JOURNAL OF SPACECRAFT AND ROCKETS

Vol. 41, No. 2, March–April 2004

Projectile Performance, Stability, and Free-Flight Motion

Prediction Using Computational Fluid Dynamics
Paul Weinacht∗
U.S. Army Research Laboratory, Aberdeen Proving Ground, Maryland 21005

With the recent development of capabilities for predicting damping derivatives, it is now possible to predict the
stability characteristics and free-flight motion for projectiles using data that are derived solely from computational
fluid dynamics (CFD). As a demonstration of the capability, results are presented for a family of axisymmetric
projectiles in supersonic flight. The particular configuration selected has been extensively tested in aeroballistic
ranges, and high-quality experimental data have been obtained. Thin-layer Navier–Stokes techniques have been
applied to compute the attached viscous flow over the forebody of the projectile and the separated flow in the
projectile base region. Parameters that characterize the in-flight motion are subsequently evaluated using the
predicted aerodynamic coefficients, including the gyroscopic and dynamic stability factors and the projectile’s fast-
Downloaded by BEIHANG UNIVERSITY on February 23, 2022 | http://arc.aiaa.org | DOI: 10.2514/1.1037

and slow-mode frequencies and damping coefficients. These parameters are then used to predict the free-flight
motion of the projectile. In each case, the computational approach is validated by comparison with experimental
data, and very good agreement between computation and experiment is found. It is believed that this demonstration
represents the first known instance of a viscous CFD approach being applied to predict all of the necessary data
for performance of linear aerodynamics stability and trajectory analyses.

Nomenclature α = angle of attack

a∞ = freestream speed of sound β = yaw angle
CD = drag coefficient γ = cosine of total angle of attack
C D0 = zero-yaw drag coefficient λ1 , λ2 = fast and slow mode damping rates
Cl = net roll moment coefficient ξ, η, ζ = transformed coordinates in N–S equations
Cl p = roll damping moment coefficient ξ̃ = complex angle of attack
Cm α = slope of the pitching moment coefficient ρ = density
with angle of attack φ10 , φ20 = phase angle of fast- and slow-mode arms
Cm q + Cm α̇ = pitch–damping moment coefficient sum φ1 , φ2 = fast- and slow-mode frequencies
C̃m = pitching moment coefficient
C Nα = slope of the normal force coefficient Superscripts
C Nq + C Nα̇ = pitch–damping force coefficient sum
· = rate of change with respect to time
Cn pα = Magnus moment coefficient
 = rate of change with respect to space
C̃n = side moment coefficient
˜ = referenced to nonrolling coordinate frame
CY pα = Magnus force coefficient
D = projectile diameter
Ê, F̂, Ĝ = flux vectors in transformed coordinates
Introduction
e = total energy per unit volume
Ĥ
K 10 , K 20
=
=
source term in Navier–Stokes (N–S) equations
amplitude of fast- and slow-yaw arms P REDICTION of the in-flight motion of projectiles requires the
determination of the aerodynamic forces and moments that act
on the body during the course of its trajectory. These aerodynamic
L = body length
M = freestream Mach number forces and moments may be determined from free-flight aerodynam-
p = spin rate, as used in roll equations ics ranges,1 wind-tunnel experiments,2 or theoretical approaches,
q = vector of dependent variables such as computational fluid dynamics (CFD). In the supersonic-
Re = Reynolds number, a∞ ρ∞ D/µ∞ flight regime, static aerodynamic coefficients such as drag and pitch-
Ŝ = viscous flux vector ing moment have been routinely computed using CFD for a number
s = distance down range of years.3,4 Much less routine are predictions of dynamic aerody-
sd = dynamic stability factor namic coefficients, such as the pitch damping and Magnus coeffi-
sg = gyroscopic stability factor cients. A computational capability for predicting the Magnus forces
u, v, w = velocity components in x, y, z directions and moments for smooth medium- and large-caliber flight bodies
V = freestream velocity has existed since the early 1980s.3,4 With the recent development
x, y, z = axial, horizontal, and vertical coordinates of capabilities for predicting damping derivatives,5 it is now possi-
ble to predict the stability characteristics and free-flight motion for
projectiles using data that are derived solely from CFD.
Whereas previous research work has focused on predictions of the
Received 6 March 2003; revision received 29 May 2003; accepted for individual aerodynamic coefficients, the focus of this paper is on the
publication 30 May 2003. This material is declared a work of the U.S. predictions of aerodynamic performance, in particular, aerodynamic
Government and is not subject to copyright protection in the United States. stability. An attempt is made to demonstrate, through a particular
Copies of this paper may be made for personal or internal use, on condition
that the copier pay the $10.00 per-copy fee to the Copyright Clearance
case study, the suitability and accuracy of Navier–Stokes computa-
Center, Inc., 222 Rosewood Drive, Danvers, MA 01923; include the code tional techniques for the prediction of projectile performance and
0022-4650/04 $10.00 in correspondence with the CCC. stability. Until recently, this was not possible because methods had
∗ Aerospace Engineer, Aerodynamics Branch, Ballistics and Weapons not been implemented to predict all of the necessary coefficients.
Concepts Division, Weapons and Materials Research Directorate. Associate As a demonstration of the capability, results are presented for
Fellow AIAA. a family of axisymmetric projectiles in supersonic flight. The
257
 258 WEINACHT

