International Journal of Impact Engineering 108 (2017) 3
26

Contents lists available at ScienceDirect

International Journal of Impact Engineering

journal homepage: www.elsevier.com/locate/ijimpeng

Analytical models for penetration mechanics: A Review

2XTagedPD1 XCharles
XD E. Anderson Jr.D4X3 XXD
TagedPCEA Consulting, San Antonio, TX 78240 USA

TAGEDPA R T I C L E I N F O TAGEDPA B S T R A C T

Article History: A review of analytical penetration models has been conducted and summarized. Models include the Ponce-
Received 8 December 2016 let equation, hydrodynamic theory, modified hydrodynamic theory, Recht-Ipson, Tate-Alekseevskii, cavity
Revised 22 March 2017 expansion, Ravid-Bodner, Walker-Anderson, and similitude modeling. These models describe, depending
Accepted 22 March 2017
upon assumptions, rigid-body penetration, eroding penetration, steady-state and transient penetration, and
Available online 23 March 2017
perforation. Model assumptions are highlighted, and examples are provided of model predictions against
experimental data. The manuscript has 30 figures, many of which compare model results to experimental
TagedPKeywords:
data; 59 reference citations are included.
Penetration mechanics
Analytical modeling
© 2017 Elsevier Ltd. All rights reserved.
Plastic-flow fields

1. Introduction TagedPvalidity is limited by the experimental data used to determine the

model parameters. Secondly, the model provided a major advance,
TagedPSeveral good articles exist in the literature that review pene- and was not a slight modification of an existing model. Any errors of
tration, perforation, and other aspects of terminal ballistics. omission are those solely by the author, either because the model
Backman and Goldsmith [1] present an extensive survey (278 did not meet the selection criteria, or simply because we were
references) of the interaction of projectiles and targets. They unaware of the model.
looked at semi-infinite targets, thin plate penetration and perfo- TagedPIn today's world of large-scale numerical simulations, it can be
ration, and intermediate and thick targets, covering the entire asked: “Why is there a need for analytical models?” There are several
velocity range. Jonas and Zukas [2] reviewed available analytical answers to this question. If an analytic model can be developed
methods for the study of projectile-armor interactions, with an that captures the essence of the phenomenology, then understand-
emphasis on numerical simulations (167 references), focused ing has been demonstrated, and the essential and relevant physics/
on the velocity range of 0.5 to 2.0 km/s. A special issue of the mechanics have been applied/incorporated. This fundamental
International Journal of Engineering Science [3] was dedicated to understanding leads to improved designs. Further, analytical models
penetration mechanics, covering experimental investigations, are typically hundreds to thousands times faster in execution than
analytical modeling, and numerical modeling. Anderson and numerical simulations. Thus, analytical models provide tools for pre-
Bodner [4] reviewed analytical and numerical modeling of bal- dictions, rapid design studies, first-order optimization, and risk
listic impact (88 references). Goldsmith [5] provides an exten- assessments. For example, a typical risk assessment might run thou-
sive review of non-ideal projectile impact on targets (367 sands (maybe millions) of Monte Carlo scenarios, which would be
references). The above articles concentrate, in varying detail, on prohibitively time-consuming for numerical simulations, but which
experimental procedures/results, data trends, analytical model- are well suited for analytical models.
ing, numerical modeling, and the mechanics of projectile-target TagedPWith the above as background, the arrangement of the paper
interactions. is largely chronological by model development. The focus is on
TagedPThere has been considerable progress in analytical modeling metallic targets.
since the publication of Refs. [1
4]. This article provides a review of
analytical models that have been developed to describe the process 2. Poncelet equation
of penetration into metallic targets, subject to several constraints.
Firstly, the model has to provide significant insight into the mechan- TagedPJean-Victor Poncelet (1788
1867) was a French engineer and
ics of penetration. This constraint eliminates empirical models that mathematician. He developed an ordinary differential equation to
require experimental data to fit an assumed functional form. It is describe penetration of rigid projectiles. Newton's second law gives
noted that empirical models can be very useful, but their range of the projectile deceleration as a result of a resisting force:
dv

E-mail address: [email protected]
M D ¡F D ¡ A C Bv2 ð1Þ
dt

http://dx.doi.org/10.1016/j.ijimpeng.2017.03.018
0734-743X/© 2017 Elsevier Ltd. All rights reserved.
4 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26

NOTATION ɛ_ p plastic strain rate

g strain hardening coeff., Eq. (47)
A, a model constants
λ dynamic effect parameter, Eq. (38)
‘ instantaneous projectile length
A(r,u) vector potential, Eqs. (50) and (59)
m (rt /rp)1/2, Eq. (32a)
a projectile radius (Fig. 14), Eq. (39)
B, b model constants
mf friction coefficient, Eq. (40)
Bmax maximum hardness at bottom of crater, Eq. (16)
n Poisson’s ratio
c bar wave speed
r density
co bulk sound speed
ft dynamic yield strength
D projectile diameter
S strength difference, Eq. (14)
Dij rate of deformation tensor, Eq. (62)
s stress
s efs effective flow stress
d hole diameter
b
er, b
eu , b
ef unit vectors in r, u , and ’ directions
sn normal stress, Eq. (39)
E Young’s modulus
st tangential stress
Ed deformation energy, Eq. (19)
s xz target shear stress, Eqs. (48) and (49)
F, Fn, Ft Force, normal force, tangential force
s Yt target strength
G shear modulus
s Yp projectile strength
h distance between projectile nose to target rear Subscripts
(back) surface p projectile
K bulk modulus t target
k shock-particle velocity constant
L projectile length
M, Mp projectile mass
TagedPThe A is a static target resistance term, and the Bv2 term
m mass of L/D 1 projectile, Eq. (16)
states that the resisting force is proportional to the square of the
mp plug mass, Eq. (17)
velocity. In traditional fluid mechanics, this v2 term is usually
P penetration depth
called a drag term. Since the projectile is non-deforming, the
Pc penetration depth of a L/D D 1 projectile, Eq. (16)
cross-sectional area remains constant, and Eq. (1) can be written
Pr pressure
as:
p Lambert-Jonas exponent, Eq. (26)
R projectile radius dv

rp L D ¡ a C bv2 ð2Þ
Rc penetration channel (crater) radius dt
Rt target resistance
e (L is the effective length of the projectile if the cross-section
R (ac
1)Rc, Eq. (60)
is not a constant.) Eq. (2) can be integrated to find the total
s extent of plastic zone in projectile
depth of penetration (P) by using the chain rule of differentiation,
T plate thickness
dv/dt D (dv/dz)(dz/dt) D v(dv/dz):
t time
Z P Z 0
u penetration velocity 1 vdv
dz D ¡ ð3Þ
uback target rear surface velocity V a C bv
rp L 0 2

udebris velocity of eroded projectile debris, Eq. (37)

TagedPThus,
us shock velocity
 
v projectile velocity P rp b V2
D ln 1 C ð4Þ
vc critical velocity, Eq. (33) L 2b a
vtr transition velocity, Eq. (34)
TagedP(A more general case, with a resisting force proportional to the
V impact velocity
velocity as well as the velocity squared, in addition to the constant
V nondimensional velocity, Eq. (65)
term, can be written. Subsequent integration of the equation results
Vr residual velocity
in an arctangent term in addition to the logarithmic term. This form
V50 ballistic limit
of the resisting force is generally not used.)
W plastic work
TagedPForrestal and Piekutowski [6] measured the depth of penetration
Ws work in shear, Eq. (20)
for maraging steel projectiles with an ogival nose penetrating 6061-
Wt plastic work in target, Eq. (24)
T6 aluminum as a function of impact velocity. The depths of penetra-
Y flow stress (strength)
tion, normalized by the length of the projectile, are plotted in Fig. 1.
Yd dynamic flow stress, Eq. (47)
The Rockwell C hardness of the projectiles varied between Rc38 and
z direction of penetration
Rc53, depending upon the specific alloying and heat treatment. A
a (rp/rt)1/2, Eq. (65)
least-squares regression fit through the experimental data gives
ec
a (1 C h/Rc), Eq. (61)
the values of the parameters a and b, given by the solid line in Fig. 1
a, h axial and radial extent of plastic flow field (Ravid-
(several points, above 1500 m/s, that fall below the general trend
Bodner model)
were ignored in the regression analysis; they will be discussed
as Eq. (43)
later).
acRc extent of plastic flow field (Walker-Anderson
TagedPFor small values of bV 2/a, the natural logarithm can be expanded
model)
to give:
b blending parameter, Eq. (60)
bs Eq. (43) P rp V 2
D ð5Þ
x, d hardening exponents, Eq. (47) L 2a
e, ecrit strain; critical strain to failure which is shown as the dashed line in Fig. 1. Thus, the depth of
penetration is proportional to the square of the impact velocity. The
 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26 5

Fig. 1. Normalized penetration for very hard steel projectiles into 6061-T6 aluminum. Data from Ref. [6].

TagedPresistive force, Eq. (2), is approximately a constant, independent TagedP xperiments, recovered projectiles of the deleted data points show
e
of the impact velocity, since the parameter b does not enter into considerable deformation.
Eq. (5). TagedPOne last comment is worth mentioning before proceeding.
TagedPThe normalized penetration for blunt-nose tungsten-alloy pro- The small argument expansion, given by Eq. (5), is only valid
jectiles into a 7000 series aluminum target is shown in Fig. 2. The to about 0.3 km/s in Fig. 2, given the values of a and b (shown
experimental data are from two data sets generated by researchers in the figure). Thus, for this projectile-target combination, the
at the Ernst-Mach-Institut [7,8]. A least-squares regression was con- assumption of a constant resisting force is only valid to about
ducted on all data below 0.7 km/s; the result is shown with the solid 0.3 km/s; at higher impact velocities, a “drag” resistive force
line. However, if four data points that are circled are deleted from must also be included to reproduce the experimental data.
the analysis, the resulting fit is shown with the dashed line. The fit- TagedPWe will see that the Poncelet equation will manifest itself
ted standard error is 45% less for the dashed curve than for the solid several times in the development of penetration models.
curve. Because we know that at sufficiently high impact velocities
that the projectile will begin to deform, this Poncelet analysis is sug- 3. Hydrodynamic penetration and modified theory of
gestive that the operative mechanics (that is, the assumption of hydrodynamic penetration
rigid-body penetration) is not true for these four omitted data
points. Note that without experimental evidence, it can only be said 3.1. Hydrodynamic theory
that this analysis is suggestive. Although not very satisfying, an alter-
native explanation is that the Poncelet parameters change during TagedPThe hydrodynamic theory of penetration emerged during
penetration. However, a similar analysis of Poncelet fits to the data WWII and was applied to shaped-charge jet penetration. The
for a completely different set of experiments, described in Section theory was developed by Birkhoff, MacDougall, Pugh, and Taylor
6.3, also show that a better fit to the experimental data is achieved if [9]; the authors acknowledged in a footnote that independent
some of the higher velocity data points are not included. In these work of Hill, Mott and Pack in England led to the same results.

