NUMERICAL MODELLING IN LARGE STRAIN PLASTICITY

WITH APPLICATION TO TUBE COLLAPSE ANALYSIS

By
JOSEsM. GOICOLEA RUIGOMEZ
Ing. Caminos
A Thesis submitted for the degree of
Doctor of Philosophy
In the Faculty of Engineering of the
University of London
King 's College London
October 1985
2
ABSTRACT
Numeri cal methods are proposed f or t he anal ysi s of 2 or 3-
di mensi onal l ar ge s t r a i n p l a s t i c i t y pr obl ems. A Fi n i t e Di f f er ence
program, wi t h 2- di mensi onal cont i nuum el ement s and e x p l i c i t t i me
i nt egr at i on, has been developed and appl i ed t o model the axi symmetri c
crumpl i ng of ci r cul ar tubes.
New t ypes of mi xed el ements ( Tr i angl es- Quadr i l at er al s f or 2-D,
T e t r a h e d r a - Br i c k s f o r 3-D) ar e pr oposed f o r t he s p a t i a l
di scr et i zat i on. These elements model accurat el y i ncompressi bl e pl ast i c
f l ow, wi t hout unwanted " zer o- ener gy" def or mat i on modes or t angl i ng
over of t he mesh. El a s t i c - p l a s t i c , r at e dependent l aws are model l ed
wi t h a "r adi al r et ur n" al gor i t hm. The t ransmi ssi on of heat generated
by p l a s t i c work and mat er i al dependence on t emper at ur e are al so
i ncl uded, enabl i ng a f ul l y coupled thermo-mechanical anal ysi s.
A 2-D and axi symmet r i c computer program has been devel oped,
i mpl ement i ng t he numer i cal t echni ques descr i bed. Comput at i onal
e f f i c i e n c y was e s s e n t i a l , as l ar ge scal e, c os t l y appl i c at i ons were
i ntended. An i mportant part of the program was the contact al gor i t hm,
enabl i ng the model l i ng of i nt er act i on between surfaces.
The axi symmet r i c cr umpl i ng of tubes under axi al compressi on
( " concer t i na" mode) has been anal yzed Numer i cal l y. Quasi - st at i c
experiments on Aluminium tubes were model l ed, usi ng vel oci t y scal i ng.
Mery l ar ge s t r ai ns are devel oped i n t he cr umpl i ng pr ocess; wi t h t he
help of tensi on t est s, mat eri al laws val i d f or such st r ai n ranges were
developed. Good agreement was obtained between numerical predi ct i ons
and experimental r esul t s. Model l i ng choices such as mesh ref i nement ,
element type and vel oci t y scal i ng were st udi ed, and found t o have an
i mpor t ant i nf l uence on t he numer i cal pr edi c t i ons . F i n a l l y , a l ar ge
scal e i mpact anal ysi s of a st eel t ube at 176m/s was per f or med. The
r esul t s compared wel l wi t h experi ment, i ndi cat i ng di f f erences wi t h the
behaviour of low vel oci t y crumpl i ng mechanisms.
To concl ude, Fi n i t e Di f f er ence procedures wi t h e x p i i c i t t i me -
3
marching techniques are proposed f or l arge st r ai n pl as t i c i t y problems,
at low or medium i mpact v e l o c i t i e s . A f a i r l y r obust code has been
developed and appl i ed successf ul l y t o a range of l arge st r ai n and tube
crumpl i ng problems.
4
ACKNOWLEDGEMENTS
I wish t o express my si ncere gr at i t ude t o the f ol l owi ng:
Dr. G.L. England who super vi sed t h i s work, f or hi s const ant support
and guidance throughout the peri od of research;
Dr. J. Mar t i f rom Pr i n c i p i a , f or hi s i nval uabl e exper t advi ce on
t h e o r e t i c a l and numer i cal mat t er s, and hi s f r i e n d l y support and
suggesti ons;
Or. . Al ar con, f or hi s hel p and mot i vat i on i n t he i n i t i a l stages of
work;
The s t a f f and t echni ci ans of t he Ci v i l Engi neer i ng Department at
Ki ng' s Col l ege, f or t hei r help and cooperat i on;
My wi f e, Tereca Mari n, f or her help i n the preparati on of f i gur es, and
her unending pati ence and encouragement duri ng the research peri od.
I would al so l i k e t o thank the f ol l owi ng bodies f or sponsoring my
research at t he var i ous st ages: The B r i t i s h Counci l , The Spani sh
Government ( Mi ni st r y of Educati on), Pr i nci pi a Mechanica Lt d. (UK), and
the Committee of Vi cechancel l ors and Pr i nci pal s (UK).
TABLE OF CONTENTS
Pa
TITLE
ABSTRACT
ACKNOWLEDGEMENTS
TABLE OF CONTENTS
BASIC NOTATION 10
CHAPTER 1 - INTRODUCTION 13
1.1 Objectives 14
1.2 Non- l i near model l i ng 16
1.3 Layout 17
CHAPTER 2 - CONTINUUM MECHANICS DESCRIPTIONS 18
2.1 Introduction 19
2.2 Kinematics 20
2.2.1 Configurations 21
2.2.2 Deformation tensors 21
2.2.3 Deformation and spin rates 23
2.2.4 Strains 23
2.2.5 Transformations 24
2.3 Stress 25
2. 3. 1 Cauchy 25
2.3.2 Pi ol a- Ki r chhof f 25
2.4 Balance laws 26
2. 4. 1 Balance of momentum 26
Pa
2.4.2 Balance of mass 26
2.4.3 Balance of energy 27
2.5 Constitutive relations 28
2.5.1 Rate equations 29
2.5.2 Elasticity 30
2.5.2.1 Hyperelastic materials 31
2.5.2.2 Hypoelastic materials 31
2.5.3 Plasticity 32
2.5.3.1 Von Mises model 34
2.5.3.2 Other plasticity models 35
CHAPTER 3 - NON-LINEAR NUMERICAL MODELS FOR SOLID MECHANICS 37
3.1 Introduction 38
3.2 Finite Difference methods 39
3.3 Finite Element methods 41
3.4 Mesh descriptions 44
3.4.1 Lagrangian 44
3.4.2 Eulerian 45
3.4.3 Arbitrary Lagrangian-Eulerian 45
3.5 Large displacement formulations 46
3.5.1 Total Lagrangian 46
3.5.2 Cauchy stress - velocity strain 47
3.5.3 Updated Lagrangian 49
3.6 Time integration 49
3.6.1 Central Difference (explicit) 50
3.6.2 Trapezoidal rule (implicit) 52
3.6.3 Operator split methods 54
3.7 Practical considerations for discrete mehes 55
3.7.1 "Locking-up" for incompressible flow 55
3.7.2 "Hourglassing" 58
3.8 Conclusions 59
CHAPTER 4 - EXPLICIT FINITE DIFFERENCE NUMERICAL MODEL 61
4.1 Introduction 63
Pa
4.1.1 General methodology 63
4.2 Spatial semidiscretization 66
4.2.1 Constant Strain Triangles and Tetrahedra (CST elements) 67
4.2.2 Mixed Discretization (MTQ, MTB elements) 72
4.2.3 Prevention of negative volumes (MTQC, MTBC elements) 75
4.2.4 Mass lumping procedure 78
4.3 Momentum balance 78
4.4 Central difference time integration 79
4.5 Constitutive models 79
4.5.1 Hypoelasticity 80
4.5.2 Plasticity; radial return algorithm 80
4.5.3 Hardening and uniaxial stress-strain laws 84
4.5.4 Objective stress rates 87
4.6 Damping 88
4.7 Stability of time integration 90
4.8 Modelling of contacts 91
4.8.1 Contact interface laws 92
4.8.2 Contact detection algorithm 94
4.9 Heat conduction 96
4.10 Energy computations 100
4.11 Implementation into Fortran program 102
CHAPTER 5 - BENCHMARKS AND VALIDATION EXAMPLES 106
5.1 I nt r oduct i on 107
5.2 Wave propagation 107
5.2.1 El ast i c waves i n bars 107
5.2.2 El ast i c- pl ast i c waves 110
5.2.3 El ast i c waves i n cone 112
5.3 Vi brat i on of a cant i l ever 114
5.4 St at i c el ast i c- pl ast i c problems 117
5. 4. 1 Punch t est 117
5.4.2 El ast i c- pl ast i c sphere under i nt er nal pressure 121
5.5 Heat conduction 123
5. 5. 1 Coupled thermomechanical anal ysi s 123
5.5.2 Temperature r edi st r i but i on i n a sl ab 124
5.6 Large st r ai ns and r ot at i ons 124
Pa
5.7 Impact of cyl i nder 126
5.8 Conclusions 128
CHAPTER 6 - TENSION TESTS ON ALUMINUM BARS 129