respect to the freestream velocity vector and undergoing a rotation at

a constant angular velocity about a line parallel to the freestream ve-
locity vector and coincident with the projectile’s c.g. Coning motion
is, in fact, a specific combination of two orthogonal planar pitching
motions plus a spinning motion. From linear flight mechanics the-
ory, it can be shown that the pitch–damping coefficients are related
to the side force and moment due to steady coning motion.5
The flowfield predictions of the projectile in steady coning motion
have been performed in a novel rotating coordinate frame that rotates
at the roll rate or coning rate of the projectile. The fluid flow relative
to the rotating coordinate frame does not vary with time, allowing the
steady (non-time-varying) Navier–Stokes equations to be applied.
To implement the rotating coordinate frame, the governing equations
have been modified to include the effect of centrifugal and Coriolis
Fig. 1 Schematic of Army–Navy spinner rocket (ANSR) configura- forces. The steady, thin-layer Navier–Stokes equations are
tion. ∂ Ê ∂ F̂ ∂ Ĝ 1 ∂ Ŝ
+ + + Ĥ = (1)
particular configuration selected for this computational study has ∂ξ ∂η ∂ζ Re ∂ζ
been extensively tested in aeroballistic ranges and high-quality ex- Here Ê, F̂, and Ĝ are the inviscid flux vectors and Ĥ is the source
perimental data have been obtained.1 The configuration consists of
Downloaded by BEIHANG UNIVERSITY on February 23, 2022 | http://arc.aiaa.org | DOI: 10.2514/1.1037

term containing the Coriolis and centrifugal force terms, which

5-, 7- and 9-caliber length versions. Three different centers of grav- result from the rotating coordinate frame. Each of these matrices
ity were tested for each body length by varying the mass distribution. is a function of the dependent variables represented by the vec-
A schematic of the configuration is shown in Fig. 1. tor q(ρ, ρu, ρv, ρw, e). Further details of the flux terms and the
The computational results were obtained by applying a parabo- source term containing the centrifugal and Coriolis forces are given
lized Navier–Stokes (PNS) approach to compute the attached vis- in Refs. 6 and 5, respectively.
cous flow over the forebody of the projectile. Aerodynamic forces The thin-layer equations are solved using the PNS technique of
and moments were then extracted from the computed flowfields by Schiff and Steger.6 Following the approach of Schiff and Steger,
integration of the predicted pressure and shear stresses acting on the governing equations, which have been modified here to include
the projectile surface. To compute the recirculating flow at the base the Coriolis and centrifugal force terms, are solved using a conser-
of the projectile and subsequently the base drag, a time-marching vative, approximately factored, implicit finite difference numerical
Navier–Stokes approach was applied. algorithm as formulated by Beam and Warming.7 A fully turbu-
Parameters that characterize the in-flight motion are subsequently lent boundary layer has been modeled using the Baldwin–Lomax
evaluated using the predicted aerodynamic coefficients, including turbulence model.8
the gyroscopic and dynamic stability factors and the projectile’s The computations presented here were performed using a shock-
fast- and slow-mode frequencies and damping coefficients. These fitting procedure reported by Rai and Chaussee.9 This procedure
parameters are then used to predict the free-flight motion of the solves the five Rankine–Hugoniot jump conditions, two geometric
projectile. In each case, the computational approach is validated by shock-propagation conditions, and one compatibility equation to
comparison with experimental data. determine the values of the five dependent variables immediately
behind the shock, as well as the position of the shock. The shock-
Computational Approach fitting procedure was appropriately modified when coning motion
The supersonic viscous flowfield about the projectile configu- was employed.5
ration was obtained by applying a PNS technique to determine The computational results presented here were obtained using a
the attached flowfield over the projectile forebody and a time- grid that consisted of 60 points between the body and the shock.
marching Navier–Stokes technique to determine the recirculating In the circumferential direction, gridding was performed over the
flow in the base region. Both of these numerical techniques are entire body (360 deg) using 36 points spaced at 10-deg increments.
briefly described. Approximately 75 axial marching steps were used for each caliber
of body length. Because several of the aerodynamic coefficients
PNS Approach are strongly influenced by viscous effects, the boundary layer was
Computation of the attached viscous flowfield about the projectile adapted by monitoring the nondimensional boundary-layer coordi-
forebody was accomplished by solving the thin-layer Navier–Stokes nate y + along the leeside of the body and adjusting the normal grid
equations using the PNS technique of Schiff and Steger.6 With use of spacing so that the y + was about 3. This approach was has been suc-
the PNS technique, computational results are obtained by marching cessfully applied in previous studies.3 The computations have been
through the grid from the projectile nose to the base. This tech- performed on a variety of computer platforms. Typical solutions for
nique is applicable in the supersonic flow regime and requires that a single set of conditions (Mach number, coning rate, angle of attack,
the flowfield contain no regions of flow separation in the axial di- etc.) required 5–10 min total CPU time on a Cray SV1 computer.
rection. Because the computational approach requires only a single The selection of grid parameters for the current study is consistent
sweep through the computational grid, it is very efficient, compared with those used in the previous studies that focused on the individual
with time-marching approaches that require many sweeps through aerodynamics coefficients.3−5 The grid requirements shown in these
the grid. previous studies indicate that accurate results for the complete set
As is standard practice, prediction of the static aerodynamic co- of aerodynamics coefficient can be obtained with a single grid. As
efficients, such as pitching moment and normal force coefficients, part of the current effort, a grid refinement study for the L/D = 9
are performed by computing the flowfield about the projectile at a body, middle c.g. position was also performed. A more refined grid,
constant angle of attack. The Magnus coefficients are obtained from with 90 points between the body and the shock, 72 circumferential
a computation of the flowfield at a constant angle of attack and spin points and 150 axial marching steps for each caliber of body length,
rate. Because the static aerodynamics for spinning axisymmetric was utilized. A comparison of the results obtained with the baseline
bodies are generally invarient with spin rate, it is possible to ob- and refined grids indicated a maximum difference of less than 1.5%
tain the static aerodynamic coefficients and the Magnus coefficients between the two complete sets of aerodynamic coefficients.
from a single computation. The pitch–damping coefficients are de-
termined from the side force and moment acting on the projectile Time-Marching Approach
undergoing steady coning motion. Steady coning motion is defined For the recirculating flow in the base region of the flowfield,
as the motion performed by a missile flying at a constant angle with the PNS technique can no longer be used, and a more generally
 WEINACHT 259