Fig. 2. Normalized penetration for tungsten-alloy projectiles into a structural grade aluminum. Date are from Refs. [7,8].
6 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26

Fig. 3. Schematic of an incompressible, hydrodynamic jet penetrating an incompressible hydrodynamic target.

TagedPThe operative equation is derived from conservation of momen- TagedPSince penetration is steady state,2 the time of penetration is:
tum, assuming that the strengths and viscosity of target and pen- L
etrator materials can be neglected, that is, the problem can be tD ð12Þ
v¡u
treated by hydrodynamics. Thus, the momentum equation is:
and the penetration depth is P D ut. After rearranging, the normal-
@v 1
 1 ized depth of penetration is:
C r v2 ¡ v £ ð r £ vÞ D ¡ r Pr ð6Þ
@t 2 r qffiffiffiffiffiffiffiffiffiffiffiffiffi
P u
where r is the vector operator. With the assumption that the jet and D D rp =rt ð13Þ
L v¡u
target materials are incompressible, the term on the right-hand side
can be written as r(Pr/r). A further simplification is done by using a
coordinate transformation, as shown in Fig. 3. It is assumed that the 3.2. Modified hydrodynamic theory
jet moves at a constant velocity v, and that the penetration velocity,
u, is also a constant; thus the projectile-target interface does not TagedPShaped-charge jets have a velocity gradient. Thus, the jet can be
move in Fig. 3b. The momentum equation, Eq. (6), then becomes partitioned into a series of connected jets with different velocities.
  The hydrodynamic theory is then applied to each section of the
1 2 Pr
½v £ ð r £ vÞ D r v C ð7Þ jet. This procedure overpredicted the final depth of penetration.
2 r
Eichelberger [10] reasoned that at sufficiently low velocities, partic-
with the assumptions of hydrodynamic and incompressible materi- ularly into strong (steel) targets, that target strength effects could
als under steady-state conditions. The dot product of both sides of not be neglected. He modified Eq. (10):
the equation by the vector velocity gives:
  1 1
r ðv¡uÞ2 D rt u2 C S ð14Þ
1 2 Pr 2 p 2
v r v C D0 ð8Þ
2 r where S D s Yt
s Yp represents the resistance to plastic deforma-
since [v £ (r £ v)] is perpendicular to v. Thus, we have that the gra- tion. The penetration velocity, Eq. (11), is modified by the addition of
dient of the term in the parenthesis is perpendicular to v. The term S to the right-hand side of Eq. (14). A more general expression for
within the parenthesis is usually called Bernoulli's equation. It is the penetration velocity will be given when the Tate model is
important to note that while the units of the terms within the paren- described. Eichelberger determined that S was 1 to 3 times the
thesis in Eq. (8) have units of specific energy, Eq. (8) was derived uniaxial yield strength of the target material (as the jet is typically
from the momentum equation.1 copper, s Yp is very small compared to s Yt).
TagedPAlong the projectile-target centerline, Eq. (8) becomes: TagedPAllen and Rogers [11] applied the modified hydrodynamic theory
    to high-velocity impacts of six different rod projectile materials
@ 1 2 Pr @ 1
v C D rv2 C Pr D 0 ð9Þ (gold, lead, copper, tin, aluminum, and magnesium) into 7075-T6
@z 2 r @z 2
aluminum. Instead of S, they used the symbol ft, which they called
TagedPEq. (9) is integrated from the back of the projectile to the projec- the dynamic yield strength of a solid target relative to a fluid jet.
tile-target interface, and then from the interface into the target. The They found that ft  1.89 GPa, or 3.9 times the yield strength of
pressure at the back of the projectile is zero, and the target is 7075-T6 aluminum (0.48 GPa). But they also determined that ft
assumed to be semi-infinite (the pressure at infinity is zero). Using had to be written as a function of impact velocity to reproduce the
the nomenclature from Fig. 3b, so that the velocity at the projectile- final depth of penetration.
target interface is zero, and the target velocity at infinity is
u, TagedPThese modifications to the hydrodynamic theory apply to rela-
integration of Eq. (9) gives: tively weak projectiles. The assumption is that the projectile is
1 1 completely consumed, that is, no projectile material remains at the
r ðv ¡ uÞ2 D rt u2 ð10Þ bottom of the penetration channel. This assumption is true only
2 p 2
for high velocity impacts and/or weak projectiles. Additionally,
TagedPThis is the hydrodynamic, steady-state, incompressible model for
the above modifications apply to relatively long projectiles where
penetration. Solving for the penetration velocity gives:
steady-state penetration is applicable.
v TagedPChristman and Gehring [12] described the four phases of
u D qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð11Þ
1 C rt = rp high velocity penetration: (I) a transient phase, (II) the primary

2
Specifically, the shock phase of penetration is neglected, that is, steady state is
1
Most elementary physics textbooks derive Bernoulli's equation using conserva- reached instantaneously; and there is not a terminal phase of penetration, i.e., pene-
tion of energy methods. tration stops as soon as the last particle of the jet has struck the target.
 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26 7

Fig. 4. Phases of penetration for a L/D 20 projectile impacting at 3.0 km/s.

TagedPphase, (III) the secondary phase, and (IV) a recovery phase. Although TagedPadded the crater depth, Pc, obtained for a L/D D 1 rod. Christman
the concept was correct, their original figure is highly distorted. and Gehring correlated Pc with experimental data:
These phases of penetration are shown in Fig. 4 for an L/D 20 tung-  1=3 1 !1=3
rp 2
sten-alloy projectile impacting a steel target at 3.0 km/s. In Fig. 4, 2 mv
Pc D 0:13 ð16Þ
the phases are renamed: (I) shock phase, (II) steady-state phase, (III) rt B max
terminal (transient) phase, and (IV) recovery (elastic rebound of the
where Bmax is the maximum hardness at the bottom of the crater (to
bottom of the penetration channel). The relative time duration of
account for strain hardening of target material). The correlation of
phases II and III depend on the aspect ratio of the projectile. The
Eqs. (15) and (16) with the measured penetration depths is shown
shock phase persists for a very short time (that depends upon the
in Fig. 5 for rods with aspect ratios of 1 to 25 at impact velocities
projectile diameter), and penetration during this phase is typically
of 2.0
6.7 km/s, for aluminum and steel projectiles impacting
included in the steady-state phase.
aluminum and steel targets (only a subset of the data in Ref. [12]
TagedPChristman and Gehring then made a very interesting modifica-
have been plotted).
tion to the hydrodynamic theory that accounted for terminal tran-
sient effects; thus, their modification is applicable to small-aspect-
ratio projectiles. They separated penetration into a primary (steady- 4. Conservation of energy and momentum model
state) phase and a secondary (transient) phase. They stated penetra-
tion for the transient phase could be approximately by the penetra- 4.1. Recht-Ipson model
tion of an L/D D 1 projectile. Therefore, for high impact velocities,
total penetration is given by: TagedPUp to this point, the models have considered the targets to be
 1=2 semi-infinite, although assumptions have been invoked by various
rp
P D ðL¡DÞ C Pc ð15Þ investigators in using these models to account for finite-thick tar-
rt gets. Recht and Ipson [13] developed a model to estimate the resid-
where penetration of the primary (steady-state) phase is given by a ual velocity of chunky projectiles (L/D » 1) impacting relatively thin
rod that is one diameter shorter than its initial length. Then they plates using conservation of momentum and energy considerations.

Fig. 5. Measured penetration vs. application of Eqs. (15 and 16); the figure is reproduced from Ref. [12].
8 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26

TagedPThe authors considered impact at ordnance velocities where the TagedPto deform (mushroom) the projectile. To reproduce the analytical
projectile stays intact, although it can deform (mushroom) slightly. results in Ref. [13], the authors must have assumed that d/D
An experimental observation for impacts against thin plates is that a increases with impact velocity; that is, the projectile mushrooms
plug is ejected and is attached to the projectile. Application of the and pushes out a plug with a larger diameter than the original pro-
conservation of momentum gives: jectile. To reproduce the Recht-Ipson results for T/L D 0.29 and 0.44,

 it was assumed that d/D goes from 1.0 to approximately 1.4 linearly
Mp V D Mp C mp Vr ð17Þ
with impact velocity as the velocity goes from V50 to approximately
where the large M and small m refer to the masses of the projectile 950 m/s. The dashed lines nominally replicate the analytical results
and plug, respectively Application of conservation of energy gives: of Fig. 3 in Ref. [13].
1 1
 TagedPRecht and Ipson [13] examined oblique perforation of thin plates
Mp V 2 D Mp C mp Vr2 C Ws C Ed ð18Þ and perforation of thick plates by cylinders. Eq. (21) must be modi-
2 2
fied and additional assumptions and experimental data are required
where Ws is the energy used to shear the plug and Ed is the energy
to make analytical predictions. Thus, some of the elegance of the
associated with deformation and heating (plastic work). Ed is simply
normal impact against a thin plate is lost.
the difference between the initial and final kinetic energies, which,
TagedPRecht and Ipson also examined the case of an armor-piercing (AP)
using Eq. (17) is:
  bullet perforating a thick plate [13]. In this case, no plug is ejected;
1 mp and there is no deformation of the bullet. Thus, all kinetic energy
Ed D Mp V 2 ð19Þ
2 Mp C mp loss is through plastic deformation of the target:
TagedPThe energy loss due to shearing the plug is estimated from 1 1
Mp V 2 D Mp Vr2 C Wt ð23Þ
Eq. (18) by finding the minimum velocity that gives a residual 2 2
velocity of zero. This minimum velocity is V50. With Vr D 0, and TagedPIf it is assumed that the plastic deformation is relatively indepen-
substituting Eq. (19) into Eq. (18), the shear energy is calculated: dent of the impact velocity, then
 
1 Mp 2 1
Ws D Mp V50 ð20Þ Wt D 2
Mp V50 ð24Þ
2 Mp C mp 2
Substituting Eqs. (19) and (20) into Eq. (18) and solving for the resid- and after some rearranging, Eq. (23) becomes:
ual velocity gives: !1=2
  Vr V2
Mp
2 
2 1=2 D ¡1 ð25Þ
Vr D V ¡V50 ð21Þ V50 2
V50
Mp C mp
T ssuming that the projectile is nominally a right-circular cylin-
agedPA TagedPExperimental data for an AP bullet into a steel target is shown
der, then Eq. (21) can be rewritten in terms of the geometric para- in Fig. 7, along with the analytical prediction given by Eq. (25).
meters, with d/D being the ratio of plug diameter to projectile
diameter, and T/L the ratio of plate thickness to projectile length:
4.2. Conservation of energy
1
2 
2 1=2
Vr D V ¡V50 ð22Þ
rt
d  2 T TagedPA remark is in order concerning the use of conservation of energy
1C r D L
p
in penetration modeling. It is difficult to account for all the various
The authors note that Eq. (22) is valid for T/L < 1/2 and T/D < 1/2. mechanisms that dissipate energy. Further, the proportion of energy
TagedPResults for three sets of experiments for steel cylinders into mild dissipation (energy transfer mechanisms) changes with impact
steel plates are shown in Fig. 6. At relatively low impact velocities, velocity. In particular, as the impact velocity increases, projectile
the hard steel fragment does not deform. The authors argue that the kinetic energy is transferred to the target in terms of target kinetic
effect of the ratio d/D is small, and since d is not usually known, they energy and elastic compression energy; this compression energy is
set the ratio to 1.0. This statement is true only for T/L D 0.15. As dissipated by plastic work at later times [14]. Walker [15] showed
the impact velocity increases and the plate thickness increases, the that it is the transfer of this energy at the time of penetration that
stresses at the projectile-target interface become sufficiently large defines the forces on the projectile, and for an energy rate balance to

Fig. 6. Comparison of Recht-Ipson model for chunky projectiles into thin plates with experimental data. Data from Ref. [13].
 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26 9

Fig. 7. Comparison of Recht-Ipson model for an armor-piercing bullet against a thick target. Data from Ref. [13].

