78

while the range of angles of attack had been extended to


the post-stall values, compared with methods applied at
previous design steps. Calculations have provided a vast
scope of very useful quantitative and qualitative data.
Results obtained by all methods during the aerodynamic
development of NLA, including the latest CFD
calculations, have shown fair agreements for practical
engineering purposes. This way, the requirement posted
in contemporary airplane design, that the calculation tools
and methods applied at all different design levels should
generally confirm, supplement and fine-tune each-other,
has been satisfied.

REFERENCES
[1] Bertin, J., Smith, M. (1989). Aerodynamics for
engineers, Prentice - Hall International Editions,
ISBN: 0-13-018227-3, Englewood Cliffs, NJ
[2] Katz, J., Plotkin, A. (1991). Low Speed
Aerodynamics - From Wing Theory to Panel
Methods, McGraw-Hill, ISBN: 0-07-100876-4, New
York
[3] Stefanovi, Z. & Kosti, (2010). A Vortex Lattice
Method Application in Aerodynamic Analysis and
Design of Light Aircraft, Proceedings of the Sixth
International Symposium about Forming and
Designin in Mechanical Engineering, KOD 2010,
Pali - Serbia, ISBN: 978-86-7892-278-7, University
of Novi Sad - Faculty of Technical Sciences, pp. 8592, Graphic Center GRID, Novi Sad

Fig.11. Comparisons of drag polars

4. CONCLUSION
Aerodynamic analyses presented in this paper have been
performed using a CFD method which takes into account
viscous, but also compressibility effects (no matter how
small in case of light aircraft). The calculation model uses
mesh with a reasonably small number of elements, and a
very sophisticated physical model, based on the RANS k SST equations for turbulence. By this, and applying the
half-model analysis, the CPU time for calculations of
different angles of attack had been substantially reduced,
78

[4] Stefanovi, Z.; Kosti, I. & Kosti, O. (2011).


Efficient Evaluation of Preliminary Aerodynamic
Characteristics of Light Trainer Aircraft, Proceedings
of the International Conference of Innovative
Technologies, IN-TECH 2011, Bratislava, ISBN:
978-80-904502-6-4,
World
Association
for
Innovative Technologies, pp. 520-523, Tisk AS,
Jaromer
[5] Stefanovi, Z.; Kosti, I. & Kosti, O. (2012).
Primary Aerodynamic Analysis of a New Light
Aircraft in Symmetrical Flight Cinfigurations,
Proceedings of the Seventh International Symposium
- Machine and Industrial Design in Mechanical
Engineering, KOD 2012, Balatonfured - Hungary,
ISBN: 978-86-7892-399-9, University of Novi Sad Faculty of Technical Sciences, Slovak University of
Technology in Bratislava - Faculty of Mechanical
Engineering, ADECO, pp. 97-104, FTS Graphic
Center GRID, Novi Sad
[6] Stefanovi, Z. (2005) Aeroprofili (Airfoils),
University of Belgrade, Faculty of Mechanical
Engineering, ISBN: 86-7083-506-1, Belgrade
[7] Wilcox, D. C. (2006) Turbulence Modeling for CFD,
DCV Industries, Inc. ISBN 978-1-928729-08-2, San
Diego, California
[8] ANSYS FLUENT 14.0 (2011): Theory Guide,
ANSYS, Inc., Canonsburg, PA
[9] ANSYS FLUENT 14.0 (2011): User's Guide,
ANSYS, Inc., Canonsburg, PA
[10] ANSYS FLUENT 14.0 (2011): Tutorial Guide,
ANSYS, Inc., Canonsburg, PA

You might also like