applicable approach such as the time-marching technique must be terized as a combination of aerodynamic forces and moments due
applied. The time-marching technique is based on the unsteady or to simple specific motions. The individual aerodynamic coefficients
time-dependent Navier–Stokes equations. For steady flow appli- can provide insight into the aerodynamic performance of the flight
cations, such as those discussed here, the solution is obtained by vehicles.
marching an initial “guessed” solution to a final converged solution
in a time-iterative manner. This requires the entire flowfield to be Aerodynamic Moments
updated at each time step, and numerous time steps are required as For symmetric missiles, the aerodynamic moments are typically
the solution evolves to the converged steady-state result. Because of modeled using the following moment expansion, which is cast in
this, the time-marching approach is much more computationally in- a nonrolling coordinate frame.14 The moment formulation employs
tensive in terms of its run times and computer memory requirements complex variables to separate the tranverse moment components
than the PNS technique. C̃m and C̃n , which produce rotations in the vertical and horizontal
For the predictions presented here, the time-dependent approach planes, respectively. The moment about the longitudinal axis of the
was applied to compute the axisymmetric (zero-angle-of-attack) vehicle, Cl , can be treated separately. Thus,
base flow as a means of evaluating the zero-yaw drag. The small,


but measureable, effect of yaw on drag was not considered in this C̃m + i C̃n = ( p D/V )Cn pα − iCm α ξ̃ − (i/γ ) Cm q + γ Cm α̇ ξ̃ 
study. The effect of the base flow on the other coefficients was also
assumed to be small. (3)
The particular time-marching technique applied here is the im-
Cl = ( p D/V )Cl p (4)
plicit, partially flux-split, upwind numerical scheme developed by
Ying et al.10 and Sahu and Steger.11 The technique is based on
Downloaded by BEIHANG UNIVERSITY on February 23, 2022 | http://arc.aiaa.org | DOI: 10.2514/1.1037

In the moment formulation, the pitching moment coefficient Cm α

the flux-splitting approach of Steger and Warming12 and is often re- produces a moment proportional to the complex angle of attack ξ̃ .
ferred to as the F3D technique. The upwind scheme provides natural For small-angle-of-attack flight, the cosine of the total angle of
numerical dissipation, better stability properties, and better com- attack γ is nearly one. This allows the pitch–damping moment
putational efficiency compared with implicit central-differencing coefficient, Cm q + Cm α̇ , which produces moments proportional to
schemes.10−12 the angular rate ξ̃  , to be treated as a single coefficient. The Magnus
The technique is based on the time-dependent thin-layer Navier– moment coefficient Cn pα accounts for a side moment due to flow
Stokes equations asymmetries from a combination of spin and angle of attack.
∂ q̂ ∂ Ê ∂ F̂ ∂ Ĝ 1 ∂ Ŝ Over the past decade and a half, CFD techniques have been devel-
+ + + + Ĥ = (2) oped to predict the various aerodynamic coefficients just discussed.
∂t ∂ξ ∂η ∂ζ Re ∂ζ
The pitching moment coefficient Cm α has been routinely determined
The flux-splitting algorithm of the F3D approach reduces the for various configurations and flight regimes for well over a decade.
computational requirements by eliminating one of the three implicit The prediction of the Magnus coefficients for axisymmetric projec-
inversions associated with the central difference Beam–Warming tiles at supersonic velocities was demonstrated in the early 1980s.3
algorithm.7 The two-factor implicit algorithm involves two sweeps This development was motivated primarily by the failure of the-
through the grid at each time step. The first sweep involves invert- oretical methods to properly predict the Magnus coefficients for
ing the block tridiagonal system of equations in the ζ direction boattailed projectiles,3 although reasonable predictions using the-
along grid lines of constant ξ and η to determine the intermedi- oretical methods are possible for cylindrical afterbodies in some
ate solution variable. During the second sweep, a second block tri- flight regimes.15,16 The Magnus problem is still a research topic for
diagonal system of equations is inverted in the η direction along complicated vehicles and at transonic velocities.
grid lines of constant ξ and ζ and to determine the dependent More recently, techniques for predicting the pitch–damping co-
variable. efficient, Cm q + Cm α̇ , based on the solution of the Navier–Stokes
The particular version of the code applied here has a multizone equations have been developed.5 Although the development of a
capability that allows the computational grid to be decomposed into capability to predict each individual aerodynamic coefficient can
smaller blocks. This allows greater flexibility in gridding of flight be considered an important research accomplishment, it was not
bodies, particularly when it is desirable to preserve sharp corners. possible to predict the stability characteristics of projectiles using
The multizone capability also allows the computer memory require- CFD until a complete capability for predicting all of the aerody-
ment for the computation to be reduced because only one zone needs namic coefficients was developed. The prediction of the pitch–
to be in memory at a time and the other zones can be stored on disk. damping coefficient represented that final step. Theoretical meth-
This capability was implemented by Sahu and Steger.11 ods such as slender body theory can also be applied to predict the
The zonal gridding approach was used to compute the recirculat- pitch–damping coefficients, although only qualitative comparisons
ing viscous flow in the base region. A large grid zone was located can be made with experimental and computational results.17 Semi-
in the base region and extended axially from the projectile base empirical methods18,19 offer an alternative approach for aerody-
6.35 calibers downstream. In the radial direction, the grid extended namic coefficient prediction, and these methods continue to evolve.
outward from the axis of symmetry to a radial location 3.5 calibers Note that the availability of the recent CFD predictions of pitch–
above the base corner. To allow the flow to expand properly around damping has aided in the improvement of these semi-empirical
the base corner, a small zone, one body diameter in length, was methods.19
placed on the forebody upstream from the base corner. This grid Figures 2–4 show results of CFD predictions for pitching mo-
had a one-plane overlap with the base grid and a one-to-one map- ment, Magnus moment, and pitch-damping moment coefficients
ping of the grid points in the overlap region. The dimensions of the coefficients performed as part of the current study. Results for a
base grid were 120 axial points, 50 radial points along the base, and free-flight Mach number of 1.8 are shown for each of the three c.g.
50 radial points above the base corner. The forebody grid had 30 and body lengths. In general, the CFD predictions appear to be in
axial points and 50 radial points. Clustering of the grid points in good agreement with the experimental data obtained from range
the radial direction was performed to resolve the boundary layer on firings1 of these projectiles. Results obtained at Mach 2.5 also show
the forebody and the shear layer in the base region. The algebraic similar agreement with the range data.
turbulence model of Chow13 was applied in the base region. The aerodynamic moment about the longitudinal axis of an ax-
isymmetric vehicle can be modeled using a single aerodynamic
Results coefficient. The roll–damping coefficient Cl p represents the aero-
As a first step in evaluating the aerodynamic performance of a dynamic moment that opposes the spinning motion of the vehicle.
flight vehicle, the aerodynamic force and moment coefficients must For axisymmetric flight bodies, the roll–damping coefficient is pro-
be evaluated. The aerodynamic coefficients allow forces and mo- duced purely by the viscous stresses acting circumferentially on
ments due to complicated three-dimensional motions to be charac- the body surface. From a modeling standpoint, proper prediction
 260 WEINACHT