TagedPbe successful, it must include transfer of energy stored in the target TagedP ardnesses) [17]. A subset of the data is shown in Fig. 8. The Lam-
h
as elastic compression. Although this is done automatically within bert-Jonas equation [18] provides for a general expression for the
numerical simulations, this would seem to be intractable within residual velocity as a function of impact velocity:
the context of an analytical model. (

p 1=p
TagedPIn summary, conservation of energy can be useful in analytical a V p ¡V50 V > V50
Vr D ð26Þ
modeling, but generally only over a limited velocity range and/or 0 VV50
projectile-target combination. For example, Recht and Ipson [13] TagedPA least-squares regression analysis is used to determine fit
made assumptions on energy dissipation, but then commented on parameters a, p, and V50. Note that if the parameter p is equal to 2.0,
the restrictions and range of applicability. Eq. (26) is nominally an expression for the conservation of energy
[similar to Eqs. (23)
(25) above], but including the parameters a
5. Similitude analysis and p allow for a more general (better) curve fit to experimental
data.
TagedPSimilitude analysis, sometimes called scale modeling, is a proce- TagedPDividing Eq. (26) by V50 gives:
dure for casting results in a nondimensional form; for example, see  p 1=p
Vr V
Ref. [16]. Eq. (25) is an example of a relationship between a normal- Da ¡1 ð27Þ
ized residual velocity and a normalized impact velocity. Often, non- V50 V50
dimensionalization allows what would seem to be disparate data This suggests that the experimental data should be plotted as
to be collapse into a single response curve. One example for pene- Vr/V50 versus V/V50. The data in Fig. 8 are plotted using these non-
tration mechanics will be given here, and then another will be dimensional quantities in Fig. 9. Besides the experimental data in
described later. Fig. 8, twelve other projectile-target combinations are also plot-
TagedPAn experimental study examined the ballistic performance of 17 ted as the light gray hexagons in Fig. 9. All the experimental data
different projectiles (a tungsten alloy and six different steels with collapse to a single curve; the expression for the curve is given in
different heat treats) into two different armor steel targets (different the figure.

Fig. 8. Residual velocity vs. impact velocity for various types of projectiles into two hardnesses of armor steel [17].
10 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26

Fig. 9. Normalized residual velocity vs. normalized impact velocity for data in Fig. 8 [17].


T lthough nondimensionalization, using the correct choice of
agedPA TagedPu D v
¡
Rt ¡ Yp
rp D rt D r ð32bÞ
parameters, permitted collapse of the experimental data, it does 2 rv
not necessarily explicitly allow one to determine the operative The solution for u in Eq. (14) is given by Eq. (32) with Yp · 0, and Rt
mechanics. Nevertheless, the results of Fig. 9 demonstrate that a replaced by S.
substantial amount of basic information concerning projectile and TagedPFrom Eq. (32), it is seen that there is a critical velocity, vc,
target material response (including deformation and failure) is con- were u D 0:
tained in V50. "
#1=2
2 R t ¡ Yp
vc D ð33Þ
6. Tate-Alekseevski model rp
That is, when the rod velocity drops to the critical velocity, there is
6.1. Eroding penetration
no further penetration. We will examine procedures for estimating
Yp and Rt after the next couple of subsections.
TagedPIn contrast to shaped-charge jets, projectiles (rods) decelerate
while penetrating a target. Independently, Tate [19] and Alekseevski
6.2. Rigid-Body penetration and transition from rigid-body to eroding
[20] proposed a modified Bernoulli equation:
penetration
1 1
r ðv ¡ uÞ2 C Yp D r u2 C Rt ð28Þ
2 p 2 t TagedPIn the previous subsection, Rt > Yp and there is erosion of the
where Yp is considered to be the dynamic flow stress of the projectile rod. Tate examined the case where Yp > Rt [21].3 For this situa-
and Rt is the target resistance to penetration. Separating the rigid tion, there exists a transition velocity, vtr, at and below which
potion of the rod from that part that is undergoing plastic deforma- the rod no long erodes, that is, penetration occurs in rigid-body
tion (mushrooming), the force acting to decelerate the rigid portion mode. For rigid-body penetration, u D v. The transition velocity
of the rod of length ‘ is given by: is found from Eq. (32a) by equating u and v · vtr, and solving
dv for vtr:
rp ‘ p R2p D ¡ p R2p Yp ð29Þ
dt 
1=2
2 Yp ¡Rt
dv Yp vtr D ð34Þ
D¡ ð30Þ rt
dt rp ‘
TagedPWhen penetrating as a rigid projectile, the stress on the end of
T he initial length of the rod is L, but since the rigid portion of the
agedPT the rod is given by the right-hand side of Eq. (28). Thus, the decelera-
rod is moving faster than the penetration velocity, the rod is getting tion is:
shorter with time:  
dv 1 1
d‘ D¡ rt v2 C Rt ð35Þ
D ¡ ðv ¡ uÞ ð31Þ dt rp L 2
dt
Thus, the three coupled equations, (28), (30), and (31) are solved Eq. (35) has the same form as Eq. (2), Poncelet's equation, with a D Rt,
simultaneously to determine the time history of v, u, and ‘. and b D rt /2. Thus, Eq. (4) is rewritten as:
 
TagedPEq. (28) can be rearranged to solve for the instantaneous penetra- P rp r V2
tion velocity in terms of the projectile velocity (which is changing D ln 1 C t ð36Þ
L rt 2Rt
with time) and the material characteristics of the projectile and
TagedPNormalized penetration versus impact velocity is shown in
target:

 Fig. 10 for a steel (7.9 g/cm3) projectile penetrating an aluminum
v ¡ m v2 C A 1=2
u D rp 6¼ rt
1 ¡ m2 3
Notwithstanding that the model described in Section 2.5.1 is usually called the
!1=2

 ð32aÞ Tate-Alekseevki model, Addison Tate extensively explored the modified Bernoulli
rt 2 Rt ¡Yp 1¡m2 equation, including rigid-body penetration, transition of rigid-body to eroding pene-
mD AD tration, and procedures for estimating Yp and Rt. Thus, the coupled equations (28),
rp rt
(30), and (31) can truly be called Tate's model.
 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26 11

Fig. 10. Normalized penetration vs. impact velocity for various Yp
Rt combinations for a steel projectile penetrating an aluminum target.

Fig. 11. Comparison of Tate-Poncelet equation with experimental data for steel projectiles penetrating 6061-T6 aluminum targets. The dashed line is given by Eq. (36). Data are
from Refs. [24, 25].

agedP(T 2.7 g/cm3) target. A variety of hypothetical Yp
Rt combinations are 6.3. Rigid-body, eroding, and secondary penetration
shown. Since Yp > Rt for all the examples shown, penetration at
low impact velocities is initially in the rigid-body mode. But at TagedPThe projectile will begin to erode, for the case of Yp > Rt, when the
some point, given by Eq. (34), the rod begins to erode. The X’s in velocity is sufficiently high that the penetration pressure is larger
the figure mark the transition velocity from rigid-body to eroding than the strength of the projectile. As the impact velocity increases,
penetration. it can be seen in Fig. 10 that P/L approaches the hydrodynamic limit
TagedPExperimental data from Forrestal et al. [22,23], are plotted in from above, except for when Yp is only slightly larger than Rt.5
Fig. 11. The results for three nose shapes are plotted: hemispherical, TagedPAlthough the transition velocity for rigid-body to eroding pene-
conical, and an ogive. T-200 and C-300 are types of maraging steel. tration is given by Eq. (34), this is overly simplistic. Depending upon
A least-squares regression analysis was conducted to find Rt in the ductility of the projectile, substantial mushrooming can occur at
Eq. (36); the resulting fit is shown in Fig. 11 with Rt D 1.62 GPa. With the projectile-target interface, which shortens the rod, prior to fail-
Yp D 3.0 GPa, vtr is 0.95 km/s, and if Yp D 4.0, vtr is 1.28 km/s. Regard- ure of projectile material (erosion). Further, as the projectile mate-
less, the Tate-Poncelet equation cannot reproduce the conical- and rial is made stronger, to resist deformation and maintain rigid-body
ogival-nose data at the higher impact velocities. Nose shape can penetration, the projectile material tends to become more brittle,
result in deeper rigid-body penetration than can be accounted for in which means that small strains lead to material failure and erosion.
a simple theory.4 The results for three strong steel projectiles with an ogival nose into
6061-T6 aluminum are shown in Fig. 12. The experimental data are
from Ref. [24]. The solid curve is a fit to the rigid-portion of the P/L
4
The two-term Poncelet equation, Eq. (4), can be used to fit the whole range of
5
experimental data, but the fit parameters do not have any significant relationship/ This is in contrast to Rt > Yp, where P/L approaches the hydrodynamic limit from
interpretation to physical parameters. below as the impact velocity increases.
12 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26

Fig. 12. Normalized penetration for hard steel ogive projectiles into 6061-T6 aluminum depicting the rapid transition from rigid-body to eroding penetration. Data from Ref. [24].

Fig. 13. Normalized penetration for tungsten-alloy projectiles into 7020 aluminum targets. Data are from Ref. [7,8].

