COMPUTATIONAL

BALLISTICS III

WIT Press publishes leading books in Science and Technology.

Visit our website for the current list of titles.
www.witpress.com

WITeLibrary
Home of the Transactions of the Wessex Institute.
Papers presented at Computational Ballistics III are archived
in the WIT eLibrary in volume 45 of WIT Transactions on
Modelling and Simulation (ISSN 1743-355X). The WIT eLibrary provides the
international scientific community with immediate and permanent access to individual
papers presented at WIT conferences.
http://library.witpress.com.
 THIRD INTERNATIONAL CONFERENCE ON
COMPUTATIONAL BALLISTICS

COMPUTATIONAL BALLISTICS III

CONFERENCE CHAIRMEN

C.A. Brebbia
Wessex Institute of Technology, UK

A.A. Motta
Brazilian Navy Research Institute, Brazil

INTERNATIONAL SCIENTIFIC ADVISORY COMMITTEE

C. E. Anderson N. Ishikawa D. W. Pepper
R. Barauskas Y. Kato A. Peratta
M. Beppu S. Kubota H. Rothe
J. L. Drummond Alves W-S. Lee J. Sahu
D. Greatrix

ORGANISED BY
The Wessex Institute of Technology, UK

SPONSORED BY
WIT Transactions on Modelling & Simulation
 WIT Transactions on Modelling and Simulation

Transactions Editor

Carlos Brebbia
Wessex Institute of Technology
Ashurst Lodge, Ashurst
Southampton SO40 7AA, UK
Email: [email protected]

Editorial Board

C Alessandri M A Atherton
Universita di Ferrara South Bank University
Italy UK
J Baish C D Bertram
Bucknell University The University of New South Wales
USA Australia
D E Beskos M Bonnet
University of Patras Ecole Polytechnique
Greece France
J A Bryant M B Bush
University of Exeter The University of Western Australia
UK Australia
M A Celia A H-D Cheng
Princeton University University of Mississippi
USA USA
J J Connor D E Cormack
Massachusetts Institute of Technology University of Toronto
USA Canada
D F Cutler E R de Arantes e Oliveira
Royal Botanic Gardens Insituto Superior Tecnico
UK Portugal
G De Mey J Dominguez
Ghent State University University of Seville
Belgium Spain
Q H Du S Elghobashi
Tsinghua University University of California Irvine
China USA
A El-Zafrany P Fedelinski
Cranfield University Silesian Technical University
UK Poland
S Finger J I Frankel
Carnegie Mellon University University of Tennessee
USA USA
M J Fritzler L Gaul
University of Calgary Universitat Stuttgart
Canada Germany
G S Gipson S Grilli
Oklahoma State University University of Rhode Island
USA USA
 K Hayami J A C Humphrey
National Institute of Informatics Bucknell University
Japan USA
D B Ingham N Kamiya
The University of Leeds Nagoya University
UK Japan
D L Karabalis J T Katsikadelis
University of Patras National Technical University of Athens
Greece Greece
H Lui W J Mansur
State Seismological Bureau Harbin COPPE/UFRJ
China Brazil
R A Meric J Mikielewicz
Research Institute for Basic Sciences Polish Academy of Sciences
Turkey Poland
K Onishi E L Ortiz
Ibaraki University Imperial College London
Japan UK
M Predeleanu D Qinghua
University Paris VI Tsinghua University
France China
S Rinaldi T J Rudolphi
Politecnico di Milano Iowa State University
Italy USA
G Schmid A P S Selvadurai
Ruhr-Universitat Bochum McGill University
Germany Canada
X Shixiong P Skerget
Fudan University University of Maribor
China Slovenia
V Sladek T Speck
Slovak Academy of Sciences Albert-Ludwigs-Universitaet Freiburg
Slovakia Germany
J Stasiek S Syngellakis
Technical University of Gdansk University of Southampton
Poland UK
M Tanaka N Tosaka
Shinshu University Nihon University
Japan Japan
T Tran-Cong W S Venturini
University of Southern Queensland University of Sao Paulo
Australia Brazil
J F V Vincent J R Whiteman
The University of Bath Brunel University
UK UK
Z-Y Yan K Yoshizato
Peking University Hiroshima University
China Japan
G Zharkova
Institute of Theoretical and Applied
Mechanics
Russia
COMPUTATIONAL
BALLISTICS III

Editors

C.A. Brebbia
Wessex Institute of Technology, UK

A.A. Motta
Brazilian Navy Research Institute, Brazil
Editors:
C.A. Brebbia
Wessex Institute of Technology, UK
A.A. Motta
Brazilian Navy Research Institute, Brazil

Published by
WIT Press
Ashurst Lodge, Ashurst, Southampton, SO40 7AA, UK
Tel: 44 (0) 238 029 3223; Fax: 44 (0) 238 029 2853
E-Mail: [email protected]
http://www.witpress.com
For USA, Canada and Mexico
WIT Press
25 Bridge Street, Billerica, MA 01821, USA
Tel: 978 667 5841; Fax: 978 667 7582
E-Mail: [email protected]
http://www.witpress.com

British Library Cataloguing-in-Publication Data

A Catalogue record for this book is available

from the British Library.

ISBN: 978-1-84564-077-4
ISSN: (print) 1746-4064
ISSN: (on-line) 1743-355X
Library of Congress Catalog Card Number:

The texts of the papers in this volume were set

individually by the authors or under their supervision.
Only minor corrections to the text may have been carried
out by the publisher.

No responsibility is assumed by the Publisher, the Editors and Authors for any injury and/
or damage to persons or property as a matter of products liability, negligence or otherwise,
or from any use or operation of any methods, products, instructions or ideas contained in the
material herein.

© WIT Press 2007

Printed in Great Britain by Athenaeum Press Ltd.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means, electronic, mechanical, photocopying,
recording, or otherwise, without the prior written permission of the Publisher.
 Preface

The study of computational ballistics is one of the most challenging scientific

endeavours, not only because of the diversity of time and space scales involved in
practically all problems, but due to its interdisciplinary character.
Interest in ballistics has also grown considerably recently as a result of the
complexity of the political and military situation in many parts of the world. The
classical subdivision into internal, external and terminal ballistics now encompasses
numerous applications on how to model and protect against accidental or man made
explosions. Many other civilian applications of ballistics are also of growing
importance as impact, crash and blast problems are a by-product of more
sophisticated techniques involving dimensions and speeds impossible to achieve a
few years ago.
The computational solution of problems in ballistics require specialised tools.
Problems with very short time scales are particularly difficult to solve accurately as
well as those involving changing of state or material and geometry conditions.
Computational ballistics solutions frequently require substantial computer power
and the use of specialised algorithms.
The lack of sufficient number of open meetings in the field also preclude the
sharing of important information amongst researchers and hence a way to avoid
choosing the wrong alternatives. Because of that, conferences such as this are
particularly important.
This conference started in Rio in 2003 and was reconvened in Cordoba, Spain in
2005 before the meeting that took place in the New Forest.
This volume covers the following topics:
• Fluid flow aerodynamics
• Interior ballistics
• Terminal ballistics
• Experimental mechanics/ballistics and field testing
• New developments in computational techniques
• Systems and technology
The Editors are grateful to the members of the International Scientific Advisory
Committee and other colleagues who helped to review the papers included in this
book. They are also indebted to all authors for their presentations.

The Editors
New Forest, 2007
This page intentionally left blank
 Contents

Section 1: Fluid flow aerodynamics

Unsteady computational fluid dynamics modeling of free-flight

projectile aerodynamics
J. Sahu ...................................................................................................................3

3D transient numerical simulation for sabot separation

D. Touati & S. Peles............................................................................................13

Fast calculation of direct fire trajectories taking the earth’s rotation

into account
W. Roetzel, W. Czarnetzki & T. Maier ................................................................21

Generating an aerodynamic model for projectile flight simulation

using unsteady time accurate computational fluid dynamic results
J. Kokes, M. Costello & J. Sahu..........................................................................31

Section 2: Interior ballistics

Suppression of axial combustion instability in solid rocket motors

D. R. Greatrix......................................................................................................57

ICE-ALE modelling for the combustion of homogeneous

solid propellant
A. Peratta & C. Peratta.......................................................................................67

Stochastic study of 60-mm gun-projectile responses

M. Chen ...............................................................................................................77
Section 3: Terminal ballistics

On the linkage of impact damage to modeling of

ballistic performance
J. M. Wells...........................................................................................................89

A method to increase the tip velocity of a shaped charge jet using

a hollow cavity
D. R. Scheffler & W. P. Walters..........................................................................99

A computational viscoelasticity/damage/plasticity model for

high-explosive materials
P. Rangaswamy & R. M. Hackett......................................................................109

Section 4: Experimental mechanics/ballistics and field testing

Offshore platforms survivability to underwater explosions: part I

A. A. Motta, E. A. P. Silva, N. F. F. Ebecken & T. A. Netto .............................123

Ballistic tests and numerical simulations for containment capability

characterisation of Waspaloy® alloy
M. Di Sciuva, M. Degiovanni, M. Mattone, M. Gherlone,
N. Bonora, A. Ruggiero & C. Frola..................................................................131

Experimental survey of the pressure cone and temperature field of

an A/T missile system
K. Krallis, Ν. G. Orfanoudakis, M. Founti, D. T. Hountalas,
A. Fatsis & N. Vlachakis ...................................................................................141

Section 5: New developments in computational techniques

Safety checks and hit probability computation in numerical trajectory

calculation using error propagation technology
A. Kuhrt & H. Rothe..........................................................................................153

Scalable parallel computation of explosively formed

penetrators (EFPs)
E. L. Baker, C. Chin, B. E. Fuchs & J. Pham ...................................................163

Data mining in military systems

A. A. Motta, A. S. Alves & N. F. F. Ebecken .....................................................171
Computer simulated fragmentation arena test for assessing
lethality and safety separation distances of explosive
fragmentation ammunitions
V. M. Gold, E. L. Baker & J. M. Pincay ...........................................................181

Section 6: Systems and technology

Evaluations of computational techniques for the engraving

of projectiles
J. South, B. Powers & M. Minnicino.................................................................193

Long range electromagnetic cannon

E. Schmidt & J. Tzeng .......................................................................................203

Analysis of weapon systems protecting military camps against

mortar fire
M. Graswald, I. Shaydurov & H. Rothe............................................................213

Empirical relationship for muzzle exit pressure in a 155 mm

gun tube
D. Carlucci & J. Vega.......................................................................................225

Author Index ...................................................................................................231

This page intentionally left blank
 Section 1
Fluid flow aerodynamics
This page intentionally left blank
 Computational Ballistics III 3

Unsteady computational fluid dynamics

modeling of free-flight projectile aerodynamics
J. Sahu
U.S. Army Research Laboratory, APG, MD, USA

Abstract
This paper describes a new multidisciplinary computational study undertaken to
model the flight trajectories and the free-flight aerodynamics of both a finned
projectile at supersonic velocities and a spinning projectile at subsonic speeds
with and without aerodynamic flow-control. Actual flight trajectories are
computed using an advanced coupled computational fluid dynamics (CFD)/rigid
body dynamics (RBD) technique. An advanced time-accurate Navier–Stokes
computational technique has been used in CFD to compute the unsteady
aerodynamics associated with the free-flight of the finned projectile at supersonic
speeds and the spinning projectile at subsonic speeds. Computed positions and
orientations of the projectile have been compared with actual data measured
from free-flight tests and are found to be generally in good agreement. Predicted
aerodynamics forces and moments compare well with the forces and moments
used in the six degree freedom fits of the results of the same tests. Unsteady
numerical results obtained from the coupled method show the flow field, the
aerodynamic forces and moments, and the flight trajectories of the projectile.
Keywords: unsteady CFD, projectile aerodynamics, couple CFD/RBD method,
flight trajectory.

1 Introduction
Understanding the aerodynamics of projectiles, rockets, and missiles is critical to
the design of stable configurations and contributes significantly to the overall
performance of weapon systems [1-3]. The prediction of aerodynamic
coefficients for these weapon systems is essential in assessing the performance
of new designs. Numerical simulations have the potential of greatly reducing
design costs while providing a detailed understanding of the complex

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070011
4 Computational Ballistics III

aerodynamics associated with each change. Recently, we have made progress in

coupling computational fluid dynamics and flight dynamics to perform required
multidisciplinary simulations for moving body problems. This involves real-
time multidisciplinary-coupled computational fluid dynamics/rigid body
aerodynamics computations for the entire flight trajectory of a complex guided
projectile system. It can lead to accurate determination of aerodynamics that is
critical to the low-cost development of new advanced guided projectiles, rockets,
missiles, and smart munitions.
Improved computer technology and state-of-the-art numerical procedures now
enable solutions to complex, 3-D problems associated with projectile and missile
aerodynamics. Modern guided munitions for future combat systems require the
use of complex control surfaces (fins and canards), control mechanisms, and/or
the use of flow technologies such as microjet gas generators to provide maneuver
authority. The aerodynamic flow fields over these Army weapons are complex
involving non-linear flow-physics especially during and after control maneuvers.
For maneuvering munitions however, very limited data is available during and
after control maneuvers [4,5], and there is a lack of knowledge and
understanding of the associated unsteady aerodynamics. Accurate numerical
modeling of this unsteady aerodynamics has been found to be challenging both
in terms of time-accurate solution techniques and computing resources required.
Our goal is to be able to perform time-accurate multidisciplinary-coupled
computational fluid dynamics (CFD) and rigid body dynamics (RBD)
computations for complex guided projectiles with control maneuvers using
microjets and/or control surfaces such as fins/canards. As part of a DOD High
Performance Computing Grand Challenge Project, the present work is focused
on the coupling of CFD and rigid body dynamics (RBD) techniques for
simultaneous prediction of the unsteady free-flight aerodynamics and the flight
trajectory of projectiles. In other words, can we perform physics-based fly-outs
of the projectiles on the supercomputers and accurately predict the unsteady
aerodynamics and flight behavior of projectiles in actual flights?

2 Technical approach
A real-time accurate approach is used in the present work; however, time-
accurate computations require much greater computer resources. The real-time
accurate approach also requires that the six-degrees-of-freedom (6 DOF) body
dynamics be computed at each repetition of the fluid flow solver. In three-
dimensional space, a rigid object has 6 DOF: three translations and three
rotations. The 6 DOF code computes linear and angular velocities as well as the
orientation of the missile, which are used as input to the computational fluid
dynamics code. In turn, the aerodynamic forces and moments obtained from the
flow solver are used to solve the 6 DOF body dynamics before moving on to the
next time step. This procedure allows one to perform real-time
multidisciplinary-coupled computational fluid dynamics/rigid body
aerodynamics computations for the partial or entire flight trajectory of a complex
guided projectile system.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 5

The CFD capability used here solves the Navier-Stokes equations [6-9] and
incorporates advanced boundary conditions and grid motion capabilities. The
present numerical study is a big step forward and a direct extension of that
research which now includes numerical simulation of the actual fight paths of the
projectile using coupled CFD/RBD techniques using real-time accurate
approach. The complete set of 3-D time-dependent Navier–Stokes equations is
solved in a time-accurate manner for simulations of actual flights. The basic
numerical framework in the code contains unified-grid, unified-physics, and
unified-computing features. The user is referred to these references for details of
the basic numerical framework. The 3-D time-dependent Reynolds-averaged
Navier–Stokes (RANS) equations are solved using the finite volume method [7]:
∂
WdV + ∫ [F − G]⋅ dA = ∫ HdV
∂t V∫
(1)
V

where W is the vector of conservative variables, F and G are the inviscid and
viscous flux vectors, respectively, H is the vector of source terms, V is the cell
volume, and A is the surface area of the cell face.
Second-order discretization was used for the flow variables and the turbulent
viscosity equation. The turbulence closure is based on topology-parameter-free
formulations. Two-equation [6] and higher order hybrid RANS/LES [10,11]
turbulence models were used for the computation of turbulent flows. The hybrid
RANS/LES approach is well suited to the simulation of unsteady flows and
contains no additional empirical constants beyond those appearing in the original
RANS and LES sub-grid models. With this method a regular RANS-type grid is
used except in isolated flow regions where denser, LES-type mesh is used to
resolve critical unsteady flow features. The hybrid model transitions smoothly
between an LES calculation and a cubic k-ε model, depending on grid fineness.
For computations of unsteady flow fields that are of interest here, dual time-
stepping as described below was used to achieve the desired time-accuracy [12].
An unique feature of the present coupled approach is the full grid motion
capability that allows the grid to move translate and rotate as the projectile flies
down the rage, since the grid velocity is assigned to each mesh point. To
account for rigid body dynamics, the grid point velocities are set as if the grid is
attached to the rigid body with 6 DOF. As shown schematically in Figure 1, the
6 DOF comprise of the three spatial coordinates (x,y,z) and the three Euler
angles, roll, pitch, and yaw (φ, Φ, Ψ). For the rigid body dynamics, the coupling
refers to the interaction between the aerodynamic forces/moments and the
dynamic response of the projectile/body to these forces and moments. The
forces and moments are computed every CFD time step and transferred to a 6
DOF module which computes the body’s response to the forces and moments.
The response is converted into translational and rotational accelerations that are
integrated to obtain translational and rotational velocities and integrated once
more to obtain linear position and angular orientation. From the dynamic
response, the grid point locations and grid point velocities are set. Both CFD
and RBD computations are performed at every time step in a fully coupled
manner.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
6 Computational Ballistics III

φ
X
θ
ψ

Y Z

Figure 1: Rigid body dynamics schematic.

Figure 2: Finned configuration.

3 Results
Time-accurate unsteady numerical computations were performed using Navier–
Stokes and coupled 6 DOF methods to predict the flow field and aerodynamic
coefficients, and the flight paths of both a finned projectile at supersonic speeds
and a spinning projectile at subsonic speeds. In all cases, full 3D computations
were performed and no symmetry was used.
The supersonic projectile modeled in this study is an ogive-cylinder-finned
configuration (see Figure 2). The length of the projectile is 121 mm and the
diameter is 13 mm. The ogive nose is 98.6 mm long and the afterbody has a
22.3 mm, 2.5° boat-tail. Four fins are located on the back end of the projectile.
Each fin is 22.3 mm long and 10.16 mm thick. The computational mesh for the
25-mm projectile model is an unstructured one. In general, most of the grid
points are clustered in the afterbody fin region. The total number of grid points
is about 4 million for the full grid.
Here, the primary interest is in the development and application of coupled
CFD and RBD techniques for accurate simulation of the free-flight
aerodynamics and flight dynamics of the projectile in supersonic flight. The first
step here was to obtain the steady state results for this projectile at a given initial
supersonic velocity. Also imposed were the angular orientations at this stage.
Corresponding converged steady state solution was then used as the starting
condition along with the other initial conditions for the computation of coupled
CFD/RBD runs. Numerical computations have been made for these cases at an
initial velocity of 1034 m/s. The simulations were started a small distance away

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 7

from the muzzle. The corresponding initial angle of attack was, α = 4.9o and
initial spin rate was 2500 rad/s. Fig. 3 shows the computed pressure contours at
a given time or at a given location in the trajectory. It clearly shows the
orientation of the body at that instant in time and the resulting asymmetric flow
field due to the body at angle of attack. The orientation of the projectile of
course changes from one instant in time to another as the projectile flies down
range. Figure 4 shows the variation of the Euler pitch angle with distance
traveled. As seen in this figure, both the amplitude and frequency in the Euler
angle variation are predicted very well by the computed results and match
extremely well with the data from the flight tests. One can also clearly see the
amplitude damped out as the projectile flies down range i.e. with the increasing
x-distance. Although not shown here, similar behavior is observed with the
Euler yaw angle and it damps out with the increasing x-distance. The computed
results again compare very well with the measured data from the flight tests.
The time histories of the pitch and yaw angles are often customarily presented
as a motion plot where the pitch angle is plotted versus the yaw angle during the
flight of the projectile. It represents the path traversed by the nose of the
projectile during the flight trajectory (looking forward from the back of the
projectile). Such a plot is shown in Figure 5. This figure shows the comparison
of the motion plots obtained both from the numerical simulations and the 6 DOF
analysis of the flight results from ARFDAS [13]. Computed results match very
well with the experimental flight test results.
8.00

6.00
Experimental Data
CFD (2M grid)
CFD (4M grid)
4.00 CFD (6M grid)

2.00
T h eta, d eg

0.00
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0
-2.00

-4.00

-6.00

-8.00
Range, m

Figure 3: Computed pressure Figure 4: Euler pitch angle vs. range.

contours.

As stated earlier, coupled CFD/RBD simulations have been performed for a

spinning projectile at subsonic speeds as well. The subsonic projectile is a 1.8-
caliber ogive-cylinder configuration (see Figure 6). A computational grid
expanded near the vicinity of the projectile is shown in Figure 7. Grid points are
clustered near the jet as well as the boundary layer regions to capture the high
gradients flow regions. The computational grid has 211 points in the streamwise
direction, 241 in the circumferential direction, and 80 in the normal direction.
The unsteady simulations took thousands of hours of CPU time on the Xeon PC
cluster system at ARL MSRC running with either 32 to 64 processors.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
8 Computational Ballistics III

( ) (a) y
8

-2

-4

-6

-8
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12
Psi

(b)

Figure 5: Motion plot (a) computation, (b) flight test.

Lift

Microjet

Figure 6: Spinning projectile geometry.

The first step here was to obtain the steady state results for the same projectile
without the jet with the grid velocities prescribed. Also imposed were the
angular orientations at this stage. Corresponding converged jet-off steady state
solution was then used as the starting condition along with the other initial
conditions for the computation of coupled CFD/RBD runs. Synthetic jets were
activated at a specified point in the trajectory. The jet locations on the projectile
are shown in Figure 6. The jet conditions were specified at the exit of the jet
(sinusoidal variation in jet velocity). The jet conditions specified include the jet
pressure, density and velocity components. The flow field inside of the tiny jet
cavity is not computed. For the unsteady jets, time-dependent jet boundary
conditions are applied at the jet exit. Numerical computations have been made
for these jet cases at an initial Mach number, M = 0.39, initial angle of attack, α

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 9

= 2o, and an initial spin rate of 434 Hz. The jet width was 0.32 mm, the jet slot
half-angle was 18o, and the peak jet velocity used was 110 m/s operating at a
frequency of 1000 Hz. The jet-off coupled calculations are started from the same
initial conditions.

Y
Jet-on

Figure 7: Computational grid. Figure 8: Schematic showing the

jet actuation in a spin
cycle).

22000.00

18000.00
ARFDAS Fit
Experimental Data
14000.00 CFD
Ph i, d eg

10000.00

6000.00

2000.00

0.0 10.0 20.0 30.0 40.0 50.0 60.0

-2000.00
Range, m

Figure 9: Comparison of computed roll angle, jet-off.

Unsteady time-accurate CFD computations require huge computer resources.

All the results presented here were obtained from unsteady numerical
computations with a single synthetic jet on a 40–mm subsonic grenade (Figure
6). Looking from the back of the projectile (see Figure 8), the projectile spins
clockwise. The jet actuation corresponds to one-fourth of the spin cycle from -
45° to +45° with zero degree being the positive y-axis. The jet is off during the
remaining three-fourths of the spin cycle. The unsteady CFD modeling
technique required about 180 time steps to resolve a full spin cycle. The
unsteady synthetic jet operates at a high frequency of 1000 Hz. For the part of
the spin cycle when the jet is on, the jet operated for approximately four cycles.
Trajectories were computed for both jet-off and jet-on conditions. For the jet-on
case12, an unsteady synthetic jet was applied while the projectile is spinning and
the jet was activated at approximately 41 meters into the trajectory. Figure 9

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
10 Computational Ballistics III

shows the computed roll angles as a function of the x-distance or the range. The
roll angles shown here are the accumulated values to include 360 degrees for
every spin cycle of the projectile. Computed roll angle are compared with the
experimentally observed values as well as data obtained the 6 DOF analysis of
the flight results from ARFDAS [13]. and are found to be in good agreement.
8.00

7.00

6.00

5.00

4.00

3.00

2.00
Th eta-FP , d eg

1.00

0.00
0.0 10.0 20.0 30.0 40.0 50.0 60.0
-1.00

-2.00
-3.00

-4.00

-5.00
ARFDAS Fit
-6.00 Experimental Data
CFD
-7.00 CFD(SM-DT)
-8.00
Range, m

Figure 10: Computed Euler pitch angle, jet-off for different time steps.

Figure 10 shows the variation of the Euler pitch angle with distance traveled.
As seen in this figure, both the amplitude and frequency in the Euler pitch angle
variation are predicted very well by the computed results and match well with
the data from the flight tests for up to a distance of 20 meters. For the rest of the
trajectory, the frequency of the motion is predicted very well as seen by the
comparison with the data, but the amplitude doesn’t compare as well with the
data. Two sets of CFD results are shown in this figure, one obtained with the
original time step and another one with a smaller time step reduced by half. As
seen in this figure, reducing time-step by half has negligible effect on the
accuracy of the computed results. Further analysis and investigation are required
and are underway. Figure 11 shows similar behavior with Euler yaw angle with
x-distance. The agreement between the computed results and the experimental
data again is generally good except for the amplitude comparison at greater
ranges. In both figures 10 and 11, we have both slow and fast mode frequencies
and both compare well with the data and the ARFDAS fits. Figure 12 shows the
variation of the Euler pitch angle with distance traveled both for jet-off and jet-
on conditions. As seen in this figure, the effect of the jet is small and
incremental. The computed Euler angles are again compared with the
experimental range test data shown in dark circles for the jet-off case. As seen
here, the computed data predicts the frequency rather well; however, the
amplitudes in some parts are under-predicted in the computations.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 11

8.00

7.00 ARFDAS Fit

Experimental Data
6.00
CFD (SM_DT)
5.00 CFD
4.00

3.00

2.00
Psi-FP, deg

1.00

0.00
0.0 10.0 20.0 30.0 40.0 50.0 60.0
-1.00

-2.00

-3.00

-4.00

-5.00

-6.00

-7.00

-8.00
Range, m

Figure 11: Comparison of the computed Euler yaw angle, jet-off for different
time steps.
10.00

8.00

6.00

4.00
Theta (degree)

2.00

0.00
0.0 10.0 20.0 30.0 40.0 50.0 60.0
-2.00

-4.00

-6.00
Experimental Data
CFD - Jetoff
-8.00
CFD - Jeton

-10.00
Range (m)

Figure 12: Comparison of computed Euler pitch angle, jet-off and jet-on.

4 Concluding remarks
This paper describes a new coupled CFD/RBD computational study undertaken
to determine the flight aerodynamics of both a finned projectile and a spinning
projectile. A three-dimensional unsteady Navier–Stokes solver is employed to
compute the time-accurate aerodynamics associated with the free-flight of the
finned projectile at supersonic velocities and the spinning projectile at subsonic
speeds. Computed positions and orientations of the projectile have been
compared with actual data measured from free-flight tests and are found to be
generally in good agreement. Predicted aerodynamics forces and moments also
compare well with the forces and moments used in the six degree freedom fits of

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
12 Computational Ballistics III

the results of the same tests. This work demonstrates a coupled method to
accurately predict the time-accurate unsteady aerodynamics and the flight
trajectories of projectiles at various speeds. The present CFD/RBD simulations
clearly show the capability of the coupled approach and form the basis for future
multidisciplinary, time-dependent computations of advanced maneuvering
munitions.

References
[1] Sahu, J., K. R. Heavey, and E. N. Ferry, “Computational Fluid Dynamics
for Multiple Projectile Configurations.” Proceedings of the 3rd Overset
Composite Grid and Solution Technology Symposium, Los Alamos, NM,
October 1996.
[2] Sahu, J., K. R. Heavey, and C. J. Nietubicz, “Time-Dependent Navier–
Stokes Computations for Submunitions in Relative Motion.” 6th
International Symposium on Computational Fluid Dynamics, Lake Tahoe,
NV, September 1995.
[3] Meakin, R. L., “Computations of the Unsteady Flow about a Generic
Wing/Pylon/Finned-Store Configuration.” AIAA 92-4568-CP, August
1992.
[4] Smith B. L., and A. Glezer, “The Formation and Evolution of Synthetic
Jets.” Journal of Physics of Fluids, vol. 10, No. 9, September 1998.
[5] Davis, S. A. and Glezer, A., “The Manipulation of Large- and Small-
Scales in Coaxial Jets using Synthetic Jet Actuators”, AIAA Paper No.
2000-0403, January 2000.
[6] Avancha, R. and Pletcher, R. H., “Large Eddy Simulation of the Turbulent
Flow Past a Backward Facing Step”, AIAA paper No. 2000-0542, January
2000.
[7] Arunajatesan, S. and Sinha, N., “Towards Hybrid LES-RANS
Computations of Cavity Flowfields”, AIAA Paper No. 2000-0401,
January 2000.
[8] Batten, P., U. Goldberg and S. Chakravarthy, "Sub-grid Turbulence
Modeling for Unsteady Flow with Acoustic Resonance", AIAA Paper 00-
0473, 38th AIAA Aerospace Sciences Meeting, Reno, NV, January 2000.
[9] Peroomian, O., S. Chakravarthy, and U. Goldberg, “A ‘Grid-Transparent’
Methodology for CFD.” AIAA Paper 97-07245, 1997.
[10] Peroomian, O., S. Chakravarthy, S. Palaniswamy, and U. Goldberg,
“Convergence Acceleration for Unified-Grid Formulation Using
Preconditioned Implicit Relaxation.” AIAA Paper 98-0116, 1998.
[11] Goldberg, U. C., O. Peroomian, and S. Chakravarthy, “A Wall-Distance-
Free K-E Model with Enhanced Near-Wall Treatment.” ASME Journal of
Fluids Engineering, Vol. 120, pp. 457-462, 1998.
[12] Sahu, J., “Unsteady CFD Modeling of Aerodynamic Flow Control over a
Spinning Body with Synthetic Jet.” AIAA Paper 2004-0747, Reno, NV, 5-
8 January 2004.
[13] Arrow Tech Associates. “ARFDAS Technical Manual.” South
Burlington, VT, 2001.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 13

3D transient numerical simulation for sabot

separation
D. Touati & S. Peles
Israel Military Industries, Central Laboratory Division, P.O.B. 1044,
Ramat Hasharon, 47100, Israel

Abstract
This paper presents a sabot discard model to predict sabots separation. The
model uses the FSI (Fluid Structure Interaction) numerical packages of
MSC-DYTRAN to determine the evolution of sabot separation motion. The 3D
model enables us to calculate the asymmetric sabots discard trajectory using
Euler Lagrange coupling. We neglect the blast region effect on the sabot discards
process near the muzzle. Physical quantities such as linear and angular
displacement, velocities, acceleration and exerted forces on the sabot structures
are computed. Simulated results and experimental data show close agreement.
Keywords: sabots petals, discard, kinetic energy penetrator.

1 Introduction
The kinetic energy penetrator is a widely used anti tank munitions. Its lethality is
due to the kinetic energy imparted by the penetrator to the target with impact
velocities of between 1.4 to 1.8 km/s. The penetrator is launched by means of a
sabot assembly, consisting of three aluminum sabot petals, required to minimize
the penetrator’s in bore balloting. The process of sabot separation begins as the
projectile leaves the gun tube. Because of the transverse motion of the projectile
within the gun, energy is stored in the elastic sabot petals. As the projectile
leaves the muzzle, the constraints of the gun tube are released and the sabot
elements are able to move laterally outward. Aerodynamic forces acting on each
of the sabot petals causes them to lift up and disengage from the buttress grooves
of the penetrator rod to permit its unconstrained, low drag flight to the target. It
has been demonstrated by Schmidt and Shear [1] that aerodynamic interference
generated by the sabot components can be a significant source of projectile

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070021
14 Computational Ballistics III

launch disturbance leading to unacceptable loss of accuracy at the target.

Perturbations to the projectile trajectory are magnified by geometric asymmetry
in the discard pattern and by the extended periods during launch when the sabot
components are in close proximity to the projectile. Erengil [2] used analytical
expressions to estimate transient sabot surface pressure distributions, which are
then integrated numerically to yield force and moments to determine sabot
discard trajectories. Cayzac et al [3] investigated the launch dynamics of a 44
mm APFSDS. The 2D Euler equations have been used to predict sabot
separation in an intermediate ballistic flow field. Nusca [4] simulated a quasi
steady sabot discard process that was also tested in a wind tunnel. First he
computes the aerodynamic forces and moments exerted on the sabot petals at
specific positions, and then this loading was used to compute the dynamic
motion of the sabots. The quasi steady simulation ignores the flow time
dynamics and does not link the aerodynamic forces to sabot motion.
An insight into the transient sabots separation processes was provided in this
paper via a detailed finite element model, representing the sabot and penetrator
subjected to a free stream flow field. The model was prepared using exact
geometric part features, which include the buttress grooves details on both the
sabot and penetrator.
We used the FSI numerical feature of MSC – DYTRAN, and utilize an Euler
Lagrange coupling to correlate unsteady aerodynamic forces and the sabot body
motion.

Computational `
domain
(a) ` (b)

Figure 1: Slice of the finite elements model of the sabot and projectile rod
subjected to free stream flow field.

2 Numerical model
The configured numerical model included a stationary fins stabilized penetrator
at a zero angle of attack with respect to the free stream flow field, and three 1200
sector sabots petals encompassing the penetrator. The sabot discard simulations
are performed for a radial separation of the sabot from the projectile surface.
Based on the above assumption, the model utilizes only one half of the
computational domain as shown in figure 1a. The flow field characteristics were

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 15

calculated using the Euler equation and the angle of attack was used as the only
source of asymmetry in the free flow stream initial condition. A slice of the finite
elements model of sabot and penetrator, respectively, subjected to the free stream
flow field is shown in Figure 1b.
The numerical model neglected the blast region effect on the sabot discards
process near the muzzle based upon experimental results which revealed a short
range spatial influence of the blast region on the sabot discard process. The
discard process comes to an end at a distance of 15 m from the muzzle exit while
the blast region influence is limited to the close proximity of the muzzle at a
distance of 1 m. The sabots and the kinetic projectile rod were modeled using 3
– Dimensional 8 – node Hexagonal elements. The sabot and the penetrator were
assumed to be rigid.
The interface at the buttress grooves between the sabot and the penetrator is
modeled using contact features of the software to allow for realistic load transfer
from the sabot to the penetrator. The interface between the penetrator and sabot
buttress groove is shown in enlarged view in figure 2.

Figure 2: Enlarged view of the sabot and projectile kinetic rod buttress
groove interface.
The air surrounding the sabot and main rod structures is modeled using a
Eulerian mesh with cylindrical outer shape border. Air inflow and outflow
boundary conditions were applied and the air velocity at the boundary was the
projectile muzzle velocity which was kept constant throughout the sabot discard
simulation. An ideal gas equation of state was used to simulate the air behavior.
The dynamic air pressure interacted with the sabots structure by means of a fluid
– structure interacting algorithms.

3 Simulation results
Figure 3 shows the pressure contours of the air within close proximity of one of
the sabot petals.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
16 Computational Ballistics III

(a) (b)

(c)

Figure 3: (a) Shadowgraph photo of sabot discard experiment 3.5 ms after the
projectile exits from bore muzzle. (b) Simulation results of air
density plot in close proximity to the sabot region at 3.5 ms. (c)
Contour map of density plots (simulation results).

Figure 4: Contours of air pressure surrounding the sabot structure.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 17

The figure presents the sabot’s simulated and experimental attitudes with
respect to the projectile rod at 6 m beyond the muzzle exit. There is a close
resemblance between the simulated and experimental shock wave patterns.
Figure 4 presents air pressure contours surrounding the sabot structure at
different time frames during the discard process.
Figure 5 depicts the linear velocities components of the sabot’s c.g during the
discard process at (00 angle of attack).

Figure 5: Sabot C.G velocities components Vx (blue) and Vy (red).

Figure 6: Rotation angle vs. time of upper sabot at three different yaw angles.

Figure 6 shows the sabot’s angular displacement during the sabot discard
process for different Yaw angles of the free flow. The measured rotation angle
around the sabot’s c.g. point decreases as the Yaw angle increase. The angular

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
18 Computational Ballistics III

differences are due to Yaw angle effects which increase or decrease the petal
area which is exposed to the flow field, causing the aerodynamic lift and drag
force to discard the sabot at a slower or faster rate correspondingly. As may be
seen, the interaction time between the sabot and the kinetic projectile rod does
not exceed 2 ms for the specific sabot geometry and the given yaw angles.
Figure 7 shows the aerodynamics force components exerted on the sabot
during the discard process. The sabot separation process involves a continuous
growth of the sabot surface area exposed to the free stream flow which results in
an increase in the corresponding aerodynamic force magnitude in the x direction
from 1 KN up to almost 10 KN while the resultant aerodynamic force magnitude
in the y direction decreases.

Figure 7: Aerodynamic forces exerted on sabot structure.

4 Summary
An insight into the transient sabots separation processes was provided in this
paper via a detailed finite element model, representing the sabot and penetrator,
subjected to a free stream flow field. The model was prepared using exact
geometric part features, which include the buttress groove details on both the
sabot and penetrator.
It was written in the FSI numerical package of MSC - DYTRAN and utilizes
an Euler Lagrange coupling to correlate the unsteady aerodynamic forces and the
sabots’ rigid body motions. In retrospect we determined that the fluid Euler
equation are sufficient for describing the sabot’s entire spatial motion and the
general characteristics of the flow field over the sabot. The magnitude of the
exerted aerodynamic interfacial forces were assessed, and it was further revealed
that the interaction time between the sabot and the projectile rod does not exceed

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 19

2 ms for the specific sabot geometry. The most significant attribute of this model
is that it does not rely on any empirical data and is therefore widely applicable.
The simulation results are in close agreement with the experiments.

References
[1] Schmidt E.M., Shear D.D., Aerodynamic Interface During Sabot Discard,
Journal of Spacecraft and Rockets, AIAA, Vol 15, No 3, May-June, 1978,
pp. 235-240.
[2] Erengil M. E., Sabot Discard Model For Conventional and
Electromagnetic Launch Package, 19th International Symposium of
Ballistics 7-11 May 2001, pp 213-221 Interlaken, Switzerland.
[3] Cayzac R., Carette E., Alziary de Roquefort T. Intermediate Ballistics
Unsteady Sabot Separation: First Computations and Validation, 19th
International Symposium of Ballistics 7-11 May 2001, pp 297-305
Interlaken, Switzerland.
[4] Nusca M. J., Numerical Simulation of Sabot Discard Aerodynamics, Army
Research Laboratory ARL-TR-204 September 1993.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
This page intentionally left blank
 Computational Ballistics III 21

Fast calculation of direct fire trajectories

taking the earth’s rotation into account
W. Roetzel1, W. Czarnetzki2 & T. Maier2
1
Helmut-Schmidt-Universität / Universität der Bundeswehr Hamburg,
Germany
2
Hochschule Esslingen, Germany

Abstract
An analytical approach is developed for the subsequent consideration of the
Coriolis effect. The target is regarded as a moving target in the star-fixed
coordinate system. The stationary atmosphere creates a nonuniform cross-wind
which reduces the normal moving target deflection of the projectile. The
approach is implemented in a previously developed analytical fast calculation
method and tested against numerical calculations with good results.
Keyword: Coriolis effect, analytical solution, power law, drag coefficient, Mach
number.

1 Introduction
Several analytical solutions of the point mass equation of motion have been
developed for the fast calculation of direct fire trajectories [1–4], which are
based on the power law

m m m/2
a  a*   T  (1)
cD = C ⋅ (Ma )
−m
= C ⋅   = C ⋅   ⋅  * 
v  v  T 

for the drag coefficient variation with Mach number. McCoy [1] published flat
fire solutions for m= 0, ½, 1. The later solutions [2–4] allow for arbitrary values
of m and any angle of sight β [3, 4]. Wind can be considered using a coordinate
transformation [2, 4]. For uphill and downhill firing the change of pressure and
temperature along the trajectory should be considered. Usually mean values of

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070031
22 Computational Ballistics III

pressure and temperature are applied [3]. A more sophisticated way [4] is the
correction of the Mach number exponent, which takes not only the changing
pressure and temperature but also the curvature of the trajectory approximately
into consideration.
The above mentioned fast calculation methods do not allow for the earth’s
rotation. In the present paper a simple analytical method is described [5] with
which the Coriolis effect can subsequently be considered. The same coordinate
system and nomenclature is used as in the foregoing paper [4].

2 Numerical method
For the precise calculation of the trajectory the complete equation of motion has
to be integrated numerically. With the previously [2,4] defined ballistic
coefficient
1− m / 2
π d² π d² p T* 
ρ * (a * ) ⋅ (2)
m
D ( p, T ) = C ⋅ ρ am = C ⋅ ⋅ 
8M 8M p *  T 

G
the equation of motion with wind velocity w can be written as [1]
G G G G G G G
v = g − D ⋅ ( v − w ) ⋅ (v − w) − b (3)
1− m

with the Coriolis acceleration

G G G
b = −2ω × v (4)
G
containing the earth’s vector ω of rotation

 cosψ cos γ 
G  . (5)
ω =ω sinψ 
− cosψ sin γ 
 

In eqn (5) the angle ψ is the latitude and the angle γ the azimuth of fire (x-axis),
measured clockwise from north. Assuming the standard temperature drop with
altitude ∆T/∆y = -0.0065 k/m [6] and regarding the atmosphere as perfect dry air
yields the ballistic coefficient D as function of height y
4.25593+ m / 2
 0.0065  , (6)
D = D0 ⋅ 1 − ⋅ y 
 T0 

where index “0” indicates the firing site and origin of the earth-fixes coordinate
system.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 23
G
Starting at the origin with τ = 0 and L = {x, y, z} = 0 , the local projectile
G
velocity v can stepwise be calculated as function of time. Integrating
G
simultaneously over the time yields L as function of time. This numerical
calculation is later carried out in order to test the new approach which is derived
and described in the following.

3 Analytical approach
The equation of motion without the Coriolis acceleration is valid in a star-fixed
not rotating system which moves with uniform velocity. Therefore the coordinate
system in which the Coriolis-free solutions [1–4] are valid is considered to move
G
on uniformly with the fixed circumferential (not rotating) velocity u0 of the
firing site (origin of coordinate system) at the instant of firing τ = 0 . For τ = 0
this star-fixed coordinate system coincides with the earth-fixed system, for τ > 0
they separate from each other.

3.1 The moving centre of gravity

In this star-fixed system a flying projectile experiences an additional

G
gravitational acceleration ∆g in the horizontal direction, as the centre of gravity
does not remain exactly perpendicular below the projectile. With the simplifying
assumption of constant mean velocity components vx and vz and the radius R =
6.37·106m of the globe, the additional time dependent acceleration can be
expressed as

u x + v x  − ω z ⋅ R + x1 / τ 
G g ⋅τ   g ⋅τ  . (7)
∆g = −  0 =−  0 
R   R  ω ⋅ R + z /τ 
u z + v z   x 1 

Substituting ωx and ωz according to eqn (5) and integrating twice over the time of
flight yields the displacement of the projectile

ω cosψ sin γ + x1 / Rτ 1 
G g 3 . (8)
∆L1g = − ⋅ τ 1  0 
6 ω cosψ cos γ + z / Rτ 
 1 1

The vertical components ∆gy = 0 and ∆L1gy = 0 since cos ωτ 1 ≈ 1 . The derivation
of eqn (8) is correct under vacuum conditions. The effect of eqn (8) is very week
in normal cases and a sufficiently accurate approach.