6.1 Introduction 130
6.1.1 Constitutive idealization 132
6.1.2 Tension tests - a review 133
6.2 Theoretical interpretation of tension tests 136
6.2.1 Strain distribution at minimum neck section 137
6.2.2 Stress distribution 138
6.3 Tension tests 140
6.3.1 Specimens and material 140
6.3.2 Programme 141
6.3.3 Procedure 142
6.3.4 Results 142
6.3.5 Microhardness tests 144
6.4 Material hardening law 147
6.5 Numerical calculations for tension tests 151
6.5.1 Model 151
6.5.2 Analysis 153
6.5.3 Results 154
6.6 Conclusions 156
CHAPTER 7 - CONCERTINA TUBE COLLAPSE ANALYSIS 164
7. 1 I nt r oduct i on 166
7. 1. 1 Scope 166
7.2 Overview of energy di ssi pat i ng devices 167
7. 2. 1 Def i ni t i on and c r i t e r i a 167
7.2.2 Types of energy di ssi pat i ng devices 168
7.2.3 Tubes as energy absorbers 169
7. 2. 3. 1 Lat eral compression 169
7. 2. 3. 2 Axisymmetric axi al crumpl i ng 170
7. 2. 3. 3 Diamond-fold axi al crumpl i ng 171
7. 2. 3. 4 Tube i nversi on 171
Page
7.3 Quasi -st at i c concerti na tube col l apse mechanisms 172
7.3.1 Related experimental work 172
7.3.1.1 Experimental programme and method 173
7.3.1.2 Typical experimental results 174
7.3.1.3 Microhardness tests 177
7.3.1.3.1 Equipment and procedure 177
7.3.1.3.2 Derivation of material strength, Y 178
7.3.2 Numerical model 182
7.3.2.1 Discretization and material 183
7.3.2.2 Velocity scaling 184
7.3.2.3 Interpretation of output 185
7.3.3 Results form numerical calculations and experiment 187
7.3.3.1 Tube geometry A: ID=19.05mm, t=1.64mm, L=50.8mm 188
7.3.3.2 Tube geometry B: ID=19.05mm, t=1.17mm, L=50.8mm 193
7.3.3.3 Tube geometry C: 0D=38.10mm, t=1.65mm, L=50.8mm 197
7.3.3.4 Tube geometry D: 0D=25.40mm, t=0.95mm, L=25.4mm 203
7.3.4 Parametric studies in numerical analyses 208
7.3.4.1 Influence of friction 208
7.3.4.2 Influence of velocity scaling 213
7.3.4.3 Influence of mesh refinement 216
7.3.4.4 Influence of element type 223
7.3.5 Discussion 228
7.4 Medium velocity (176m/s) tube impact analysis 233
7.4.1 Description of problem 233
7.4.2 Numerical idealization 235
7.4.3 Numerical results 237
7.4.4 Discussion 243
7.5 Conclusions 249
CHAPTER 8 - CONCLUSIONS AND SUGGESTIONS FOR FURTHER RESEARCH 251
8.1 Conclusions 252
8.2 Suggestions for further research 253
8.2.1 Theoretical and numerical developments 254
8.2.2 Additional applications 255
REFERENCES
256
10
BASIC NOTATION
A Area; Mass damping coef f i ci ent (eqn. 4.56)
B St i f f nes s damping c o e f f i c i e n t (eqn. 4. 56); s t r a i n - r a t e
parameter (eqn. 4.50)
B (B.j-) Lef t Cauchy-Green deformation tensor (eqn. 2.12)
B.:j Gradient operator f or Fi ni t e Elements (eqn. 3.6)
c Stress wave vel oci t y
C (Cj -j ki ) Const i t ut i ve tensor f or Jaumann rat e of Cauchy stress (eqn.
3.23)
C ( CJJ) Right Cauchy-Green tensor (eqn. 2.12)
C Damping matri x (eqn. 3.30)
(^ (Cj-jki ) Cons t i t ut i v e t ensor f or Tr uesdel l r at e of Cauchy st r ess
( sect . 3.5.2)
CST Constant st r ai n elements
d (d.j-j) Rate of deformation tensor (eqn. 2.15)
d (d.j ) Penetrati on i n contact (eqn. 4.68)
D ( DJJKL) Const i t ut i ve tensor ( t ot al Lagrangian) (eqns. 2.47, 3.21)
D Diameter
E Young's modulus of El ast i ci t y
E ( Ej j ) Green's st r ai n tensor (eqn. 2.17)
F Force; Yi el d f unct i on (eqn. 2.54)
F (F j ) Deformation gradi ent tensor (eqn. 2.7)
FD Fi ni t e Di f f erence (method)
FE F i n i t e Element (method)
G El as t i c shear modulus ( eqn. 2. 45)
g (g-j-j) Met r i c t ensor ( eqn. 2. 2)
h, h
a
, h
Y
, h' Hei ght ; Pl as t i c hardeni ng modul i ( eqns. 4. 44, 4. 45)
h ( h
i
) Heat f l ow r at e ( eqn. 2. 32)
I dent i t y tensor
Inner diameter
Jacobian of motion (eqn. 2.10)
St i f f ness, st i f f ness matri x (eqns. 3.10, 3.11)
Length
Vel oci t y gradi ents (eqn. 2.14)
Mass, mass matrices (eqn. 3.8)
Mixed Di scr et i zat i on (sect . 4. 2. 2)
MTB(C) Mixed Tetrahedra-Bri ck (Corrected) elements
I
ID
J
K,K
L
1 Oi
m , M , m
MD
j >
i,M
11
MTQ(C)
N,
n
N
OC
P
q
r,
R
s,
s
s
t
T
u
u
U
n
i
R
( Ri j )
s
(
S
I J )
( s
i d
)
( U- j )
( UI J )
Mixed Tr i angl es- Quadr i l at er al (Corrected) elements
Normal vectors
Time i nst ant corresponding t o n t
Shape f unct i ons f or FE (eqn. 3.5)
Outer Diameter
I nt er nal forces
Body heat supply
Radius; External force (eqn. 3.8)
Rotati on tensor (eqn. 2.11)
Surface; Distance along a curve
2nd Pi ol a- Ki r chhof f stress tensor (eqn. 2.25)
Cauchy devi at or i c stresses (eqn. 2.59)
Time; Thickness
Temperature
I nt ernal energy
Displacements
Right st r et ch tensor (eqn. 2.11)
V Volume
V (Vi-j) Left stretch tensor (eqn. 2.11)
v,v (v.j ) Velocity (eqn. 2.6)
w (w-j ^) Spin tensor (eqn. 2.15)
W Work, Energy
x (x
n
- ),
(x,y,z) Spatial coordinates (eqn. 2.5)
X Particle (sect. 2.2.1)
X (Xj) Lagrangian coordinates (eqn. 2.4)
X Vector product (eqn. 2.30)
V Yi el d stress
a Thermal expansi on c o e f f i c i e n t ; Back st r ess (eqn. 2.61);
Mixed Di scr et i zat i on cor r ect i on coef f i ci ent (eqn. 4.19)
(3 Pr opor t i on of c r i t i c a l dampi ng; Radial ret urn coef f i ci ent
(eqn. 4.32)
T Pl ast i c fl ow ar bi t r ar y mul t i pl i er (eqn. 2.55)
A Increment
5-j -; Kronecker del t a
eP Ef f ect i ve pl ast i c st r ai n (eqn. 2.60b)
e(e-j-j) Small st r ai n tensor (eqn. 2.43)
X Lame's El ast i c constant
\i Coef f i ci ent of Coulomb f r i c t i o n (eqn. 7.11)
12
v
P
9
a (cr
i : j
)
d
x
0)
Poisson' s r at i o
Mass densi t y; Radius of curvature
Angular coordi nate
Cauchy stress tensor (eqn. 2.24)
Par t i al der i vat i ve
Pul l -back, push-forward of tensors (eqns. 2. 21, 2.22^
Mesh coordi nates ( sect . 3.4)
Angular frequency
13
CHAPTER 1
INTRODUCTION
1.1 OBJECTIVES
1.2 NON-LINEAR MODELLING
1.3 LAYOUT
14
1.1 OBJECTIVES
Thi s work consi st s of a t h e o r e t i c a l par t ( mat hemat i cal and
numer i cal model s, chapt er s 2- 4) , and a p r a c t i c a l par t ( appl i c at i ons
and numer i cal s i mul at i on of l ar ge s t r a i n t ube col l apse anal y s i s ,
chapters 5-7).
The mot i vat i on f or the t heor et i cal part of the work l i es i n the
author' s i nt er est i n non-l i near sol i d mechanics model l i ng, understood
broadly as encompassing the f ol l owi ng phenomena:
- l arge st r ai ns and l arge displacements (geometric nonl i near i t i es) ;
- pl ast i c and vi scopl ast i c behaviour (materi al nonl i near i t i es) ;
- contacts and impact (nonl i near boundary condi t i ons) ;
- thermomechanical coupl i ng.
On the pr act i cal si de, the source of mot i vat i on was the research
program on t ube col l apse mechanisms bei ng c ar r i ed out at t he Ci v i l
Engi neer i ng depart ment of Ki ng' s Col l ege, Uni v er s i t y of London
(Andrews, England and Ghani , 1983). Such mechanisms are e f f i c i e n t
energy di ssi pat i ng systems (Johnson and Reid, 1978), f or use i n impact
si t uat i ons. Addi t i onal l y, t ubesar e frequent st r uct ur al components f or
aerospace vehi cl es and other equipment or components which may suf f er
acci dental c o l l i s i o n s .