TagedPresults using Eq. (4), with a D 1.956 GPa and b D 0.0518 g/cm3. Note TagedP pproximately 0.5 and 0.7 km/s, it is believed that the projectile
a
that although the Poncelet equation provides a good overall fit to is mushrooming.6 By 0.8 km/s, the projectile begins to erode, and
the experimental data, all the low velocity results tend to lie above there is a significant decrease in normalized penetration for a
the curve fit. small change in impact velocity.
TagedPIt is observed that the decrease in penetration efficiency (P/L) TagedPThe velocity of the erosion debris, udebris, is (2 u
v), as shown
is not gradual as indicated in Fig. 10. Vertical lines in Fig. 12 in Fig. 3. The direction of this velocity depends upon the ratio of
denote the approximate impact velocity where the two projec- target to projectile density. Ignoring strength effects, if rp << rt,
tiles begin to erode. The onset of erosion occurs over a relatively then u  0, and udebris 
v. If rp D rt, then udebris D 0; and if rp > rt,
narrow velocity range; this is particularly evident for the VAR then udebris is into the target. Then, if
4340 projectile, where this transition velocity interval is approxi-
1
mately 75 m/s. The AerMet 100 appears to be deforming at about r u2 > Rt ð37Þ
2 p debris
1.81 m/s, but unfortunately, the next data point is at 2.04 km/s.
there is additional penetration, called secondary penetration [11].
However, similar data using spherical-nose projectiles also show
TagedPThe penetration velocity u can be estimated from the Tate model
a very narrow velocity interval for the transition from rigid-body
if Yp and Rt are known. Given the uncertainties in Yp and Rt, it is esti-
to eroding penetration [25].
mated that secondary penetration will occur between 1.25 and
TagedPExperimental data for a flat-nosed, tungsten-alloy projectile
1.40 km/s (Yp D 2.0 GPa, and Rt between 2.15 GPa and 2.35 GPa). The
into 7020 aluminum (rt D 2.8 g/cm3) are shown in Fig. 13 [7,8].
vertical dotted lines in Fig. 13 denote where secondary penetration
Tungsten alloy is more ductile than the maraging steels used in
the experiments described above, and thus, will plastically
deform considerably more before failure. Only the data to an 6
Wickert [8] shows photographs of recovered projectiles that impacted the alumi-
impact velocity of approximately 0.54 km/s were used for num target at 0.540 km/s and 0.664 km/s. The projectile had not deformed at the
the Tate-Poncelet curve fit; at impact velocities between lower impact velocity, but there is deformation at the higher impact velocity.
 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26 13

Table 1
Quasi-static cavity expansion solutions.

Spherical Cylindrical
h
i h
i h pffiffi
i h pffiffi
i
Incompressible Rt D 2s3Yt 1 C ln ð1 C Entt Þs Yt ! D 2s3Yt 1 C ln 32E t
s Yt Rt D 2s3Yt 1 C ln 2ð1 C3nEttÞs Yt ! D 2s3Yt 1 C ln 3s3EYtt
h
i h pffiffi
i
Compressible 2s Yt Et
Rt D 3 1 C ln 3ð1 C nt Þs Yt Rt D spYtffiffi3 1 C ln 6ð1 C3nEttÞs Yt
h
i h
i
Tate solenoidal model Rt D s Yt 23 C ln ð4¡e2E¡λt Þs ! λ D 0:7 s Yt 23 C ln 0:57E
s Yt
t
Yt

TagedPbegins. Of course, if the left-hand side of Eq. (37) is only a little TagedPA problem, though, is that Rt is not a material property, although
greater than Rt, one would not expect substantial secondary penetra- it does depend upon material properties. Anderson, Littlefield, and
tion. But as the impact velocity increases, secondary penetration Walker [14] showed that Rt also depends upon the impact velocity,
keeps the P/L response from asymptotically approaching the hydro- just as Allen and Rogers showed that ’t was a function of impact
dynamic limit; instead, P/L significantly exceeds the hydrodynamic velocity [11].
limit as shown in Fig. 13.
6.5. Summary of the Tate model
6.4. Estimates for Yp and Rt
TagedPThe Tate model is a one-dimensional model that predicts the
T ate realized that for the modified hydrodynamic model to be
agedPT time history of penetration, including projectile deceleration. The
useful, he needed a methodology for estimating Yp and Rt. Two model provides insights into the dependence of penetration on
articles summarized his efforts [26,27]. The term Yp is associated material properties (for example, density and strength). It predicts a
with the deceleration of the projectile, and is estimated from the critical velocity below which no (additional) penetration occurs. It
flow stress, s Yp, by: also predicts rigid-body penetration, and a threshold velocity for

 the transition from rigid-body penetration to eroding penetration.
Yp D 1 C λ s Yp ð38Þ
Tate provided engineering estimates for evaluating Yp and Rt;
where λ is a constant independent of velocity and accounts for the primary weakness of the model is obtaining a priori accurate
dynamic material effects. estimates for Rt, which depends upon the impact velocity as well as
TagedPRt is the resistance to plastic deformation of the target flow. Vari- material properties.
ous methods have been proposed to estimate Rt. In particular, cavity
expansion theory has been used [28]. The concept is to assume an 7. Rigid-Body penetration revisited
incompressible or compressible plastic region and an elastic region.
A cavity from zero radius to a radius R is opened quasi-statically. A 7.1. Nose shape and dynamic cavity expansion
similarity solution is obtained; the solution leads to a plastic region
and an elastic region, and allows calculation of the stress at the inter- TagedPForrestal and his co-authors examined rigid-body penetration.
faces. Spherical cavity expansion or cylindrical cavity expansion can Although primarily interested in penetration into geologic materials,
be assumed. The various solutions are shown in Table 1. If the mate- they were uncertain about the accuracy of their model because of
rial is assumed to be incompressible, then n · 0.5, as indicated in scatter in experimental data, inherent in brittle materials. Therefore,
the table. for model validation, they decided to study penetration into 6061-
TagedPTate conducted a cavity expansion on a flow field inspired by the T6 aluminum. In their work, they calculated the force on the projec-
magnetic flow lines in a solenoid (the impact velocity is sufficiently tile nose, examining different nose shapes such as hemispherical,
high that hydrodynamic flow is established, with a cavity radius conical, and an ogive. An example for calculating the force on a
larger than that of the projectile) Assuming that the target material hemispherical-nose projectile will be shown here; the procedure for
is incompressible, that the J2 flow law applies, and when yielding, other nose shapes is found in Ref. [22].
the material is perfectly plastic (no account for rate effects or micro- TagedPAssuming the direction of penetration is in the z-direction, the
structural features, etc.), he derived a different expression for Rt, as incremental normal force (dFn) on the hemispherical nose, see
shown in the third row of Table 1. Note that this solution depends Fig. 14, is given by:
upon the value of λ in Eq. (38). After analyzing experimental data,
Tate determined that λ D 0.7 was an appropriate value to account for dFn D 2pRas n ðv; uÞdu R D a sinu ð39Þ
dynamic effects [27].
TagedPAs an example, consider that the target is an armor-like steel,
with s Yp D 1.0 GPa, nt D 0.3 (nt D 0.5 for an incompressible material),
and Et D 200 GPa. The values calculated for Rt are shown in Table 2.
Note that the resistance to plastic deformation is 3 to 6 times the
flow stress of the target material; the target material resists the
opening of a cavity (or penetration channel) due to self-confinement.
When comparing model predictions to experimental data, Tate's
solenoidal model provides a much more realistic value for Rt [29].

Table 2
Estimates for Rt for an armor-like steel.

Spherical Cylindrical

Incompressible 3.93 GPa 3.32 GPa

Compressible 3.74 GPa 3.15 GPa
Tate solenoidal model 5.44 GPa
Fig. 14. Rigid projectile geometry with hemispherical nose.
14 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26

Fig. 15. Normalized penetration as a function of impact velocity for various dynamic spherical cavity solutions and experimental data. Data from Refs. [22,23].

TagedPwhere s n(Vz,u) represents the normal stress on the projectile nose, TagedPwhere M is the projectile mass. The authors found that the total
which is a function of the penetration velocity and u. They also depth of penetration was relatively insensitive to friction coeffi-
assume that there is frictional (tangential) force (dFt) proportional to cients between 0.02 and 0.1 for the hemispherical-nose projectile,
the normal stress, s t D mf s n , then but predictions for the conical and ogival nose projectiles were more
sensitive to values of mf [22].
dFt D 2pRamf s n ðv; uÞdu ð40Þ
TagedPUsing the nomenclature of Forrestal et al. [23], the mass of
TagedPTo obtain the component of the normal and tangential forces in the projectile, where L is the length of the shank and a is the
the z-direction, dFn and dFt are multiplied by cosu and sinu, respec- spherical nose radius, so that the total projectile length is (L C a),
tively. Integrating over the hemisphere gives the retarding force: is given by:
Z p=2 h i 2
Fz D pa2 s n ðv; uÞ sin2u C 2mf sin2 u du ð41Þ M D r p p a2 L C r p a3 ð45Þ
0 3 p
It now is required to calculate s n(v,u). Thus, Eq. (44) can be rewritten as:
TagedPThe authors used dynamic cavity expansion [30], including strain  
P rp b V2
hardening and rate effects, to calculate the force on the projectile D ln 1 C s ð46Þ
ðL C 2a=3Þ bs as
nose. Similar to quasi-static cavity expansion, there is an incom-
pressible or compressible plastic region and an elastic region. The TagedPTheir work perhaps culminated in an article by Warren and For-
difference is that in dynamic cavity expansion, the cavity is opened restal that considered strain hardening and strain-rate sensitivity on
at a constant velocity. A similarity solution leads to a plastic bound- the penetration of aluminum targets with spherical-nosed rods [32].
ary, an elastic boundary (plastic, elastic, and undisturbed regions); They considered the constitutive relationship:
and stresses at the interfaces. 8
>
< Eɛ s  Yd  d
TagedPThe math becomes more complicated for compressible materials,  x  d ɛ_
sD Eɛ ɛ_ Yd D Y C g ð47Þ
and for materials where a more realistic constitutive response is >
:Y Cg s > Yd ɛ_ 0
assumed, for example, strain hardening and strain rate effects, e.g., Yd ɛ_ o
Ref. [31,32]. Here, only the incompressible spherical dynamic cavity where the exponents x and d, and the proportionality constant g ,
result is shown for an elastic-perfectly plastic material: are determined from fitting experimental stress-strain data. The dot
  
2Y 2Et 3 symbolizes the strain rate. The solutions for incompressible and
s r D t 1 C ln C rt v2 D As C Bv2 ð42Þ
compressible spherical cavity expansion,7 with and without rate
3 3Yt 2
effects, as a function of impact velocity are shown in Fig. 15.
The first term, As, is the quasi-static spherical cavity expansion term
TagedPExperimental data from [22,23] are plotted in Fig. 15. It is seen
(Table 1). Note that the stress (or resistance to penetration) includes
that the compressible model with rate effects provides the best over-
strength and inertial effects. Also note that the resistance is propor-
all prediction of penetration, specifically going through the experi-
tional to the square of the cavity (expansion) velocity. The stress
mental data at the higher impact velocities.
normal to the projectile nose surface is then given by Eq. (41) using
the component of the velocity normal to the surface, i.e., vcosu.
TagedPInserting Eq. (42) into Eq. (41) and integrating gives: 7.2. Frictional effects
     
mf p 3rt mf p 2
 TagedPInitially, Forrestal and colleagues were concerned about friction
Fz D pa2 As 1 C C 1C v D p a2 a s C b s v 2 ð43Þ
2 4 4 effects between the projectile and the penetration cavity wall, so
TagedPNote that the force has the form of the Poncelet equation with the they included frictional forces in their model, e.g., Eq. (40). However,
coefficients calculated from physical properties. The final depth of Warren and Forrestal [32] had to neglect frictional forces in order for
penetration, following a similar procedure as done for Eqs. (1)
(4) the analytical model to reproduce the experimental data, Fig. 15.
gives: The reason is described in the paragraphs below.
 