3.2 The moving target

The main effect of the earth’s rotation is the fact that in the uniformly moving
star-fixed system the earth-fixed target appears as a moving target. We look first
at the poles of the globe. There the circumferential velocity is zero, but the earth-

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
24 Computational Ballistics III

fixed coordinate system rotates relatively to the considered star-fixed system. At

the north pole any target will travel from west to east, i.e. from right to left.
A negative angle of lead α < 0 would be required to hit the moving target. The
G
shift ∆L1ω of the target during the time of flight τ 1 can be expressed as

  G 
G
[ (
G G G
∆L1ω = τ 1 ω × L1 + ∆L1g ) ] =τ (ωG × L ) + ωG × ∆L
1
G
1 1g 


(9)
  ≈0 

which leads with eqn (5) to

 z1 sinψ + y1 cosψ sin γ 

G  . (10)
∆L1ω = ω ⋅ τ 1 − x1 cosψ sin γ − z1 cosψ cos γ 
 y cosψ cos γ − x sinψ 
 1 1 

At other latitudes ψ ≠ ±90° an additional target movement takes place in the

star-fixed system which is caused by the growing distance between the origins of
both coordinate systems. However, a detailed analysis shows that this additional
shift can be omitted. Its effect is compensated by the fact that in the usual
standard free-fall acceleration g = 9.80665 m/s2 (at sea level) the centrifugal
acceleration due to the earth’s rotation is included (subtracted from the mass
attraction force). Thus, eqn (10) is valid for all values − 90° < ψ < +90° .
The same shift (10) with negative sign is found by the twofold integration of
the Coriolis acceleration eqns (4,5) using a constant mean projectile velocity
G G
v = L1 / τ 1 . The negative sign shows that in the earth-fixed system the projectile is
apparently accelerated and displaced in the opposite direction as the target is in
the star-fixed system. One could suppose that the final Coriolis deflection of the
projectile in the earth-fixed system could be expressed as
G G G
∆L1 = ∆L1g − ∆L1ω . (11)

However, this is only a rough approximation under normal shooting conditions

which will be discussed later in this paper. Eq (11) holds true in a vacuum. The
G
contribution ∆L1g is normally relatively small. In the special case of the
G G
(
downward free-fall in a vacuum at the equator ψ = 0, L1 = {0,− L1 ,0} ∆L1g is )
decisive. For γ = 0 and with the time of flight τ 1 = (2 L1 / g ) 1/ 2
eqn (11) yields the
correct [7, p.214] east drift
1/ 2 1/ 2
 2L   1   2L  2
∆L1z = ω ⋅ L1 ⋅  1  ⋅  − + 1 = ω ⋅ L1 ⋅  1  ⋅ . (12)
 g   3   g  3
G
The term –1/3 is due to ∆L1g . Neglecting it would cause an overestimation of
50%.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 25

3.3 The Coriolis wind

The moving target problem under consideration differs from the usual case in so
far as the air follows the moving target. If no wind is present, the stationary air in
the earth-fixed system rotates in the star-fixed system together with the target.
The moving target gets tailwind of equal velocity. The wind is not uniformly
G
distributed but the cross-wind velocity wc is zero at the firing site and grows
G
linearly with the distance to its maximum value w1 at the target
G
G x G x − ∆L1
wc = ⋅ w 1 = ⋅ . (13)
x1 x1 τ 1

This Coriolis wind pushes the projectile towards the moving target and reduces
the deflection according to eqn (11). Since the consideration of uniform wind is
no problem, a suitable mean uniform Coriolis wind velocity wG cm is defined and
derived which produces the same effect as the actual linearly growing Coriolis
wind does:
G G G
G G ∆L1 ∆L1ω − ∆L1g . (14)
w cm = ⋅ w 1 = − ⋅
f f == ⋅
f
τ1 τ1

A horizontal flat shot in the x-direction (vx = v) towards a target at distance x1

with cross-wind of velocity wz = wc is considered. Gravity is neglected. For
wZ << v

v  x 
v z = −  w1 ⋅ − v z  . (15)
v x1 

From eqns (1–3) one can derive [2]

vm 
[ ] 
m
x
1 − 1 + D(1 − m)v0 ⋅ τ
1− m
= 0 1− m
 (16)
x1 Dmx1  

and
v − Dv 10− m . (17)
=
v 1 + D (1 − m)v 10− m ⋅ τ

Substituting eqns (16) and (17) into eqn (15) yields a first order linear
inhomogeneous differential equation for vz (τ ) . Solving first the homogeneous
differential equation and applying the variation of the constant yields with the
boundary condition τ = 0 : vz = 0 the projectile velocity component vz as
function of time. Integrating vZ from τ = 0 to τ 1 gives the deflection z1 of the
projectile at the distance x1. The same deflection z1 has to be produced by the

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
26 Computational Ballistics III

uniform wind of velocity wcm, defined with eqn (14). This deflection can be
found by replacing in eqn (15) the variable velocity w1 ·x/x1 by the constant
velocity f·w1. The solution leads to the known formula of Didion [1]
f ⋅ w1
z1 = (v0 ⋅ τ 1 − x1 ) . (18)
v0

Equating both deflections z1 and solving for f yields finally the implicit formulas
for the determination of the correction factor f:

 m 
V+
V  ,
1− 2 m

f (V , m) = ⋅ 1 − m + V − V [1 + (1 − m)η ] 1− m 
(19)
V − 1  m ⋅η m(1 − m)(1 − 2m) ⋅ η 2 
 
 

m ⋅η
V = m
. (20)
1 − [1 + (1 − m)η ]m −1

For given values of m and guessed values of η the dimensionless number V and
the factor f can be calculated. Iteratively one can determine f as function of m and
V. The number V is defined as
v x0 ⋅ τ 1 τ
V= = 1 ≥ 1. (21)
x1 τ 1, p =0

It is the ratio of the horizontal components of the muzzle velocity and the mean
velocity. It can be determined once the trajectory has been calculated
numerically or analytically (neglecting the Coriolis effect). In a vacuum V → 1.
The right-hand definition as the ratio of the times of flight under normal and
vacuum conditions is more appropriate when both vx0 and x1 turn to zero. (e.g.
free-fall in a vacuum).
In the special case m = ½ the variable η in eqns (19, 20) can be eliminated
yielding the explicit formula
2
 V   V + 1 2 ⋅ ln V  . (22)
f m =1 / 2 =   ⋅ − 
 V −1  V V −1 

In the special cases m = 0 and m = 1 the eqns (19, 20) cannot be applied directly
and the following implicit equations have to be used

V  2(V − 1)  λ , (23)
f m=0 = ⋅ 1 − , V =
V −1  λ  ln(1 + λ )

V  2(V − 1)  µ . (24)
f m=1 = ⋅ − 1 , V =
V − 1  µ  1 − e −µ

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 27

For m ≠ ½ the determination of f is inconvenient and time consuming. Therefore

the following empirical explicit formula has been developed:

1 2
= 1 + E , E = 0.48(1 + m) + 0.08m . (25)
2

f V

For the limiting value V → 1 the exponent E = (l +m)/2 is more precise. Equation
(25) is sufficiently accurate for all direct fire applications.
Once the correction factor f is found, the Coriolis deflection of a projectile
G
can subsequently be calculated as follows: First ∆L1 is calculated from eqs (8,
G
10, 11). This vector represents the deflection of the projectile from the target L1
G
neglecting the Coriolis wind. The mean Coriolis wind velocity wcm is calculated
using eqn (14) with f from eqn (25). If a real uniform wind is present, the
Coriolis wind velocity has simply to be added to the real wind velocity. The sum
of both winds can be taken into account by a coordinate transformation [2, 4] or
other known methods.
If no real wind is present the application of Didion´s formula is
G
recommended. Combining the shift ∆L1 and the Coriolis wind effect according
to Didion leads to the following formula for the Coriolis drift of the projectile

G G 2 + V E −1
∆L1c = ∆L1 ⋅ . (26)
2 +V E

A horizontal flat shot has been assumed in the above derivations. For uphill and
downhill fire the exponent m in eqs. (19–26) has to be corrected for changing air
pressure and temperature along the trajectory [4]. The exponent m has to be
replaced by m* = m – n with n according to [4, eq. (26)]:

v 0m sin β  g (1 − m ) (0.0065)( 4.25593 + m / 2)  . (27)

n=  2
+ 
D0  v0 T0 

4 Test of the analytical approach against numerical

calculations

The analytical Coriolis approach has been implemented into a fast calculation
programme, based on the previously developed explicit Coriolis-free analytical
solution [4, chapter 3.2 and 4]. The programme is constructed in such a way that
for a given target (x1, y1) or (L1, β) and wind velocity (wx, wy, wz) the angle of
lead and the gun elevation angle above line of sight is calculated iteratively. The
extended programme has been tested against numerical calculations described in
chapter 2 of this paper. Four typical examples are presented in the following.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
28 Computational Ballistics III

4.1 Example 1

The 185 Grain match projectile Lapua Scenar .30 GB 423, d = 7.8232 mm, M =
11.988 g is considered with m = 0.4855 and C = 0.4286 (eq. (1)). The projectile
is fired with v0 = 930 m/s at sea level, p0 = 1013.25 mbar and t0 = 15° C. The
angular velocity of the earth is taken as ω = 7.292 . 10-5 rad/s [1]. The muzzle
Mach number Ma0 = 2.73. It reduces to Ma1 = 1.25 (in all examples) after the
fixed time of flight τ1 = 1.6 s. Table 1 shows calculated Coriolis deflections for
flat fire from the north pole and the equator. The same results are obtained (with
millimeter accuracy) from the analytical and the numerical method. The
analytical deflections are calculated using eq. (26) yielding the same values as
the fast calculation programme.

4.2 Example 2

The same projectile and data are used as in example 1 with the exception of the
gun elevation angle φ0 = 46° and the latitude ψ = 45°. Table 2 shows the results.
The values in brackets are analytically calculated deflections which deviate from
the corresponding numerical values. The largest deviation is 1 mm in height at
the distance (bee-line) of about 1 km.

4.3 Example 3

A projectile of identical shape (equal values of m and C) and mass density is

considered, which has the threefold diameter d = 23.47 mm. The mass is
enlarged by the factor 27, yielding M = 323.68 g. From the analytical solution [4]
one can see that with the same muzzle velocity and threefold time of flight τ 1 =
4,8 s the end velocity will remain the same at the threefold distance. So Ma =
2.73 to 1.25. One can also predict that for flat fire the elevation angle above line
of sight will increase by the factor 3. Table 3 shows the flat fire results for φ0 =
2°, τ 1 = 4.8 s and L1 = 3019.26 m. The largest deviations are 3 mm.
Concerning the superelevation angles: For example 1 the Coriolis-free
superelevation angle ε1 = 9.7866 mils and for example 3 ε3 = 29.3598 mils =
3 . ε1.

Table 1: Flat fire vertical ( ∆ y) and horizontal ( ∆ z) Coriolis deflections,

calculated numerically and analytically. Projectile Lapua Scenar
.30 GB 432, 185 Grain. Gun elevation angle φ0 = 1°. Coriolis-free
range x1 = 1006.2 m, hitpoint height y1 = 7.71 m, distance L1 =
1006.2 m. Muzzle velocity v0 = 930 m/s, time of flight τ 1 = 1.6 s.

ψ(°) 90 0 0 0 0
γ (°) 0÷360 0 90 180 270
∆y (mm) 0 0 +102 0 -102
∆z (mm) +102 -1 0 +1 0

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 29

Table 2: Uphill fire Coriolis deflections, calculated numerically and

analytically (deviating analytical results in brackets). Projectile
Lapua of Table 1, φ0 = 46°, x1 = 705.45 m, y1 = 720.58 m, L1 =
1008,41 m, v0 = 930 m/s, τ 1 = 1.6 s.

ψ(°) 90 45 45 45 45
γ (°) 0÷360 0 90 180 270
∆y (mm) 0 0 +51 0 -52 (-51)
∆z (mm) +72 -1 +51 +103 +51

Table 3: Flat fire Coriolis deflections, calculated numerically and

analytically (deviating analytical results in brackets). Projectile of
Lapua geometry and mass density but threefold diameter (d = 23.47
mm, M = 323.68 g), φ0 = 2°, x1 = 3019.22 m, y1 = 16.75 m, L1 =
3019.26 m, v0 = 930 m/s, τ 1 = 4.8 s.

ψ(°) 90 0 0 0 0
γ (°) 0÷360 0 90 180 270
∆y (mm) 0 0 +920 0 -920
∆z (mm) +920 -14(-17) 0 +14(+17) 0

Table 4: Uphill fire Coriolis deflections, calculated numerically and

analytically (deviating analytical results in brackets). Projectile of
Table 3, φ0 = 47°, x1 = 2116.19 m, y1 = 2178.37 m, L1 = 3037.03 m,
v0 = 930 m/s, τ 1 = 4.8 s.

ψ(°) 90 45 45 45 45
γ (°) 0÷360 0 90 180 270
∆y (mm) 0 0 +469(+463) 0 -469(-463)
∆z (mm) +655 -20(-22) +463 +947(+948) +463

4.4 Example 4

The same data are used as in example 3, however the elevation angle changes to
φ0 = 47° and the latitude to ψ = 45°. The Coriolis-free distance is now L1 =
3037.03 m, which is slightly longer than in example 3, although gravity pulls the
projectile back. The reason is the decreasing air pressure along the trajectory.
Table 4 again shows very good agreement. The largest error occurs in height
with 6 mm at a distance of about 3 km. As mentioned with eq. (26), in the
analytical approach the corrected exponent m has to be used. In this particular
case m = 0.4855, n = 0.3574 (eq. (27)) and the corrected exponent m* = 0.1281.

5 Conclusions
The developed analytical approach yields sufficiently accurate vertical and
horizontal Coriolis deflections and represents a simple and useful extension of

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
30 Computational Ballistics III

the previously developed [4] analytical fast calculation method for direct fire
applications.

References
[1] McCoy, R.L., Modern Exterior Ballistics / The Launch and Flight
Dynamics of Symmetric Projectiles, Schiffer Publishing Ltd: Atglen, PA
19310, 1999.
[2] Roetzel, W., Analytische Berechnung gestreckter Geschossflugbahnen.
Mitteilungen der Schießsport-Arbeitsgemeinschaft an der Universität der
Bundeswehr Hamburg Nr. 1, Ed. H. Rothe, Helmut-Schmidt-Universität,
Universität der Bundeswehr Hamburg, 2004.
[3] Kuhrt, A., Rothe H., The use of computer algebra and nonlinear
optimization for realtime computation of fire orders for direct fire,
Computational Ballistics II, pp. 337 -346, Eds. V. Sanchez-Galvez, C.A.
Brebbia, A.A. Motta, C.E. Anderson, WIT Press 2005.
[4] Roetzel, W., Analytical calculation of trajectories using a power law for
the drag coefficient variation with Mach number. Computational Ballistics
II, pp. 303 – 312, Eds. V. Sanches-Golvez, C.A. Brebbia, A.A. Motta,
C.E. Anderson, WIT Press 2005.
[5] Roetzel, W., Verfahren zur nachträglichen Berücksichtigung von Coriolis-
Effekten bei der Bestimmung der Flugbahnen von Geschossen,
unpublished report, Hamburg, 2005.
[6] Rogers, G.F.C., Mayhew, Y.R., Thermodynamic and Transport Properties
of Fluids, Basil Blackwell Ltd. Oxford, 4th Edition, 1988.
[7] Szabó, I., Einführung in die Technische Mechanik, 3. Auflage, Springer-
Verlag, Berlin / Göttingen / Heidelberg, 1958.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 31

Generating an aerodynamic model for

projectile flight simulation using unsteady time
accurate computational fluid dynamic results
J. Kokes1, M. Costello2 & J. Sahu3
1
Orbital Sciences Corporation, USA
2
Georgia Institute of Technology, USA
3
U.S. Army Research Laboratory, USA

Abstract
A method to efficiently generate a complete aerodynamic description for
projectile flight dynamic modeling is described. At the core of the method is an
unsteady, time accurate computational fluid dynamics simulation that is tightly
coupled to a rigid body dynamics simulation. A set of n short time snippets of
simulated projectile motion at m different Mach numbers is computed and
employed as baseline data. For each time snippet, aerodynamic forces and
moments and the full rigid body state vector of the projectile are known. With
time synchronized air loads and state vector information, aerodynamic
coefficients can be estimated with a simple fitting procedure. By inspecting the
condition number of the fitting matrix, it is straight forward to assess the
suitability of the time history data to predict a selected set of aerodynamic
coefficients. To highlight the merits of this technique it is exercised on example
data for a fin stabilized projectile. The technique is further exercised for a fin
and spin stabilized projectile using simulated data from a standard trajectory
code.
Keywords: projectile, flight dynamics, aerodynamic coefficients, CFD.

1 Introduction
There are roughly four classes of techniques to predict aerodynamic forces and
moments on a projectile in atmospheric flight: empirical methods, wind tunnel
testing, computational fluid dynamics simulation, and spark range testing.
Empirical techniques aerodynamically describe the projectile with a set of

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070041
32 Computational Ballistics III

geometric properties (diameter, number of fins, nose type, nose radius, etc) and
catalog aerodynamic coefficients of many different projectiles as a function of
these features. The database of aerodynamic coefficients as a function of
projectile features is typically obtained from wind tunnel or spark range tests.
This data is fit to multivariable equations to create generic models for
aerodynamic coefficients as a function of these basic projectile geometric
properties. Examples of this approach to projectile aerodynamic coefficient
estimation include Missile DATCOM, PRODAS, and AP98 [1–6]. The
advantage of this technique is that it is a general method applicable to any
projectile. However, it is the least accurate method of the four methods
mentioned above, particularly for new configurations that fall outside the realm
of projectiles used to form the basic aerodynamic database. The empirical
method has been found very useful in conceptual design of projectiles where
rapid and inexpensive estimates of aerodynamic coefficients are needed. In wind
tunnel testing, a specific projectile is mounted in a wind tunnel at various angles
of attack with aerodynamic forces and moments measured at various Mach
numbers using a sting balance. Wind tunnel testing has the obvious advantage of
being based on direct measurement of aerodynamic forces and moments on the
projectile. It is also relatively easy to change the wind tunnel model to allow
detailed parametric effects to be investigated. The main disadvantage to wind
tunnel testing is that it requires a wind tunnel and as such is modestly expensive.
Furthermore, dynamic derivatives such as pitch and roll damping as well as
Magnus force and moment coefficients are difficult to obtain in a wind tunnel
and require a complex physical wind tunnel model. Wind tunnel testing is often
used during projectile development programs to converge on fine details of the
aerodynamic design of the shell [7,8]. In computational fluid dynamics (CFD)
simulation, the fundamental fluid dynamic equations are numerically solved for a
specific configuration. The most sophisticated computer codes are capable of
unsteady time accurate computations using the Navier–Stokes equations.
Examples of these tools include, for example, CFD++, Fluent, and Overflow-D.
Over the past couple of decades, tremendous strides have been made in the
application of CFD to prediction of aerodynamic loads on air vehicles, including
projectiles. CFD is based on first principles and does not involve physical
testing. It is a general method that is valid for any projectile configuration.
However, CFD is computationally expensive and requires powerful computers to
obtain results in a reasonably timely manner [9–22]. In spark range aerodynamic
testing, a projectile is fired through an enclosed building. At a discrete number
of points during the flight of the projectile (< 30) the state of the projectile is
measured using spark shadowgraphs [23–27]. The projectile state data is
subsequently fit to a rigid 6 degree-of-freedom projectile model using the
aerodynamic coefficients as the fitting parameters [28–30]. Spark range
aerodynamic testing is considered the gold standard for projectile aerodynamic
coefficient estimation. It is the most accurate method for obtaining aerodynamic
data on a specific projectile configuration. It usually the most expensive
alternative, requires a spark range facility, and strictly speaking is only valid for
the specific projectile configuration tested.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 33

Various researchers have used CFD to estimate aerodynamic coefficient

estimation of projectiles. Early work focused on Euler solvers applied to steady
flow problems while more recent work has solved the Reynolds averaged
Navier–Stokes equations and Large Eddy Simulation Navier Stokes equations
for both steady and unsteady conditions [9–22]. For example, to predict pitch
damping Weinacht prescribed projectile motion to mimic a typical pitch
damping wind tunnel test in a CFD simulation to estimate the different
components of the pitch damping coefficient of a fin stabilized projectile [31].
Excellent agreement between computed and measured pitch damping was
attained. Algorithm and computing advances have also led to coupling of CFD
codes to projectile rigid body dynamics codes for simulation of free flight
motion of a projectile in a time accurate manner. Aerodynamic forces and
moments are computed with the computational fluid dynamics solver while the
free flight motion of the projectile is computed by integrating the rigid body
dynamic (RBD) equations of motion. The ability to simulate the flight of a
projectile using first principles has led to the notion of “virtual fly outs” where
the simulation tools above are used to replicate a spark range test. Along these
lines, Sahu achieved excellent agreement between spark range measurements and
a coupled CFD/RBD approach for a finned stabilized projectile [32]. Projectile
position and orientation at down range locations consistent with a spark range
test were extracted from the output of the CFD/RBD software to compute
aerodynamic coefficients. Standard range reduction software was utilized for
this purpose with good agreement obtained when contrasted against example
spark range results.
While coupled CFD/RBD simulation is now capable of replicating time
accurate projectile motion, computing time for this type of analysis is
exceedingly high and does not currently represent a practical method for typical
flight dynamic analysis such as impact point statistics (CEP) computation where
thousands of fly outs are required. Furthermore, this type of analysis does not
allow the same level of understanding of the inherent underlying dynamics of the
system that rigid body dynamic analysis using aerodynamic coefficients yields.
However, the coupled CFD/RBD approach does offer an ideal way to rapidly
compute the aerodynamic coefficients needed for rigid 6 degree-of-freedom
simulation. During a time accurate CFD/RBD simulation, aerodynamic forces
and moments and the full rigid body state vector of the projectile are generated at
each time step in the simulation. This means that aerodynamic forces,
aerodynamic moments, position of the mass center, body orientation,
translational velocity, and angular velocity of the projectile are all known at the
same time instant. With time synchronized air load and state vector information,
the aerodynamic coefficients can be estimated with a simple fitting procedure.
This paper creates a method to efficiently generate a complete aerodynamic
model for a projectile in atmospheric flight using n short time histories at m
different Mach numbers with an industry standard time accurate CFD/RBD
simulation. The technique is exercised on example CFD/RBD data for a small
fin stabilized projectile. The technique is further exercised for a fin and spin
stabilized projectile using simulated data from a standard trajectory code.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
34 Computational Ballistics III

Parametric trade studies investigating the number of time snippets and the length
of each time snippet to obtain accurate aerodynamic coefficients are reported.

2 Projectile CFD/RBD simulation

The projectile CFD/RBD algorithm employed here combines a rigid six degree
of freedom projectile flight dynamic model with a three dimensional, time
accurate CFD simulation. The RBD dynamic equations are integrated forward in
time where aerodynamic forces and moments that drive motion of the projectile
are computed using the CFD algorithm.
The RBD projectile model allows for 3 translation degrees of freedom and 3
rotation degrees of freedom. As shown in Figure 1, the I frame is attached to
G
the ground while the B frame is fixed to the projectile with the I B axis
G G
pointing out the nose of the projectile and the J B and K B unit vectors forming
a right handed triad. The projectile state vector is comprised of the inertial
position components of the projectile mass center ( x, y, z ), the standard
aerospace sequence Euler angles ( φ , θ ,ψ ), the body frame components of the
projectile mass center velocity ( u , v, w ), and the body frame components of the
projectile angular velocity vector ( p, q, r ). Both the translational and rotational
dynamic equations are expressed in the projectile body reference frame. The
standard rigid projectile, body frame equations of motion are given by Equations
1 through 4.

 x  cθ cψ sφ sθ cψ − cφ sψ cφ sθ cψ + sφ sψ   u 
    
 y  =  cθ sψ sφ sθ sψ + cφ cψ cφ sθ sψ − sφ cψ   v  (1)
 z   − s sφ cθ cφ cθ  w 
   θ

φ  1 sφ tθ cφ tθ   p 
    
θ  = 0 cφ − sφ   q  (2)
ψ  0 s / c cφ / cθ   r 
   φ θ

 u   X / m   0 − r q   u 
      
 v  =  Y / m  −  r 0 − p  v  (3)

 w   Z / m   − q p 0   w 
    

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 35

 p    L   0 − r q   p 
  −1      
 q  = [I ]   M  −  r 0 − p [I ] q  (4)

 r    N   − q p 0   r 
 
Note that the total applied force components ( X , Y , Z ) and moment
components ( L, M , N ) contain contributions from weight and aerodynamics.
The aerodynamic portion of the applied loads in Equations 3 and 4 is computed
using the CFD simulation and passed to the rigid body dynamic simulation.

Figure 1: Reference frame and position definitions.

On the other hand, the CFD flow equations are integrated forward in time
where the motion of the projectile that drives flow dynamics are computed using
the RBD algorithm. The complete set of 3-D time-dependent Navier–Stokes
equations is solved in a time-accurate manner for simulation of free flight. The
commercially available code, CFD++17-20, is used for the time-accurate
unsteady CFD simulations. The basic numerical framework in the code contains
unified-grid, unified-physics, and unified-computing features. The 3-D, time-
dependent Reynolds-averaged Navier–Stokes (RANS) equations are solved
using the following finite volume equation.

∂
WdV + v∫ ( F − G ) dA = ∫ HdV
∂t V∫
(5)
V

where W is the vector of conservative variables, F and G are the inviscid and
viscous flux vectors, respectively, H is the vector of source terms, V is the cell

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
36 Computational Ballistics III

volume, and A is the surface area of the cell face. A second-order discretization
is used for the flow variables and the turbulent viscosity equation. The
turbulence closure is based on topology-parameter-free formulations. Two-
equation higher order RANS turbulence models are used for the computation of
turbulent flows. These models are ideally suited to unstructured book-keeping
and massively parallel processing due to their independence from constraints
related to the placement of boundaries and/or zonal interfaces.
A dual time-stepping approach is used to integrate the flow equations to
achieve the desired time-accuracy. The first is an “outer” or global (and
physical) time step that corresponds to the time discretization of the physical
time variation term. This time step can be chosen directly by the user and is
typically set to a value to represent 1/100 of the period of oscillation expected or
forced in the transient flow. It is also applied to every cell and is not spatially
varying. An artificial or “inner” or “local” time variation term is added to the
basic physical equations. This time step and corresponding “inner-iteration”
strategy is chosen to help satisfy the physical transient equations to the desired
degree. For the inner iterations, the time step is allowed to vary spatially. Also,
relaxation with multigrid (algebraic) acceleration is employed to reduce the
residues of the physical transient equations. It is found that an order of
magnitude reduction in the residues is usually sufficient to produce a good
transient iteration.
The projectile in the coupled CFD/RBD simulation along with its grid moves
and rotates as the projectile flies downrange. Grid velocity is assigned to each
mesh point. For a spinning and yawing projectile, the grid speeds are assigned
as if the grid is attached to the projectile and spinning and yawing with it.
In order to properly initialize the CFD simulation, two modes of operation for
the CFD code are utilized, namely, an uncoupled and a coupled mode. The
uncoupled mode is used to initialize the CFD flow solution while the coupled
mode represents the time accurate coupled CFD/RBD solution. In the uncoupled
mode, the rigid body dynamics are specified. The uncoupled mode begins with a
computation performed in “steady state mode” with the grid velocities prescribed
to account for the proper initial position ( x0 , y0 , z0 ), orientation ( φ0 ,θ 0 ,ψ 0 ),
and translational velocity ( u0 , v0 , w0 ) components of the complete set of initial
conditions to be prescribed. After the quasi-steady state solution is converged,
the initial spin rate ( p0 ) is included and a new quasi-steady state solution is
obtained. A sufficient number of time steps are performed so that the angular
orientation for the spin axis corresponds to the prescribed initial conditions. This
steady state flow solution is the starting point for the coupled solution. For the
coupled solution, the mesh is translated back to the desired initial position
( x0 , y0 , z0 ) and the remaining angular velocity initial conditions ( q0 , r0 ) are
then added. In the coupled mode, the aerodynamic forces and moments are
passed to the RBD simulation which propagates the rigid state of the projectile
forward in time.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 37

Figure 2: Projectile orientation definitions.

3 Flight dynamics projectile aerodynamic model

The applied loads in Equations 3 and 4 are expressed in the body reference frame
and split into contributions due to weight and body aerodynamic force as shown
below.

 
C X 0 + C X 2 (v 2 + w2 ) / V 2 
X   − sθ   
    π 2 2  pD  (6)
 Y  = W  sφ cθ  − ρV D C NAv / V − CYPA w / V 
Z  c c  8  2V 
   φ θ  pD 
C NA w / V + 2V CYPAv / V 

The air loads can be further split into a steady air loads component that acts at
the center of pressure and a Magnus air loads component that acts at the center of
Magnus. The terms containing CYPA constitute the Magnus air loads component
while the terms containing C X 0 , C X 2 , C NA define the loads acting at the center
of pressure. The externally applied moment about the projectile mass center is
composed of an unsteady aerodynamic moment along with terms due to the fact
that the center of pressure and center of Magnus are not located at the mass
center.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
38 Computational Ballistics III

 pD 
 CLDD + CLP 
L 2V
 
  π 2 3 w qD pD v 
 M  = ρV D  CMA + CMQ + CNPA  (7)
  8  V 2V 2V V 
N  v rD pD w
−CMA V + 2V CMQ + 2V CNPA V 
 

The terms involving CMA accounts for the center of pressure being located off
the mass center while the terms involving CNPA accounts for the center of
Magnus being located off the mass center. In Equations 1 and 2, the
aerodynamic coefficients and the distances from the aerodynamic force
components to the projectile mass center are all a function of local Mach
number. Typically in flight dynamic trajectory computer codes, this dependence
on Mach number is handled through a table look-up scheme.

4 Aerodynamic coefficient estimation

The time accurate coupled CFD/RBD simulation provides a full flow solution
including the aerodynamic portion of the total applied force and moment
( X , Y , Z , L, M , N ) along with the full state of the rigid projectile
( x, y, z , φ , θ ,ψ , u , v, w, p, q, r ) at each time step in the solution. The rigid
state of the projectile is used to obtain the weight portion of the applied force so
that the aerodynamic force can be isolated. Using the information provided by
the coupled CFD/RBD simulation, it is desired to compute all aerodynamic
coefficients: C X 0 , C X 2 , C NA , CYPA , CLDD , CLP , CMA , CMQ , C NPA . For a fin-
stabilized projectile, the Magnus force and moment are usually sufficiently small
so that CYPA and C NPA are set to zero and removed from the fitting procedure to
be described below.
To estimate the aerodynamic coefficients near a particular Mach number, a
set of n time accurate coupled CFD/RBD simulations are created over a
relatively short time period. The initial conditions for the set of n time histories
are generated to produce a rich database of aerodynamic loads and projectile
states so that a unique solution can be obtained for the aerodynamic coefficients.
The initial conditions for the rigid projectile states are Gaussian random numbers
with a mean and standard deviation selected to cover normal operating
conditions for the projectile. Since the aerodynamic coefficients to be estimated
depend on local Mach number, the set of n time histories is repeated at m
different Mach numbers of interest. Thus a total of n * m short time accurate
coupled CFD/RBD trajectories are generated to support computation of a
complete set of aerodynamic coefficients for flight dynamic simulation.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 39

Since initial conditions for a given set of n time histories are randomly
generated and because Mach number changes during a simulation, Mach number
varies slightly even for a set of time histories intended to be generated at a
particular Mach number. Hence, at a particular Mach number, all aerodynamic
coefficients are assumed to vary linearly with Mach number. Parameters at
intermediate values of Mach number were linearly interpolated as shown,

M − M−
C ( M ) = C− + ( C + − C− ) (8)
M+ − M−

where C− and C+ are the aerodynamic coefficient values at Mach numbers

slightly less than ( M − ) and slightly greatly than ( M + ) the target Mach number.
This general form for the aerodynamic coefficients is then substituted into the
aerodynamic force and moment equations. Note that all aerodynamic
coefficients that are to be estimated appear in the force and moment equations in
a linear fashion suggesting a linear least squares approach to estimate the
aerodynamic coefficients at each Mach number.
Define the vectors PX , PYZ , PL , and PMN as vectors containing all the
unknown aerodynamic coefficients that are to be estimated at a given target
Mach number.

PX = C X− 0 C X+ 0 C X− 2 C X+ 2  (9)

PYZ = CNA
− +
CNA −
CYPA +
CYPA  (10)

PL = CLP
− +
CLP −
CLDD +
CLDD  (11)

PMN = CMA
− +
CMA −
CMQ +
CMQ −
CNPA +
CNPA  (12)

Denote the total number of unknowns as j . For a fin stabilized projectile,

j = 14 while for a spin stabilized projectile j = 18 . Assuming each time
history contains k time simulation output points, then k * n linear equations in
j unknowns are generated at each target Mach number.

AX PX = BX (13)

AYZ PYZ = BYZ (14)

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
40 Computational Ballistics III

AL PL = BL (15)

AMN PMN = BMN (16)

Provided the matrices AX , AYZ , AL , and AMN are maximal rank, a unique
solution for PX , PYZ , PL , and PMN exists. Thus, properties of the matrices
above, such as the rank or singular values, can be used as an indicator of the
suitability of the CFD/RBD simulation data in estimating the aerodynamic
coefficients at the target Mach number.

5 Results
In order to exercise the method developed above, a detailed comparison of the
aerodynamic forces and moments for a typical finned projectile are shown in
Figures 3–8. Mach number varied from 3.03 to 2.97, covering 0.07 s and
containing 3317 points. The projectile had the following properties: mass of
4.84x10-1 kg, length of 1.259x10-1 m, diameter of 1.319x10-2 m, and axial inertia
of 7.4x10-7 kg-m2. The estimated data is generated using Equations 5 and 6 with
the identified aerodynamic coefficients. Aerodynamic forces and moments agree
well, particularly the Y and Z components which appear coincident in the plotted
data. The CFD/RBD data appears slightly noisy in both the axial force and
rolling moment. However, the estimated data removes the noise.
Figures 9–17 present estimation results for an example finned projectile and
an example spin stabilized projectile. The finned projectile is a 120 mm direct
fire kinetic energy round while the spin stabilized projectile is a 155 mm shell.
Synthetic CFD/RBD data was generated using a rigid 6 DOF trajectory
simulation. The solid lines correspond to the coefficient values used to generate
the synthetic CFD/RBD data. The “square” symbols represent the finned
projectile aerodynamic coefficient estimations and the “circle” symbols represent
the spin stabilized aerodynamic coefficient estimations. In Figure 14, Cmq
values for the finned projectile are scaled by a factor of 1/40 so that data for both
rounds is easily viewed. The spin stabilized projectile cover a Mach range from
0.6 to 4, while the finned projectile covers a Mach range from 1 to 4. One
thousand output points were used from each of the 25 runs at each Mach number
of interest. For the finned projectile, each run had random initial conditions for
angular rates with a zero mean and a Gaussian standard deviation of 3 rad/s for
roll rate and 2 rad/s for pitch and yaw rates. For the spin stabilized projectile,
each run had random initial conditions for roll rate with a mean value of 900
rad/s and Gaussian standard deviation of 10 rad/s.
Parametric trade studies were conducted to determine the effect of the number
of runs needed for convergence of the aerodynamic coefficients as well as the
effect of the length of the time snippet on convergence. Figures 18–23 present
results for CX2 and CMQ as a function of the number of runs used for
estimation. The number of runs at each Mach number was varied to values of 4,

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 41

6, and 10. The number of output points was 1 for all runs. Note that for 10 runs,
the aerodynamic coefficients are converged. The effect of the number of data
points in each time snippet was also investigated. Three values were considered:
1, 2, and 5. The number of runs at each Mach number is 6 and the snippet length
equals 0.1 s. While not shown, all aerodynamic coefficients converged with 5
data points.

Figure 3: Estimated (dashed) and CFD/RBD (solid) body axis axial force
versus time.

Figure 4: Estimated (dashed) and CFD/RBD (solid) body axis side force
versus time.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
42 Computational Ballistics III

Figure 5: Estimated (dashed) and CFD/RBD (solid) body axis vertical force
versus time.

Figure 6: Estimated (dashed) and CFD/RBD (solid) body axis rolling

moment versus time.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 43

Figure 7: Estimated (dashed) and CFD/RBD (solid) pitching moment versus

time.

Figure 8: Estimated (dashed) and CFD/RBD (solid) yawing moment versus

time.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
44 Computational Ballistics III

Figure 9: CX0 versus mach number (square = spin stabilized projectile, circle
= finned projectile; solid line = data, symbols = estimated).

Figure 10: CX2 versus Mach number (square = spin stabilized projectile,
circle = finned projectile; solid line = data, symbols = estimated).

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 45

Figure 11: CNA versus Mach number (square = spin stabilized projectile,
circle = finned projectile; solid line = data, symbols = estimated).

Figure 12: CLP versus Mach number (square = spin stabilized projectile,
circle = finned projectile; solid line = data, symbols = estimated).

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
46 Computational Ballistics III

Figure 13: CLDD versus Mach number (square = spin stabilized projectile,
circle = finned projectile; solid line = data, symbols = estimated).

Figure 14: CMQ versus Mach number (square = spin stabilized projectile,
circle = finned projectile; solid line = data, symbols = estimated).

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 47

Figure 15: Stationline center of pressure versus Mach number (square = spin
stabilized projectile, circle = finned projectile; solid line = data,
symbols = estimated).

Figure 16: CYPA versus Mach number for spin stabilized projectile (dashed
line = data, symbols = estimated).

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
48 Computational Ballistics III

Figure 17: Stationline center of Magnus force versus Mach number for spin
stabilized projectile (solid line = data, symbols = estimated).

Figure 18: CX2 versus Mach number: 1 data point per time snippet, 4 time
snippets.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 49

Figure 19: CX2 versus Mach number: 1 data point per time snippet, 6 time
snippets.

Figure 20: CX2 versus Mach number: 1 data point per time snippet, 10 time
snippets.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
50 Computational Ballistics III

Figure 21: CMQ versus Mach number: 1 data point per time snippet, 4 time
snippets.

Figure 22: CMQ versus Mach number: 1 data point per time snippet, 6 time
snippets.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 51

Figure 23: CMQ versus Mach number: 1 data point per time snippet, 10 time
snippets.

6 Conclusions
Using a time accurate computational fluid dynamics simulation that is tightly
coupled to a rigid body dynamics simulation, a method to efficiently generate a
complete aerodynamic description for projectile flight dynamic modeling is
described. A set of n very short time snippets of simulated projectile motion at
m different Mach numbers is computed and employed as baseline data. The
combined CFD/RBD analysis computes time synchronized air loads and
projectile state vector information, leading to a straight forward fitting procedure
to obtain the aerodynamic coefficients. The estimation procedure decouples into
4 sub problems that are each solved via linear least squares. By inspecting the
condition number of each fitting matrix, the suitability of the time history data to
predict a selected set of aerodynamic coefficients can be assessed. The overall
method has been shown to work well for both fin and spin stabilized projectiles.
As would be expected, convergence of the aerodynamic coefficients is strongly
influenced by the number of time snippets and the number of points in each time
snippet. This technique provides a new means for the CFD analyst to predict
aerodynamic coefficients for flight dynamic simulation purposes. While
CFD/RBD is computationally intensive, the method described in this paper
provides a means to monitor progress in estimating aerodynamic coefficients
through the individual fitting matrices.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
52 Computational Ballistics III

Nomenclature
x, y, z : Components of position vector of mass center in an inertial reference
frame.
φ ,θ ,ψ : Euler roll, pitch, and yaw angles of box.
u , v, w : Components of velocity vector of mass center in body reference frame.
p, q, r : Components of angular velocity vector in body reference frame.
Fx , Fy , Fz : Total applied force components in body reference frame.
M x , M y , M z : Total applied moment components about mass center in body
reference frame.
V : Magnitude of relative aerodynamic velocity vector of mass center.
ρ : Air density.
D : Projectile diameter.
α : Aerodynamic angle of attack.
Cx0: Zero yaw drag aerodynamic coefficient.
Cx2: Yaw drag aerodynamic coefficient.
Cna: Normal force due to angle of attack aerodynamic coefficient.
Cypa: Magnus force aerodynamic coefficient.
Clp: Roll damping aerodynamic coefficient.
Cldd: Fin cant aerodynamic coefficient.
Cmq: Pitch damping moment aerodynamic coefficient.
Dcop: Distance from the mass center to the center of pressure.
Dmag: Distance from the mass center to the center of Magnus.
CFD: Computational Fluid Dynamics.

References
[1] J. Sun, R. Cummings, “Evaluation of Missile Aerodynamic
Characteristics Using Rapid Prediction Techniques,” Journal of
Spacecraft and Rockets, Vol 21, No 6, pp 513-520, 1984.
[2] F. Moore, “The 2005 Version of the Aeroprediction Code (AP05),” AIAA
2004-4715, AIAA Atmospheric Flight Mechanics Conference,
Providence, Rhode Island, 2004.
[3] T. Sooy, R. Schmidt, “Aerodynamic Predictions, Comparisons, and
Validations Using Missile DATCOM and Aeroprediction 98,” AIAA-
2004-1246, AIAA Aerospace Sciences Meeting and Exhibit, Reno,
Nevada, 2004.
[4] J. Simon, W. Blake, “Missile DATCOM – High Angle of Attack
Capabilities,” AIAA-1999-4258, AIAA Atmospheric Flight Mechanics
Conference, Portland, Oregon, 1999.
[5] A. Neely, I. Auman, “Missile DATCOM Transonic Drag Improvements
for Hemispherical Nose Shapes,” AIAA-2003-3668, AIAA Applied
Aerodynamics Conference, 2003.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 53

[6] W. Blake, “Missile DATCOM – 1997 Status and Future Plans,” AIAA-
1997-2280,” AIAA Applied Aerodynamics Conference, Atlanta, Georgia,
1997.
[7] A. Dupuis, C. Berner, “Wind Tunnel Tests of a Long Range Artillery
Shell Concept,” AIAA-2002-4416, AIAA Atmospheric Flight Mechanics
Conference, Monterey, California, 2002.
[8] C. Berner, A Dupuis, “Wind Tunnel Tests of a Grid Fin Projectile
Configuration,” AIAA-2001-0105, AIAA Aerospace Sciences Meeting,
Reno, Nevada, 2001.
[9] J. Evans, “Prediction of Tubular Projectile Aerodynamics Using the
ZUES Euler Code,” Journal of Spacecraft and Rockets, Vol 26, No 5, pp
314-321, 1989,
[10] W. Sturek, C. Nietubicz, J. Sahu, P. Weinacht, “Applications of
Computational Fluid Dynamics to the Aerodynamics of Army Projectiles,
Journal of Spacecraft and Rockets, Vol 31, No 2, pp 186-199, 1994.
[11] M. Nusca, S. Chakravarthy, U. Goldberg, “Computational Fluid
Dynamics Capability for the Solid-Fuel Ramjet Projectile,” Journal of
Propulsion and Power, Vol 6, No 3, 1990.
[12] S. Silton, “Navier-Stokes Computations for a Spinning Projectile from
Subsonic to Supersonic Speeds,” Journal of Spacecraft and Rockets, Vol
42, No 2, pp 223-231, 2005.
[13] J. DeSpirito, M. Vaughn, D. Washington, “Numerical Investigation of
Canard-Controlled Missile with Planar Grid Fins,” Journal of Spacecraft
and Rockets, Vol 40, No 3, pp 363-370, 2003.
[14] M. Graham, P. Weinacht, J. Bennett, “Numerical Investigation of
Supersonic Jet Interaction for Finned Bodies, Journal of Spacecraft and
Rockets, Vol 37, No 5, pp 675-683, 2000.
[15] P. Weinacht, “Navier-Stokes Prediction of the Individual Components of
the Pitch Damping Sum,” Journal of Spacecraft and Rockets, Vol 35, No
5, pp 598-605, 1998.
[16] B. Guidos, P. Weinacht, D. Dolling, “Navier-Stokes Computations for
Pointed, Spherical, and Flat Tipped Shells at Mach 3,“ Journal of
Spacecraft and Rockets, Vol 29, No 3, pp 305-311, 1992.
[17] P. Weinacht, W. Sturek, “Computation of the Roll Characteristics of a
Finned Projectile,” Journal of Spacecraft and Rockets, Vol 33, No 6, pp
769-775, 1996.
[18] S. Park, J. Kwon, “Navier-Stokes Computations of Stability Derivatives
for symmetric Projectiles, AIAA-2004-0014, AIAA Aerospace Sciences
Meeting, Reno, Nevada, 2004.
[19] N. Qin, K. Ludlow, S. Shaw, J. Edwards, A. Dupuis, “Calculation of Pitch
Damping for a Flared Projectile,” Journal of Spacecraft and Rockets, Vol
34, No 4, pp 566-568, 1997.
[20] P. Weinacht, “Coupled CFD/GN&C Modeling for a Smart Material
Canard Actuator,” AIAA-2004-4712, AIAA Atmospheric Flight
Mechanics Conference, Providence, Rhode Island, 2004.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
54 Computational Ballistics III

[21] S. Park, Y. Kim, J. Kwon, “Prediction of Dynamic Damping Coefficients

Using Unsteady Dual Time Stepping Method,“ AIAA-2002-0715, AIAA
Aerospace Sciences Meeting, Reno, Nevada, 2002.
[22] J. DeSpirito, K. Heavey, ”CFD Computation of Magnus Moment and
Roll-Damping Moment of a Spinning Projectile,” AIAA-2004-4713,
AIAA Atmospheric Flight Mechanics Conference, Providence, Rhode
Island, 2004.
[23] K. Garon, G. Abate, W. Hathaway, “Free-Flight Testing of a Generic
Missile with MEMs Protuberances,” AIAA -2003-1242, AIAA Aerospace
Sciences Meeting, Reno, Nevada, 2003.
[24] B. Kruggel, “High Angle of Attack Free Flight Missile Testing,” AIAA-
1999-0435, AIAA Aerospace Sciences Meeting, Reno, Nevada, 1999.
[25] J. Danberg, A. Sigal, I. Clemins, “Aerodynamic Characteristics of a
Family of Cone-Cylinder-Flare Projectiles, Journal of Spacecraft and
Rockets, Vol 27, No 4, 1990.
[26] A. Dupuis, “Free-Flight Aerodynamic Characteristics of a Practice Bomb
at Subsonic and Transonic Velocities,” AIAA-2002-4414, AIAA
Atmospheric Flight Mechanics Conference, Monterey, California, 2002.
[27] G. Abate, R. Duckerschein, W. Hathaway, “Subsonic/transonic Free-
Flight Tests of a Generic Missile with Grid Fins, AIAA-2000-0937, AIAA
Aerospace Sciences Meeting, Reno, Nevada, 2000.
[28] G. Chapman, D. Kirk, “A Method for Extracting Aerodynamic
Coefficients from Free-Flight Data,” AIAA Journal, Vol 8, No 4, pp 753-
758, 1970.
[29] G. Abate, A. Klomfass, “Affect upon Aeroballistic Parameter
Identification from Flight Data Errors,” AIAA Aerospace Sciences
Meeting, Reno, Nevada, 2005.
[30] G. Abate, A. Klomfass, “A New Method for Obtaining Aeroballistic
Parameters from Flight Data,” Aeroballistic Range Association Meeting,
Freiburg, Germany, 2004.
[31] P. Weinacht, W. Sturek, L. Schiff, “Projectile Performance, Stability, and
Free-Flight Motion Prediction Using Computational Fluid Dynamics,“
Journal of Spacecraft and Rockets, Vol 41, No 2, pp 257-263, 2004.
[32] J. Sahu, “Time-Accurate Numerical Prediction of Free-Flight
Aerodynamics of a Finned Projectile,” AIAA-2005-5817, AIAA
Atmospheric Flight Mechanics Conference, San Francisco, California,
2005.
[33] G. Abate, A. Klomfass, “Affect upon Aeroballistic Parameter
Identification from Flight Data Errors,” AIAA-2005-0439, AIAA
Aerospace Sciences Meeting, Reno, Nevada, 2005.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Section 2
Interior ballistics
This page intentionally left blank
 Computational Ballistics III 57

Suppression of axial combustion instability in

solid rocket motors
D. R. Greatrix
Department of Aerospace Engineering, Ryerson University, Canada

Abstract
A numerical model for the prediction of undesirable axial combustion instability
symptoms in solid-propellant rocket motors is utilized in the process of
demonstrating various approaches for suppressing this instability behaviour. The
unsteady interior one- or two-phase flow in the motor chamber, the oscillation of
the surrounding structure of the propellant and motor casing, and the
corresponding transient combustion process, are all incorporated within this
comprehensive model. Example numerical results are presented for a cylindrical-
grain motor having differing characteristics with respect to particle loading and
grain geometry (fore and aft cross-sectional area changes), in illustrating the
effectiveness of the particular suppression technique being applied.
Keywords: solid rocket motor, combustion instability, particles, geometry.

1 Introduction
Over the last fifty years or so, there has been a number of research efforts
directed towards understanding the physical mechanisms, or at least the
surrounding factors, behind the appearance of symptoms associated with
nonlinear axial combustion instability in solid rocket motors (SRMs). The
motivation for these studies was and is of course to bring this better
understanding to bear in more precisely suppressing, if not eliminating, these
symptoms. Research towards predicting instability symptoms in SRMs
necessitates comprehensive numerical models for interior ballistic simulation
under dynamic flow and combustion conditions. An effective model combines
the effects of the unsteady flow, the transient combustion process, and the
structural dynamics of the surrounding propellant/casing structure. In the present

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070051
58 Computational Ballistics III

Figure 1: SRM model setup.

investigation, a numerical model incorporating the above attributes is used in the

prediction of the unsteady instability-related behaviour in a reference cylindrical-
grain motor (sustained axial pressure wave presence, in this case), and various
means for instability suppression are in turn demonstrated.

