AIAA 2017-3398

AIAA AVIATION Forum

5-9 June 2017, Denver, Colorado
35th AIAA Applied Aerodynamics Conference

Quasi-Steady Simulations for the Efficient Generation of

Static Aerodynamic Coefficients at Subsonic Velocities

Sidra I. Silton 1
U.S. Army Research Laboratory, Aberdeen Proving Ground, MD, 21005

A previously developed quasi-steady sweep procedure that allows determination of the

static aerodynamic coefficients for a range of angles of attack in a single simulation is extended
to subsonic Mach numbers. Given a steady-state solution, the procedure can be used to
generate multiple numerical simulations over a select range of angles of attack for a given set
of flight conditions using the time-accurate Reynolds-Averaged Navier–Stokes equations.
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

Computational Fluid Dynamics simulations are completed for a canard-controlled, fin-

stabilized projectile at Mach 0.65 for a moderate range of angles of attack using a quasi-steady
sweep procedure. Separate steady-state solutions are also computed at various angles of attack
for comparison. Comparison of the results showed that the quasi-steady sweep procedure can
provide accurate static aerodynamic coefficients efficiently if convergence at each time step is
ensured.

Nomenclature
𝛼𝛼 = angle-of-attack
𝛿𝛿 = canard deflection angle
ε = turbulent dissipation rate
BF = body-fin configuration
BFC0 = body-fin-canard configuration, undeflected canards
BFC4 = body-fin-canard configuration, 𝛿𝛿 = 4°
𝐶𝐶𝑙𝑙𝑜𝑜 = roll torque coefficient
𝐶𝐶𝐷𝐷 = drag coefficient
𝐶𝐶𝐿𝐿 = lift coefficient
𝐶𝐶𝑚𝑚 = pitching moment coefficient
𝑑𝑑𝑑𝑑 = angle-of-attack increment
k = turbulent kinetic energy
N = number of inner iterations
𝑅𝑅𝑡𝑡 = undamped eddy viscosity
𝑦𝑦 + = non-dimensional wall spacing

Superscripts
C = canard component
F = fin component

Introduction
An accurate understanding of fundamental flow physics and accurate prediction of aerodynamic coefficients is
critical to the development of precision munitions.1-3 The flow fields associated with precision munitions can be very
complex, involving shock-boundary layer interaction, unsteady wakes, and body, canard and tail-fin interactions.4-11
Simple empirical and analytical techniques such as PRODAS,12 Missile Datcom,13 MISL3,14 and AP0915 are not
always able to accurately determine the complex 3-D flow interactions and associated nonlinear flow physics.
Computational Fluid Dynamics (CFD) offers an alternative to compute these nonlinear interacting flow fields,
providing detailed understanding of the associated nonlinear aerodynamic processes and predicting the associated
aerodynamic coefficients that are required to determine if performance requirements can be met. CFD has continued

1
Aerospace Engineer, Weapons and Materials Research Directorate, RDRL-WML-E, AIAA Associate Fellow

This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.
 to emerge as a critical technology for the aerodynamic design and assessment of flight vehicles. Bringing CFD into
the earlier stages of design and development is necessary for it to be more effective; to bring CFD into the earlier
stages, though, the results must be obtained as quickly and accurately as possible. However, even with the
advancement of high-performance computing, CFD computing time is still much greater than the time required to
obtain solutions using engineering-level codes. This means that CFD is not implemented until further along in the
process, often after a problem is discovered.
Steady-state and/or time-accurate CFD methods can be used to obtain a complete set of aerodynamic coefficients
(static and dynamic derivatives) in all flight regimes, from subsonic to supersonic speeds. In general, most of the
aerodynamic coefficients of a projectile or missile in the supersonic regime can be generated efficiently using steady-
state methods; only pitch damping may need to be calculated in a time-accurate manner. Time-accurate methods are
sometimes required to accurately predict base flows and dynamic derivatives of these munitions, especially in the
subsonic and transonic flight regimes.16-27 The static aerodynamic coefficients can be obtained from a steady-state
method. A number of researchers22,26,28 found the pitch-damping coefficient derivative could also be determined using
a steady-state coning methodology. More recent advances in computational modeling have also led to coupling of
CFD codes to rigid-body dynamics (RBD) codes for the time-accurate simulation of free-flight motion, which allows
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

determination of both the static aerodynamic coefficients and dynamic derivatives using the same numerical
simulation.29-38
When designing a new precision munition, especially in the early stages, it is the static aerodynamic coefficients
that are of greatest importance to ensure the munition can meet its range and maneuverability requirements. This
requires the development of an aerodynamic database that likely includes a range of angles of attack and either roll
angles or side-slip angles at a given (or multiple) Mach number(s). While the new coupled CFD–RBD method would
require fewer total simulations, it is quite time intensive as it must be solved using the time-dependent Reynolds-
Averaged Navier–Stokes (RANS) equations with a time step small enough to accurately capture the flow physics. The
traditional steady-state solution uses the same governing RANS equations, but without the time dependence, such that
the flow physics are captured in the mean. However, a large number of these steady-state solutions would be required
to generate the aerodynamic database when nonlinear aerodynamic behavior is present (canard stall, vortex
interactions, etc.) in the range of angle of attack of interest causing this method to be time intensive as well.
Recently, Sahu and Heavey39 demonstrated an alternative method for rapid determination of static aerodynamic
coefficients using the time-accurate RANS equations; a quasi-steady sweep procedure that allows determination of
the coefficients for an entire range of angles of attack. Although their research was on a complex non-axisymmetric
missile configuration, it was limited to the transonic and supersonic flight regimes. The work presented in this paper
extends their work to the subsonic regime. Numerical computations using the quasi-steady and time-accurate sweep
procedures are performed at Mach 0.65 on the high-maneuverability airframe (HMA) being developed at the US Army
Research Laboratory (ARL) as a demonstrator platform for the precision-munitions research area.1,10,40,41 Steady-state
simulations were also performed for comparison with the sweep results. Computational resource requirements are
included to show the efficiency of the quasi-steady sweep procedures.

