AIAA Atmospheric Flight Mechanics Conference AIAA 2010-7637

2 - 5 August 2010, Toronto, Ontario Canada

Some applications of a panel method to aerodynamic

modelling of missiles

Antti Pankkonen*
Aalto University School of Science and Technology
(Helsinki University of Technology), Espoo, Finland

A high-order panel method is applied to compute longitudinal aerodynamic

characteristics of several missile configurations for which wind-tunnel test data are
available. Feasibility and limitations of the panel method in missile aerodynamic modelling
are then assessed based on comparison of computational and experimental results. An effort
is made to extend the range of applicability of the panel method to moderate angles of attack
by modelling separated vortices. Results show that non-linear aerodynamic characteristics of
conventional missile configurations can be predicted within preliminary design accuracy at
low to moderate angles of attack.

Nomenclature
Cm = pitching moment coefficient (Cm = CMY)
CN = normal force coefficient (CN = -CZ)
CX = force coefficient in the x direction (forward)
CY = force coefficient in the y direction (right)
CZ = force coefficient in the z direction (down)
CMX = moment coefficient around the x axis
CMY = moment coefficient around the y axis
CMZ = moment coefficient around the z axis
D = maximum body diameter
M = Mach number
Xcp = location of centre-of-pressure
α = angle-of-attack
δ = pitch control deflection
φ = roll orientation

Reference dimensions of the aerodynamic coefficients are maximum cross section area and maximum diameter of
missile body. Delta Δ preceding a symbol indicates difference.

I. Introduction

A lthough it has been decades since the panel methods were on the cutting-edge of computational aerodynamics,
these methods still find use as rapid modelling tools. Short set-up and execution time on a modern computer
and ability to handle arbitrary geometries make panel methods an attractive alternative as a midway between state-
of-the-art CFD and handbook methods. Coupled with automatic grid generation, panel methods can rapidly produce
aerodynamic data for flight simulation and other applications.
Several limitations and approximations must be accepted when utilising panel methods. This is especially true in
the supersonic regime, where the geometry must be sharp-nosed and relatively slender. In supersonic cases the
maximum local slope of solid walls is constrained by the Mach angle, which yields an upper limit for Mach number
on a given geometry. The panel methods being based on the linearised potential equation means that the transonic
regime must be considered off-limits, although computation is possible very near Mach number 1. The same is true

*
Research Engineer, Department of Applied Mechanics, Flight Mechanics Research Group. AIAA Member.
1
American Institute of Aeronautics and Astronautics

Copyright © 2010 by Antti Pankkonen. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
for flow cases that are dominated by viscous effects. For more comprehensive overview and theoretical background
of panel methods the reader is referred to Erickson's paper1. Previous applications of a panel method to missile
aerodynamic modelling have been presented by Fornasier and d'Espiney2.

II. Computational methods

A.Panel method software

The PANAIR program3 is utilised to solve the potential flow around missile configurations. The program is
capable of solving both subsonic and supersonic flow around an arbitrary geometry subject to the underlying
limitations of panel methods in general. The program has been modified to improve computational performance by
replacing several temporary files with memory storage. Also handling of wake network abutments has been made
more robust so that partially coinciding wake networks may exist in the grid without the program trying to create
unwanted abutments.
The geometry input to PANAIR is composed of topologically rectangular panel networks. A variety of different
surface singularity types and boundary condition specifications are available to formulate the problem. Impermeable
surfaces are generally modelled using panels with quadratically varying doublet strength and/or linearly varying
source strength. Continuity of source and doublet strength is enforced at panel network abutments, and user-
specified boundary conditions are enforced on panel centre control points.
The program computes surface pressure coefficients from perturbation velocities using linear, slender body,
second order, or isentropic pressure formula according to the analyst's choice. The isentropic formula is used
throughout this study.

B.Batch-run utility
Building an aerodynamic data set generally requires dozens or hundreds of computations in varying flight
conditions. In order to utilise the panel method efficiently and conveniently, a batch-run utility has been developed
to invoke grid generation and solution of the cases, and collection and post-processing of the results. Aerodynamic
coefficients can be tabulated for plotting or formatted as input for flight simulation.