Fig. 2 Pitching moment coefficient vs c.g. position, M = 1.8, ANSR. Fig. 5 Roll–damping moment coefficient vs body length, M = 1.8,
ANSR.
Downloaded by BEIHANG UNIVERSITY on February 23, 2022 | http://arc.aiaa.org | DOI: 10.2514/1.1037

Fig. 3 Magnus moment coefficient vs. c.g. position, M = 1.8, ANSR. Fig. 6 Normal force coefficient vs body length, M = 1.8, ANSR.

Fig. 4 Pitch–damping moment coefficient vs c.g. position, M = 1.8, Fig. 7 Magnus force coefficient vs body length, M = 1.8, ANSR.
ANSR.

of the roll–damping coefficient requires adequate grid resolution of The normal force coefficient C Nα acts normal to the longitudinal
the boundary layer. A comparison of the predicted roll–damping axis of the body in the pitch plane. The force expansion also con-
coefficient with range data is shown in Fig. 5. Both the computa- tains a damping force coefficient, C Nα̇ + C Nq , that produces a force
tional predictions and the range data show a nearly linear increase in proportional to the yawing rate ξ̃  . The Magnus force coefficient
the roll–damping coefficient with increasing body length. The pre- CY pα accounts for the side force produced by a combination of spin
dicted result shows an overprediction of about 10%. Similar results and yaw.
were observed at Mach 2.5. The predictions showed less than a 1% Figures 6 and 7 show comparisons of the predicted and exper-
variation in the roll–damping coefficient at angles of attack to 6 deg. imentally determined normal force and Magnus force coefficients
at Mach 1.8 as a function of body length. The predicted normal
force coefficient is predicted to within the scatter of the range data
Aerodynamic Forces
and shows little influence of body length. Both the predicted and
The aerodynamic forces can also be described in a manner that experimentally determined Magnus force coefficient increase with
is analogous to the aerodynamic moments. The transverse force body length. The Magnus force determined from the variation of the
components CY and C Z act along the ỹ and z̃ axes of the non- Magnus moment with c.g. location is about 30% less than the pre-
rolling coordinate frame and consist of components proportional to dictions for the L/D = 7 and L/D = 9 bodies. The Magnus force
the complex yaw ξ̃ and yawing rate ξ̃  . The third force component, coefficients determined from the swerving (c.g.) motion for indi-
which acts along the axial direction, is loosely coupled to the trans- vidual shots shows significant scatter, indicating the difficulty in
verse forces and can be treated separately. This force component accurately determining this coefficient. The pitch–damping force
consists predominantly of the drag force. Thus, coefficient is not shown because it does not influence the projectile


 motion in a measurable manner. Comparisons of predictions with
C̃Y + i C̃ Z = − C Nα + i( p D/V )CY pα ξ̃ − (1/γ ) C Nq + γ C Nα̇ ξ̃  experimental data are presented in Ref. 5.
(5) The drag coefficient is also very important for predicting projec-
tile flight performance because it is one of the primary aerodynamic
C X = −C D (6) coefficients affecting range. The drag coefficient is also required
 WEINACHT 261

trajectory, the equations can be integrated and closed-form solutions

for the projectile motion can be obtained.
For a symmetric missile, the closed-form solution for the angu-
lar response as a function of the distance down range, s, takes the
following form14 :

ξ̃ = ξ̃g + K 1 + K 2 ≈ β + iα

K 1 = K 10 exp[i(φ10 + φ1 s) + λ1 s)]