Computational Approach
The commercially available code CFD++ v12.1.1,42 by Metacomp Technologies, Inc., is used for the CFD
simulations. CFD++ is a finite-volume, unstructured solver capable of computing CFD solutions for a wide range of
aerospace applications. In this study, CFD++ is used in both its steady-state and time-accurate modes to numerically
solve the 3-D, compressible RANS equations to compute the flow solution.
Initially, the Goldberg 3-equation k-ε-Rt turbulence model44 was implemented in this study based on the findings
by Silton and Fresconi.10 This 3-equation model solves the transport equations for undamped eddy viscosity, Rt, in
addition to turbulent kinetic energy (k) and its dissipation rate (ε). This solution methodology accounts for
nonequilibrium conditions and avoids free-stream-turbulence decay under shear-free flow conditions. The 2-equation
realizable k-ε , SST and one-equation SA turbulence models were also used during the course of the study.
Initialization of the turbulence transport for both models was completed by setting the turbulence intensity to 2% and
the turbulent-to-molecular viscosity ratio to 50, as the length scale was not known.

A. Steady-State Simulation
The flow solution was advanced toward steady-state convergence using a point-implicit time-integration scheme
with local time-stepping, defined by the Courant–Friedrichs–Lewy (CFL) number. A linear ramping schedule was
used to gradually increase the CFL number over the first 100 iterations after which the CFL remained constant until
convergence was reached. As only the subsonic flight regime (0.3 < M < 0.8) is considered in this paper, the suggested

2
 CFL ramping is from 1 to 100. However, depending on the angle of attack being investigated, the maximum CFL
number for the ramping was limited to 25 or 50. The multigrid W-cycle method with a maximum of 4 cycles and 20
coarse grid levels was used to accelerate convergence. Implicit temporal smoothing was applied for increased stability,
which is especially useful where strong transients arise. The spatial discretization function was a second-order, upwind
scheme using a Harten–Lax–van Leer-Contact (commonly, HLLC) Riemann solver and Metacomp’s multi-
dimensional Total-Variation-Diminishing (commonly, TVD) flux limiter.42 Double-precision format was used for all
computations.
Solution convergence (typically within 3,000 iterations) was determined by 1) a several-orders-of-magnitude
reduction of the cell-averaged residuals of the RANS equations until the residuals plateaued and 2) the total, tail-fin,
and canard forces and moments reaching asymptote and unchanging to within 0.1%.

B. Quasi-Steady Sweep Simulation

Using the equivalent of a series of steady-state simulations, the flow solution for the quasi-steady sweep is
generated using the CFD solver’s time-accurate mode to obtain the static aerodynamic coefficients across a range of
angles of attack in a single numerical simulation. The time-accurate mode utilizes the implicit solver with dual time-
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

stepping. Dual time-stepping employees 2 time-steps; an “outer” or global time-step that corresponds to the time
discretization of the physical time variation and an “inner” or local time variation that helps to satisfy the physical
transient equations.
The “outer” time-step is chosen to provide the desired time accuracy and is applied to every cell. Typically, this is
set to a small value to ensure that the flow features are sufficiently resolved in time. In the present, sweep simulation,
the time-accurate, dual time-step method is used to determine a “quasi-steady” solution at each time-step, which
corresponds to an increment of rotation, in an efficient manner. Therefore, a large global time-step of 1.0 s is employed.
This large time-step drives the rate of change of the flow variables with respect to angle of attack to zero as quickly
as possible. Thus, the solutions are not transient, as is typical of a time-accurate simulation, but rather multiple static
solutions.
The “inner” time-step is determined internal to the solver via the CFL number that is equal to the final maximum
CFL number from the steady-state solution. The inner time-step (or iteration) is a local, nonphysical time-step used to
converge the time-accurate RANS equations at each physical time-step. The time-step for the inner iteration is allowed
to vary spatially. For the inner iteration, the multigrid W-cycle method and implicit temporal smoothing are applied
to accelerate convergence of the inner iterations and therefore the outer global time-step. If the inner iterations are not
sufficiently converged, the solution is not converged and there is a loss in accuracy. The number of inner iterations is
chosen such that approximately a 2-order-of-magnitude reduction in the cell-averaged “inner” residuals was achieved.
Typically, 100 inner iterations are found to be adequate in the present simulations using the quasi-steady sweep
procedure. These requirements are significantly larger than for a “standard” time-accurate solution where only 5–20
inner iterations are required for a 1–2-order-of-magnitude drop in the cell-averaged inner residuals and a converged
solution.
The steady-state solution with the mesh rotated to the lowest desired angle of attack serves as the initial condition
for the initial sweep solution. For each step of the sweep solution, the computational grid is rotated by a small
increment and a new solution is computed. In CFD++, the rotation takes place at the end of the time-step; the first
time-step is run at the starting angle of attack. This process continues across the desired range of angles of attack.
Figure 1 shows the location of the body at various stages of the sweep. In this example, the rotation point is at the
projectile’s center of gravity. The rotation angle must be chosen small enough to capture any nonlinear aerodynamic
effects that may be present. In the present sweep simulations, a rotation-angle increment of 0.25° was able to suitably
capture the nonlinear aerodynamic effects.

