AD-A140 695 COMPUTERSIMULATION FOR THE TRANSIENT

DYNAMIC RESPONSE f/
OF LOCKNGMAERAS(U GEORGIA NS OF TECHA ANTA
S HANAGUONOV82 0 TE A -638 AMMRC-TR-82-60

UCAED hAG6hE-C0hhEFG 1/EII

mohhhEEmhhhmhE
*uuuuuuuwmmo
 136.

• .I,,.,$.-
1.25 11..- J

ZIA&
. .. . . ... . I' . ..
 [AD
I .

AD-A1 4 0 695

AMMRC TR 82-60

COMPUTER SIMULATION FOR THE TRANSIENT

DYNAMIC RESPONSE OF LOCKING MATERIALS

November 1982

S. HANAGUD
Georgia Institute of Technology
School of Aerospace Engineering
Atlanta, Georgia 30332

FINAL REPORT Contract Number DAAG46-79-C-0O14

B DTIC
LL Approved for public mlearns; distribution unlimited. ELECTE

U
S MAYZ2B 84
Prepared for
ARMY MATERIALS AND MECHANICS RESEARCH CENTER
Watertown, Massachusetts 02172

84 04 30 068
 UNCLASSIFIED
SECURITY CLAMPICATION OF TbIS PAGE (Who. Data En#e,ed)

REPORT DOCUMENTATION PAGE COMUCnTIoNs

READCOMPLETING FORM
BEFORE
1.RPOTNM 12. 4OVT ACCESSION NO. S. RECIPIENTS CATALOG HUNGER

AMMRC TR 82-60
(and SAAtf9e)
.TITLE
- &f
5. TYPE OF REPORT a PE-OO COVERED

Feb Reor ET R
COMPUTER SIMULATION FOR THE TRANSIENT
C.
DYNAMIC RESPONSE OF LOCKING MATERIALS 6. PERFORMING ONG. REPORT NUMBER
E-16-638
7. AUT"Oft~s) U. CONTRqACT ON GRANT NOUMBENs)

SS. HngdDAAG46-79-C-0014

9. PEAVoRMNG ORGANZATON'.NAME ANDADDESS 10 PROGRAM

ARA
ErLEMENT.PROJECT. TASK
& WORK UNIT NUMUERS

Georgia Institute of Technology D/A Project: 1Ll61102AH42

School of Aerospace Engineering AM(S Code: 611102.H420011
Atlanta, Georgia 30332
II. CONTROLLING OFFICE NAME ANO AOORESS I2. REPORT DATE

Army Materials and Mechanics Research Center November 1982

ATT: DRXMR-K 13. NUMBEROF PAGES
atertown. Massachusetts 02172
14. MONITORING AGENCY NAME a AOORESS(I dhifemut frmm coiitulaIng
Offers) IS. SECURITY CLASS. (of thl report)

Unclass .J
I. OECLASSIFICATION 'OOWNGRAOING
SCHEOULE

16. ISTRIUTION STATEMENT (ol this Report)

Approved for public relear.-; distribution unlimited.

17. OISTRIBUTION STATEMENT (of the bstrft entmred In Black 20. it dIfferentfrom Report)

IS. SUPPLEMENTARY NOTES

I KEY WORDS (Cntlawe on reverse side efnrossowy and Identifyby Work numbe.')

Wave propagation
Finite difference theory
.Numerical analysis
Dynamic response
20. AISTRACT (Contiue on revse side Ofneoerse seedlentely by Wo.k maewe)

(SEE REVERSE SIDE)

DD ,o,,1473
Ape 1 7 EDITION, 455 O~o.*,
o NOV ,, SOIPTE UNCLASSIFIED
SECURITY CL OSSI ICA ION OF TNIS PAGE (0ba Dee "see)
 UiNCLSIFIED
StCusMTV CLASMSICArW OP TiyS PASgfWa' -8-t.O

Block No. 20

\'AB$'T"MCT

The concept of a locking material and the potential effectiveness of

the locking material for possible use as a countermeasure to specified
mine blasts have been investigated in this report. The effectiveness of
the countermeasure has been demonstrated by the use of analytical methods
and second order accurate computer codes. Computer codes have been devel-
oped to investigate and evaluate such a design that uses the counter-
measures. These computer codes include programs in one space dimension,
and axisymmetric coordinates.

As a result of the investigation, specific design parameters, candi-

date materials and fields for further study have been identified.

K
UNCLASSIFIED
$1Cub v CL A10PC*ATIUO Or To$$%PAe UMAN0.. I..e..,

-_
I

SUMMARY

The concept of a locking material and the potential effectiveness of the locking
material for possible use as a countermeasure to specified mine blasts have been
investigated in this report. The effectiveness of the countermeasure has been
demonstrated by the use of analytical methods and second order accurate computer
codes. Computer codes have been developed to investigate and evaluate such a design
that uses the countermeasures. These computer codes include programs in one space
dimension, and axisymmetric coordinates.
As a result of the investigation, specific design parameters, candidate materials
and fields for further study have been identified.

AccessiOn For

Jz'j

AvainabilitV Codes__
IAvaU md/or
Dist specl_

• _- ...... ...
 I.I

CONTENTS

Page
I. INTRODUCTION . . . . . . . . . . . . . . . . I

2. RESEARCH PLAN . ............... 2

3. STATE-OF-THE-ART ....... ................ 7

4. ANALYSIS UNDER CONDITIONS

OF ONE-DIMENSIONAL STRAIN ....... ............. 10

3. ELASTIC-PLASTIC-LOCKING
MATERIALS UNDER CONDITIONS OF AXISYMMETRY . . . . . . 16

6. CONCLUSIONS . . . . . . . ............. 29

7. REFERENCES . . . . . . . ........ . . . . 31

M -
 1. INTRODUCTION

This report presents the results, to date, of investigations concerning the

development of a countermeasure to mine blasts. In particular, the investigations have
been conducted to develop computer simulation techniques that are capable of simulating
the transient dynamic response of Locking materials 12or locking material-structure
combinations. These deveoloped techniques have been used to

(a) examine if a locking material shield has the potential of protecting a given
structure from dynamic loads that have a time history similar to that of mine
blasts, and
(b) provide a tool for designing the locking material shield-structure
combination.

Terms such as foams, distended materials 3 or nonreactive porous solids4 ,1have been
used to describe the locking materials. Typically, a locking material is characterized by
a hydrostatic pressure-density curve similar to that shown in Figure 1.1. The model
described in Figure 1.1 is an idealized behavior. In practice, a more realistic model may
be necessary. A material, which follows the model shown in Figure 1.2 behaves like an
elastic solid below a pressure PE' For pressure P> PEI the pore spaces collapse and the
material locks at a density p, . The subsequent behavior of the material is that of an
incompressible material. In practical shock interaction and attenuation calculations by
numerical techniques, the model shown in Figure 1.1 is very often replaced by a pressure
density relationship similar to those shown in Figures 1.2 or 1.3. In Figure 1.3, after
reaching the locking density, the material is assumed to follow a pressure-density
behavior of the corresponding solid. Furthermore, unloading will follow different paths
as shown in the Figures 1.2 and 1.3. Such unloading paths are necessary to reflect the
fact that a collapsed porous space can not be recovered. An exception to this rule is the
3
behavior of graphite foam
The theoretical foundation for expecting a locking material to be an effective peak
stress attenuator or a countermeasure to mine blasts can be summarized as follows.
Materials, that approximate a pressure-density behavior of Figures 1.1, 1.2 or 1.3, imply
a large ratio of unloading or rarefaction wave velocity to the loading or shock velocity.
The large ratio is the result of the fact that the velocity of the loading shockwave Is
related to the slope of a straight line connecting the initial and final states In the
 2

pressure density diagram. The velocity of the unloading wave or rarefaction wave is
related to the slope of the pressure-density curve for the solid at the locked state. The
attenuation is caused by the unloading or rarefaction wave overtaking the loading shock
wave because of the relatively higher velocity of the unloading waves. In practical
situations, the effects of shear stresses and yielding need to be considered.

1.1 Candidate Materials

In principle, locking material can be produced from any parent solid material. The
needed process of production requires the formation of a nonreactive porous solid of a
density lower than that of the parent material. A commerical aluminum locking material
MO-AL (Emerson and Cummings) has been available in the past. This is usually
inhomogeneous. Homogeneous aluminum locking material can be produced by
techniques 3 such as hot pressing of aluminum powders, cold pressing of aluminum
powders followed by sintering or a repeated sequence of hot pressing that is followed by
sintering of aluminum powders and the use of silica microballons. Commerically,
graphite locking materials are available from companies such as National Carbon
Company. Of course, a very common locking material of comparatively low threshold
pressure capability that is easily available is the styro-foam. Depending on the
environment and the peak stresses that need to be attenuated, different types of locking
materials can be chosen or designed for a particular application. For example, in the
case of a tank that is subjected to mine blasts, a locking material made of steel may be
an interesting possibility. In some cases, a combination of different locking materials or
a sandwich construction may be more practical.