K 2 = K 20 exp[i(φ20 + φ2 s) + λ2 s)] (8)

The complex variable ξ̃ consists of the vertical and horizontal com-

ponents of the angle of attack, α and β. The angular motion contains
Fig. 8 Zero-yaw drag coefficient vs c.g. position, M = 1.8 and M = 2.5, a gravity term ξ̃g , which is small and for the remainder of the discus-
ANSR. sion this term will be ignored. The second and third terms of Eq. (8)
are expressions of the amplitudes of two components of the angular
for the stability assessment, although its effect is small. The total motion, referred to as the fast and slow modes, respectively. The
drag on the projectile is composed of contributions from the fore- complex angle of attack ξ̃ can be displayed graphically along with
body and the base. The forebody drag consists of both pressure and the fast- and slow-mode arms, K 1 and K 2 . Graphically, the solution
Downloaded by BEIHANG UNIVERSITY on February 23, 2022 | http://arc.aiaa.org | DOI: 10.2514/1.1037

viscous drag components. The base drag results from the pressure is shown by Fig. 9 and is referred to as a damped epicycle. It is
acting on the projectile base. The forebody drag can be predicted helpful to regard Fig. 9 as the path that the nose of the projectile
using the PNS technique. The recirculating flow in the base region would follow as it flies down range. The fast- and slow-mode fre-
is predicted using the time-marching technique from which the base quencies of the motion, φ1 and φ2 , and the damping rates, λ1 , and λ2 ,
drag can be determined. are functions of the aerodynamic coefficients, projectile spin rate,
A comparison of the predicted zero-yaw drag with range results and inertial properties of the body. Generally speaking, the frequen-
is shown in Fig. 8. Drag predictions at Mach 1.8 and 2.5 are shown cies are influenced by the pitching moment coefficient, whereas the
as a function of body length. Both the experimental results and the damping is influenced by the Magnus and pitch–damping coeffi-
computational results show a slight increase in zero-yaw drag with cients. The equation also contains the initial amplitudes, K 10 and
increasing body length and decreasing Mach number. The computed K 20 , and phase angles, φ10 and φ20 , which result from applying the
drag coefficient is predicted to within the accuracy of the range data. initial conditions during the integration of the equations.
The increase in the drag with body length is due primarily to the Using the predicted aerodynamic coefficients, the fast- and slow-
increase in the wetted surface of the longer body. The predicted mode damping and frequencies were computed for the various body
results showed only minor variations in the base drag with body lengths, c.g., launch Mach numbers, and spin rates. Figure 10 shows
length. the predicted slow-mode frequency plotted against the experimen-
At Mach 1.8 for the L/D = 9 body, the pressure and viscous tally determined value. (The results are plotted in this fashion be-
forebody drag account for 39% and 25% of the total drag with the cause of the large number of parameters considered here.) Points
base drag accounting for the remaining 36% of the drag. Predictions
of the forebody pressure drag are within 1% of previous predictions
using empirical20 or inviscid1 approaches.
An estimate of the predicted skin-friction drag C Dv can be ob-
tained using the relation developed by Charters and Kent21 :

C Dv = −4Cl p (7)

The estimated skin-friction drag determined from the predicted roll–

damping coefficient underestimates the predicted skin-friction drag
by 3–10% for the predictions presented here. (This comparison gives
an indication of the accuracy of the Charters and Kent relation.)
The estimated skin friction obtained from the range measurements
of the roll–damping coefficient is 13–22% less than the predicted
skin friction. Although these estimates indicate a potential 10%
overprediction of the skin-friction drag, the contribution to the total Fig. 9 Projectile angular motion epicycle.
drag would result in an overprediction of only a few percent.
Design codes20 can provide a more cost-effective alternative for
predicting the base drag, thus, eliminating the need for a time-
dependent base drag computation. The semi-empirical approach of
McCoy20 appears to overpredict the base drag by 10–20% com-
pared to the CFD result. When the estimated base drag is added to
the forebody-drag CFD predictions, the total drag is overpredicted
by less than 9% compared to the CFD results. For stability com-
putations, where the drag coefficient plays a minor role, acceptable
accuracy can be obtained using base drag estimates.

Stability and Free-Flight Motion

Once the aerodynamic forces and moments have been determined,
the equations of motion can be integrated and the projectile’s mo-
tion can be determined for various initial conditions. Because the
equations of motion are nonlinear, numerical integration is usu-
ally required to predict the projectile motion from launch to impact. Fig. 10 Comparison of computed slow-mode frequency with mea-
These nonlinearities are typically weak, and for portions of the flight sured slow-mode frequency, ANSR.
 262 WEINACHT

statically unstable, the two stability criteria can be satisfied by ad-

justing the spin rate of the projectile, as long as the dynamic stability
factor is between zero and two. If the dynamic stability factor is out-
side this bound, the projectile cannot be stabilized using spin. A third
condition for stability essentially requires the pitch–damping coef-
ficient, Cm q + Cm α̇ , to be negative. This condition is rarely violated
in practice.
The constraints on the gyroscopic stability and dynamic stability
factors discussed earlier can be shown graphically on what is often
termed a stability plot. Figures 13 and 14 show results as measured in
the aerodynamics range and obtained from the computation. Shown
are results for each of the three c.g., the three different length bodies,
various spin rates, and Mach numbers of 1.8 and 2.5. For each firing,
experimental and computational values of the gyroscopic stability
factor sg and the dynamic stability factor sd were computed. It is
Fig. 11 Comparison of computed fast-mode damping with measured important to realize that for a given set of conditions (body length,
fast-mode damping, ANSR. spin rate, launch Mach number), the computational results will yield
a unique value. However, because of the scatter associated with the
experimental data, each firing yields a slightly different value. Thus,
the experimental results gives the appearance of having more data.
Downloaded by BEIHANG UNIVERSITY on February 23, 2022 | http://arc.aiaa.org | DOI: 10.2514/1.1037

It is important to realize this when comparing Figs. 13 and 14 data.