3
 Fig. 1 Location of the projectile at various stages of the sweep procedure
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

Geometry

A. Model Geometry
The geometry utilized for this study was the HMA designed as the platform to demonstrate advanced guided
munitions using low-cost technologies. Both the body–fin configuration and the body–fin–canard configuration were
considered (Fig. 2).
The geometric model was created in SolidWorks44 and has a body reference diameter of 83 mm (1 caliber). This
was a simplified geometry with the gaps and slots in the body needed to accommodate the tail fins suppressed. The
body geometry consisted of a hemispherical nose cap; a nearly cylindrical body 320.38 mm (3.86 cal.) long; and a
66.4-mm (0.8-cal)-long, 7° boattail. The overall projectile length was 427.23 mm (5.15 cal.). Ten flat-plate fins having
a span of 217.22 mm and an average chord of 20.5 mm, with the trailing edge located 4.48 mm from the base of the
projectile, were included in order to maintain static stability. The tail fins were shaped such that they can store flush
in the body for gun launch while maximizing area for stability. A 7-mm-high, 3.8-mm-deep tab at the tip of each tail
fin was bent at a 15° angle such that the projectile would roll clockwise (looking from rear of airframe).
For the body–fin–canard configuration, 4 NACA0015 airfoils, rotated 45° off plane, were added to the model.
Each of the 4 canards had a chord of 18.86 mm and a semispan of 96.36 mm with the quarter chord located 144.7 mm
aft of the nose of the projectile. Canard deflection angles of 0° and 4° pitch up about the quarter chord were considered.
For simplicity, the model had the canards directly attached to the body; the body slots were suppressed.

Fig. 2 Subsonic geometries: body–fin configuration (top) and body–fin–canard configuration (bottom)

4
 B. Numerical Grids
The grids used in the numerical simulation were created with MIME v4.145 by Metacomp Technologies. MIME is
an unstructured mesh generator that allows triangular or quadrilateral dominant cells for the surface mesh. Once an
adequate surface mesh is generated, prism layers can be specified and created when the volume mesh (tetrahedral
dominant) is generated.
MIME was used to generate the volume mesh for the quasi-steady simulations without canards and with the
canards directly attached to the body. Figure 3 shows the extent of the outer boundary. It extended approximately 35
body lengths from the projectile in all directions. Cylindrical density boxes were placed in the wake of the projectile
and tail fins to ensure proper resolution in this area. An additional density box (density Box 3) was placed between
the canards and the tail fins, when the canards were present, to ensure the canard-tip vortices would not dissipate
significantly due to inadequate mesh resolution prior to reaching the tail fins. The surface mesh and boundary-layer
growth was restricted to a ratio of 1.2 or less. The growth ratio was relaxed to 2.0 when creating the remainder of the
volume mesh. The prism-layer spacing was chosen such that wall-function spacing (nondimensional wall distance,
30≤y+≤60) was present on the body, while a solve-to-wall mesh y+≈1 was used on and around the tail fins and canards.
To achieve this, a first cell spacing for the prism layer was specified at 3.5×10-3 mm on the tail fins and canards, 0.25
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

mm on the body, and 0.2 mm on the nose. This produced y+≈1 on the tail fins and canards as well as on the body
immediately surrounding them. On the nose of the airframe and remainder of the body, the y+ varied between 25 and
45. MIME allowed for a smooth transition between the specified spacings. Mesh-refinement studies, which showed
that this mesh achieved grid independence, have been conducted for this body–fin geometry10 and are not included in
this paper. Mesh-refinement studies for the body–fin–canard geometry have not yet been completed. The
computational domain is fully 3-D with no symmetry. The body–fin mesh consisted of approximately 26 million cells,
while the body–fin–canard mesh consisted of approximately 45 million cells.

C. Boundary Conditions
The entire far-field boundary was set as “characteristics-based” inflow/outflow. This boundary condition takes the
specified free-stream conditions and solves a Riemann problem at the boundary using the supplied data as a virtual
state outside the domain. The walls of the projectile were specified as an adiabatic, no-slip, viscous boundary. Wall
functions were used to compute the boundary layer on the projectile’s surface except where there was adequate prism-
layer resolution to solve to the wall, as was the case on the tail fins and canards.

Fig. 3 Extent of outer mesh boundary with close-up of projectile showing locations of density boxes

5
 Results
Numerical simulations were performed to assess the ability and efficiency of the quasi-steady sweep procedure to
predict the flow field and aerodynamic coefficients for a complex gun-launched-projectile configuration in the
subsonic flight regime. All computations were completed using a free-stream temperature and pressure of 288.15 K
and 101325 Pa, respectively. The study was conducted at Mach 0.65 (220 m/s). Angles of attack between –14° and
14°, inclusive, were considered. The flow domain was initialized using free-stream conditions everywhere.