1.2 Theoretical Foundation

First, a locking material in a state of one dimensional strain along x-axis is

considered.* A pressure-density relationship as shown in Figure 1.4 has been assumed.
Initially, stresses and velocities in the region x>O, have been assumed to be equal to zero.
At time t = 0, it is assumed that a stress -Ox = Go is applied at the left boundary. The
left boundary is initially at x = x0 (t = 0) = 0 (figure 1.5). In addition to the pressure
density behavior, the behavior in shear is assumed to be elastic-plastic with the following
yield condition.
 3

(o' I " C"2 )2 + (a2 - a 3) 2 +

(a3 "1 ) 2 2Y 2 (1.1)

In this equation, a 1' a 2, 03 are the principal stresses and Y is the yield stress in
simple tension. For the assumed one-dimensional strain conditions, a o x -p relationship
can be derived from the pressure-density relationship and the assumed yield condition.
The resulting ax - p relationship is illustrated in Figure 1.6. The slope of the r x -p curve
for p > p Lis derived from the mechanical behavior of the solid that constitutes the parent
material for the given locking material of initial density po. For example, p o for an
aluminum locking material can be in the range of 1.1 gms/c.c. to 2.3 gms/c.c. The value
of p is approximately equal to 2.7 gms/c.c. In practice, p., is usually slightly below the
value of the solid density of the parent material.
For t>O, the applied stress at the left boundary results in two stress waves
propagating in the positive x-direction. The first wave is an elastic forerunner that
carries the stress discontinuity that corresponds to Y*. The second wave is a shock wave
that carries the stress discontinuity that corresponds to the difference between a and
0
Y*. The shockwave velocity is denoted by k. The following jump conditions are valid
across the elastic forerunner and the shock wave.

Pe(ke v) Pe(kj- v e ) (1.2)

Y Pe(ke -ve)ve + P (k -v )v 0 (1.3)

Pt kL- vL= ekv) . e<,-e)v (1.4)

a7 - i 0e(k V )v (1.5)

In these equations v* and v are the particle velocities ahead of the elastic
forerunner wave and behind the forerunner wave respectively. The velocity of elastic
forerunner wave itself is equal to and vL is the particle velocity behind the shock
wave. From initial conditions, vo = 0. From equations (1.4) and (1.5) the shock wave
velocity ktis given by the following equation.

KC 2

YjK ) Y
 4

Usually, ve is very small and

Pe p0
Then

t P

This is the velocity of the loading shock wave in the locking material. For
example, for aluminum locking material of initial density P0 1.5 gms/c.c.,
Pz= 2.7 gms/c.c., Y = 37,500 psi (258.6 mpa) and a0 = 125,000 psi (862 mpa) the velocity
of the loading shock wave is 3354 feet per second (1022 m/sec) which is approximately
20% of the longitudinal elastic wave velocity of the solid aluminium. These examples
qualitatively illustrate the slow loading velocities in a locking material of low initial
density. However, it is to be noted that the shock wave velocity in a solid is slower than
the correspondifig longitudinal elastic wave velocity. Even then, the shock wave velocity
can be shown to be slower in locking materials than in the corresponding solid.

1.3 Experimental Background

In the past, experimental investigations have been conducted 3 on several locking

materials. In these investigations, metals, plastics, graphite and ceramics have been
used as parent materials to produce locking materials. Experiments have been conducted
.4 by using light gas guns and flyer plates. In these experiments, attempts have been made
to simulate conditions of one dimensional strain. From the paint of view of the present
study, a significant result from these studies concerns a comparison of the shock wave
velocities in locking materials with the corresponding wave velocities in solid materials.
In most of these studies, a two wave pattern has been observed in locking materials
and solids. In locking materials, a forerunner wave carries stresses in the material
before the phase transition from the initial density p 0 to the locked density p.,. The
second slow wave corresponds to the shock wave that was discussed in the previous
section. This shock wave carries stresses that exceed the stresses necessary for phase
transition. Solid materials also exhibit a two wave pattern. In this case, the stresses
carried by the forerunner wave correspond to the yield limit and the elastic behavior of
 5

the solid before yield. This forerunner wave travels at a velocity corresponding to the
longitudinal elastic wave velocity in the isotropic solid. The second wave carries stresses
that exceed the elastic limit. The experiments confirm the fact that this second wave in
the solid is much faster than the second shock wave in the locking material. For
example, in aluminum locking materials of initial density p0 = 2.1 gms/c.c., the second
shock wave velocity varied from 0.7 to 1.29 mm/psec. The second wave velocity, in the
corresponding solid was in the range of 4.62 mm/Usec. Similarly, the first wave
velocities in the locking materials were smaller than the first wave velocities in the
corresponding solid. The first wave velocities, in aluminum locking materials of initial
density p0o = 2.1 gms/c.c., varied in the range 1.6 mm/Isec to 2.0 mm/ 1 isec. The
corresponding first wave in the solid travelled with a velocity of 6.11 mm/1i sec.
The theoretical foundation and the results of the experimental studies confirm the
potential benefits that can be derived from locking materials when they are used as
protective structures.

2. RESEARCH PLAN

2.1 Problems Associated With Engineerin! Design

In the point of view of engineering designs, the results of this project should
provide tools to design a countermeasure or a protective structure for the mine blasts
that will be specified. In this report, however, the protective structure for mine blasts is
assumed to be made of a locking material. Specifically, such a design involves the
selection of a parent material from which the locking material is produced, the thickness
of the protective structure, other geometrical parameters and methods of joining the
countermeasure to the structure. In addition, the selection of an initial density for the
locking material and the specification of the desired microstructure constitute important
aspects of the design. The selected Initial density, the specified production technique
and the resulting microstructure, usually determine the range of pressures at which the
phase transition takes place from the initial density to the locked density. The
microstructure is also responsible tor the elastic-plastic stresb-strain behavior and the
associated yield conditions.
An essen
al design tool for the selection of the various design parameters, is a
compe er sims .tion that is capable of providing the transient dynamic response of the

I. .._ _ - ..
_ ,* - . .
 6

locking material or the locking material and structure combinations. The specified mine
blasts and the various design parameters can be supplied as an input to the resulting
computer program. The output can be obtained as deformations and stresses at various
locations as a function of time. The output data can then be used

(a) to check if the selected design parameters provide the desired margin of
safety for various failure modes,
(b) to iterate or to modify the design parameters and
(c) to establish the damage tolerance and reliability of the design.

The primary objective of the reported investigations is to develop such computer

simulations and the resulting computer programs. To date, most of the reported
investigations have been restricted to one space dimension and time. The results have
been mainly used to simulate the experimental results. The programs have not been used
to study the effectiveness of the locking material as a countermeasure to resist mine
blasts. A study of the transient dynamic response of a locking material that involved two
space dimensions and time has been reported by the author4 . This study described the
impact of compactible plates under axisymmetric conditions. The constitutive
relationship that was used for the compactible plate is very specialized and cannot be
easily generalized for the locking materials of concern here. The one dimensional
analysis and the compactible plate analysis have used the finite difference technique
developed by Von Neuman, 2 '3 and Wilkins 3 . Certain difficulties were encountered in
the application of the Von Neuman's approach to locking materials.

(1) Very small time steps were needed to account for the phase transition from
the initial density to the locked density.
(2) An accurate computer simulation of the precursor wave was very difficult.
(3) Computational efficiency and the level of accuracy decreased when two or
more space dimensions were considered. Similar problems were encountered
in considering multiple reflections.

Other investigators 3 ' 4 also experienced similar difficulties in applying the Von
Neumann's technique to the study of the dynamic response of locking materials and phase
 7

transition. In order to reduce these difficulties and to develop the needed computer
simulation techniques for transient dynamic response of locking materials, the following
research plan was adapted.

(I) The first step was to prepare a state-of-the-art review of finite difference
methods for the numerical solution of hyperbolic differential equations.
(2) The second step was to select a computationally efficient and accurate finite
difference method.
(3) The selected second order accurate method was then applied to study the
transient response of locking material structures under the conditions of one
dimensional strain. The purpose of the analysis was to demonstrate the
capabiltiy of locking material as a countermeasure to mine blasts.
(4) Then, a computer simulation of locking material was developed under
conditions of axisymmetry and finite deformations.

3. STATE-OF-THE-ART

The field of the analysis of the transient dynamic response of elastic-plastic-

locking materials is in the general field that is concerned with seeking solutions to
nonlinear hyperbolic differential equations. For both linear and nonlinear differential
equations, that are encountered in the field of solid mechanics, very few analytical
solutions have been obtained. As a consequence, alternative methods of solutions have
been sought for many practical problems of solid mechanics. Integral tranform
5-7
techniques have been very effective tools in solving linear elastodynamic problems
Primarily, integral transforms, such as Laplace tranforms, have been used to remove
time as an independent variable. This effectively reduces the equation to an elliptic
type. After the solutions have been obtained to the reduced equation, in tranformed
variables, the next operation is to obtain the inverse tranform to return the solution to
the time domain. One of the particularly successful techniques of obtaining the inverse
8
transforms for elastodynamic problems, is the Cagniard-deHoop technique . Basically,
this method involves the modification of the contour of integrations that are encountered
in the process of obtaining inverse transforms, in such a way that the integrals are
rendered to a form with known solutions. This technique has been successfully applied to
9 22.
many solid mechanics problems -
These tranform techniques have limitations when applied to nonlinear problems and

k-
 8

structures with finite boundaries. For this class of problems, other numerical methods
are necessary. Numerical methods that use the method of characteristics, have been
primarily restricted to the class of one space dimensional problems. Butler 2 3 and
Clifton 2 4 have reported the applications to two space dimensional problems. Most of the
developments in the last two decades have used finite difference and finite element
methods to solve the transient dynamic response problems in solids.