C.Grid generation
An automatic grid generator has been developed to rapidly produce surface and wake grids for conventional
missile configurations. Wake geometry depends on angle-of-attack, roll orientation, and Mach number as described
in section II.D. Hence, a single grid is not sufficient to produce aerodynamic data in varying flight conditions, and
automatic grid generation is necessary to reduce workload of the analyst. The grid generator produces a complete
PANAIR input file that contains the flow case specification and program control parameters in addition to the
geometry and boundary conditions.
Missile configuration is built up of an axially symmetric body and up to five wing groups. Wings are modelled
as zero-thickness doublet networks, whereas body surface is represented by composite source-doublet networks.
Body surface is divided into panel networks at wing leading-edges and trailing-edges and surface slope discontinuity
points in the lengthwise direction, and along wing root chord lines in the transverse direction. Boundary conditions
are set to enforce velocity impermeability at panel centre control points, and body surface panels also enforce zero
perturbation potential inside the body.

D.Wake modelling
A prescribed wake is used to model the trailing vorticity of the lifting configuration. As a basic rule, the grid
lines of wake panel networks are aligned with the free-stream whenever this is possible without intersecting body
surface networks. Special treatment is required for wake networks that emanate from wings on the free-stream side
of the body. Near the body these wakes are deflected outwards from the centre-line to keep them outside the body.
Wing trailing-edge wakes are extended as constant-strength networks to the sideline of the body behind the wing
in order to avoid the singularity that would otherwise present itself at the junction of wing trailing edge and body. In
case of a deflected wing the root chord becomes separated from the body, and the wake is therefore extended along
the root chord to support non-zero doublet strength along the edge. Otherwise a high doublicity gradient, or
vorticity, would concentrate near the root chord causing unrealistic local velocities.
Wake networks are also attached to wing tips and leading edges when necessary to represent the side-edge or
leading-edge vortices of low-aspect-ratio wings and delta-wings. The separated vortices are not discrete, indeed it is
impossible to model a concentrated vortex with a continuous doublet network, and no effort is made to exactly

2
American Institute of Aeronautics and Astronautics
represent the shape of separated leading-edge vortices. Instead, a set of generic rules is used to locate maximum
vorticity above and slightly behind the leading-edge in order to model the gross effect of a separated vortex.
Leading-edge vortices are included only in cases where the leading-edge is subsonic and the inflow angle in a plane
perpendicular to the leading-edge exceeds 10º.

III. Results

A.Missile with strake-type wings and tail

In Reference 4, comprehensive wind tunnel measurements were carried out on a missile model with strake-type
wings and tail. Figure 1 depicts the geometry and surface panelling of this missile. Composition of the surface and
wake panel networks at α=10º and φ=45º is shown in Figure 2.
Longitudinal aerodynamic coefficients of this configuration are computed at four Mach numbers (0.6, 1.18, 1.7,
and 2.86), two roll orientations (0º and 45º), and six angles-of-attack from 0 to 18 degrees. A half-grid is sufficient
for the computations because only symmetric cases are considered. The half-grid consists of 932 surface panels at 0º
roll orientation as the wings coinciding with the plane of symmetry are omitted, and 1084 surface panels at 45º roll
orientation. A sequential single-process batch-run of the 48 cases takes about 10 minutes of computer time on a
commonplace workstation. Results at various Mach numbers and roll orientations are plotted in Figures 3-5. Figure
6 illustrates the distribution of surface pressure coefficients in one of the cases.

Figure 1. Surface panelling of missile model with strake-type wings and tail.

Figure 2. Surface and wake panel networks of the missile model.

3
American Institute of Aeronautics and Astronautics
 Direct comparison of computational axial force results
of Fig. 3 with wind tunnel measurements is obviously
pointless since the panel method solution represents
inviscid flow. Even so, it can be seen that supersonic
pressure drag increment and effects of angle-of-attack are
predicted within reasonable accuracy despite the zero-
thickness wings.
On the other hand, normal force and pitching moment
coefficients of Figures 4 and 5 are comparable to wind
tunnel results, and show fair agreement with the
experiments of Ref. 4. The reference results are not
reproduced here.
Non-linear behaviour of normal force and pitching
moment coefficients is evident in Figures 4 and 5, and
results from the side-edge vortices of the strake-type
wings and variation of wing-tail interference due to wake Figure 3. Axial force coefficients at various Mach
geometry. numbers.

Figure 4. Normal force (CN=-CZ) and pitching moment (CMY) coefficients at various Mach numbers.

Figure 5. Normal force (CN=-CZ) and pitching moment (CMY) coefficients at various roll angles.

4
American Institute of Aeronautics and Astronautics
 Figure 6. Top-side surface pressure coefficient (difference on wings) at α=10º, φ=0º, M=1.7.