M bs V 2
PD ln 1 C ð44Þ 7
The modeling work of Forrestal and colleagues largely focused on using spherical
pa2 bs as cavity expansion.
 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26 15

TagedPHill [33], in revisiting work done during WWII, makes the follow- TagedP greement with experimental data by plotting P D V 2 =ð2aÞ,
a
ing statement as he outlined the assumptions for his work: “The fric- albeit the value for a was obtained by analyzing the experimental
tional component can be disregarded because of surface melting.” data.
Tate made an estimate of temperature effects during rod penetration TagedPRubin et al. [40], develop a model for the drag force using what
[34]. He found that thermal conduction is significant only very close they believe to be a more realistic flow field in the target. They find
to the projectile-target interface (during the timeframe of penetra- that the force is a constant below a critical impact velocity, called
tion). Further, that when distances are scaled relative to the crater the separation velocity Vs, which is what Hill calls the cavitation
diameter, the temperature distribution is independent of the impact velocity [33].8 Since the force is a constant, then the deceleration is a
velocity, and the temperature approaches the melting temperature constant. They calculate the resisting force that is larger than that
in a small region which is of the same order of size as the conduction calculated from cavity expansion. Thus, Refs. [39-40] state that the
zone. Wilkins [35] did not initially include frictional forces in the one-dimensional cavity expansion solutions, whether cylindrical or
Lagrangian hydrocode HEMP [36], and had good agreement with spherical, do not adequately represent the two-dimensional flow
experiments. When he added friction, the calculated results became field around the penetrating projectile, and therefore, the cavity
worse. expansion solutions predict a value for the resisting stress that is too
TagedPLastly, Camacho and Ortiz [37] performed detailed finite element low. In consequence, they conclude that Warren needed to include
simulations of some of the Forrestal et al., experiments using a new target inertia to increase the resisting stress as the impact velocity
adaptive meshing technique and a constitutive material law that increased.
included strain hardening, rate-dependent plasticity, heat conduc- TagedPWarren responds [41] that Hill's comment concerns a projectile
tion, and thermal-mechanical coupling. The simulations showed a that is moving at constant velocity, which is not the case here
very thin melted layer in the target next to the projectile that because the projectile is decelerating due to target resistance; fur-
resulted in a nearly frictionless interface. ther, Section 4 of Hill's paper [33] states “that if the velocity is not
TagedPThus, it is concluded that friction in penetration mechanics is a constant, then the deceleration of the projectile will have an effect
term added to an analytical model to improve agreement with on the target resistance” [41]. Warren then shows experimental
experiments, whose sole justification is that the friction coefficient data (with the tunneling region the same diameter as the projectile
is on the order of 0.01
0.2, and thus appears reasonable. But, when so that there is no cavitation) where the deceleration is not con-
an accurate constitutive model is used, there is no need to include stant. Warren concludes that “when the strength term overshad-
friction for ballistic penetration modeling. ows the inertial term. . .the target inertia term can be neglected in
the penetration model. However, when the strength does not over-
7.3. Applicability of cavity expansion for rigid-body penetration shadow the target inertia effects, then a target inertia term must
be included” [41].
TagedPWarren [38] examines rigid-body penetration for an elastic, TagedPHaving laid out these arguments, it is observed that Warren
strain-hardening aluminum using quasi-static spherical cavity has modeled the penetration process using an accurate constitu-
expansion and dynamic spherical cavity expansion solutions to tive description of the target material. On the other hand, the
obtain the force on an ogival projectile. Warren compares the analytical model of Rubin et al., cannot explicitly account for the
depth of penetration as a function of penetration velocity against influence of compressibility, strain hardening, strain rate, and
a number of different data sets. He concludes that quasi-static tar- temperature of the target material, and must rely on numerical
get strength dominates penetration resistance at the lower impact simulations for the influence of these effects [40]. Further, for
velocities (where target inertial effects have been ignored), but accurate replication of the P versus V response, the deceleration
that at higher velocities, target inertial effects must be included to parameter a must be determined from experimental data.
match the experimental data. In effect, although more simplistic
than the analysis in Ref. [38], the quasi-static solution is similar to
Eq. (5), whereas, the dynamic solution is similar to Eq. (4). Warren
8. Dynamic plasticity and the Ravid
Bodner model
concludes: “Final depth of penetration predictions. . .that include
target inertia are found to be in excellent agreement with all of
TagedPThe next major advance in analytical penetration modeling was
the experimental penetration data for all striking velocities, pro-
the introduction of a multi-stage penetration/perforation model,
jectile geometries and ogive-nose shapes, projectiles densities,
developed by Ravid and Bodner [42]. The authors proposed five
and material properties” [38]. He further notes: “Additionally, it
stages of penetration/perforation for a rigid projectile, as shown in
is observed that target inertia must eventually be included at
Fig. 16: 1) penetration, 2) bulge formation, 3) bulge advancement, 4)
some particular striking velocity that is. . .dependent on the pro-
plug formation and exit, and 5) projectile exit.
jectile geometry, nose shape. . .for the spherical cavity-expansion
TagedPThe authors established plastic flow fields within the target.
approximation model to accurately predict final depths of pene-
As shown in Fig. 17, there are three zones where target material
tration.” [38].
is moving (there is no target motion in zone IV). Velocity distri-
TagedPRosenberg and Dekel [39] and Rubin et al. [40], take issue
butions are assigned to each zone, subject to compatibility and
with the applicability of the cavity expansion approximation
continuity conditions, and then the plastic work rates (including
applied to rigid-body penetration. They state that the resisting
strain-rate effects) are computed. Work hardening is considered
stress on a rigid projectile is constant, and that deceleration dur-
indirectly by taking the flow stress to be the average value, with
ing penetration is independent of velocity [39]. They argue that a
respect to plastic work, over the full strain range. The radial and
constant penetration channel diameter implies that the deceler-
axial extent of the plastic zone fields (hR and aR, where R is
ation is a constant [39], and they support their argument by
the projectile radius) are solved for by minimizing the plastic
examining data from Piekutowski et al. [24]. They also note
work rate. Projectile deceleration is computed from an energy
that Hill [33] states that a penetration channel with a constant
rate balance.
diameter, which is equal to that of the projectile, is an indication
for a constant resisting stress on the projectile [39]. Using exper- 8
At the critical velocity, the penetration channel begins to open up (and is no lon-
imental data from Ref. [24], Rosenberg and Dekel calculate an ger the same diameter as the projectile), penetration resistance increases, and target
average deceleration, a D V 2 =2P, where V is the impact velocity inertia must be included to determine the deceleration force on the projectile. Vs
and P is the final depth of penetration. They then show excellent depends upon nose shaped, target material, etc.
16 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26

Fig. 16. Multi-stage penetration/perforation model by Ravid and Bodner [42].

TagedPStage I (the penetration stage) continues until the plastic flow TagedP epends upon the failure mode; the specific failure mode is target
d
field reaches the rear of the target, at which point the target begins material dependent.
to form a spherical bulge, Fig. 18. Uplifting of target material at the TagedPAn example of the model is shown in Fig. 20. A 7.62-mm armor-
surface ceases, and all target motion is in the direction of penetra- piercing (AP) bullet impacted a 12-mm-thick steel plate at 855 m/s.
tion. Target mass motion, combined with the incompressibility con- The force on the projectile, the projectile velocity, and bullet dis-
dition of plastic flow, determines the plastic flow velocity field (thus, placement are plotted in the figure. The various stages of penetration
no minimization procedures are required). are also shown in the figure. The model predicted a plug velocity of
TagedPAt some point, the bulge no longer expands radially (the end of 424 m/s and a bullet exit velocity of 364 m/s. The experimental exit
Stage II), and instead, the bulge advances in the direction of penetra- velocity was 345 § 45 m/s.
tion, Stage III, Fig. 19. The geometry is determined by the plastic TagedPRavid, Bodner, and Holcman [43] extended the concept of
flow incompressibility condition. dynamic plastic flow fields to high velocity impact (greater than
TagedPThe target can possibly fail during Stages II or III. Ravid and Bod- 1 km/s). The analysis considered impact shock generation, plastic
ner considered a number of possible failure mechanisms, Stage IV. In flow of the target, and mushrooming of the projectile for normal
their approach, it is possible to estimate the strains in the zones and impact by a rod. The analysis continues until the shock effects in
between zone boundaries, which allows them to consider the forma- the rod become relatively unimportant (a few microseconds after
tion of adiabatic shear bands, brittle failure in shear, conventional impact). This model would be the first stage of a postulated
plug formation, and ductile failure. The length and shape of the plug seven-stage model, which was never fully developed.
C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26 17

Fig. 17. Ravid-Bodner penetraton model, Stage I (from Ref. [42]).

Fig. 18. Bulge formation, Stage II (from Ref. [42]).

18 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26

Fig. 19. Bulge advancement, Stage III (from Ref. [42]).

Z Z C1
9. Walker
Anderson model TagedP rp
zi
@uz @uz 1
 1
dz C rt dz C rp u2 ¡v2 ¡ rt u2
zp @t zi @t 2 2
Z C1 ð49bÞ
9.1. Semi-infinite targets @s xz
¡2 dz D 0
zp @x
TagedPThe momentum equation along the projectile-target centerline is
where zp and zi denote the positions of the rear of the projectile and
given by:
the projectile-target interface, respectively; the target is assumed to
@uz 1 @ðuz Þ2 @s zz @s xz be semi-infinite in the original model. To integrate Eq. (49b) explic-
r C r ¡ ¡2 D0 ð48Þ
@t 2 @z @z @x itly, three major assumptions were invoked: (1) the velocity profile
Walker and Anderson integrated this equation along the projectile- along the centerline in the projectile and target are specified (so that
target centerline [44]: the first two integrals can be integrated); 2) the rear of the projectile
Z z1 Z C1 is decelerated by elastic waves (similar to Tate); and 3) the shear
@uz @uz
rp dz C rt dz behavior of the target material is specified (so that the gradient of
zp @t zi @t
  C1 C 1 Z C1 ð49aÞ the shear stress can be evaluated).
1  zi 1   @s xz TagedPThe velocity profiles were motivated by the results of numerical
C rp u2z  C rt u2z  ¡ s zz  ¡2 dz D 0
2 zp 2 zi zp zp @x simulations [29]. The velocity profile in the projectile is assumed to

Fig. 20. Example of the Ravid-Bodner model: a 7.62-mm AP bullet impacting a 12-mm-thick steel plate at 855 m/s.
 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26 19

Fig. 21. Centerline velocity profiles of Walker-Anderson model [44].