3.1 Finite Difference Methods

The research activity of the sixties and seventies has proved that the finite
element methods are superior to other numerical methods in solving elastostatic
problems. However, the same superiority does not apply to linear or nonlinear transient
dynamic response problems in solids. In fact, the finite difference methods offer a viable
choice to the solution of transient dynamic response problems. It has been son5that,
for selected problems, finite difference methods provide better accuracy and computa-
tional efficiency when compared with the solutions obtained by using finite element
methods. The rest of this review of the state-of-the-art is restricted to finite difference
methods and their application to the study of the linear and nonlinear transient dynamic
response problems in solids.
Early work in the field of the application of the finite difference methods to
transient dynamic response of solids, consisted of the application of a finite difference
scheme developed by Von Neumann 2 for use in hydrodynamics. The Von Neumann's
scheme is an explicit finite difference approximation of the time derivatives and is often
27 paper in this field is by Wilkins 28 to solve
classified as "Leap Frog" scheme 2 7 . A classic
two and three-dimensional transient2 elastodynamic problems. Wilkins has also
considered ideal plasticity and large deformation effects in his studies.
In Wilkin-s approach, as is done in all leap frog schemes, the dependent variables
are staggered in space and time to satisfy the stability requirements and maintain second
order of accuracy. This results in the calculation of only one set of field variables at any
particular mesh point. This means that only stresses or velocities are computed at a
selected mesh point. Furthermore, such a staggered scheme sometimes results in
computationally induced oscillations 3 0 . These cornputatio. .ally induced oscillations can
be minimized by using an artificial viscosity 26 . However, the use of the artificial
viscosity results in a finite difference scheme that is not optimally stable3 0 The time
steps are often reduced by a factor of three in two-dimensional problems 2 2 9
 9

Variations of Wilkins! approach3 1 -36 appear throughout the literature in the field of
the study of transient elastodynamic problems. Clifton 2 4 has proposed a different
approach to the problem. He has extended the difference scheme developed by Butler 2 3 ,
for hydrodynamics, to study two-dimensional elastodynamic problems. This finite
difference scheme has been developed by formulating integration procedures along
bicharacteristics. The resulting procedure is an explicit finite difference scheme.
Unlike the leap frog scheme, all the variables are calculated at all mesh points.
Lax and Wendroff3 7 have developed a second order accurate scheme for
hydrodynamic problems. In this scheme, all the dependent variables are calculated at all
mesh points. At interior points Clifton's scheme and Lax-Wendroff's scheme are
identical. Both these schemes, however, are not optimally stable, in the sense of
Courant, Friedrich and Levy, for more than one space dimension
38 . This leads to
computational procedures that are less than optimally efficient. Also the Lax-Wendroff
schemes involve computation of squares of certain matrices and result in a complicated
algorithm. Smith 3 9 has applied the Lax-Wendroff procedures and the early time splitting
procedures of Strang4 0 to two-dimensional problems. Smith's work is concerned with the
comparison of the relative efficiency of the schemes and is not a complete initial
boundary value problem.
However, an improved version of Lax-Wendroff's scheme, that has the potential of
providing an improved computational efficiency and accuracy, can be applied to both
linear and nonlinear transient dynamic response problems in solid mechanics. The
improvement can be incorporated in three different areas. The first improvement is in
4 0
the modification of the scheme to make it optimally stable . This is due to Strang 4 0 ' 4 1
and somewhat similar to that used by Smith. The second improvement is to improve the
computational efficiency by providing second order accuracy at every other time step.
This improvement follows the developments of Gottlieb 3 , Gourlay 4 2 , Morris 4 2 and
Mitchell
43 . The third improvement is in the incorporation of McCormack's 444 two step
procedure. This third improvement results in programming simplicity and eliminates the
need for squaring the matrices that is needed in Lax-Wendroff or Strang's procedures.
This improvement is very useful in solving nonlinear problems.
An improvement in the implementation of the finite difference schemes is the
concept of point condition codes.2 Attempts at improvement of osillations, overshoots
and smearing of contact discontinuities are also possible

_-mood
 10

4. ANALYSIS UNDER CONDITIONS OF ONE-DIMENSIONAL STRAIN

4.1 Development of the Procedure

The proposed second order accurate method of analyzing the behavior of the
locking material has received little attention when applied to solid mechanics problems.
The efficiency and the resulting accuracy of the method have been studied as a part of
another sponsored research project at Georgia Tech. In this project, one and two-
dimensional problems of transient dynamic analysis have been investigated. At present,
nonlinear problems have been studied. In the previous analyses, the nonlinear behavior
has not been studied.
In this section, the proposed second order accurate difference scheme has been
applied to the transient dynamic response of locking materials. The application is
restricted to one-dimensional strain in the x-direction. The numerical results are
presented for the specific cases of loading for aluminum locking materials. This study is
undertaken for the following reasons.

(1) one-dimensional problem is relatively simple

(2) the analysis provides an understanding of how the proposed second order
accurate difference schemes can handle the phase transition
(3) the effect of different initial densities for the various locking materials can
be qualitatively understood with regard to a one-dimensional analysis.
(4) The method of analysis for one-dimensional problems would indicate the
possible areas of difficulties that may be encountered when extended to two
and three-dimensional problems of transient dynamic analysis of locking
materials.

The one-dimensional equations of motion have been written in Lagrangian

coordinates. The initial positions of the body have also been selected as the Lagrangian
coordinates. Large elastic-plastic deformations of the locking material have been
considered. The resulting equations are as follows:

-{u} A(p) fu (4.1)

where

U)i = v(x,t), pOx't), ar N(xt), ay(x~ t)} (4.2)

 11

and

0 0 p 0

f P). 0 0 0

2G 0 0 0

In these equations, it has been assumed that the stress tensor can be separated into
hydrostatic pressure and stress deviators. The hydrostatic pressure has been assumed to
be related to the changes in density and follow a locking behavior as shown in Figure 1.3.
The stress deviators are assumed to follow an elastic-ideal plastic behavior with
Von Mises' yield condition.
All the assumptions of the preceding paragraph are approximations. However,
these approximations have been used, in the past, to express the mechanical behavior of
locking materials 3 ,4 ,52 and other solids5 2 at very high stresses. On the basis of
experimental results, the approximations have been found to be reasonable5 2 . In the
absence of any available constitutive relationships on the basis of second Pioa-Kirchhoff
stresses and Green-Lagrange strains, this paper has considered similar approximations.
The constitutive relationships have been written in terms of Cauchy stresses, particle
velocity gradients and density. Then,

a x = p + 4S x (4.4)
~ x

o = -p + Sy (4.5)

and

p = fl(P) 0 < P <

0 P"
p = f 2 (p) PI < 0 < t • (4.6)
p = f 3 (p) Pt < P

i ---
 12

The quantities f I f 2 and f3 are as follows (Figure 1.3):

fl(/ )
~~P
\o0-"- -I PI

f,° -p K(4)-R
P,+- Po 1
2 Lo 1l R I

f 3 (p) = A(P/PG - 1) + B (P/G - 2

(4.7)

Similarly,

=-2Sy=-2S .. !GE
x z p (4.8)

with the yield condition

2 2 2

Sx +2Sy 2 - Y2
2 ! 0. (4.9)

4.2 Numerical Analysis

A numerical integration of the hyperbolic partial differential equation (4.1) by

a
procedure similar to that of Lax requires that the field variables I u at t + A t should
be
calcualted from a knowledge of the field variables u and its spatial derivatives f
at t.
To a second order accuracy,

= ~J t At ' 8t--} 2 t * 0(At) • (.0

The differential equation (4.1) can be used to express the quantities a y/ t and a 2
/a t 2 in
terms of the spatial derivatives of Uu. Then,

t +At = t [A()I (au

(+ -u +
I~2% a)2 2
x2 2
aarI
I3X

2 + [tA(p)J[ (4.11)
j1
[LAI 111It

a ... a .. . . i
 13

The equation (4.11) is very complicated. It involves [A(p) 2 and multiples of first and
second partial derivatives of Uu with respect to x. This equation can be considerably
simplified by using a two step approximation due to McCormack, Richtmyer or Gottlieb.
For example, McCormack two step formulation is as follows.
t
= t) At

= }(Ii
u" )* (4.12)

It is easy to verify the accuracy of (4.12) by expansion. Now, the time derivatives on the
right hand sides of (4.12) are replaced by spatial derivatives. Thus,

• au t
u = Jul [A( 0 )] ax At.

a..u
IuIt+ 't = I u Al - • (4.13)

It is to be noted that [A(p) I* and ay/ ax * are to be evaluated by using the star values
of the field variables. Also, it can be seen that the formulation described in the
equations (4.13) is much simpler than the formulation described in (4.11). To complete
the finite difference formulation, the spatial derivatives of (u } and { u } are expressed
in terms of their spatial finite difference approximations. A central, forward or
backward difference is used depending on whether the point under consideration is an
interior point, left boundary point or right boundary point.