A. Body–Fin Configuration
Steady-state simulations were completed for the body–fin configuration (BF) at α = –14°, 0°, 2°, 3°, 5°, 8°, 10°,
and 14°. The quasi-steady sweep procedure was performed for BF using the k-ε-Rt turbulence model. The quasi-
steady simulation uses the steady-state solution at α=-14° as the initial condition. The critical parameters were varied
to determine the requirements for an accurate solution. The critical parameters varied were angle-of-attack increments
(dα) = 0.5°, 0.25°, and 0.1°; and number of inner iterations (N) = 50, 100, and 200.
Initially, the CFL was set to 10 (a small CFL is typically warranted when determining time-accurate solutions)
with 50 inner iterations,39 and 𝑑𝑑𝑑𝑑 = 0.25°. This combination of parameters produced approximately 1-order-of-
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

magnitude drop in residual (Fig. 4). The residuals did not level off as would be expected of a completely converged
solution.

Fig. 4 Time history of solution residual as a function of number of inner iterations for CFL = 10, dα = 0.25°;
N = 50.

Comparison of the resulting lift and drag coefficients (𝐶𝐶𝐿𝐿 and 𝐶𝐶𝐷𝐷 , respectively), as well as the pitching moment
coefficient (𝐶𝐶𝑚𝑚 ), obtained from the quasi-steady sweep procedure to those obtained from the steady-state methodology
were reasonable. For the roll torque coefficient, 𝐶𝐶𝑙𝑙𝑜𝑜 , the agreement was not as good as the author would have liked it
to be. (Fig. 5). Because the scale is quite enlarged, the most concerning issue is that the quasi-steady sweep results
are not symmetric about α=0° as is expected for this symmetric configuration. Therefore, the convergence of the inner
iterations for dα=0.25° was investigated.
As expected, increasing the number of inner iterations allowed further convergence in the solution residuals for
the inner iterations. Nearly 2 orders of magnitude drop are achieved for 200 inner iterations, although they still do not
level off as was desired. With the improved convergence at each time step, a change in 𝐶𝐶𝑙𝑙𝑜𝑜 over the first 40 steps of
the sweep procedure can be seen (Fig. 6). Although the actual values do not change significantly (the scale is
expanded), the shift to the left indicates a better solution can be achieved with an increase in the number of inner
iterations. Even with 200 inner iterations, convergence has not been achieved. Thus, the quasi-steady sweep
procedure was not going to be as efficient as it was for a supersonic configuration.

6
 Fig. 6 Roll-torque coefficient for BF for varying
number of inner iterations, CFL = 10 and dα =
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

Fig. 5 Roll-torque coefficient for BF comparing

0.25°
for dα = 0.25°, CFL = 10 and N = 50

It was hypothesized that the quasi-steady sweep procedure could be completed in a more efficient manner if a
larger CFL number – similar to that used for the steady-state simulations – was chosen, rather than using a small CFL
as would be the case if a truly time-accurate solution was being sought. Use of a larger CFL in a steady-state
simulation typically allow convergence in fewer iterations. Since a quasi-steady sweep is solving for a “steady-state”
solution at each time-step, it was thought that convergence at each time-step could be achieved with fewer inner
iterations if a larger value for the CFL number was chosen. As the CFL number for some of the steady-state simulations
at higher angles of attack had to be limited to 50 in order for convergence to be achieved for the BF (if a CFL greater
than 50 was chosen, convergence was not possible), it was desirable to know the improved efficiency of sweep
procedure and the required parameters if the CFL was set to 50 for the quasi-steady sweeps.
The CFL number was set to 50 rather than the value of 100 suggested by CFD++ as some of the steady-state
simulations at higher angles of attack had to be limited to 50 in order for convergence to be achieved for the BF (if a
CFL greater than 50 was chosen, convergence of some of the steady-state solutions was not possible as residual drops
of only 2 – 3 orders of magnitude could be obtained). Again, fifty inner iterations were chosen as a minimum based
on the findings of Sahu and Heavey;39 50 inner iterations was the maximum number that could be chosen in CFD++ at
the time that study was completed. The findings for each 𝑑𝑑𝑑𝑑 were the same; there were small differences in force and
moment coefficients when the results of N = 50 and N = 100, but virtually no differences when the results of N = 100
and N = 200 were compared. The convergence history of the inner iterations for the 𝑑𝑑𝑑𝑑 = 0.25° case (Fig. 7) shows
why this is the case. There is a decrease in the residuals during the inner iterations from approximately 2E-2 to 1E-2
when N = 100 is used rather than N = 50. When N is further increased to 200, the inner-iteration residuals plateau at
1E-2 indicating that the solution has converged. Thus, N = 100 was chosen for all 𝑑𝑑𝑑𝑑 investigated.
With inner iteration convergence achieved, the results of the quasi-steady sweep for 𝑑𝑑𝑑𝑑 = 0.5°, 𝑑𝑑𝑑𝑑 = 0.25° and
𝑑𝑑𝑑𝑑 = 0.1° were compared. Most of the forces and moments for the total configuration, as well as the body components
and fin components (CF ) were identical and in good agreement with the steady-state solutions. Figure 8 shows the fin
contribution to the forces and moments. Only 𝐶𝐶𝑙𝑙𝑜𝑜 shows any difference, i.e., for 𝑑𝑑𝑑𝑑 = 0.5°, however, it is negligible
(less than 1%). The total value of roll torque shown in Fig. 8 is equivalent to the fin contribution as it is the only
contribution. Thus, agreement in the total forces and moments predicted using the sweep procedure and the steady-
state solutions is achieved for all of the increments investigated, indicating that for the BF at subsonic velocity an
increment as large as dα=0.5° is acceptable. This does, of course, assume a sufficient number of inner iterations is
utilized. It becomes the researcher’s choice of the desired increment, based on efficiency and desired resolution.
For the BF configuration, completing the quasi-steady sweep allows a range of angles of attack to be investigated
in significantly less time than if individual steady-state simulations were used (Table 1). When completing the quasi-
steady simulations, it is critical to ensure the inner iterations are converged. The study conducted by Sahu and Heavy39
found only 25–50 inner iterations were required to achieve time-step convergence in the transonic and supersonic
regimes. In the subsonic regime investigated here, 100 inner iterations were required for convergence regardless of
angular resolution.