A quantitative comparison of the results at various roll orientations (Fig. 5) with experimental results of Ref. 4
is presented in Figure 7. Normal force coefficients are seen to agree within ±5%. The disagreement of centre-of-
pressure locations seems rather severe at first glance, but the maximum difference, 0.67 diameters, is actually about
5% of the length of the missile and occurs at low angle-of-attack where the net aerodynamic force is small.
0.06 0.2
0.04 0
0.02
-0.2
0
ΔCN /CN Δxcp/D -0.4
-0.02
0
-0.04 0 -0.6
45
45
-0.06 -0.8
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
α α

Figure 7. Relative difference of experimental and computational normal forces and centre-of-pressure
locations at various roll orientations.

To illustrate the effect of wake modelling on the aerodynamic coefficients, the computations at various Mach
numbers are repeated with a fixed wake aligned with the body centre-line. Normal force and pitching moment
results are plotted in Figure 8. As expected the coefficient slopes at zero angle-of-attack are predicted reasonably
well with the fixed wake, but at increasing angle-of-attack the coefficients deviate wildly from the test data of Ref. 4
and the results of Figure 4 obtained with more comprehensive wake model. The non-linearity of aerodynamic
coefficients in Figure 8 arises mostly from the isentropic formula used in surface pressure computation.
Figure 9 illustrates the surface pressure coefficient in one case with fixed wake model. Steep pressure gradients
on the tail are due to the body-aligned wake that coincides with the tail and causes strong interference effects.
Comparison with Figure 6 indicates that aligning the wake with the free-stream reduces interference strength as the
wake is deflected away from the tail. The increased downwash from the side-edge vortices of the strake-wings
causes reduction in tail normal force, but this is more than compensated by marked increase in wing and rear body
normal force due to the vortices.

5
American Institute of Aeronautics and Astronautics
Figure 8. Normal force and pitching moment coefficients with fixed wake model.

Figure 9. Top-side surface pressure coefficient (difference on wings) at α=10º, φ=0º, M=1.7 with fixed
wake model.

B.Double-canard missile nose

Supersonic wind tunnel measurements have been
previously carried out at Helsinki University of
Technology using a missile nose model with two
different nose shapes and several wing
configurations5. One of these configurations, namely
a sharp-nosed double-canard configuration shown in
Figure 10, is chosen for comparison with panel
method results. The forward wings are fixed and the
rearward wings are deflected for pitch control.
Computations were made at Mach number 1.6 with
various control deflections. At this Mach number the
leading-edges of the wings are supersonic, and
separated vortices are therefore not included in the
wake model. Again a half-grid is used due to
symmetry, the number of surface panels being 524.
The total of 21 cases is completed in just over one Figure 10: Missile nose model panelling.
minute of computer time.
Computational results are shown in Figure 11, and corresponding experimental results in Figure 12. Agreement
of the results is fair at small control deflections, but the effect of large control deflection is underestimated at low
angle-of-attack and, although the relative differences remain small as seen below, becomes qualitatively incorrect at
large angle-of-attack.

6
American Institute of Aeronautics and Astronautics
Figure 11. Computational normal force and pitching moment coefficients coefficients at various control
deflections.

Figure 12. Experimental normal force and pitching moment coefficients at various control deflections.5

Quantitative assessment of the above results is presented in Figure 13, which shows the relative difference of
experimental and computational normal forces and difference of centre-of-pressure locations. Differences in zero-
control-deflection normal force coefficients are similar in magnitude to the previous case, and larger local
differences are evident with increasing control deflections. Good agreement of centre-of-pressure locations is in part
due to the short overall length of the model, which results in short moment arms of the components and small
pitching moment coefficients.
0.15 0.10
0.1 0
0.05 10 0.05
20
ΔCN /CN 0
Δxcp/D 0.00
-0.05
0
-0.1 -0.05 10
-0.15 20
-0.2 -0.10
0 5 10 15 20 25 30 0 5 10 15 20 25 30
α α

Figure 13. Relative difference of experimental and computational normal forces and centre-of-pressure
locations at various control deflections.