4.3 Stability Requirements

The equations (4.13) can be written in the form of finite difference operators.
Then,
L =rl+ k[AJA
x
L= -11+ k[AIAx) . (4.14)
I~~~ , , II 2I ~
 14

In these equations, Ax represents the spatial finite difference approximations. The

quantity I is the identity operator and At has been replaced by k. Then,

iut +At =-L+LL) i (4.15)

2

r
The operators 1, LI and L, operating on I u
change I I to I It+A
in the manner shown in equation (4.15)
This is an explicit scheme and is subject to the usual
restrictions on k =At to maintain the stability of the computational scheme. In
particular, it can be shown that At 4Ax/c where c is the fastest of the local elastic wave
velocities.

4.4 Plasticity Effects and Yield Conditions

The numerical method that has been used in this paper is an explicit method. The
field variables at t.+at depend only on the field variables, at time t. The particle
velocities and strains at t + At can be computed from the current field variables at t.
The stresses computed at t.+at may violate the yield condition (4.9). However, these
stresses can be adjusted along appropriate normals to the yield surface back to the yield
surface' 2 . The already calculated velocities and strains that depend only on the field
variables at t remain unchanged. The only regions where iterative calculations are
needed are at the boundaries.

4.5 Discussion of Results

In order to achieve the same accuracy as a first order accurate method, a time step
equal to the square root of At that is used in the first order method is needed. The
quantity At is usually less than one and hence the second order accurate method uses a
larger time step. Hence, a larger At can be chosen to obtain the same order of accuracy
as a first order method. This leads to a smaller number of the finite difference cells and
increased computational efficiency. The developed computer program has been used to
examine the efficiency of locking materials in attenuating stresses and protecting
structures. In this analysis, specified items include the peak impact stress and the
unloading pattern of the stress wave impinging on the locking materials. The locking
 15

density pI is usually fixed for a given locking material. The results of the numerical
analysis can be used to explain the peak stress and the stress distribution for various
times after impact and unloading at the boundary.
In the numerical example, aluminum locking materials have been selected. Some of
the properties of aluminum locking materials have been extensively investigated3 . A
pressure density relationship as shown in Figure 1.3 has been assumed. The locking
density pIis equal to the 2.72 gmslc.c. Various initial densities have been considered.
Results have been presented for p0 z 1.39 and 2.1 gms/c.c here. There are three distinct
branches of the pressure-density relationship. Equations for these branches are the same
as in equations (4.6) and (4.7).
The quantity PG is the intersection of the p - 0 curve with p axis as shown in Figure
1.3. The appropriate constants are selected from reference 3. For all p > pt, the
unloading is assumed to follow the slope of the solid p - p curve as shown in Figure 1.3.
Two types of loading have been considered. The first type of loading consists of a
step loading followed by step unloading. It has been assumed that a peak compressive
stress of a00 is applied at time t =0 to the left boundary of the slab of a locking material.
For purposes of illustration, a value of - ao = 100,000 psi (689.5 MPa) has been assumed.
The applied stress is reduced to zero by step unloading at time i=0.? uisec. This loading
pattern is illustrated in Figure 4.1 and will be called loading pattern 'a'. The second type
of loading consists of a step loading of a compressive stress of magnitude - a0at t = 0
followed by an exponential unloading at the left boundary. For purposes of illustration,
-ahas been assumed to be equal to 100,000 psi (689.5 MPa)- The applied stress is
assumed to be maintained at a = -aO for a duration of 0.02 pi sec. Then the exponential
decay of the applied load at the left boundary decreases the magnitude of a0 to .05 a at
0.2 pzseCs. This unloading pattern is illustrated in Figure 4.2 and will be called loading Wb.
First, a solid material slab of initial thickness 0.3 inches (12.7 mm) has been
considered. The thickness has been divided into 100 cells. In this case the solid density
is 0z = 2.72 gms/c.c. The transient response of this slab to the loading pattern 'a' has
been studied by using the developed computer program and the results are illustrated in
Figure 4.3. This figure is a plot of the stress ax as a function of the thickness at two
different instants of time t a 0.21 p~secs and t a 0.68 11secs. As seen in the figure, the
stress at a distance of 0.14 inch (3.6 mm) from the initial left boundary is still
approximately 100,000 psi (689.5 MPa). This is equal to the applied peak stress. No
significant attenuation has taken place during the travel of the stress wave through the
 16

thickness equal to 0.14 inch (3.6 mm).

Next, a locking material of initial density po = 2.1 gms/c.c. and locking density
P1= 2.72 gms. c.c. has been considered. The thickness and division into cells are
identical to those for the solid. The transient response to the loading pattern 'a' has been
computed and illustrated in Figure 4.4. The stress distribution as a function of the
distance from the left boundary has been illustrated for t = 0.13 jasec, 0.54 Pasec,
1.15jasec, and 2.22)asec. It can be seen that the peak stress has reduced by
approximately 35% over a distance of 0.09 inch (2.3 mm) from the left boundary. In this
figure, the elastic forerunner wave can also be seen. Similar stress distributions are
observed for locking materials of initial density po = 1.818 gms/c.c., 1.604 gms/c.c. and
1.39 gms/c.c. If these locking materials are considered as countermeasures in front of a
given structure, the peak impact stress has been reduced by an amount as much as 50%
when the shock wave has propagated a distance of 0.013 inch (0.33 mm) (Figure 4.5). It
can also be observed that the locking material of lower initial density has the potential
of attenuating the impact peak stress by a larger percentage when compared with the
locking material of higher initial density.
As a next step, the loading pattern ' has been considered. The results of the study
of transient response through a solid has been illustrated in Figure 4.6. The results are
similar to those for the loading pattern 'a'. The peak impact stress has not attenuated
during the passage of the stress wave over a distance of 0.15 inch (3.8 mm). Similar
stress distribution for loading pattern b' has been illustrated in Figures 4.7 and 4.8 for
different locking materials of initial density p 0 = 2.1 gms/c.c. and 1.39 gms/c.c. The
Figure 4.7 is for a locking material with initial density 2.1 gms/c.c. Over a shock wave
traverse of .045 inch the peak stress has been reduced by 60%. However, the locking
material with initial density of 1.39 gms/c.c. can attenuate the peak stress by almost
90%, during the shock traverse of 0.05 inch (1.3 mm). This is illustrated in Figure 4.8.
The unloading waves and the elastic forerunners can be identified in all these figures. In
Figure 4.9, the effect of the plasticity on the transient dynamic response is shown. A
yield stress of 80,000 psi is assumed for purposes of illustration.

5. ELASTIC - PLASTIC - LOCKING MATERIALS UNDER CONDITIONS OF AXISYMMETRY

3.1 Introduction

In this section, the numerical scheme that has been discussed in Sections 3 and 4

* 1_ _ _ _
 17

has been modified and applied to study the transient dynamic response of elastic-plastic
locking materials under conditions of axisymmetry. A cylindrical polar coordinate
system r, Oz has been used. Initial positions of various material points have been used as
Lagrangian coordinates. The problem has been formulated as a finite deformation
problem with Cauchy stresses that are defined in a deformed coordinate system. In order
36
to represent the equations in a form 1that is suitable for the use of Gottlieb -
MacCormack 4 4 -Strang4 0 '4 1
type of scheme, the constitutive relationships have been
used in a rate form. The Jaumann stress' 3 rate (or the corotational stress rate) has been
used to satisfy the principle of objectivity.

5.2 Governing Equations

The governing equations are then written as follows.

Kinematic equations:

= w (5.1)

f u (5.2)

Equations of motion:

I azz I Tzr I Tzr

P 3z P ar p r

I aTzr I Orr I arr-06

paz
-T -. r (5.4)

Constitutive equations:

(p +z Gzr ( G) G (.)

___ ip3z r P p+ z

) k u + . . . ...... ). +Ip~ T . . . A
ll-l
G ) LI+ ( -
dp 3 r m .... z , dp , ip IIIr z P
 18

AR~ - 2 G +L (pdD - Z G) u (pog + G) U (5.7)

dp3 a0d ar dp 3 r

w 'r (5.8)

Continuity equation:

=
-P L-W
3 w - 0 L'
3u -PRu (5.9)

It has been assumed that

P = P(P) (5.10)

follows the equations similar to (4.6) that represent a locking behavior. A Von Mises
yield condition and ideal plasticity have been assumed.