7
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

Fig. 7 Time history of solution residual as a function of number of inner iterations for CFL = 50, dα = 0.25°;
N = 50 (top), N = 100 (middle), and N = 200 (bottom)

8
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

Fig. 8 Fin contribution to aerodynamic coefficients for BF with CFL = 50, N = 100, dα comparison

Table 1 Comparison of time required to obtain converged solution for –14° ≤ α ≤ 14° at dα indicated

Simulation dα (°) N CPU hours

Steady-state 1.00 ... 26,000
0.10 100 13,286
Quasi-steady 0.25 100 5,555
0.50 100 2,995

B. Body–Fin–Canard
The body–fin–canard configuration is investigated for canards in the x-orientation only (i.e., 45° rotation out of
the lift plane). The canards are considered in their undeflected position (𝛿𝛿 = 0°, BFC0) as well as for a 4° deflection
about its quarter chord for a pitch up maneuver (𝛿𝛿 = 4°, BFC4). Two deflections were considered as it was unknown
if variations in canard separation, canard stall, canard-root and/or tip vortex shedding, and vortex–fin interactions due
to canard deflection would require different values of the critical parameters of the sweep procedure to obtain
agreement with the steady-state solutions.
As was done for the BF configurations, steady-state simulations were completed first. The steady-state simulations
were completed at 𝛼𝛼 = 0°, ±2°, ±3°, ±5°, ±8°, ±10°, and ± 14°. The sweep procedure was performed using the
steady-state solution at 𝛼𝛼 = −14° as the initial condition. The critical parameters were varied to determine the
requirements for an accurate solution. The critical parameters varied were 𝑑𝑑𝑑𝑑 = 0.25° and 0.1°; and N = 100 and 200.
The CFL number was set to 50.
As was the case for quasi-steady sweeps on the BF configuration, N = 100 was found to produce a converged
solution for both 𝑑𝑑𝑑𝑑 investigated (0.25°, 0.1°). The inner-iteration convergence histories for BFC0 and BFC4 (Fig. 9)
were similar to those found for the BF (Fig. 7).

9
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

Fig. 9 Inner-iteration convergence history for BFC4 with CFL = 50, N = 100; dα = 0.25° (top) and dα = 0.1°
(bottom)

For the BFC0, very good agreement was found between the static results and the quasi-steady sweep results for
both 𝐶𝐶𝑚𝑚 and 𝐶𝐶𝑙𝑙𝑜𝑜 (Fig. 10). Both the prediction of the body and fin contributions to 𝐶𝐶𝑚𝑚 by the sweep procedure and
canard stall was properly captured (when compared to the steady-state solutions) even with 𝑑𝑑𝑑𝑑 = 0.25° (Fig. 11).
Although small differences (expanded scale) exist in 𝐶𝐶𝑙𝑙𝑜𝑜 , 𝑑𝑑𝑑𝑑 = 0.25° was determined to be adequate; additional 𝑑𝑑𝑑𝑑
were not investigated.

10
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

Fig. 9 Pitching-moment and roll torque coefficient for BFC0 with CFL = 50, N = 100, dα comparison

Fig. 10 Single canard contribution to pitching-moment coefficient for BFC0 with CFL = 50, N = 100, dα
comparison

For the BFC4, all force and moment coefficients (Fig. 11) show very good agreement with the solutions from the
steady-state methodology. Even the nonlinearity in the moment coefficients is accurately captured. Figure 12 presents
𝐶𝐶
𝐶𝐶𝑚𝑚 as a function of α for each canard. Part of the analysis of this curve is to isolate the aerodynamic behavior of the
canard, which includes at what α each canard is stalling. If the individual steady-state solutions are used for this, two
fits are determined; one prior to stall and one after stall. To approximate the α at which stall occurs, these two fits are
set equal to each other; the error is only as good as the fits. To determine canard stall angle for the quasi-steady sweep
solution, no interpolation and the answer is known to within the 𝑑𝑑𝑑𝑑.
For 𝑑𝑑𝑑𝑑 = 0.25°, 36.5 wall-clock hours on 384 cores of the SGI ICE X cluster (approximately 14,000 CPU hours)
were required. For 𝑑𝑑𝑑𝑑 = 0.1°, 86.3 wall-clock hours on 384 cores of the SGI ICE X cluster (approximately 33,000
CPU hours) were necessary. Compare this to a required wall-clock time of 3.8 h (1,460 CPU hours) for one steady-
state solution. The break-even point is 10 steady-state solutions. If more than 10 steady-state solutions are required
to adequately resolve the static coefficients, which is likely the case of a highly nonlinear flow such as this, the sweep
procedure is definitely the more efficient method. The increment is dependent on the desired resolution as was the
case with the BFC0. The physics is being adequately captured.