The computational and experimental results presented in Figs. 13
and 14 show that this family of projectiles is statically unstable
(sg > 0), but has been made gyroscopically stable by appropriate
selection of the spin rate (sg > 1). The results also show that all of
the L/D = 5 projectiles are dynamically stable. (The region of dy-
namic stability is denoted by the shaded area in Figs. 13 and 14.)
The results indicate that several of the L/D = 7 and L/D = 9 pro-
jectiles are marginally stable, although these can be stabilized by
selection of a higher spin rate because the dynamic stability factor
is between zero and two. Several of the L/D = 9 bodies are dynam-
ically unstable and cannot be stabilized by changing the spin rate.
Fig. 12 Comparison of computed slow-mode damping with measured Further examination of the data revealed that the L/D = 9 bodies
slow-mode damping, ANSR. that could not be stabilized all had the c.g. in the aftmost position.
Note that the computational approach correctly predicts the stabil-
that fall close to the solid line indicate that the predicted value is ity characteristics of the various configurations for the conditions of
close to the experimentally determined value. dynamic stability, marginal dynamic stability, and dynamic insta-
There is good agreement between prediction and experiment. The bility. This demonstrates the utility of the computational approach
slow-mode frequency represents the critical comparison for the fre- for the purposes of stability prediction.
quencies because the fast-mode frequency is dominated by the spin
rate, which is essentially an input parameter.
Figures 11 and 12 show comparisons of the predicted slow- and
fast-mode damping of each of the bodies. The data are scattered
about the solid line, indicating that the prediction is generally in
good agreement with the experiment. No consistent bias between the
computation and experimental data is seen. This indicates that there
is no consistent underprediction or overprediction of the damping
rates. Much of the scatter in the data is attributed to the uncertainty in
the experimental data rather than inaccuracies in the computational
approach.
Generally, from a design standpoint, predicting and controlling
damping rates, rather than the frequencies, is a primarily consider-
ation. One desirable requirement is that the damping rates λ1 and
λ2 be negative so that the angular motion does not grow during the
projectile’s flight. When the equations for the damping rates are al- Fig. 13 Stability plot for ANSR from range data.
gebraically manipulated, three necessary conditions can be derived,
which will ensure that the damping rates are negative. Two of these
conditions are
 
sg > 1, sg = f p, Cm α (9)
 
sg < sd (2 − sd ), sd = f Cn pα , Cm q + Cm α̇ (10)

The first condition states that the gyroscopic stability factor sg must
be greater than one for the projectile to be gyroscopically stable.
The second condition also places an additional constraint on the
gyroscopic stability factor through the dynamic stability factor sd .
As shown in these equations, the gyroscopic stability and dynamic
stability factors are functions of the aerodynamic coefficients. Ad-
ditionally, the gyroscopic stability factor is also a function of the
projectile spin rate p. For typical spin-stabilized projectiles that are Fig. 14 Stability plot for ANSR from CFD predictions.
 WEINACHT 263

comparison, the in-flight angular motion for individual firings was

compared with range data and showed good agreement. The results
are believed to properly represent the capability for predicting the
linear aerodynamics for the high-speed low-angle-of-attack flight of
aerodynamically smooth axisymmetric projectiles. Additional com-
putational effort (and, in some cases, additional research) may be
required to establish the same capability at subsonic and transonic
velocities or for complex geometries.

References
1 Murphy, C. H., and Schmidt, L. E., “The Effect of Length on the Aerody-
namics Characteristics of Bodies of Revolution in Supersonic Flight,” U.S.
Army Ballistic Research Lab., Rept. 876, Aberdeen Proving Ground, MD,
Aug. 1953.
2 Nietubicz, C. J., and Opalka, K., “Supersonic Wind Tunnel Measure-
ments of Static and Magnus Aerodynamic Coefficients for Projectile Shapes
with Tangent and Secant Ogive Noses,” U.S. Army Ballistic Research Lab.,
Memorandom Rept. ARBRL-MR-02991, Aberdeen Proving Ground, MD,
Fig. 15 In-flight angular motion, 5-caliber body, M = 1.8. Feb. 1980.
3 Sturek, W. B., and Schiff, L. B., “Computations of the Magnus Effect for
Slender Bodies in Supersonic Flight,” Proceedings of the AIAA Atmospheric
Downloaded by BEIHANG UNIVERSITY on February 23, 2022 | http://arc.aiaa.org | DOI: 10.2514/1.1037

Flight Mechanics Conference, AIAA, New York, 1980, pp. 260–270.