(a + p)2 ( 2 + (a + p)2 + 2 (.11y

zz + (arr + P) 60 p Tzr.5-3

The notation used in the equations are as follows:

z,r current positions of the coordinate

w velocity in z-direction
u velocity in r-direction
a normal stress in z-direction
Srr normal stress in r-direction

a88 normal stress in -direction

T
zr shear stress
p density
p hydrostatic pressure
G modulus of rigidity
Y yield stress in simple tension
L length, in the z-direction
R radius of the structure

I1
 19

The equations (5.1) to (5-.9) can

be rewritten in the folJowing form

[A]= IA
u), + EB] f U 1' + [C ] {u (512)
where
u zor, w, u, rr' ae FTzrPI "
The matrices [A],[B] and [C]
are not constants in the finite
deformation problem of
elastic-plastic-locking materials.
Specifically

o 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
o 0 0 0 I/P 0 0 0 0
0 0 0 0 0 0 0 1/p 0
0 0 - zr 0 0 0 0 0
d4p 32G (5.13)
d "T Tzr 0 0 0 0 0

0 0 do 0 0 0 0 0 0
0 G+2 azz -a rr) 0 0 0 0 0

0 a 0 0 0 0 0 0

o 0 0 0 0 0 0 / 0
o 0 0 0 0 0 0 0 0
0 0 a 0 0 0 0 /P 0
o 0 0 0 0 /o 0 0 0

0 0 0 d p. 2G
0 0 0 0 0
o 0 G z -ao + T 0 0 0 0 0
d000 0 0
2 zz rr
0 0 Ip 0 0 0 0 0
 20

and

0 0 1 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0
0 0 0 0 0 0 0 1/pr 0
0 0 0 0 0 I/or -I/pr 0 0

[c]= 0 0 0 AR - L)Ir 0 0 0 0 0 *(5.15)

o 0 0 (p9P-- )/r 0 0 0 0 0

0 ~ +L)r 0 0 0 0 0
0 0 0 0 0 0 0 0 0
0 0 0 -p/r 0 0 0 0 0

5.3 Numerical Method

For the hyperbolic differential equation (5.12), the Lax-Wendroff second order
operator can be written as follows.

{u} t + At = L {u}t (5.16)

where
Lzr C] + At(IAA.z + B] A + Cc])

+tz[ 2 * [B] 2 rr ([A] [B] [A]) Ar,

+ [A][ A], u &zA + [1B][A],. A, A.

* [A],[A]A, A, [A],.[B] Ar Az + tA],[C] A A

" ([A][C]+ CC] [A])

"
 21

+ [A]EB], 8zAr + EB][B],u

+ [B],uEA]Az Ar + EB],uEB] ArA

r

+ EB],u [C] Ao Ar + ([BJEC] + [CJ B])Ar

+ [A] c],o ' z Ao +I] [c], Ar A0 + [c],o[A] Az Ao

+ [C],[]A&o + [c],o[C]Ao [C]2 (5.17)

In this equation, the quantities

%z{u} .. a{u

{u {u},

Az{} u} -I {u}, (5.18)

az
2
ar

and

represent the finite difference approximations. Because the Lagrangian meshes deform,
a central finite difference operator is not always suitable. The central difference has
been replaced by a contour difference that maintains a second order accuracy.
The finite difference operator is not optimally stable. The maximum value of the

.a.
 22

time step is restricted by the following equation.

At max(X I
m i n ( A , -z
Ar) A' X /8

where XA' XB are the eigenvalues of A and B respectively. In an optimally stable

scheme, the right hand side, which is also often called Courant, Friedrichs and Levy
number (CFL), should be 1, instead of lIv/8. Furthermore the operator (5.17) is
complicated and the increased number of computations, in many cases, negates any
computational efficiency obtained by resorting to second order accurate methods. The
0
Lax-Wendroff operator can be replaced by a modified Strang's4 ,4 operator to make the
Lax-Wendroff's scheme optimally stable. The modification of Strang's scheme involves
replacing the central difference operators of Strang by contour differences and
accomodating the axisymmetric equations. Then,

UPt+ At= LAt/2 LAt/ 2 LzAt/ 2

LrAt/2 { u
It (5.19)

{ }~+t r z z r~u

where

L~t ' ]+ t ([A]A, + [D]Ao)

+ ttLA] AA + [A][D] [D][A]

* [ A][A]u A. + [A],u[A ]. + [A]u[D] o, Az

* [D] 2 *+ A][D] , + [D],u[A]A z + [D],u[D] Ao} (5.20)

 23

2i
rA /2 Ij +VL([B] +{E])

4 BJL+ [ LIf 'FF + LJ

B Si'u A,

+ B]u [B] Ar + [B],u[E] A.) Ar +([E][E]+ BEE],u Ar

+ [E],u[BlA r + CE~ E] )
A52

and

[D]+[E]=[C] (5.22)

Even though the Strang's type of operator (5.20) is optimally stable, the
computational efficiency can be substantially increased by restoring the second order
accuracy at every other time step by following the procedures similar to those of
Gourlay, Morris and Mitchell 42,43
2 ,
Then

L t+ 2At = L At LAt Lztt LAt f u(

rz r z z r f(.23)

The finite difference operators (5.23), (5.20) and (5.21) still contain multiplication of
matrices, such as, I A IB), IAI,] I A) etc. The operation can be simplified by using
44
two step procedures . Thus, for example

Lz =[II + At (A] Az + At (DI

2!(I L +a
* A J ~ *A+ 1 (5.24)
 24

where I A I* and I D ]* are evaluated by using the value of { u }*. A similar expression
can be written for Lr in r-direction. Suitable spatial contour difference forms that
maintain second order accuracy and are suitable for use in equations of the type (5.24)
have been developed at Georgia Institute of Technology . These are different from those
used in references (28, 29)
For an interior point, a typical member V of the field vector {u} is evaluated as

follows.

V(1 ) t+ At(BA r + E)Ut

V(2 = (U t +V( ) + I At(B(IV + E(I)) V(

V()
(3 VV(2)+ At(A(2 Az
(2)z D+)D
(2)) V(2)
)

= V + A V+D )
(4) V(2) +V(3 ) - V * D(3 ) V(3 )

(5.25)
V(5) = V(4 ) + At(A(4 ) Az + D(4 ))V(4 )

V(6 ) = (V( 4 ) ( + At(A (5)Vz + (5 ) (5 )

V(7 ) = V(6 ) + At(B(6) Ar + E(6 ))V(6 )

V~~~~~6 rVV
+iA(
(8) = 2(6) + (7) 2At(B( 7 )V E( 7 ))V( 7 )

ut +2At V

In this equation Ar and Vr represent the contour difference operators that represent the
equivalent forward and backward difference operators, that are needed in the two step

*Chen, H. P., Ph.D. Thesis

I
 25

procedures. At the boundaries, these equations have been modified to include the
appropriate boundary conditions.

5.4 Numerical Results and Discussion

The material considered for the detailed numerical analysis has the following
properties:

Solid density, p, = 2.72 gms/cc

Young's Modulus, E = 1.1 x 107 psi (75845 MPa)
Poisson's ratio, v = 0.3

In the following analysis, three different types of material behavior have been
considered. These are: (i) elastic behavior, (ii) elastic-plastic behavior and (ii) elastic-
plastic behavior in a locking material. For the study, a yield stress Y = 20,000 psi
(137.9 MPa) has been assumed. For the locking material, the initial density is chosen to
be 2.1 gms/cc. There are three distinct branches of the pressure-density relationship.
These are shown in Figure 5.1. For this analysis, first two branches are assumed to be
straight line segments. The values chosen for p of 1, p1 etc. (figure 5.1), for the present
numerical studies, are

P0 = 2.1 gms/cc P1 6600 psi (45.5 MPa)

P1 2.11 gms/cc pe = 66600 psi (459.2 MPa)

and

a n= 2.7
2 g m s/cc

The equations of the lines corresponding to the three branches of the pressure-density
relations are

P(
p-
=l-o 0 pp
 26

P = P1 + Po-P (p PI P < <,

p = A (o/P G - ) + B (P/p G 1) 2 p > t

The constants A, B, P G are selected from reference 3.

Two different time variations of loading are considered. The first type is a step
loading (without unloading) as shown in Figure 5.2. The second type is also a step loading
but with an unloading at a time t tL This is shown in Figure 4.1. The magnitude of the
peak compressive stress %has been assumed to be 100,000 psi (689.5 MPa). The
geometry of the structure and the corresponding applied loading are shown in Figures
5.3a and 5.3b. The boundary conditions at the unloaded edges are also shown in the same
figures. These boundary conditions are the same for both the cases of applied loading.
Physically, these boundaries can be considered to be "frictionless-rigid". This means
that, on these edges, the velocities in the direction normal to the plane of edges and the
shear stresses are zero. Referring to figures 5.3a and 5.3b, these boundary conditions are
expressed as W =0, T zr z 0 at the boundary which can be denoted by z(t =0) =L and
U=0, T zr Oat the boundary denoted by r(t = 0) = R. On the left boundary, represented
0
by z(t = 0) =0, there are two types of applied loading. One type corresponds to the load
on the entire edge as shown in Figure 5.3a and the other corresponds to a partial load on
the edge as shown in Figure 5.3b. There are two cases of loading type shown in Figure
5.3a. These cases are denoted as case 5a and case 5b. In both these cases, the loading,
the geometry and the boundary conditions are such that a one-dimensional strain
condition is created in the axisymmetric problem that consists of two space dimensions
and time. The problems under these cases are solved in order to simulate a one-
dimensional problem of transient dynamic analysis that can be compared with known
solutions. The difference between case 5a and case 5b is that case 5a has only a step
loading while case 5b has a step loading followed by a step unloading. The case Sb can
also be called a pulse loading.
Similar to the above cases, there are also two different cases under the partial
loading condition shown in Figure 5.3b. These cases are identified as case 5c and case
3d. The partial loading creates two-dimensional conditions. Again, the difference
between case 3c and case 5d is that case Sc represents a step loading and case 5d
represents a pulse loading.
 27