2 Method
A simplified schematic diagram of the physical system of an SRM placed on a
static test stand is provided in fig. 1. In this example, the cylindrical-grain
motor is free to vibrate radially without any external constraint (i.e., only
constrained as indicated by a thick steel static-test sleeve surrounding the
aluminum flightweight casing), while axial motion is constrained to a large
degree by the thrust-measuring load cell (represented here as a spring/damper) at
the lefthand boundary. Under nominal operating conditions, the internal gas
flow moves smoothly from the burning propellant surface through and beyond
the exhaust nozzle.
With respect to modelling the internal flow within the motor, a number of
studies for SRMs having larger length-to-diameter (L/D) ratios have
incorporated the use of a one-dimensional unsteady flow model, e.g., as done by
Loncaric et al [1] and Greatrix [2,3] using a higher-order random-choice method
(RCM) for the one- or two-phase flow solution. The effect of such factors as
turbulence can be included through one or more additional equations that employ
the information from the bulk flow properties arising from the one-dimensional
equations of motion for the gas (and for the particles, if a two-phase flow
solution). In those cases where two- or three-dimensional flow effects can
influence a particular aspect of the motor’s interior ballistic behaviour, one might
consider a multi-dimensional flow solution approach if a one-dimensional
solution does not capture (or capture adequately) the phenomenon in question.
For the purposes of the present study, a one-dimensional RCM approach will be
utilized for the unsteady flow solution.
Structural vibration can play a significant role in nonsteady SRM interior
ballistic behaviour, as evidenced by observed changes in combustion instability
symptoms as allied to changes in the structure surrounding the internal flow
(e.g., propellant grain configuration, wall thickness, material properties). The

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 59

level of sophistication required for modelling the motor structure (propellant,

casing, static-test sleeve, nozzle) and applicable boundary conditions (load cell
on static test stand) can vary, depending on the particular application and motor
design. Loncaric et al [1] employed a finite-element approach towards the
structural modelling of a star-grain propellant configuration. In the present
study, a cylindrical-grain configuration allows for a simpler approach from thick-
wall theory, as reported by Greatrix and Harris [4]. The radial deformation
dynamics of the propellant/casing/sleeve are modelled by a series of independent
ring elements along the length of the motor. Axial motion along the length of
the structure is modelled via beam theory, and bounded by the spring/damper
load cell at the motor’s head end. Viscous damping is applied in the radial and
axial directions. Reference structural properties are assumed for an ammonium-
perchlorate/hydroxyl-terminated polybutadiene (AP/HTPB) composite
propellant surrounded by an aluminum casing and steel sleeve. For greater
accuracy, some properties like the propellant/casing/sleeve assembly’s natural
radial frequency may be predetermined via a finite-element numerical solution,
rather than via theoretical approximations.
The transient burning rate for the regression of the core periphery of the
cylindrical propellant grain may be modelled through the Zeldovich-Novozhilov
approach; see [5]. This phenomenological modelling technique applies an
energy conservation criterion in coupling the heat conduction within the solid
phase (the propellant) to the heat produced in the gas phase above the propellant
surface. Conveniently, empirical or semi-empirical steady-state burning rate
information may be used in place of more complex dynamic flame-based
reaction rate equations. The principal equation for the nominal (or
unconstrained) instantaneous burning rate rb*, tying the solid phase to the gas
phase, is given by:
0
1 ∂
rb∗ = rb,qs −
(Ts − Ti − ∆H s / C s ) ∂t ∫ ∆Tdy
−∞
(1)

where rb,qs is the quasi-steady burning rate (value for burning rate as estimated
from steady-state information for a given set of local flow conditions), Ts is the
burning surface temperature at the spatial position y = 0, Ti is the initial
propellant temperature (y < 0 moving deeper into the propellant), and T = T(y,t)
– Ti is the temperature distribution within the propellant. The transient heat
conduction in the solid phase can be solved by an appropriate finite-difference
scheme. As reported by Greatrix [5], the actual instantaneous burning rate rb
may be found as a function of rb* through the rate limiting equation:

drb
= K b (rb∗ − rb ) (2)
dt

The rate limiting coefficient Kb effectively damps the unconstrained burning rate
rb* when for a finite time increment ∆t:

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
60 Computational Ballistics III

1
Kb < (3)
∆t

The quasi-steady burning rate rb,qs can be found as a function of various

parameters; in this study, as a function of local static pressure p, core flow
velocity u (erosive burning component), and normal/lateral/longitudinal
acceleration such that:

rb,qs = r p + ru + ra (4)

The equations needed for the coupled solution of rb,qs may be found in [1] and
[3].

3 Results and discussion

The reference SRM for this investigation is a cylindrical-grain motor employing
a nonaluminized AP/HTPB propellant, with characteristics, such as the solid
propellant’s specific heat Cs, as reported by Greatrix and Harris [4]; some
parameters have been updated or added for the current effort. A thin steel sleeve
of 4.7-mm thickness is in place for the reference static-test simulations. In fig. 2,
one can view the general combustion response (here, in terms of nondimensional
limit magnitude M A , defined by

rb, peak − rb,o

MA = , (5)
rb,qs , peak − rb,o

where unity would indicate a flat response) of the reference solid propellant, at
differing values of the net surface heat release ∆Hs (sign convention: positive
with exothermic output of heat). The propellant’s resonant frequency fr of
approximately 1 kHz is established via the value set for Kb (20000 s-1). The
value of 1 kHz is consistent with regard to being within the range of observed
values for composite propellants at typical base burning rates, and will allow for
close to the worst-case scenario in being similar to the fundamental axial
resonant pressure wave frequency f1L. In an actual test firing for evaluation of an
SRM’s susceptibility to nonlinear axial combustion instability symptoms, after
the initial travelling pulse disturbance is introduced into the flow of the motor
chamber, an unstable motor will exhibit a sustained axial compression wave, on
occasion accompanied by a substantial increase in the base chamber pressure
(referred to as a dc shift). An initial pulsed-firing simulation run was completed
as a starting reference for this study, in which no particles are present or any
other suppression technique being applied. In fig. 3 for head-end pressure pc as a
function of time, one can see that at some point, the principal compression wave
reaches its quasi-equilibrium strength from an initial disturbance pressure ∆pd of
2 atm, the sustained compression wave front arriving about every 1 ms,

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 61

5
Limit Magnitude

∆Ηs = +30000 J/kg

∆Ηs = +100000 J/kg
∆Ηs = +150000 J/kg

0
0 1000 2000 3000
Frequency, Hz

Figure 2: Frequency response of ref. propellant (Kb = 20000 s-1; differing

values for ∆Hs; rb,o = 1.27 cm/s) in terms of nondimensional
magnitude.

oscillating at the fundamental frequency f1L of 1 kHz. The base pressure is not
appreciably elevated over the nominal operating chamber pressure. The effect of
normal acceleration on the burning process (related to the radial vibration of the
motor propellant/casing; see [2,3]) has been nullified for this simulation (in order
to isolate frequency-dependent Z-N combustion response as the predominant
instability symptom driver for this and subsequent runs discussed in this paper),
a factor in reducing the development of a dc shift. One can note that the limit
pressure wave magnitude (∆pw, peak-to-trough) is decreasing gradually with time
after first reaching its quasi-equilibrium level, as the cylindrical grain burns back
and the base pressure rises.
One can refer to fig. 4 for the pressure-time profile for the same motor, but
now with 5% particle loading (by mass) of inert spherical aluminium particles
having a 12-µm diameter. Suppression of axial wave development after an initial
2-atm pulse is near-complete (limit magnitude of the sustained pressure wave, at
0.26 s, is about 0.042 MPa [∆pw], as compared to 1.42 MPa for the 0% loading
case noted earlier [∆pw,peak], giving a nondimensional attenuation Ma, defined by

∆p w, peak − ∆p w
Ma = , (6)
∆p w, peak

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
62 Computational Ballistics III

13
Pressure, MPa

12

11

10

0.10 0.15 0.20 0.25

Time, s
Figure 3: Predicted head-end pressure-time profile, reference motor (Kb =
20000 s-1, ∆Hs = 150000 J/kg, ∆pd = 2 atm, αp = 0%), acceleration
nullified.
a value of 0.97, noting a value of unity is complete suppression). Historically,
suppression of high-frequency tangential and radial pressure waves in SRMs by
the use of particles in the range of 1 to 3% loading by mass has been in general
largely successful. In the case of axial pressure waves, the effectiveness of
particles from 1% to over 20% loading in suppressing wave development has
been less consistent, relative to the previously mentioned transverse cases. In the
case of fig. 4, remembering that acceleration as a factor has been nullified in the
combustion process, a loading of 5% at 12 µm does appear to effectively
suppress axial wave development in this particular motor, at this point in its
firing.
An attenuation map for the above motor is provided in fig. 5. Here, various
Ma curves are displayed as a function of aluminium particle diameter, with each
curve being at a particular loading percentage. The case of fig. 4 is the
approximate peak Ma point of the 5% loading curve in fig. 5. The x-axis in
essence represents the 0% loading case (no suppression at any particle diameter).
The same juncture in the firing simulations is the time reference for each point
(0.26 s). Dobbins and Temkin [6], via acoustic (weak-wave) theory for low-
Reynolds-number flow of a gas-particle mixture, reported the following well-
known relation for the optimum particle diameter dopt in attenuating pressure
waves:

9µ
d opt = (7)
πfρ m

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 63

13
Pressure, MPa

12

11

10

0.10 0.15 0.20 0.25

Time, s

Figure 4: Predicted head-end pressure-time profile, reference motor (Kb =

20000 s-1, ∆Hs = 150000 J/kg, ∆pd = 2 atm, αp = 5%, dm = 12 µm),
acceleration nullified.

Here, µ is the gas absolute viscosity, ρm is the particle solid density, and f is the
frequency of the oscillating axial pressure wave. For the present motor, eq. (7)
would suggest a dopt of around 9.2 µm, while fig. 5 suggests peak attenuation
occurs at particle diameters of around 10 to 25 µm at very low loadings (around
1% for αp, peak appears as a plateau range), narrowing and dropping somewhat
towards 12 µm as one increases the loading from 1% towards 5%, and then again
a spreading range (plateau peak) from that figure of 12µm at even higher
loadings. It is not surprising that there would be a difference from the predicted
acoustic-theory attenuation, given that the low-speed, weak-wave assumptions
inherent in eq. (7) are not being presumed in generating the higher-flow-speed
numerical results of fig. 5, specifically where both laminar- and turbulent-flow
drag between the spherical particles and gas may in fact be present as a
predominant attenuation/loss mechanism.
An alternative or complementary approach to the use of particles is to change
the internal grain geometry to improve the suppression of observed symptoms.
In practice, one sometimes notes that c.i. symptoms only occur within a certain
timeframe within a given motor design’s firing, suggesting that the grain shape
during that interim period is somehow more susceptible to pressure-wave or dc
shift development. Conversely then, outside of that time period, something
about the grain’s shape is acting to suppress symptom development. As an
example, in fig. 6, one can see a comparison in limiting wave strength (measured

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
64 Computational Ballistics III

1
αp = 1%
αp = 2%
αp = 3%
αp = 4%
αp = 5%
Attenuation

0
0 10 20 30 40 50
Diameter, µm

Figure 5: Nondimensional attenuation as function of particle diameter and

loading, reference motor, acceleration nullified.

at the head end) at a later point in the firing with the introduction of a roughly
2:1 cross-sectional core area expansion (moving left to right) through the
halfway point on the grain of the reference motor, reducing the pressure wave to
about 25% of its reference strength. Note that the nozzle throat diameter was set
at 1.7 cm (vs. the reference 1.6 cm) for the expansion case, for a comparable
chamber pressure. Positioning the internal grain expansion at the one-quarter
and three-quarter grain-length positions respectively was not nearly as effective.
Moving to the case of contracting the grain internal core cross-sectional area
as one moves left to right, one can see from fig. 7 that at the one-quarter position,
this technique is quite impressive in almost fully suppressing wave activity. The
nozzle throat diameter was set at 1.68 cm (vs. the reference 1.6 cm) for the
contraction case, for a comparable chamber pressure. Positioning the internal
grain contraction at the halfway position is still very much effective in
suppressing wave activity, while much less effective at the three-quarter
position.
The analysis undertaken for this paper provides a few examples for how one
might suppress, to some degree, instability symptoms within a given motor’s
design. By no means are these the only techniques available to the motor
designer. Ramohalli [7] provides a good review of a number of techniques that
have been attempted over the years.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 65

Pressure, MPa 13

12

ref. straight cyl.

0.5L/2:1 expansion

11
0.260 0.265
Time, s

Figure 6: Predicted head-end pressure-time profiles, reference motor (Kb =

20000 s-1, ∆Hs = 150000 J/kg, ∆pd = 2 atm, αp = 0%), grain
expansion case, acceleration nullified.

11.0
Pressure, MPa

10.5

0.10 0.15
Time, s

Figure 7: Predicted head-end pressure-time profiles, reference motor (Kb =

20000 s-1, ∆Hs = 150000 J/kg, ∆pd = 2 atm, αp = 0%), grain
contraction case, acceleration nullified.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
66 Computational Ballistics III

4 Concluding remarks
The implications of such factors as particle loading and grain geometry on
nonlinear axial combustion instability symptom development and suppression
have been demonstrated by the numerical simulation results presented in this
study of a reference cylindrical-grain solid rocket motor. Unquestionably,
further work remains to be done in establishing a more complete understanding
of the various mechanisms involved in driving instability symptoms, so as to
allow for more refined and targeted techniques in suppressing said symptoms.

References
[1] Loncaric, S., Greatrix, D.R. and Fawaz, Z., Star-grain rocket motor —
nonsteady internal ballistics. Aerospace Science & Technology, 8(2), pp.
47-55, 2004.
[2] Greatrix, D.R., Nonsteady interior ballistics of cylindrical-grain solid
rocket motors. Computational Ballistics II, WIT Press, pp. 281-289, 2005.
[3] Greatrix, D.R., Predicted nonsteady internal ballistics of cylindrical-grain
motor. AIAA Paper No. 2006-4427, 2006.
[4] Greatrix, D.R. and Harris, P.G., Structural vibration considerations for
solid rocket internal ballistics modeling. AIAA Paper No. 2000-3804,
2000.
[5] Greatrix, D.R., Transient solid-propellant burning rate model:
comparisons to experimental data. AIAA Paper No. 2006-4426, 2006.
[6] Dobbins, R.A. and Temkin, S., Measurements of particulate acoustic
attenuation, AIAA Journal, 2(6), pp. 1106-1111, 1964.
[7] Ramohalli, K., Technologies and techniques for instability suppression in
motors, Nonsteady Burning and Combustion Stability of Solid
Propellants, edited by De Luca, L., Price, E.W. and Summerfield, M.,
Vol. 143, Progress in Astronautics & Aeronautics series, AIAA,
Washington D.C., Ch. 20, 1992.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 67

ICE-ALE modelling for the combustion of

homogeneous solid propellant
A. Peratta & C. Peratta
Wessex Institute of Technology, Southampton, UK

Abstract

The paper presents a hybrid Finite Volume Method (FVM) for the laminar
combustion of a double-base (homogeneous) solid propellant burning inside a
two-dimensional narrowed pipe. The approach for the hydrodynamics in the gas
phase is based on an Implicit-Continuous Eulerian method, implemented with
the Arbitrary Lagrangian-Eulerian FVM, while the condensed phase is modelled
with an Implicit Eulerian FVM approach. The hypothesis of pre-mixed laminar
flame, and homogeneous solid phase are assumed. The formulation is based on the
conservation equations of mass, species, momentum and energy.
The processes of degradation and pyrolysis are both taken into account in the
solid phase by means of two finite chemical reactions. The complete chemical
model is implemented with 7 unidirectional global finite rate chemical reactions
and 12 reactive species. The outcomes of the model are pressure, temperature,
density, heat release, species concentration in gas and solid phase, gas speed and
bulk burning rate. The numerical model is time dependent, and it is able to describe
the main production of the flame structure, including the induction (dark) zone.
The numerical results obtained for the burning rate are in good agreement with the
empirically based Saint-Roberts law which relates the burning rate withe the bulk
pressure by means of an exponential law. The method has also been proved suitable
for solving time dependent two dimensional chemically reactive flows problems
presenting sub-sonic and super-sonic regimes in the same integration domain.
Keywords: combustion, solid homogeneous propellant, double-base, ale, finite
volumes.

1 Introduction
Solid energetic materials are mostly used for rocket propulsion systems in either
space or defence industry. Other applications include pyrotechnic actuators,
boosters, manoeuvring and safety devices, as well as air-bags for automobile safety

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070061
68 Computational Ballistics III

[1] and gas generators. Among the large family of solid energetic materials are
the “smokeless” homogeneous propellants, developed several decades ago, which
are still in use in many of the above mentioned applications. The homogeneous
propellant, also known as Double-Base propellant (DBP) carries both oxidising
and reducing chemical agents in the same molecular structure, usually composed
of a homogeneous mixture of Nitroglycerine (NG) and Nitrocellulose (NC)
(N G = C3 H5 O3 (N O2)3 , and N C = C6 H7 O2 (N O3 )3 , respectively) with
some plasticiser and other minor ingredients [2, 3]. Because of this feature, the
combustion of DBPs can be regarded as preliminary mixed from the beginning
of the process. Although DBP do not usually have high specific impulses
(Is
210s under normal conditions), they are still on high demand due to
its smokeless combustion, and their relatively easy manufacturing, they can be
made by solvent or semi-solvent processes and extruded in various shapes. Hence,
there is no need to keep into account mixing processes between oxidiser and
reducer. Much effort has been done in research oriented to the CFD applied to
solid propellant combustion (see for example [4, 5]). The aim of this work is
to develop a comprehensive numerical modelling for the unsteady chemically
reactive compressible multi-species flow interacting with the solid reactive burning
surface of the DBP. The paper is organised as follows, in section 2 the theoretical
formulation is established. Section 3 provides a brief description of the DB under
study. Section 4 describes the interface boundary conditions, Section 5 presents
the numerical scheme. Section 6 shows the numerical results obtained in a one-
dimensional implementation and finally Section 7 elaborates the corresponding
conclusions.

2 Mathematical formulation
The governing equations [6] are basically the following average conservation of
mass, species concentration, momentum and energy equations for a mixture of Ne
species and Nr global chemical reactions:
∂ρ
Mass + ∇ · (ρu) = 0 (1)
∂t
∂ (ρYk )

Species + ∇ · ρuYk + qD k = ωk (2)
∂t
∂ (ρu)

Momentum + ∇ ut ρu + p + σ̄ = 0 (3)
∂t
∂ (ρε) Ne
Energy + ∇ · (ρuε + qe ) + p∇ · u = − ωk ∆hTk r , (4)
∂t
k=1

where subscript k denotes species, ωk is the reaction constant, Yk the mass

fraction, p pressure, u, ρ and ε are the average velocity, density and
 internalenergy,
Ne Ne
respectively, defined as: ρu := k=1 ρk uk and ε := k=1 Yk εk − εTk r , where
T
εk = εTk r + Tr cv,k (T
)dT
, being T and Tr the actual and reference temperatures,
respectively.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 69

3 Chemical model, thermodynamic and transport properties

The specific rate of j-th global reaction κj,β is calculated by means of Arrhenius
expressions of the form: κj = Bj exp [−E0 /R0 T ], where E0 is the activation
energy, R0 is the universal gas constant, and Bj ≈ cteT α is the frequency factor
with 0 < α < 0.5 and index j. Thus, j-th rate reaction can be calculated by means
of: 
Ne
ν∗
ωjR = κj,f Ck kj ,
k=1
∗
where νjk is the stoichiometric coefficient of the reactive species k in reaction j,
and Ck is the mole concentration of the species.
The following equation of state (EOS) is considered in the gas phase

Ne
p = ρR0 T Yk /Mk
k=1
where c2s = p/ρ (1 + Rmix /cvmix ); Mk is the molecular mass, Rmix :=
Ne Ne
R0 k=1 Yk /Mk the mixture constant, cvmix = k=1 Yk cpk − Rmix the mixture
specific heat, and cs the speed of sound. The condensed phase (CP) is considered
incompressible, i.e. infinite sound speed. The specific heat at constant pressure of
each component is evaluated using: cpk (T ) = Ak +Bk T +Ck T 2 +Dk T 3 +Ek /T 2 ,
where A, B, C, D and E are constants obtained N from [7, 8]. Then, the specific
internal energy of the mixture become: cv,mix = k=1 e
Yk cpk − Rmix .
Transport coefficients of species involved in energetic materials are not always
available in published tables, therefore, they have to be approximated from the
classical statistical theory of gases. The diffusive effects considered in eqs. (1)
to (4) are viscosity, Fick’s diffusion and Fourier law for heat conduction. In this
way qD k = −ρDk,mix ∇Yk in (2) is the Fick’s
 e
mass diffusive flux, σ is the viscous
tensor in (3), and qe = −λmix ∇T − ρ N k=1 hk Yk uk − ∇(σ̄u) is the diffusive
flux of internal energy. The binary diffusion coefficient between species A and
B is calculated by mean of the Chapman–Enskog (C–E) theory and the transport
Boltzmann equation [9].
For the multi-component mixture, the effective diffusion coefficient (Dk,mix ) of
k with respect to the mixture is related to binary one DAB by means of Wilke’s
mixing rule [9, 10]:  

Ne
X
Dk,mix = (1 − Xk )/ 
j 
.
Dkj
j=k

Viscosity for a single component

√ is calculated using C–E theory and L–J potential
according to: η = 5/16 πmkB T /(πσ2 ΩD ). C–E theory can also be applied to
obtain the thermal conductivity λ of pure mono-atomic
√ species with no internal
degrees of freedom [9, 11], yielding λ = 25/32 πmkB T cv /(πσ2 ΩD ). Then,
Eucken’s correction factor given by λM η−1 = 1.32cv + 1.77R0 , is used to
consider the internal degrees of freedom in polyatomic molecules, and finally,
for the multi-component mixture, Wilke’s rule is used in the same way as for
the viscosity coefficient (η) above. Reference values for the calculation of cross

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
70 Computational Ballistics III

sections, dipole moments, and collision integrals can be obtained from refs.
[9, 11, 12].

4 The DB propellant
In this work the EC1 formulation (52 % NG + 43% NC + 5% of minor additives)
contemplated by A. Kubota [2], with 1100cal/g of heat of explosion has been
adopted. Because of its relatively large stability under room conditions, an external
source of heat (igniter) must be provided to initiate the DBP burning. During
the ignition process, heat is transported by conduction inside the solid and the
degradation reaction of NG and NC begins to accelerate generating reactive
species like N O2 and aldehydes.

Condensed phase Interface Gas phase

NO2 CO Induction
NG (dark) zone
(Foam zone)

CH2O CO 2 2850K
NC
(CHO) 2 C
NO H2O 1600K
Preheating
zone N2 Secondary
NO
700K Primary CO flame
flame CO2 zone
293K (Fizz zone) H2O

Figure 1: Temperature, species and regions in condensed and gas phase near the
burning surface of the DBP.

Then, further recombination takes place into the solid phase, yielding extra
heat to reinforce the ignition. We call condensed phase (CP)to the region
between gas and solid phase where solid propellant coexists with degradation and
recombination products.(The CP consists of a pre-heated zone and a superficial
degradation layer where molecular degradation of the propellant and further
recombination of the decomposed products takes place.) As degradation evolves,
the solid concentration drops down from one to zero. The closer to the burning
surface the higher degradation rates. When solid concentration reaches zero
at the interface, convection process drags the unburnt condensed phase from
the cold zone of the DBP thus replacing the decomposed material. Then, a
convective chemically reacting flow emerges from the burning surface toward
the gas phase with a velocity ruled by eq.(1). The emerging molecules react
with each other in the gas phase, developing an important amount of heat in
the flame zone. The distribution of heat release along the flame zone is ruled
by the coupling between the gas-phase reactions, hydrodynamics, and transport
phenomena, thus yielding a particular flame structure as qualitatively shown in
Fig. 1. Meanwhile, the heat released in the primary zone provides some feedback
energy to maintain the degradation reaction in the CP. The combustion can be
either self-sustained steady, unsteady, oscillatory, or vanishing, depending on this
energetic loop. The modelling involves Nr = 7 chemical reactions, [10, 12] and
the following chemical species: DB, N O2 , CH2 O, (CHO)2 , CO, N O, CO2 ,

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 71

H2 O, H2 , C2 H4 , N2 and C. The corresponding rate reactions, specific enthalpy,

and Arrhenius coefficients were obtained from [12–14].

5 Interface boundary conditions

Gas and CP equations couple to each other throughout the burning surface
(interface) by means of the following mass and species conservation equations:

[ρu]gas = ρp rb , and [ρuk · n̂Yk ]gas = ρp rb Yk , (5)

respectively, and energy according to:

Ne
 Ne

∂T  ∂T 
−λmix +ρ Yk hk uk · n̂ = −λs + ρrb cp (Ts − Tr ) + Yk ∆hk ,
∂ n̂ k=1
∂ n̂ k=1 (6)
gas s

where subscript s refers to the CP, rb is the local propellant burning rate, Ts is the
interface temperature, h is enthalpy and ∆h the enthalpy at reference temperature.
Because reactions in the CP have relatively high activation energy and occur in a
very thin region (few microns in foamy zone), a matched asymptotic expansion
technique as described in refs. [15, 13, 14] can be used in order to estimate the
burning rate formula. The result for a steady self-sustained combustion is:
 
2
Ts2 −Ec
rb ≈ β exp
Ts − T0 − Qs /(2cs ) R0 T s
where β
:= λs BcR0 /(cS ρs Ec ) is a constant, T0 ≈ 293K is the initial
temperature equal to the temperature at x → ∞ in the cold zone of the propellant.
The following values were assumed: ρs = 1.6g/cm3 , cs = 0.4cal/(gK),λs =
5.1 × 10−2 cal/(smK), Bc = 1017 s−1 , and Ec = 40kcal/mol. Thus rb ≈
10−2 m/s, and consequently: δxc ≈ 37µm, τc ≈ 4ms, δxd ≈ 2µm, τd ≈ 0.2ms,
where δx indicates characteristic length, τ residence time, and subscripts c and
d pre-heated and foamy zone in the CP, respectively. Note that the typical Mach
number in the flame zone is M ≈ 10−3 , this involves an incompressible flow
regime which might potentially limit the numerical scheme.

6 Numerical scheme
The gas phase is solved with the Implicit Continuous Eulerian (ICE) method [16]
and an Arbitrary Lagrangian–Eulerian (ALE) technique. The region involving the
solid (condensed)-gas phase interface has been modelled with a one-dimensional
approach oriented perpendicular to the burning boundary. The scheme for both
gas and condensed phases is based on the FVM in which each physical process
is integrated in an independent way during one calculation time step. The FVM
is implemented with a staggered grid approach where diffusive fluxes (q) are
assigned to cell faces, flow velocities (u) to cell vertices and the other field
variables such as εmix , T , p, cell volume V , and Yk are assigned to cell centres.
The different objects participating in the discretisation are basically cell vertices,

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
72 Computational Ballistics III

faces and centres. An update of the cell dynamic property ψ between time level
m and m + 1 is generally written as: ψ m+1 = ψm + δψD + δψH + δψS , where
δψD , δψH , and δψS represent the contribution to the total variation of ψ due to
diffusive, hydrodynamic and source terms, respectively. Time variations δψD of
either mass or energy due to either Fick and Fourier diffusion, or species migration
is calculated in a Lagrangian way with an Alternating Direction Implicit (ADI)
approach [17]. During the ADI step, all other terms contributing to the variation of
ψ apart from the diffusive ones, are considered as frozen. The ICE approach [12,
16] is employed in order to update field variables due to the hydrodynamic process
(δψH ), i.e. for integrating the pressure-velocity coupling and the kinetic energy of
the system. Time variations due to source terms (δψS ) are integrated explicitly in
time. For example, the increment of species concentration and total internal energy
in the cell, due to the source terms appearing in (2) and (4) arising from chemical
ωm m+1 Ne m
reactions become: δYk,i m+1
= ρk,i
m δt, and δ (mε)i = Vim k=1 ωk,i ∆hk δt,
i
respectively, where mi and Vi represent the total mass and volume of cell i.
During the Lagrangian step, all convective terms are zero, and the vertices move
with the fluid in order to preserve constant the total mass inside a cell. Because
of that, after a few time steps the grid may become excessively distorted. In
addition the cells near any outflow boundary condition may result pushed away
from the integration domain. In order to overcome these technical difficulties all
mesh vertices are moved backwards, i.e. usually opposite to the flow motion, with
an arbitrary numerical velocity ua given by the superposition:
 
x∗i − xi
ua = −u + τ (7)
δt
where u is the fluid velocity, x∗i are the 2D coordinates of the grid vertices at the
new time level, xi represent the actual location of the vertices, and τ is a relaxation
factor. When τ = 0 the method resembles an Eulerian one. The second term in the
right hand side of eq. (7) represents an arbitrary velocity which is added in order to
adapt the grid space resolution to the gradient of the developing solution. The new
location of the grid points is computed in order to obtain a uniform distribution of
the Lagrangian coordinates Li associated with i-th cell, i.e.:
∗ ∗
x1 x2
1
Li = w(x
)dx
1 dx
2 = , (8)
Nc
0 0

where Nc is the number of cells, and w = c (1 + ∇ · (b1 T + b3 YCO2 )b2 ωk ∆hk )

is a density distribution function, b1 , 
b2 and b3 are scaling constants, and c
Nc
is a normalisation factor that makes i=1 Li = 1 the integral (8) over the
whole integration domain. As a consequence, convective terms are artificially
restored, and the mesh tends to increase its resolution in regions with steep
gradients of CO2 concentration and temperature and local peaks of heat release. A
detailed explanation of the scheme and the corresponding pseudo-code is presented
in [10, 12].
WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 73

Figure 2: Conceptual model.

7 Results
The ICE-ALE FV method has been applied to the combustion of a DBP providing
inlet flow conditions to the two-dimensional domain shown in Fig. 2. The
geometrical parameters are length L = 16 cm, inflow and outflow aperture
Din = Dout = 5 cm, convergent and divergent angle α = 30◦ . The throat width
and curvature radius are D = 3 cm and R = 2 cm, respectively.
The burning surface is located at x = 0 where the FV version of matching
conditions (5), (6) are prescribed. The cold end of the DBP has T = 293K,
YDB = 1, and Yk = 0 for k = DB. The top and bottom surfaces are considered
as adiabatic walls. The corresponding boundary conditions are implemented by
mean of phantom cells and vertices. Smooth outflow boundary conditions like
∂(ux , ε, Yk )/δx = 0 are prescribed at the gas exit. The initial conditions are
YDB = 1,YN2 = 0, T = 293K at the DB propellant, and YDB = 0,YN2 = 1,
T = 650K at the gas phase. An external source of heat has been added in order to
maintain a temperature of 650K during 1ms so that pyrolisys and decomposition
endotermic reactions in the solid can start relatively quickly. This external source
can be regarded as a simplified representation of an hypothetical ignition system.
This transient condition is removed after the combustion becomes self sustained.
It has been observed from the numerical experiments that all transient effects
occurring near the burning front due to ignition and flame structure development
vanish after approximately t = 0.2s when the process becomes stable. Fig. 3
shows the concentration profile of all species in gas phase and CP at t = 0.2s.
The main properties of steady state flame structure is relatively insensible to
the geometry and physical conditions of the bulk of the domain, apart from the
bulk pressure which affects the characteristic length scale of the flame. The results
obtained for temperature and species concentration are in excellent agreement with
the ones previously obtained in [14] and others qualitative descriptions of the flame
structure [15].
Figure 4 shows the temperature distribution, and its development towards the
steady state regime. At time t = 0.09543 s a pseudo-steady state is achieved for
the temperature distribution.
The characteristic velocity near the burning zone is of the order of a few m/s,
and the temperature is typically 700◦ C. Therefore the Mach number is of the order
10−3 , thus representing a quite incompressible behaviour.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
74 Computational Ballistics III

1 1e-06 1e-05 0.0001 0.001 0.01 0.1

Yk 0.45
CO
NO 0.4

<---------------- Burning surface (x=0) ---------------->

CO2
H2O 0.35
0.8 COND. H2
PHASE GAS PHASE 0.3
C2H4
N2 0.25
C
0.2
0.6
0.15
DB 0.1
(CHO)2
CH2O 0.05
0.4 NO2
0
0.35
(CHO)2 0.3
CH2O
NO2 0.25
0.2
0.2
0.15
0.1
0.05
0 0
-0.05
1e-5 2e-6 1e-06 1e-05 0.0001 0.001 0.01 0.1
x [m] x [m]

Figure 3: Species concentration profile near the burning surface in gas and
condensed phases at t = 0.2s.

Figure 4: Temperature at different time levels after the ignition.

In order to test the ability of the ICE method to deal with chemically reacting
subsonic, transonic and supersonic regime at the same time, the DBP was replaced
by a constant inflow boundary condition supplying hot reactive gases at 450 m/s.
Note that this is an extraordinary high value in comparison with the typical
velocity found in the gas phase near the burning surface. However, it is useful
as a benchmark test for the hydrodynamic part of the problem. The chemical
composition of the flow at the inlet has been imposed in order to match the
corresponding flame composition at steady state in the interface, found in one-
dimensional simulations (see Table 1). The temperature at the inlet section was
fixed to 700 K.
Figure 5 shows the Mach number (M = |u|/cs ) distribution in the steady state.
As expected, the transonic region is located in the narrowed section of the domain,
the convergent region is subsonic and the convergent region is supersonic.
WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 75

Table 1: Mass concentration of the inlet flow.

Species CO NO CO2 H2 O H2 C2 H4 N2 C
Y 0.384 0.003 0.318 0.157 0.009 0.003 0.126 0.0

Figure 5: Mach number in a compressible regime.

0.03
Burning Rate [m/s]

0.025
0.02
0.015
0.01 Saint-Roberts’s law
Numerical
0.005
1 2 3 4 5 6
Pressure [MPa]

Figure 6: Burning rate vs. pressure. Comparison between numerical finding and
the Saint-Robert’s law.

Figure 6 shows the burning rate plotted against the bulk pressure. The
dots represent the values explored during the numerical calculation, while the
continuous line represents the Saint-Robert’s law (Vieille’s law) given by rb =
A(p/p0 )n . The exponent n which best fits the numerical relation in a least squares
sense is n = 0.76. This value is in good agreement with the expected indices for
DBPs.

8 Conclusions
A hybrid approach based on the FVM was developed and tested in order to
study time dependent combustion problems involving homogeneous Double-Base
propellants. The method is based on an Implicit Continuous Eulerian formulation
with an Arbitrary Lagrangian Eulerian technique and mesh optimisation. The
present approach is in principle oriented to compressible flow, however, it resulted
quite satisfactory for modelling very low Mach numbers, i.e. order O(10−3 ). The
method is suitable for dealing with subsonic, transonic and supersonic regimes at
the same time. After the flame has achieved a steady state, the comparison between
Saint-Robert’s burning law and the numerical findings are very good. In addition,
the scheme has proven to be stable and provided accurate description of the flame
structure and flame dynamics near the burning surface of the propellant. Hence,
it can be used to study a variety of 2D time dependent combustion processes
including transient ignition and combustion instabilities.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
76 Computational Ballistics III

References
[1] P. P. Cord and F. Beteille. Versatility of a two chamber inflator for
airbag application. In Airbag 2004, 7th Int. Symposium and exhibition on
Sophisticated car Occupant Safety Systems, Karlsruhe, Germany, 2004.
[2] Aoki N. Kubota. Combustion wave structures of high - and low energy
double-base propellants. AIAA J., 20(1):100–105, 1982.
[3] E. Dick, J. Bulgheroni, R. Crespo, J. Padula, E. Runstein, and L. Urdiain. La
propulsion de cohetes y misiles (in spanish). Technical Report 685, CITEFA,
Bs As, Argentina, 1989.
[4] M. T. Health, R. A. Fiedler, and W. A. Dick. Simulating solid propellant
rockets at csar. AIAA Paper, pages 2000–3455, 2000.
[5] H. J. Gibeling, R. K. Madabhushi, F. J. De Long, and J. S. Sabnis. Advanced
flow field model. Technical report, U.S.Air Force Philips Lab., 1993. CRPL-
TR-3005, Edwards AFB CA.
[6] J. Warnatz, U. Maas, and R.W. Dibble. Combustion. Springer Verlag, 1996.
[7] D.R. Stull and H. Prophet. Janaf tables of thermochemical properties, 2nd ed.
Technical report, National Standard Ref. Data System, Washington, 1971.
[8] Nat. Inst. of Standards and Tech. The NIST Chemistry WebBook.
http://webbook.nist.gov/, 2005. U.S. Secretary of Commerce.
[9] Hirschfelder, Curtis, and Bird. Molecular Theory of gases and Liquids. John
Wiley & Sons, Inc., New York, 1954.
[10] A. Peratta, C. González, and E. Dick. Numerical analysis for double-base
propellant combustion. In V. Sanchez-Galvez and C.A. Brebbia, editors,
Computational Ballistics II, volume 40 of Transactions on Modelling and
Simulations, pages 269–278. Wessex Institute, WIT-Press, 2005.
[11] B. E. Poling, J. M. Prausnitz, and John P. O’ Connel. The Properties of gases
and Liquids. McGraw Hill, 5 edition, 2000.
[12] A. Peratta. CFD Aplicado a la Combustión de Propulsantes Sólidos
Homogéneos. Master’s thesis, University of Buenos Aires, Faculty of Exact
and Natural Sciences, October 2001. in spanish.
[13] V. Yang, Thomas Brill, and Wu-Zhen Ren, editors. Solid Propellant
Chemistry, Combustion, and Motor interior Ballistics, volume 185.
American Institute of Aeronautics and Astronautics, New York, 2000.
[14] Tae-Seong Roh, I-Shih Tseng, and Vigor Yang. Effects of acoustic
oscillations on flame dynamics of homogeneous propellants in rocket motors.
J. of Propulsion and Power, 11(4), July August 1995.
[15] G. Lengellé, A. Bizot, J. Duterque, and J.F. Trubert. Fundamentals of solid
propellant combustion. In K.K.Kuo and M. Summerfield, editors, Progress in
Astronautics and Aeronautics, AIAA, volume 90, page 361, New York, 1984.
[16] C.W. Hirt, A.A. Amsdem, and J.L.Cook. An arbitrary lagrangian-eulerian
method for all flow speeds. J. of Computational Physics, 14:227–253, 1974.
[17] G.I. Marchuk. Handbook of Numerical Analysis, volume 1, chapter Splitting
and alternating direction methods, pages 197–462. Elsevier, North-Holland,
Amsterdam, 1990.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 77

Stochastic study of 60-mm

gun-projectile responses
M. Chen
Army Research Laboratory, USA

Abstract
Gun propulsion modeling has been under development for many decades.
Starting from lumped parameter computer code, then 1D and 2D approaches and
finally by multi-dimensional and multi-phase interior ballistics codes, one has
been able to estimate in-bore pressure-time history at a fairly accurate level.
However, some underlying assumptions among the models exhibit certain levels
of uncertainties, for instance: the time-varying friction between obturator and
bore surface; the granular shape variations of propellant charges; the packaging
deviations of each propellant load, etc. This study was to investigate in-bore
responses of a 60-mm projectile subjected to the inherent randomness of
propulsion pressures. The pressure-time curve was modeled with 47 Gaussian
variables. A total of 100 pressure samples were then generated through Monte
Carlo simulation techniques. Subsequently, the time histories of the highest
in-bore velocity, peak acceleration and the maximum von Mises stress responses
of the projectile were obtained respectively corresponding to each pressure
simulation. The goals were to gain more understanding of how the projectile
performs in response to the changes of chamber pressures, and how much
variation of the responses should be expected from experimental results under
normal circumstances. The significance of the random variables on the projectile
responses was outlined as well.
Keywords: interior ballistics, 60-mm projectile, IBHVG2, Monte Carlo
simulation, stochastic responses.

1 Introduction
The modeling of complex gun propulsion schemes has evolved at a great stride
over the past two decades. In the mid 1980s, a lumped-parameter computer code
IBHVG2 (Interior Ballistics of High Velocity Gun, version 2), developed by the

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070071
78 Computational Ballistics III

US Army Research Laboratory (ARL, formerly Ballistic Research Laboratory),

was available for the calculation of interior ballistic trajectories including time-
dependent gas pressure, projectile displacement and velocity [1]. In early 1990s
the efforts were extended to 1-D/2-D interior ballistic modeling, where a number
of experimental studies have demonstrated space-time-dependent flame
spreading processes [2,3]. More details in determining gun distributed pressure
field were further taken into account, which led to initial development of next
generation model named NGEN [4]. The 3-D, multiphase and computational
fluid dynamics based NGEN codes have been used for a number of applications
in recent years [5,6]. Nowadays, in-bore pressure-time curve can be estimated at
a fairly accurate level.
Nevertheless, there are a few factors that are inherently uncontrollable and
exhibit certain level of variations when calculating in-bore pressure history.
Some of the examples are time-varying frictions between obturator and bore
surface, gun barrel manufacturing tolerance, granular shape variations of
propellant charges, packaging deviations of each propellant load, distinctive
flame spreading path, changing ambient temperature, etc. Many experimental
shootings have demonstrated the pressure variations, where the differences from
one shot to another may be attributed to one or more of these factors.
Understandably, the shot deviations would become even more apparent at war
field. Thus, it is important to study how the performance of a projectile is
influenced by the stochastic excitations. The objective of this paper was not to
model the random nature of each factor individually. Instead, the variations of all
the contributing factors were considered as a whole. A nominal pressure-time
curve, where the pressure level at each time step was modeled as a Gaussian
variable, was utilized for this study.

2 Description of the 60-mm projectile system

A 60-mm gun projectile was chosen for the study. The projectile possessed a
total length of 317 mm from nose to tail and an outer diameter of 23.5 mm in the
body. It was equipped with a windscreen and a penetrator in the front, having an
ogive length and radius of 70.5 mm and 1,380 mm, respectively. Four fins for
stabilization were embedded in the tail with fin span of 50 mm. Figure 1
illustrates material configuration from a cross-sectional view of the projectile
system. The sabot and windscreen cover were composed of 7075-T651
aluminum alloy. Tungsten material was used for the penetrator. The gun barrel,
projectile body and fins were modeled with 17-4 PH stainless steel. The inside of
its body was divided into two cavity areas. The forward cavity may carry high
explosive payload while the rear cavity was designed to accommodate electronic
equipments. The projectile was intended to hit and destroy hostile objects, such
as mortars, rockets and artillery. Because of the precision-sensitive mission, it
becomes vital to investigate the stochastic responses of the projectile system.
A 64-caliber smooth bore gun tube with an inner diameter of 60 mm was used
to simulate the projectile firing. The barrel has a total length of 3,840 mm, i.e. in
bore travel distance for the projectile. M2 propellants with geometry of 7 perf

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 79

grain were used for the propulsion. Given a chamber volume of 1.3 liter, a peak
breech pressure of 470 MPa was derived from the interior ballistics code
IBHVG2. The in-bore structural dynamic analysis of the projectile system except
optimized sabot has been previously performed [7]. For computational efficiency
in stochastic study, the windscreen and stabilized fins were substituted with
equivalent weight such that the center of gravity of the projectile system
remained at the same location. The simplification that avoided very fine mesh
significantly reduced computational time. The projectile configuration and grids
are displayed in Figure 2. This model contained solely hexahedral elements with
a total number of 42,984. The entire mass of the projectile system including
sabot was approximately 1 kg.

Fin (Steel) Body

(Steel) Nose
(Electronics)

Windscreen
(Aluminum)

Cavity Sabot Penetrator

(Electronics) (Aluminum) (Tungsten)

Figure 1: Material configuration of a 60-mm projectile.

Figure 2: Display of 60-mm projectile geometry and mesh.

3 Stochastic modeling
A nominal base pressure-time curve shown in Figure 3 was adopted for the
study. A peak pressure of 315 MPa occurred at 2 ms from ignition and the total
pressure duration was 4.7 ms. Given the random nature of pressure generation
through propellant burning process, the pressure level at each time step was

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
80 Computational Ballistics III

considered as a normally distributed function. As a result, a total of 47 Gaussian

variables were required to represent the profile of the stochastic pressures. This
modeling may account for the fact that the measured pressure level of every
experimental shot differed even with the same gun, the same cartridge and the
same projectile. Because the differentiation was not substantial, a coefficient of
variation of 3% for the random variables was employed [8]. Note that no
additional energy was imposed on the projectile system as opposed to the
original pressure curve since the ensemble mean remained the same.
To better illustrate the pressure deviations, Figure 4 shows the Gaussian
distributions of the base pressures at the time steps of 1.8, 1.9 and 2.0 ms from
ignition, respectively. It is a local magnified view in the area of Figure 3. The
middle line represents the mean value of the pressure, and the lines above and
below stand for one standard deviation from the mean. The bell shapes depict the
spread of the random variables. In this paper, HyperStudy software [9] was
adopted to perform Monte Carlo simulations. A total of 100 samples were
generated for each random variable. As a result, 100 design cases, each with a
distinctive time-pressure curve, were constituted. The pressure curves for the
first three design cases are shown in Figure 5. Due to low values in the initial
stage, the variations are not visible because of scale. The uncommon small
double-bumps in high pressure area have to be achieved with complex
configurations of propellant charges, which is beyond the scope of the paper.

350

300

250
Base Pressure (Mpa)

200

150

100

50

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
millisecond

Figure 3: A nominal base pressure-time history.

In terms of sampling distribution, Figure 6 gives a histogram plot for the

random variable #20, i.e. the peak pressure at 2 ms travel time. The histogram
passed a normality test at 95% confidence level although a bit right skewed is
seen on the chart. In addition, the computed sample average was close to the
ensemble average of 315 MPa, indicating that the number of 100 samples was
sufficient, i.e. a good representation of the pressure profile. Based on the 100
sampled pressures, the whole design cases were solved with LS-DYNA on Linux

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 81

Networx Evolocity II cluster, JVN, at the ARL High Performance Computing

Center. Each analysis took approximately two hours of CPU time on 8-thread
parallel execution. In summary the gun- projectile system was subjected to the
described stochastic excitations, the system responses would be stochastic even
with a deterministic projectile system.

340

330

320

310
σ
300 µ
290
−σ
280
1.7 1.8 1.9 2 2.1
m ilis e c o nd
millisecond

Figure 4: Gaussian distributions of base pressures at the time steps of 1.8, 1.9
and 2.0 ms from ignition, respectively.
Case 1 Case 2 Case 3
350

300

250
Base pressure (MPa)

200

150

100

50

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (ms)

Figure 5: Pressure curves of the first three design cases.

4 Stochastic responses

The dispersion of in-bore projectile responses including peak velocity, peak

acceleration, von Mises stress and travel distance, was obtained. The scattering
of the values provided insight into what variations and ranges of the projectile
responses should be expected even in a normal circumstance. The derived

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
82 Computational Ballistics III

statistical in-bore data are important for the design of the projectile system and
may be further utilized for the analysis of exterior ballistics and terminal effect
evaluations.

Figure 6: Histogram of random variable #20, i.e. peak pressure distribution.

A histogram of the maximum in-bore velocity is showed in Figure 7. The

average and standard deviation of the velocity were approximately 1551 m/s and
12.4 m/s, respectively. Due to a very low coefficient of variation (COV) of 0.8%,
most computed data were centralized at the mean. Given an assumption of
Gaussian response, which is acceptable in this case, there was a 99.7% likelihood
that the axial velocity responses would fall into the range between 1514 m/s and
1588 m/s, i.e. within three standard deviations of the mean.
Figure 8 displays the distribution of peak accelerations. The mean value of
the response was 85,000 g. Unlike velocity, the accelerations exhibited high
spread, which was because of inconsistent occurrence time, i.e. the peak
accelerations occurred at the time step where the peak pressures took place. One
can see the time distinction for the peak values of the pressure curves as
demonstrated in Figure 5. As a result, the peak accelerations appeared to be a
non-Gaussian distribution. In terms of stress responses, a point of the projectile
body in which the highest stress happened was selected. The histogram of the
von Mises stress distribution of the point with a mean value of 952 MPa is
shown in Figure 9. As expected, a wide range and a non-Gaussian distribution of
the responses from 894 MPa to 1020 MPa were derived.
The response statistics of the projectile including mean, COV, minimum and
maximum values are summarized in Table 1. Note that the acceleration and
stress responses had higher COV of 3.1% and 3.0%, respectively. The second
order derivatives exhibited equivalent COV level compared with the applied
pressure. The displacement response showed a COV of 0.67%, a much smaller

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 83

Figure 7: Histogram of maximum axial velocity distribution.

Figure 8: Histogram of peak axial acceleration distribution.