TagedPbe bi-linear, see Fig. 21. The velocity field in the target is derived TagedPwhere it is assumed that the decelerating elastic waves run between
from a vector potential A: the tail and the plastic zone in the projectile (a material nonlinearity)
(  ) at a bar wave speed of c, (Ep /rp)1/2. Lastly, the length of the projectile
ur ac Rc 2
Aðr; uÞ D
2  ¡1 sinu ð50Þ shortens with time:
2 ac ¡1 r
L_ D ¡ðv¡uÞ ð56Þ
1 @ 1@ TagedPEqs. (54)
(56) are solved simultaneously with an equation that
v D r £Aðr; uÞ b
ef D ðA sinuÞb
er ¡ ðrAÞb
eu ð51Þ
r sinu @u r @r provides for the extent of the plastic flow field (ac), which is esti-
where Rc is the radius of the penetration channel. Thus, mated by dynamic compressible cylindrical cavity expansion [44]:
 2 2     
vr D 2
u ac Rc
¡1 cosu ð52aÞ r u2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi r u2 a2c qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2ffi
ac ¡1 r 2 1C t Kt ¡rt u2 a2c D 1 C t Kt ¡rt u ð57aÞ
Yt 2Gt
u  
vu D sinu ð52bÞ u
a2c ¡1 K t D Ko 1 C k ð57bÞ
co
TagedPThe velocity in the z-direction is given by: TagedPCompressibility of the target is reflected in ac. Eq. (57b) is an heu-
ristic expression to represent the stiffening of the bulk modulus with
vz D vr cos u¡vu sin u ð53Þ
increasing penetration velocity [44], where the shock velocity is
which gives the expression for the velocity field in the target, related to the particle (penetration) velocity by us D co C ku, and
along the centerline, in Fig. 21. The velocity profiles are shown Ko D roco2. As the penetration velocity increases, ac decreases as
in Fig. 21, along with a comparison to results of a numerical sim- shown in Fig. 22. This prediction was confirmed by numerical simu-
ulation. In the figure, the symbol s represents the extent of plas- lations in Ref. [14]. As already stated, ac is dependent on the penetra-
tic flow in the projectile (from the projectile-target interface); tion velocity, and not the initial impact velocity. Therefore, ac
and acRc is the extent of the plastic flow field in the target. changes as the penetration velocity changes. The elastic energy
The specific comparison is for a tungsten-alloy projectile that stored in compression is extremely important as describe in Section
impacted a steel target at 1.5 km/s (the projectiles has deceler- 4.2; ac reflects this compression. For example, the dashed curve in
ated to approximately 1.4 km/s at the time that the velocity pro- Fig. 22 represents ac for rigid-body penetration into aluminum.
file was plotted). As the penetration velocity decreases, ac becomes larger, effectively
TagedPAssuming a von Mises yield surface and that the shear stress is accounting for less energy being stored in elastic compression, and
proportional to the rate of deformation (rigid plasticity), the integral more energy being directly dissipated in target plastic work.
of the gradient of the shear stress gives ¡ 76 Yt lnðac Þ [44]. Taking the TagedPFor eroding long-rod penetration, the penetration velocity is
time derivatives of the velocity profiles in Fig. 21, and assembling approximately constant during the primary phase of penetration, so
the various terms in Eq. (49b), the centerline momentum balance ac is approximately a constant. It then increases during the terminal
becomes: phase when there is rapid deceleration of the remaining projectile.
 TagedPIt is interesting to note that Chocron, Anderson, and Walker [45]
ac ¡1 d v¡u
s2 2Rc u
_
rp vðL¡sÞ C u_ rp s C rt Rc C rp C rt a_ c compared the results for compressible cylindrical cavity expansion
ac C 1 dt s 2 ðac C 1Þ2
 ð54Þ and compressible spherical cavity expansion to compute ac. They
1 1 7 found that the use of the compressible cylindrical model, using val-
D rp ðv¡uÞ2 ¡ rt u2 C Yt ln ac
2 2 3 ues for the material flow stresses consistent with the constitutive
where the dot over a symbol implies the time derivative. Decelera- behavior of the materials, gave better agreement with experimental
tion of the rear of the projectile is given by: data at impact velocities below 2 km/s. (Cavity expansion is used to
 provide an estimate of the extent of plastic flow in the target. This is
sp v¡u s_
v_ D ¡ 1C C ð55Þ in contrast to the discussion in Section 7 where cavity expansion
rp ðL¡sÞ c c is used to calculate directly the force on the projectile. The
20 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26

Fig. 22. ac as a function of penetration velocity for steel and aluminum targets.

TagedPhemispherical flow field in the Walker-Anderson model is a good TagedP iscrepancy in the current model, the Walker-Anderson model pro-
d
approximation of the applicable mechanics. Any error is estimating vides a reasonable estimate of the total depth of penetration for
the extent of plastic zone in the target using cavity expansion is L/D D 1 projectiles (as shown in Ref. [44]), probably due to the fact
somewhat mitigated by the fact it enters the Walker-Anderson that the target mechanics (the plastic flow zone) is being treated
model as ln(ac).) correctly combined with the robustness of the conservation of
TagedPThe Walker-Anderson model uses material properties for the momentum approach.
projectile and target; the only empirical input is the crater radius TagedPIn the limit where s!0, Young's modulus of the projectile
Rc as a function of impact velocity [44]. Examples of the model, becomes large (c!1), and the three-dimensional terms are
compared to experimental data, are shown in Fig. 23; other exam- removed (Rc!0), Eqs. (54)
(56) reduce to Tate's model, Eqs. (28),
ples are provided in Ref. [44]. Although comparison with experi- (30), and (31), but with an explicit expression for Rt:
mental data is excellent, there is a discrepancy for eroding 7
penetration that has not been resolved. As already discussed, the Rt D Yt lnðac Þ ð58Þ
3
projectile is decelerated by elastic waves. However, when the
TagedPIt is noted that s p in Eq. (55) is the dynamic flow stress of the pro-
length of the projectile has decreased to approximately one projec-
jectile material; whereas, Tate heuristically accounted for dynamic
tile diameter, the projectile material within the plastically deform-
effects by adjusting the parameter λ in Eq. (38). More will be dis-
ing region, and the material is “sees” the high pressures associated
cussed about the appropriate value for Yt in Section 9.3.
with the penetration front (for example, see Ref. [47]). This high
pressure decelerates the remaining projectile sooner than calcu-
lated by the analytical model, as shown in Fig. 24 where the nose 9.2. Finite targets
and tail velocities are plotted versus time. The result for the Tate
model is also shown for comparison. For this specific example, TagedPInitially, the Walker-Anderson model was extended to finite-
the Walker-Anderson model, the Tate model, and the simulation thick targets by combining the failure modes of the Ravid-Bodner
have the same total depth of penetration. Notwithstanding this model with the Walker-Anderson model [48]. In this approach,

Fig. 23. Results for Walker-Anderson model for L/D 10 tungsten-alloy and steel projectiles penetrating two different armor steels. Data are from Ref. [50].
 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26 21

Fig. 24. Comparison of nose and tail velocities vs. time for an L/D D 10 tungsten-alloy projectile into an armor steel target at an impact velocity of 1.5 km/s.

TagedPthe Walker-Anderson model is considered operative until the

3
~ :
ðTagedP 2=3ÞpR
presence of the target rear surface begins to influence projectile  2  3
3 h 1 h
penetration, which was assumed to occur when the plastic zone bD ¡ ð60Þ
2 R~ ~
2 R
reached the effected target rear surface covering the projected
area of the projectile (pR2). The modified model predicted bulg- In practice, h is the distance between the projectile nose and the tar-
ing, failure of the target, and an estimate of the residual velocity get rear surface, and R ~ D ðac ¡1ÞRc . b equals 1 until the plastic zone
as a function of impact velocity, including the ballistic limit reaches the rear surface of the target. As the distance between the
velocity. projectile nose and the target rear surface decreases, b decreases
TagedPThen, in 1999, Walker extended the Walker-Anderson model to until it reaches zero at h D 0, and the velocity field is pure shear.
include explicitly bulging [49] and failure [50]. When the flow field TagedPAfter considerable math, and a few minor approximations for
reaches the rear surface of the target, the flow field is achieved simplification, modifications for the back surface is effected by
through a multiplicative blending of the potentials for hemispherical replacing the last terms in Eq. (54):
flow (Adeep), Eq. (50), and shear flow (Ashear):  
1 7 1 7
b 1¡b
¡ rt u2 C Yt lnðac Þ by ¡ rt ðu¡uback Þ2 C Yt lnða~ c Þ ð61Þ
Aðr; uÞ D Adeep Ashear b
ef ð59Þ 2 3 2 3
where a ~ c D 1 C (distance to back surface)/Rc. The change in the tar-
where b determines how much of the potential is deep versus shear. get resistance term reflects the fact that there is less target material
The velocity flow fields are found by taking the curl of Eq. (59), as being plastically deformed. The velocity of the rear (back) surface is
was done for Eqs. (50) and (51). Details are provided in Ref. [49]. The obtained explicitly from the flow fields, and integrating the velocity
volume of the material in the truncated hemisphere with radius R ~ over time provides the position of the rear surface. A comparison of
~
and height h is V D ph2 ðR¡h=3Þ [51]. The weighting is determined by the Walker-Anderson model with bulging to a numerical simulation
dividing this volume by the original volume of the hemisphere, is shown in Fig. 25.

Fig. 25. Comparison of Walker-Anderson model with bulging: red lines denote model predictions superimposed on a numerical simulation.
22 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26

TagedPSince the flow field is known, it is possible to evaluate the equiva- TagedPThe model is compared to two experiments in Fig. 26. Both
lent plastic strain rate at any point of the target: experiments are near the V50 limit for the specific target thickness (a
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   tungsten-alloy projectile into an armor steel). For the thinner target,
2 1 @ui @uj
ɛ_ p D Dij Dij Dij D C ð62Þ the impact velocity is just below V50 of 1250 m/s. The projectile has
3 2 @xj @xi perforated the thicker target at the V50 impact velocity. Penetration,
where Dij is the strain rate tensor, and the ui’s and xj’s are the velocity bulging, and perforation are well predicted by the analytical model.
and coordinate directions, respectively. However, Eq. (62) is difficult TagedPIt was found that a critical strain to failure calculated in this man-
to evaluate for a general location within the target. It is straightfor- ner permits accurate predictions of residual velocities for relatively
ward, however, to evaluate the equivalent plastic strain rate along ductile targets. However, this failure criterion is too simplistic for
the projectile-target centerline where the expression for the flow very hard targets that fail by brittle shear or strain localization [50].
field is algebraically simple [50]:
9.3. Similitude analysis revisited and the effective flow stress
 
4 R c 2λ b u
ɛ_ p D ð63Þ
3 r r TagedPNormalized penetration as a function of impact velocity is shown
for five sets of data in Fig. 27 [46]. The projectiles were flat-nosed
Integration over time provides the equivalent plastic strain. with an aspect ratio (L/D) of 10. There is a tungsten alloy (WA) pro-
TagedPThe strains were calculated at specified locations along the back jectile into an armor steel, a maraging (W8) steel, and a mild steel
surface [50], and it was found that the maximum equivalent plastic (Steel 52). There is also a moderately strong steel projectile into a
strain occurs at a distance of approximately two projectile radii from mild steel (Steel 37/52) and into an armor steel. A nonlinear regres-
the centerline, and not on the centerline. Nevertheless, the simplest sion was used to fit each individual set of experimental data to an
failure criterion that can be posed is to invoke target failure (that is, equation with the form:
perforation) when the centerline strain at the rear surface exceeds
P a2
some critical strain, ecrit. It is necessary to use an experimental data D a1 C ð64Þ
point to determine ecrit, such as the ballistic limit velocity. Once L 1 C 10a3 ða4 ¡VÞ
determined, then ecrit is considered to be a constant for other targets where the ai’s are determined from the regression analysis; the solid
that use the same material. lines represent the curve fits [53].