7
American Institute of Aeronautics and Astronautics
C.Missile with delta-wing and tail
In Reference 2, a delta-wing missile model was used for comparisons between the HISSS panel code and
experimental results at Mach numbers 0.8 and 2.0. Incidentally, the HISSS program shares the theoretical
background and methodology of PANAIR. Full configuration test cases are repeated here using PANAIR and a
more comprehensive wake model as described in section II.D. Surface geometry and panelling of this model is
depicted in Figure 14. Sweep angle of the delta-wing is 70º and thus the leading-edge remains subsonic also in the
supersonic case. Wake networks, whose geometry at α=15º is shown in Figure 15, are therefore attached to leading-
edges to represent separated vortices. The half-grid used in the computations consists of 930 surface panels, and the
solution of the 16 cases requires about 2.5 minutes of computer time.
Comparison of the results shown in Figure 16 with Reference 2 indicates that the effect of angle-of-attack on
wake geometry and inclusion of prescribed leading-edge vortices, even if their exact shape is not realistically
modelled, brings forth the non-linear character of aerodynamic coefficients. The result is a remarkably better
approximation of the experimental data than is attainable with simpler fixed wake geometry.
Despite the improvement due to vortex modelling, a systematic under-prediction of normal force coefficients is
evident in Figure 17. Differences in centre-of-pressure locations appear reasonable, the maximum being about 3% of
the missile length.

Figure 14. Surface panelling of missile model with delta-wing and tail.

Figure 15. Detail of wake networks of delta-wing missile.

8
American Institute of Aeronautics and Astronautics
Figure 16. Normal force coefficients and centre-of-pressure locations at various Mach numbers.

0 0.5
0.8
0.4
2
-0.05
0.3
ΔCN /CN Δxcp/D
0.2
-0.1
0.1 0.8
2
-0.15 0
3 4 5 6 7 8 9 10 11 12 13 3 4 5 6 7 8 9 10 11 12 13
α α

Figure 17. Relative difference of experimental and computational normal forces and centre-of-pressure
locations at various Mach numbers.

IV. Conclusion
A panel method was applied to compute longitudinal aerodynamic characteristics of various missile
configurations. Comparison of the computational results with experimental results shows that aerodynamic
coefficients can be predicted within preliminary design accuracy up to moderate angle-of-attack and small control
deflection. The highest angle-of-attack computed herein was 24º and the results were satisfactory up to 10º positive
control deflection. The panel method is therefore considered a useful tool in aerodynamic modelling. It is possible to
compose a comprehensive simulation model of a missile using mostly panel method results, provided that exact
behaviour at transonic speeds or high angles-of-attack is not significant. The only piece of data required from other
sources is the zero angle-of-attack drag curve, which can be combined with the panel method results to create a
comprehensive axial force representation as a function of Mach number, angle-of-attack, and control deflection.
Successful computation of aerodynamic coefficients at moderate angles-of-attack requires some a-priori
knowledge of the essential flow phenomena in the case at hand. Separated vortices may have major effect on the
aerodynamics and therefore their existence and approximate geometry must be known, and wake-type networks
must be present in the computational grid in order to model the effects of these vortices. The present results suggest
that the exact shape of the separated vortices is not critical as far as the total aerodynamic coefficients are concerned,
although prediction improvement might be realisable with more accurate vortex modelling especially in the delta-
wing case.
The panel method is non-iterative when a prescribed wake model is used, and thus free from instability and
convergence issues inherent in iterative numerical schemes. Therefore, coupled with automatic grid generation it is
suitable for batch-runs which robustly provide comprehensive aerodynamic data sets without intervention from the
analyst. Of course, review of the results is always necessary to detect any anomalous solutions that may arise, for
example, due to locally invalid panel geometry.
Future applications of the panel method will include more general, non-cruciform configurations and asymmetric
9
American Institute of Aeronautics and Astronautics
flight conditions, where examination of the lateral-directional aerodynamic characteristics is of interest. Also,
investigation of dynamic stability derivatives is possible using the panel method.

Acknowledgements
The present study was conducted as a part of a research project funded by the Finnish Air Force. The author
wishes to thank the FiAF for continuing support of research activities at Helsinki University of Technology.

References
1
Erickson, L. L., "Panel Methods - An Introduction," NASA, TP 2995, 1990.
2
Fornasier, L., d' Espiney, P., "Prediction of Missile Stability Using the HISSS Panel Code," Recherche Aérospatiale 1989-4,
1989.
3
Saaris, G. R., "A502 User's Manual - PAN AIR Technology Program for Solving Problems of Potential Flow about Arbitrary
Configurations," Boeing, 1992.
4
Allen, J. M., "Aerodynamics of an Axisymmetric Missile Concept Having Cruciform Strakes and In-Line Tail Fins From Mach
0.6 to 4.63," NASA, TM-2005-213541, 2005.
5
Siiropää, V., "Ohjusten siipien aerodynamiikka suurilla kohtauskulmilla," Helsinki University of Technology, Laboratory of
Aerodynamics, Report T-179 (Unpublished, In Finnish), 2002.

10
American Institute of Aeronautics and Astronautics