Figure 9: Histogram of peak von Mises stress distribution at projectile body.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
84 Computational Ballistics III

Table 1: Response statistics of a 60-mm projectile subject to random

excitations.
Response Mean Coefficient Minimum Maximum
of Variation
Peak velocity (mm/s) 1.55x106 0.80% 1.506x106 1.583x106
Peak acceleration 8.38x108 3.1% 7.80x108 9.09x108
(mm/s2)
von Mises stress at 9.52x105 3.0% 8.94x105 1.02x106
projectile body (KPa)
Travel Distance 3770 0.67% 3699 3847
(mm)

variations as expected. The mean value of the travel distance was 3770 mm with
a standard deviation of 25.3 mm. It implies that the time duration of the pressure
curve should have been a bit longer than 4.7 ms for the projectile to reach the
muzzle. Based on the simulated results, the likelihood to travel through the barrel
within the time frame was less than 10%. The discrepancy was probably due to
the slight difference in the total mass of the launch package used in computer
modeling and IBHVG2.
The relationships among responses were of interest as well in this study.
Figure 10 provides the scatter plot of the peak acceleration against the von Mises
stress responses. Overall speaking, a positive trend is seen from the plot.
However, the level of correlation was not considerably high because the
locations where the nodal acceleration and the element stress were obtained were
not identical. Another scatter plot that represented the relationships between the
maximum axial velocity and the travel distance is given in Figure 11. As
anticipated, the data demonstrated a very strong correlation. The relationships
should help derive required muzzle velocity given a certain length of gun barrel.
It should be noted that having scatter plots between random variables and
responses are not very meaningful because the 47 random variables were not
completely independent in terms of their contributions to the total responses. No
clear pattern was found between any instant pressure variable and any of the
projectile responses.

5 Summary
A pressure-time curve that consisted of 47 data points was adopted for the
stochastic study. The pressure at each data point was considered as a normally
distributed variable. The variations of the pressure were used to account for the
uncertainties from a number of inherently uncontrollable factors that are
associated with propulsion thrust generation. A slight 3% coefficient of
variations for each random variable was employed to support the evidence that
the dispersion of in-bore projectile responses exists in daily experimental
shooting even with the same gun, the same cartridge and the same projectile
system.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 85

Figure 10: Scatter plot of von Mises stress vs. maximum acceleration.

Figure 11: Scatter plot of maximum axial velocity vs. total travel distance.

A deterministic 60-mm projectile system and a 64-cal smooth bored gun

were used. The composite projectile system was simplified in the nose and fin
areas in order to increase computational efficiency. A total of 100 distinctive
pressure curves that consisted of 47 Gaussian variables were created based on
Monte Carlo simulation techniques. The corresponding in-bore projectile
responses to each pressure sample were then obtained.
Given 95% likelihood, the highest in-bore velocity should fall into the range
from 1526 m/s to 1575 m/s, the peak acceleration from 80170 g to 90735 g, the
maximum von Mises stress of the projectile body from 894 MPa to 1009 MPa.
The variations should serve as a guideline for the design of the 60-mm gun-
projectile system. In addition, the computed response ranges should be
anticipated from experimental results even under regular shooting environment.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
86 Computational Ballistics III

Acknowledgement
The author would like to thank Dr. Lang-Mann Chang, a retiree from the
Propulsion Science Branch of the US Army Research Laboratory for his valuable
advice on the physics of gun propellant combustion.

References
[1] R. Anderson & K. Fickie, IBHVG2 (Interior Ballistics of High Velocity
Guns, Version 2) – A user’s guide, Technical Report BRL-TR-2829, US
Army Research Laboratory, Aberdeen Proving Ground, MD, July 1987.
[2] P.S. Gough, The XNOVAKTC code, Contractor Report BRL-CR-627, US
Army Research Laboratory, Aberdeen Proving Ground, MD, February
1990.
[3] C. Ruth & T. Minor, Experimental study of flame spreading processes in
155-mm, XM 216 modular propelling charges, Memorandum Report
BRL-MR-3840, US Army Research Laboratory, Aberdeen Proving
Ground, MD, June 1990.
[4] P.S. Gough, Initial development of the core module of next generation
model NGEN, Contractor Report ARL-CR-234, US Army Research
Laboratory, Aberdeen Proving Ground, MD, June 1995.
[5] M. Nusca, High-performance computing and simulation for advanced
armament propulsion, Technical Report ARL-TR-3215, US Army
Research Laboratory, Aberdeen Proving Ground, MD, June 2004.
[6] M. Nusca, A. Horst & J. Newill, Multidimensional, two phase simulations
of national telescoped ammunition propelling charge, Technical Report
ARL-TR-3306, US Army Research Laboratory, Aberdeen Proving
Ground, MD, September 2004.
[7] M. Chen, Structural Design and Analysis of Hit-to-Kill Projectile, 9th
International LS-DYNA Users Conference, Dearborn, Michigan, USA,
Session 2, 13-22 (2006)
[8] L.-M. Chang, Personal communication, 2 August 2006, Propulsion
Science Branch, Army Research Laboratory, USA.
[9] Help Manual of HyperStudy 7.0, Altair Engineering Inc. (2005)

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Section 3
Terminal ballistics
This page intentionally left blank
 Computational Ballistics III 89

On the linkage of impact damage to modeling

of ballistic performance
J. M. Wells
JMW Associates, Mashpee, MA, USA

Abstract
The assessment of terminal ballistic performance has historically been strongly
biased toward the penetration resistance of the target architecture and its
component materials. However, penetration modeling alone does not provide
sufficient knowledge to create new and/or improved armor ceramics materials
capable of mitigating and preventing penetration. It is apparent that physical
impact damage occurs prior to, and strongly affects, the occurrence and progress
of the penetration process and as such impact damage needs to be explicitly
included in ballistic performance modeling. One aspect of this situation has been
the difficulty in attaining a detailed volumetric characterization of actual
sub-surface bulk impact damage. The x-ray computed tomography, XCT,
diagnostic and 3D visualization techniques presently appear as the only effective
nondestructive evaluation, NDE, modality for high resolution volumetric impact
damage interrogation, spatial characterization, quantification, visualization, and
3D analysis. An overview of the current XCT impact damage diagnostic
capabilities and results are discussed along with some remaining challenges and
the need to incorporate 3D physical damage features into future computational
models for predictive ballistic performance.
Keywords: X-ray Computed Tomography, XCT, damage diagnostics, ballistic
impact damage, computational modelling, terminal ballistics, NDE, cracking,
embedded fragments, impact-induced porosity.

1 Introduction
The historical focus of ballistic computational modeling on the penetration
phenomenon is understandable and logical in that no friendly armor system
should be penetrated. However, penetration-only based models lack the

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070081
90 Computational Ballistics III

capability to provide the essential insight into extrinsic target architectural or

intrinsic material modifications required for the design of vastly improved
ceramic armor systems. Penetration models generally use a 2D axisymmetric
geometry approach which has proven computationally tractable and successful in
creating accurate penetration simulations. The measurement of penetration
parameters is relatively straight forward and hence the validation and verification
of penetration-only models is quite possible. Conversely, comparatively little
progress has been made in developing physical damage-based computational
models that accurately describe real ballistic impact damage.
It is apparent that physical impact damage occurs prior to, and likely affects,
the occurrence and progress of the penetration process and as such needs to be
explicitly included in ballistic performance modeling [1]. Impact damage has
received renewed attention over the past decade with the evolution and
application of non-invasive industrial x-ray computed tomography, XCT,
damage diagnostic and 3D voxel visualization techniques. XCT presently
appears as the only effective NDE modality for high resolution volumetric
ballistic impact damage interrogation, spatial characterization, quantification,
visualization, and 3D analysis and has been successfully demonstrated on armor
ceramics [1-8] and Ti-6Al-4V monolithic metallic armor materials [9]. Its
inclusion in predictive ballistic impact damage modeling [1,6] has been
proposed.

2 Background
The traditional “post-mortem” method of examining ballistic damage within the
interior of impacted opaque ceramic targets has been destructive metallographic
sectioning, polishing, and optical and electron microscopy. While permitting
damage observations at multiple length scales, this method is irreversible and is
essentially limited to observations on individual 2-D planar sections. Traditional
non-destructive examination, NDE, modalities such as digital x-ray and
ultrasound are currently more capable of damage “detection” than they are of the
detailed characterization of defect size, morphology, volumetric location, and the
discrimination between multiple damage variants which are superimposed in a
complex assemblage.
Thus an essential need exists for a diagnostic in-situ approach which can
provide the qualitative and quantitative characterization details of actual impact-
induced damage. The most satisfactory “post-mortem” solution demonstrated to
fulfil this need to date is industrial x-ray computed tomography, XCT. Yet, other
than the work of the author and/or his collaborators [1-9], no other published
application of this approach for ceramic impact damage diagnostics is known.

3 XCT impact damage results

3.1 Impact damage
There is no universally agreed upon definition of impact damage in terminal
ballistics testing of armor ceramics. Most of the post-mortem studies using

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 91

destructive metallographic sectioning focus their description of impact damage

on traditional types of radial, ring, conical, and laminar cracking. These
observations are made on a limited number of 2D planar sections and do not
provide a full spatial characterization of these cracking types. Computational
ballistic models such as the Johnson-Holmquist [10] model invoke a less
physically descriptive, yet more mathematically convenient damage definition by
relating cumulative inelastic strain to a normalized “damage function” which
ranges from a value of 0 for no damage up to the fully damaged condition with a
value of 1.0. The use of this simplistic damage definition has worked reasonably
well for the computational model of the penetration phenomenon. Such
mathematically unsophisticated definitions of impact damage do not incorporate
the distinct differences in the various details of actual ballistic impact damage.
Granted the inclusion of such realistic details of the actual physical impact
damage is a most difficult computational modeling challenge. However, the
author considers it ultimately necessary to achieve a computational methodology
which more closely incorporates critical aspects of real impact damage features.

3.2 The XCT approach to impact damage diagnostics

XCT is a powerful and functional 3D NDE diagnostic modality which allows the
in-situ interrogation of the external and internal design surfaces and/or damage
features within the bulk of the original object. The triangulation of volumetric x-
ray absorption data is utilized to construct a completely digitized “density” map
of a solid object thus accurately representing the dimensional and structural
features of that object. Resolution levels achievable with XCT are a function of
both the object’s size and density, and the x-ray source and detector system. The
nominal resolution level for the meso-scale tomography of a modest size
medium density laboratory ballistic target is < 0.250 mm. Higher micro-
tomography resolution levels of < 20 microns are achievable only on relatively
less dense and/or smaller objects. If higher resolution levels are still required,
then destructive sectioning and electron microscopy techniques may augment the
results of XCT. After digitization of the physical target by XCT scanning,
subsequent interrogations of the impacted target are then conducted in the virtual
diagnostic domain. The original XCT scan data file is reconstructed into a virtual
3D solid object using an advanced voxel analysis and visualization software
package [11]. Several sophisticated image processing tools and routines are then
used to obtain the XCT diagnostic results that contribute to the enhanced
understanding of various physical impact damage details. The XCT diagnostic
results are reproducible, accessible, and amenable to digital archival file storage.

3.2.1 Impact damage observations in TiB2

An earlier ballistic experiment conducted in 1996 at the US Army Research
Laboratory involved a target of a short cylindrical TiB2 ceramic measuring
72 mm in diameter by 25 mm in thickness [2,4]. Examples of some of the XCT
diagnostic results obtained on this sample are shown in figures 1 to 4. Various
impact surface and near-surface impact damage feature observations are shown
in figure 1. Perhaps the most dramatic damage features shown here are the three

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
92 Computational Ballistics III

distinct circular “steps” which surround the large impact cavity. These steps
appear to be an agglomeration of intermixed debris from both the impacting
projectile and the target ceramic. The height of these steps decreases from
~3.5 mm for the inner circular step to ~ 1.2 mm for the outermost step. Their
grey levels directly reflect their density which was found to be intermediate
between than that for the projectile and for the TiB2 ceramic. Figure 2 reveals a
large tungsten alloy projectile fragment mass and the segmented cracking impact
damage features observed in a virtually transparent 3D view.

Figure 1: Examples of surface and near surface damage features.

Figure 2: Segmented and virtual transparent views of a large projectile

fragment mass and impact cracking.
Further investigation of the large embedded tungsten alloy projectile fragment
revealed that it is not a homogeneous mass of projectile composition, but rather
is a complex agglomeration of multiple small tungsten alloy fragments

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 93

intermixed with debris from both the ceramic rubble and fine rubble from the
projectile. The intermixed debris is identified from its grey level (primarily a
density indicator) found to be intermediate between that of the high density
projectile material and the lower density of the host ceramic material. Also, the
presence of impact-induce porosity was observed entrapped within the large
agglomerated fragment mass. In figure 3 these features are observed in the
virtual longitudinal section and the four axial slices taken at locations from
above the exit face (A-20) to below the impact face (A-135).

TiB2

A-135
A-100
A-50
A-20

3D Transparent View TiB2 A-100

Segmented fragment

TiB2 A-50

TiB2 A-20

Half-section vertically Axial Section Views

Figure 3: Fragment details revealing the multiple smaller w-alloy fragments

(white) with enclosed porosity and intermixed ceramic debris
(darker grey areas).

Figure 4: Three separate impact damage modes in the axial slice image on
the left; both spiral and “hourglass” shaped cracking morphology is
evident on the image on the right.

Two additional and very revealing images of damage features are shown in
figure 4. First, the image on the left is an axial slice (A-137) of the TiB2 target in
which can be seen several small projectile fragments in the center of the sample.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
94 Computational Ballistics III

Also seen in this image are several significant voids (dark grey level) about the
embedded fragments. Finally, one can distinguish four concentric circular cracks
surrounding the central voids and fragments. These circular cracks are actually
the intersections of 3D conical cracks with the axial plane of this image. The
image on the right side of figure 4 shows a profile view of a virtually surfaced
point cloud of the complex cracking morphology within the TiB2 ceramic. Two
observations are made from this unique image: first that a spiral “thread-like”
orientation of the cracking exists, and secondly that the cracking diameter is
larger at both the impact and exit surfaces and narrows near the mid-thickness of
the target. This narrowing “hourglass” shaped cracking profile likely results from
the interaction of incident and reflective shock waves and demonstrates
extremely complex cracking damage morphology.

3.2.2 Impact damage observations in Al2O3

Addition impacted ceramic targets examined include an Al2O3 target made
available for XCT diagnostic study from the earlier work of D.A. Shockey et al.
[12-14]. This target consisted of a half-cylindrical longitudinal section measuring
approximately 100 mm in height by 102 mm in width. It has a large cavity on its
impact surface and had been totally penetrated. Figure 5 includes virtual
reconstructed images of this target revealing both the virtually segmented semi-
continuous projectile fragments through the entire thickness of the target and the
presence of an inhomogeneous distribution of impact-induced porosity in close
proximity to the fragment segments. A further complexity of the fragment
material is revealed in the virtual axial slices shown on the left hand side of
figure 6. As observed in the smaller TiB2 target above, the inner structure of the
large fragment sections consists of multiple smaller tungsten alloy fragments
(whitest grey levels) and intermixed debris of ceramic and projectile fine rubble.

Figure 5: XCT virtual images showing segmented projectile fragments and

adjacent impact-induced porosity.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 95

Figure 6: The complex inner-structure of the large fragment is revealed in the

five designated virtual axial slices as shown on the left.

Figure 7: Frontal view of the B4C ceramic target reveals several

manifestations of impact damage.

3.2.3 Impact damage observations in B4C

Another impacted ceramic target sample made available from the work of
Shockey et al. [12-14] was a longitudinally sectioned half-cylindrical target of
B4C ceramic measuring approximately 113 mm in height by 103 mm in width.
This target has a modest sized cavity on its impact surface and had also been
fully penetrated by the high energy kinetic energy projectile. Several damage
features are observed on the virtual rendered frontal section shown in figure 7
including multiple high density (bright white) projectile fragments, medium grey

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
96 Computational Ballistics III

intermixed areas of B4C ceramic and finer tungsten alloy debris, and enclosed
porosity (fully dark grey level). Impact-induced cracking, although in somewhat
low contrast, is also observed. The large dark horizontal bands appearing in
figure 7 occur at the identical heights as the larger high density fragments and are
considered artifacts created in the original scanning process.
With the application of the virtual transparency image processing technique,
the opacity of the B4C ceramic is adjusted to zero and embedded fragment
sections are observed discontinuously through the thickness of the target. Also,
with the defect analysis tool, the presence of localized impact-induced porosity is
seen adjacent to all major projectile fragments as shown in figure 8.

Figure 8: Oblique 3D images show segmented projectile fragments from the

impact to the exit face. Images on the right show localized porosity
adjacent to fragments.

3.3 Discussion of observed impact damage features

With the information developed from the above three impacted ceramic targets
samples, as well as from additional targets examined by the author but not
described herein, several somewhat revealing impact damage observations can
be made as follows:
a. Several types of physical ballistic impact damage occur in terminal
ballistic ceramic targets, including complex cracking morphologies,
impact-induced voids and porosity, and multiple embedded tungsten alloy
projectile fragments smaller than first apparent.
b. Additional mesoscale cracking morphologies beyond the traditionally
reported ring, radial, conical, and laminar cracking types have been
observed as spiral cracking, and with multiple superimposed 3D
hourglass-shaped cracking. Such 3D cracking morphologies are generally
asymmetric and are not unrelated to traditional cracking forms normally
viewed in 2D. Some limited success has been achieved to date with the
segmentation of ensemble cracking morphologies. Fragmentation of the
ceramic targets is observed but in-situ metrology of the ceramic fragments
has yet to be obtained.
c. Embedded residual projectile fragments have been segmented and
observed in isolation from the opacity of the surrounding host ceramic.
Such embedded fragments are observed to have an inner structure

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 97

considerably more complex than simply a monotonic tungsten alloy only

segment. In addition to multiple smaller constituent projectile fragments, a
surrounding volume of intermixed fine debris consisting of both ceramic
and projectile rubble is frequently observed. Porosity is often found within
the intermixed fragment masses. Such features can be segmented and in-
situ size metrology provides dimensions, surface area, and volume of
designated fragment masses.
d. Impact-induced porosity is observed in many instances adjacent to
fragment masses along the center line of the projectile path. There is a
considerable size scale distribution to this porosity and it is generally
inhomogeneous throughout the bulk of the target ceramic. Similar
porosity is observed along cracking boundaries as well. The size and
location of individual pores are obtainable with XCT diagnostics as well
as the spatial and statistical distribution of pore sizes.
e. Initial developments in the quantification and 3D mapping of the local
damage fraction has been successful for the case of axisymmetric damage
[4]. However, the observed damage features are generally asymmetrical,
and further development work is needed for the 3D asymmetrical mapping
of impact damage features. Also, the process of creating the 3D mapping
of the impact damage needs to be simplified and its construction
accelerated.

4 Summary
XCT diagnostics were demonstrated to provide unprecedented results of the in-
situ characterization and visualization of complex ballistic impact damage
features. Future developments and refinements of this technology are
realistically anticipated. The need exists for this damage diagnostics technology
to be integrated into future damage-based computational modeling efforts. The
most feasible way for this to occur is to establish collaborative interchanges
between the practitioners of both disciplines. Such collaborations can maximize
the functionality of both the content and the format of impact damage diagnostic
knowledge and accelerate its incorporation into more realistic predictive models.

Acknowledgements
Acknowledgements are gratefully extended to W.H. Green of ARL for the XCT
scans, Dr. D. A. Shockey of SRI for making the Al2O3 and B4C targets available,
and Dr. C. Reinhart of Volume Graphics GmbH for assistance with the
StudioMax v.1.2.1 voxel analysis and visualization software.

References
[1] J.M. Wells, On the Role of Impact Damage in Armor Ceramic
Performance. Proc. of 30th Int. Conf. on Advanced Ceramics &
Composites-Advances in Ceramic Armor, (2006), (In Press)

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
98 Computational Ballistics III

[2] J.M. Wells, Progress in the Nondestructive Analysis of Impact Damage in

TiB2 Armor Ceramics. Proc. of 30th International Conf. on Advanced
Ceramics & Composites-Advances in Ceramic Armor, 2006, (In Press)
[3] J. M. Wells, Progress on the NDE Characterization of Impact Damage in
Armor Materials. Proc. of 22nd Int. Ballistics Symp., ADPA, v2, pp. 793-
800, 2005.
[4] H.T. Miller, W.H. Green, N. L. Rupert, and J.M. Wells, Quantitative
Evaluation of Damage and Residual Penetrator Material in Impacted TiB2
Targets Using X-Ray Computed Tomography. 21st Int. Symp. on
Ballistics, Adelaide, Au, ADPA, v1, pp. 153-159, 2004.
[5] J. M. Wells, N. L. Rupert, and W. H. Green, Progress in the 3-D
Visualization of Interior Ballistic Damage in Armor Ceramics. Ceramic
Armor Materials by Design, Ed. J.W. McCauley et al., Ceramic
Transactions, v134, ACERS, pp. 441-448, 2002.
[6] J.M. Wells, On Incorporating XCT into Predictive Ballistic Impact
Damage Modeling. Proc. of 22nd Int. Ballistics Symp., ADPA, v2, pp.
1223-1230, 2005.
[7] J.M. Wells, On Continuing the Evolution of XCT Engineering
Capabilities for Impact Damage Diagnostics, Proc. 31st Intn’l Conf. on
Advanced Ceramics & Composites, ACERS, 2007, In Press.
[8] J.M. Wells, N.L. Rupert, W.J. Bruchey, and D.A. Shockey, XCT
Diagnostic Evaluation of Ballistic Impact Damage in Confined Ceramic
Targets. 23rd Intn’l Symp. on Ballistics, Tarragona, Spain, ADPA v2,
2007, In Press
[9] J.M. Wells, W.H. Green, N.L. Rupert, J. R. Wheeler, S.J. Cimpoeru, and
A.V. Zibarov, Ballistic Damage Visualization & Quantification in
Monolithic Ti-6Al-4V with X-ray Computed Tomography. 21st Int. Symp.
on Ballistics, DSTO, Adelaide, Australia, ADPA 1, pp. 125-131, 2004.
[10] T.J. Holmquist and G.R. Johnson, Modeling Prestressed Ceramic and its
Effect on Ballistic Performance. Int. Jnl. of Impact Eng’g, 31, pp. 113-
127, 2005.
[11] Volume Graphics StudioMax v1.2.1, www.volumegraphics.com
[12] D.A. Shockey, A.H. Marchand, S.R. Skaggs, G.E. Cort, M.W. Burkett and
R. Parker, Failure Phenomenology of Confined Ceramic Targets and
Impacting Rods. Int. Jnl of Impact Eng’g, 9 (3), pp. 263-275, 1990.
[13] D.A. Shockey, D.R. Curran, R.W. Klopp, L. Seaman, C.H. Kanazawa,
and J.T. McGinn. Characterizing and Modeling Penetration of Ceramic
Armor. ARO Report No. 30488-3-MS, 1995.
[14] D.A. Shockey, A.H. Marchand, S.R. Skaggs, G.E. Cort, M.W. Burkett and
R. Parker, Failure Phenomenology of Confined Ceramic Targets and
Impacting Rods, Ceramic Armor Materials by Design, Ed. J.W. McCauley
et al., Ceramic Transactions, v134, ACERS, pp. 385-402 2002.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 99

A method to increase the tip velocity of a

shaped charge jet using a hollow cavity
D. R. Scheffler & W. P. Walters
U.S. Army Research Laboratory,
Weapons and Materials Research Directorate, USA

Abstract
A technique is presented to increase the tip velocity of a conventional shaped
charge (SC) device by employing an air cavity wave shaper which overlaps the
liner. Thus, the device is a combination of an SC and an air cavity charge.
Two-dimensional, axisymmetric CTH simulations were used to design the
warhead. CTH is a family of computer programs developed at Sandia National
Laboratories for modeling solid dynamics problems involving shock-wave
propagation, multiple materials, and large deformations in one, two, or three
dimensions. The baseline case was a standard 7.5-cm-diameter liner. The liner
is basically a 42° copper, conical liner with a tip velocity of 8 km/s. The air
cavity, which overlaps the liner, increases the tip velocity of the coherent portion
of the liner to about 10 km/s, with hypervelocity jet particles traveling in front of
the tip with a velocity of about 14 km/s.
Keywords: shaped charge jet, wave shaper, numerical simulation, hydrocode,
explosive, cavity.

1 Introduction
Currently used shaped charge (SC) designs (i.e., in weapon systems, oil-well
completion, or drilling operations) may be designed to provide a deep hole in a
target material and maximize crater volume. Increasing the tip velocity of the jet
increases the depth of penetration into many targets, including most metals and
geological materials. The depth of penetration is critical for most military targets
and for releasing the flow of gas or oil in the oil-well completion problem. SC
performance, i.e., depth of penetration, may also be improved by using a wave
shaper. A wave shaper is a device that can contour or shape the detonation

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070091
100 Computational Ballistics III

wave, causing it to impact the liner at a more favorable angle, namely nearly
normal or perpendicular to the liner. Wave shapers can thus be used to alter the
impact angle of the detonation wave to improve performance of the SC. They
can also be used to decrease the amount of explosive required to form a fast jet.
Wave shapers have been made of many materials, including metals, plastics,
concrete, and air. Usually, the intent is to contour the detonation wave to move
around the wave shaper to impact the liner at a favorable angle. Here, a hollow
wave-shaper concept is examined where the hollow volume overlaps the apex of
the SC liner. The goal is to produce low-mass, high-velocity particles to
simulate micrometeorite particles which may impact spacecrafts. A series of five
hydrocode simulations are presented, where one simulation is a baseline SC
without a hollow cavity to which all other simulations are compared. The SC
concept that uses a hollow cavity overlapping the apex of the SC liner is further
detailed in Walters and Scheffler [1].
The simulations were performed using the CTH hydrocode [2], a state-of-the-
art, second-order accurate, Eulerian hydrocode developed by Sandia National
Laboratories. This code, which is capable of solving complex problems in shock
physics in one, two, or three dimensions, provides several constitutive models,
including an elastic-perfectly plastic model with provisions for work hardening
and thermal softening, the Johnson–Cook model [3], the Zerrilli–Armstrong
model [4], the Steinberg–Guinan–Lund model [5,6], an undocumented power-
law model, and others. High-explosive detonation can be modeled using the
programmed burn model, the Chapman–Jouguet volume burn models, or the
history variable reactive burn model [7]. Several equation-of-state (EOS) options
are available, including tabular (i.e., SESAME), analytical (ANEOS), Mie-
Grüneisen, and Jones–Wilkins–Lee (JWL) [8]. Material failure occurs when a
threshold value of tensile stress or hydrostatic pressure is exceeded. In addition,
the Johnson–Cook failure model [9] is also available. When failure occurs in a
cell, void is introduced until the stress state of the cell is reduced to zero.
Recompression is permitted. To reduce the diffusion typically encountered in
Eulerian simulations, several advanced material interface tracking algorithms are
provided, including the high-resolution interface tracking algorithm (available
for two-dimensional [2-D] simulations only), the simple line interface
calculation algorithm [10], and the Sandia-modified Youngs’ reconstruction
algorithm [11].

2 Simulation setup
A series of five 2-D simulations using the March 1999 version of the CTH
hydrocode were performed using cylindrical symmetry for each of the SC
geometries shown in Figures 1–5. Figure 1 shows the baseline SC. The baseline
SC consists of a copper liner that has an apex angle of 42°, a liner thickness of
0.225 cm, and a base diameter of 7.5 cm. The SC liner had a mass of 232 g. The
unconfined explosive was 78/22 Octol; the JWL EOS was used. Explosive
detonation was accomplished using a simple programmed burn model and point
initiated at the rear of the SC.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 101

All simulations employed axisymmetry and used the same computational

mesh. The mesh consisted of 207 × 2163 cells, with the origin of the coordinate
system at the base of the SC liner. A constant grid of purely square cells, with a
cell size of 0.0241 × 0.0241 cm, was used. Simulations were carried out until the
lead jet particles left the computational mesh.

3 Simulation results
Figure 6 shows the shaped charge jet (SCJ) for the baseline case. At 60 µs, the
SCJ is still coherent and has a typical shape for a jet before it particulates. The
left side of Figure 6 is a pressure plot showing only compression, and the right
side is a material plot showing only the jet material (the explosive material has
been discarded from the simulation at an earlier time). The axial velocity profile
is shown in Figure 7. Figure 7 shows the tip of the SCJ travelling at
approximately 8 km/s. Also examined was the maximum axial pressure
experienced by the SCJ, as seen in Figure 8. Figure 8 shows the maximum
pressure of 85 GPa at 20 µs. (It should be noted the axial pressure plots were
examined at 1-µs intervals and that the actual peak pressure may have been
missed.) The SCJ tip velocity and maximum pressure are used for comparing
the remaining examined charges. It was postulated that the high pressure in the
air cavity resulted in the higher jet tip velocity.

Figure 1: Baseline SC configuration. Figure 2: Case 1: SC with baseline

cavity overlapping liner.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
102 Computational Ballistics III

Figure 3: Case 2: SC with cavity wave shaper, no overlap.

Figure 4: Case 3: SC with reduced and modified cavity overlapping liner.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 103

Figure 5: Case 4: SC with modified baseline cavity overlapping liner.

Figure 6: Baseline SC at 60 µs.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
104 Computational Ballistics III

Figure 7: Axial velocity for baseline SC at 60 µs.

Figure 8: Axial pressure for baseline SC at 20 µs.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 105

Figure 9 shows the geometry and pressure plot for the Case 1 SCJ at 40 µs, as
shown in Figure 2. Case 1 shows the wave shaper cavity overlapping the SC
liner. Unlike the baseline SC, Figure 9 shows a coherent SCJ, with
hypervelocity particles ahead of the main jet tip. The axial velocity profile of the
SCJ, shown in Figure 10, shows that the tip velocity is approximately 10 km/s,
with the leading hypervelocity particles travelling much faster. The maximum
velocity of the lead particle is approximately 14 km/s. Figure 11 shows the axial
pressure profile along the jet at 12 µs, the time when maximum pressure was
observed. The maximum pressure was approximately 185 GPa. All other SC
cases with wave shapers, whether the hollow cavity overlaps the apex of the SCJ
liner or not, had similar jet profiles to that shown in Figure 9. Due to page
limitations, the results of all cases are simply summarized in Table 1.

Figure 9: Case 1 SC at 40 µs.

The Case 2 SC, Figure 3, does not have a cavity that overlaps the apex of the
copper liner. Rather, the conical cavity’s apex meets at the liner’s apex. From
Table 1, the Case 2 SCJ gives a slightly higher SCJ tip velocity than the baseline
case, namely 8.7 km/s. Though the cavity of Case 2 did not overlap the copper
liner, it still produced hypervelocity particles similar to those shown for Case 1
in Figure 9. The approximate maximum particle velocity was 12 km/s. The
maximum pressure seen for the Case 2 SCJ along its axis occurred at 9 µs and
was 135 GPa. While this charge represented an improvement in performance
over the baseline SC, it used more explosive than Case 1 and did not match the
Case 1 SC performance.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
106 Computational Ballistics III

Figure 10: Axial velocity for Case 1 SC at 40 µs.

Figure 11: Axial pressure for Case 1 SC at 12 µs.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 107

Table 1: Comparison of simulation results.

Charge Max. Particle Jet Tip Max. Pressure

Case # Mass Velocity Velocity & Time at Max.
(g) (km/s) (km/s) (GPA @ µs)
Baseline 1406 N/A 8.0 85 @ 20
1 1363 14.0 10.0 185 @ 12
2 1392 12.0 8.7 135 @ 9
3 1373 11.2 8.7 165 @ 10
4 1340 12.5 8.9 115 @ 11
N/A = not applicable.

For the Case 3 SC, the cavity again overlaps the apex of the liner and has
additional explosive removed along the liner, as shown in Figure 4. This
represents a reduced and modified cavity from Case 1 as shown in Figure 2, and
thus has more explosive mass than Case 1 (see Table 1). From Table 1, it can be
seen that while the velocity-based performance, maximum hypervelocity particle
velocity, and pressure does not match the performance of Case 1, it nevertheless
represents an improvement over the baseline SC. The jet tip velocity was the
same as Case 2, which did not overlap the liner apex; however, the maximum
hypervelocity particle velocity was less, about 11.2 km/s compared to 12 km/s.
For Case 4, Case 1 (Figure 2) was modified by removing explosive material
along the liner, as shown in Figure 5. Of all the SCs examined, Case 4 used the
least amount of explosive mass, about 1340 g, as listed in Table 1. The Case 4
SC was the second best in performance when measured in terms of maximum
particle velocity, jet tip velocity, and maximum pressure. It was exceeded only
by Case 1. However, probably none of the cases studied represented the optimal
air cavity shape.

4 Conclusions
The results of five axisymmetric simulations were presented. The first
simulation represented a standard SC without a hollow wave shaper. The second
simulation, Case 1, represented the baseline hollow wave shaper cavity which
overlaps the apex of the SC liner. While wave shapers, including hollow
cavities, have been in use for some time, it was shown that by using a hollow
cavity overlapping the apex of the liner, improvement in SCJ performance, as
measured in terms of jet tip velocity, could be obtained. The generation of
higher velocity particles preceding the jet tip could also be obtained. The
expansion of the overlapping cavity by removing explosive along the side of the
liner showed no real advantage.

References
[1] Walters, W.P. & Scheffler, D.R., Shaped charge explosive device and
method of making the same. Patent 6,983,698, 2006.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
108 Computational Ballistics III

[2] McGlaun, J.M., Thompson, S.L. & Elrick, M.G., CTH: a three-
dimensional shock wave physics code. International Journal of Impact
Engineering, 10(1-4), pp. 351-360, 1990.
[3] Johnson, G.R. & Cook, W.H., A constitutive model and data for metals
subjected to large strains, high strain rates and high temperatures. Proc. of
the 7th Int. Symp. on Ballistics, The Hague, The Netherlands, pp. 541-548,
1983.
[4] Zerilli, F.J. & Armstrong, R.W. Dislocation-mechanics-based constitutive
relations for material dynamics calculations. Journal of Applied Physics,
61(5), pp. 1816-1825, 1987.
[5] Steinberg, D.J., Cochran, S.G. & Guinan, M.W., A constitutive model for
metals applicable at high-strain rate. Journal of Applied Physics, 51(3),
pp. 1498-1504, 1980.
[6] Steinberg, D.J. & Lund, C.M., A constitutive model for strain rates from
10-4 to 106 s-1. Journal of Applied Physics, 65(4), pp. 1528-1533, 1989.
[7] Kerley, G.I., CTH equation of state package: porosity and reactive burn
models. SAND92-0553, Sandia National Laboratories, Albuquerque, NM,
1992.
[8] Lee, E.L., Hornig, H.C. & Kury, J.W., Adiabatic expansion of high
explosive detonation products. UCRL-50422, Lawrence Livermore
National Laboratory, Livermore, CA, 1968.
[9] Johnson, G.R. & Cook, W.H., Fracture characteristics of three metals
subjected to various strains, strain rates, temperatures, and pressures.
Journal of Engineering Fracture Mechanics, 21(1), pp. 31-48, 1985.
[10] Noh, W.F. & Woodward, P., SLIC (simple line interface calculation).
Lecture Notes in Physics, 59, Springer-Verlag, 1976.
[11] Bell, R.L. & Hertel Jr., E.S., An improved material interface
reconstruction algorithm for eulerian codes. SAND92-1716, Sandia
National Laboratories, Albuquerque, NM, 1992.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 109

A computational
viscoelasticity/damage/plasticity model for
high-explosive materials
P. Rangaswamy & R. M. Hackett
Los Alamos National Laboratory, Los Alamos, New Mexico, USA

Abstract
It is important to be able to accurately predict the behavior of high-explosive
(HE) materials because of the sensitivity with which they respond in the
environment in which they are designed to perform, and because they are also
extremely sensitive to accidental mechanical shock, which might occur during
handling. Thus it is imperative that a capability exists to develop a constitutive
material model and simulate the response of HE materials subjected to a variety
of thermomechanical loading conditions, a response that is characterized as
highly complex. It has been demonstrated through uniaxial strain-rate
experiments conducted at the Los Alamos National Laboratory (LANL) that the
initial response of HE material is rate-dependent, with the growth of microcracks
continuing under load until the yield strength is reached. In a material such as
this, the initial stages of deformation involve both elastic and viscous
mechanisms. It has been postulated that this observed behavior could be
categorized as nonlinear viscoelasticity, coupled with continuing microcracking.
A computational material model containing viscoelasticity, internal damage
accumulation and rate-dependent plasticity has been formulated for the purpose
of simulating the behavior of these types of material systems. The formulation
has been incorporated as a user-defined material (UMAT) into the ABAQUS
finite element code. It is demonstrated in this paper how the material model can
be employed by calibrating it to the experimental response of the material Mock
900-21, a high-explosive material mock utilized at LANL.
Keywords: computational, viscoelasticity, damage, fracture mechanics,
plasticity, user material model.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070101
110 Computational Ballistics III

1 Introduction
It is important to be able to accurately predict the behavior of HE materials
because of the sensitivity with which they respond in the environment in which
they are designed to perform, and because they are also extremely sensitive to
accidental mechanical shock, which might occur during handling. Thus it is
imperative that a capability exist to model the response of HE materials
subjected to a variety of thermo mechanical loading conditions, a response that is
highly complex, requiring sophisticated material modeling to simulate.
It has been demonstrated through uniaxial strain-rate testing conducted at
LANL that the initial response of HE material, along with the yield strength, is
rate-dependent, with the growth of microcracks continuing under load until the
yield strength is reached. In a material such as this, the initial stages of
deformation involve both elastic and viscous mechanisms and are truly
viscoelastic. It has been postulated that this observed behavior could be
categorized as nonlinear viscoelasticity, coupled with continuing microcracking.
However, since it is extremely difficult to obtain experimental data to support
current nonlinear viscoelasticity theory, and since internal damage accumulation
modeling has been successfully employed in numerous circumstances in lieu of
nonlinear viscoelasticity modeling [1], a linear viscoelasticity/damage approach
has been taken here in modeling the behavior of the HE material. Also, since
certain HE materials have been observed to exhibit flow characteristics
subsequent to reaching the damage-controlled yield stress, an applicable
plasticity theory is presented and developed.
The theory presented herein is used to develop a UMAT subroutine for the
ABAQUS implicit finite element code. Mock materials are developed to
simulate the laboratory response of HE materials because the mock materials can
be conventionally tested whereas the HE materials usually cannot be. The
developed UMAT is calibrated for Mock 900-21, a mock material for the high-
explosive material PBX 9501, and uniaxial tests performed on Mock 900-21 are
simulated. The calibrations are described in detail and simulation results are
compared with test results.

2 Model formulation
The time dependence of microcrack growth in polymeric materials has been
studied extensively, both theoretically and experimentally, for quite some time.
The linear viscoelasticity/damage formulation presented here is based primarily
upon the work of Addessio and Johnson [2]. During deformation, the
distribution of microcracks in the material is assumed to be random, with the
initial distribution being exponential.

2.1 Linear viscoelasticity/damage

Using Cartesian tensor index notation, strain is given by the kinematic

relationship

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 111

1
ε ij = (u j ,i + ui , j ) , i, j = 1,2,3 (1)
2
where the ui are the material deformation variables. It can be decomposed into
deviatoric and mean components
ε ij = eij + ε mδ ij (2)
where δ ij is the Kronecker delta; the mean strain is defined by
1
ε m = ε ii (3)
3
The stress can likewise be decomposed into deviatoric and mean components
σ ij = sij + σ mδ ij (4)
where the mean stress is defined by
1
σ m = σ ii (5)
3
and is related to the mean strain through the expression
σ m = 3 Kε m (6)
where K is the bulk modulus of the material.
In a deviatoric Maxwell model, a single spring and dashpot in series, the
relationship between stress and strain is
s ij = 2Geije (7)
where G is the shear modulus, and the relationship between stress and viscous
strain rate is
s ij = 2ηeijv (8)
where η is the viscosity of the dashpot and where the dot over the variable
indicates the time rate of change of the variable. Taking the partial derivative of
Equation (7) and combining the result with Equation (8), i.e. adding elastic and
viscous strain rates, gives
sij
sij = 2Geijve − (9)
τ
where the relaxation time τ is given by
η
τ= (10)
G
For the viscoelastic solid, represented by a generalized deviatoric Maxwell
model, with the strain being common for all elements of the model and the
stresses for the individual elements being additive, i.e.
N
sij = ∑ sij( n ) (11)
n =1

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
112 Computational Ballistics III

where N is the number of elements in the generalized Maxwell model and

sij( n ) is the deviatoric stress component for the n th element, the relationship
between the deviatoric stress rate and the viscoelastic deviatoric strain rate and
the deviatoric stress, referring to Equation (9), is given by
N sij( n )
sij = ∑ (2G e − ( n ) ve
) (12)
τ (n)
ij
n =1

where G and τ are the shear modulus and relaxation time, respectively, for
(n ) (n )

the nth Maxwell element.

The cracking deviatoric strain versus deviatoric stress relationship, see
Addessio and Johnson [2], is
eijcr = βc 3 sij (13)
where c is the value of the average microcrack radius and β is a parameter that
relates the shear modulus to the initial flaw size through the expression
1
β= (14)
2Ga3
where a is the limiting crack size and
N
G = ∑ G(n) (15)
n =1
Combining Equations (13) and (14) yields a relationship between the deviatoric
strain and the average crack radius
3
c (16)
2Geijcr =   sij
a
Taking the partial derivative of Equation (16) with respect to time gives the
expression
2 3
c c c
2Geijcr = 3  sij +   sij (17)
a a a
The total deviatoric strain, the sum of the viscoelastic deviatoric strain and the
cracking deviatoric strain, is given by
eij = eijve + eijcr (18)
Combining Equations (12), (17) and (18) gives an expression for the deviatoric
stress in terms of combined viscoelastic and microcracking response:
sij = ψeij − θ ( sij + λij ) (19)
where
2G (20)
ψ=
1 + (c / a ) 3
3(c / a ) 2 (c / a ) (21)
θ=
1 + (c / a )3
and

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 113

∑
N (n)
s / τ (n)
λij = n =1 ij (22)
2
3(c / a ) (c / a )
The expression for the deviatoric stress rate for the nth Maxwell element
is given by
sij( n ) G ( n )   c  2 c c 
3
(23)
sij( n ) = 2G ( n ) eij − ( n ) −   
3 s +   sij 
τ
ij
G   a  a  a  

2.2 Fracture mechanics

The application of fracture mechanics to time-dependent materials is relatively

new [3]. An evolution equation defining crack growth rate is required.
Consistent with traditional fracture-mechanics reasoning, and following the
observations and conclusions of Dienes [4], it is assumed that the growth rate of
the average crack radius is functionally dependent upon the stress intensity. The
equations below are from Dienes [5] and Dienes and Kershner [6], with
modifications for tension and friction provided by Johnson [7]. The cracking
rate is either
m
K 
c = vres  I  , KI < K′ (24)
 K1 
or
  K0µ 
2

c = vres 1 −    , KI ≥ K′ (25)
  K I  
where
1/ 2
 3πc  , σm < 0 (compression) (26)
KI =  sij sij 
 2 
or
1/ 2
 3πc  , σm ≥ 0 (tension) (27)
KI =  σ ijσ ij 
 2 
and
1/ 2
 2
K ′ = K 0 µ 1 +  (28)
 m 
1/ m
 m (29)
K 1 = K ' 1 + 
 2
1/ 2
 πµ ' σ m c1 / 2  µ ' σ m c1 / 2  
K 0µ = K 0 1 − 1 −  (30)
 K0  K 0 
1/ 2
 45  (31)
µ' =  2 
µs
 2 (3 − 2 µ )
s 

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
114 Computational Ballistics III

where νres is the value of the rate of growth of the average crack radius Ko is the
threshold value of stress intensity, m is a cracking parameter and µs is the static
coefficient of friction.
For the N-component Maxwell model, the viscous work rate is
N N sij( n ) sij( n )
w ve = ∑ sij( n ) eijve = ∑ (32)
n =1 n =1 2G ( n )τ ( n )
and the cracking damage work rate is
1   c  c 
2 3
c
w = s e = 3  sij sij +   sij sij 
cr cr
ij ij (33)
2G   a  a a 
Employing the constitutive model formulation presented, the incremental form
of the equations, facilitating a computational solution, will be developed.
Combining Equations (4), (6) and (19) yields the expression
θσ ij + σ ij = 3K (θε m + εm ) + ψeij − θλij (34)
Rearranging Equation (34) somewhat yields the expression
 ψ
θσ xx + σ xx = θKε V +  K − εV + ψε xx − θλ xx (35)
 3
for the global x-direction on a plane x = constant, with similar expressions for
the y- and the z-direction, and
ψ
θσ xy + σ xy = ε xy − θλ xy (36)
2
for the x-direction on the plane y = constant, with similar expressions for the y-
and the z-direction, where
ε V = ε ii (37)
The factor of 2 in the denominator in Equation (36) indicates that the strains are
engineering.
Applying the central difference operator to Equations (35) and (36) yields
 θ∆ t    θ∆t  ψ 
1 + ∆σ xx =  K 1 +  −  ∆ε V + ψ∆ε xx + θ∆t ( Kε V − σ xx − λ xx ) t (38)
 2    2  3
and
 θ∆t  ψ
1 + ∆σ xy = ∆ε xy − θ∆t (σ xy + λ xy ) t (39)
 2  2
where ∆ indicates incremental change, so that the Jacobian matrix D then has
the terms
∂∆σ xx 2ψ
D(1,1) = =K+ (40)
∂∆ε xx 3(1 + θ∆t / 2)
∂∆σ xx ψ ,etc. (41)
D(1,2) = =K−
∂∆ε yy 3(1 + θ∆t / 2)
and

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 115

∂∆σ xy ψ
D(4,4) = = , etc. (42)
∂∆σ xy 2(1 + θ∆t / 2)
crack growth. In a simulation of a uniaxial strain-rate test, the value of vres is
changed until the yield stress of the simulation matches that of an actual test run
at the same strain-rate. This procedure is repeated for a test run at a different rate
of strain until a wide range of strain-rate tests have been covered. This results in
a collection of say five data points, each of which relates a value of vres to a
given value of strain-rate. A log-log plot of these data can usually be expected to
yield a linear relationship between log ( vres ) and the log of the effective strain
rate (for three-dimensionality) where the effective strain rate is given by

2 3
εeff = [(ε xx − ε yy ) 2 + (ε yy − ε zz ) 2 + (ε zz − ε xx ) 2 + (ε xy2 + ε yz2 + ε zx2 )]1 / 2 (43)
2(1 + υ ) 2
where the strains are engineering and where υ , Poisson’s ratio, is given by
ψ
K−
3(1 + θ∆t / 2) (44)
υ=
ψ
2K +
3(1 + θ∆t / 2)

2.3 Plasticity

Plasticity theory adaptable to computational modeling is generally well

developed [8, 9]. An example of combining damage growth and plasticity is
reported by Khaleel et al. [10], and the coupling of yield surface to damage is
discussed by Johansson and Runesson [11]. The yield condition modeled herein
is basically the flow rule of von Mises’ theory, the well-known Prandtl–Reuss
relationship [12]. The effective plastic strain increment ∆ε eff is defined by
p

{Q}T [D]
W  = (45)
( H + {Q} [D]{Q})
T

3  s xx s yy s zz 
{Q} =  , , , s xy , s yz , s zx  (46)
σ eff  2 2 2 
2
σ eff = [(σ xx − σ yy ) 2 + (σ yy − σ zz ) 2 + (σ zz − σ xx ) 2 + 6(σ xy2 + σ yz2 + σ zx2 )]1/ 2 (47)
2
and where H is the hardening (or softening) modulus. The Prandtl–Reuss
relationship states that plastic strain increments ∆ε { } result when the effective
p

plastic strain increment occurs under the state of stress {Q} , i.e.
{∆ε } = {Q}∆ε
p p
eff (48)
The corresponding stress increments are obtained from the relationship

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
116 Computational Ballistics III

{∆σ } = [D]({∆ε } − {∆ε p }) (49)

where
{∆σ } = {∆σ xx , ∆σ yy , ∆σ zz , ∆σ xy , ∆σ yz , ∆σ zx } (50)
It is assumed that, consistent with plasticity theory in general, the hardening
modulus depends on the strain and the strain rate [13]. The dependence upon
strain rate is expressed as a shift function with regard to a basis value of strain
rate. It has the form
s1 (εeff − εeff
0
)
aS = (51)
s2 + (εeff − εeff )
0

where εeff is the basis value of strain rate and s1 and s 2 are constants to be
0

determined from curve fitting. The dependence of the hardening modulus on

strain is expressed as
H = a1 + a2 (ε eff − aS ) + a3 (ε eff − a S ) 2 (52)
where
2 3 (53)
ε eff = [(ε xx − ε yy ) 2 + (ε yy − ε zz ) 2 + (ε zz − ε xx ) 2 + (ε xy2 + ε yz2 + ε zx2 )]1 / 2
2(1 + υ ) 2

10.0
10.0
Log10 Compressive Young's Modulus - Pa

Log10 Tensile Young's Modulus - Pa

9.5
9.5
9.0

8.5 9.0

8.0

8.5
7.5
-1
-1
0.001 sec
1.0 sec
7.0 -1
-1 0.0001 sec
0.1 sec 8.0
-1 -1
6.5 0.01 sec 0.00001 sec
-1
0.001 sec
6.0 7.5
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5
Log τ - sec Log τ - sec
10 10

Figure 1: Log compressive (left) and tensile (right) Young’s modulus versus
Log relaxation time for Mock 900-21.