Fig. 26. Bulging and failure of the Walker-Anderson model. Data from Ref. [52].

Fig. 27. Normalized penetration for five sets of projectile-target combinations as a function of impact velocity. Data from Ref. [50].
 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26 23

Fig. 28. Normalized penetration vs. a nondimensional impact velocity [54].

Fig. 29. Root mean square error as s t is varied for the steel projectile steel 37/52 target.

TagedPThe data are replotted in Fig. 28 using nondimensional variables TagedPThe regression coefficients are shown in Fig. 28.
[54]. The ratio of the projectile to target densities (a) is used to TagedPKey to this analysis is the selection of the effective flow stress,
account for density differences in normalized penetration. A nondi- s efs. s efs was determined by conducting a parametric study on P/L
mensional velocity is defined, rpV 2/s efs, where s efs is the effective as a function of velocity for various values of s efs. The valued
flow stress. rpV 2 is proportional to the penetration pressure, and selected for s efs was the one that minimized the root mean square
s efs is related to the strength of the target. Thus, the nondimensional error on P/L, see Fig. 29 for an example [53]. This was done for each
velocity is a ratio of penetration pressure to target resistance. Simili- of the experimental data sets. s efs for Steel 37, Steel 52, and
tude theory does not specify the relationship between the nondi- German armor steel were found to be 0.797 GPa, 0.858 GPa, and
mensional variables, and it was found that better correlation was 1.41 GPa, respectively.
obtained by also including the ratio of the densities on the abscissa, TagedPThe target flow stress used in the Walker-Anderson model (Yt),
raised to a small power. The exponents were determined by adjust- which is a single value, must somehow encompass the volumetric
ing them to provide a relatively tightly collapsed data set.9 effect of the target flow stress that is a function of strain, strain rate,
TagedPThe form of the equation used to fit the collapsed data is similar and temperature; and which varies with time and location in the tar-
to that of Eq. (64), but now it is P/aL as a function of V [54]: get. Adiabatic stress strain curves,10 using the Johnson-Cook consti-
!0:2 tutive model [55], are shown for two materials at various strain
P a2 rp V 2 qffiffiffiffiffiffiffiffiffiffiffiffiffi
D a1 C VD a¡0:14 a D rp =rt rates in Fig. 30: a mild steel (A36) and 4340 steel hardened to Rock-
aL 1 C 10½ 3 ð 4 Þ
a  a ¡V s efs well C 30. A36 is similar to the German mild steels 37 and 52. The
ð65Þ long horizontal dashed lines represent s efs for Steel 37 and 52, deter-
mined by the minimization procedure. The dashed-dotted line
represents an average equivalent stress for A36 with some strain
9
It is interesting to note that the strength of the projectile is not included in the
10
analysis. This is because the penetration pressures greatly exceeds the flow stress of Plastic work is converted to temperature, which then results in thermal
the projectile material for these projectile-target combinations and impact velocities. softening.
24 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26

Fig. 30. Equivalent stress vs. equivalent plastic strain for two materials.

TagedPhardening; this dashed-dotted line is only 8% lower than the calcu- tTagedP hat they are supposed to predict. Specifically, Backman and
lated s efs for Steel 37. Goldsmith state that the next two topics can contribute to less
TagedP4340 steel, hardened to Rc30, is often used as an experimental empirical procedures. I will comment after item number 3.
surrogate for armor steel. The average equivalent stress at a strain TagedP2 The improved description of interface phenomena (e.g., the forma-
rate of 103 s¡1 is 1.2 GPa. This is the value used for Yt for one of the tion of the penetration channel, interface phenomena involving
steels in Fig. 23. s efs for German armor steel was determined to oblique impact, nonzero flight orientations, deforming projec-
be 1.4 GPa, which is the value for Yt used for the stronger steel in tiles).
Fig. 23. Thus, Yt in the Walker-Anderson model and s efs are equiva- T The improved analytical modeling of force contributions (a need to
agedP3
lent; Yt, i.e., the effective flow stress, encompasses an averaged volu- extend modeling to thicker targets, and to include dishing and
metric flow stress for the target material. And it is seen that an bulging of target elements, plug displacement, etc.).
estimate for Yt can be made from a stress-strain curve.
TagedPIn fact, Walker developed a procedure for estimating the target TagedP In general, the models described above do not use empirical fits
flow stress [56] using Johnson-Cook constitutive parameters. The for model parameters; instead, procedures have been developed
strain field in the target can be calculated, and the plastic strain rate to estimate model parameters from independent laboratory
can be determined directly from the velocity field. If no thermal soft- measurements, such as density, flow stress, etc. A major
ening is present, the resulting equations can be integrated explicitly; advance, particularly for penetration and perforation of thick
if thermal softening is taken into account, then the target resistance targets, was the inclusion of dynamic plasticity, i.e., the descrip-
term in Eq. (58) must be solved in the presence of thermal softening tion of the plastic flow field(s). Thus, considerable progress has
[56]. been made in these three areas for normal impact of an
unyawed projectile.
10. Summary
TagedP4 Effects of obliquity. Typically, the line-of-sight thickness is used
T his article has reviewed analytical penetration models for
agedPT to account for target obliquity. This is a reasonable approxima-
metallic targets. The models range in applicability from rela- tion. However, at large obliquities, particularly for thick targets,
tively low impact velocities (rigid-body penetration) to very the presence of the free surface results in asymmetric loading
high-velocity impacts (hydrodynamic approximation). It was on the projectile nose, which can lead to bending and possibly
seen that the oldest of the models, represented by the Poncelet fracture of the projectile. Although this can be modeled quite
equation, Eq. (1), is derived naturally from the Tate model. The nicely in three-dimensional numerical simulations, this is cur-
Poncelet equation, where the force is proportional to a static rently beyond the ability of first-principles analytical models.
term and a velocity-squared term, was also derived from calcu- TagedP5 Effects of flight orientation of the projectile. It is well-known that
lating the decelerating force on the projectile using dynamic yaw and pitch can dramatically affect penetration response,
cavity expansion. which is amplified by target obliquity, as demonstrated by
TagedPThe objective of all of these models is to replicate experimental Anderson, Behner, and Hohler [57]. Although analytical model-
data using real material properties while minimizing empirical fac- ing of some of the initial conditions can be performed, the com-
tors. Exploring the applicability and accuracy of the models has led plete projectile-target interaction is probably not analytically
to insights as well as limitations. tractable. However, approximations can be made. An example is
TagedPBackman and Goldsmith [1] highlighted eight areas that merited a paper by Walker, Anderson, and Goodlin [58] that combines
further investigations to improve modeling of penetration mechan- several analytical expressions (based on experimental data) for
ics. It is instructive to comment on these to see what progress has predicting the depth of penetration as a function of velocity, L/D
been made, and where there remain outstanding issues and require- ratio, and impact inclination.
ments for further research. TagedP6 Improved constitutive representations. Forrestal, Warren, and col-
leagues explicitly use the stress-strain behavior of the target
TagedP The development of more rational bases of specifying modeling
1 material, including work hardening and strain-rate effects, in
parameters. Model parameters should be specified beforehand, their modeling. In the Tate and Walker-Anderson models, a sin-
i.e., they should not be determined from the experimental data gle value for the projectile flow stress, including strain-rate
 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26 25