The dimensions pertaining to the problem are as follows (Figure 5.3)

L = 3.2 in (81 mm), R = 0.8 in (20 mm), R1 = 0.32 in (8 mm)

In order to apply the finite difference scheme, the region bounded internally by the edges
z(t = 0) = 0, L and r(t = 0) = 0, R is divided into 400 cells. The dimensions of these cells
are: Az = Ar = 0.08 in (2 mm). This mesh pattern is used for all the cases. Since the
finite difference scheme proposed in this section is optimally stable, the CFL number is
chosen to be 1. However, the time step at is not always a constant. All the plots display
the variation of the stress a zz with repsect to the deformed z-coordinate. Also, the
symbols E, E-P and E-P-L in the figures refer to the elastic solid, elastic-plastic solid
and elastic-plastic-locking material respectively.
As a first step, some typical results of the stress variations obtained by using a leap
frog scheme (programed at Georgia Institute of Technology) with linear or quadratic
artificial viscosities are compared with those obtained by the second order accurate
method. These are shown in Figures 5.4a, 5.4b and 5.4c. In these figures, step loading
condition of case 5a has been assumed and only elastic behavior is considered. The
addition of artificial viscosity in a leap frog scheme results in a scheme that is no longer
optimally stable. Suitable CFL numbers must be chosen in these numerical calculations.
In Figure 5.4a the descriptions of the plots correspond to the CFL numbers and times t as
given in the following table.

Method used CFL number Elapsed time t after

the application of load

present method 1.0 9.25 1jsecs

leap frog scheme 0.3334 9.26 tsecs

with linear viscosity

leap frog scheme 0.3334 9.51 Usecs

with quadratic viscosity

It can be seen from Figure 3.4a that the a zz- z variation by the present second order
accurate method has a very sharp wave front, and indicates very little oscillation behind
the shock wave. A modest amount of overshooting of the shock front can also be
observed. The curve obtained by leap frog scheme with linear viscosity has a very wide
 28

transition layer instead of a steep shock front, although it has no overshooting and
oscillations behind the shock wave. In the plot that represents the results of the leap
frog scheme with quadratic viscosity, oscillations behind the shock wave are significant.
The solution represents a problem with one-dimensional strain and hence no physically
significant oscillations are present. All oscillations are computationally induced. A
larger overshooting is observed when compared with that of the second order accurate
method. The shock wave front is also not steep. The results of using higher CFL
numbers in the leap frog scheme with artificial viscosity are shown in Figures 5.4b and
5.4c. For the results shown in Figure 5.4b, the CFL number used is 0.66 for both the
linear and quadratic viscosity cases. Obviously, the results indicate instability.
Oscillations behind the wave front, large overshooting and wide transition shock layers
can be observed. In Figure 5.4c, the values of CFL = 0.7 and 0.9 are used in the leap frog
scheme with linear and quadratic viscosities respectively. These results are also not
satisfactory because of the numerical instability. From this elementary analysis, it can
be concluded that the second order accurate finite difference method suggested in this
report is a desirable choice for using larger time steps.
Elastic, elastic-plastic and elastic-plastic locking behavior are compared in Figures
5.5 and 5.6. In both these figures, the stress a has been plotted against the current z
coordiantes that were originally located along r(t = 0) = 0. The Figure 5.5 corresponds to
approximately 7 usecs after loading at the boundary z(t = 0) = 0. The corresponding time
for Figure 5.6 is approxiamtely equal to 12 usecs. The specific times for the elastic
body, elastic-plastic and elastic-plastic-locking bodies in the Figure 5.5 are 6.61 Iusecs,
6.60 Uisecs and 6.74 lhsecs. Similar times in Figure 5.6 are 11.09tisecs for the elastic
body, 11.86 Iecs for the elastic-plastic body and 11.96 Iusecs for the elastic-plastic
locking material. The calculated values were available at these different times because
of the different values of At. The three different wave fronts can be seen for the
elastic-plastic locking material. The single wave front for the elastic body and a two
wave front system for elastic-plastic body can also be seen. There are no oscillations
behind the wave front because of the simulation of one-dimensional strain conditions by
a combination of loading, geometry and boundary conditions. The velocity of the locking
wave is the slowest of the waves.
In Figure 5.7, the results for the case of a pulse loading are shown. The loading
consists of a step loading followed by a step unloading. Again the loading, geometry and
the boundary conditions simulate the one-dimensional strain conditions. The applied load
 29

is set to zero at t =t. The value of t = 1.33 secs for the elastic body, t =1.33 Ij secs
for the elastic-plastic body and t =1.42 p~ secs for the elastic-plastic locking material
have been selected. The figure clearly indicates that the attenuation of peak stresses is
maximum in the locking material. A reduction of peak stress is also seen in the elastic-
plastic body without locking. The reduction, however, is not significant in comparison to
the locking material. In the elastic body the peak stresses are not reduced. These
results support the fact that locking materials are effective stress attenuators.
The loading for cases 5c and 5d correspond to a partial loading on the surface
z(t = 0) =0. This corresponds to a two-dimensional (r - z) problem. The variation of z
against z are shown in Figures 5.8 to 5.10. These are for a step loading without
unloading. The two Figures 5.8 and 5.9 correspond to lines r(t = 0) = 0 and 0.56 in
(14.2 mm). It is observed that the elastic-plastic locking material displays a steeper
wave front in comparison with an elastic-plastic material. The oscillations behind the
elastic wave, due to boundary effects in this two-dimensional problem, can be seen. The
stress magnitudes in Figure 5.9 are small. This is due to the fact that the region is away
from the area of the load application. The variations of a zzwith z, for r(t = 0) = 0 and
the locking material, are shown in Figure 5.10 for increasing values of times. The
development of three distinct wave fronts can be seen.
The Figures 5.11 and 5.12 correspond to pulse loadings on a part of the surface
specified by z(t = 0) =0. Again, two-dimensional effects are present. The Figure 5.11 is
for r(t = 0) = 0. The Figure 5.12 is for r(t =0) = 0.56 in (14.2 mm) which is located away
from the loading region. Again, observation regarding the attenuation of peak stresses
can be made.
6. CONCLUSIONS

In this report, it has been shown that a locking material shield has the potential of
protecting a given structure from dynamic loads that have a time history similar to that
of mine blasts. This conclusion is based on the analysis under conditions of one-
dimensional strain and the studies under conditions of axisymmetry. In order to provide a
tool for designing the locking material shield, a second order accurate, computationally
efficient numerical scheme has been developed. The numerical scheme is capable of
considering large deformations and locking phase transition and provide accurate results
in comparison to the present state-of-the-art. Point condition codes are also developed
for locking material-structure combinations. The accuracy and the reliability of the
resulting computer program have been checked by comparing the results with known
 30

solutions.
To follow the present work, some additional tasks are recommended. One such task
is to improve the accuracy and the computational efficiency during phase transition and
unloading. Furthermore, the shockwave can be steepened by employing new techniques
such as the method of artificial compression. Another area of suggested work is the non-
ideal plasticity. A very important additional task will be to investigate the accuracy of
methods such as the state space approach with its inherent exactness and simplicity of
reducing three-dimensional problems to those involving only two-dimensions.

[ ________
 31

REFERENCES

1. Hanagud, S., "A Contribution to the Theory of Isotropic Licking" Sudaer Report
No.152, Stanford University, California, 1963.

2. Hanagud, S., "Finite Amplitude Spherical Shock Waves in Locking Solids" Proc. 5th
U.S. National Congress, 1966. (Invited Sectional Lecture).

3. Linde, R. K., and D. N. Schmidt, "Attenuation of Shockwaves in Distended Solids"

AFWL-TR-66-13, 1966.

4. Hanagud, S., G. S. Sidhu and B. Ross, "Elastic-Plastic Impact of Compactible Rods

and Plates" Israel Journal of Technology, 1969.

5. Miklowitz, 3., The Theory of Elastic Waves and Waveguides, North-Holland, 1978.

6. Achenbach, J. D., "Wave Propagation, Elastodynamic Stress Singularities, and

Fracture", Theoretical and Applied Mechanics, ed. W. T. Koiter, North-Holland
Publishing, 1976, pp. 71-87.

7. Fung, Y. C., Foundations of Solid Mechanics, Prentice-Hall, 1965.

8. de Hoop, A. T., "A Modification of Cagniard's Method for Solving Seismic Pulse
Problems", Appl. Sci. Res., Sec. B, 8, 1959, pp. 349-356.

9. Ang, D. D., "Transient Motion of a Line Load on the Surface of an Elastic Half-
Space", Quart. Appl. Math., 18, 1960, pp. 251-256.

10. Mitra, M., "Disturbance Produced in an Elastic Half-Space by Impulsive Normal

Pressure", Proc. Camb. Phil. Soc., 60, 1964, pp. 683-696.