3 Application
The formulation thus presented has been used to develop a User MATerial
Subroutine (UMAT) for the ABAQUS implicit finite element code. Mock
materials are developed to simulate the response of HE materials because the
mock materials can be conventionally tested whereas the HE materials usually
cannot be. For example, Mock 900-21 is designed to simulate the strength and

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 117

density properties of the high-explosive material PBX 9501. The

theory/formulation described herein is applied to Mock 900-21 for the purpose of
validating the material model.

9
3.5 10
9 3.5 10

9
3.0 10 9
3.0 10
K G
Shear Modulus - Pa

9
2.5 10

Shear Modulus - Pa
9
2.5 10
9
2.0 10
9
9
2.0 10
1.5 10

9 9
1.0 10 1.5 10

8
5.0 10 9
1.0 10

0
0.0 10
-8 -6 -4 -2 0 2 4 8
5.0 10
Log τ - sec -8 -6 -4 -2 0 2 4
10
Log τ - sec
10

Figure 2: Log compressive (left) and tensile (right) Young’s modulus versus
Log relaxation time for Mock 900-21.
-1.0
0.05000

y = -1.1914 + 0.92119 x
-1.5
0.04000
Log10Vres - m/sec

-2.0
Vres
Vres - m/sec

0.03000

-2.5

0.02000
-3.0

0.01000
-3.5 Vres

0.000
-4.0 0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012
-3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5
-1
-1 Effective Strain Rate - sec
Log Effective Strain Rate - sec
10

Figure 3: Relationship between log of rate of growth of crack radius and log
of effective strain rate in compression (left) and tension (right) for
Mock 900-21.

Compressive strain rate data were obtained from tests conducted by Cady
[14], and tensile strain-rate test data were obtained from tests conducted by
Thompson [15]. Compressive tests were for rates of 0.001, 0.01, 0.1 and 1.0 per
second. Tensile tests were for rates of 0.00001, 0.0001 and 0.001 per second.
These tests were used exclusively to determine the viscoelastic, damage and
plasticity parameters employed in the material model to predict the response of
the Mock 900-21. Each set of uniaxial strain-rate data was converted from a true
stress versus true strain data set to a Young’s modulus (Pa) versus relaxation

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
118 Computational Ballistics III

time (sec) data set. The relaxation time was obtained by dividing strain by strain
rate. Each of these data sets was then converted to a log Young’s modulus (Pa)
versus log τ (sec) data set. These data are shown plotted in Figure 1. Fitting the
upper-most data points in Figure 1 yields expressions for the compressive
relaxation modulus and the tensile relaxation modulus, respectively.
With a value of Poisson’s ratio of 0.30, and 10 generalized deviatoric
Maxwell model elements, each separated by a decade of relaxation time, as
shown in Figure 2, the linear viscoelastic models for the material are established.
It was found, and is thus noted, that the elastic/viscoelastic material properties
for Mock 900-21 are different in compression and tension. For the compressive
strain-rate tests, using the procedure described earlier, a linear relationship
between log (vres ) ) and log (εeff ) was established for Mock 900-21. This is
shown in Figure 3. The yield behavior of Mock 900-21 in compression and in
tension is also quite different, as one might expect. For tensile strain-rate tests,
again using the procedure described earlier, a relationship between vres and
εeff was established; this relationship is shown in Figure3. The numerical values
of the plasticity parameters found in Equations (51) were s1 = −4.0322 x10−5 ,
s2 = 1.1178 x10−3 sec-1, and εeff
0
= 0.001 sec-1, and the values of those found in
Equations (54) were a1 = 4.7863x103 Pa, a2 = 6.1878 x106 Pa, and
a3 = 2.1976 x10 Pa. 9

Finally, comparisons between model predictions and test results are shown in
Figures 4 and 5, Figure 4 for compressive loading and Figure 5 for tensile
loading.

7
2.5 107 2.5 10

1.0 sec-1 -1
-1 1.0 sec
2.0 10
7 0.1 sec 2.0 10
7
-1
-1 0.1 sec
0.01 sec -1
-1 0.01 sec
0.001 sec
T ru e S tre s s - P a

T ru e S tre s s - P a

-1
1.5 10
7
1.5 10
7
0.001 sec

7 7
1.0 10 1.0 10

6 6
5.0 10 5.0 10

0 0
0.0 10 0.0 10
0 0.02 0.04 0.06 0.08 0.1 0.12 0 0.02 0.04 0.06 0.08 0.1 0.12

True Strain True Strain

(a) (b)

Figure 4: Uniaxial compressive stress-strain relationship for Mock 900-21:

(a) test data, (b) simulation.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 119

6
3.0 10
3.0 106
-1
-1
0.001 sec
0.001 sec 2.5 10
6
-1
2.5 106
-1 0.0001 sec
0.0001 sec
-1
-1 0.00001 sec
0.00001 sec 2.0 10
6

T ru e S tre ss - P a
6
T ru e S tre s s - P a

2.0 10

6
1.5 10
6
1.5 10

6
1.0 10
6
1.0 10

5
5.0 10
5
5.0 10

0
0.0 100
0.0 10 0 0.0005 0.001 0.0015
0 0.0005 0.001 0.0015
True Strain
True Strain
(b)
(a)

Figure 5: Uniaxial tensile stress-strain relationship for Mock 900-21: (a) test
data, (b) simulation.

4 Conclusions
Results obtained from the analyses performed using this material model for the
Mock 900-21material indicates that the model performs satisfactorily. Since
Mock 900-21 exhibits flow characteristics subsequent to reaching its damage-
controlled yield strength in tension, the plasticity theory developed for
computational modeling in conjunction with the coupled viscoelastic/damage
model was applicable to this material.

References
[1] Schapery, R.A., Nonlinear Viscoelastic Solids, International Journal of
Solids and Structures, 37, pp 359-366, 2000.
[2] Addessio, F.L. & Johnson, J.N., A Constitutive Model for the Dynamic
Response of Brittle Materials, Journal of Applied Physics, 67(7), pp.
3275-3286, 1990.
[3] Anderson, T.L., Fracture Mechanics, Fundamentals and Applications,
Second Edition, CRC Press, Inc., pp. 31-99, 205-220, 1995.
[4] Dienes, J.K., A Unified Theory of Flow, Hot-Spots, and Fragmentation
with an Application to Explosive Sensitivity. High-Pressure Shock
Compression of Solids II, eds. L. Davison, D.E. Grady & M. Shahinpoor,
Springer Verlag: New York, Chapter 14, pp. 366-398, 1996.
[5] Dienes, J.K., Strain-Softening via SCRAM, Los Alamos National
Laboratory Report, LA-UR-98-3620, 1998.
[6] Dienes, J.K. and Kershner, J.D., Multiple-Shock Initiation via Statistical
Crack Mechanics, Proceedings of the 11th Symposium on Detonation, eds.
J.M. Short & J.E. Kennedy, Snowmass, Colorado, USA, 1998.
[7] Johnson, J.N. Personal communication, May 2004, Technical Staff
Member, Los Alamos National Laboratory, Los Alamos, USA.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
120 Computational Ballistics III

[8] Lubliner, J. Plasticity Theory, Macmillan Publishing Company, 1990

[9] Simo, J.C. & Hughes T.J.R., Computational Inelasticity, Springer-Verlag
New York, 1998.
[10] Khaleel, M.A., Zbib, H.M., & Nyberg, E.A., 2001, Constitutive Modeling
of Deformation and Damage in Superplastic Materials, International
Journal of Plasticity, 17, 277-296.
[11] Johansson, M. & K. Runesson, K., Viscoplasticity with Dynamic Yield
Surface Coupled to Damage, Computational Mechanics, 20, pp 53-59,
1997.
[12] Steeb, H. and E. Ramm, A General Framework for Local Error Estimation
Applied to Material Nonlinear Problems, European Congress on
Computational Methods in Applied Sciences and Engineering, ECCOMAS
2000, Barcelona, 11-14 September, p. 1-14, 2000.
[13] Fish, J. & Shek, K., Computational Plasticity and Viscoplasticity for
Composite Materials and Structures, Composites Part B, 29B, pp613-619,
1998.
[14] Cady, C.M. Personal communication, Aug 2004, Technical Staff Member,
Los Alamos National Laboratory, Los Alamos, USA.
[15] Thompson, D.G., Personal communication, Aug 2004, Technical Staff
Member, Los Alamos National Laboratory, Los Alamos, USA.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Section 4
Experimental
mechanics/ballistics
and field testing
This page intentionally left blank
 Computational Ballistics III 123

Offshore platforms survivability to

underwater explosions: part I
A. A. Motta1, E. A. P. Silva2, N. F. F. Ebecken2 & T. A. Netto2
1
Brazilian Navy Research Institute, Rio de Janeiro, Brazil
2
Federal University of Rio de Janeiro, Rio de Janeiro, Brazil

Abstract
Recent developments in the Brazilian oil industry led to the necessity to conduct
offshore platforms underwater explosions (undex) survivability studies. The
ongoing research has been segmented in parts, including theoretical and
experimental correlated studies. Part of this study involves computer simulation,
and, therefore, the necessary validation of the developed models used in such
simulations. The existence of benchmark problems helped to establish a
guideline to be followed in the study of underwater explosions effects on
structural elements and simple and well-controlled experiments are being
conducted to be used to calibrate the computer models.
In this work, one of these benchmark problems is presented, along with its
experimental implementation: a submerged aluminum cylinder submitted to the
undex effects. This problem was proposed by Kwon and Fox in 1993. Results
will be collected for the direct shockwave load, the bubble-produced shockwave
load, and for both, although the primary interest relies on the shockwave effects.
Keywords: underwater explosion, offshore platforms, simulation, survivability.

1 Introduction

The underwater explosions (undex) terminal effects on semi-submerged steel

structures, such as offshore platforms, is a very complex problem, in which
individual parts failure can lead to buoyancy or stability problems that can result,
in extreme cases, in the loss a platform or in irreversible environmental damage.
Since the cost scale in this segment is in order of millions of US dollars, the
savings are obvious, and it is imperative, thus, to understand how undex can

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070111
124 Computational Ballistics III

affect these intricate structures and how design modifications can be introduced
in order to improve their survivability probability. This problem was presented in
[2].
The Submarine Technology Laboratory of the Federal University of Rio de
Janeiro started the study of such phenomenon. General guidelines were
established based on the available information in the literature and recent
developments in current related research. It was decided that the starting point of
this study would be on-site small-scale undex experiments, which are currently
under development. A tank has been built and the necessary instrumentation is
being assembled to perform the tests, similar to available benchmark problems,
which are to be conducted in conjunction with computer simulation using
Abaqus [3], providing validation material.
In this work, the problem presented in reference 1 is outlined, and its current
experimental implementation is presented. Although there are no results
available to be presented at this time, it is expected that important data will result
from it, allowing the research to move on to more elaborate problems.

2 Underwater explosions and exponential approximation

Underwater explosions and their effects on submerged structures were
approached in [2]. When an undex happen, a shockwave is instantaneously
produced, spreading in all directions. The resulting detonation products form a
bubble that pulses, expanding and contracting, while moving toward the surface,
and emitting a pressure pulse every time there is an inversion from expansion to
contraction and vice-versa. Detailed information on undex and underwater
shockwaves are given by Cole [4] and other studies, [5–11], among many other
references available in the literature.
As shown in [2], the exponential approximation is widely accepted to
represent the shockwave produced in an undex. Various models have been
proposed by different authors, all leading to very similar results. In this
approximation, the shockwave is represented by an impulse (in which the
amplitude is distance-dependent) with exponential decay in time, according to:

P(t) = Pmáx * e -( t / τ ) [Pa] (1)

where:
Pmáx = K1 * ( W1/3 / d ) * A1[Pa] (2)

τ = K2 * W1/3 * ( W1/3 / d ) * A2 [s] (3)

W - is the explosive mass, in TNT-equivalent

τ - time constant
d - distance from the explosion source
K1, K2, A1, A2 - parameters suggested by the author
Models have been introduced for various types of explosives and media, and
the one adopted in this work was the model presented by Felippa et al [5] and
Shin and Geers [6].

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 125

3 The reference problem

The model problem herein presented is derived from the one proposed by Kwon
and Fox [1]. In the original problem, a submerged T6061-T6 aluminum cylinder
is exposed to the shockwave produced by the detonation by 60 lb of HBX-1,
positioned at a 7.62 m distance from the cylinder surface. The cylinder
particulars are: overall length of 1.067 m; outside diameter of 0.305 m; wall
thickness of 6.35 mm, and 24.5 mm thick welded endcaps. The cylinder is
suspended horizontally in a 40 m deep fresh water test quarry, and the cylinder
and the explosive charge are both placed at a depth of 3.66 m. These distances
were selected so that cavitation and the bubble pulses can be neglected, as well
as reflected waves on the quarry walls and on the free surface. The test profile is
presented in figure 1.

3.66 m

cylinder explosive
7.62 m charge

Figure 1: Undex test profile proposed by Kwon and Fox [1].

18.00
16.00
14.00
12.00
Pressure (MPa)

10.00
8.00
6.00
4.00
2.00
0.00
0.0000 0.0005 0.0010 0.0015 0.0020 0.0025
Time (s)

Figure 2: Pressure load curve obtained by Kwon and Fox [1].

A series of strain gauges are placed on the cylinder outer surface and the
experimental data is filtered in 2,000 Hz. Two pressure transducers are used to
measure the shockwave intensity at 7.62 m from the source, away from the
cylinder for validation purposes. Figure 2 shows the pressure curve as
determined by the authors. The cut-off time is set to 2 ms.
In the Abaqus explicit Finite Elements model, the cylinder was discretized in
2,400 quadrilateral shell elements, and the fluid in 45,586 acoustic tetrahedral
elements, in a total of 23,337 active degrees of freedom, and the transient

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
126 Computational Ballistics III

analyses is run for 8 ms. The outer fluid boundary radius was established as 6
times the cylinder radius, which corresponds to an added mass error of about 6%
for infinite cylinders when compared to an infinite boundary. This is necessary
so that the added mass associated with the low-frequency beam bending modes
of the cylinder is represented adequately. Table 1 shows the added mass ratio for
various fluid boundaries / cylinder radius ratio. Figures 3 and 4 present the
complete mesh and more detailed information regarding the cylinder (figure
4(a)) and the fluid (figure 4(b)).

Table 1: Added mass ratio with respect to the radius ratio.

Fluid Boundary / Added Mass Ratio (External

Cylinder Radius Ratio Boundary/Infinite Domain)
1.5 2.600
2.0 1.667
4.0 1.133
6.0 1.057
8.0 1.032
16.0 1.008
24.0 1.004

Figure 3: Complete mesh for the Finite Elements implementation for the
problem proposed by Kwon and Fox [1].

4 The ongoing experiment

The ongoing experiment is very similar to the proposed by Kwon and Fox [1].
The main differences are the cylinder characteristics and material (commercial 2
inches T6061 aluminum pipe), the explosive type and amount, and the distances
involved. The media is about the same (still water).
Figure 5 shows the tank under construction where the experiments are to be
conducted. Its characteristics and location are such that will allow full control of
testing conditions, including the water temperature and salinity.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 127

(a) (b)
Figure 4: Mesh details for the Finite Elements implementation of the problem
proposed by Kwon and Fox [1], showing the fluid and the cylinder.

Figure 5: General view of the tank being built to perform the undex
experiments.
In figure 6, a general schematics view of the tank is presented. The water
mass depth will be 1500 mm, and the aluminum tube, the explosive charge (in
the center of the tank), and the transducers are to be positioned in a 750 mm deep
horizontal plane (this will maximize the time for the reflected waves to reach the
target, to be set as cut-off time for the experiments).
For this small-scale experiment, charges up to 10 g are to be used, positioned
from 200 to 700 mm away from the 1500 mm aluminum air-filled cylinder. The
cylinder ends will be welded to thick plates that will be fixed in an “inertial”
device, as sketched in figure 7, to ensure that they will be fully constrained.

5 Expected results and comments

As long as the cylinder does not collapse, it is of interest to know its residual
strain. Figure 8 presents the accumulated equivalent plastic strains (PEEQ) for
the reference problem, as simulated with Abaqus.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
128 Computational Ballistics III

Figure 6: General schematics of the tank being built to perform the undex
experiments. The aluminum tube, the explosive charge (in the
center of the tank), and the transducers are to be positioned as
shown, at half the depth of the water mass.

Figure 7: The cylinder ends are to be welded to thick plates that will be
mounted in an “inertial” device.

Figure 8: The accumulated equivalent plastic strains (PEEQ) for the

reference problem, as simulated with Abaqus.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 129

It is noticeable that important information can be obtained from the numerical

simulation, and Abaqus has shown to be suited for this purpose. The experiments
outcome will be used to validate the simulation and, hopefully, while providing
new insights about the phenomenon, they will point in the next step direction,
where more complex models will be used to provide further information, and so
on.

6 Final remarks
Structural elements response to underwater explosions are being studied to
provide insight to the more complex problem of how these impulsive loads affect
offshore platforms. An environment to perform real experiments is under
construction, and small-scale tests will be used along with numerical analysis,
starting from simple configurations and following into more complex problems
and steps are concluded.
It is expected to obtain, from this study, data enough to allow the introduction
of design modifications for new platforms to increase their survivability chances
if exposed to such loads, as well as the elaboration of contingency plans for
platforms already in use and for which modifications are not feasible.

References
[1] Kwon, K. W., and P. K. Fox, “Underwater Shock Response of a Cylinder
Subjected to a Side-On Explosion,” Computers and Structures, Vol. 48,
No. 4, 1993.
[2] Ballistics studies applied to offshore platforms, Second International
Conference on Computational Ballistics, eds. C. A. Brebbia, V. Sanchez-
Galvez, A.A. Motta, and C. E. Anderson, WIT Press: Southampton and
Boston, pp. 371-380, 2005.
[3] Abaqus v.6.5, Abaqus, Inc, (http://www.abaqus.com).
[4] Cole, R. H., Underwater Explosions, Princeton University Press:
Princeton, 1948.
[5] Felippa, C. A., Geers, T. L. and DeRuntz, J. A., The Underwater Shock
Analysis Code: A Reference Manual, Lockheed Palo Alto Research
Laboratory, Palo Alto, 1980.
[6] Shin, Y. S. and Geers, T. L., Response of Marine Structures to
Underwater Explosions, Short Course, 1993.
[7] Shin, Y. S., Underwater Explosion Research, Volume I – The Shock
Wave, Office of Naval Research, US Department of the Navy, 1950.
[8] Shin, Y. S., Application of the USA Code to Underwater Shock Problems,
Shock and Vibration Information Analysis Center, 1997.
[9] Shin, Y. S., Experimental Study of Underwater Explosion Phenomena,
NDRC, AD 268.905, Vol. 8.
[10] Motta, A. A., Underwater explosive charges optimization with respect to
their terminal effects on submerged steel structures, Dissertation, Federal
University of Rio de Janeiro, Rio de Janeiro, 2004.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
130 Computational Ballistics III

[11] Motta, A. A. and Ebecken, N. F. F, Steel structures optimization to resist

underwater shockwaves - explosive charges optimization, Structures
Under Shock and Impact VIII, eds. N. Jones and C. A. Brebbia, WIT
Press: Southampton and Boston, pp. 121-128, 2004.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 131

Ballistic tests and numerical simulations for

containment capability characterisation of
Waspaloy
alloy
M. Di Sciuva1 , M. Degiovanni1, M. Mattone1 , M. Gherlone1 ,
N. Bonora2 , A. Ruggiero2 & C. Frola3
1 Aerospace Engineering Department, Politecnico di Torino, Italy
2
Industrial Engineering Department, Università di Cassino, Italy
3 Engineering Advanced Technology, Avio S.p.A., Italy

Abstract
This article presents the results of an experimental and numerical activity for
the characterisation of the containment capability of Waspaloy
alloy, used for
the case in aircraft engines. The ballistic facility is described and the impact
experiments outcomes are discussed; some preliminary numerical simulations are
also briefly presented.
Keywords: blade containment, ballistic tests, airgun, Waspaloy, high strain rate,
impact response.

1 Introduction
Turbine casings are the heaviest elements of an aircraft engine, thus the lightening
of the containment case, without jeopardizing the structural integrity level, would
lead to great advantages such as the reduction of the absolute weight of the whole
engine and to an increase of the engine performance (specific thrust per unit
weight), Kelvin [1].
In casing design, the minimum thickness allowed for the released blade
containment requirements [2, 3] is usually determined via the use of experimental
ballistic curves and, eventually, endorsed by blade out event FEM simulations
[1, 4–7]. A possible work plan for casing containment design could be: (1) to
obtain, through experiments, both the static and the dynamic characterisation of
the material (in the strain rate range of interest), (2) to perform the ballistic tests

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070121
132 Computational Ballistics III

on the specimens with which looking for the minimum containment thickness, (3)
to validate the FEM model of the ballistic tests (using the material characterisation
coming from step (1)) in order to simulate with an appropriate accuracy the
experimental results, (4) to obtain, via a set of reliable FEM simulations, a
containment curve to be used during the design process. While step (1) and (2)
are mainly experimental in nature, steps (3) and (4) can be performed using a
design-by-analysis approach based on extensive use of numerical simulations.
Experimental testing is mandatory because of two main reasons, at least: (a) to
characterize material behaviour at high strain rates and (b) to validate the
numerical simulation framework. Also FEM simulations are required in order to
describe a containment curve without recurring to experimental tests. Anyway,
both the ballistic tests and the FEM simulations should be limited to a reasonable
number: tests are very expensive both for the set up of the facility and for the
cost of specimens, numerical calculations need much time for modelling and for
the analysis. Thus, by means of response surface techniques based on a reduced
number of tests/simulations, it is possible to describe the dependence of the
containment thickness from the key parameters of the problem (kinetic energy
of the projectile, material, etc.) [4, 8, 9].
In this paper we present the ballistic tests conducted on some Waspaloy
alloy
plates. The experimental set-up and results are presented. Static and dynamic
material characterisation is briefly presented; a detailed discussion on these results
may be found in Bonora et al [10]. A preliminary FEM simulation including a
simplified model of the material static and dynamic behaviour is performed and
compared with the experimental results.

2 Ballistic tests
2.1 Airgun facility

The ballistic tests have been conducted by means of an airgun facility at the
AERMEC - Aeromechanical Structural Systems laboratory of the Aerospace
Engineering Department of Politecnico di Torino. The experimental set-up consists
of a pressure tank (air filled, maximum operating pressure: 14 bar), a shooting
control system, a support structure for targets and a velocity measuring system that
allows also to control the trajectory and attitude of flight of the projectile. Some
protection shields are located around the impact area as well to stop the bullet after
it has hit the target. In fig. 1 an overall functional scheme of the airgun facility
is presented. The facility includes a set of barrels of various lengths allowing to
adequately cover a wide range of speed until 250 m/s; barrels’ inside diameter is
38, 1 mm (1, 5 inch).
The projectile is made up of a steel cylinder attached on the top of a polyurethane
foam cylinder (later on called sabot), the all wrapped with thermo-shrinking
packing like shown in fig. 2. The dimensions of the projectile are reported in
table 1; the projectile’s mass is that of a small turbine blade (engines for executive
airplanes). The sabot allows to guide the projectile inside the barrel avoiding

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 133

Figure 1: Overall functional scheme of the airgun facility.

rotations and scratches, while the thermo-shrinking packing warrants the projectile
integrity and its constant length as well; the last feature is particularly important
for a correct measurement of the projectile’s velocity before the impact.
The shooting control system is based on a Mylar
membrane that separates the
firing section from the pressure tank and barrel; Mylar
membrane is designed
to hold out the tank’s maximum operating pressure and its burst is induced from
an electrical resistance by Joule effect. Two EPDM rubber packings act as electric
insulator and as air seal. Both EPDM rubber packings and Mylar
membrane are
punching so that they can be inserted in the tightening system as shown in fig. 3.

Table 1: Projectile’s dimensions and masses.

diameter (mm) length (mm) mass (g)

steel cylinder 33 10 67
sabot 37 50 5
thermo–shrinking packing - - 4, 5

The projectile is inserted into the rear end of the barrel then the diaphragm is
inserted in the tightening system which is closed; compressed air is pumped into

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
134 Computational Ballistics III

Figure 2: Pictures of projectile made up of a steel cylinder, a polyurethane foam

cylinder and a thermo-shrinking packing.

the pressure tank. At firing, the electrical resistance bursts the Mylar
membrane
and the projectile is accelerated along the barrel; then it leaves the muzzle and it
passes the velocity measurement station before that it impacts the target after about
0, 5 m of free flight.
The target’s support structure allows to set the plate with an angle of 60◦
between the normal to its surface and the projectile’s trajectory, Arace [8]. The
plate is simply supported by a rectangular steel frame. Four stoppers are set in
the frame’s corners to prevent plate’s sliding during the impact. Also, the target’s
support structure is designed to realize an adequate connection with the airgun
box so as to transmit the impact forces on the box itself; the high support structure
stiffness assures that great part of the strain energy is charged to the panel as well.
A picture of target’s support structure is shown in fig. 3.
Three laser-light sources on one side of the projectile path and three diode
detectors on the other side are used and mounted in lines to match themselves
and to measure the projectile’s velocity. When the projectile passes between the
sources and detectors, laser-light is interrupted and signals are given and stored to
a digital oscilloscope: as projectile’s length and distances between laser-beams are
known, it is possible to estimate both “instantaneous” velocity and average velocity
of the projectile just before the impact. The first laser-beam works also as a trigger
for the digital high-speed camera system used in the tests. PCO SENSICAM QE
digital camera system is used together with three flash-lamps system to control the
trajectory and attitude of flight of the projectile both before and while it impacts
the target; the airgun box is equipped with two framing windows to allow the high-
speed camera to photograph and the flash-lamps to light properly both projectile
and target. The system is able to take some snapshots after the projectile leaves the
muzzle and while it impacts the target, as shown in fig. 4.
Measurement of velocity is the main feature in experimental ballistic tests. The
equipment described above has been purpose-built for this particular application
and it was therefore necessary to check it; tests were carried out where projectiles
were fired through the impact chamber at different velocities without a target

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 135

(a) (b)

Figure 3: (a) Picture of target’s support structure; (b) Sketch of the shooting control
system.

Figure 4: PCO SENSICAM QE digital high-speed camera system. On the left, first
frame with four snapshot of the projectile travelling versus the target; on
the right, second frame with single snapshot of the impact.

plate. Good agreement was found between the velocity estimated by the laser-
beams system and the images obtained from high-speed camera.
2.2 Experimental results

At first, a firing tuning process has been done through many preliminary shoot tests
that have resulted in a pressure–velocity calibration curve of the system, constantly
updated by new tests.
A ballistic campaign has been carried out on four plates of Waspaloy
alloy
whose geometric characteristics are indicated in table 2. All plates are of the same
thickness so the experimental goal is to determine a kinetic energy range inside
which the limit value resides for the thickness in examination.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
136 Computational Ballistics III

The speed with which to start the ballistic campaign has been evaluated through
an energetic method based on an empirical equation obtained for correlation with
experimental data, Di Sciuva et al [4]

t = a · (Ekin )b (1)

where t is the plate’s thickness, a and b are obtained by experimental data fitting
and Ekin is the kinetic energy of the projectile.
In table 2 are reported all information about several ballistic tests, particularly
the projectile’s characteristics, the impacts’ average speeds, therefore the kinetic
energy and the experimental result (C = containment, NC = not containment) are
pointed out; all values of velocity and kinetic energy are dimensionless because
they are confidential company data. In particular, velocities and kinetic energies
are divided by the corresponding (containment) limit values.
In figs. 5–6 are shown some photos of the plates #2 and #4.

Table 2: Plates’ geometric characteristics and ballistic experimental results.

Plate Projectile
# a (mm) b (mm) h (mm) m (g) Ṽ Ẽkin C/NC
1 300 300 1, 00 76, 5 1, 398 1, 955 NC
2 300 300 1, 00 76, 5 0, 917 0, 841 C
3 300 300 1, 00 76, 5 1, 203 1, 448 NC
4 299 299 1, 00 76, 5 1, 083 1, 173 NC

3 Static characterisation
Quasi-static tensile tests were carried out according to ASTM standards [11,
12, 13] to identify the material mechanical properties of the Waspaloy
. Seven
specimens were used in order to make a proper statistical survey. Both target’s
plates and specimens were obtained from the same sheet.
The engineering and true mechanical properties for the target are given in
table 3; fig. 7 shows the quasi-static tensile engineering stress-strain curve for the
target; such information are required in order to calibrate both the analytical model
(ref. eqn. 1) and the numerical simulation.
Furthermore, the dynamic response of Waspaloy
at high strain rate has been
investigated by means of direct tension Hopkinson pressure bar (DTHPB). Using
round bar samples with different reference lengths, the material response in the
strain rate range (from 700 s−1 up to 3200 s−1 ) expected to occur under dynamic
impact conditions, has been experimentally determined and used to derive the
constitutive model to be used in the FEM simulations.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 137

Figure 5: Experimental results: Waspaloy
alloy plate #2, 300 × 300 × 1, 00 mm,
Ẽkin = 0, 841.

Figure 6: Experimental results: Waspaloy
alloy plate #4, 299 × 299 × 1, 00 mm,
Ẽkin = 1, 173.

4 Preliminary FEM simulation

A preliminary estimation of numerical capability to reproduce the experimental
ballistic results, using MSC.Dytran
, has been done (Degiovanni [14]). A finite
element model representative of the true experimental tests has been realized;
it is made up of about 270.000 HEXA8 solid elements (with 3 elements in the
thickness) and a failure model based on a maximum equivalent plastic strain is
used. For these first analyses the material model employed considers in simplified
way the mechanical characteristics, both static and dynamic, obtained from the
respective tests.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
138 Computational Ballistics III

Table 3: Quasi-static mechanical properties for Waspaloy
alloy.

Engineering Data True Data

Property Unit mean value standard mean value
deviation

Density (kg/m3 ) 8090 - -

Tensile ultimate strength (MPa) 1264 8, 8733 1590
Tensile yield strength (MPa) 692 12, 5884 695
Elongation at break (mm/m) 288, 85 5, 239 25
Poisson coefficient 0, 2871 0, 0119 -
Young Modulus (MPa) 226930 2423, 4 -

Figure 7: Quasi-static tensile engineering stress-strain curves of Waspaloy
alloy.

Some results are shown in fig. 8 for the velocity of two different projectiles,
respectively about experimental limit events of containment and not containment.
As preliminary outcomes they show a good agreement with the ballistic tests.

5 Conclusions

The design of aircraft engines has also to face containment problems; any object
(released blade, disk fragment) impacting the protection case must be contained
according to regulations. The evaluation of minimum containment thickness is
normally based on simple thickness vs kinetic energy curves; anyway, these curves
are obtained through extensive campaigns of experimental tests and numerical
simulations. The present article describes some results of a ballistic tests campaign
on Waspaloy
alloy plates and some preliminary numerical simulations of the
impact experiments; these show that good correlation with tests is obtained (due

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 139

(a) Ẽkin = 0, 841 (b) Ẽkin = 1, 173

Figure 8: Numerical results of preliminary FEM simulation of Waspaloy

containment capacity (MSC.Dytran
).

to a reliable experimental evaluation of the material mechanical properties, both

quasi-static and at high strain rates).

Acknowledgements
The authors from Politecnico di Torino gratefully acknowledge Messrs Giuseppe
Ruvinetti and Lorenzo Mascarello for the helpful and constant technical support
during ballistic tests. Also, they acknowledge Regione Piemonte for funding their
activity (“Multidisciplinary optimisation of aerospace structural systems”, Project
E57) in the framework of “Bando Regionale per la Ricerca Scientifica 2004”.
The author from Avio S.p.A. acknowledges Dr Antonino Vassallo for his
contribution to the activity.

References
[1] Kelvin Y. N., Turbine Rotor Burst Containment Analysing using Msc/Dytran,
Hamilton Standard (United Technologies), Internal Report, 1996.
[2] FAR 33.94 – Blade containment and rotor unbalance tests, Federal Aviation
Administration, 1984.
[3] AC 20–128A – Design Considerations for Minimizing Hazards caused
by Uncontained Turbine Engine and Auxiliary Power Unit Rotor Failure,
Federal Aviation Administration, 1997.
[4] Di Sciuva M., Frola C. & Salvano S., Low and high velocity impact
on Inconel 718 casting plates: ballistic limit and numerical correlation,
International Journal of Impact Engineering, 28, pp. 849–876, 2003.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
140 Computational Ballistics III

[5] Carney K., Pereira M., Revilock M. & Matheny P., Jet engine Fan Blade
Containment using Two Alternate Geometries, Proc. of the 4th European LS
Dyna Users Conference, Ulm (Germany), 2003.
[6] Sarkar S. & Atluri S. N., Effects of multiple blade interaction on the
containment of blade fragments during a rotor failure, Finite elements in
Analysis and Design, 23(2), pp. 211–223, 1996.
[7] Krauss A. & Frischbier J., Containment and penetration simulation in case
of blade loss in a low pressure turbine, MTU Aero Engines GmbH, Internal
Report, Monaco, Germany, 2002.
[8] Arace F., Gherlone M. & Frola C., A simplified methodology for containment
case design, Proc. of the XVIII AIDAA National Congress Volterra (Italy),
2005.
[9] Liu D. & Stronge W.J., Ballistic limit of metal plates struck by blunt
deformable missiles: experiments, International Journal of Solids and
Structures, 37(10), pp. 1403-1423, 2000.
[10] Bonora N., Ruggiero A., Di Sciuva M., Degiovanni M., Mattone M.,
Gherlone M. & Frola C., Nickel base superalloy containment case design:
constitutive modeling and computational analysis, to be submitted to Int. J.
of Solids and Structures.
[11] ASTM E 8 – Standard Test Methods for Tension Testing of Metallic
Materials. ASTM International.
[12] ASTM E 1012 – Standard Practice for Verification of Test Frame
and Specimen Alignment Under Tensile and Compressive Axial Force
Application. ASTM International.
[13] ASTM E 111 – Standard Test Method for Young’s Modulus, Tangent
Modulus, and Chord Modulus. ASTM International.
[14] Degiovanni M., Indagine numerico sperimentale di soluzioni strutturali
ottimizzate per casing di motori aeronautici, Report DIASP-259, Politecnico
di Torino, 2006

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 141

Experimental survey of the pressure cone and

temperature field of an A/T missile system
K. Krallis1, Ν. G. Orfanoudakis2, M. Founti3, D. T. Hountalas3,
A. Fatsis2 & N. Vlachakis4
1
Heron Consultant Engineers, Har. Trikoupi 107, 11473 Athens, Greece
2
Technical University of Chalkis,
Department of Mechanical Engineering, 34400 Chalkis, Greece
3
National Technical University of Athens,
Department of Mechanical Engineering, Zografou Campus,
Athens 15780, Greece
4
Engineering Science and Mechanics Department,
Virginia State University, USA

Abstract
This paper describes the characteristics of the temperature and pressure fields
measured downstream of the launcher of the Russian “FAGOT” anti-tank missile
system. The experimental data were used to establish the proper location of the
launcher on a vehicle used by the Greek Army.
The pressure field developing at the rear of the missile launcher has the
characteristics of a conical shock wave. The field was measured with an array of
piezoelectric pressure transducers and a fast computer-based data acquisition
system. The extent of the temperature field was established with high speed
video and infra-red still photography and was confirmed with miniature
thermocouple sensors.
The results showed that the pressure cone intersects the ground level at a
distance of 1.5 to 2.5 m behind the launcher. Typical overpressure of the shock
wave was of the order of 1 to 1.8 bar. The conical shock wave breaks down into
several smaller shock waves at a distance of 3 m and extends up to 15 m behind
the launcher. The diameter of the cone at this location is 2.5–3.0 m.
Keywords: missile launcher, measurements, pressure field, temperature field.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070131
142 Computational Ballistics III

1 Introduction
Most of existing information on the characteristics of the pressure cone and
temperature field from missile systems is restricted and partially presented in
International Symposia on Ballistics [1–5].
Temperature measurements at a 7.62 mm gun exit (muzzle flash) were
performed under laboratory conditions by Klingenberg and Mach [1] using
spectroscopy, absorption spectroscopy or high-speed photography with pre-
calibrated film. Film techniques can be usually applied for field measurements,
but should be used under low light conditions and with use of high-speed
cameras.
Measurements of density and exit gas velocity in a 20 mm gun as presented in
[3, 4], have been performed with the use of Schlieren photography and laser
Doppler anemometry methods. Both the above methods have serious
shortcomings in case that they are applied in larger than laboratory scale. The
application of Schlieren and laser Doppler anemometry in field measurements is
prohibitive due to high costs and inaccuracies induced by limitations of the
methods themselves.
However, pressure measurements using piezoelectric transducers have been
performed by Klingenberg and Schroder [2], Klingenberg [3] and Klingenberg et
al. [4] with very good response, especially at the frequency of 20Hz. Pressure
transducers are widely used for in cylinder measurements of internal combustion
engines commercially available.

2 Instrumentation and experimental method

The work reported here has two major parts: the first refers to flow visualisation
with the use of photography and the latter to pressure and temperature
measurements.
Various types of equipment have been used for the visualisation of the flow
field developing at the rear of the launcher during launching conditions. High
speed (24 frames/sec), high-resolution S-VHS video cameras allowed fast, real
time recording of the phenomena. In addition, three professional SLR-cameras
have been used: one panoramic Linhof 612, one Technorama, one EOS 1N and
one high resolution ΕΟS 1NRS. The cameras were equipped with motor drives
and zoom lenses of 28 -70 mm, F2.8L and 35 - 70 mm, F3.5-F5.6. During
preliminary launches, amateur SLR-cameras with 100 ASA and infra-red films
have been used to allow for representation of the different temperature zones.
The flow field was seeded with 0-0.3 mm diameter, lightweight (density equal to
60 kg/m3), expanded perlite particles that could travel together with the exhaust
gases at the tail end of the missile launcher. Perlite had been chosen to support
the flow visualisation because is light and inert, as well as environmental-
friendly.
The pressure and temperature measurements have been performed with
piezoelectric transducers and thermocouples, respectively. They have been
mounted in selected positions on a specially designed and constructed metallic

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 143

frame, placed downstream of the gas exit of the launcher. The research team has
designed the measuring system and has calibrated the measuring equipment and
the necessary data acquisition system. The metallic frame has been constructed
by 301 Technical Corps Unit of the Greek Army.
In order to allow adequate representation of the pressure cone characteristics
and at the same time to reduce the initial cost of the measurements, it was
decided to use 5 transducers located axially and radially, at critical points, with
respect to the launcher. It has been assumed that the pressure and temperature
cones develop axi-symmetrically downstream the tail edge of the missile
launcher.
The signals from the piezoelectric transducers were amplified and together
with the signals from the thermocouples were recorded using two independent
data acquisition systems in order to avoid repetition of the costly measurements
(i.e. launching of a FAGOT missile) associated with failure of the data
acquisition systems.

Table 1: Characteristics of the measuring system.

Type of Parameter Typical Characteristics

Equipment
Pressure Operation limit 0-20 bar or 0-200 bar
transducers: Sensitivity 25mV/bar
Kistler Eigen frequency > 50 KHz
Piezotron linearity <1%FSO
Tranducer
Nominal current 2...18 mA
Max. signal voltage 20 Vpp
Pressure Amplification 0.5...5
transducers' Frequency response +5% from DC to 220 KHz
amplifier: (at 1 meg load & 10 Vpp)
Kistler Max. Signal Voltage 20 Vpp for unit amplification
Curent + 10mA
Impedance 20Ω
Noise 250 µ for unit amplification

Sampling frequency 100 KHz

Number of analogue 16 single inputs, 8 differential
inputs + 10 Volt
Input Voltage 1,2,4
Amplification 100 Megohms, 10pF (channel
Input Impedance OFF)
Analog-digital 100 Megohms, 10pF (channel
converter Resolution ON)
Differential non linearity 12 Bit
System accuracy < 1LSB
+0.05FSR

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
144 Computational Ballistics III

The data acquisition frequency was 1 and 2 kHz, established by literature

review [1–4] and the signal cables were shielded to avoid noise and interference.
For each launch 16000 data points have been recorded and stored. The
electricity supply was provided by a military diesel generator, which in turn was
connected to an UPS (uninterrupted power system), to avoid power surges. The
characteristics of the pressure transducers, of the amplifiers and the data
acquisition system used for the measurements, are shown in Table 1.
Two very sensitive thermocouples (Φ5µm), with high response to temperature
variations, were placed close to the gas exit and on the axis of symmetry. The
thermocouples were installed on an old-used carrier. The temperature measured
was actually the temperature of the carrier and not of the gases. Nevertheless, it
is strongly connected to that of the gas, if the gas density and thermal
conductivity, the heat transfer coefficient (from gas to thermocouple) and the
thermal conductivity of the thermocouple are considered. The temperature
measurements were recorded simultaneously with the pressure data using the
same data acquisition system, as described above.
There was generally a problem to synchronise the cameras and the
instruments with the launcher itself because there was always a 1–1.5 s lag
between triggering and launching. Installation of an external triggering adapter
was not possible due to lack of technical information in relation to the launcher
and the missile. Triggering of the cameras and of the data acquisition system was
done manually.
During the last group of measurements, four launches have been conducted
and data recording had started well before and stopped well after the completion
of the phenomenon, to allow optimum data analysis.

3 Results

3.1 Flow visualisation and photographic work

Preliminary determination of the pressure cone had been conducted by use of the
photographic material collected during preliminary and main launches. The
pressure, temperature measurements and the flow visualisation were, at most
times, conducted simultaneously. This assisted the extraction of conclusions and
comparison of the photographic and measurement data
Flow visualisation allowed initial determination of the angle of launching and
pressure cone. Expanded perlite of low density (60 kg/m3) has been used to fill
the half section of a Φ120 mm and 1.5m long pipe, which in turn had been
placed on the ground at the back of the launcher. Vertical staffs with red
markings have been positioned every meter and allowed quick estimation of the
cone height. The missile and the long pipe filled with expanded perlite particles
are shown in Figure 1. It can be clearly seen that milliseconds (in fact 5 ms) after
launching the pipe broke into pieces exactly at the point where the shock wave
intersected the ground level. At this location the perlite particles got elevated due
to the exerted pressure forces and thereafter followed the exhaust gas path-lines.
A series of such photographs (e.g. Figure 2) allowed visual determination of the

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 145

pressure cone length and height, when launching is from the ground. Figure 1,
also shows red-hot smithereens that ejected during and due to the launch flying
along the pressure cone.
The results from processing the photographic material can be summarized as
shown in Table 2.
Flow visualisation and photography were used to interpret the phenomena and
to verify the pressure transducer and thermocouple results. Photographic work
has served to establish the following:
The angle of the thermal and pressure cone are not strongly dependent on the
angle the launcher forms with the ground.
The ‘front flame’ (i.e. the flame formed at the direction of the fly of the A/T
missile) has a conical shape; it is ca. 0.4 – 0.9 m long, with diameter ca. 0.3 –
0.5 m.
The pressure cone attains characteristics of intense turbulence mixing
downstream the first 3 m from the gas exit. The cone has about 15m length and
2.5 to 3m height.

Figure 1: Visualisation of the pressure cone at the rear of the launcher.

Figure 2: Long distance view of the visualised pressure cone at the rear of the
launcher.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
146 Computational Ballistics III

Table 2.

Geometry of the gunshot

Height of the launcher from ground 40 cm
Angle of launching 50
Mean flight velocity 180 m/s
Measurements
Area of max. ground pressure 2.0 – 2.5 m downstream the gas exit
Area of max. temperatures 0.9 – 1.2 m downstream the gas exit
Pressure cone downstream of the gas length ca. 12 – 15 m, height ca. 2.5 –
exit 3 m.
Thermal cone downstream of the gas length 0.9 m, diameter 0.6 – 1.0 m
exit
High light intensity Cone upstream length 0.4 – 0.9 m, diameter 0.3 –
of the launcher 0.5 m

H i g h p re ssu re zo n e

D ire c tio n o f
flig h t

F la m e c o n e -
F la m e c o n e o n th e
h ig h
g a s sid e - H ig h
te m p e ra tu re
z on e tem p e ra tu re zo n e
U n its in (m )
B o u n d a rie s o f sm ith e ree n s e je c tio n
zo n e . A re a o f lo w p re ssu re

Figure 3: Determination of danger zones during the launch of the FAGOT

A/T missile system.
The process of the flow visualisation showed that the angle, of the pressure
cone with the ground is about 9o.
Based on the photographic and experimental data Figure 3 was drawn in order
to designate safe angles of launching (for personnel and vehicles). In this figure,
the most important characteristics of the pressure and thermal cone are defined.
In addition, areas where smithereens were ejected during and due to the launch
are also designated.
3.2 Pressure and temperature measurements
The measurements of temperature and pressure are shown in Figures 4–6. For
each measurement campaign, five different pressure transducers were used.
Figure 4 represents the evolution of pressure and temperature with time for
various positions of the transducers.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 147

position curve Max.

(x,y) pressure
(bar)

(0.5, 0) 0.0206
(1.0, 0) 0.28466
(1.5, 0) 0.3446
(2.5, 0) 1.075
(0.8, 0.3) 0.020
Sampling frequency: 2 kHz

Figure 4: Variation of pressure (above) and temperature (below) at various

positions downstream the launcher.

Based on the maximum-recorded pressure at each known position of the

pressure transducers the variation of maximum pressure with spatial location is
plotted in Figures 5 and 6.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
148 Computational Ballistics III

Shot

Shot
Shot
Shot

Estimate
Pressure (bar)

Distance from launcher on the ground (m)

Figure 5: Variation of maximum pressure along the axis of the launcher,

which was positioned at 50 cm from the ground.