11
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

Fig. 11 Force and moment coefficient for BFC4 with CFL = 50, N = 100, dα comparison

Fig. 12. Individual canards’ pitching-moment contributions with dα = 0.25°, CFL = 50, N = 100

C. Further Comments
The above results for the BF, BFC0 and BFC4 were all obtained using the 3 equation k-ε-Rt turbulence model.
Although significant time was spent determining the required parameters through the determination of the
aerodynamic coefficients using both steady-state and quasi-steady sweep procedures, the quasi-steady sweep
procedure did provide a more efficient methodology. As mentioned in the beginning of the section “Computational

12
 Approach,” the 3 equation k-ε-Rt turbulence model was chosen based on the results of a turbulence model study
completed prior to adding canards to the configuration. As wind tunnel results became available during the course of
the present study, a turbulence model study was conducted.
It was determined that the newly validated quasi-steady sweep procedure would be implemented to rapidly obtain
numerical data for multiple turbulence models to compare to the wind tunnel data. The parameters of CFL=50, N=100,
𝑑𝑑𝑑𝑑 = 0.25° were initially utilized for each turbulence model for the BFC4 configuration. The resulting aerodynamic
moment coefficients (𝐶𝐶𝑚𝑚 and 𝐶𝐶𝑙𝑙𝑜𝑜 ) can be seen in Fig. 13. What immediately is apparent is that the other turbulence
models investigated via quasi-steady sweeps (SST and SA) have a high frequency oscillation in 𝐶𝐶𝑙𝑙𝑜𝑜 at moderate α,
which is not apparent in the k-ε-Rt solution. There are also oscillations present in Cm, although they are not as
apparent. These high frequency oscillations were not expected with a quasi-steady solution, so additional steady-state
solutions were performed to see if the results of the different turbulence models agreed; they did. Significant effort
was taken to look at the effect of decreasing 𝑑𝑑𝑑𝑑 and increasing the number of inner iterations. After decreasing to
𝑑𝑑𝑑𝑑 = 0.1° and increasing up to 400 inner iterations to ensure convergence as well as to try and alleviate the oscillations
(Fig. 14), it was determined that the quasi-steady sweep procedure may not be as efficient in the subsonic and transonic
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

regimes. The convergence criteria is obviously not universal and may have to be evaluated every time a piece of the
study varied. Another option, which is currently being evaluated is the time-accurate sweep. It is hypothesized that
the sweep rate and time step required to achieve time converge results would be based on non-dimensional flow
parameters that would be applicable to all geometries in the subsonic flow regime. Additionally, any transients that
appear in the flow would be inherently resolved.

Fig. 13 Total moment coefficient for BFC4 for multiple turbulence models, CFL = 50, N = 100, 𝒅𝒅𝒅𝒅 = 𝟎𝟎. 𝟐𝟐𝟐𝟐°

Fig. 14 Total moment coefficient for BFC4 for SST turbulence models, various quasi-steady sweep
parameters.

13
 Conclusions
A quasi-steady sweep procedure that allows determination of the static aerodynamic coefficients for a range of
angles of attack in a single simulation is extended to subsonic Mach numbers. Given a steady-state solution, the
procedure can be used to generate multiple quasi-steady numerical simulations over a select range of angles of attack
for a given set of flight conditions. Computational fluid-dynamics simulations are completed for the ARL-designed
HMA, a canard-controlled, fin-stabilized projectile, at Mach 0.65 for a moderate range of angles of attack (−14° ≤
𝛼𝛼 ≤ 14°) using the sweep procedure. Separate steady-state solutions were also computed at various angles of attack
within this range for comparison.
Inner-iteration convergence is critical to accurate solutions for the quasi-steady sweep procedure. Using a CFL
number suitable for converging the steady-state solution (CFL = 50), 100 inner iterations were found to be necessary
to ensure convergence for the 𝑑𝑑𝑑𝑑 investigated as part of this study. Once simulation convergence was confirmed, the
quasi-steady sweep procedure was found to produce accurate force and moment coefficients for the range of angles
of attack considered in significantly less time. In the time it would take to complete 10 steady-state solutions, a
solution with 112 angles was generated.
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

For a less complex configuration such as the BF, where the forces and moments are known to be linear, the
increased resolution of the sweep procedure is not necessary; greater efficiency could be achieved by just choosing a
small number of angles of attack to simulate with the steady-state method and using standard curve fit procedures to
generate data at the other angles. For a complex configuration that may potentially have nonlinear forces and
moments, this increased resolution is critical. However, for extremely complex configurations or for more complex
flows due to turbulence closure models, quasi-steady sweeps may not be the most efficient process. Other methods,
such as time-accurate sweeps, may need to be considered.

Acknowledgments
This work was supported in part by a grant of high-performance computing time from the U.S. Department of
Defense (DOD) High Performance Computing Modernization program at the Army Research Laboratory DOD
Supercomputing Resource Center (DSRC), Aberdeen Proving Ground, MD and at the Air Force Research Laboratory
DOD DSRC, Wright Patterson Air Force Base, OH. The author would like to thank Dr. James DeSpirito, ARL Flight
Sciences Branch, for multiple conversations relating to the sweep procedure and the CFD simulations. The author
would also like to thank Mr. Prasanth Kachgal, Metacomp Technologies, for multiple discussions to correctly
implement the quasi-steady sweep procedure within CFD++.