The obj ect i ve of t h i s work was t he devel opment of numer i cal
methods of si mul at i on f or nonl i near anal y s i s , capabl e of model l i ng
tube col l apse mechanisms. More s p e c i f i c a l l y , t he at t ent i on was
r e s t r i c t e d t o col l apse t hr ough axi symmet r i c sequent i al f o l d i n g
(Concertina mode). Numerical predi ct i ons f or tube col l apse should be
obtained and compared t o experimental r esul t s, avai l abl e from previous
work on al umi ni um tubes by Ghani (1982). Thi s obj ec t i v e posed some
i mportant chal l enges, such as the development of a numerical model f or
l arge st r ai ns and l arge di spl acements, wi t h el ast i c- pl ast i c behaviour,
capabl e of model l i ng a r b i t r a r y t ube- t ube and t ube- pl at en cont act s
(chapt ers 2, 3, 4). Rel i abl e data woul d have t o be obt ai ned f or t he
15
TIME = 0.00 ms
TIME = 1.37 ms
TIME = 0.77 ms
TIME = 2.09 ms
3=
TIME = 2.93 ms TIME = 3.62 ms
i g u r
1, TYPICAL RESULTS FOR AXISYMMETRIC TUBE COLLAPSE ANALYSIS
0D = 38..1mm.t.= 1.22mm.L=88.9
mm
- TUBE9, GEOMETRY F (SEE TABLE 7.4)
16
c o n s t i t u t i v e behavi our of t he Al umi ni um at l ar ge s t r ai ns ( chapt er 6) .
Wi t h t h i s o b j e c t i v e i n mi nd, a F i n i t e Di f f e r e n c e met hod w i t h
e x p l i c i t t i me i n t e g r a t i o n was chosen. The method uses a F i n i t e El ement
t o p o l o g y , and can t h e r e f o r e be a p p l i e d t o i r r e g u l a r meshes i n
a r b i t r a r y cont i nua. An e f f i c i e n t cont act l ogi c was essent i al f or t he
success of t he s i mul at i ons . A t y pi c al exampl e of t he r es ul t s obt ai ned
i s shown i n f i gu r e 1. 1.
1.2 NON-LINEAR MODELLING
Wi t h i n t he past decade t h e r e has been c o n s i d e r a b l e i n t e r e s t i n
n o n l i n e a r s o l i d mechani cs s i mu l a t i o n s , due t o t he gr eat p r o b l e m-
sol vi ng power av ai l abl e f r om t he new gener at i ons of d i g i t a l comput er s.
The p o s s i b i l i t y of det ai l ed s o l u t i o n s f o r h i g h l y compl ex n o n - l i n e a r
probl ems has occassi oned renewed i n t e r e s t and pr essur e f or power f ul
mat hemat i cal descr i pt i ons and numer i c al t ec hni ques whi ch i mpl ement
them e f f i c i e n t l y .
In many aspect s of non- l i near numer i cal model l i ng, choi ces are
a v a i l a b l e : e x p l i c i t or i m p l i c i t t i me i n t e g r a t i o n , Lagr angi an or
Eul er i an meshes, t o t a l Lagrangi an or Cauchy st r ess f or mul at i ons . Each
of t hes e c hoi c es has i t s own advant ages and dr awbacks. On t h e o t h e r
hand, non- l i near mechani cs i s a f i e l d under const ant devel opment , and
new approaches are bei ng expl or ed whi ch at t empt t o combi ne e f f i c i e n t l y
t he advant ages of d i f f e r e n t t e c h n i q u e s ( e. g. El ement By El ement
method, Ar b i t r a r y Lagrangi an Eul er i an des c r i pt i ons ) .
The procedure chosen f or t h i s work ( Ex p l i c i t F i n i t e Di f f er ence)
i s i d e a l l y s u i t e d f o r s t e e p l y n o n - l i n e a r , s h o r t d u r a t i o n t r a n s i e n t
phenomena (wave pr opagat i on t ype pr obl ems) . A nat ur al appl i c at i on i s
f o r i mpac t s c e n a r i o s , e. g. t he mi s s i l e i mpac t t e s t s done at UKAEA
Wi n f r i t h r e p o r t e d by Barr ( 1983a) ( s e c t . 7. 4) . E x p l i c i t F i n i t e
Di f f er ence t echni ques are al so usef ul f or sl ow l oadi ng phenomena, i n
order t o t ake advantage of t he non- l i near robust ness and c a p a b i l i t i e s ,
t hrough t he use of v e l o c i t y scal i ng or dynamic r el ax at i on ( chapt er s 4,
6, 7) .
17
1.3 LAYOUT
In chapter 2 a number of essent i al sol i d mechanics concepts are
i nt r oduced and di scussed b r i e f l y . Non- l i near numer i cal models and
t echni ques t o i mpl ement t hose concept s i nt o numer i cal codes are
revi ewed i n chapt er 3. The Ex p l i c i t Fi n i t e Di f f er ence model and
computer code developed here are described i n chapter 4, whi l e chapter
5 contai ns some val i dat i on examples which t est the main aspects of the
f or mul at i on. Chapter 6 concerns t he der i v at i on of a mat er i al
const i t ut i ve law f or Aluminium al l oy through t ensi l e t est s, wi t h some
appl i c at i ons t o t he numer i cal s i mul at i on of t he t e n s i l e t est s
themselves. Chapter 7 contai ns appl i cat i ons t o tube col l apse anal ysi s,
comparing the r esul t s wi t h experimental data f or quasi - st at i c col l apse
of Al umi ni um t ubes. The c o n s t i t u t i v e law f rom chapt er 6 i s used f or
the numerical pr edi ct i ons. An anal ysi s f or a medium vel oci t y (176 m/s)
t ube i mpact i s al so descr i bed. F i n a l l y , concl usi ons and some
suggestions f or f ur t her work are given i n chapter 8.
CHAPTER 2
CONTINUUM MECHANICS DESCRIPTIONS
2.1 INTRODUCTION
2.2 KINEMATICS
2.2.1 Configurations
2.2.2 Deformation tensors
2.2.3 Deformation and spin rates
2.2.4 Strains
2.2.5 Transformations
2.3 STRESS
2.3.1 Cauchy
2.3.2 Piola-Kirchhoff
2.4 BALANCE LAWS
2.4.1 Balance of mass
2.4.2 Balance of momentum
2.4.3 Balance of energy
2.5 CONSTITUTIVE RELATIONS
2.5.1 Rate equations
2.5.2 Elasticity
2.5.2.1 Hyperelastic materials
2.5.2.2 Hypoelastic materials
2.5.3 Plasticity
2.5.3.1 Von Mises model
2.5.3.2 Other plasticity models
19
2.1 INTRODUCTION
The advances made i n di gi t al computing wi t hi n the l ast decades
have opened up new f i e l d s f or engi neer s and s c i e n t i s t s . Problems
previ ousl y regarded as unsol vabl e, only approached through experiments
and si mpl i f i ed empi ri cal f ormul ae, can now be analyzed numeri cal l y i n
gr eat d e t a i l . In t he f i e l d of cont i nuum mechanics t h i s has gr eat l y
increased the i nt er est i n det ai l ed mathematical descr i pt i ons, amenable
t o be used i n numer i cal models wi t h d i s c r e t i z a t i o n t echni ques (e.g.
Fi ni t e Element or Fi ni t e Di f f erence methods).
Having sai d t h i s , t her e s t i l l ex i s t s a c er t ai n degree of
c o n f u s i o n i n t he s p e c i a l i s t l i t e r a t u r e . On t he par t of t he
mathematicians, ri gorous mechanical descri pt i ons are often presented
i n ways d i f f i c u l t t o be grasped by engi neer s and i mpl ement ed i n
numerical producti on codes. As a r esul t , many engineers s t i l l cl i ng on
t o out dat ed and much l ess power f ul not at i ons. On t he ot her hand,
t heor et i c al pr esent at i ons are not uni que, causi ng some degree of
confusi on t o researchers f i r s t approaching ser i ousl y these t opi cs.
An e f f o r t has been made i n t h i s chapt er t o present a br i ef
overview of cer t ai n continuum mechanics concepts, i ndi spensabl e i n a
ri gorous t reat ment , wi t hout unnecessary mathematical f uss. The purpose
of t hi s exposi t i on i s :
- t o i nt r oduce t he nomencl at ure and d e f i n i t i o n s of concept s used i n
l at er chapt ers;
- t o di scuss t he s i gni f i c anc e of and i n t e r p r e t some concept s wi t h a
view t o numerical model l i ng (basis of t hi s work);
- t o ensure c er t ai n compl et eness f or t he i deas pr esent ed i n t h i s
t hesi s.