4 Sturek, W. B., and Mylin, D. C., “Computational Parametric Study of
the Magnus Effect on Boattailed Shell at Supersonic Speeds,” AIAA Paper
81-1900, Aug. 1981.
5 Weinacht, P., Sturek, W. B., and Schiff, L. B., “Navier–Stokes Predictions
of Pitch Damping for Axisymmetric Projectiles,” Journal of Spacecraft and
Rockets, Vol. 34, No. 6, 1997, pp. 753–761.
6 Schiff, L. B., and Steger, J. L., “Numerical Simulation of Steady Super-
sonic Viscous Flow,” AIAA Journal, Vol. 18, No. 12, 1980, pp. 1421–1430.
7 Beam, R., and Warming, R. F., “An Implicit Factored Scheme for the
Compressible Navier–Stokes Equations,” AIAA Journal, Vol. 16, No. 4,
1978, pp. 85–129.
8 Baldwin, B. S., and Lomax, H., “Thin Layer Approximation and Al-
gebraic Model for Separated Turbulent Flows,” AIAA Paper 78-257, Jan.
1978.
9 Rai, M. M., and Chaussee, D. S., “New Implicit Boundary Procedure:
Theory and Applications,” AIAA Paper 83-0123, Jan. 1983.
10 Ying, S. X., Steger, J. L., Schiff, L. B., and Baganoff, D., “Numeri-
Fig. 16 In-flight angular motion, 7-caliber body, M = 2.5. cal Simulation of Unsteady, Viscous, High-Angle-of-Attack Flows Using a
Partially Flux-Split Algorithm,” AIAA Paper 86-2179, Aug. 1986.
11 Sahu, J., and Steger, J. L., “Numerical Simulation of Three-Dimensional
With the prediction of the frequencies and damping rates of the
angular motion, the angular motion itself can be predicted and Transonic Flows,” International Journal for Numerical Methods in Fluids,
Vol. 10, June 1990, pp. 855–873.
compared with the motion measured in the aerodynamics range. 12 Steger, J. L., and Warming, R. F., “Flux Vector Splitting of the Invis-
Figures 15 and 16 show the angular motion of a 5-caliber projectile cid Gasdynamic Equations with Application to Finite-Difference Methods,”
launched at Mach 1.8 and a 7-caliber projectile launched at Mach Journal of Computational Physics, Vol. 40, April 1981, pp. 263–293.
2.5. These three-dimensional plots show the vertical and horizontal 13 Chow, W. L., “Improvement on Numerical Computation of the Thin-

components of angle of attack, α and β, as a function of the distance Layer Navier–Stokes Equation With Special Emphasis on the Turbulent Base
downrange, Z . The thickened curve displays the three-dimensional Pressure of a Projectile in Transonic Flight Condition,” Univ. of Illinois,
motion. Two-dimensional projections of the motion in the vertical Contract Rept. DAAG29-81-D-0100, Urbana, IL, Nov. 1985.
14 Murphy, C. H., “Free Flight Motion of Symmetric Missiles,” U.S. Army
and horizontal planes are also shown. It is useful to regard the curves
Ballistic Research Lab., Rept. 1216, Aberdeen Proving Ground, MD, July
shown here as the path traversed by the nose of the projectile as it flies 1963.
downrange. The motion predicted using the CFD derived aerody- 15 Vaughn, H. R., and Reis, G. E., “A Magnus Theory,” AIAA Paper 73-
namic coefficients is denoted by the solid lines, whereas the motion 124, Jan. 1973.
obtained from the range firings is shown by the circular symbols. The 16 Sturek, W. B., Dwyer, H. A., Kayser, L. D., Nietubicz, C. J., Reklis, R. P.,

agreement between the two motions is excellent. Comparisons of and Opalka, K. O., “Computations of Magnus Effect for a Yawed, Spinning
the in-flight motion at other flight conditions and for the other body Body of Revolution,” AIAA Journal, Vol. 16, No. 7, 1978, pp. 687–692.
17 Weinacht, P., “Navier–Stokes Predictions of the Individual Components
geometries also showed good agreement with the experimental data.
of the Pitch-Damping Sum,” Journal of Spacecraft and Rockets, Vol. 35,
No. 5, 1998, pp. 598–605.
Conclusions 18 Whyte, R. H., “ ‘Spinner’—A Computer Program for Predicting the

CFD provides an alternate means for determining the aerody- Aerodynamic Coefficients of Spin Stabilized Projectiles,” General Electric
namic performance of projectiles. This will allow projectiles to be Co., Class 2 Rept. 69APB3, Burlington, VT, Aug. 1969.
19 Moore, F. G., and Hymer, T. C., “The 2002 Version of the Aeropredic-
designed with a reduced requirement for manufacturing models and
tion Code: Part I—Summary of New Theoretical Methodology,” U.S. Naval
performing test firings. As a test case, the aerodynamic performance
Surface Warfare Center, Rept. NSWCDD/TR-01/108, Dahlgren, VA, March
of a family of axisymmetric projectiles in supersonic flight was ex- 2002.
amined. The predictions showed good agreement with range data for 20 McCoy, R. L., “ ‘MC DRAG’—A Computer Program for Estimating
individual aerodynamic coefficients. Subsequently, the frequencies the Drag Coefficients of Projectiles,” U.S. Army Ballistic Research Lab.,
and damping rates of the angular motion were computed using these Technical Rept. ARBRL-TR-02293, Aberdeen Proving Ground, MD, Feb.
coefficients. The results showed good correlation with the experi- 1981.
21 Charters, A. C., and Kent, R. H., “The Relation Between the Skin Fric-
mental data for various body lengths, c.g. locations, Mach numbers
and spin rates. The aerodynamic coefficients were also used to pre- tion and the Spin Reducing Torque,” U.S. Army Ballistic Research Lab.,
dict gyroscopic and dynamic stability factors. The computational Rept. 287, Aberdeen Proving Ground, MD, July 1942.
approach correctly distinguished the conditions of dynamic insta- R. Cummings
bility, marginal dynamic stability, and dynamic stability. As a final Associate Editor
 This article has been cited by:

1. Jian Shen, Shao-bo Fan, Ya-xin Ji, Qing-yu Zhu, Ji Duan. 2020. Aerodynamics analysis of a hypersonic electromagnetic
gun launched projectile. Defence Technology 16:4, 753-761. [Crossref]
2. V. de Briey, I. Ndindabahizi, B. G. Marinus, M. Pirlot. 2019. Aerodynamical CFD Study of a Non-Lethal 12-Gauge Fin-
Stabilized Projectile. Human Factors and Mechanical Engineering for Defense and Safety 3:1. . [Crossref]
3. Véronique de Briey, Benoit G. Marinus, Marc Pirlot. Aerodynamic Characterization of a Non-Lethal Finned Projectile at
Low Subsonic Velocity . [Citation] [PDF] [PDF Plus]
4. Frank Fresconi, Ilmars Celmins, James Maley, Bryant Nelson. Experimental Flight Characterization of a Canard-
Controlled, Subsonic Missile . [Citation] [PDF] [PDF Plus]
5. Frank Fresconi, Ilmars Celmins. Experimental Flight Characterization of Spin-Stabilized Projectiles at High Angle of
Attack . [Citation] [PDF] [PDF Plus]
6. Vishal A. Bhagwandin. 2016. High-Alpha Prediction of Roll Damping and Magnus Stability Coefficients for Finned
Projectiles. Journal of Spacecraft and Rockets 53:4, 720-729. [Abstract] [Full Text] [PDF] [PDF Plus]
Downloaded by BEIHANG UNIVERSITY on February 23, 2022 | http://arc.aiaa.org | DOI: 10.2514/1.1037

7. Guillaume Strub, Simona Dobre, Vincent Gassmann, Spilios Theodoulis, Michel Basset. 2016. Pitch-Axis Identification
for a Guided Projectile Using a Wind-Tunnel-Based Experimental Setup. IEEE/ASME Transactions on Mechatronics 21:3,
1357-1365. [Crossref]
8. Graham Doig, Shibo Wang, Harald Kleine, John Young. 2016. Aerodynamic Analysis of Projectiles in Ground Effect at
Near-Sonic Mach Numbers. AIAA Journal 54:1, 150-160. [Abstract] [Full Text] [PDF] [PDF Plus]
9. Wanchai Jiajan, Randy S.M. Chue, Tue Nguyen, Yin Yin Pey, Simon C.M. Yu. 2015. Aerodynamic characteristics of high
performance rounds at Mach 1.8 to 4. Aerospace Science and Technology 40, 62-74. [Crossref]
10. Xu Liu, Wei Liu, Yun Fei Zhao. 2014. Dynamics Characterization of Missiles with Control Flaps Based on CFD. Advanced
Materials Research 1016, 221-227. [Crossref]
11. Graham Doig, Shibo Wang, John Young, Harald Kleine. Aerodynamics of Transonic and Supersonic Projectiles in Ground
Effect . [Citation] [PDF] [PDF Plus]
12. Sidra Silton, Vishal Bhagwandin. Computational Analysis of a Spinning Fin-Stabilized Projectile . [Citation] [PDF] [PDF
Plus]
13. Sidra Silton. Navier-Stokes Predictions of Aerodynamic Coefficients and Dynamic Derivatives of a 0.50-cal Projectile .
[Citation] [PDF] [PDF Plus]
14. Mark Carpenter, Roy Hartfield, John Burkhalter. A Comprehensive Approach to Cataloging Missile Aerodynamic
Performance using Surrogate Modeling Techniques and Statistical Learning . [Citation] [PDF] [PDF Plus]
15. Graham Doig, Tracie J. Barber, Eddie Leonardi, Andrew J. Neely, Harald Kleine. 2010. Aerodynamics of a Supersonic
Projectile in Proximity to a Solid Surface. AIAA Journal 48:12, 2916-2930. [Citation] [PDF] [PDF Plus]
16. Jubaraj Sahu, Karen Heavey. Time-Accurate Computations for Rapid Generation of Missile Aerodynamics . [Citation]
[PDF] [PDF Plus]
17. Carlo Montalvo, Mark Costello. Estimation of Projectile Aerodynamic Coefficients Using Coupled CFD/RBD Simulation
Results . [Citation] [PDF] [PDF Plus]
18. James DeSpirito, Sidra I. Silton, Paul Weinacht. 2009. Navier-Stokes Predictions of Dynamic Stability Derivatives:
Evaluation of Steady-State Methods. Journal of Spacecraft and Rockets 46:6, 1142-1154. [Citation] [PDF] [PDF Plus]
19. Jenna Stahl, Mark Costello, Jubaraj Sahu. Projectile Aerodynamic Coefficient Estimation Using Integrated CFD/RBD
and Flight Control System Modeling . [Citation] [PDF] [PDF Plus]
20. M Costello, J Sahu. 2008. Using computational fluid dynamic/rigid body dynamic results to generate aerodynamic models
for projectile flight simulation. Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering
222:7, 1067-1079. [Crossref]
21. James DeSpirito, Sidra Silton, Paul Weinacht. Navier-Stokes Predictions of Dynamic Stability Derivatives: Evaluation of
Steady State Methods . [Citation] [PDF] [PDF Plus]
22. Paul Weinacht. Characterization of Small-Caliber Ammunition Performance Using a Virtual Wind Tunnel Approach .
[Citation] [PDF] [PDF Plus]
23. James DeSpirito, Peter Plostins. CFD Prediction of M910 Projectile Aerodynamics: Unsteady Wake Effect on Magnus
Moment . [Citation] [PDF] [PDF Plus]
 24. Mark Costello, Stephen Gatto, Jubaraj Sahu. Using CFD/RBD Results to Generate Aerodynamic Models for Projectile
Flight Simulation . [Crossref]
Downloaded by BEIHANG UNIVERSITY on February 23, 2022 | http://arc.aiaa.org | DOI: 10.2514/1.1037