11. Norwood, F. R., "Exact Transient Response of an Elastic Half Space Loaded Over a
Rectangular Region of its Surface', 3. Appl. Mech., 36, 1969, pp. 516-522.
12. Gakenheimer, D. C., and 3. Miklowitz, "Transient Excitation of an Elastic Half
Space by a Point Load Traveling on the Surface", 3. Appl. Mech., 36, 1969, pp. 505-
515.

13. Gakenheimer, D. C., "Numerical Results for Lamb's Point Load Problem", J. App.
Mech., 37. [970, pp. 522-524.

14. Chao, C. C., "Dynamical Response of an Elastic Half-Space to Tangential Surface

Loadings", 3. App. Mech., 27, 1960, pp. 559-567.

15. Eason, G., "The Displacements Produced in an Elastic Half-Space by a Suddenly

Applied Surface Force", 3. Inst. Maths. Applics., 2, 1966, pp. 299-326.

16. Craggs, 3. W., "On Axially Symmetric Waves", Proc. Camb. Phil. Soc., 59, 1963, pp.
803-809.
 32

17. Craggs, 3. w, "On Two-Dimensional Waves in an Elastic Half-Space", Proc.

Cambridge Phil. Soc., 56, 1960, pp. 269-285.

18. Eason, G., "The Stress Produced in a Semi-Infinite Solid by a Moving Surface
Force", Int. 3. Engng. Sci., 2, 1965, pp. 581-609.

19. Ungar, A., "The Propagation of Elastic Waves from Moving Normal Point Loads in
Layered Media", Pageoph, 114, 1976, pp. 845-861.

20. Miles, 3. W., "Homogeneous Solutions in Elastic Wave Propagation", Quart. Appl.
Math., 18, 1960, pp. 37-59.

21. Norwood, F. R., "Interior Motion of an Elastic Half-Space Due to Normal Finite
Moving Line Load on its Surface", Int. 3. Solids Struct., 6, 1970, pp. 1483-1498.

22. Handelman, G. H., and L. A. Rubenfeld, "Diffraction of Horizontal Shear Waves by

a Half Plane", 3. Appl. Mech., 36, 1969, pp. 873-874.

23. Butler, D. S., "The Numerical Solution of Hyperbolic System of Partial Differential
Equations in Three Independent Variables" Proc. Royal, Soc. (London), Series A, 17,
1964, pp. 381-398.

24. Clifton, R. 3., "A Difference Method for Plane Problems in Dynamic Elasticity",
Quart. Appl. Math., 25, 1968, pp. 97-116.

25. Hanagud, S., and R. L. Latham, "Interaction of Stress Waves and Cracks", Technical
report, Georgia Institute of Technology (AFOSR Grant, 1982)

26. Von Neumann, 3. and R. Richtmyer, "A Method for the Numerical Calculation of
Hydrodynamic Shocks", 3. Appl. Phys., 21, 1950, pp. 232-237.

27. Roache, P. 3., Computational Fluid Dynamics, 2nd edition, Hermosa, 1976.

28. Wilkins, M., "Finite Difference Scheme for Calculating Problems in Two Space
Dimensions and Time", 3. Comput. Phys., 5, 1970, pp. 406-414.

29. Chen, Y. M., and M. L. Wilkins, "Numerical Analysis of Dynamic Crack Problems",
Elastodynamic Crack Problems, ed., G. C. Sih, Noordhoff, 1977, p. 295-345.

30. Richtmyer, R. D., and K. W. Morton, Difference Methods for Initial Value
Problems 2nd edition, Interscience-Wiley, 1967.

31. Wilkins, M. L., "Calcualtion of Elastic-Plastic Flow", Methods in Computational

Physics 3, 1964, Academic Press, pp. 211-263.

32. Andrews, D., "A Numerical Study of Tectonic Stress Release by Underground
Explosions", Bull. Seismol. Soc. Am., 63, No. 4, pp. 1375-1391.

33. Andrews, D., "Rupture Propagation With Finite Stress in Antiplane Strain", 3ournal
of Geophysical Research, 81, No. 20, 1976, pp. 3575-3582.

A_
 33

34. Madariaga, R., "Dynamics of an Expanding Circular Fault", Bull. Seism. Soc. Am.,
66, No. 3, 1976, pp. 639-666.

35. Alterman, Z. and F. Karal, "Propagation of Elastic Waves in Layered Media by

Finite Difference Methods", Bull. Seismol. Soc. Am., L8, No. 1, pp. 367-398.

36. Hanagud, S., B. Ross and G. S. Sidhu, "Axisymmetric Impact of Compactible Rods
Subjected to Large Deformations", Israel 3. of Technology, 1972.

37. Lax, P., and B. Wendroff, "Difference Schemes for Hyperbolic Equations with High
Order of Accuracy", Comm. Pure App. Math., 17, 1964, pp. 381-398.

38. Gottlieb, D., "Strang-Type Difference Schemes for Multidimensional Problems",

SIAM 3. Numer. Anal., 9, 1972, pp. 650-661.

39. Smith, P. D., "On Some Numerical Schemes for the Solution of Wave Propagation
Problems", Int. 3. Num. Meth. Engng., 8, 1974, pp. 91-102.

40. Strang, G., "On the Construction and Comparison of Difference Schemes", SIAM 3.
Numer. Anal. 5, No. 3, 1965, pp. 506-517.

41. Strang, G., "Accurate Partial Difference Methods 1: Linear Cauchy Problems",
Arch. Rat. Mech. Anal., 12, 1963, pp. 392-402.

42. Gourlay, A. R., and 3. L. Morris, "On the Comparison of Multistep Formulations of
the Optimized Lax-Wendroff Method for Nonlinear Hyperbolic Equations in two
Space Dimensions", 3. Comp. Phys., 5, 1970, pp. 229-243.

43. Gourlay, A. R., and A. R. Mitchell, "A Classification of Split Difference Methods
for Hyperbolic Equations in Several Space Dimensions". SIAM 3. Numer. Anal., 6,
No. 1, 1966, pp. 62-71.

44. MacCormack, R. W., and A. 3. Paullay, "The Influence of the Computational Mesh
on Accuracy for Initial Value Problems with Discontinuous or Nonunique Solutions",
Comp. Fluids, 2, 1974, pp. 339-361.

45. Gottlieb, D., and Turkel, E., "On Acceleration of MacCormack's Scheme", 3. Comp.
Phys., 26 1978, pp. 252-256.

46. Glimm, 3., "Solution in the Large for Nonlinear Hyperbolic Systems of Equations",
Comm. Pure Appl. Math., 18, 1965, pp. 697-715.

47. Chorin, 3. A., "Random Choice Solution of Hyperbolic Systems", 3. Comp. Phys.,
22, 1976, pp. 517-533.

48. Harten, A., "The Artifical Compression Method for Shocks and Contact
Discontinuities", Comm. Pure Appl. Math., 30, 1977, pp. 611-630.

49. Harten, A., "The Method of Artificial Compression" U.S. Atomic Energy
Commission, contract report-( 1-)-3077, 1979.
 34

50. Gottlieb, D. and Turkel, E., "Boundary Conditions for Multistep Finite-Difference
Methods for Time-Dependent Equation", 3. Comp. Phys., 26, 1978, pp. 181-196.

51. McGuire, G. R., and Morris, J. L., "Restarting Order of Accuracy for Multilevel
Scheme for Nonlinear Hyperbolic System in Many Variables", J. Inst. Math. Appl.,
17, 1976, pp. 53-67.

52. Varley, E. (Ed.), "Propagation of Shcok Waves in Solids", AMD, 17, ASME, New
York 1976.

53. Truesdell, C. (Ed.) Continuum Mechanics III, Foundations of Elasticity Theory,

International Science Review Series, 1965.

Ii

. S
 35

p p

0 p
o p

Fig.l.2 An Elastic Locking Solid

Fig.l.l A Simple Locking Material

p before yieldingp
p p = -K ln

P1

0 P 0

4
3 Fig.l. Pressure-Density Relationship
Fig.l. Experimental p-p diagram
for Locking Materials

go
Go /

2 )y

3
/
x° 0
0
Fig.l.5 Material Geometry and P
Loading PO P,
Fig.l.6 Ox - p Relationship
 36

=
(1 -po P

2.72 ;/cc
.56

o t .42- .68 lsecz

Fig.4.1 Step Loading - Case 'a' Stress -. 22 ltsec

(GPa) 2
028
.1

0 0X()
- t- 0 1.5 3.0
Fig.4•3 Stress Distribution
Fig.4.2 Step Loading - Case 'W i4. (Laiga) 'a')
(Loading

.70 .70

Po - 2.1 gms/cc Po = 1.39 gms/cc

- 2.72 gms/cc PL = 2.72 gms/cc

.56 *--13.&sec .56

• 4 l sec

1.17 /Asec
.42 .42 .17 1Asec

St.;.ss 2.22S c Stress

(GPa) (GPa)
.28 .28 .37 .Asec

.•
. 4

0 _ _____ ___ __ x(mm) 0 _ _ _ _ _ __ _ _ _ _

1.5 3.0 .15 .3 .45 .6

Fig.4.4 Stress Distribution Fig.4.5 Stress Distribution
(Loading 'a') (Loading 'a')
___......
 37

.70 .70

Po 2. Ic P0 w 2.1 gins/cc
2.7 cs/ccP, u 2.72 gms/cc

.56 .56

.42 .42 .18/A.Bec

Stress .67 /usec Stress
(GPa) (GPa) .85 jssec
.28 .28 1.51 psec
.27 j.Asec 2.26 j4sec

.14 .14

0 _______ _ x(no) 0 x(mm)

x_________

0 1.5 3.00 15 3.
Fig.4.6 Stress Distribution Fig.4.7 Stress Distribution
(Loadig 'b')(Loading 'b')I

.70

-1.39 gms/cc
P
P, 2.72 gms/cc .70 -
.56

.56-
.42

.42 t .6Ae
Stress
(a).28 .22 j~sec Stress
(GPa)
.28

0 Ditiuio0 2

Fig.4.8 Stress DitiuinFig.4.9 Plasticity Effects

(Loading bW) (Yield Stress -. 56 GPa)
 38

p 0

Go

0 P, P 0 t
Po Pl

Fig.5.1 Simplified p -p Relationship Fig.5.2 Step Loading without

Unloading

R 00

-20 ,
L

Fig.5.3a Geometry and Loading with Fully Loaded Edge

Fig.5.3b Geometry and Loading with Partially Loaded Edge

 39

- Leap Frog(quadr.visc.)