I t must be st r essed, however, t hat t h i s exposi t i on does not
pretend t o be complete. Only the concepts which are rel evant f or the
r est of t h i s t hesi s wi l l be dwel t upon. In p a r t i c u l a r , emphasis i s
l a i d on s ol i d mechanics and el a s t i c - p i a s t i c behavi our . A number of
r es ul t s wi l l be present ed wi t hout pr oof . For a more compl et e
20
di scussi on of these t opi cs, the i nt er est ed reader i s r ef er r ed t o Fung
(1965), Malvern (1969), and Bi l l i ngt on and Tate (1981) f or the general
concept s, and t o Marsden and Hughes (1983) f or a more det ai l ed
and up-t o-dat e mathematical descr i pt i on.
In the f ol l owi ng pr esent at i on, the ambient space i s assumed t o be
an Eucl i dean poi nt space ( i . e. i n t e r i o r product def i ned) , and where
necessary t hi s wi l l be par t i cul ar i zed t o R3. The coordi nate bases may
be c u r v i l i n e a r and a r b i t r a r y , al t hough when equat i ons are gi ven i n
component f orm, of t en orthonormal bases (not necessar i l y car t esi an)
are assumed f or si mpl i ci t y. The usual conventions f or tensor not at i on
are empl oyed: repeat ed i ndi ces i ndi c at e summati on over t h e i r range
unless ex pl i c i t l y st at ed, and commas i ndi cat e covari ant der i vat i ves.
Vectors and tensors are represented by boldface charact ers. Superposed
dots i ndi cat e materi al ti me der i vat i ves.
Given two t ensor s A and B t he product AB i s under st ood t o be
cont r act i ng the near i ndi ces wi t h opposite vari ance:
( AB)
i j
= A
i k
B
k
j
(2. 1)
I f t he i ndi ces have t he same var i ance, e.g. both are cont r avar i ant ,
the metri c tensor g i s necessary t o lower one:
(AB)
1J
- A
l k
g
k l
B
1 j
= A
l k
8
k
j
(2.2)
When t he t ensor components are r ef er r ed t o or t honor mal bases t he
ver t i cal posi t i on of the i ndi ces i s i r r el evant , as the metri c tensor
i s uni t y.
A colon i ndi cat es doubly contracted product:
A:B = A ^ B ^ (2. 3)
2.2 KINEMATICS
A body (or cont i nuum) i s a set whose el ement s, c al l ed mat er i al
p a r t i c l e s , have a one- t o- one correspondence wi t h a r egi on V of t he
21
Euclidean point space. The following kinematics concepts are intended
to provide a description of the motion of deformable bodies.
2.2.1 CONFIGURATIONS
Each particle X of the body B may be identified by its position X
in the original configuration, V
0
, which is taken as reference:
X = k(X) (2.4)
X (components Xi) are cal l ed Mat eri al or l agrangi an coordi nates of the
par t i cl e. The motion of the body at a l at er t i me i s given by the t i me-
dependentpositions x of the par t i cl es i n the current conf i gur at i on,
V:
x = x ( X, t ) ( 2 . 5 )
x (components xi ) are the spat i al or Eul eri an coordi nat es. Hereafter
upper case i ndi ces shal l r ef er t o Lagrangi an coor di nat es, and l ower
case t o Eul er i an. The vel oci t i es are defi ned as
: 2. 6'
where the dot si gni f i es a mat eri al ti me der i vat i ve, i .e. f ol l owi ng the
par t i cl e X.
2.2.2 DEFORMATION TENSORS
Central to deformation measurements is the deformation gradient
tensor:
F = dx/dX (2.7)
wi t h components F
n
j = x
1
j .
The t ensor F i s used as t he base f or a number of s t r a i n and
def or mat i on measures. An el ement of a curve dX i s t r ansf or med by
22
dx = FdX. The inverse of F gives the spatial gradients of the material
coordi nates:
F"
1
= dX/dx
(2.9)
( ( F "
1
)
1
! = X
!
s i
)
F const i t ut es a t wo- poi nt tensor. Another i nt er pr et at i on t hat rel at es
F t o t ransf ormat i ons between conf i gur at i ons i s given i n secti on 2.2.5.
The Jacobian of the motion i s
J = det(F) (2.10)
The polar decomposition of F gives
F = RU = VR (2.11)
where R i s an or t hogonal ( r ot at i on) t ensor , RRT = I . U and V are
posi t i ve def i ni t e, and are cal l ed the r i ght and l ef t st r et ch tensors
r es pec t i v el y . Equat i ons (2.11) r epr esent t wo ways t o v i s ual i z e t he
def or mat i on: f i r s t s t r et c hi ng (U) and t hen r o t a t i n g (R), or f i r s t
r ot at i ng (R) and then st r et chi ng (V)
Other deformation measures are the Cauchy-Green t ensors:
C = F'F ( r i ght Cauchy-Green)
(2.12)
B = FF
1
( l ef t Cauchy-Green)
The l engt h of an el ement of curve i s gi ven by ds2 = dxdx i n t he
current conf i gur at i on, and dS2 = dXdX i n the or i gi nal conf i gur at i on.
The si gni f i cance of C and B i s given by the r el at i ons
ds2 = dXCdX
(2. 13)
dS2 = dxB-ldx
23
2.2.3 DEFORMATION AND SPIN RATES
The spatial velocity gradient tensor is defined as:
1 = dv/dx (2.14)
which can be decomposed i nt o symmetric and skew-symmetric par t s:
d = ( l +l T) / 2
(2.15)
w = ( l - l T) / 2
These are cal l ed the rate of deformati on (or vel oci t y st r ai n) and spin
rate tensors respect i vel y. The rat e of change of l ength of an element
of curve i s given by
ds = (dxddx)/ds (2.16)
2.2.4 STRAINS
A measure of t he t o t a l s t r a i n i s gi ven by t he Green s t r a i n
tensor, defi ned as
E = (C - I ) / 2 (2.17)
where I i s the I dent i t y t ensor. I t i s t r i v i a l t o see t hat
ds2-dS2 = 2dXEdX (2.18)
and t hat the rate of E i s given by
E = FTdF ( 2. 19)
24
2. 2. 5 TRANSFORMATIONS
For t he t ensor s def i ned above, some of t he i ndi ces r ef er t o t he
or i gi nal conf i gur at i on (upper case), whi l e others are rel at ed to the
current conf i gur at i on (lower case). Here some t ransf ormat i on laws are
gi ven t o f i n d t he c o r r e s p o n d i n g t ens or i n t he a l t e r n a t i v e
conf i gur at i on.
Although the transformed tensors wi l l be considered as di f f er ent
t ens or i al e n t i t i e s , one way t o v i s u a l i z e t he t r ans f or mat i on i s as a
mere change of base. I magi ne a base ( O. e^ f i x ed i n space t hr oughout
t he mot i on, and anot her base ( 0 ' ( t ), e ' i U ) ) whi ch def or ms and
t r ans l at es wi t h t he body. In t h i s convect ed c u r v i l i n e a r base, t he
coordi nates of a mat eri al poi nt remain constant throughout the moti on,
and equal t o the mat eri al coordi nat es, XI . The spat i al components of F
pr ovi de t he mat r i x f or t he change of coor di nat es between t he t wo
bases. Given a 2nd order cont r avar i ant t ensor a by i t s convect ed
mat eri al components, a_U, the spat i al components are
a
i j
= Ox Vd x MOx J / a x ^ a U
( 2
. 20)
Hence F provides a means f or t ransf ormi ng between spat i al and mat eri al
coordi nates, a"iJ and a_U are the components i n di f f er ent bases of the
same t ensor , a. I f now one assumes components a_U t o appl y t o t he
spat i al basi s, a new tensor i s obtai ned:
A = a_
IJ
ej Bej (2.21a)
where s si gni f i es a t ensor i al product. A i s cal l ed the pul l -back of a,
and may be obtained as
A = F
_1
aF"
T
= (d(a) (2.21b)
whi l e the push-forward i s defi ned by the i nverse r el at i on:
a = 0
t
*( A) = FAFT (2.22)
These r el at i ons may be t r i v i a l l y general i zed t o tensors of any rank.
25
El ement s of ar ea and vol ume i n r e f e r e n c e and c u r r e n t
conf i gurat i ons are transformed by the f ol l owi ng t r anspor t formul ae:
nda = JF"
T
NdA (2.23a)
dv - JdV (2.23b)
These r el at i ons may be used t o express i nt egr al bal ance l aws ( sect .
2.4) i n ei t her c o n f i g u r a t i o n . Eqn. (2.23a) condi t i ons t he f orm of t he
Pi ol a t ransf ormat i ons f or the stress tensor (eqn. 2.25)
2.3 STRESS
2.3.1 CAUCHY
The concept of st r ess r est s upon t he Cauchy post ul at e t hat t he
acti on of the rest of the mat eri al upon any volume element of i t i s of
t he same f orm as d i s t r i b u t e d sur f ace f or ces. A t r a c t i o n vect or t ( n )
may be d e f i n e d at each p o i n t , as t he f o r c e e x e r t e d per u n i t
i nf i ni t esi mal area, f or each or i ent at i on n.