References
1Fresconi F, Celmins I, Fairfax L. Optimal parameters for maneuverability of affordable precision munitions. Aberdeen Proving
Ground (MD): Army Research Laboratory (US); 2011 Aug. Report No.: ARL-TR-5647.
2Coyle CJ, Silton SI. CFD aerodynamic characterization of a high maneuverability airframe. Paper presented at: 33rd AIAA

Applied Aerodynamics Conference; 2015 Jun 22–26; Dallas (TX). Paper No.: AIAA-2015-3015.
3Silton SI, Fresconi FE. Effect of canard interactions on aerodynamic performance of a f in-stabilized projectile. J Space Roc;

2015;52(5):1430–1442.
4Sahu J. Numerical computations of transonic critical aerodynamic behavior. AIAA Journal. 1990;28(5):807–816.
5DeSpirito J, Vaughn, M, Washington. Numerical investigation of canard-controlled missile with planar grid fins. J Space Roc.

2003;40(3):363–370.
6Silton S. Navier-stokes computations for a spinning projectile from subsonic to supersonic speeds. J Space Roc.

2005;42(2):223–231.
7DeBonis JR, Oberkampf WL, Wolf RT, Orkwis PD, Turner MG, Babinsky H, Benek JA. Assessment of computational fluid

dynamics and experimental data for shock boundary-layer interactions. AIAA Journal. 2012;50(4):891–903.
8Bhagwandin VA. Numerical prediction of shock-boundary layer interaction between a pair of fins in hypersonic flow. Paper

presented at: 44th AIAA Fluid Dynamics Conference; 2014 Jun 16–20; Atlanta (GA). Paper No.: AIAA-2014-3337.
9Scheuermann E, Costello M, Silton S, Sahu, J. Aerodynamic characterization of a microspoiler system for supersonic finned

projectiles. J Space Roc. 2015;52(1):253–263.

10Silton SI, Fresconi FE. High maneuverability airframe: initial investigation of configuration aft end for increased stability,

range, and maneuverability. Aberdeen Proving Ground (MD): Army Research Laboratory (US); 2013 Sep. Report No.: ARL-TR-
6585.
11Silton SI, Coyle CJ. Effect of canard deflection on fin performance of a fin-stabilized projectile. Paper presented at: 33rd

AIAA Applied Aerodynamics Conference; 2015 Jun 22–26; Dallas (TX). Paper No.: 2015-2586.
12Arrow Tech Associates. PRODAS V3 user manual Burlington (VT); 2015.
13Rosema C, Doyle J, Auman L, Underwood M, Blake WB. Missile DATCOM user’s manual–2011 revision. Wright–Patterson

AFB (OH): Air Force Research Laboratory (US); 2011 Mar. Report No.: AFRL-RB-WP-TR-2011-3071.

14
 14Lesieutre DJ, Love JF, Dillenius MFE. Recent applications and improvements to the engineering-level aerodynamic

prediction software MISL3. Paper presented at: 40th AIAA Aerospace Sciences Meeting and Exhibit; 2002 Jan 14–17; Reno (NV).
Paper No.: AIAA-2002-0275.
15Moore FG, Moore LY. 2009 version of the aeroprediction code: AP09. J Space Roc. 2008;45(4):677–690.
16Oktay E, Akay H. CFD predictions of dynamic derivatives for missiles. Paper presented at: 40th AIAA Aerospace Sciences

Meeting and Exhibit; 2002 Jan 14–17; Reno (NV). Paper No.: AIAA 2002-0276.
17Park SH, Kwon JH. Comparisons of steady and unsteady methods for pitch-damping predictions. Paper presented at: 21st

AIAA Applied Aerodynamics Conference; 2003 Jun 23–26; Orlando (FL). Paper No.: AIAA-2003-3671.
18DeSpirito J, Heavey KR. CFD computations of magnus moment and roll-damping moment of a spinning projectile. Paper

presented at: AIAA Atmospheric Flight Mechanics Conference and Exhibit; 2004 Aug 16–19; Providence (RI). Paper No.: AIAA-
2004-4713.
19Murman SM. Reduced-frequency approach for calculating dynamic derivatives. AIAA Journal. 2007;45(6):1161–1168.
20DeSpirito J, Plostins P. CFD prediction of M910 projectile aerodynamics: unsteady wake effect on magnus moment. Paper

presented at: AIAA Atmospheric Flight Mechanics Conference and Exhibit; 2007 Aug 20–23; Hilton Head (SC). Paper No.: AIAA-
2007-6580.
21DeSpirito J. CFD prediction of magnus effect in subsonic to supersonic flight. Paper presented at: 46th AIAA Aerospace
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

Sciences Meeting and Exhibit; 2008 Jan 7–10; Reno (NV). Paper No.: AIAA-2008-427.
22DeSpirito J, Silton SI, Weinacht P. Navier-stokes predictions of dynamic stability derivatives: evaluation of steady-state

methods. J Space Roc. 2009;46(6):1142–1154.