55, eap rog~in~vsc.)Second Order Acc.

(GPa)

16 32 48 64 zm

Fig.5.4a IU - z Variation with Different Schemes (Elastic Case)

r.Leap Frog (quadr.visc.)

* Second order Acc.

Uzz Leap Frog (lin.visc.)1-i%
(GPa)

16 32 48 64zm)

Fig.5.4b U - z variation with Different Schemes (Elastic Case)

1.38 Leap Frog(quadr.visc.)

1.03
11;.:;v 4 Second Order Acc.

I Leap Frog(lin.visc.)

16 32 48 64zi)
Fig.5.4c azz z Variation with Different Schemes (Elastic Case)
 40
.83

.55-
E-P

(GPa)
.28 N.
E-P-L

0z (mmn)
16 32 48

Fig.5.5 Ozz - z Variation for Step Loading

.83
_ .... . .,..o....... ...... ..... ..............
E

E-P N
azz
(GPa) E-P-L
.28.

16 32 48 64

Fig.5.6 ZZ - z Variation for Step Loading

.8.-
A

.6

i~E
5!
azz . Uzz
(GPa) (GPa) .43 A.

9 P-L 16 E-1 .~ . \ " /

.2&
0 - - 3 z(nu) .14. -P-L

z Variation for
Fig.5.7 ZZ -
o - (m)
Pulse Loading 16 24

Fig.5.8 Gas - z Variation for Step

Loading
 41

.83

.11
.69

.08

SJ ,azz
a z..06 Af (GA .41.

(GPa) E-.
03 E-P-L .28
, | l " .248 ra

8 16 24

-.03 _ 0
8 16 24 32

Fig.5.9 z - z Variation for Fig.5.10 azz - z Variation for

Step Loading Step Loading

.10
.62 Ei

::'" ," E.07

.41
(GPa) E--P-L (r'-,FE--

.21 -

00 1 Jzmm
0
I z. .

Fig.5.11 ozz - z Variation for Fig.5.12 fzz - z Variation for

Pulse Loading Pulse Loading

....... ....... OW "

 DISTRIBUTION LIST

No. of
Copies To

I Office of the Under Secretary of Defense for Research and

Engineering, The Pentagon, Washington, DC 20301

12 Commander, Defense Technical Information Center, Cameron

Station, Building 5, 5010 Duke Street, Alexandria, VA 22314

1 Metals and Ceramics Information Center, Battelle Columbus

Laboratories, 505 King Avenue, Columbus, OH 43201

1 Deputy Chief of Staff, Research, Development, and

Acquisition, Headquarters Department of the Army,
Washington, DC 20301
1 ATTN: DAMA-ARZ

Commander, Army Research Office, P.O. Box 12211,

Research Triangle Park, NC 27709
1 ATTN: Information Processing Office
I Dr. F. W. Schmiedeshoff

Commander, U.S. Army Materiel Development and

Readiness Command, 5001 Eisenhower Avenue,
Alexandria, VA 22333
I ATTN: DRCLDC

Commander, U.S. Army Materiel Systems Analysis

Activity, Aberdeen Proving Ground, MD 21005
I ATTN: DRXSY-MP, H. Cohen

Commander, U.S. Army Communications-Electronics

Command, Fort Monmouth, NJ 07703
1 ATTN: DRSEL-LE-R

Commander, U.S. Army Missile Command,

Redstone, AL 35809
1 ATTN: DRSMI-TB, Redstone Scientific Information Center

Commander, U.S. Army Natick Research and Development

Laboratories, Natick, MA 01760
1 ATTN: Technical Library

Commander, U.S. Army Satellite Communications Agency,

Fort Monmouth, NJ 07703
1 ATTN: Technical Document Center

Commander, U.S. Army Tank-Automotive Command,

Warren, MI 48090
1 ATTN: DRSTA-RKA
I DRSTA-UL, Technical Library
No. of
Copies To

Commander, U.S. Army Armament Research and Development

Command, Dover, NJ 07801
2 ATTN: DRDAR-TSS, Technical Library

Commander, White Sands Missile Range, NM 88002

1 ATTN: STEWS-WS-VT

Commander, U.S. Army Armament Research and Development

Command, Aberdeen Proving Ground, MD 21010
1 ATTN: DRDAR-QAC-E

Director, U.S. Army Ballistic Research Laboratory,

Aberdeen Proving Ground, MD 21005
1 ATTN: DRDAR-TSB-S (STINFO)

Commander, Harry Diamond Laboratories, 2800 Powder Mill Road,

Adelphi, MD 20783
1 ATTN: Technical Information Office

Commander, Watervliet Arsenal, Watervliet, NY 12189

1 ATTN: DRDAR-LCB, Dr. T. Davidson

Commander, U.S. Army Foreign Science and Technology Center,

220 7th Street, N.E. Charlottesville, VA 22901
1 ATTN: Mr. Marley, Military Tech

Director, Eustis Directorate, U.S. Army Air Mobility

Research and Development Laboratory, Fort Eustis,
VA 23604
I ATTN: Mr. J. Robinson, DAVDL-E-MOS (AVRADCOM)

Commander, U.S. Army Engineer Waterways Experiment Station,

Vicksburg, MS 39180
1 ATTN: Research Center Library

Commander, U.S. Army Research and Technology Laboratory,

Air Mobility Research and Development Laboratory,
Ames Research Center, Moffett Field, CA 94035
I ATTN: Dr. R. Foye

Director, Structural Mechanics Research, Office of Naval

Research, 800 North Quincy Street, Arlington, VA 22203
1 ATTN: Dr. N. Perrone

David W. Taylor Naval Ship Research and Development Center,

Annapolis, MD 21402
I ATTN: Dr. H. P. Chu

Naval Research Laboratory, Washington, DC 20375

I ATTN: C. D. Beachem, Head, Adv. Mat'Is Tech Br. (Code 6310)
I Dr. Jim C. I. Chang
No. of
Copies To

Chief of Naval Research, Arlington, VA 22217

I ATTN: Code 471

I Ship Research Committee, Maritime Transportation Research

Board, National Research Council, 2101 Constitution Avenue, N.W.,
Washington, DC 20418

Commander, U.S. Air Force Wright Aeronautical Laboratories,

Wright-Patterson Air Force Base, OH 45433
2 ATTN: AFWAL/MLSE, E. Morrissey
1 AFFDL/FB, Dr. J. C. Halpin

Air Force Flight Dynamics Laboratoryi Wright-Patterson

Air Force Base, OH 45433
1 ATTN: AFFDL (FBS), C. Wallace
I AFFDL (FBE), G. D. Sendeckyj

National Aeronautics and Space Administration,

Washington, DC 20546
I ATTN: Mr. B. G. Achhamner

National Aeronautics and Space Administration, Marshall Space

Flight Center, Huntsville, AL 35812
1 ATTN: R. J. Schwinghammer, EHOI, Dir., M&P Lab

National Aeronautics and Space Administration, Langley

Research Center, Hampton, VA 23665
1 ATTN: Mr. H. F. Hardrath, Mail Stop 188M
I Dr. R. J. Hayduck, Mail Stop 243

National Aeronautics and Space Administration, Lewis

Research Center, 21000 Brookpark Road, Cleveland, OH 44135
I ATTN: Dr. J. E. Srawley, Mail Stop 105-1

Lockheed-Georgia Company, 86 South Cobb Drive,

Marietta, GA 30063
1 ATTN: Materials and Processes, Eng. Dept. 71-11, Zone 54

National Bureau of Standards, U.S. Department of Commerce,

Washington, DC 20234
I ATTN: Mr. J. A. Bennett

Director, Army Materials and Mechanics Research Center,

Watertown, MA 02172
2 ATTN: DRXMR-PL
I DRXMR-PR
I DRXMR-K
1 DRXMR-FD
40 DRXMR-SE, Dr. Tsui