Appl yi ng e q u i l i b r i u m consi der at i ons, i t may be deduced t hat a
stress tensor a must exi st , such t hat f or every or i ent at i on n
t ( n) = n a (2.24)
a i s cal l ed the Cauchy or t rue stress t ensor, and i t i s rel at ed
t o the current conf i gur at i on.
2.3.2 PIOLA-KIRCHHOFF
I f both t he f or ce and t he area components of t he concept of
st r ess are t r ansf or med back i nt o t he o r i g i n a l c o n f i g u r a t i o n , a new
stress tensor i s obt ai ned:
S = JF
-1
(7 F
_ t
= J0
t
* ( cr ) (2.25)
26
This r el at i on i s cal l ed the backward Pi ol a t r ansf or mat i on. I t defi nes
t he 2nd Pi ol a- Ki r c hhof f st r ess t ensor S, whi ch i s a st r ess measure
ref erred t o the or i gi nal conf i gur at i on.
2. 4 BALANCE LAWS
Balance laws (mass, momentum, angular momentum, and energy) may
be st at ed a l t e r n a t i v e l y i n i nt egr al f orm or as f i e l d equat i ons.
I nt egral forms provide "weaker" expressions f or the same pr i nci pl es.
This wi l l be commented f ur t her i n sect i on 4. 3.
2 . 4 . 1 BALANCE OF MASS
Conser vat i on of mass i mpl i es t hat t he mass of t he mat er i al
ocupyi ng a c er t ai n r egi on V of t he body remai ns const ant t hr oughout
the moti on:
(d/dt)fpdV = 0 (2.26)
where p is the mass density.
As a field equation , balance of mass is expressed by the
continuity equation:
p + pdiv(v) = 0 (2.27)
2 . 4 . 2 BALANCE OF MOMENTUM
For a region V of the body wi t h boundary S, the i nt egr al form of
the equation of l i near momentum balance i s
( d/ dt ) f pvdV = I pfdV + f nadS (2.28)
V V S
where f i s t he body f o r c e per u n i t mass. The c o r r e s p o n d i n g
27
di f f er ent i al expression i s Cauchy's equation of moti on,
pv = di v( a) + p f (2.29)
The i nt egr al s i n eqn. (2.28) i nvol ve vect or s, and as poi nt ed out
by Marsden and Hughes (1978), may not provi de a covari ant statement of
the momentum balance pr i nci pl e i n a general mani f ol d. However, f or the
Euclidean space t o which t hi s exposi t i on r ef er s, the obj ect i on i s not
r el evant . The i nt egr al expr essi on i s bet t er sui t ed f or f i n i t e
di f f erence numerical models (secti on 4.3), f or which weak var i at i onal
global expressions (as employed i n Fi ni t e Elements) are not obtai ned.
Taki ng moments i n eqn. (2.28) wi t h r espect t o t he o r i g i n , t he
balance of angular momentum i s expressed by
( d/ dt ) /p(xXv)dV = f p(xXf )dV + /xX(nc7)dS (2.30)
where xXv denotes vector product of x and v. The corresponding field
equation states simply the symmetry of O":
U = G
J
(2.31)
Symmetry of S may be deduced from eqns. (2.25) and (2.31).
2.4.3 BALANCE OF ENERGY
In a continuum, the first law of thermodynamics may be expressed
as
(d/dt) JpudV = / ( p q +a :d ) d V + fhndS (2.32;
V
where u is the internal energy per unit mass
q is the rate of body heat supply per unit mass
h is the heat flux vector; for an oriented infinitesimal area
the heat flow rate is given by H = hndS
The corresponding field equation is
28
pu = aid + pq + di v( h) ( 2. 33)
The t er m aid r epr esent s t he st r ess work per uni t volume and
t i me, a and d are sai d t o be conjugate stress and st r ai n measures. An
al t er nat i ve represent at i on of t he energy bal ance p r i n c i p l e i nvol ves
the use of S and E, also conj ugate:
p
0
u = S:E + p
0
q + &- di v( h) ( 2. 34)
2.5 CONSTITUTIVE RELATIONS
The bal ance l aws pr ovi de a set of equat i ons whi ch are not
suf f i ci ent t o determine the behaviour of a mat eri al body. Some f ur t her
equat i ons are necessary, s t at i ng t he r el at i on between ki nemati c and
dynamic vari abl es ( const i t ut i ve equati ons).
Const i t ut i ve equations are based on judgement, a- pr i or i knowledge
of how the mat eri al behaves. However, cer t ai n general pr i nci pl es must
be sat i sf i ed i n t hei r f or mul at i on. For our purpose, the most i mportant
p r i n c i p l e i s t hat of o b j e c t i v i t y , whi ch st at es t hat c o n s t i t u t i v e
equations must be i nvar i ant under changes of reference frame, i n order
t o represent the mat eri al behaviour obj ect i vel y.
For a homogeneous mat eri al i t may be seen ( Bi l l i ngt on and Tate,
1981) , t h a t an o b j e c t i v e r e l a t i o n bet ween Cauchy s t r e s s and
deformation takes the form:
a = R p( C
t
( s) , T
t
( s) ) R
T
(2.35)
where R i s the r ot at i on tensor (eqn. 2.11) and T the temperature. The
not at i on Ct ( s) s i g n i f i e s t he hi s t or y of C (eqn. 2.12) f rom - o<s<t .
Note t hat i n general the complete hi st or y of the deformati on C (or of
E equi val e n t l y , eqn. (2.17)) are r equi r ed, whi l e f or R onl y t he
instantaneous current value i s used, f or r ot at i ng the stresses.
The second Pi ol a- Ki r chhof f stress S i s obj ect i ve as such (being
r el at ed t o a f i x ed r ef er ence c onf i gur at i on) . In t erms of i t eqn.
29
(2.35) may be rephrased as
S -
B
( Ct ( s ) , Tt (
s
) ) (2.36)
The advantage of usi ng the 2nd Pi ol a- Ki r chhof f stress tensor f or t ot al
formul ati ons i s evi dent (see also sect . 3. 5. 1) .
Some t ypes of El as t i c and Pl as t i c r at e equat i ons are di scussed
below. For si mpl i ci t y, at t ent i on i s centred on the isothermal case.
2. 5. 1 RATE EQUATIONS
Mat eri al s wi t hout memory or wi t h smooth memory may be described
wi t h rate equati ons, e. g.
a= g( d, a, F) (2.37)
where a i s a stress rate which i s obj ect i ve f or r i gi d body r ot at i ons.
The choi ce of obj ec t i v e r at e i s not uni que. A v ar i et y of opt i ons are
avai l abl e, the two most wi del y used being the Jaumann rate
a = a + a w + wTa (2.38)
and the Truesdel l r at e,
G = <7 - ( j l
T
- 1(7+ a t r ( l ) (2.39)
In equat i on (2.37) a pr ovi des t he c o n s t i t u t i v e par t of t he
st r ess i ncr ement . Eqns. (2.38) or (2.39) def i ne t he r emai ni ng t er ms
t hat must be added f or o b j e c t i v i t y . For mul at i ons based on ei t her
obj ect i ve rate may be made equi val ent by adj ust i ng the const i t ut i ve
l aw, g. However, i f g i s post ul at ed a - p r i o r i , i ndependent l y of t he
choi ce of obj ect i ve r at e, both f or mul at i ons gi ve r i s e t o d i f f e r e n t
const i t ut i ve behavi our.
The need f or speci al obj ect i ve rates i s avoided i f the equations
are formul ated i n a materi al set t i ng, e. g.
30
S = g.(E,E,S) (2.40)
S i s the mat eri al t i me rate of a tensor on the current conf i gur at i on
(2nd Pi ol a- Ki r chhof f ) , which i s already obj ect i ve.
2.5.2 ELASTICITY
El ast i c mat er i al s are t hose f or whi ch a n a t u r a l , s t r es s - f r ee
st at e exi st s, t o which the body returns upon removal of al l external
f or ces. The stress depends on the deformation from t hi s natural st at e:
S = f ( C, t ) (2.41)
A perfect memory of the natural st at e, wi t h no memory of i nt ermedi at e
st at es, i s exhi bi t ed.