23Silton SI. Navier-stokes predictions of aerodynamic coefficients and dynamic derivatives of a 0.50-cal projectile. Paper

presented at: 29th AIAA Applied Aerodynamics Conference; 2011 Jun 27–30; Honolulu (HI). Paper No.: AIAA-2011-3030.
24Da Ronch A, Vellespin D, Ghoreyshi M, Badcock KJ. Evaluation of dynamic derivatives using computational fluid dynamics.

AIAA Journal. 2012;50(2):470–484.

25Bhagwandin VA. Numerical prediction of roll damping and magnus dynamic derivatives for finned projectiles at angle of

attack. Paper presented at: 30th AIAA Applied Aerodynamics Conference; 2012 Jun 25–28; New Orleans (LA). Paper No.: AIAA-
2012-2905.
26Bhagwandin VA, Sahu J. Numerical prediction of pitch damping stability derivatives for finned projectiles. J Space Roc.

2014;51(5):1603–1618.
27Klatt D, Hruschka R, Leopold F. Numerical and experimental investigation of the magnus effect in supersonic flows. Paper

presented at: 30th AIAA Applied Aerodynamics Conference; 2012 Jun 25–28; New Orleans (LA). Paper No.: AIAA-2012-3230.
28Weinacht P, Sturek WB. Navier-stokes predictions of pitch damping for finned projectiles using steady coning motion. Paper

presented at: Flight Simulation Technologies Conference and Exhibit; 1990 Sep 17–19; Dayton (OH). Paper No.: AIAA-1990-
3088.
29Sahu J. Coupled CFD and rigid body dynamics modeling of a spinning body with flow control. Paper presented at: 2nd AIAA

Flow Control Conference; 2004 Jun 28–Jul 1; Portland (OR). Paper No.: AIAA-2004-2317.
30Sahu J. Time-accurate numerical prediction of free-flight aerodynamics of a finned projectile. J Space Roc. 2008;45(5):946–

954.
31Sahu J. Computations of unsteady aerodynamics of a spinning body at transonic speeds. Paper presented at: 27th AIAA

Applied Aerodynamics Conference; 2009 Jun 22–25; San Antonio (TX). Paper No.: AIAA-2009-3852.
32Sahu J. Unsteady aerodynamic simulations of a canard-controlled projectile at low transonic speeds. Paper presented at:

AIAA Atmospheric Flight Mechanics Conference; 2011 Aug 8–11; Portland (OR). Paper No.: AIAA-2011-6336.
33Sahu J, Fresconi F. Aeromechanics of control of projectile roll using coupled simulation techniques. J Space Roc.

2015;52(3):944–957.
34Costello M, Gatto S, Sahu J. Using CFD/RBD results to generate aerodynamic models for projectile flight simulation. Paper

presented at: AIAA Atmospheric Flight Mechanics Conference and Exhibit; 2007 Aug 20–23; Hilton Head (SC). Paper No.: AIAA-
2007-6582.
35Camargo J, Lopez O, Ochoa-Lleras N. A computational tool for unsteady aerodynamic flow simulations coupled with rigid

body dynamics and control. Paper presented at: 30th AIAA Applied Aerodynamics Conference; 2012 Jun 25–28; New Orleans
(LA). Paper No.: AIAA-2012-3034.
36Montalvo C, Costello M. Estimation of projectile aerodynamic coefficients using coupled CFD/RBD simulation results. Paper

presented at: AIAA Atmospheric Flight Mechanics Conference; 2010 Aug 2–5; Toronto, Ontario, Canada. Paper No.: AIAA-2010-
8249.
37Stahl J, Costello M, Sahu J. Projectile aerodynamic coefficient estimation using integrated CFD/RBD and flight control

system modeling. Paper presented at: AIAA Atmospheric Flight Mechanics Conference; 2009 Aug 10–13; Chicago (IL). Paper
No.: AIAA-2009-5715.
38Wang G, Zheng Z, Suo Q. Trajectory simulation of a spinning projectile based on variable step size CFD/RBD method. Paper

presented at: AIAA Atmospheric Flight Mechanics Conference; 2015 Jan 5–9; Kissimmee (FL). Paper No.: AIAA-2015-0522.
39Sahu J, Heavey KR. Time-accurate computations for rapid generation of missile aerodynamics. Paper presented at:

Atmospheric Flight Mechanics Conference; 2010 Aug 2–5; Toronto, Ontario, Canada. Paper No.: AIAA-2010-8248.
40Fresconi F, Celmins I, Silton SI. Theory, Guidance and flight control for high maneuverability projectiles. Aberdeen Proving

Ground (MD): Army Research Laboratory (US); 2014 Jan. Report No.: ARL-TR-6767.

15
 41Silton SI, Fresconi F, Celmins I. High maneuverability airframe: investigation of fin and canard sizing for optimum

maneuverability. Aberdeen Proving Ground (MD): Army Research Laboratory (US); 2014 Sep. Report No.: ARL-TR-7052.
42Metacomp Technologies, Inc. CFD++ user manual. Agoura Hills (CA); 2011.
43Goldberg U, Peroomian O, Chakravarthy S. A wall-distance-free k-ε model with enhanced near-wall treatment. J Fluids Eng.

1998;120(3):457–462.
44Dassault Systemes SolidWorks Corporation. SolidWorks 2012; Waltham (MA).
45Metacomp Technologies, Inc. MIME user manual. Agoura Hills (CA); 2010.
Downloaded by UNIVERSITY OF ILLINOIS on June 30, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2017-3398

16