0.5-1
0-0.5

1
0.5
0.5

Pressure
1
1.5

(Bar)
2

0 0
2.5
3
3.5
4

Axial distance from

launcher (m)
Transverse distance
from launcher
(m )

Figure 6: Constant pressure lines for launcher position at 50 cm from ground.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 149

Figure 5 demonstrates that the maximum value of the pressure occurs in a

range between 1.5 and 2 m downstream of the gas exit. This statement is in
agreement with the photographic results. The maximum measured pressure value
was 1 ± 0.8 bar. This value is expected to vary with the change of the height of
launcher from the ground (as for example, when the launcher would be installed
on the military vehicle). Figure 6 shows the lines of constant pressure
downstream of the gas exit, for launcher position at height 50 cm from the
ground. This figure is representative for the installed height of the launcher.

4 Conclusions
This paper described the tasks undertaken to investigate the temperature and
pressure fields behind the launcher of the Russian “Fagot” anti-tank missile
system. The measurements were conducted in order to allow decision about the
proper location of the launcher on a vehicle used by the Greek Army. The main
conclusions can be summarised, for launch position at 50cm height from the
ground, as follows:
The pressure and thermal cone have characteristics independent from the
angle of launching.
Flame is generated at both sides of launcher; the characteristics of these
flames are strongly dependent on the kind of the launch (successful and not).
The maximum measured pressure on the ground was found to be at distance
between 1.5 and 2.5 m downstream of the gas exit and it found to be 1 ± 0.8 bar.
The pressure cone is extended 15 m downstream the gas exit, at a height of
2.5–3m, has turbulent characteristics.
The angle between the cone and the ground is about 9o.
The installation of the A/T launcher on a vehicle is safe under certain
conditions.

References
[1] Klingenberg, G. and Mach, H.: “Investigation of Combustion Phenomena
associated with the flow of Hot Propellant Gases - I: Spectroscopic
Temperature Measurements Inside the Muzzle Flash of a Rifle”,
Combustion and Flame, 27, pp. 163-176, 1976
[2] Klingenberg, G. and Schroder, G.A.: “Investigation of Combustion
Phenomena associated with the flow of Hot Propellant Gases - II: Gas
Velocity Measurements by Laser Induced Gas Breakdown”, Combustion
and Flame, 27, pp. 177-187, 1976.
[3] Klingenberg, G.: “Investigation of Combustion Phenomena associated
with the flow of Hot Propellant Gases - III: Experimental Survey of the
Formation and Decay of Muzzle Flow Fields and of Pressure
Measurements”, Combustion and Flame, 29, pp. 289-309, 1977.
[4] Klingenberg, G., Mach, H. and Smeets, G.: “Flow Field Measurements of
an Unsteady Reacting Muzzle Exhaust Flow”, Transactions of ASME,
105, pp 884-888, 1983.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
150 Computational Ballistics III

[5] Stiefel, L., Editor: “Gun Propulsion Technology”, American Institute of

Astronautics and Aeronautics, Progress in Astronautics and Aeronautics,
109, AIAA, Washington 1988.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Section 5
New developments in
computational techniques
This page intentionally left blank
 Computational Ballistics III 153

Safety checks and hit probability computation

in numerical trajectory calculation using error
propagation technology
A. Kuhrt & H. Rothe
University of the Federal Armed Forces, Hamburg, Germany

Abstract

In fire control systems using numerical trajectory integration, safety checks and
hit probability information still is based on empirical parameters. The need for a
modern and precise security testing on a defined level of confidence and tailored
hit probability computation were the motivation to develop the technology outlined
in this paper. The methods are based on trajectory calculation using a modified
point mass model that considers drag, lift, Magnus and Coriolis forces, gravity and
all atmospheric influences. The Gaussian law of the propagation of the stochastic
error within a system of dependent and independent variables was used to generate
a vector of standard deviations for all error afflicted parameters every integration
time step. This approach leads to a flightpath dataset of the projectile, not only
giving information about position and velocity, but also about their standard
deviations. The check for a minimum height over ground or distance to no-fly
zones is very easy and can be performed with a changing level of confidence
throughout the trajectory. Hit probability calculation is done automatically and
considers all current data that has influence on the projectile’s trajectory.
Keywords: error propagation, safety checks, numerical integration, crest
clearance, hit probability, multivariate statistics.

1 Introduction
In fire control systems using numerical trajectory integration, safety checks and
hit probability information still is based on empirical parameters. Even modern
systems using time consuming routines on calculating numerical derivatives for a
trajectory perturbation in order to gain hit probability information. The need for a

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070141
154 Computational Ballistics III

precise security testing on a defined level of confidence and tailored hit probability
computation were the motivation to develop the technology outlined in this paper.
The methods are based on trajectory calculation using a modified point mass model
that considers drag, lift, Magnus and Coriolis forces, gravity and all atmospheric
influences.

2 Fire control software

Modern fire control software mostly is strongly unitized to cover a wider range of
different weapon systems or ammunition types. This leads to a data-handling that
has to be taken into consideration. Trajectory calculation has to meet the demands
on both, computation time and accuracy. That influences the numerical method
and the physical model used.

2.1 Trajectory calculation

The trajectory calculation used in modern Fire Control Software is mostly based
on a set of differential equations (DEQ) incorporating a modified point mass model
[1, 2]. These differential equations are solved by a numerical integration scheme
to generate a flightpath containing the important information such as position,
velocity and angle for each time step from leaving the muzzle to the impact.

Drag V V Height Air Density

CD Mach Number Vel of Sound

Figure 1: Example for the Dependencies of integration parameter Drag.

Common integration schemes used are explicit or implicit single-step methods,

such as Heun or Runge-Kutta. Due to the good conditioned DEQs faster schemes
without error-estimation were preferred to keep computation time short. The
algorithms described below are developed for an Euler scheme, but were also
successfully tested on a Heun scheme.
During numerical integration of the trajectory parameters change, depending
on current projectile position and velocity. Atmospheric data gained from
measurements or meteorological prognosis depends strongly on the height over
ground. Ballistic coefficients may depend on the Mach number, which on its part
depends on the actual speed of sound and the projectile velocity. There are many
parameters in the DEQs of motion that depend on the results of the last performed
integration step or that are linked to external data sets (e.g. fig. 1).

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 155

2.2 Requirements for safety checks

In modern fire control software an important part is security. Therefore, a

wide range of tests were provided to check the interaction of a trajectory with
other objects or zones. These zones can vary between traditional checks against
trajectory-ground interaction and checks against violation of temporary no-fly
zones. Today’s checks are mostly performed by perturbation mathematics and their
application on the trajectory [3]:
Adding a perturbation on one initial condition and calculating the difference of
the hit point allows to compute the numerical partial derivative with respect to
the initial condition perturbed. The linear combination of these partial derivatives
multiplied with their variances leads to the absolute error probability usually given
as a CEP. The mathematical background is the Gaussian law of the propagation of
the stochastic error [4]:
N 

2  2  2  2
∂f ∂f ∂f ∂f
σy2 = σi2 = σ12 + σ22 + σ32 + · · · (1)
i=1
∂xi ∂x1 ∂x2 ∂x3

This approach is also suitable for crest clearance checks:

Two perturbed trajectories were calculated which enclosure the original trajectory.
One of this new trajectories - the lower one - is used for checking against projectile-
ground-interaction. The higher one is obsolete.
Another approach is to rise the ground to a safety level needed and then check
the original trajectory against projectile-ground-interaction using the raised ground
level.
Both algorithms can only handle defined and fixed error budgets for a limited
number of parameters because every parameter to be perturbed causes an
additional complete trajectory calculation. These parameters are limited to initial
conditions of the numerical integration and no dependencies were taken into
consideration.

3 A measurement approach
In measurement theory basically all measurands are error afflicted. These errors are
quantified by two characteristic values: mean value of the sample (x̄) and standard
deviation of the sample (sx ). All error afflicted measurands are assumed Gaussian
distributed.
Looking at the parameters affecting a trajectory integration, three main
categories can be differentiated:
1. Parameters directly affecting the initial conditions, such as gun position, gun
height, muzzle velocity, azimuth and elevation.
2. Parameters indirectly affecting the initial conditions, like propellant
temperature and barrel abrasion.
3. Parameters directly affecting the projectile in flight containing all
atmospheric properties depending on height (temperature, air density, air

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
156 Computational Ballistics III

pressure, wind, speed of sound) and all projectile depending coefficients

(drag, lift, Magnus and Coriolis forces, spin, yaw).
From a measurement point of view, all these parameters are error afflicted and
therefore, every parameter can be quantified by its mean and standard deviation.
The aim is to get a statement for the certainty - or better uncertainty - of the
projectile’s position and velocity. The approach is to propagate the error of the
position and the velocity vector not by perturbing the initial conditions, but
by recalculating the standard deviation of position and velocity vector every
integration time step.

Figure 2: Resulting fuzzy position of a projectile.

The main advance is that a data set of projectile’s uncertainty for each point of
the trajectory calculated is produced. Changing standard deviation of error afflicted
parameters during the trajectory is possible and therefore assures a good flexibility.
This computation is extraordinary fast compared to an equivalent perturbation
approach.

4 Error propagation
In general, the Gaussian Law (eqn. (1)) covers most of normal measurement cases.
But because modern fire control softwares integration methods and object handling
splits up the computation of position and velocity vector into subroutines, a more
general definition of the error propagation has to be used. The variances and their
depending variables to be added at the end of the calculation cannot be assumed to
be independent anymore. To consider these dependencies the general law for error
propagation has to be used [5]: Let f (x1 , x2 , . . . , xn ) be a function which depends
on n variables x1 , x2 , . . ., xn . The uncertainty of each variable is given by σxj
multiplied with the quantile for a level of confidence of α (remark, that in military
applications times of PE is the common quantile): If the variables are correlated,

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 157

the covariance between variable pairs, Ci,k := cov(xi , xk ), enters eqn. (1) with a
double sum over all pairs (i, k):


 n n  


∂f ∂f
σf =  Ci,k , (2)
i=1
∂xi ∂xk
k=1

where ∂x∂f
j
designates the partial derivative of f for the j-th variable and Ci,i =
var(xi ) = σx2i . After calculating σf , the value of the function with its uncertainty
is:

f ± uα/2 σf . (3)

The estimation of the correlation matrix is discussed in section 5. Within

numerical integration schemes error propagation does not follow this law.

4.1 Error propagation in Euler schemes

In Euler schemes every discretized equation can be written as:

xi = xi−1 + gi ∆ti , (4)

with ∆ti as the not error afflicted, but varying integration increment, gi as a
function, varying and error afflicted, that incorporates the discretized DEQ within
increment i.
Analysis following the method described in section 5 shows that the correlation
between xi−1 and gi is negligible. This would enforce a Gaussian error
propagation approach. Trying this one will end up with wrong results, because
this approach considers only the increment ∆ti ∈ [0, ∆ti ], but neglects the actual
position within the integration process, which would be t ∈ [0, t]. Instead an error
propagation similar to the integration scheme used is postulated:

 
2
σxi = σxi−1 + σx20 + t2 σg2i − σx20 + (t − ∆ti ) σg2i (5)

In fact, the equation shown is a steady continuation of the existing error with a
gradient with respect to the actual integration increment.
The main advantage is to consider the whole integration time t and the initial
error. The following figures show the differences of a Gaussian and this approach
for the error in velocity and range of a constantly slowed down vehicle from
30 m/s with an error in v0 and in the friction co-efficient:

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
158 Computational Ballistics III

1.04 1.04

1.03 1.03

sv in m
sv in m

1.02 1.02

1.01 1.01

1.0 1.0
0 10 20 30 0 10 20 30
t in s t in s

Figure 3: Resulting standard deviation of v of analytical solution (grey) and of

Euler scheme (black); Left: Gaussian law Right: New approach.

30 30

20 20
sx in m
sx in m

10 10

0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
t in s t in s

Figure 4: Resulting standard deviation of x of analytical solution (grey) and of

Euler scheme (black); Left: Gaussian law Right: New approach.

5 Estimation of correlation coefficients

For the calculation of the error propagation, the covariances between pairs of
parameters must be known. The general definition of the covariance is:

cov(X1 , X2 ) = E(X1 · X2 ) − E(X1 ) · E(X2 ) (6)

= ρ · σ12 · σ22 (7)
 −∞  −∞
= x1 · x2 f (x1 , x2 )dx1 dx2 − E(X1 ) · E(X2 ) (8)
∞ ∞

with X1 , X2 as variates, E(. . .) as the expectant and ρ as the the Pearson

product-moment correlation coefficient (PMCC). Mind that the probability density
function (PDF) f (x1 , x2 ) is a bivariate Gaussian distribution with the Gaussian
distributions of each error afflicted parameter as marginal distributions. The
difficulty to estimate the covariance of two parameters is, that the transformation
of bi-variate distribution to the marginal distributions is irreversible. The bi-variate

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 159

Gaussian PDF is defined by 5 parameters: µx1 , µx2 , σx1 , σx2 , ρ.


1 (x1 −µx1 )2 (x1 −µx1 )(x2 −µx2 ) (x2 −µx2 )2
− −ρ +
1−ρ2 2
2σx σ x1 σ x2 2
2σx
e 1 2
f (x1 , x2 ) = (9)
2πσx1 σx1 1− ρ2
The extraction of the marginal distributions is easy and like:
(x1 −µx1 )2 (x2 −µx2 )2
− 2 − 2
fx1 (x1 ) = √ 1 e 2σx
1 , fx2 (x2 ) = √ 1 e 2σx
2 , (10)
2πσx1 2πσx2

but the reconstruction of the bivariate PDF is impossible without the defining
parameter ρ [6]. The approach is to determine this PMCCs using a pseudo empiric
approach.

5.1 Pseudo-empiric correlation matrix

Although the error afflicted parameters used in fire control software are
populations, the method to estimate the PMCC is derived from a method developed
for samples. The way to determine an empiric correlation matrix, as described in
[7], is:
j variates are determined in i samples. The values of the variates are
standardized by subtracting the arithmetic mean value of the variate and dividing
it by the standard deviation of the variate throughout the samples.
xij − x̄j
zij = (11)
sj
with
zij : standardized value of the variate j in sample i.
x̄j : arithmetic mean of the variate j: x̄j = n1 ni=1 xij .
sj : empiric standard deviation
 of the variate j:
1 2 n
sj = n−1 i=1 xij − x̄j .
This standardization results in a n · m matrix Z:
 
z11 z12 · · · z1m
 
 z21 z22 · · · z2m 

Z = .  (12)

 .. 
zn1 zn2 ··· znm

The empiric correlation matrix R is then:

1
R= ZT Z (13)
n−1
The expectation value of the empiric correlation matrix R is the correlation
matrix of the population P = E(R).

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
160 Computational Ballistics III

The approach is to generate samples out of the error afflicted equation using the
conditional equations of the error afflicted parameters.
1. Each error afflicted parameter is defined as a variate.
2. Each error afflicted parameter appearing in a conditional equation generates
a sample.
3. A sample is generated by perturbing the parameter defined in 2. within all
conditional equations by its standard deviation σ.
4. The results are standardized using the means and standard deviations of the
variate populations.
The resulting matrix consists of n · m samples z, n variates fxn , depending on
m parameters a:

 
z11 = fx1 (a1 + σa1 , · · · , am ) ... z1m = fx1 (a1 , · · · , am + σam )
 .. .. 

Z = 
. . 
zn1 = fxn (a1 + σa1 , · · · , am ) . . . znm = fxn (a1 , · · · , am + σam )

The sum of each column is 1. Each sample adds the part of the PMCC generated
by the perturbed parameter. Perturbed parameters adding parts to more than one
variate cause correlations. Calculating R using eqn. (13) provides an equation
for every PMCC. The division by n − 1 drops out because the use of standard
deviation generates theoretical infinite sized samples and the resulting matrix is
in fact the correlation matrix of the population P . This approach was verified and
successfully tested using Monte-Carlo simulations [8].

6 Example
As an example for safety testing of trajectories an artillery shell was chosen. It is
fired with a quadrant elevation (QE) of 800− . The trajectory was computed using a
predictor-corrector integration scheme for a modified point mass model (MPMM).
Only a few parameters were considered error afflicted, such as the drag coefficient,
projectile mass, QE and muzzle velocity. Using the error propagation technology,
outlined here, a plot can be generated that shows the position of the projectile at
any point of the trajectory on a defined level of confidence. Figure 5 shows areas
of positional probability on a level of confidence of 10 PE which is equivalent to
99.93 %. In the first plot elipsoids for 10 PE for range and height are given. The
second plot shows elipsoids for range and deflection.
For safety tests the most important information is the distance between projectile
and ground measured in times of PE. Figure 6 shows on the left the trajectory and
the corresponding ground profile. On the right the times of PE are plotted together
with the changing safety level regarding the projectile’s traveling over own (20 PE)
or foe (10 PE) ground or while approaching the target (0 PE).

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 161

1
0.9
0.8

normalized height over ground 0.7

0.6
0.5
0.4
0.3
0.2
0.1
0
-0.1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
normalized range

1
normalized
deflection

0.5

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

normalized range

Figure 5: Probable projectile position on a C.L. of 50 PE.

7000

6000

5000

4000
height in m

3000

2000

1000

-1000

-2000
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
normalized range

100

90

80

70
distance projectile-ground in PE

60

50

40

30

20

10

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
normalized range

Figure 6: Safety in Crest checking in times of PE.

7 Conclusions

The technology of error propagation outlined in this paper is a new approach

in calculating PEs. The current technology of overall numerical derivation with
respect to some parameters is replaced by an analytical derivation with respect to
all parameters and performed within each integration time step. This allows the
change of standard deviation throughout the flightpath of the projectile.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
162 Computational Ballistics III

A new method for the propagation of error afflicted parameters within first and
second order numerical integration schemes has been developed and successfully
tested. The correlation between parameters assumed error afflicted has been
determined using a new quasi empiric approach. The resulting PMCC matrices
vary throughout the flightpath and can be recalculated every integration time step.
In result, flightpath data with information about error probability for position
and velocity for each integration time step is given. Calculation of the probable
error at the target is done automatically and with good accuracy even for deflection,
that was determined empirical before. Safety testing against trajectory - ground
interaction can be performed on a high level of flexibility by changing safety
quantiles during the flight. Checks against no-fly zones now can be performed
by a fraction of the effort formerly necessary.
The usage of the technology outlined is not restricted to trajectory calculation,
but offers the opportunities for many other branches using time critical numerical
initial value solvers.

References
[1] NATO, STANAG 4355 - The Modified Point Mass and five Degrees of Freedom
Trajectory Models. 4th edition, 2006.
[2] McCoy, R.L., Modern Exterior Ballistics. Schiffer Military History: Atglen,
1999.
[3] Molitz, R., Hellmuth; Strobel, Äußere Ballistik. Springer: Berlin, 1963.
[4] Adunka, F., Meßunsicherheiten - Theorie und Praxis. Vulkan-Verlag: Essen,
2000.
[5] Irle, A., Wahrscheinlichkeitstheorie und Statistik. Teubner: Stuttgart, Leipzig,
Wiesbaden, 2001.
[6] Cook, S., R. Dennis; Weisberg, Applied Regression including Computing and
Graphics. Wiley: New York, 1999.
[7] Eckey, R.R.M., Hans-Friedrich; Kosfeld, Mulitvariate Statistik: Grundlagen -
Methoden - Beispiele. Gabler: Wiesbaden, 1st edition, 2002.
[8] Kuhrt, A., Crest Clearance und ACA Check Technologie für numerische
Flugbahnberechnungen - Theoretische Grundlagen. Research Report:
Hamburg, 2007.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 163

Scalable parallel computation of explosively

formed penetrators (EFPs)
E. L. Baker, C. Chin, B. E. Fuchs & J. Pham
Energetics and Warheads Division, U.S. Army, ARDEC, Picatinnyl,
NJ, USA

Abstract
Under the DoD High Performance Computer Modernization Program, the
ARDEC Energetics and Warheads Division has successfully developed and
demonstrated scalable parallel warhead mechanics modeling of an advanced
explosively formed penetrator (EFP) warhead device. Warhead development
commonly uses advanced computer modeling which can take extreme amounts
of computer time. The recent application of scalable parallel computers has
demonstrated the capability to achieve approximately a 120 times speed-up of
EFP warhead modeling. Speed-ups were demonstrated using modeling of a
copper lined EFP anti-armor warhead using the PCTH computer program on a
number of multiple processor computers including a 256 node IBM SP2, 128
node Silicon Graphics Origin 2000 and smaller multiple processor Silicon
Graphics workstations. This pioneering work clearly demonstrates the feasibility
of applying parallel computers for warheads development.
Keywords: explosively formed penetrator, scalable computing, parallel
computing.

1 Background
Continuum simulation of high rate explosive applications using finite element,
finite difference and discrete particle methodologies have received considerable
attention since the advent of digital computers (Von Neumann and Richtmyer
[1], Mader [2], Wilkins [3]). Explicit high rate continuum simulation is
extremely computationally intensive. As a result, less computationally
demanding analytic models of high explosive systems have been widely
developed and used (Baker [4], Yates et al [5]). However, analytic models rely

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070151
164 Computational Ballistics III

on many simplifying approximations and empirical relationships. Continuum

modeling techniques allow the inclusion of advanced material descriptions and
attempt to simulate detailed distributed material responses. The resolution of
detailed wave dynamics and highly path dependant nonlinear material response
are often very important aspects of high explosive application simulations that
simply cannot be reasonably approximated using simplified analytic modeling.
As computational capabilities have progressed, so has the application of
continuum simulations of high explosive applications including shaped charge
and explosively formed penetrator (EFP) warheads.

2 Warheads modeling
Previous efforts have demonstrated the successful application of explicit high
rate finite difference and finite element computer modeling to the simulation and
design of shaped charges (Baker and Campbell [6]) and EFPs (Baker and Liu
[7], Fong et al [8]). This modeling has been primarily axisymmetric, as the
initial geometries are often axisymmetric. Figure 1 presents an axisymmetric
shaped charge simulation calculated using CALE. Figure 2 presents an
axisymmetric EFP simulation calculated using DYNA2D. CALE (Tipton [9]) is
an arbitrary Lagrangian Eulerian finite difference program. DYNA2D (Hallquist
[10]) is a Lagrangian finite element program.

0µs

10µs

Explosive 20µs

30µs
Jet
Slug
60µs

Figure 1: Axisymmetric shaped charge simulation.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 165

DYNA2D

X-RAY

Figure 2: Axisymmetric EFP simulation and experimental flash x-ray.

The liner collapse and jet formation of axisymmetric shaped charges are
normally considered to be a primarily axisymmetric event. In contrast, the
folding of an EFP liner into a desirable projectile geometry is inherently three
dimensional, as the liner normally buckles during the folding process. This
buckling behavior results in a final projectile configuration with buckling at the
rear flare section. This means that accurate modeling of EFP warheads should
include full three dimensional continuum simulation. Three dimensional
continuum simulation is at least an order of magnitude more intensive than two
dimensional simulation. As a result, full three dimensional explicit continuum
simulation requires considerable computer resources. In particular, Eulerian
finite difference modeling has historically been extremely computationally
intensive, but often required in order to resolve highly distorted geometries or
eroding material that cannot be resolved using traditional Lagrangian
methodologies.

3 Scalable software
Under the DoD High Performance Computer Modernization Program, Common
High Performance Computing Software Support Initiative (CHSSI) a scalable
parallel version of the CTH Eulerian high rate finite difference program
(McGlaun et al [11]) was developed. This scalable software suite is for the
simulation of multidimensional, multi-material, large deformation, strong shock
wave propagation problems in solid mechanics. As a part of this development,
the ARDEC Energetics and Warheads Division is assessing the performance and
feasibility of performing parallel modeling for the design and development of
warheads technology. A part of this assessment includes benchmarking of
warheads demonstration calculations.

4 Scalable EFP calculations

The ARDEC Energetics and Warheads Division has a long history of warheads
design and technology development. A part of that technology development
includes EFP warheads. As a result, a number of baseline EFP designs have

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
166 Computational Ballistics III

been developed for general investigations and applications. One of these

baseline designs was used for a computational comparison to assess the
performance and feasibility of performing parallel EFP modeling for the
development of warheads technology. This relatively conservative copper lined
EFP warhead produces a final projectile with length to diameter ratio of about
three. A demonstration calculation was run on the 128 node U.S. Army ARL
Silicon Graphics Origin 2000 and the 256 node U.S. Air Force Aeronautical
Systems Center IBM SP2. Figure 3 presents a visualization of the three
dimensional EFP simulation. The three dimensional liner buckling is clearly
evident during the formation of the projectile flare. The flare is desirable for
aerodynamic stability of the projectile. In addition, the liner edge erosion
material process is clearly evident. Figure 4 presents the three dimensional EFP
simulation final shape. The final buckling of the flare region is clearly evident.
Figure 5 presents an experimental streak photograph of the EFP taken at the
ARDEC small scale test area. The experimental result clearly shows the three
dimensional buckling on the projectile flare, as well as the erosion material
debris. The simulation agreement to experiment is excellent.

5 Computational performance

Run time comparison using varied number of CPUs was conducted for scalable
computational performance benchmarking. Although the simulation was run to
completion on both the Silicon Graphics Origin 2000 and the IBM SP2, the
benchmarking was done with a shorter run time in order conserve computer time.
The calculated speed-ups on the Silicon Graphics Origin 2000 are presented on
Table 1. The calculated speed-ups on the IBM SP2 are presented on Table 2.
Based on the results, both machines can be expected to achieve computational
speed-ups of about 120 times using 1024 processors. This is a considerable
increase, as current warheads calculations are typically completed at ARDEC
using a single processor of a Silicon Graphics Origin server using the same
processor and processor speed as the ASC Origin 2000.

Table 1: Scalable computational speed-up comparison for the Silicon

Graphics Origin 2000.

Silicon Graphics Origin 2000

# CPUs Speed Up
1 1.0
4 1.5
8 2.1
32 5.6
1024 119.5*
*=Projected

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 167

0µs

20µs

50µs

100µs

200µs

Figure 3: Three dimensional EFP simulation.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
168 Computational Ballistics III

Figure 4: Three dimensional EFP simulation final projectile shape.

Cu, V=2Km/s

Figure 5: Experimental streak photograph of the same EFP.

Table 2: Scalable computational speed-up comparison for the IBM SP2.

IBM SP2
# CPUs Speed Up
1 1.0
8 1.1
16 2.0
32 3.9
1024 121.7*
*=Projected

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 169

6 Conclusions
The ARDEC Energetics and Warheads Division have successfully developed
and demonstrated scalable parallel warhead mechanics modeling of an advanced
EFP using the scalable parallel CTH computer program. This recent application
of scalable parallel computers has demonstrated approximately that a 120 times
computational speed-up would be expected in a 1024 node machine using the
current scalable CTH software on either a Silicon Graphics Origin 2000 or an
IBM SP2. This pioneering work clearly demonstrates the feasibility of applying
parallel computers for warheads development. Ongoing efforts are concentrating
on modeling of highly computationally intensive continuum simulation of three
dimensional shaped charge behavior.

Acknowledgement
This work was supported in part by a grant of HPC time from the DoD HPC
Centers.

References
[1] Von Neumann, J. and R.D. Richtmyer. 1950. J. Appl. Phys. 21, 232.
[2] Mader, C.L. 1964. “The Two-Dimensional Hydrodynamic Hot Spot.”
Los Alamos Scientific Laboratory report LA-3077.
[3] Wilkins, M.L. 1965. “The Use of One- and Two-Dimensional
Hydrodynamic Machine Calculations in High Explosive Research.”
Proceedings of the Fourth Symposium (International) on Detonation,
Naval Ordnance Laboratory, White Oak, Silver Spring, MD (October).
[4] Baker, E.L. 1991. “An Application of Parametric Optimization to
Analytical Shaped Charge Modeling.” Proceedings of the SCS 1991
Eastern Multiconference, Ballistics Conference, New Orleans, LA (April).
[5] Yates, J.E., M.E. Majerus and R.E. Brown. 1995. “A Panel Code Model
of the Shaped Charge Liner Collapse Process.” Proceedings of the 15th
International Symposium on Ballistics, Jerusalem, Israel (May).
[6] Baker, E.L. and R. Campbell. 1993. “Optimized Wave Shaping of a
Molybdenum Conical Lined Shaped Charge.” Proceedings of the SCS
1993 Eastern Multiconference, Ballistics Conference, Arlington, Va., pp.
3-6 (March).
[7] Baker, E.L., and Liu, A.Y. 1995. “Numerical Optimization of Tantalum
Material Properties for Two-Dimensional Lagrangian Modeling of a
TNAZ Loaded EFP.” Proceedings of the 6th TACOM Combat Vehicle
Survivability Symposium, Monterey, CA (March).
[8] Fong, R., B. Rice and W. Ng. 1998. “Hydrocode 3D Simulation of Non-
AxiSymmetric (NAS) Explosively Formed Penetrator (EFP) Warhead.”
Proceedings of the 17th International Symposium on Ballistics, Midrand,
South Africa (March).

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
170 Computational Ballistics III

[9] Tipton, R. 1991. “CALE User's Manual.” CALE Documentation,

Lawrence Livermore National Laboratory, Livermore, CA, (July).
[10] Hallquist, J.O. 1984. “User’s Manual for DYNA2D -- An Explicit Two-
Dimensional Hydrodynamic Finite Element Code with Interactive
Rezoning.” Lawrence Livermore National Laboratory Report UCID-
18756, Rev. 2.
[11] McGlaun, J.M., Thompson, S.L., and Elrick, M.G. 1990. “CTH: A
Three-Dimensional Shock Wave Physics Code.” Int. J. Impact
Engineering, Vol. 10, pp. 351-360.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 171

Data mining in military systems

A. A. Motta1, A. S. Alves1 & N. F. F. Ebecken2
1
Brazilian Navy Research Institute, Rio de Janeiro, Brazil
2
Federal University of Rio de Janeiro, Rio de Janeiro, Brazil

Abstract
The discovery of knowledge in databases has gained great importance and
interest in recent years, especially in strategical areas where the access to the
information is restricted many times.
There has been a substantial growth in the amount of stored information in
digital media, mainly due to its increasing easiness and cost reduction. However,
the large amount of stored data becomes impracticable for specialists to analyze
through conventional methods. The difficulty for a more accurate analysis of
these data means they may become accumulated useless information. On the
other hand, it is known that in large amounts of data an enormous potential of
extractable information can exist.
This work presents the application of a data mining tool in military databases.
A case study is shown where the mining of rules of association and classification
capable of finding useful information for the improvement of an anti-missile
system in ships was carried out.
The results show that the methodology reached its main objectives, that is,
rules were generated from large databases. However, it is important to remember
that the generation of these rules represents only the initial stage of the system
improvement with the use of data mining.
Keywords: data mining, association rules, classification rules, anti-missile
systems, chaff.

1 Introduction

The finding of new knowledge from a mass of operational data is only possible if
it is analyzed in a joint form. The large amount of stored data can make it
impracticable for specialists to analyze them through conventional methods.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070161
172 Computational Ballistics III

Moreover, difficulty for a more accurate analysis means that this data is
transformed into an accumulation of information without utility, even though it
is known that there might be a great potential of information in it.
The majority of data mining tasks, in special classification and association
rules mining, suffers from strong restrictions in its implementation due the
diverse factors, like, among others, the iterative character of the mining
algorithms, allied to the often very large amount of data available, that limit the
hardware platforms where these can be processed.
Although the generality of the majority of the algorithms developed for
mining tasks of classification and association rules, it is evident that the more
robust and efficient implementations are those where it is taken into account not
only the available hardware, the knowledge of the data and its nature [1].
Therefore, it is believed that the biggest motivation for implementations of data
mining tools for specific applications inhabits in the fact of getting greater
domain of the system, thus facilitating future adaptations and additions of
functionalities, as they show necessary.
Data mining systems can be used in military applications as described by
Cantoni et al [2]. This technique can be especially important for electronic
countermeasures (ECM) development, where the knowledge of a certain threat
parameters, for which the countermeasure will be developed, is sometimes very
limited. Then, the use of techniques and systems of data mining can provide
important information that can contribute for the development effectiveness.
Amongst the ECM used for ships to defend themselves from fire-and-forget
anti-ships missiles, it is the employment of chaff rockets, as described by Vakin
et al [3]. Chaff rockets are rockets loaded with of metallized filaments that, once
in suspension in the atmosphere, form a radar-reflective cloud that provide a
target with the intention to confuse or seduce the missile, as illustrated in figure 1
[4].

Figure 1: Shunting line of route of the missile for cloud action of chaff.

Chaff clouds must be formed in key position so that incoming missiles are
deviated. For this to be accomplished, a chaff rockets launching system must
exist that determines, through the analysis of diverse parameters, the position,
the load opening time, and the amount of rockets necessary to deviate the
missile. The biggest difficulty in this type of countermeasure system
development is the lack of necessary specific threat information, normally kept

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 173

secret. Thus, the mining of data of tactical scenarios involving adverse

conditions and different threats behaviors, meteorological conditions and ships
parameters can lead to the discovery of important information, hopefully useful
to improve the ECM systems.
The computational data analysis possibilities, combined to military
applications restrictions, has been the main motivation for the use of data mining
tools to search new knowledge for the anti-missile systems improvement. The
application of data mining tasks can be extended to other military systems as
well, for, where there is data collected, certainly exists useful implicit
information.
The remaining of this work is organized of the following form: section 2
presents the anti-missile system description, with its parameters and tactical
operation modes; section 3 presents a brief conceptualization of classification
and association rules; section 4 describes as the mining of data is applied to the
systems databases and how it can contribute to their improvement; section 5
shows some experiments and results; and, finally, section 6 presents the final
considerations.

2 Chaff ECM systems

The mission of an anti-missile ECM system is to defend a ship against missile
attacks. In the case of radar-guided missiles, one possible countermeasure is the
coordinated launching of chaff rockets. There are several other types of ECM
available, but this work will be restricted to chaff employment.

2.1 Tactical employment modes

To defend a surface ship against a radar-guided anti-ship missile, a chaff

countermeasure system can, according to a tactical software or a pre-generated
table, produce fire solutions that will depend on the ship behavior,
meteorological data, rocket launchers to be assigned, number and type of rockets
to be launched from each launcher, and chaff load release time, among other
factors, so that the seduction, distraction, or confusion defensive modes are
applied to deviate the incoming missile.
2.1.1 Seduction
Seduction is the tactical mode in which action is taken in the moment that the
missile acquires the target, in the accompaniment phase, also called final
guidance phase. It can also be used against fire control radars to break their
accompaniment. There are two types of seduction usually applied in these ECM
systems: centroid displacement seduction and accompaniment window
displacement (DUMP) seduction [5].
In centroid displacement seduction, chaff cloud(s) is(are), with bigger radar
cross-section than the ship, formed at close distance(s) from the ship, to form a
single initial radar echo that includes the ship and then, relatively, moves away
from it, seducing the missile guidance radar. This mode often demands ship to
maneuver to dislocate the centroid.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
174 Computational Ballistics III

In DUMP seduction, a coordinated action is carried by radar tricking

equipment and chaff rockets. Initially, a calculated radar signal is emitted to
dislocate the displacement window, the direction, or both, of the missile
guidance radar, producing a fake position, whilst chaff rockets are launched, with
the lesser reaction time, in this fake position area to create a cloud that will be
assumed as the target to be pursued by the missile. For this to work, a precise
choice and programming of the chaff rockets is necessary, combined with an
efficient evasive maneuver to move the ship as far as possible from that area.
2.1.2 Distraction
This tactical way is efficient only during the search phase of the missile guidance
radar. It consists of launching a series of chaff rockets in accordance with
optimized patterns for each threat type, presetted in the system, that will offer to
the missile guidance radar a great number of fake targets available, with a high
probability for it to acquire one of these fake targets instead of the ship, initiating
its final attack phase against it.
2.1.3 Confusion
Tactical mode characterized by acting before a missile is launched. Its objective
is to generate fake targets, confusing the enemy surveillance system and thus
making it difficult to assign a target. The chaff cloud role in this mode demands
their placement at the distances up to 10 km from the unit to be protected.
2.2 System requirements
In order to be possible to ECM systems to perform one or more countermeasure
modes, they must comply with certain requirements. The same is observed to the
ammunition, the chaff rockets. These systems present some intrinsic
characteristics that determine and limit their applicability. Amongst others, the
main characteristics are: the number of rocket launchers and their relative
alignment with respect to the ship; the number of simultaneously launchable
rockets per launcher; the rise angle for fixed launchers or, in the case of mobile
launchers, the boundary-values of training and rise; the amount of simultaneous
launchings; the reaction time between the threat detection and the chaff rockets
launching; the time interval between launchings from the same launcher; and the
ammunition characteristics (since the amount of chaff load, cloud opening time,
the rocket range, among others characteristics, have a significant role in the
system performance).

3 Data mining tasks

Amongst the diverse research areas in the data mining, one of most important
and that has received much dedication is the mining of frequent items (or
Frequent Items Mining, FIM) [6]. FIM is the search of subgroups that appear,
with certain frequency, in a data mass, and it appears as main subproblem in
other data mining tasks of great relevance, such as association and classification
rules mining, stated that rules are the most expressive and intelligible forms of
human knowledge [7–9].

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 175

3.1 Association rules

According to Liu et al [10], “association rule mining finds all the rules existing
in the database that satisfy some minimum support and minimum confidence
constraints”. They represent standards where the occurrence of events in a set is
high [6]. It can be said that it is the probability of a set of items to appear in a
transaction, given that another set is present. An association rule is an
implication of form R: X⇒Y, where X, the antecedent, and Y, the consequence,
are subsets of disjoint items.
The support of a rule R in a database D is defined as sup(X ∪ Y). Thus, such
rule can be understood as a prediction, since if a transaction T supports X, it will
also support Y, with certain probability, which is called rule reliability and
denoted as conf (R):

P (Y ⊆ T ∧ X ⊆ T ) sup( X ∪ Y ) (1)
conf (R ) = P (Y ⊆ T | X ⊆ T ) = =
P( X ⊆ T ) sup( X )

The confidence of a rule discloses how much this rule is applicable, whereas
its support indicates how much this rule is trustworthy. Therefore, for a rule to be
relevant, it is necessary to have sufficient support and confidence, and it is said
that a rule is relevant or strong if its support and confidence are above some
predetermined minimum values [7].

3.2 Classification rules

According to Liu et al [10],”classification rule mining aims to discover a small

set of rules in the database that forms an accurate classifier”. It is the process of
class prediction for records present in a database. Considering a standard as a set
of attribute values, a classification rule is the implication of a standard with a
class. It must be capable of making a record prediction depending on if its
antecedent is a subset of this record. A rules-based classifier is composed of a set
of rules set ordered by some anteriority norm and a standard class (its default).
The problem of classification involves the discovery of relationships (or
dependences) between class variables other variables, using such relationships to
classify unknown cases. These relationships are stored as classification models,
in form of rules [10, 11].

4 Data mining employment in ECM

ECM systems possess specialized software responsible for the execution of its
tactical modes. This tactical software receives, from the others systems and
sensors in the ship, the necessary parameters for the scenario creation, or threat
instancing. From the interpretation of this scenario, the tactical mode designated,
ammunition availability, and the status of the system units, the software
calculates a solution that consists, basically, of the selection of the number of

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
176 Computational Ballistics III

chaff rockets to be launched from each launcher and the opening time of the
loads (or fuze temporization for each rocket).
There might have cases in which, due to reaction time requirements, lack of
available ammunition in one or more launchers, or even other diverse failures, it
is not possible to calculate or to execute a solution, and, for these cases, these
systems are endowed with tables (look-up tables) composed of diverse tactical
scenarios with their respective standard solutions, representing an alternate way
to find a possible defensive solution. Computer simulation is intensively used to
validate to tactical software and to elaborate the look-up tables.

4.1 Data simulation

Simulation is a fast and non-expensive method to test a system behavior.

However, it suffers from an important limitation: the validity of the employed
models. It is extremely difficult to model any complex system, its difficulty
growing with its complexity, and it must be established a commitment between
accuracy and feasibility.
The simulator that generates the data to be used by the countermeasure
system simulates the attack of a missile against a ship, in which the attacking
missile radar tracks the target, while the ship searches to defend herself from the
incoming threat through the use of chaff rockets. For that, diverse factors are
considered, including surrounding meteorological data; missile-target geometry;
attacking missile characteristics (e.g., physical, aerodynamical, sensors, and
control system); own ship characteristics (e.g., physical, electromagnetic,
hydrodynamic, propulsion and maneuver systems); countermeasure system
characteristics (e.g., launchers type and positioning in the ship, launching control
system); rockets characteristics (e.g., physical, aerodynamics, propulsion); and
rockets payload characteristics (e.g., physical, electromagnetic, aerodynamics).
The modeling of the entities that composes the simulation environment is the
great challenge in the simulation process, because, to ensure useful results, it is
necessary to well model these entities (e.g., own ship, threat, chaff cloud, etc).
For this, some usually unavailable data are necessary, such as the ship radar
cross-section for all possible angles and the exact characteristics of the threat
search radar, among others.

4.2 Data mining contributions

Considering the limitations of simulation and the amount of involved variables in

the problem, it is easy to conclude that any analysis for simple data inspection
hardly leads to some type of contribution, what motivates the application of data
mining tasks in this problem.
Diverse data mining data can be applied to simulation results and contribute,
somehow, for the countermeasures systems improvement. In this work,
association and classification rules are applied to improve the survivability
chances of a ship equipped with a chaff system under attack by an anti-ship
missile.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 177

4.2.1 Association rules mining

Association rules constitute a tool of non-supervised computational learning. It is
based on observations and discoveries, searching for occurrences of standards,
trends and generalizations in the data mass without user intervention and the
application of this task can lead to surprising discoveries.
Its most immediate application in the current problem (launching of chaff
rockets to defense a ship under an anti-ship missile attack) is the use of the
mined information as an evaluation instrument for the simulation outcome. They
can disclose, for instance, that a solution or even a simulation behavior
associated to events characterize a wrong trend of a model, showing that the
model must be reviewed and corrected or improved and that certain assumptions
might not be applicable, what can improve the ECM system effectiveness, since
the data are used to produce a defensive solution or an entry in the look-up table.
Association rules can also reveal occult characteristics in the problem (such
as, for instance, unnoticed details regarding the missile search pattern) that can
contribute not only to improve the defense system, but lead to tactical or doctrine
re-evaluation on chaff employment. In addition, based in the knowledge of the
available data, some user interventions can be applied, like the use of some type
of hierarchy between the records, and possibly leading to other useful
discoveries.
Association rules can be used to create parameters to help to understand
threats behavior, to identify affinities between solutions and tactical scenarios, to
foresee solutions for new scenarios, to perform behavior analysis for fault
detection, etc, demonstrating that their use in such databases can contribute for
the ECM systems improvement, and that the application possibilities are
innumerable.

4.2.2 Classification rules mining

Liu et al [10] showed that it is possible to integrate association and classification
rules mining, so that it is possible, starting from association rules, to perform
classification tasks implementing a few algorithm changes that will allow the
employment of both mining rules without performance loss. It must be pointed,
however, that classification is a supervised method, that is, it is learned with the
use of examples, being applied for prediction.
In this work, classification of tactical scenarios was applied in classes of
defense sectors. The area near the ship is divided in sectors in accordance with
the launcher to be used in a scenario solution, so that, in ships equipped with four
launchers, four classes were created. Neglecting the combined solutions (those
where the use of more than one launcher for the defense of the ship is necessary),
the result of the analysis of this process can be used, for instance, in the
reloading politics procedure for situations where there is not plenty of
ammunition available.
There are instances of the problem (scenarios) for which it is not possible to
find a solution, and the grouping of these instances in two classes (with solution
and without solution) can help in the chaff rockets application doctrines
development. In addition, a real time scenario classification can allow the ship to

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
178 Computational Ballistics III

engage in other defense method. Other parameters can be used for the scenarios
separation in classes, such as, for instance, the necessity or not of maneuvering
by the ship. Even in cases for which missiles do not directly hit the target, nearby
explosions often cause expense damage to the ship systems, especially in the
detection and communications antennas.
Like the association mining rules, classification mining rules can collaborate
for the ECM systems improvement and its use doctrine.

5 Case of study and speed up

One of the biggest challenges in the mining rules tasks use is the algorithm
implementation. Due to their iterativity aspect, to the great number of generated
combinations, and mainly the usually very large size of the databases,
implementations must be constructed carefully, using optimized data structures
so that computational costs do not make them unfeasible.
A tool has been developed in C programming language, capable of effective
execution of the rules generation task for application in the databases generated
for a chaff rockets employment environment simulator. Its initial task is a data
pre-processing, carried through the databases produced by the simulator, in
which this data is converted to a proper format, suitable for the mining tasks
implementations, some parameters discretization, data cleanness, ordering, and
attributes choice.
The produced databases did not present a priori problems for the serial
mining rules tasks execution (in one machine). After verification that the rules
were being correctly mined and to ensure that a reduced set of classification rules
was extracted, as suggested by Liu et al [11], a parallel version was developed, to
reduce to overall consumed time with direct impact in its feasibility.
The experiments were carried in a PC cluster with 16 Pentium IV HT 2.8GHz
with 1GB of RAM computers, 8 Pentium IV 2.4GHz with 1GB of RAM
computers, and 16 AMD Dual-Core Opteron with 2GB of RAM. In the
executions, the node selection was intentionally set as automatic, so that there is
no identification of the used processors in each run. The implementation includes
the development of a parallel implementation of CBA algorithm [11], including
associations-based classification – successfully strategy already in use in serial
implementations. Figure 2 shows the speed up obtained with the use of up to 10
processors and databases varying from small to large sizes. It can be observed
that, as the database size increases, the speed up approaches the theoretical value.
Due to confidentiality issues, it is not possible to disclosure further details
regarding the database.
It is important to point out that, in very dense databases, there is a risk of
combinatory explosion in association or classification rules mining applications,
making its usage impracticable quite often. However, the results showed that the
methodology worked fine, being possible to extract rules from the experimental
databases and that, from the results of the speed up study, there is an indication
that its application in bigger databases is not only possible, but feasible.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 179

Speed Up According to the Number of Processors

for Databases with Various Number of Records
10

8
Speed UP

0
1 2 3 4 5 6 7 8 9 10
Number of Processors

Theoretical 2558 20463 61387 83204 94949

Figure 2: Speed up according to the number of processors for databases with

various numbers of records.

6 Final remarks
The association or classification rules mining from simulated databases
represents the initial stage of this work. For the attainment of results that come to
effectively contribute for the ECM systems improvement, it is necessary to give
continuity to the work. It is necessary to establish more precisely the parameters
of minimum support and minimum confidence, which directly influence the
methodology effectiveness. Very high values of minimum support can prevent
the mining of important rules or reduce too much the classifier precision. Thus,
after tuning the rules generator tool, it will be necessary to perform an analysis of
the disclosed information by the data mining tasks so that the new knowledge
can be added to the application.
In addition, it is possible to include tasks execution improvements, such as
data hierarquization or the inclusion of other data mining tasks, among others,
extending, in this way, the possibilities of discovery of new knowledge in this
area of application. Finally, this work introduces the application of data mining
tasks with the purpose to improve anti-missile systems for ships. However, the
methodology is not restricted to the present case study, rather non-military
problems can also benefit from this methodology, and the methodology just
needs to be applied to other individual problems.