For l i near e l a s t i c i t y and smal l s t r ai ns t he r e l a t i o n i s as
fol 1ows:
o = c:e (2.42)
(in component form er
1J
= c ^-j e*')
c is termed the elasticity tensor, and e is the small strain tensor:
e
i
ij
=
(
u
i,j
+ u
j , i >
/ 2
(
2
'
43
>
where u are displacements. For i sot r opi c mat er i al s, and provided a
and e are both symmetric, c must take the form
c
i j k l " X i j
5
k l
+ 2 G
5
i k
S j l (2.44)
where X and G are c al l ed Lame's const ant s. Thi s gi ves r i se t o t he
cl assi c general i zed Hooke's l aw:
ff
ij = ^ k k
5
i j
+ 2G e
i j (
2
-
4 5
)
31
2. 5. 2. 1 HYPERELASTIC MATERIALS
The concept of Hyper el ast i ci t y was i ntroduced by Green and given
i t s present name by Tr uesdel l (e.g. Tr uesdel l and Toupi n, 1960). I t
postul ates the exi stence of a st rai n-energy f unct i on from which the
stresses may be deri ved as
S = ft, ( dW/ dE) (2.46)
Assuming t he necessary d i f f e r e n t i a b i 1 i t y , t he e l a s t i c i t y t ensor i s
defined as
D = P
0
(d^/d^) (2.47)
and a rate equation may be written as
S = D:E (2.48)
(i n component form S = D \ \ E )
For a const ant val ue of D, a l i near hyper el ast i c t o t a l equat i on
i s obtai ned:
S = D:E (2.49)
2.5.2.2 HYPOELASTIC MATERIALS
The term Hypoel asti c, also i ntroduced by Truesdell (Truesdell and
Toupi n, 1960), char act er i zes a mat er i al f or whi ch t he behavi our i s
def i ned i n t he cur r ent c onf i gur at i on by an i ncr ement al l y l i near
r el at i onshi p of the form:
0= c:d (2.50)
(i n component form &
1J
= c
1J
k
-]d )
An obj ect i ve st r ess r at e must be used f or eqn. (2.50) (see sect i on
32
2 . 5 . 1 ) .
Hypoelastic behaviour i s very convenient f or descri pt i ons based
on t he cur r ent c o n f i g u r a t i o n . Mat er i al data based on t r ue st r ess -
natural st r ai n r el at i onshi ps (see secti on 6.1) give r i se nat ur al l y t o
hypoel asti c i nt er pr et at i ons.
For i sot r opi c mat eri al s eqn. (2.50) takes the form
a = \ t r ( d ) I + 2Gd (2.51)
( i n orthonormal components o^- = A d ^ 8^ + 2Gd^ )
2.5.3 PLASTICITY
For most sol i ds, behaviour may be assumed el ast i c only wi t hi n a
c er t ai n st r ess range. Beyond t he e l a s t i c r ange y i e l d o c c u r s ,
def or mat i ons bei ng char act er i zed by permanent changes occasioned by
sl i p or di sl ocat i ons at the atomic l evel ( Pl ast i c f l ow) .
Af t er y i e l d , El as t i c and Pl as t i c def or mat i ons are assumed t o
happen concur r ent l y ( El a s t i c - Pl a s t i c mat er i al s ) . More r e s t r i c t i v e
i d e a l i z a t i o n s are pr ovi ded by r i g i d - p l a s t i c models (onl y p l a s t i c
deformati ons). An addi t i ve decomposition of the rate of deformati on i s
assumed here:
d = d
e
+ dP (2.52)
where super i ndi ces e and p i ndi c at e e l a s t i c and p l a s t i c components
r es pec t i v el y . Addi t i ve decomposi t i on of st r ai ns i n t hi s fashion was
proposed by Hi l l (1950). Lee (1969) has proposed a mu l t i p l i c a t i v e
decomposition of deformati on gradi ents i nst ead, F = Fepp, whi l e Green
and Naghdi (1965) have advocated an addi t i v e decomposi t i on of t o t a l
s t r a i n , E = Ee + EP.
Cl assi cal p l a s t i c i t y i s f or mul at ed i n t erms of t he cur r ent
c onf i gur at i on ( H i l l , 1950). Hence t he popul ar i t y of an addi t i v e
decomposition of the rates of def ormat i on, eqn. (2.52), coupled wi t h
33
hypoel asti c behaviour, f or El ast i c- Pl ast i c mat eri al descri pt i ons (e.g.
Wi 1 kins (1964), Hi bbi t , Marcal and Rice (1970)). In t hi s case
<7= c : d
e
= c: ( d- dP) ( 2. 53)
The y i el d c r i t er i on determines the l i mi t of the el ast i c range:
F( a, Q) = 0 ( 2. 54)
where Q i s a set of pl ast i c hardening parameters. For F<0 the mat eri al
behaves el ast i cal l y. Two addi t i onal sets of r el at i ons must be provided
t o determine f ul l y the st r ess- st r ai n behaviour:
- Flow rul e dP = 7R( a,Q) (2.55a)
- Hardening rul e Q= 7H( a ,Q) (2.55b)
where 7 i s an ar bi t r ar y mul t i pl i er , whose value i s determined from
o
the simultaneous sol ut i on of eqns. (2.54), (2.55). An obj ect i ve rat e Q
must be used i n eqn. (2. 55b).
Drucker (1951) post ul at ed a c r i t er i on f or st abl e work-hardening
mat er i al s. This i nvol ves the work done by a set of sel f - equi l i br at i ng
f or ces, r equi r i ng:
a: d > 0 ( 2. 56a)
a: d
p
^ 0 ( 2. 56b)
The equal si gn i n (2.56b) hol ds f or p e r f e c t l y p l a s t i c mat er i al s (no
hardening). A consequence of Drucker's post ul at e i s the associ at i vi t y
of pl ast i c f l ow: f or a smooth part of the y i el d surf ace,
d
p
= 7 ( d F / d a ) ( 2. 57)
whi ch i n a ni ne- di mensi onal st r ess space may be i nt er pr et ed as t he
normal i ty of dP t o the surface F( c, Q) .
34
2. 5. 3. 1 VON MISES MODEL
Par t i cul ar l y usefu] and si mpl e models are deri ved from the y i el d
cr i t er i on of Von Mises (1913). This may be wr i t t en as
F = ( 3/ 2) s: s - Y2 = 0 (2.58)
where s are the deviatoric Cauchy stresses,
s = a - (l/3)tr(a)I (2.59a)
( i n orthonormal components, s ^ = a^- - (1/ 3) o- ^ 5 . . (2.59b)
Y i s t he y i e l d st r engt h of t he ma t e r i a l , whi ch coi nci des wi t h t he
y i e l d st r ess i n uni ax i al t ensi on (see sect i on 6.1). The Von Mises
y i e l d c ondi t i on i s i ndependent of vol umet r i c st r esses, whi ch are
assumed to behave el as t i c al l y .
An i sot r opi c hardening model i s obtained by making Y a f unct i on
of the ef f ect i ve pl ast i c s t r ai n,
e
P:
Y = Y(e
p
) (2.60a)
wi t h
e
P = J 6 e
p
= J A 2/3) dp: dP dt (2.60b)
A more general hardening model, i ncor por at i ng Bauschinger ef f ect ,
may be obtained by combining i sot r opi c hardening wi t h the ki nemati c
hardeni ng proposed by Prager (1956) and Zi egl er (1959), gi vi ng t he
yi el d condi t i on
F = ( 3 / 2 ) ( s - a ) : ( s - a ) - Y2 (2.61)
o: i s c al l ed t he back- st r ess and r epr esent s a ki nemat i c hardeni ng
parameter ( t r ansl at i on of the Von Mises c i r c l e) . The associ at i ve f l ow
ru 1 e i s
dP = 7 ( s - a ) (2.62)
35
and the hardening laws,
a= (2/3)h
a
dP (2.63a)
Y = h
y
eP (2.63b)
I mposi ng t he consi st ency c ondi t i on (F = 0) dur i ng l oadi ng, and
combi ni ng eqns. (2. 51), (2. 53), ( 2. 61) - ( 2. 63) , t he s t r e s s - s t r a i n
r el at i on i s found t o be
a = c:Ld - ( s - a ) ( 3 / 2 ) d : ( s - a ) / Y2 ( i +h ' / 3 G ) J (2. 64)
where h' = hy
+ n
a ( pl as t i c modul us). Pur el y i s ot r opi c hardeni ng i s
obtained wi t h h
a
=0, and purel y kinematic wi t h h
Y
=o.
In a uni ax i al t e s t , law (2.64) w i l l pr ovi de an El as t opl as t i c
hardening modulus of
h = 1/L1/E + 1/h' J (2.65)
where E - G(3 X+2G)/( X +G ) (Young's modulus of elasticity).
2.5.3.2 OTHER PLASTICITY MODELS
Pl a s t i c i t y i n s o i l s i s gener al l y consi derabl y more compl i cated
than the above Von Mises model. Pressure dependent y i e l d , ani sot ropy,
d i l a t a t i o n and n o n - a s s o c i a t i v i t y , hyst er et i c cycl i c behaviour, pore
pr essur e, are i mpor t ant f eat ur es f or s oi l p l a s t i c i t y . An excel l ent
review of t hi s t opi c has been given by Marti and Cundall (1980).
The mat hemat i cal t heor y of p l a s t i c i t y i s a f i e l d s t i l l under
development. Very ref i ned phenomenological models have been proposed
(e.g. Mroz (1967), Prevost (1978)). These models are based on
mul t i pl e- s ur f ac e i deal i z at i ons . They provi de el aborate st r ess- st r ai n
l aws r equi r i ng consi der abl e c omput at i onal c os t f o r numer i c al
model l i ng, t hus i n pr ac t i c e t hey are har dl y used. Thi s f ac t has been
acknowl edged by Or t i z and Popov (1983), who propose s i mpl er , one-
surface models f or metal pl as t i c i t y .