TagedPeffects, accurately predicts deceleration of the projectile. For the ITagedP n: Galvez F, Sanchez-Galvez V, editors. In: Proc. 23rd Int. symp. ballistics. . Gra
fi-
Walker-Anderson model, a single value for the target flow stress cas Couche; 2007. p. 1437–52.
TagedP [9] Birkhoff G, MacDougall DP, Pugh EM, Taylor G. Explosives with lined cavities. J
is used, but it was shown that this value is related to a volumet- Appl Phys 1948;19:563–82.
ric average of the target stress that includes work hardening, TagedP[10] Eichelberger RJ. Experimental test of the theory of penetration by metallic jets. J
strain-rate effects, and thermal softening. Backman and Gold- Appl Phys 1956;27(1):63–8.
TagedP[11] Allen WA, Rogers JW. Penetration of a rod into a semi-infinite target. J Franklin
smith's comment also had to do with materials other than met- Inst 1961;272:275–84.
als. The focus of this article was metallic targets; ceramics, TagedP[12] Christman DR, Gehring JW. Analysis of high-velocity projectile penetration
glasses, fabrics, composites, etc., are beyond the scope of this mechanics. J Appl Phys 1966;37(4):1579–87.
TagedP[13] Recht RF, Ipson TW. Ballistic perforation dynamics. J Appl Mech 1963;30:384–
article.
90.
TagedP7 Thermo-mechanical coupling and heat dissipation. Aside from the TagedP[14] Anderson Jr. CE, Littlefield DL, Walker JD. Long-rod penetration, target resistance,
concept of the effective flow stress in the Walker-Anderson and hypervelocity impact. Int J Impact Eng 1993;14:1–14.
TagedP[15] Walker JD. Hypervelocity penetration modeling: momentum vs. energy and
model, thermo-mechanical coupling, including localization, is
energy transfer mechanisms. Int J Impact Eng 2001;26:809–22.
probably beyond the capability of analytical models without TagedP[16] W.E. Baker, P.S. Westine, and F.T. Dodge, Similarity methods in engineering
some empiricism. However, thermo-mechanical coupling is dynamics: theory and practice of scale modeling, Spartan Books, Rochell Park, NJ,
readily modeled in numerical simulations, although the issue of 1973.
TagedP[17] Anderson Jr. CE, Hohler V, Walker JD, Stilp AJ. The influence of projectile hard-
different length scales of various phenomena remains an issue. ness on ballistic performance. Int J Impact Eng 1999;22(6):619–32.
T Failure criteria. Prediction of failure is the really tough problem.
agedP8 TagedP[18] Zukas JA, Nicholas R, Swift HF, Greszczuk LB, Curran DR. Impact dynamics, chap-
See the paragraph below. ter 5. NY, NY: John Wiley & Sons, Inc; 1982.
TagedP[19] Tate A. A theory for the deceleration of long rods after impact. J Mech Phys Solids
1967;15:387–99.
TagedP[20] Alekseevskii VP. Penetration of a rod into a target at high velocity. Fizika Gore-
T nresolved at this time are accurate predictions for the transition
agedPU niya I Vzryva 1966;2(2):99–106 translated Combustion, Explosion, and Shock
from rigid-body to eroding penetration, which depends upon projec- Waves, pp. 63-66, July 1965.
TagedP[21] Tate A. Further results in the theory of long rod penetration. J Mech Phys Solids
tile material properties, projectile nose shape, and target strength,
1969;17:141–50.
although some interesting work has been done by Segletes [59]. TagedP[22] Forrestal MJ, Okajima JK, Luk VK. Penetration of 6061-T651 aluminum targets
Another area of uncertainty is target failure, which depends upon with rigid long rods. J Appl Mech 1988;55:755–60.
TagedP[23] Forrestal MJ, Brar NS, Luk VK. Penetration of strain-hardening targets with rigid
the target material and loading conditions. Thus, it is failure
spherical-nose projectiles. J Appl Mech 1991;58:7–10.
(whether projectile and/or target) that is difficult, if not impossible, TagedP[24] Piekutowski AJ, Forrestal MJ, Poormon KL, Warren TL. Penetration of 6061-T6511
to predict a priori. This is also true of numerical simulations. It may aluminum targets by ogive-nose steel projectiles with striking velocities
be that some empiricism will always be required to calibrate a between 0.5 and 3.0 km/s. Int J Impact Eng 1999;23:723–34.
TagedP[25] Forrestal MJ, Piekutowski AJ. Penetration experiments with 6061-T6511 alumi-
model for accurate predictions. num targets and spherical-nose steel projectiles at striking velocities between
TagedPAnother area that requires further work is multi-material layered 0.5 and 3.0 km/s. Int J Impact Eng 2000;24(1):57–67.
targets. The stress field precedes the penetration front, thus, the TagedP[26] Tate A. Long rod penetration models—Part I. A flow field model for high speed
long rod penetration. Int J Mech Sci 1986;28(8):535–48.
next layer will affect the rate of penetration. This effect will increase TagedP[27] Tate A. Long rod penetration models—Part II. Extensions to the hydrodynamic
as the difference in material properties of the two materials theory of penetration. Int J Mech Sci 1986;28(9):599–612.
increases. This is true for a metallic layered target, and the effect TagedP[28] Bishop RF, Hill R, Mott NF. The theory of indentation and hardness. Proc R Soc
1945;57(3):147–59.
may become even larger when other materials such as ceramics, TagedP[29] Anderson Jr. CE, James, Walker D. An examination of long-rod penetration. Int J
fabrics, or composites are used. Impact Eng 1991;11(4):481–501.
TagedPIt was observed that when models are based on conservation of TagedP[30] Hopkins HG. Dynamic expansion of spherical cavities in metals. In: Sneddon I,
Hill R, editors. Progress in solid mechanics, 1. NY: North-Holland; 1960. p. 85–
momentum, such as the Tate and Walker-Anderson models, they
164.
tend to be robust and applicable to a wide range of materials and TagedP[31] Luk VK, Forrestal MJ, Amos DE. Dynamic spherical cavity expansion of strain-
impact velocities. Models based on conservation of energy typically hardening materials. J Appl Mech 1991;58:1–6.
TagedP[32] Warren TL, Forrestal MJ. Effects of strain hardening and strain-rate sensitivity on
have limited applicability because energy sinks/dissipation change
the penetration of aluminum targets with spherical-nosed rods. Int J Solids
with impact velocity and materials. Struct 1998;35(28-29):3737–53.
TagedPThis article focused on analytical models for metallic targets. An TagedP[33] Hill R. Cavitation and the influence of headshape in attack of thick targets by
article is planned on reviewing analytical models for non-metallic non-deforming projectiles. J Mech Phys Solids 1980;28:249–63.
TagedP[34] Tate A. A theoretical estimate of temperature effects during rod penetration. In:
targets, such as glasses, ceramics, yarns and composites. Proc. 9th Int. symp. ballistics, 2; 1986. p. 307–14.
TagedP[35] M. Wilkins, personal communication.
Acknowledgements TagedP[36] Wilkins ML. Calculations of elastic-plastic flow. In: Adler B, Fernback S,
Rotenberg M, editors. Methods of computational physics, 3. NY, NY: Academic
Press; 1964.
TagedPThe author is very grateful for the interactions over the years TagedP[37] Camacho GT, Ortiz M. Adaptive Lagrangian modeling of ballistic penetration of
with most (but not all) of the authors referenced in this article. metallic targets. Comput Methods Appl Mech Eng 1997;142:269–301.
TagedP[38] Warren TL. The effect of target inertia on the penetration of aluminum targets by
rigid ogive-nosed long rods. Int J Impact Eng 2016;91:6–13.
References TagedP[39] Rosenberg Z, Dekel E. “A comment on ‘The effect of target inertia on the penetra-
tion of aluminum targets by rigid ogive-nosed long rods’ by T. L. Warren,” Letter
TagedP [1] Backman ME, Goldsmith W. The mechanics of penetration of projectiles into tar- to the Editor. Int J Impact Eng 2016;93:231–3.
gets. Int J Eng Sci 1978;16(1):1–100. TagedP[40] Rubin MB, Kositski R, Rosenberg Z. Essential physics of target inertia in penetra-
TagedP [2] Jonas GH, Zukas JA. Mechanics of penetration: analysis and experiment. Int J Eng tion problems missed by cavity expansion models. Int J Impact Eng 2016;98:97–
Sci 1978;16(11):879–904. 104.
TagedP [3] Int J Eng Sci 1978;16(11):793–920. TagedP[41] Warren TL. Response to: ‘A comment on ‘The effect of target inertia on the
TagedP [4] Anderson Jr CE, Bodner SR. Ballistic impact: the status of analytical and numeri- penetration of aluminum targets by rigid ogive-nosed long rods by T. L. Warren’
cal modeling. Int J Impact Eng 1988;7(1):9–35. by Z. Rosenberg and E. Dekel,” Letter to the Editor. Int J Impact Eng 2016;93:
TagedP [5] Goldsmith W. Non-ideal projectile impact on targets. Int J Impact Eng 1999;22 234–5.
(2-3):95–395. TagedP[42] Ravid M, Bodner SR. Dynamic perforation of viscoplastic plates by rigid projec-
TagedP [6] Forrestal MJ, Piekutowski AJ. Penetration experiments with 6061-T6511 alumi- tiles. Int J Eng Sci 1983;21(6):577–91.
num targets and spherical-nose steel projectiles at striking velocities between TagedP[43] Ravid M, Bodner SR, Holcman I. Analysis of very high speed impact. Int J Eng Sci
0.5 and 3 km/s. Int J Impact Eng 2000;24:57–67. 1987;25(4):473–82.
TagedP [7] Stilp AJ, Hohler V. Long rod penetration mechanics Chapter 5. In: Zukas JA, edi- TagedP[44] Walker JD, Anderson Jr. CE. A time-dependent model for long-rod penetration.
tor. High velocity impact dynamics. NY, NY: John Wiley & Sons; 1990. Int J Impact Eng 1995;16(1):19–48.
TagedP [8] Wickert M. Penetration data for a medium caliber tungsten sinter alloy penetra- TagedP[45] Chocron S, Anderson Jr. CE, Walker JD. Long-rod penetration: cylindrical vs.
tor into aluminum alloy 7020 in the velocity regime from 250 m/s to 1900 m/s. spherical cavity expansion for the extent of plastic flow. In: Van Niekerk C,
26 C.E. Anderson Jr. / International Journal of Impact Engineering 108 (2017) 3
26

TagedPeditor. In: Proc 17th symp. ballistics. . South African Ballistics Organisation; TagedP[52] Anderson Jr. CE, Hohler V, Walker JD, Stilp AJ. Time-resolved penetration of long
1998. p. 319–26. rods into steel targets. Int J Impact Eng 1995;16(1):1–18.
TagedP[46] Anderson Jr. CE, Orphal DL. Analysis of the terminal phase of penetration. Int J TagedP[53] Riegel III JP, Anderson Jr. CE. Target effective flow stress calibrated using the
Impact Eng 2003;29:69–80. Walker-Anderson penetration model. In: Proc. 28th Int. symp. ballistics. DESTech
TagedP[47] Ravid M, Bodner SR, Walker JD, Chocron S, Anderson Jr. CE, Riegel III JP. Modifica- Publications, Inc.; 2014. p. 1242–53.
tion of the Walker-Anderson penetration model to include exit failure modes TagedP[54] Anderson Jr. CE, Riegel III JP. A penetration model based on experimental data.
and fragmentation. In: Van Niekerk C, editor. In: Proc. 17th Int. symp. ballistics. . Int J Impact Eng 2015;80:24–35.
South African Ballistics Organisation; 1998. p. 267–74. TagedP[55] Johnson GR, Cook WH. A constitutive model and data for metals subjected to
TagedP[48] Walker JD. An analytic velocity field for back surface bulging. In: Reinecke WG, large strain, high strain rates, and high temperatures. In: Proc. 7th Int. symp. bal-
editor. Proc. 18th Int. symp. ballistics. Lancaster, PA: Technomic Publishing Com- listics; 1983. p. 541–8. 19-23.
pany; 1999. p. 1239–46. TagedP[56] Walker JD, Chocron S, Anderson Jr. CE, Riegel III JP, Riha DS, McFarland JM, et al.
TagedP[49] Chocron S, Anderson Jr. CE, Walker JD. A consistent flow approach to model pen- Survivability modeling in DARPA's adaptive vehicle make (AVM) program. In:
etration and failure of finite-thickness metallic targets. In: Reinecke WG, editor. Proc. 27th Int. symp. ballistics. . DEStech Publications; 2013. p. 967–79.
Proc. 18th Int. symp. ballistics. Lancaster, PA: Technomic Publishing Company; TagedP[57] Anderson Jr. CE, Behner T, Hohler V. Penetration efficiency as a function of target
1999. p. 761–8. obliquity and projectile pitch. J Appl Mech 2013;80 031801-1/11.
TagedP[50] Hohler V, Stilp AJ. A penetration mechanics database. In: Anderson CE Jr., TagedP[58] Walker JD, Anderson Jr. CE, Goodlin DL. Tungsten into steel penetration includ-
Morris BL, Littlefield DL, editors. SwRI Report 3593/001. San Antonio TX: ing velocity, L/D, and impact inclination effects. In: Crewther IR, editor. In: Proc.
Southwest Research Institute; 1992. 19th Int. ballistics symp, 3; 2001. p. 1133–40.
TagedP[51] Standard mathematical tables. 20th ed. (S. M. Selby, Ed-in-Chief ) CRC Press; TagedP[59] Segletes SB. The erosion transition of tungsten-alloy long rods into aluminum
1972. targets. Int J Impact Eng 2007;44:2168–91.