References
[1] Adamo J-M., Data Mining for Association Rules and Sequential Patterns:
Sequential and Parallel Algorithms, Springer, New York, pp 66-74, USA
2000.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
180 Computational Ballistics III

[2] Cantoni V., Lombardi L., Lombardi P., Challenges for Data Mining in
Distributed Sensor Networks. In The 18th International Conference on
Pattern Recognition. Proc. of the ICPR'06, pages 1000-1007, China, Hong
Kong, August 2006.
[3] Vakin S. A., Shustov L. N. & Dunwell R. H., Fundamentals of Electronic
Warfare. Artech House, INC. Norwood, MA, USA 2001.
[4] http://en.wikipedia.org/wiki/Chaff_%28radar_countermeasure%29, last
accessed on January 30, 2007.
[5] http://www.sew-lexicon.com, last accessed on January 30, 2007.
[6] Agrawal R, Imielinski T. & Srikant R., Mining Association Rules between
Sets of Items in Large Databases, Proc. of the ACM SIGMOD Intl. Conf.
on Management of Data, pages 207–216, Washington, USA, 1993.
[7] Agrawal R., Mannila H., Srikant R., Toivonen H. & Verkamo A. I., Fast
discovery of association rules. In U. M. Fayyad, G. Piatetsky-Shapiro, P.
Smyth, and R. Uthurusamy, editors, Fayyad U. and et al, 1995.
[8] Agrawal R. & Srikant R. Mining sequential patterns. In Proc. 11th Int.
Conf. Data Engineering, ICDE, pages 3–14. IEEE Press, 1995.
[9] Beeferman D., BERGER A., & LAFFERTY J. Statistical models for text
segmentation. Machine Learning, 34(1-3):177.210, 1999.
[10] Liu B., Hsu W., & Ma Y.. Integrating classification and association rule
mining. In Proceedings of the Fourth International Conference on
Knowledge Discovery and Data Mining (KDD-98), pages 27–31, 1998.
[11] Liu B., Ma Y., & Wong C. Improving an association rule based classifier.
In 4th European Conference on Principles and Practice of Knowledge
Discovery in Databases PKDD, pages 504–509, 2000.
[12] Girish K., Palshikar A, Mandar S., Kale B & Manoj M., Association rules
mining using heavy itemsets, In Tata Research Development and Design
Centre, Engineering and Industrial Services, Tata Consultancy Services
Ltd., Pune 411001, India, 2006.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 181

Computer simulated fragmentation arena test

for assessing lethality and safety separation
distances of explosive fragmentation
ammunitions
V. M. Gold, E. L. Baker & J. M. Pincay
U.S. Army RDEC-ARDEC, AMSRD-AAR-AEE-W, Picatinny Arsenal,
NJ, USA

Abstract
A combined analytical and experimental technique for determining explosive
fragmentation ammunition safe separation distance and lethality without costly
arena fragmentation tests is presented. The new methodology integrates
high-strain high-strain-rate computer modelling with semi-empirical analytical
fragmentation modelling and experimentation, offering warhead designers and
ammunition developers more ammunition performance information for less
money spent. The new method provides more detailed and accurate warhead
fragmentation data for ammunition safe separation distance analysis, as
compared to the traditional fragmentation arena testing approach.
Keywords: fragmentation modelling, fragmentation arena test, lethality, safety
separation distance.

1 Introduction
This paper reports on a combined analytical and experimental technique for
determining explosive fragmentation ammunition safe separation distances and
lethality without costly arena fragmentation tests, offering warhead designers
and ammunition developers more ammunition performance information for less
money spent. Since the main objective of the developed procedure is to simulate
the “data output” (i.e. warhead fragmentation performance information) from
traditional fragmentation arena tests, a brief description of the fragmentation
arena test methodology is given for completeness.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070171
182 Computational Ballistics III

Basic principles and requirements governing fragmentation arena test

procedures in the U.S.A. are stipulated in Joint Munition Effectiveness Manual
[1]. In a typical fragmentation arena test set-up, tested munitions are positioned
at the origin of the reference polar coordinate system and surrounded with series
of velocity-measuring screens and fragment-catching witness panels, all at
significant distances from the warhead. Defining the longitudinal axis of the
munition as the polar axis z, the polar altitudinal angles Θ are measured from the
munition’s nose (Θ=0°) to the tail (Θ=180°), and the azimuthal angles φ are
measured from an arbitrary projectile’s feature (φ=0°) in a counterclockwise
direction. In conventional fragmentation arena test procedures fragment
sampling and fragment velocity measuring is usually limited to relatively small
azimuthal sections, mainly because of enormous construction and data
assessment costs associated with recovering fragments from the entire
fragmenting shell. This sampling technique requires the assumption of isotropic
fragmentation properties for all azimuthal angles φ throughout the entire Θ-angle
zone (i.e. a complete altitudinal region bounded by two polar angles). By
sampling small azimuthal angles across all polar zones from the munition nose to
tail and adjusting this sample data mathematically, a prediction for entire
munition fragment characterization is obtained. Since only a small azimuthal
section of the region is sampled and scaled up, even relatively small munition
positioning errors may result in large deviations of the fragmentation data,
routinely requiring repeated testing for statistical data stability.
In this work, the assessment of the ammunition fragmentation parameters was
performed analytically employing the PAFRAG (Picatinny Arsenal
Fragmentation) code which links three-dimensional axial symmetric high-strain
high-strain-rate continuum analyses with a phenomenological fragmentation
model validated through a series of experiments including flash radiography,
Celotex™ and water test rear fragments recovery, and sawdust total fragment
spray recovery.
In fragmentation arena tests, the ammunition fragmentation characteristics are
assessed as functions of polar angles Θ identifying angular positions of
fragment-catching witness panels and velocity-measuring screens. In PAFRAG
code analyses, positions of these devices are irrelevant, and the fragmentation
characteristics are assessed in reference to the fragment trajectory angles Θ'
calculated from the CALE code [2] cell velocities at the time of the shell break-
up. Once the shell breaks up and fragments are formed, fragment velocities may
change with time due to a number of reasons, including the air drag and the rigid
body motion induced at the time of the shell break up. Assuming that the
fragment trajectory angles Θ do not change with time (that is the rigid body
motion and the lateral drift of fragments due to air resistance is relatively small)
and that the definitions of angles Θ and Θ' are approximately identical, the
PAFRAG model enables prediction of crucial characteristics of explosive
fragmenting munitions including the number of fragments, the fragment size
distribution, and the average fragment velocities.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 183

2 PAFRAG modelling
Similarly to fragmentation arena test fragment sampling assumptions, the
PAFRAG fragmentation model assumes that for any point within a fixed Θ-angle
zone the fragment number distribution Nj(m) is uniform and independent of the
altitude and the azimuth angles Θ and φ, respectively. Hence, the total fragment
number distribution is given by
L
N (m ) = ∑ N j ( m). (1)
j

In equation (1) m is the fragment mass, L is the number of altitudinal Θ-angle

zones, 0≤Θ≤π, and Nj(m) is the fragment number distribution function for the j-
th zone. For convenience, all Θ-zones are assumed to have the same altitudinal
lengths of ∆Θ=π/(L-1), except for the first and the last “half-length” zones with
lengths of ½∆Θ. In the fragmentation ammunition arena testing practice adopted
at the US Army Armament Research Development and Engineering Center, the
number altitudinal zones is usually L=37, resulting in uniform Θ-angle resolution
of ∆Θ=5°. Accordingly, the Θ-zones are identified by the middle of the zone
altitudinal angles Θ j given by the following series

 1 4 ∆Θ, j = 1,
 (2)
Θ j = ∆Θ( j − 1), 2 ≤ j ≤ L − 1,
π − 1 ∆Θ, j = L.
 4

In the case of traditional fragmentation arena testing, all individual fragment

number distribution functions Nj(m) for all polar Θ-zones are determined directly
from the test data. The main drawback of this approach is the extremely high
testing costs limiting the fragmentation arena testing to final ammunition
fragmentation characterization. Alternatively, the PAFRAG modeling and
experimentation is a relatively low-cost procedure enabling accurate assessment
of the fragmenting munition performance at the research, design, and
development phases. In the PAFRAG approach the individual Θ-zone fragment
number distribution functions Nj(m) are computed analytically from the sawdust
or water tank fragment recovery test data, N(m). Mathematically, the PAFRAG
fragmentation modeling is a solution of the inverse problem of equation (1), i.e.
determining a series of individual Nj(m)’s for given N(m). Since with PAFRAG
approach, the N(m) function is assessed based on approximately 98-99%
fragment recovery data, the accuracy of PAFRAG predictions is high.
Figure 1 shows results of high-strain high-strain-rate CALE modeling and
flash radiographic images of a representative natural fragmentation warhead at
30 and 50 µs, and at 300 and 500 µs after detonation. Upon initiation of the high
explosive, rapid expansion of high-pressure high-velocity detonation products
results in high-strain high-strain-rate dilation of the hardened fragmenting steel
shell, which eventually ruptures generating a “spray” of high-velocity steel

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
184 Computational Ballistics III

fragments. As shown in the model, the rear end of the warhead has a cylindrical
cavity for the projectile tracer material. Following the expansion of the
detonation products, the tracer holder fractures and the resulting fragments are
projected in the negative direction of the z-axis, without contributing to the
warhead lethality but posing potential danger to the gunner. As evidenced from
the series of flash radiographic images shown in Figure 1, the tracer holder
section of the warhead breaks up into a number of relatively large fragments that
may cause serious or fatal injuries to the gunner.

Fragmenting shell Fragments

Tracer v
holder Θ
Mock fuze R
z

Explosive
t=0 µsec t=12.5 µsec, V/V0=3

Rear fragments t=30 µsec

t=300µsec & 500µsec t=50 µsec

Figure 1: Results of CALE modeling and flash radiographic images of a

natural fragmentation warhead at 30 and 50 µs (test No. X-969),
and at 300 and 500 µs (test No. Y-070) after detonation.
CALE analyses had been conducted until approximately 30 µs after the
charge initiation. As shown in the figure, CALE modeling results are in very
good agreement with flash radiographic images of the fragmented warhead. The
fundamental assumption of all fragmentation analyses presented in this work was
that the fragmentation occurs simultaneously throughout the entire body of the
shell. Following Pearson [3], the fragmentation of steel shells with the idealized
cylindrical geometries occurs approximately at 3 volume expansions, the instant
of fragmentation defined as the time at which the detonation products first
appear emanating from the fractures in the shell. Accordingly, at approximately
3 volume expansions (12.5 µs), the fragmenting steel shell was assumed
completely fractured, and the CALE-code cell flow field data was passed to
PAFRAG-MOTT and PAFRAG-FGS2 fragmentation modeling.
For a large part the PAFRAG-MOTT fragmentation model is based on the
Mott’s theory of break-up of cylindrical “ring-bombs” [4], in which the average
length of the resulting circumferential fragments is a function of the radius and
velocity of the ring at the moment of break-up, and the mechanical properties of

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 185

the metal. Accordingly, in the PAFRAG-MOTT model the “random variations”

in fragment sizes of natural fragmentation warheads are accounted through the
following fragment distribution relationship

1/ 2
 
L − m 
N (m ) = ∑ N 0 j e  µj 
. (3)
j

In eqn. (3) N0j and µj represent number of fragments and one half of the average
fragment mass in the j-th Θ-zone, respectively, computed from the CALE-code
data.

All Fragments, Test X-972

Number of fragments, N/Nexp

1.2 All Fragments, Test X-973

All Fragments, PAFRAG-Mott, γ=14
(m0, N0) 0.03 All Fragments, PAFRAG-Mott, γ=20
1.0
Number of fragments, N/Nexp

All Fragments, PAFRAG-FGS2

0.02
0.8
0.01 (m2, N2)
0.6
0.00
(m1, N1)
0.4 6 9 12 15 18
Fragment mass, m/µ 0
0.2

0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5
Fragment mass, m/µ 0

Figure 2: Cumulative number of fragments versus fragment mass, N=N(m),

for small-to-moderate weight (m/µ0 < 5.5) and relatively large
(m/µ0 > 5.5) fragments.

The PAFRAG-FGS2 fragmentation model is defined in parametric form as

 L mj 
 N k (ξ k )  ∑ L
( a N 0 k + a N 1k ξ k + a N 2 k ξ k2 + a N 3 k ξ k3 ) 
(4)
 m(ξ )  =  ∑ m j .
j

 k   j 
 a + a ξ + a ξ 2
+ a ξ 3 
 m0k m1k k m2k k m 3k k 

In eqn. (4) ξ k is a non-dimensional parameter, 0 ≤ ξ k ≤ 1 , k is the curve index,

k=0,1, and sixteen coefficients aN’s and am’s are obtained by fitting two curve
segments k=0 and k=1 with conditions of curve and tangent continuity at the
adjacent ends.
Results of the PAFRAG modeling are given in figures 2 through 5. Figure 2
shows plots of the cumulative number of fragments versus fragment mass for

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
186 Computational Ballistics III

small-to-moderate weight (m/µ0 < 5.5) and for relatively large (m/µ0 > 5.5)
fragments calculated with PAFRAG-MOTT and PAFRAG-FGS2 models. As
shown in the figure, attempting to fit the sawdust fragment recovery data with
the PAFRAG-MOTT model by changing parameter γ “rotated” the curve, but did
not yield an accurate fit to the data. Accordingly, more a “flexible” PAFRAG-
FGS2 model was applied. As shown in the figure, using the PAFRAG-FGS2
model resulted in accurate fit throughout the entire range of data. Accordingly,
PAFRAG-FGS2 model was used for all further analyses.

Experimental Data, X-969, t=29.4µsec

Experimental Data, X-965, t=125.2µsec Image at 49.9 µsec
Experimental Data, Y-070, t=300.0µsec
Experimental Data, X-964, t=310.9µsec Θ
CALE-PAFRAG Analyses,
0.18 Cell Data, t=12.5µsec (V/V0=3)
CALE-PAFRAG Analyses, Average, t=12.5µsec (V/V0=3)
CALE-PAFRAG Analyses,
0.15 Adjusted, t=12.5µsec (V/V0=3)

0.12
Velocity, cm/µs

0.09

0.06

0.03

0.00 Image at 29.4 µsec

0 30 60 90 120 150 180
θ, degrees

Figure 3: Fragment velocities versus fragment spray angle Θ and flash

radiographic images at 29.4 and 49.9 µs (test No. X-969) after
detonation.

Figure 3 shows the PAFRAG model fragment velocity predictions compared

with the experimental data. The experimental values of fragment velocities of
the main fragment spay (80°≤Θ≤100°) were obtained from the flash radiographic
images at 29.4 and 49.9 µs. Velocities of the rear fragments broken off from the
tracer section of the shell (which move significantly slower than fragments from
the main spray) were assessed from the flash radiographic images at 125.2, 300.0
and 310.9 µs. PAFRAG model prediction of the “average” Θ-zone fragment
velocities was obtained from the momentum averaged CALE-code flow field
cell velocities. As shown in the figure, the agreement between the PAFRAG
model fragment velocities predictions and the data is good.
Figure 4 shows PAFRAG model predictions of the fragment mass distribution
versus the spray angle Θ; the zonal fragment mass mj and the cumulative
fragment mass M distribution functions were computed from CALE-code cell
flow field data. For representation clarity, the cumulative fragment mass
function M is defined in terms of angle 180°-Θ, not the spray angle Θ. As shown
in the figure, the PAFRAG model prediction of the cumulative fragment mass

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 187

distribution M is in good agreement with the available experimental data at

Θ=161.6° (the Celotex™ and the water test fragment recovery) and at Θ=180°
(the sawdust fragment recovery).

PAFRAG Analyses
Sawdust Recovery, Tests X-972 & X-973

Cumulative mass, M/M0

TM
z Θ=0° Θ=180° 1.2 Celotex Fragment Recovery, Test X-964
1.0
Θ 0.8
0.6
0.4 Θ=161.6o

Lethal fragments 0.2

0.3 0.0
Mass, mj/M0

0 30 60 90 120 150 180

0.2 180-Θ, degrees
No lethality Rear fragments
0.1
Θ=180° z Θ=0°
0.0
0 30 60 90 120 150 180
Θ, degrees Θ=161.6°

Figure 4: PAFRAG analyses of fragment mass distribution versus Θ.

Cumulative fragment mass distribution from PAFRAG analyses is
in excellent agreement with experimental data.

As shown in figure 4, PAFRAG modeling predicts that the majority of the

munition‘s fragment spray is projected into a relatively narrow Θ-zone in the
direction perpendicular to the projectile’s axis, approximately at angles
80°≤Θ≤100°. This is in good agreement with the flash radiography data
showing no fragments projected to the projectile’s anterior region, 0°≤Θ≤50°.
The fragment velocity “spikes” in the region of 0°≤Θ≤50° (see fragment velocity
plot, Fig. 3), are due the numerical “noise” from a few “stray” mix-material
computational cells from the CALE modeling. Because there is no considerable
fragment mass in the front Θ-zones, the overall effect of these errors is
negligible, and the “average” fragment velocity in the 0°≤Θ≤50° region should
be disregarded.
As evidenced from the flash radiographic images presented in Fig. 4, the
tracer holder portion of the warhead breaks up into a number of relatively large
fragments projected in the negative z-axis direction, back towards the gunner.
As shown in Fig. 4, in excellent agreement with the Celotex™ and the water test
fragment recovery data, PAFRAG modeling predicts that approximately 7.2% of
the total fragment mass is projected to the “rear”, in the region of
161.6°≤Θ≤180°. Since according to PAFRAG modeling and the flash
radiography data, Fig. 3, the velocities of these fragments is approximately 0.05
cm/microsecond, the broke-up pieces of the projectile’s tracer holder are capable
of causing serious injuries or death to the gunner.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 188 Computational Ballistics III

Figure 5 shows PAFRAG-FGS2 model predictions of the cumulative number

of fragments versus fragment mass, N=N(m), and of the Θ-zonal number of
fragments versus Θ, Nj=Nj(Θ), for both the total “all fragments” and the “rear
only” (161.6°≤Θ≤180°) modeling cases. The “all fragments” fragment
distribution was assessed from the sawdust fragment recovery tests that included
fragments from the tracer section together with all fragments from the entire
shell. The “rear only” fragment distribution was obtained from the Celotex™
and from the water test fragment recovery experimentation and accounted only
for fragments projected at angles greater than approximately 161.6°. The
limiting rear fragment collection angle of Θ=161.6° represents the altitudinal
angle Θ covering the fragment recovery surface area.

Sawdust Fragment Recovery, All Fragments, Test X-972

Sawdust Fragment Recovery, All Fragments, Test X-973
Celotex Fragment Recovery, Rear Fragments, Test X-964
W ater Fragment Recovery, Rear Fragments, Test Y-070
75
Cumulative number of fragments, N

PAFRAG Analyses, All Fragments, FGS2

PAFRAG Analyses, Rear Fragments, FGS2
PAFRAG, m > 10 grains, All Fragments
Number of fragments, Nj

60 6 PAFRAG, m > 20 grains, All Fragments

PAFRAG, m > 40 grains, All Fragments
PAFRAG, m > 60 grains, All Fragments
All fragments, average PAFRAG, m > 10 grains, Rear Fragments
45 4 PAFRAG, m > 20 grains, Rear Fragments
PAFRAG, m > 40 grains, Rear Fragments
PAFRAG, m > 60 grains, Rear Fragments

30 2
Θ=161.60
Rear fragments, upper bound
15 0
130 140 150 160 170 180
Θ, degrees
0
0 10 20 30 40 50 60 70
Fragment mass m, grains

Figure 5: Cumulative number of fragments versus fragment mass and

number of fragments versus Θ, for total “all fragments” and “rear
only” (Θ > 161.6°) distributions.

As shown in Fig. 5, the “rear fragments” PAFRAG-FGS2 model fragment

distribution was obtained by fitting eqn. (3) to the upper bound of the Celotex™
and water test recovery data, providing an additional “safety” margin for the safe
separation distance analyses. Since in a typical fragmentation warhead only a
few fragments are projected backward towards the gunner, establishing a
statistically robust database from the conventional fragmentation arena test
requires repeated experimentation and is expensive. In contrast, the data from
the PAFRAG modeling offers to munition designers more warhead performance
information for significantly less money spent. The PAFRAG provides more
detailed and more statistically accurate warhead fragmentation data for
ammunition safe separation distance analysis, as compared to the traditional
fragmentation arena testing approach.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 189

3 PAFRAG model assessment of lethality and safety

separation distance

The safety separation distance analyses presented in this work were performed
employing the JMEM/OSU Lethal Area Safety Program for Full Spray
Fragmenting Munitions code [6] and the Wedge model computational module.
According to [7], the safe separation distance is defined as fixed distance from
the weapon’s launch platform and personnel beyond which functioning of the
munition presents an acceptable risk of a hazard to the personnel and the
platform. Accordingly, the safe separation hazard probability had been calculated
based on the warhead’s fragment spray ability to strike and to penetrate exposed
(bare) skin tissue of unprotected gun crew personnel. According to [7], the
maximum total risk to the munition crew at safe separation distance is generally
accepted as 10-6.

Vz =0
Vz =0.25V0
Vz =0.5V0
Vz =0.75V0
Vz =V0

Θ=180° Θ=0° Θ=180° Θ=0°

0.8 0.6 0.4 0.2 0.2 0.4 0.6 0.8 4.0 3.0 2.0 1.0 1.0 2.0 3.0 4.0
R/R0 R/R0

0.1≤Pi≤1.0 Pi≤10-6

Figure 6: Areas of 0.1≤Pi≤1 and Pi≤10-6 unprotected personnel risk hazards

for varying projectile lunch velocities.

The input for the lethality and safe separation distance analyses included a
range of possible ballistic projectile trajectories and the static PAFRAG FGS2
model predictions of the fragment spray blast characteristics. Figure 6 shows
resulting plots of areas with 0.1≤Pi≤1 and Pi≤10-6 unprotected personnel risk
hazards for varying projectile lunch velocities. As shown in Fig. 6, the projectile
launch velocity has a significant effect on both the munition lethality (0.1≤Pi≤1)
and the safety (Pi≤10-6). As shown in the figure, if the gun operates normally
and launches the projectile with the nominal muzzle velocity of V0, all fragments
are projected in the forward direction, posing no danger to the gun crew.
However, if the gun misfires (Vz<< V0) and the munition is detonated, the results
may be catastrophic.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
190 Computational Ballistics III

Acknowledgements
Messrs. K. P. Ko of PM MAS and A. N. Cohen of US Army RDEC-ARDEC are
acknowledged for providing funds that made this work possible. Mr. K. W. Ng
of US Army RDEC-ARDEC and Mrs. Y. Wu of SAIC are acknowledged for
their contribution in CALE modeling. Mr. W. J. Poulos of US Army RDEC-
ARDEC is acknowledged for performing experiments. Mr. T. Fargus and Mrs.
D. L. Snyder of US Army RDEC-ARDEC are acknowledged for performing
lethality and safety separation distance analyses.

References
[1] Joint Munition Effectiveness Manual. “Testing and data reduction
procedures for high-explosive munitions”, Report FM 101-51-3, Revision
2, 8 May 1989.
[2] Tipton R. E., “CALE users manual”, Version 910201, Lawrence
Livermore National Laboratory, 1991
[3] J. Pearson, “A fragmentation model for cylindrical warheads”, Technical
Report NWC TP 7124, Naval Weapons Center, China Lake, California,
December 1990.
[4] N. F. Mott, F.R.S., “Fragmentation of steel cases”, Proc. Roy. Soc., Vol.
189, pp. 300-308, 1947.
[5] J. C. Ferguson, “Multivariable curve interpolation”, Report No. D2-
22504, The Boing Co., Seattle, Washington, 1963.
[6] Joint Technical Coordination Group for Munitions Effectiveness.
“Computer program for general full spray materiel MAE computations”,
Report 61 JTCG/ME-70-6-1, 20 December 1976, Change1: 1 April 1991.
[7] “Guidance for Army Fuze Safety Review Board safety Characterization”,
US Army Fuze Office, January, 1999.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Section 6
Systems and technology
This page intentionally left blank
 Computational Ballistics III 193

Evaluations of computational techniques for

the engraving of projectiles
J. South, B. Powers & M. Minnicino
US Army Research Laboratory, Aberdeen Proving Grounds, MD, USA

Abstract
This paper presents research on the evaluations of computational methods used
to model the engraving of small caliber projectiles. The launch mechanics of a
5.56 mm projectile fired from a rifled barrel based on a M16A2 are examined in
this study. The interaction between small caliber projectiles and rifled barrels is
evaluated using finite element analysis (FEA) simulations utilizing the 3D
explicit dynamic FEA LS-DYNA code. The calculations were performed to
understand the interaction of the projectile during the initial portion of launch,
where the projectile is subjected to the dynamic pressure from the combustion of
the charge. The modeling parameters investigated include the order of element
integration, the hourglass control scheme, and contact algorithms. The stress and
strain state within both the projectile and the barrel are evaluated and compared
to experimental data. The results show that a comparison of global hourglass
energy to internal energy can lead to improper conclusions about the validity of
the engraving model. The choice of the hourglass scheme and contact algorithm
affects the hourglass energy and may affect the internal energy. This effort has
led to increased insight into the behavior of the engraving process on small
caliber projectiles and the numerical techniques necessary to reduce modeling
difficulties.
Keywords: engraving, projectile, hourglass.

1 Introduction
Correct simulation of the engraving process is crucial for the design of
obturation and prediction of in-bore dynamics. In addition, given realistic
constitutive models, consideration of the utility of different materials for the
bullet can be made. Unfortunately the engraving process is a difficult

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070181
194 Computational Ballistics III

phenomenon to model. During the engraving process excessive element

deformations often lead to erroneous results. A possible solution is to utilize
element erosion; however, this approach assumes that the elements and the
corresponding volume of material engraving is worn away, which may not
capture the effect of material yielding and flow around the rifling of a barrel
during the launch of a projectile. Two dimensional modeling can be applied
[1,2] to evaluate the axi-symmetric interaction of a projectile with a barrel but
this approach precludes a true rifling profile with lands and grooves. Full three
dimension modeling is required to properly evaluate the engraving process.
In this study evaluations of the computational techniques for the engraving of
5.56 mm projectiles were performed. The interaction between small caliber
projectiles and rifled barrels was evaluated using finite element analysis (FEA)
simulations utilizing the 3D explicit dynamic FEA code LS-DYNA. A
simplifying assumption in these analyses was that the rifling profile was without
twist and that the interactions could be assessed with a quarter symmetry model.
Excessive element deformations and high hourglass energies were present in the
initial modeling efforts and caused numerical difficulties. In an effort to address
these issues, a series of modifications were implemented into the models to
evaluate the effects of the hourglass control scheme, order of integration of the
elements, and the contact algorithm,

2 Model development
Finite element models (FEM) were developed of a 5.56 mm projectile fired from
a rifled barrel based on a M16A2. As previously stated a simplifying assumption
in these analyses was that the rifling profile was without twist. This assumption
provided the ability to generate quarter symmetric models. The standard
projectile evaluated in the models was the M855 projectile. This projectile
contains a hardened steel penetrator backed by a lead-antimony slug encased in a
copper gilding full metal jacket body [3].

Figure 1: Graphics of the quarter symmetry model showing two different

views.

The geometry for both the projectile and the weapon [4] were obtained from
their respective technical drawing packages. The barrel was assumed to be an
ideal barrel with perfect symmetry and centerline, and a greatly reduced length to

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 195

decrease computation time since only initial engraving was to be studied. Axial
boundary conditions were applied to the base of the barrel to prevent its motion.
The base pressure-time curve for the M855 was obtained from the interior
ballistic code IBHVG2.
The simulations were performed using LS-DYNA [5]. Contact was used
between the slug-core-jacket and the jacket-barrel. All of the components within
the projectile were kinematically constrained by contact surfaces. The material
properties for the components in the model were obtained from experimental
data. A graphic of the model is show in Figure 1. The baseline model possesses
six elements thru the thickness of the copper gilding metal and thirty six
elements across the ninety degree sweep of the quarter symmetry. The baseline
contact algorithm was surface to surface with the master and slave pairs
identified by parts. The baseline hourglass control was the stiffness form.

2.1 Quasi-static model of push test

A quasi-static model was used to explore dynamic effects by comparing with the
baseline model. While the material properties defined in the model are not rate
dependent, inertial and transient effects are still present in the baseline model.
Also, the results of quasi-static experiments are available where a projectile was
slowly pushed through a barrel [6]. In these experiments an M855 projectile was
statically pushed thru an M16A2 barrel at a given displacement rate. The load,
displacement of the push rod, and hoop strain on the outer diameter of the barrel
were recorded in a manner similar to that described in [7]. This data allows for a
validation of the model with the experiments.
The geometry of the baseline dynamic model was used for the quasi-static
analysis. Mass scaling was used to artificially increase the density, in turn
increasing the time step through the Courant Condition [5]. This allows quasi-
static simulations to be completed with an explicit finite element code using a
minimum of computational time. The purpose of the “long time” simulation is
to remove any transient and inertial effects from the simulation. The time step
and termination time were arbitrarily chosen to be 5.0x10-5 s and 2.5 s,
respectively. The time scale of the pressure-time history was linearly scaled to
cover the entire defined run time of the model. Mass damping was applied to all
materials in the model to damp out any oscillations in the solution due to solver
noise. A value of 10% of critical damping at 5 Hz was used.

2.2 Hourglass control and contact algorithm effects models

The simulations to evaluate the affect of altering the contact algorithm also
evaluated the incorporation of selective reduced fully integrated elements and
viscous hourglass control. In these models the projectile was altered so that the
internal components of the projectile were steel. This was chosen since previous
work [6,8] has shown that the choice of slug material affects the radial stress
exerted on the rifling lands and the grooves. Steel was chosen so as to
exacerbate this effect as a worst case scenario.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
196 Computational Ballistics III

The contact algorithm and element integration order and their corresponding
model are presented in Table 1. The automatic option for contact in LS-DYNA
changes how the surface normals of the search algorithm on master surface are
calculated. The default option uses normals calculated at the nodes, where the
automatic option calculates the normal on element edge segments. The fully
integrated elements were only selected for the jacket and the lands and grooves.
The remainder of the barrel was set at single point integration elements. The
application of the fully integrated elements was chosen to compare the inherent
elimination of hourglass modes with the increased computation time.

Table 1: Listing of the corresponding model and contact algorithm.

Model LS-Dyna Contact Algorithm

A Surface to Surface
B Automatic Surface to Surface
C Surface to Surface with Fully Integrated Elements
D Surface to Surface with Fully Integrated Elements and
Viscous Hourglass Control
E Automatic Surface to Surface with Fully Integrated
Elements

3 Model results

3.1 Quasi-static model of push test

The quasi-static models were post processed to compare the hoop strain to the
experimental data and to evaluate the selection of model control parameters. To
verify that the simulation is quasi-static, the kinetic energy was verified to be
negligible compared to the internal energy. The hourglass energy was examined
to verify that the energy preventing the hourglass modes was negligible to the
internal energy. The ratio of the hourglass energy to the internal energy was
found to be a maximum of 2.0% after engraving started.
The predicted strain data was compared between the static and dynamic cases
at a location 1.27 cm from the chamber end of the barrel section. Figure 2 shows
the predicted hoop strain and the average experimental hoop strain plotted
against projectile displacement. For comparison, the strains could not be plotted
versus time since the projectiles in the model and experiment would not be at the
same location for a given time. The average experimental hoop strain from the
push test has a maximum of approximately 105 µε. The maximum predicted
hoop strain from the quasi-static model is 170 µε. The longitudinal shift
between the two data sets is due to the fact that the strain gauges were placed
further down the barrel in the experimental setup. This increased distance was
the result of experimental constraints within the test fixture. The maximum
dynamic strain was approximately 1.5 times greater than the quasi-static strain.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 197

300

250
Static Strain

200 Dynamic Strain

Microstrain

Average Experimental Data

150

100

50

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
Displacement [m]

Figure 2: Plot of the predicted strain for the static and dynamic simulations as
well as experimental strain from a push test.

12 0.35

10 0.3

Model A Internal Energy

Model C Internal Energy 0.25
Hourglass Energy

8
Internal Energy

Model A Hourglass Energy

Model C Hourglass Energy 0.2
6
0.15
4
0.1

2 0.05

0 0
0 0.0001 0.0002 0.0003 0.0004 0.0005
Time [sec]

Figure 3: Plot showing internal (open symbols) and hourglass (filled

symbols) energy.

3.2 Contact algorithm effects

The affect of the contact algorithm, element and material selections was
quantified by comparison of the internal energy and hourglass energy. Figure 3
shows the representative internal and hourglass energies for models A and C.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
198 Computational Ballistics III

The 0.25 ms delay in the appearance of measurable energy is due to the time
necessary to build sufficient pressure to begin engraving of the projectile. The
hourglass energies in the figure begin to increase after 0.25 ms and then decrease
as the time approaches 5 ms. The decrease is the result of the projectile exiting
the barrel section.

5.0%

4.5%
HourglassEnergy/Internal Energy

4.0%

3.5%

3.0%

2.5%

2.0%

1.5% A B
C D
1.0%
E
0.5%

0.0%
0.00025 0.0003 0.00035 0.0004 0.00045 0.0005
[Time]

Figure 4: Plot of the ratio of hourglass energy to internal energy for the six
models.

The ratio of the hourglass energy to the internal energy was plotted for the
five models. Figure 4 shows these ratios as a function of time. The data in the
figure between zero and 0.25 ms are not valid as the data is the quotient of two
small numbers. Figure 3 shows that the internal and hourglass energies are very
small between zero and 0.25 ms. Once engraving begins the ratios begin to
consolidate into groups. Models A and B, without higher integration order
elements, yielded a ratio with a maximum of 4%. Models C and E with some
fully integrated elements yielded a maximum ratio of less that 0.5%. There was
an average 40% increase in the run time between models with and without the
fully integrated elements. Model D with the viscous hourglass control
demonstrated an increasing ratio during the time that the projectile was engraved.
This value peaked at approximately 4.5%. All of the contact formulations with
stiffness hourglass controls improved with time.

3.3 Local effect of hourglass control algorithm

As shown previously, the hourglass control algorithm has a large effect on the
global hourglass energy. The engraving process has large local stresses, which
might drive local hourglassing. This local effect might not be seen in the global
energy comparisons. For this application, the state of stress in the land is of
particular interest, so the effect of the hourglass control algorithm on the land
stress is compared.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 199

The radial stress on a land in the first element forward of the forcing cone is
shown in Figure 5. All three cases show an initial peak when the projectile first
contacts the element, and then a sharp change in stress level. The Stiffness
Control case shows a tensile radial stress after the sharp change in stress level,
and then the stress sharply oscillates. The radial stress in the Viscous Control
case stays compressive and oscillates around an average value. The Fully
Integrated case has the smoothest time history. The radial stress stays
compressive, and the magnitude of the oscillations is the lowest of the three
cases.

400

200

-200
Radial Stress [MPa]

-400

-600

-800

-1000

-1200 Fully Integrated

Viscous
-1400 Stiffness

-1600
0.0002 0.00025 0.0003 0.00035 0.0004 0.00045
Time [sec]

Figure 5: The radial stress in a land as a function of time for the Fully
Integrated, Viscous Control, and Stiffness Control.

4 Discussion

4.1 Quasi-static model

The quasi-static model acts as a bridge between the dynamic model and the
experimental results. The comparison of the maximum hoop strain shows
reasonable agreement between the quasi-static model and the experimental data.
The strain was recorded over a land in the quasi-static FE model where the
maximum affect of the passing projectile would be expected. This difference
may be because the experiment uses a real barrel where the lands have a twist,
which might cause the strain measurement taken at fixed location on the barrel to
decrease faster as the projectile moves by. Also, the FE analysis only modeled a
section of the barrel 3.8 cm in length and the projectile is approximately 2.3 cm
long. The shorter barrel length is more compliant since the end faces are not
constrained, allowing the faces to move radially. Therefore, the model might
deform more as the projectile passes, than the experimental barrel.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
200 Computational Ballistics III

The agreement between maximum strains from the experiment and quasi-
static model gives confidence that assumptions made in creating the FE models
are reasonable. The dynamic strain is expected to be greater because of inertial
effects that are likely to results in projectile dilation. This is seen in the results
for the dynamic FE model, where the maximum is approximately 1.5 times
greater than the quasi-static model.

4.2 Global energy effects

The simulations to evaluate the effect of the contact algorithm, element

integration order and hourglass control yielded clear data. The affect of
including the automatic option onto surface to surface, models A and B and C and
E, had very little effect on the ratio of the hourglass to the internal energy. There
was some reduction in the ratio by use of this option but the reduction was on the
order of 0.1%. There was no change to the solution time of the model. As
expected the use of fully integrated elements for portions of the model greatly
reduced the overall hourglass energy and thus the ratio. This decrease was a
factor of eight as the ratio was reduced from 4% to 0.5%; however, this decrease
was offset by a 40% increase in run time. For more complex models, a
substantially longer runtime could be a factor in meeting design timeframe
constraints.

4.3 Local effects

One advantage of the finite element method is the ability to inspect local effects.
While the stiffness hourglass control is shown to be more effective than the
viscous algorithm from a global energy perspective, care must be taken when
validating the model. In general, viscous hourglass control is preferred for
simulations involving high rates of deformation. Stiffness hourglass control
should be used when the rates of deformation are low.
When investigating the radial stress in a land, the stiffness control gave the
least reliable results, even calculating a tensile radial stress when this is
physically impossible. This is most likely due to the element hourglassing,
making the stress prediction incorrect. In a large model such as this, a small
number of elements hourglassing will not affect a global energy comparison.
But if the primary region of interest is near those elements, the model cannot be
used for quantitative analysis.
The viscous control algorithm predicted a much more realistic time history
for the radial stress. This is not surprising since the strain rates at the contact
interface are expected to be rather high. The rest of the model, which includes
the majority of the barrel, the core, and penetrator, experiences low strain rates.
Therefore the viscous control has minimal effect on suppressing the hourglassing
in large sections of the model. A global energy comparison highlights this
problem and shows the stiffness control performing better at suppressing
hourglassing, even though the viscous control predicts a more realistic stress in
the land.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 201

The use of fully integrated elements in the land prevents the hourglassing in
the area of interest and predicts a realistic radial stress history. The
computational cost of fully integrated elements is high, which is why they were
only used locally. Another drawback is that fully integrated elements cannot
undergo large deformations like reduced integration elements. Used sparingly,
fully integrated can be used to insure trustworthy results in areas of importance.

5 Conclusions
In this study, evaluations were conducted to investigate different computational
techniques for the engraving of a 5.56 mm projectile fired from a rifled barrel
based on a M16A2. The results show the choice of the contact algorithm,
element order of integration, and hourglass control can drastically affect the
accuracy of the model as well as the run time. The baseline model was evaluated
quasi-statically and validated against experimental data. The choice in contact
algorithm can affect the hourglass energy. For this engraving model the use of a
stiffness form of hourglass control yielded the best results from a global energy
perspective, while the viscous formulation appeared to result in energy ratios that
were increasing. However, the local stress in the land shows that the viscous
control predicts more realistic stress history. The partial use of fully integrated
elements provided the expected reduction in the energy ratio however this
reduction, from 4% to 0.5%, was at the expense of a 40% increase in run time.
The results of this modeling effort have led to increased insight into the
behavior of the engraving process on small caliber projectiles and the numerical
techniques necessary to reduce numerical difficulties. The models should be
tailored to the problem being solved to provide the most accurate results. Future
work will include combinations of hourglass and viscous hourglass control in
different parts of the engraving model in an effort to save on the computational
cost of using fully integrated elements.

References
[1] Chen P.C.T., “Analysis of Engraving and Wear in a Projectile Rotating
Band,” U.S. Armament Research, Development and Engineering Center
Technical Report ARCCB-TR-99012, Watervliet, NY, USA, July 1999
[2] Chen P.C.T., “Modeling of Barrel/Projectile Interaction in a Rotating
Band,” U.S. Armament Research, Development and Engineering Center
Technical Report ARCCB-TR-01011, Watervliet, NY, USA, June 2001
[3] M855 Technical Drawing Package, Armament Research, Development
and Engineering Center Picatinny Arsenal, NJ, USA, 1980
[4] M4A1 Technical Drawing Package, Armment Research, Development
and Engineering Center, Rock Island, IL, 2001
[5] LS-DYNA Users Manual, Livermore Software Technology Corporation,
Livermore, CA, USA, 2003

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
202 Computational Ballistics III

[6] South J., Yiournas A., Minnicino M., “The Effect of Slug Material on the
Behavior of Small Caliber Ammunition,” U.S. Army Research Laboratory
Technical Report ARL-TR-3901, APG, MD, USA, 2006
[7] Siewert J., “Small Caliber Engraving Force Measurements,” Presented at
the NDIA Small Arms Systems Section Annual Symposium, Exhibition
and Firing Demonstration, Las Vegas, NV, USA, 10-13 May 2004,
www.ndia.org
[8] South J., Newill J., Kamdar D., Middleton J., Hanzl F., DeRosa G.,
“Bridging the Gap Between the Art and Science of Materials for Small
Caliber Ammunition,” Amptiac Quarterly, 8(4), pp. 57-63, 2004.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 203

Long range electromagnetic cannon

E. Schmidt & J. Tzeng
US Army Research Laboratory, Aberdeen Proving Ground, MD, USA

Abstract
Electromagnetic (EM) railguns offer the potential to fire projectiles at velocities
up to 6 km/s. This could conceivably permit ballistic trajectories to reach very
long ranges, perhaps 1000 km or more. To deliver a payload of interest and
survive the aerothermal environment associated with hypervelocity, sea level
launch, the projectile would need to be quite large, 500–1000 Kg. This in turn
implies a launcher of prodigious size and mass. This paper examines the nature
of an electromagnetic cannon required to achieve such capabilities.
Keywords: electromagnetic launch, rail guns, hypervelocity flight, long range
artillery.

1 Introduction
Historically, there has been a continuing interest in long range cannons. During
the First World War, Germany [1] employed the 210 mm, K12(E), Paris Gun
with a maximum range of 115 km. In the Second World War, they fielded the
800 mm Gustav cannon with a range of 47 km. The cannon was transported in
pieces and assembled on two parallel railway tracks by a crew of 1200 men.
Naval artillery also illustrates this trend as the bore size of guns on battleships
grew throughout the Twentieth Century culminating in the 460 mm guns on the
Japanese Yamato class with a range of 40 km. With the advances in aircraft and
missile capabilities, long range delivery of ordnance became routine and interest
in very large calibre guns has waned. However, the cost of missiles and risk of
manned delivery are of concern. Thus, worldwide efforts to increase the range
of conventional artillery, e.g., 155 mm, have lead to the introduction of
technologies such as base bleed, rocket assist, and gliding trajectories stretching
ranges out to nearly 100 km.
At the Electromagnetic Launch Symposium in May 2006 in Potsdam,
Germany, MG Nadeau (Commander, US Army RDECOM) challenged the

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070191
204 Computational Ballistics III

technical community to consider the viability of extremely long range artillery.

In particular, he was interested in a system that could send projectiles over
distances equivalent to one or more time zones. Since a time zone at 30o latitude
has a width of 1400 km or so, such an undertaking is daunting. While no
program existed to provide such a capability, Schmidt [2] conducted a
parametric study of possible electromagnetic (EM) railgun solutions to this
challenge; however, the required launcher was not examined in detail. The
present paper provides some of this information for one possible delivery
system.

2 Projectile
EM guns are not rifled; thus, a statically stable projectile is considered. Fin
stabilization is assumed; although it is recognized that aerodynamic heating may
preclude such a scheme. The projectile, Figure 1, has a body diameter of 0.20 m
and a length of 2.46 m. The projectile has a tungsten nose to improve stability
and provide some thermal protection. PRODAS [3] is used to compute inertial
and aerodynamic properties. With allowances for batteries and guidance, the
flight mass is 227 Kg and the transverse and axial moments of inertia are 101
and 1.65 Kg-m2, respectively. The payload is 57 Kg. The center of gravity is
6.67 calibers from the nose, while the center of pressure is 6.94 calibers from the
nose. With a static margin of 0.27 calibers, the projectile is barely stable;
however, this may be advantageous in providing control authority to the
guidance system.

6 Thick Clipped GNC Battery

DeltaFins

Payload Tungsten Nose

Figure 1: Schematic of notional projectile.

The muzzle velocity of interest is 4.0 km/s. At this velocity, the zero yaw
drag coefficient is CDo = 0.172. Because of the long ranges considered, the
trajectory of the projectile is computed using central body theory assuming an
exponential atmosphere throughout the flight. At the 4.0 km/s muzzle velocity
and 50o elevation, a range of 1176 km is estimated, Figure 2. This velocity is
well within the demonstrated capabilities of EM railguns using plasma
armatures. For these launch conditions, a maximum altitude of nearly 400 km is
obtained. Such a trajectory closely replicates that of an intermediate range
ballistic missile.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 205

400

300
Altitude (km)
200

100

0
0 500 1000 1500
Range (km)

Figure 2: Trajectory for Θ = 50o, vm = 4 km/s.

3 EM railgun launcher
3.1 General description

The EM railgun accelerates the projectile under the influence of the Lorentz
force, Figure 3. Current is passed down one rail, crosses over through the
projectile armature, and returns through the opposing rail. The magnetic field,
B, generated by the two rails acts on the current in the armature, J, to produce
the JxB force. This purely electrical operation results in significant advantages
compared to conventional cannon. An EM gun is capable of pushing projectiles
to very high velocity. Values 6 km/s and greater have been reported for light
masses. Conventional cannon are limited by the free expansion velocity of the
propellant gases, typically around 3.5 km/s. Practical considerations limit this to
velocities of 1.8 km/s or less. As the magnetic field acts on the armature, it also
acts on the opposing rails resulting in a Lorentz force that will cause the rails to
repel each other. Thus, confinement must be provided to hold the gun assembly
together. For the purpose of describing the size and mass of the launcher, this
confinement is given particular attention herein.

Current Armature

x x x x x x x x
Magnetic Field x x Force
x x x x x x x x Sabot

Rail Repulsion Force

Figure 3: Schematic of a typical railgun.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
206 Computational Ballistics III

It is assumed that the launch package consists of the flight body, a base push
armature, and a sabot to provide in-bore support to the projectile. Based on the
action limit, it has been postulated [4] that a parasitic mass ratio (armature and
sabot mass divided by total launch mass) of 25% is realizable. Thus, the total
launch mass is taken to be 303 Kg. At vm = 4.0 km/s, the resulting muzzle
energy is 2.42 GJ. To provide clearance for the fins, the bore diameter is 0.525
m. A circular bore geometry is assumed with a pair of GlidCop rails each having
an 80° included angle.
There are a number of constraints that must be considered in developing a
design for the railgun. Among them are the acceleration tolerance of the
payload, the projectile base pressure, the bore pressure relative to the yield
strength of the bore materials, the rise in bulk bore temperature during the shot,
and the current density to which the rails are exposed as the projectile passes.
Available technology for projectile guidance systems limits maximum
acceleration levels to less than 20 Kg.
For the projectile considered here, the maximum in-bore acceleration is
selected to be 17.5 Kg, which allows the guidance system to survive and gives a
maximum base pressure equivalent to that of conventional cargo carrying
projectiles. Based on prior data from electromagnetic guns, a reasonable value
for the acceleration ratio, average to peak, is 0.7. This leads to an estimated bore
length of 67 m.
Brody [5] gives a relation for the rail temperature rise based on the electrical
action to which the rails are exposed:
∆T = 2[ρe /(ρmCp)]mvm/L’A2 (1)
where ρe is electrical resistivity, ρm is material density, Cp is heat capacity, L’ is
the longitudinal inductance gradient, and A is rail cross-sectional area. From this
relation, the rail thickness may be computed once an allowable temperature rise
is selected and L’ is known. To permit multiple shots to be fired without serious
degradation of composite properties, ∆T = 20 K is selected. Since L’ depends
upon the rail thickness, an iterative process is required. As an initial estimate
[6], L’ = 0.5 µH/m. These values result in an initial estimate of rail thickness, t =
82.5 mm.
Ellis [7] has developed a model that integrates the electromagnetic fields
surrounding the rails to obtain a value of L’ and other properties of interest. The
corrected value for the inductance gradient is L’ = 0.57 µH/m. This results in a
thickness, t = 77.8 mm. Further iteration produces no perceptible change in L’.
An expression for the Lorentz force is F = ½ L’I2 = ma. Knowing the launch
mass, maximum acceleration, and L’, the peak current is computed to be Imax =
13.5 MA. Thus, the maximum current density at the armature location is 36.8
kA/mm, a value within the accepted limits for copper. McNab [4] calculates the
back emf for a railgun launcher as Vback = IL’vm. For the present launcher, this
predicts Vback = 30.8 kV yielding a maximum power required of 416 GW. These
values imply a pulsed power supply of impressive capabilities and equally
impressive size and mass. For space launch of a 1250 Kg launch package at vm =
7.5 km/s (Em = 35 GJ), McNab proposed using 100 pulsed alternators distributed

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 207

along the 1600 m length of the earth-to-space launcher. Treatment of the power
supply is beyond the scope of the present study; however, the details of the
launcher structure are considered in the next section.