36
F i n a l l y , a p r i n c i p l e t o be hel d pr esent when choosi ng a
p l a s t i c i t y model , i s t h a t i t can onl y be as r e l i a b l e and as
s o p h i s t i c a t e d as t he e x p e r i me n t a l i n f o r ma t i o n on whi ch t he
det ermi nat i on of the model parameters i s based. For example, there i s
l i t t l e poi nt i n usi ng anyt hi ng ot her t han a Von Mises i s ot r opi c
hardeni ng model i n a met al , i f a l l t he i nf or mat i on av ai l abl e i s a
uni axi al st r ess- st r ai n law. On the other hand, the added compl i cat i on
of some models may not be necessary i f t he l oadi ng i s mai nl y
monotonic.
CHAPTER 3
NONLINEAR NUMERICAL MODELS FOR SOLID MECHANICS
3.1 INTRODUCTION
3.2 FINITE DIFFERENCE METHODS
3.3 FINITE ELEMENT METHODS
3.4 MESH DESCRIPTIONS
3.4.1 Lagrangian
3.4.2 Eulerian
3.4.3 Arbitrary Lagrangian-Eulerian
3.5 LARGE DISPLACEMENT FORMULATIONS
3.5.1 Total Lagrangian
3.5.2 Cauchy stress - velocity strain
3.5.3 Updated Lagrangian
3.6 TIME INTEGRATION
3.6.1 Central difference (explicit)
3.6.2 Trapezoidal rule (implicit)
3.6.3 Operator split methods
3.7 PRACTICAL CONSIDERATIONS FOR DISCRETE MESHES
3.7.1 "Locking up" for incompressible flow
3. 7. 2 " Hour gl assi ng"
3.8 CONCLUSIONS
38
3.1 INTRODUCTION
The Governi ng equat i ons i n s o l i d mechanics are t he equat i ons of
mot i on, whi ch can be wr i t t e n i n component f or m as
a
i j 5 j
+ p ( f
i
- i i
i
) = 0 ( 3. 1)
where cr^j - j
s
t he Cauchy st r ess t ensor , p t he mass densi t y, f-j are body
f or ces per u n i t mass ( t y p i c a l l y g r a v i t y ) , and ui di spl acement s. These
equat i ons o r i g i n a t e f r om t he bal ance of momentum p r i n c i p l e ( sect i on
2. 4. 2) . Thi s p r i n c i p l e may be s t a t e d a l t e r n a t i v e l y i n i n t e g r a l f or m
(eqn. 2. 28) .
The p a r t i a l d i f f e r e n t i a l eqns. of mot i on ( 3. 1) depend upon 3
space and 1 t i me v a r i a b l e s . The numer i c al model s d e s c r i b e d her e
per f or m i ndependent s emi di s c r et i z at i ons i n space and t i me. Fi r s t eqns.
( 3. 1) ar e d i s c r e t i z e d i n space, y i e l d i n g a syst em of o r d i n a r y
d i f f e r e n t i a l eqns. i n t i me . These ar e t hen i n t e g r a t e d wi t h a t i m e -
st eppi ng pr ocedur e.
The d i s c r e t i z a t i o n of t he cont i nuum may be achi eved ei t her wi t h
F i n i t e El ement (FE) or F i n i t e Di f f e r e n c e (FD) met hods. Bot h met hods
have had s epar at e h i s t o r i c a l dev el opment s , a l t h o u g h some degr ee of
c o n v e r g e n c e has been r e a c h e d l a t e l y i n t h e l i t e r a t u r e ( e . g .
Be l y t s c h k o , 1983) . The t h e o r e t i c a l p r i n c i p l e s f o r bot h met hods ar e
d i f f e r e n t : l ocal t r unc at i on er r or s f or FD, gl obal er r or norms f or FE.
However, FE methods are al so based on i ndependent shape f unct i ons f or
each el ement . As a r e s u l t , FE and FD f o r mu l a t i o n s o f t e n pr oduce
equi val ent al gor i t hms (e.g. Kunar and Mi nowa, 1981).
Ot her numer i c al met hods need onl y a d i s c r e t i z a t i o n i n t he
boundar y: t he Boundar y El ement Met hods (BEM). These wer e f i r s t
pr oposed f o r s o l i d mechani cs by Ri zzo ( 1967) and Cr use ( 1969) .
C o n s i d e r a b l e a d v a n t a g e can be g a i n e d by t h e r e d u c t i o n and
s i mp l i f i c a t i o n of t he d i s c r e t i z a t i o n . For nonl i near pr obl ems, however,
BEM l ose much of t h e i r appeal . Volume i n t e g r a l s appear whi ch r equi r e
an addi t i onal d i s c r e t i z a t i o n of t he cont i nuum (e.g. Gar ci a, 1981). For
t h i s reason BEM w i l l not be r evi ewed here.
39
Time i nt egr at i on may be perf ormed ei t her by modal anal ysi s
methods or by di r ec t i n t e g r a t i o n ( t i me- mar chi ng) . Modal anal ysi s
r equi r es t r ansf or mat i ons i nt o t he f requency domain whi ch are onl y
val i d i n a l i near regime, f or which reason they must be rul ed out f or
nonl i near models. As t o ti me-marchi ng procedures two main al t er nat i ves
e x i s t , e x p l i c i t or i mp l i c i t methods. Both have advantages and
disadvantages, which wi l l be reviewed br i ef l y i n t hi s chapter. Recent
al t er nat i ve procedures based on operator s pl i t t i ng methods wi l l also
be consi dered.
3.2 FINITE DIFFERENCE METHODS
Fi ni t e Di f f er ence methods have been used f or a l ong t i me by
engineers wi t hi n r el axat i on procedures (e.g. Sout hwel l , 1940). Fi ni t e
Di f f er ence oper at or s pr ovi de l ocal appr oxi mat i ons f or a system of
coupl ed d i f f e r e n t i a l equat i ons. Due t o t h i s f a c t , a one- st ep gl obal
sol ut i on i s not possi bl e and recourse must be made t o r el axat i on and
i t er at i ve techniques. Addi t i onal l y , FD methods have been associ at ed
nor mal l y wi t h r egul ar zoni ng (at l east t o p o l o g i c a l l y r egul ar ) . For
these two reasons, FD methods were ecl i psed by the Fi ni t e Element boom
i n the 1960's f or st r uct ur al and sol i d mechanics appl i cat i ons. FD has
al ways been popul ar , however, i n ot her areas such as Eul er i an f l u i d
mechanics (e.g. Ni chol s, Hi r t , and Hotchki ss, 1980).
For a r e g u l a r mesh wi t h " I " and " J " l i n e s al ong t he t wo
coordi nate di r ect i ons, standard f i n i t e di f f erence approximations f or
the gradi ent of a vector u are given by:
u
I +l / 2 , J
+
l / 2
a
J _
( u
I
+
l , J
+
l / 2 _
u
I , J
+
l /
2 )
u
|
:
i / 2 , J
+
l / 2
=
1_
( u
I
+
l / 2 , J
+
l _
u
j
+
l / 2 , J
)
Ax.
Eqns. (3.2) r equi r e t he mesh t o be t opol o g i c a l l y
regul ar.
The use of contour i nt egr al formulas ( Wi l ki ns, 1964) al l ows the
appl i cat i on of FD approximations t o t opol ogi cal l y and geomet r i cal l y
i r r egul ar meshes. The basic idea i s t o employ Gauss' theorem i n order
(3.2)
and geomet r i cal l y
t o express t he gr adi ent of a f i e l d i n a c e l l i n t erms of a cont our
i nt egr al . Considering cel l VE enclosed by contour SE, and the gradi ent
of the displacement vector u,
J u
i 5J
dV = [ u^j dS (3.3)
If the gradient is assumed constant in the cell, then
1
f
u
i
,j = - I u^jdS (3.4)
V
E J
S
E
The contour i nt egr al may be evaluated assuming a l i near var i at i on of u
along the edge of the c e l l .
Contour i nt egr al s may be used f or any 2-0 pol ygon or 3-D
pol yhedron, t o i nt er pol at e a val ue f or t he gr adi ent at t he cent r e o
t he c e l l , knowi ng t he val ues at t he cor ner nodes. For t he p a r t i c u l a r
cases of t r i a n g l e s and t et r ahedr a, an a l t e r n a t i v e t echni que i s
avai l abl e (e.g. Mar t i , 1981), i n which the gradi ents are i nt er pol at ed
d i r e c t l y by i n v e r t i n g t he s pat i al f i n i t e di f f er enc e equat i ons. Thi s
technique has been f ol l owed i n the present work, and wi l l be det ai l ed
i n sect i on 4. 2. 1.