3.2 Structural analysis

The rails are copper and are separated by G10 (glass composite) insulators. The
containment is assumed to be constructed with carbon composite hoop wound
overwrap. In general, the stress magnitude and distribution can be adjusted by
varying the thickness of rail and containment, the arc length of rails, and the
stiffness ratio of the gun bore components. A finite element model, Figure 4, is
built to calculate the stress profiles in the bore. The rail, insulator, and
containment are shown in the region of blue, orange, and gray, respectively. The
rail thickness is 77.8 mm (3 in) and the containment thickness is 127 mm (5 in).
Based on these thickness and material assumption, the structural analysis showed
the gun bore can be designed and built with reasonable engineering effort.

Figure 4: Finite element model of the gun bore.

Hoop, radial, and shear stress profiles are illustrated in Figures 5–7,
respectively. The stress profiles are plotted in English units (i.e. psi). The
maximum tensile hoop stress of 630 MPa (90 ksi) is located in the inner radius of
rails. Accordingly, the combined hoop and radial stress at the inner radius is
critical since it is assumed that a 140 MPa (20 ksi) magnetic pressure is applied
on the surface. The combined stress is in excess of the 517 MPa yield of
GlidCop; however, evolving alumina particle reinforced coppers may be
available to meet the strength requirement. In actuality, the magnetic force is
distributed through the entire rail thickness and can only be determined by more
detail electromagnetic analysis. The hoop stress at the inner radius of insulator is
525 MPa (75 ksi) in compression. It may be critical for commonly used material
such as G10, various materials or insulator designs can mitigate these stress
problems.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
208 Computational Ballistics III

Figure 5: Hoop stress profile in the gun bore.

Figure 6: Radial stress profile in gun bore.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 209

Figure 7: Shear stress profile in the gun bore.

The radial stress is manageable but shear stress is critical in this gun bore
design. It is about 175 MPa (25 ksi) in the region of the rail/insulator interfaces.
It is slightly high for composite materials (G10 and carbon composite).
However, it again can be mitigated by using different material or varying the
composite fiber architecture and thickness.
When a high strength, hoop wound carbon composite containment is used, the
hoop stresses in the containment are quite manageable. A shortcoming of the
gun tube may be a lack of the axial strength and stiffness; however, the addition
of external truss should provide the axial requirement. This preliminary
evaluation of the rail gun design shows that a functional gun bore design can be
achieved.

3.3 Launcher dimensions

To show the scale of the launcher, a comparison between the EM Long Range
Artillery Launcher and the World War II German Gun, Gustav, is presented in
Figure 8. The EM launcher is roughly twice as long as Gustav, but has a bore of
525 mm compared to 800 mm. The outer diameters of the two cannon are
roughly equal. Given its length, the EM launcher will require an external truss
work to provide stiffness against droop. The Paris Gun of World War II used
such a truss. From the structural analysis, the weight of the EM gun tube is
calculated to be 77,560 Kg given a length of 66.6 m. This is significantly less
massive that Gustav which weighed in at 364,000 Kg.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
210 Computational Ballistics III

The recoil system is an important consideration in determining overall size

and mass. Gustav launched a 4800 Kg projectile to a velocity of 820 m/s using
roughly 704 Kg of propellant. This results in a recoil impulse of 4.22 x 106 N-s.
Surprisingly, the longer EM launcher firing a 303 Kg projectile at 4.0 km/s has a
lower recoil impulse of 1.21 x 106 N-s. Thus, the size and mass of the recoil
system and overall cannon mount could be less for the EM launcher than for
Gustav. Of course, the power supply must be given consideration. For Gustav,
it consisted of the propellant needed for each shot. For the EM launcher, a rather
substantial array of pulsed alternators or capacitors would be needed. Depending
upon the number of rounds to be fired, the volume of the pulsed power supply
could be offset by the reduction in propellant. Of course, Gustav only fired a
total of 48 rounds in anger.

EM Launcher

Gustav

Figure 8: Relative scales of EM launcher and 800 mm Gustav cannon.

4 Conclusions
A study is presented to examine the launcher required to electromagnetically fire
a projectile to very long range. The scale of the launcher is prodigious, generally
larger than any known gun capable of transport or elevation. Although not
addressed in this paper, the pulsed power supply needed to drive the launcher is
also a challenge.

References
[1] Hogg, I., Twentieth-Century Artillery, Amber Books, London, 2000.
[2] Schmidt, E., Aerodynamic Aspects of Hypervelocity Projectiles,
Proceedings of the 23rd International Symposium on Ballistics, Tarragona,
Spain, April 2007.
[3] PRODAS 2000, Arrowtech Associates, Burlington, VT, (2000).
[4] McNab, I., Launch to Space with an Electromagnetic Railgun, IEEE
Transactions on Magnetics, 39, (2003).
[5] Brody, J., Theoretical Design and Modeling of an Infantry Railgun
Projectile, Thesis, Naval Postgraduate School, Monterey, CA, 2005.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 211

[6] Werst, M., Cook, K., Kitzmiller, J., Liu, H., Price, J., and Yun, H., Design
and Testing of a Rapid Fire, Lightweight, Ultra Stiff Railgun for a Cannon
Caliber Electromagnetic Launcher System, IEEE Transactions on
Magnetics, 31 (1995).
[7] Ellis, R., Private Communication, ONR, Rosslyn, VA, February 2007.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
This page intentionally left blank
 Computational Ballistics III 213

Analysis of weapon systems protecting military

camps against mortar fire
M. Graswald, I. Shaydurov & H. Rothe
Helmut-Schmidt-University, Hamburg, Germany

Abstract
The protection of military camps that are subject to attacks by rockets, artillery
projectiles, or mortar grenades (RAM) is currently in the spotlight of Western
nations involved in overseas missions. Due to its worldwide distribution and
good combat properties for guerilla warfare, mortars are likely to be applied and
therefore selected as RAM threat in this paper. For this case, camp safety shall
be ensured by an early warning system and an air defense weapon system using
35 mm Ahead ammunition. For both systems, a key to success is the accuracy
of the sensors used, namely the radar. Therefore after providing the mathematical
background, simulations are conducted in order to find the variations of the hit
point of mortar shells. Moreover, the consumption of Ahead ammunition to engage
and destroy typical 82 mm grenades is estimated. From these results, the suitability
of present radar sensors and air defense systems and the technical requirements of
future weapon systems can be derived.
Keywords: rockets artillery mortar (RAM), 35 mm Ahead ammunition, radar
sensor, exterior ballistics, terminal ballistics, probability calculation, error
propagation, circular error probability, ammunition consumption.

1 Introduction
Accomplished missions of the Western military e.g. in Iraq or Afghanistan in the
recent past have shown that the safety of military camps is not sufficient. This is
because suitable weapon systems protecting the military installations against this
new asymmetric threat do currently not exist. Furthermore, these RAM attacks by
guerillas or terrorists are frequently undertaken from urban area and therefore,
the rules of engagement do usually not allow counterstrikes. Primarily, mortar
grenades of 60 and 82 mm caliber and unguided 107 mm rockets are distributed

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070201
214 Computational Ballistics III

all over the world in great quantities and relatively easy to obtain and thus chosen
for attacks.
This study concentrates on 82 mm mortar grenades as a typical example for the
RAM threat. Two different approaches to increase the camp safety are investigated
and requirements will be given for:
• an early warning system that alerts soldiers depending on the predicted point
of impact of the mortar shell,
• an air defense system that intercepts the shells in a safe distance with Ahead
sub-projectiles forming a fragment cone, fig. 1.

2 Ballistic model
A mathematical model is adopted estimating the ballistic coefficient of the grenade
from four radar tracks of its flight path. As a result, it allows the iterative
calculation of the trajectory from an averaged radar location to either the firing
or the hit point.

2.1 Exterior ballistics of mortar grenades

Since the mortar grenades considered are arrow-stabilized and fired on short
distances up to approximately 8 km, the 2-DOF equations of motion are derived
from Newton’s second law treating the grenade as point mass and incorporating
only gravitation and air drag as external forces. This leads to the system of
differential equations given in a path-dependant form for a right-handed frame
of reference shown in fig. 2:

dvx dy
vx
= = −c2 (M a) v(x) (1a) y
= = p (x) (1c)
dx dx
dp g dt 1
p
= =− (1b) t
= = (1d)
dx vx (x)2 dx vx (x)


where p = tan θ = vy /vx , v = vx 1 + p2 , vx – x-component of velocity, vy
– y-component of velocity, g – acceleration due to gravity (g = 9.81 m/s2 ), t –
computed time of flight, θ – quadrant elevation, c2 (M a) – coefficient of air drag
depending on Mach’s number.

2.2 Trajectory determination by radar measurements

A well-known procedure to experimentally determine the projectile’s air drag is

adopted to estimate the ballistic coefficient [1]. This coefficient is an essential
input parameter solving eqns. (1) numerically in order to calculate the trajectory.
It is iteratively computed starting at point B determined by radar measurements to
the hit point, see fig. 3.
To estimate the ballistic coefficient it is assumed that the velocity depending
part of the air drag is known by an experimentally determined reference function

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 215

YI
.......
..
Y ...... .. vy
.. .
v
....
... .. ......... ........................... ....
... . ............ .. ..
..
.....
. .................. ..
c.m.
. ........................................................θ.................vx
......... . ... ... . ..
.
Point of F W... . .. .
............... ... ....................................................................................... ... ..
ignition
... .
...
... ... .. .... .........
... .................. .β... ...... .............. ...... ............... .. X .. .. ....
..
........... ..... hK ..... .... .. Mortar FG
.. ... .......................................................................
. . .
..... ...
.
......
grenade ..........................................................................................................................
AE = ACEP
Z ZI XI

Figure 1: Fragment cone. Figure 2: Inertial reference frame.

and the part depending on the atmosphere is considered as constant. Therefore,

the term that depends on the projectile itself needs to be determined and is called
ballistic coefficient c.
The drag coefficient c2 (c can simply be derived from it) is determined by the
drag function FW = ρ2 cW v2 A, which corresponds to the difference in kinetic
energy in A and B and refers to the distance xAB :

aW
c2 = 2
, (2)
vm

where aW = 12 (vx2A − vx2B )/ xAB is the deceleration and vm = 12 (vxA + vxB )

is the averaged horizontal component of the projectile’s velocity. The horizontal
distance
of the averaged radar locations A and B is determined geometrically by
xAB = (xB − xA )2 + (zB − zA )2 .
The velocities and coordinates in x and z at the locations A and B are calculated
by two tracks of a monopulse radar, respectively. Due to the specific form of
eqns. (1), only horizontal components of the velocity and the horizontal distance
between A and B are required. In another step, these radar tracks xi , yi , zi (i =
1 . . . 4) are linked to measuring quantities of radar azimuth α, elevation ε, and
time t.

3 Error propagation
The methods of error propagation are applied to the model in Section 2 in
order to find variations of the point of impact depending on the sensor accuracy,
i. e. variations in range, deflection, and the circular error probability (CEP). All
systematic measuring errors shall be eliminated by calibration or adjustment.
Thus, the measurements of the radar azimuth, elevation and time are subject to
random errors. It is assumed that these random errors are normally distributed
with µ = 0 as mean value and the standard deviations σα , σε , and σt are given by
the measuring devices. For a radar with rotating antenna its angular velocity ω is
also subject to errors with σω as standard deviation.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
216 Computational Ballistics III

...... v
B............... B
.
...
...
..... 4
..................3 ...
v..A
YR .....
.. ....
............... ...
....
... .
A
......... 2
.
...
.... ZR 1..... ..... ..... x...AB .. ..... ..... ..
. . ...
...
... ..... ...... .
.... . ... . . .......x...R1..y1
... . .
z1.. .. ..... .............. ..... .......
.... ..R ... AP . ......
. . R.1.. .. .... ..
.
.... ..... ...ε............. ..... .. . ...
.... .... ..... ....1..R1 .. .
.... ................ .......... ..
.
.......................α .
.....1..............................................................................................................................
x1
XR

Figure 3: Radar reference frame with trajectory of mortar shell.

3.1 Variations of the predicted point of impact

As described, the ballistic coefficient c is used to determine the trajectory and
point of impact. Hence, errors of the radar measurements are propagated by the
ballistic coefficient to the hit point and determine its desired variation. In order to
calculate the variation in range, the standard deviation of the ballistic coefficient
σc is computed with the random errors of azimuth, elevation, and time by applying
the laws of propagation of uncertainties:
 2  2  2
∂c ∂c ∂c
σc = σx2AB + σv2x + σv2x (3)
∂xAB ∂vxA A ∂vxB B

where σxAB , σvxA , and σvxB are computed from given σα and σt :



 4    4   2  2
 ∂xAB 2  ∂vxA 2  ∂vxA
σxAB =  σα , σvxA =
2  σα +
2 σt2 ,
j=1
∂αj j=1
∂αj k=1
∂tk

and

 4   4  2
 ∂vxB 2  ∂vxB
σvxB = σα2 + σt2 .
j=1
∂αj
k=3
∂t k

The deviation of the angular velocity can be easily calculated with σt

2π ω2
σω = σ t = σt . (4)
t2 2π

Subsequently, the approach of varying failure parameters can be used to

calculate e.g. N = 1000 normally distributed random quantities of the ballistic
coefficient ci (i = 1 . . . N ) with the mean value µc and standard deviation σc .
These quantities are taken to determine the hit points xi by solving the system of
differential equations (1) numerically.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 217

In order to compute the variation in range with these x-coordinates of the hit
point, the averaged quadratic deflection µ is determined [1]

N 2
i=1 λi
µ= (5)
N −1

where λi = xi − xm is the so-called apparent error and xm = N1 N i=1 xi is an
averaged point of impact.
Given µ and the u-quantile of the normal distribution, the variation in range of
the hit point lP at P % confidence level (C. L.) is calculated by [1]

lP = 2 u µ = 2 P ER . (6)

With the given errors of time and azimuth of the radar sensor, the
variation in deflection is directly calculated based on the locating geometry
i = Ri sin αi . Considering the worst case, the standard deviation σz =
z
sin2 αi σR
2 + R2 cos2 α σ 2 reaches its maximum value for α = π and can
i i α i 2
be reduced to σz = σR . The range error of a radar is determined with the speed of
light c0 by σR = c20 σt .
Similar to eqn. (6), the variation in deflection bP is determined by

bP = 2 u σz = 2 P ED . (7)

As final result, the CEP of the predicted hit point is determined with the
variations in range and deflection. Gilles [2] introduces a procedure to compute
the CEP on the desired C. L. with the standard deviations in x and z as well as
the belonging covariance cov(x, z). As input parameters, the x-coordinates are
already calculated, while the z-coordinates are N = 1000 normally distributed
random quantities generated with µz = 0 and σz = σR .

3.2 CEP simulations

For the following simulations, the commissioned radar systems MWRL-SWK

(a Russian air surveillance radar for airports; σα = 0.033 deg , σR = 10 m, in
RBS mode) and COBRA (a counter battery radar by Germany, France, and UK;
σα = σR /R = 0.191 deg , σR = 50 m, system specification) may serve as typical
examples (all radar and ammunition data throughout this paper are public domain).
An 82 mm-grenade with propellant charge O-832, No. 6 launched with M1937
is chosen as RAM threat with the data v0 = 211 m/s, cnom = 2.11 m2 /kg, and
m = 3.31 kg.
Furthermore, the following data is given:

ω = 90 deg/s , α1 = 145 deg (MWRL-SWK);

α1 = 145 deg , α2 = 132, 7 deg , α3 = 114, 6 deg , α4 = 91, 2 deg (COBRA);
xAP = −2040 m , zAP = 1000 m .

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
218 Computational Ballistics III

The general simulation results are shown in fig. 4, while the data for the selected
radar sensors are tabulated in tab. 1 and integrated in fig. 4 for confidence levels of
50%, 90%, and 99%. Designing an early warning system for military camps, the
estimation of CEP = f (σα = σε , σt ) allows the definition of the edge length of
plane squares in order to enable specific areal warnings to soldiers. Since a clear
threat direction is usually unknown, the camp shall be divided into squares with its
edge length given by the CEP

lP =
2 CEP  . (8)

For a very precise radar system like MWRL-SWK, its edge length is already
46 m at 50% C. L. and 98 m for 90% C. L. These distances cannot be covered in
such short advance warning times of roughly 10 s. It also becomes clear that the
COBRA radar is insufficient for this task.

Table 1: Results of simulations for selected radar systems.

ccal. in m2 /kg σc in m2 /kg CEP in m

C. L. in % - 68.3 50.0 90.0 99.0
MWRL-SWK 2.187 0.070 22.6 48.7 75.1
COBRA 2.187 0.402 134.5 305.2 471.7

4 Ammunition consumption calculations

Intercepting a RAM target successfully requires fulfilling two conditions: at first,
the kinetic energy of the sub-projectiles is sufficient to destroy the mortar grenade
and secondly, at least one pellet hits the shell. For this purpose, 35 mm x 228 Ahead
ammunition is examined consisting of Nf = 152 sub-projectiles and featuring
fuze programming. Since the attempt of deflecting the grenade from its trajectory
with the momentum of a sub-projectile is practically not promising due to the mass
difference of factor 102 , this study concentrates on the physical destruction of the
mortar grenade.

4.1 Minimum energy for destruction of mortar grenades

The minimum energy to destruct a grenade is compounded by the kinetic energy

to penetrate the shell and the potential energy to activate the explosive. At first, the
well-known formula of de Marre [3] is resolved for the desired minimum impact
velocity vmin of an sub-projectile in order to penetrate a armored steel plate with
thickness D (in dm)
B d0.75
f D0.7
vmin (γ) = 0.5 (9)
mf cosn γ
WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 219

Table 2: Minimum energy destructing typical Russian mortar grenades.

caliber dM in mm 82 107 120 160 240

Dmax in mm 10.2 16.5 19.1 21.8 39.3
Emin in J γ = 0 deg 1160 2257 2756 3335 7563
Emin in J γ = 30 deg 1778 3467 4236 5126 11636

82 mm, O−832, No. 6

500 12
50.0% C. L.
γ = 0°
450 90.0% C. L.
γ = 30°
99.0% C. L.
10 82 mm, O−832
400 107 mm, OF−841A
120 mm, OF−843A
350 160 mm, F−853U
8 240 mm, F−864
Eges,min in kJ

300
CEP in m

250
6
200

150 4

100
2
50

0
0 0.05 0.1 0.15 0.2 0
10 15 20 25 30 35 40
σ =σ in ° D in mm
α ε

Figure 4: CEP = f (σα ), σt = 200 ns Figure 5: Emin = f (D, γ).

(∗ – MWRL, ♦ – COBRA).

where df – caliber of sub-projectile (df = 5.8 10−2 dm, estimated), mf – sub-

projectile’s mass (mf = 3.3 10−3 kg), B – empirical factor (B = 2000, averaged
value), γ – angle of impact (measured from plate’s normal) and n = 1.5 [3]. For
this case, the maximum thickness of an 82 mm shell is set for D.
The energy to ignite the grenade’s explosive is calculated with the impact
sensitivity for TNT mh = 1.5 kg m and the acceleration due to gravity. Finally,
the overall minimum energy is

mf 2
Emin (γ) = v (γ) + mh g (10)
2 min

and is shown for several Russian mortar grenades and impact angles of γ =
0 deg , 30 deg in fig. 5 as well as tab. 2.
Ideally, the internal energy Ei provided through an rigid body impact of the
sub-projectile with the grenade needs to be greater than the minimum energy
Emin = f (dM , γ) given in tab. 2. Due to the small amount of HE, the sub-
projectiles are not accelerated after ignition. Therefore, the velocity of a sub-
projectile is calculated with the velocity of the Ahead projectile at time of ignition
and decreased on its flight path because of air drag.
WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
220 Computational Ballistics III

4.2 Kill probability

In order to obtain a high kinetic energy for destruction, a frontal attack by an Ahead
sub-projectile is considered as ideal case due to the high velocity difference of the
sub-projectile and the grenade. Hence, the target area AT is calculated with the
caliber radius of the shell. However, the chance to hit the fuze of the grenade in
order to ignite it is not considered in this paper.
Calculating the hit probability of an sub-projectile at first, it is assumed
• the cannon is calibrated (µy = µz = 0),
• the ballistic deviation of cannon and ammunition is σy = σz = σb ,
• the hit pattern in Y and Z are regarded as independent (ρ = 0).
Thus, the probability pHF to hit a circle area in a Y -Z-plane with radius RM
with a single sub-projectile is based on the normal distribution [3]

R2
M
− 2(u σ
pHF = 1 − e )2b . (11)

Since the target has also a dimension in the X-Y -plane, the hit probability
pHt in this plane is determined with the Gaussian error function and the relative
target length xrel = lM /σx applying the distribution function Φ of the normal
distribution
√ 1
pHt = erf (x) = 2Φ( 2x) − (12)
2
where x is the input parameter determined as follows [1]:

u lM
x= √
2 σx


where lM is the grenade’s length and the deviation σx = σxR 2 + σ 2 consists
xG
of the converted time deviation of the radar σxR = 2 σtR and the converted time
c0

resolution of the fuze σxG = vG σtG (vG – velocity of Ahead projectile). Finally,
the desired hit probability of a single sub-projectile with the above probabilities
being independently is
pH = pHF pHt . (13)

A single effective sub-projectile shall be able to destruct a grenade. Thus, the

destruction probability is calculated with the internal energy Ei [3]
E
− Ei
pk|H = 1 − e 0 , (14)

where E0 is given by the minimum energy Emin (γ) at γ = 0 deg.

In conclusion, the kill probability of a single sub-projectile is

pK = pH pK|H . (15)

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 221

50% C. L. 99% C. L.
160 600
σ = 10 ns σ = 10 ns
t t
140 σ = 50 ns σ = 50 ns
t t
σt = 75 ns 500 σt = 75 ns
120 σt = 100 ns σt = 100 ns

400
100
s

s
80 300
N

N
60
200

40

100
20

0 0
0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
σα=σε in ° σα=σε in °

Figure 6: NS = f (σα , σt ) for 50% and 99% C. L. (∗ – MWRL-SWK).

4.3 Number of rounds

Calculating the kill probability of Nw effective sub-projectiles hitting the target

area AT , it is assumed that the base area of the fragment cone AE equals the area
of the radar CEP ACEP (see fig. 1):

AE (P ) = ACEP (P ) = π (RR u σα )2 , (16)

where RR is the distance radar – located grenade and σα is the radar’s azimuth
deviation assuming σα = σε . ACEP is valid for a given probability P (via quantile
u) that the grenade is located within this area.
The required probability that at least one out of Nw sub-projectiles hits and
destructs the target is given by [3]

pK,Nw (P ) = 1 − (1 − pK )Nw (P ) , (17)

N
where Nw (P ) = ρf (P ) AT = AE (P f
A . This leads to the wanted ammunition
) T
consumption, i. e. the number of rounds NS to destroy the target with an overall-
kill probability P
 
1
NS (P ) = . (18)
pK,Nw (P )

4.4 Simulation results

This subsection contains the simulation parameters and results to battle an 82 mm

grenade with Ahead ammunition strongly depending on the radar errors. The battle
distance is supposed to be 1000 m. Estimating the velocity of an Ahead sub-
projectile at this point and considering the slowest trajectory velocity of an 82 mm
shell fired on maximum distance, the internal energy Ei is 1580 J. Hence, the
destruction probability pk|H is 74.4%.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
222 Computational Ballistics III

Table 3: Simulation results for MWRL-SWK radar.

C. L. in % ACEP in m2 hK in m pK in % ρf in 1/m2 Nw NS
50 0.47 2.1 1.3 321 1.7 46
90 2.82 5.2 3.2 54 0.3 110
99 6.91 8.1 5.0 22 0.1 171

In addition to the parameters already given, the quantities are as follows:

lM = 329 mm , σb = 0.3 mil = 0.0169 deg , σtG = 2 ms , RR = 1000 m.

For several overall-kill probabilities P , the results are shown in fig. 6 and in
tab. 3. They reveal that even with a precise radar like MWRL-SWK an ammunition
consumption NS < 100 can be obtained only for probabilities P  90%.
Furthermore, the distance of the point of ignition to the impact point needs to
be short enough such that the sub-projectile’s kinetic energy exceeds the minimum
energy of Emin = 1160 J. This condition is not fulfilled for P = 99% in tab. 3.

5 Conclusions
The protection of military camps underlying the threat of attacks by mortar
grenades is studied by implementing an early warning system and an air defense
system to destroy an approaching shell. The basic mathematical principles are
given in order to evaluate or design sensors and ammunition with preformed or
natural fragments as well. The study yielded the following most relevant results:
• Determining the CEP = f (σα , σt ) in order to define the edge length
of plane warning squares in the military camp, the sensor precision is the
bottleneck. Even a very precise radar system like MWRL-SWK (σα = 2
,
σR = 10 m) provides edge lengths of almost 50 m at 50% and 100 m at 99%
C. L. This is impractical considering warning times in a scale of 10 s.
• Therefore, a frontal attack of 82 mm mortar grenades with a cloud
of 152 sub-projectiles originated from 35 mm Ahead ammunition is
investigated. However, the ammunition consumption at 99% C. L. amounts
to 171 shots. The low initial velocity of the sub-projectiles associated with
a small mass, a strongly limited number of sub-projectiles, and a very small
cone angle β of approximately 10 deg restricts the kinetic energy and the
hit probability severely. For these reasons, the Ahead ammunition is hardly
suited for this application.
• The artillery radar COBRA is completely inappropriate as sensor for both
supposed systems.
Considering decisive follow-on tasks in this field, the application of large-
caliber HE projectiles instead of Ahead ammunition shall be investigated. In

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 223

addition, the complex combat procedure with detection, tracking, fire control, and
interception should be simulated for different threat scenarios.

References

[1] Athen, H., Ballistik. Quelle & Meyer: Heidelberg, Germany, 1958.
[2] Gilles, J.T., Computation of the Circular Error Probability Integral. Systems
Reseach Center, University of Maryland, 1991.
[3] Germershausen, R., Waffentechnisches Taschenbuch. Rheinmetall GmbH:
Duesseldorf, Germany, 1977.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
This page intentionally left blank
 Computational Ballistics III 225

Empirical relationship for muzzle exit pressure

in a 155 mm gun tube
D. Carlucci & J. Vega
Analysis and Evaluation Technology Division,
U. S Army Armament Research, Development & Engineering Center,
Picatinny Arsenal, Dover, NJ 07806-5000, USA

Abstract
The development of guided artillery projectiles has uncovered a need for the
understanding of muzzle exit flows due to the common practice of designing
deployable fins that open upon muzzle exit. Knowledge of this base pressure
drop is also important for the electronics designer because so-called “set-
forward” (the rapid “un-springing” of the projectile as it leaves the muzzle of the
gun) has been identified as the cause of many component failures. An empirical
relationship is developed for the base pressure drop in a 155 mm gun tube. This
tube utilized a standard double-baffle muzzle brake. Several Instrumented
Ballistic Test Projectiles (IBTP) were fired at various charge zones and the base
pressure drops were compared and curve fits developed. A basic exponential
decay curve with averaged coefficients was developed from the data. This
empirical model can be utilized by researchers performing Computational Fluid
Dynamics calculations as a check on results generated for the double baffle
muzzle brake configuration.
Keywords: gun launch, gun hardening, muzzle exit, shot exit, electronics,
MEMS, projectile dynamics, in-bore dynamics.

1 Muzzle exit behavior of projectiles

With today’s focus on guided, smart projectiles, more and more designs are
being developed that employ fins or other stabilization devices that are activated
upon muzzle exit. This muzzle exit activation requires detailed knowledge of the
pressure field to insure that these devices open uniformly and reliably.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
doi:10.2495/CBAL070211
226 Computational Ballistics III

The set-forward effects or “un-springing” of the projectile as it leaves the muzzle

of the weapon is a major reliability concern for gun launch electronics [1, 2]
therefore an understanding of the time-dependent pressure field is critical.
In most instances the flow field at the muzzle of the weapon is a time-
dependent jet flow complicated by one or more of the following: the presence of
a muzzle brake; the two-phase nature of the propellant gas; the continued
reaction of the burning propellant with its gaseous products and the atmosphere;
compressibility effects and the presence of the moving projectile. The
complicated nature of this problem usually forces the designer to move directly
into a Computational Fluid Dynamics (CFD) model with some degree of
simplification. One of the challenges that occur frequently in CFD is obtaining
model confidence without hard data to validate against. The purpose of this
paper is to provide useful data for the CFD researcher to validate against. Due to
funding limitations data was only taken in a standard double baffle muzzle brake
mounted on a U.S. M198 155 mm howitzer at different charge zones. It is hoped
that this data will be useful in model validation.

2 Measurement technique
Data were obtained in 9 different firings at three different charge zones (i.e. three
different propellant amounts) using an Instrumented Ballistic Test Projectile
(IBTP). Table 1 specifies the pressures obtained in the firings while Figure 1
depicts the IBTP. The pressures were gathered from a pressure transducer
mounted in the base of the projectile.
The data obtained in these tests are shown as Figure 2. We note that for space
reasons these curves have been compressed to fit in a small area therefore the
scales are not particularly legible. They are intended to provide the reader with a
qualitative feel for the data. The interested reader is invited to contact the
authors for full sized plots. In each case the pressure drop at muzzle exit is
compared to filtered data resulting in the smooth curves. These curves were then
used to determine the exponent β tabulated in Table 1.

3 Pressure drop model

The curve fit that best described the pressure drop across all firing conditions
was an exponential decay of the form:

p s (t ) = pe e − βt (1)

Here ps(t) is the base pressure acting on the projectile, pe is the pressure acting
on the base at the instant the aft end of the projectile clears the bore of the gun, t
is the time in seconds from bore exit and β is an exponential decay factor
established from the smoothed data. Table 1 shows the values of pe and β
obtained from the test as well as peak values of pressure (pmax) that occurred in
the bore of the weapon measured with the pressure gage in the base of the IBTP.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 227

A mean value of β was determined to be 5,577 with a standard deviation of

1,117. While this may seem to be fairly uncorrelated this model has actually
helped the U.S. Army design gun launch electronics of very high reliability [4].

4 Uses of the model

There are, in general two uses for the model: CFD validation and structural
modelling of projectiles. In the former case the relationship is used directly to
compare base pressure measurements to the model results as the projectile leaves
the muzzle of the weapon. In the latter case the pressure decay curve model is
actually inserted at the proper time phasing in a finite element analysis to
examine the response of the structure to the pressure drop. Alternatively the
model can be modified as an acceleration time curve and used directly as a
forcing function on a projectile.

Figure 1: Instrumented Ballistic Test Projectile (IBTP).

Table 1: Firing data from IBTP tests.

pmax pe β
(PSI) (PSI )

TM2 53096 9848 6500

TM3 54648 9989 6500
TM4 39707 8551 5000
TM5 42440 8751 5000
TM6 51059 9364 5500
TM7 46844 9354 7000
TM8 47729 9771 7000
TM9 29556 8457 6000
TM22 44000 6708 3500

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
228 Computational Ballistics III

5 Conclusions
A pressure decay model has been established for a double baffle muzzle brake
configuration. This model has been shown to yield highly reliable projectile
structures when used as either a CFD validation tool and as a structural model
input. Future work will include pressure decay comparisons on tubes of various
calibres using differing muzzle brake configurations.

W35: W27; overplot(9848*exp(-6000*x), lred) W36: W28; overplot(14069*exp(-6000*x), lred) W37: W29; overplot(8551*exp(-5000*x), lred)

12000 16000 10000

14000
10000 8000

12000

TM2 TM3
8000

TM4
10000 6000

6000 8000
4000

6000
4000
2000
4000

2000
2000
0

0 0

-2000
-2000 0 0.0005 0.001 0.0015 0.002 0.0025 0.003 0.0035
-2000 0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014 0.0016 0.0018 0.002 0.0022 0.0024 0.0026
0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014 0.0016 0.0018 0.002 0.0022 0.0024 0.0026

W39: W31; overplot(9364*exp(-5500*x), lred) W40: W32; overplot(9354*exp(-6500*x), lred)

W38: W30; overplot(8751*exp(-5000*x), lred)
10000 12000
10000

10000
8000 8000

8000

TM5 TM6 TM7

6000 6000

6000

4000
4000
4000

2000
2000
2000

0
0 0

-2000
0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014 0.0016 0.0018 0.002 0.0022 0.0024 0.0026 0.0028 -2000
-2000 0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014 0.0016 0.0018 0.002 0.0022 0.0024 0.0026
0 0.0005 0.001 0.0015 0.002 0.0025 0.003 0.0035 0.004

W41: W33; overplot(9771*exp(-6500*x), lred) W42: W34; overplot(8457*exp(-6000*x), lred) W45: W3; overplot(6700*exp(-3500*x), lred)

12000 10000 8000

7000
10000 8000

TM9
6000

TM8
8000

TM22
6000 5000

6000 4000
4000
3000
4000

2000 2000
2000
1000
0
0
0

-2000 -2000 -1000

0 0.0005 0.001 0.0015 0.002 0.0025 0.003 0 0.0005 0.001 0.0015 0.002 0.0025 0.003 0.0035 0.004 0.0045 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007

Figure 2: Qualitative plots of pressure drop at muzzle exit.

References
[1] D. Carlucci, J. Cordes, S. Morris, R. Gast, Muzzle Exit (Set Forward)
Effects on Projectile Dynamics, Technical Report ARAET-TR-06003,
U.S. Army Armament Research Development and Engineering Center,
Dover, NJ, April 2006.
[2] D. Carlucci, J. A. Cordes, J. Hahn, A. Frydman, Electronics and the Gun
Environment, Invited Paper, U.S. Army Workshop on Advanced Active
Thin Film Materials for the Next Generation of Meso-Micro Scale Army
Applications, May 2005.
[3] M. Hollis, B. Flyash, A. Bahia, J. Potucek, D. Carlucci, Empirical
Measurements of Cannon Launch Pressures on a Finned 155-mm

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
 Computational Ballistics III 229

Artillery Projectile, 21st International Symposium on Ballistics, 19-23

April 2004.
[4] Cordes, J. A., Carlucci, D. E., Kalinowski, J., Reinhardt, L, “Design and
Development of Reliable Gun-Fired Structures,” DTIC Technical Report
ADA455406, U.S. Army Armament Research Development and
Engineering Center, Dover, NJ, June, 2006.

WIT Transactions on Modelling and Simulation, Vol 45, © 2007 WIT Press
www.witpress.com, ISSN 1743-355X (on-line)
This page intentionally left blank
 Computational Ballistics III 231

Author Index
Alves A. S................................ 171 Minnicino M............................ 193
Motta A. A....................... 123, 171
Baker E. L........................ 163, 181
Bonora N.................................. 131 Netto T. A................................ 123

Carlucci D................................ 225 Orfanoudakis Ν. G................... 141

Chen M. ..................................... 77
Chin C...................................... 163 Peles S. ...................................... 13
Costello M. ................................ 31 Peratta A. ................................... 67
Czarnetzki W. ............................ 21 Peratta C. ................................... 67
Pham J. .................................... 163
Degiovanni M. ......................... 131 Pincay J. M. ............................. 181
Di Sciuva M............................. 131 Powers B.................................. 193

Ebecken N. F. F. .............. 123, 171 Rangaswamy P. ....................... 109

Roetzel W. ................................. 21
Fatsis A. ................................... 141 Rothe H............................ 153, 213
Founti M. ................................. 141 Ruggiero A. ............................. 131
Frola C. .................................... 131
Fuchs B. E................................ 163 Sahu J. ................................... 3, 31
Scheffler D. R. ........................... 99
Gherlone M.............................. 131 Schmidt E. ............................... 203
Gold V. M................................ 181 Shaydurov I. ............................ 213
Graswald M. ............................ 213 Silva E. A. P. ........................... 123
Greatrix D. R. ............................ 57 South J. .................................... 193

Hackett R. M............................ 109 Touati D..................................... 13

Hountalas D. T......................... 141 Tzeng J..................................... 203

Kokes J. ..................................... 31 Vega J. ..................................... 225

Krallis K. ................................. 141 Vlachakis N. ............................ 141
Kuhrt A. ................................... 153
Walters W. P.............................. 99
Maier T. ..................................... 21 Wells J. M.................................. 89
Mattone M. .............................. 131
This page intentionally left blank
 ...for scientists by scientists

Fracture and Damage Computational

of Composites Ballistics II
Edited by: M. GUAGLIANO, Politecnico Edited by: V. SÁNCHEZ-GÁLVEZ, Univ
di Milano, Italy and M.H. ALIABADI, Politecnica de Madrid, Spain, C.A.
Imperial College, London, UK BREBBIA, Wessex Institute of Technology,
This book contains up-to-date contributions UK, A.A. MOTTA, Brazilian Navy
on areas such as: Compressive Strength of Research Institute, Brazil,
Laminated Composites - On Application of C.E. ANDERSON, Southwest Research
the Continuum Fracture Theory; Institute, USA
Macroscopic Crack Propagation Due to This book features most of the papers
Stress-Corrosion Cracking in Unidirectional presented at the second International
GRP Composites - Micromechanical Theory Conference on Computational Ballistics 2005.
and Its Application; Fatigue Damage of The contents stress the importance and
Particle Reinforced Metal Matrix Composites; possibilities of numerical simulation on
Modeling and Prediction of the Mechanical internal, external and terminal ballistics, to
Properties of Woven Laminates by the Finite describe, analyse, predict and subsequently
Element Method; Boundary Element reduce the experimental requirements in
Analysis of Fracture Failure in Anisotropic ballistics.
Composite Laminates; and Analysis of Ballistics, as a science, relates to a great variety
Piezoelectric Composite Laminates with of phenomena that occurs from the moment
Edge Delamination. an object or projectile is fired until its effects
are observed in a target. Ballistic studies
Series: Advances in Fracture Mechanics, include applications as varied as the study of
Vol 8 the structural and control behaviour of
ISBN: 1-85312-669-1 2006 304pp rockets and satellites; strikes on aircraft,
£97.00/US$175.00/€145.50 terrorist attacks and automobile
Find us at crashworthiness modelling, to name but a
http://www.witpress.com few. Many of the basic problems of ballistics
are similar to those in other fields of
Save 10% when you order from our encrypted applications, such as combustion, heat
ordering service on the web using your credit card. conduction, in-flight structural behaviour,
WIT eLibrary trajectory related issues, contact, impact,
penetration, structural response to shock
Home of the Transactions of the Wessex waves and many others.
Institute, the WIT electronic-library This book, which is an important addition to
provides the international scientific the literature, contains a wide variety of topics
community with immediate and permanent including: Terminal ballistics; Fluid-structure
access to individual papers presented at interaction; Perforation and penetration
WIT conferences. Visitors to the WIT mechanics; High rate loads, shock and impact;
eLibrary can freely browse and search Interior ballistics; Fluid flow and
abstracts of all papers in the collection Aerodynamics and Systems and Technology.
before progressing to download their full
text. WIT Transactions on Modelling and
Visit the WIT eLibrary at Simulation, Vol 40
http://library.witpress.com ISBN: 1-84564-015-2 2005 416pp
£146.00/US$234.00/€219.00
 ...for scientists by scientists

Advances in Fatigue, field.

Contents: Dual Boundary Element Method
Fracture and Damage for 3D Dynamic Crack Problems;
Modelling of Fracture Processes During Fast
Assessment of Crack Growth; Transient Dynamic Crack

Materials Analysis in Anisotropic Solids; Evolution

of Dynamic Fractures in PMMA –
Edited by: A. VARVANI-FARAHANI, Experimental and Numerical Investigations;
Ryerson University, Canada The Mechanism of Dynamic Crack
Propagation.
The damage analysis of engineering
materials and components is a prime concern Series: Advances in Fracture Mechanics,
in the durability assessment and integrity of Vol 9
engineering components in service. ISBN: 1-85312-948-8 2005 224pp
Engineering materials are routinely £83.00/US$133.00/€124.50
subjected to failure in a wide variety of
applications in aeronautical, electronic,
automotive, nuclear plant, petroleum and
transportation industries.
How Gears Break
This book presents the fundamental elements B. ABERŠEK and J. FLAŠKER,
and theories in fracture and damage analysis, University of Maribor, Slovenia
plus the recent research and advances in the This book provides an in-depth study of
development of the analytical and practical the failure mechanisms of gears and
approaches required to assess the materials describes how simulation models can be
damage and the durability of structures. developed to complement or replace
The wide-ranging discussion on damage experimental tests.
theories and practical issues in damage The authors first introduce the physical
analysis will be of benefit to researchers, processes of failure and explain how to
post-graduate students and engineers in analyse these using fracture mechanics.
practical sectors of durability and failure of Much of the following text is devoted to
materials and structures. mathematical models and formulations that
Series: Advances in Damage Mechanics, can be used to predict how cracks behave in
Vol 6 gears, together with special techniques
developed by the authors. The concluding
ISBN: 1-85312-836-8 2005 520pp
section describes an expert system devised
£170.00/US$290.00/€255.00
by them for the design of gear assemblies.
References are included at the end of each
chapter.
Crack Dynamics Series: Advances in Damage Mechanics,
Edited by: A. IVANKOVIC, University Vol 7
College Dublin, Ireland and ISBN: 1-85312-739-6 2004 240pp
M.H. ALIABADI, Queen Mary College, £81.00/US$130.00/€121.50
University of London, UK
Covering various aspects of dynamic All prices correct at time of going to press but
subject to change.
fractures this book contains state-of-the-art
WIT Press books are available through your
contributions from leading scientists in the bookseller or direct from the publisher.
 ...for scientists by scientists

Macro and the problems is based on relevant elementary

solutions in which the complex variable
Microscopic function method is used.
The multiple crack problem is considered as
Approach to Fracture a superposition of many single crack
problems while many more complicated
Edited by: S.-I. NISHIDA, Saga cases are covered.
University, Japan Miscellaneous problems, including the
Written by internationally recognized multiple rigid line problem and the multiple
specialists, this book covers a wide range of circular hole problem are studied. Solutions
topics associated with application and testing for three-dimensional multiple crack
methods for the assessment of risk and problems are also investigated by using the
performance behavior. A key reference for Fredholm integral equation, the
postgraduate students, lecturers and hypersingular integral equation and the
researchers working in this area, the text is variational principle. Many programs for
designed to enhance the knowledge and multiple crack problems using FORTRAN
appreciation of engineering integrity and its are featured. A CD-ROM containing
relevance to economic and strategic issues. solutions is also included.
Partial Contents: Fatigue Properties of
Austenitic Stainless Steels; Highly Accurate Series: Advances in Damage Mechanics,
Analysis of Stress Intensity Factors by the Vol 4
Crack Tip Stress Method in FEM; ISBN: 1-85312-903-8 2003
Characteristics of Fatigue Behavior of 356pp+CD-ROM
Surface Modified Materials; Macro- and £129.00/US$206.00/€193.50
Microscopic Approach to Fracture of High
Strength Steel Notched Bars; Nondestructive
Monitoring of Fracture Behavior of Materials Fatigue Damage of
using Acoustic Emission.
Series: Advances in Damage Mechanics,
Materials
Vol 3 Experiment and Analysis
ISBN: 1-85312-902-X 2004 352pp
Edited by: A. VARVANI-FARAHANI,
£105.00/US$168.00/€157.50
Ryerson University, Canada and
C.A. BREBBIA, Wessex Institute of
Multiple Crack Technology, UK

Problems in Elasticity
Focusing on recent advances in the
development of methods to predict fatigue
performance, this book contains papers from
Y.Z. CHEN, Jiangsu University, China, N.
the First International Conference on Fatigue
HASEBE, Nagoya Institute of Technology,
Damage of Materials. The coverage is broad
Japan and K.Y. LEE, Yonsei University,
encompassing a range of materials and
South Korea
structures.
“...a useful addition to the literature.”
MATHEMATICAL REVIEWS WIT Transactions on Engineering
Sciences, Vol 40
In this volume various integral equations
for multiple crack problems in plane ISBN: 1-85312-979-8 2003 528pp
elasticity are investigated. Formulation of £158.00/US$249.00/€237.00
 ...for scientists by scientists

Boundary Element Thermomechanical

Analysis of Cracks in Fatigue and Fracture
Shear Deformable Edited by: M.H. ALIABADI, Queen Mary

Plates and Shells

College, University of London, UK
Thermal and mechanical fatigue leads to
T. DIRGANTARA, Institut Teknologi initiation and growth of cracks in many
Bandung, Indonesia structures. This book brings together up-
to-date contributions from leading
In this book the author gives a new set of
specialists.
boundary element formulations for the
Partial Contents: Thermomechanical
solution of bending problems in plates and
Crack Growth using Boundary Elements;
shells. These are based on modified Reissner
Prediction of Crack Initiation and Growth
plate and two-dimensional plane stress
under Thermal Fatigue; The Saturation of
governing equations.
Thermomechanical Fatigue Damage in
Series: Topics in Engineering, Vol 43 Brittle Materials; Plasticity Induced Heating
in the Fracture and Cutting of Metals; Effects
ISBN: 1-85312-950-X 2002 272pp
of Residual Stress on Thermal Fatigue
£109.00/US$169.00/€163.50
Phenomena of Aluminium Die Casting Dies.

Nonlinear Fracture Series: Advances in Fracture Mechanics,

Vol 7
and Damage ISBN: 1-85312-549-0 2002 344pp

Mechanics
£115.00/US$173.00/€172.50

Edited by: M.H. ALIABADI, Queen Mary

College, University of London, UK
Fracture of Rock
Edited by: M.H. ALIABADI, Queen Mary
Incorporating recent research, this book College, University of London, UK
presents theoretical and computational
methods for problems in this field. This volume covers a wide range of subjects,
Partial Contents: Analysis of Crack Growth including hydraulic fracturing, blasting and
in Residual Stress Fields; Constitutive- fragmentation, transport problems and creep.
Microdamage Description of Ductile Series: Advances in Fracture Mechanics,
Dynamic Fracture; Vol 5
A Complete Gurson Model.
ISBN: 1-85312-542-3 1999 440pp
Series: Advances in Fracture Mechanics, £155.00/US$248.00/€232.50
Vol 4
ISBN: 1-85312-508-3 2001 264pp
£99.00/US$153.00/€148.50 WITPress
Ashurst Lodge, Ashurst, Southampton,
Find us at SO40 7AA, UK.
http://www.witpress.com Tel: 44 (0) 238 029 3223
Save 10% when you order from our encrypted Fax: 44 (0) 238 029 2853
ordering service on the web using your credit card. E-Mail: [email protected]