For ex pl i c i t ti me-marchi ng models the semi -di scret e equations of
motion become uncoupled. This means t hat only l ocal approximations to
t he p a r t i a l d i f f e r e n t i a l equat i ons (3.1) are perf ormed wi t h i n each
t i me- s t ep, no i t e r a t i o n s bei ng needed f or a FD oper at or . Such a f ac t
was expl oi t ed i n the development of the f i r s t FD "Hydrocodes" at the
U.S. nat i onal l abor at or i es i n t he 1950' s. These were or i ent ed mai nl y
towards sensi t i ve nuclear and defence appl i cat i ons. L i t t l e publ i ci t y
was given unt i l the 1960's (Wi l ki ns (1964), Maenchen and Sack (1964),
Noh (1964)). At t h i s t i me Fi n i t e Element Methods had j us t been
i nt r oduced f or s ol i d mechanics (Cl ough, 1960), and t echni ques were
being developed f or l i near anal ysi s. Not much at t ent i on was given t o
FD f or s o l i d mechanics by t he engi neer i ng communi t y, as i ndeed FE
methods seemed much more powerful and indeed advantageous f or l i near
systems, being able t o provi de a one-step gl obal sol ut i on.
I nt er es t i n t he nonl i near and wave- pr opagat i on regi mes f or
speci al i zed engi neer i ng appl i c at i ons i n t he l at e 1960's and 1970's
41
cr eat ed a resurgence of t he Hydrocodes (Bert hol f and Benzley (1968),
Wi l ki ns (1975)), and a cer t ai n degree of convergence between FD and FE
l i t e r a t u r e ( Bel yt schko (1978), Kr i eg and Key (1976), Goudreau and
Hal l qui st (1982)). Expl i ci t f i ni t e- di f f er ence methods were popul ari zed
t o wi der sect or s of t he engi neer i ng communi t y, and new codes were
cr eat ed such as PISCES (Hancock, 1976), t he rock mechanics codes of
Cundall and Marti (1979), and PR3D f or sol i d mechanics impact by Marti
(1981).
3. 3 FINITE ELEMENT METHODS
The f i r s t appl i cat i on of Fi ni t e Element techniques f or conti nua
was by Clough (196U), al t hough t he t h e o r e t i c a l bases f or t he method
had already been set by Courant (1943) and appl i cat i ons t o st r uct ur al
anal ysi s had been proposed ear l i er (Argyri s and Kelsey, 1954).
Fi ni t e Element di scr et i zat i ons r el y on two essent i al i ngr edi ent s:
a v a r i a t i o n a l or weak f orm of t he eqns. of mot i on ( 3. 1) , and a
c ons t r uc t i on of appr oxi mat e sol ut i ons based on gener al i zed nodal
coordi nates and independent element shape f unct i ons.
The domain V i s subdi vi ded i nt o el ement s VE, i nt er connect ed by
nodes. An approximate sol ut i on i s constructed wi t hi n an element E as a
product of shape f unct i ons N^x) and the nodal displacements uf ( t ) :
u ( x , t ) = u ^ ( t ) Nj ( x ) ( 3 . 5 )
where I i s summed over the nodes of the element. The shape f unct i ons
are chosen so t hat u i s cont i nuous over t he el ement boundar i es,
al t hough i t s gr adi ent need not be cont i nuous (CO c o n t i n u i t y ) . The
shape f unct i ons Ni
s o
def i ned are i ndependent of t i me ; eqn. (3.5)
c o n s t i t u t e s i n f a c t a l o c a l s e p a r a t i o n of v a r i a b l e s
(semi di scr et i zat i on ).
The di scr et e form of the gradi ent operator may be wr i t t en as
u
i ' j "
B
j l
u
i l
(3.6)
42
where
B
Ji
dx
i
Let t he s o l i d cont i nuum be V wi t h boundary S, c ons i s t i ng of S
y
and Sj '
w h e r e
u = u* on S
u
an = T*
o n
s
T
A weak f orm of t he eqns. of mot i on (3.1) may be obt ai ned by usi ng
ei t her Gal erki n weighed resi dual s or the vi r t ual work pr i nci pl e, both
of which yi el d the same r esul t :
f
V i
,
j C T i j
dV + r
p
v
i
u
i
d V = JpVi f i dV + [v-jT^dS (3. 7)
V V V S
T
where v i s the t est f unct i on (or var i at i on) and u the t r i a l f unct i on.
Eqns. (3.7) r equi r e onl y C c o n t i n u i t y f or both t r i a l and t es t
f unc t i ons , as opposed t o (3. 1), f or whi ch Cl c o n t i n u i t y i s needed. I f
the approximations defi ned i n (3.5) are used f or u and v, and because
(3.7) must hol d f or a r b i t r a r y v, t he gl obal di s c r et e equat i ons are
deduced:
Mu + P(u) = R (3. 8)
where t he gl obal c o e f f i c i e n t mat r i ces M, P, R are assembled f rom
i ndi vi dual element matrices t hat take the form:
M
E
= / pNj Nj &j j dV (mass mat ri x;
V
E
P
E
= / Bj j ^ j d V ( i nt er nal f orces) (3. 9;
J
V
E
R
E
= / f-j NjdV +f N
I
T
i
dS (external f orces)
v
J
s$
43
(3.8) i s a system of ordi nary di f f er ent i al equations of second order
i n t i me; i n t e g r a t i o n of t hese i s di scussed i n sect i on 3.6. For a
l i near model (small deformations and el ast i c behavi our), (3. 8) becomes
Mii + Ku = R (3.10)
The st i f f ness mat ri x K i s assembled from element matri ces of the type
K
E
= / 6
j j C i j k l
B
1 K
d V (3.11)
J v
E
In st at i c anal ysi s the i ner t i a term may be dropped from eqns. ( 3. 8) :
P(u) = R (3.12)
which for the linear case becomes
Ku = R (3.13)
For l i near anal y s i s , a s ol ut i on i s obt ai ned merel y by i n v e r t i n g t he
s t i f f n e s s mat r i x K i n eqns. (3.10) or (3.13). I n a nonl i near case,
eqns. (3.8) or (3.12) must be solved i n a number of steps usi ng Newton-
Raphson or i t er at i ve techni ques. I t i s i nt er est i ng t o note the a b i l i t y
of FE t o gi ve a one-st ep s ol ut i on t o t he l i near pr obl em, whi ch F0
methods l ack, havi ng t o approach t he gl obal s ol ut i on t hr ough
r el ax at i on and i t e r a t i o n . Hence t he popul ar i t y of FE f or l i n e a r
problems. For nonl i near behaviour, however, t hi s advantage disappears'*
as both FE and FD have t o perform some sort of i t er at i ons.
For l arge systems, the assemblage of mat ri x K i s undesi rabl e, as
core memory l i mi t s may be exceeded and recourse must be made t o sl ow,
cost l y disk Input/Output. This f act accounts f or the popul ar i zat i on of
r el ax at i on t echni ques f or equat i on s o l v i n g ( e. g. Fl anagan and
Bel ytschko, 1981a), which avoid the assemblage of global coef f i ci ent
mat r i ces. The FE oper at or s are used onl y at a l ocal l e v e l . In t h i s
case FE become conceptual l y very si mi l ar t o FD methods wi t h general
topol ogy, speci al l y as they of t en pr ovi de equi val ent al gor i t hms f or
the l ocal approximations (Kunar and Minowa, 1981).
44
3.4 MESH DESCRIPTIONS
Let a p a r t i c l e X of body B be def i ned by i t s pos i t i on at t =0
( r ef er ence c o n f i g u r a t i o n ) , X . At t i me t ( cur r ent c onf i gur at i on) t he
posi t i on of the par t i cl e wi l l be
x = x( X, t ) ( 3. 14)
x are called the spatial coordinates, and X the material coordinates
of X. Eqns. (3.14) describe the motion of 8.
For the discretization of B three types of meshes may be used,
depending on the motion of the nodes of the mesh. The position of a
point of the mesh, initially coincident with particle X, will be given
by X = X ( X ,t ) .
3.4.1 LAGRANGIAN
In a lagrangian description the mesh follows the motion of the
body,
X(X,t) = x(X,t) (3.15)
A gi ven node remai ns coi nci dent wi t h t he same mat er i al p a r t i c l e
throughout the moti on. Each element wi l l contai n the same domain of
mat eri al throughout the def or mat i on, t hus enf or ci ng i mp l i c i t l y t he
cont i nui t y equat i on.
Mot i on of t he boundary does not present d i f f i c u l t i e s , as i t
al ways coi nci des wi t h t he mesh boundary. For a scal ar f i e l d g( X, t ) ,
the mat eri al t i me der i vat i ve ( i . e. f ol l owi ng the par t i cl e) coi nci des
wi t h the par t i al ti me der i vat i ve:
dg
g = - ( 3. 16)
dt
The only disadvantage of t hi s descr i pt i on comes from the f act t hat the
mesh can become excessi vel y d i s t o r t e d f or c e r t a i n probl ems (e.g.
f l u i d s , hi gh v el oc i t y i mpact ) . In some cases, " r ezoni ng" t echni ques
45
may be used to circumvent this problem (e.g. Kalsi and Marti, 1985).
3.4.2 EULERIAN
In an Eulerian description the mesh is fixed in space, i.e.
X(X,t) = X (3.17)
Nodes are no longer coi nci dent wi t h mat eri al par t i cl es through t i me,
and the mat eri al f l ows through the cel l s. Cont i nui t y must be enforced
e x p l i c i t l y . The mat er i al t i me der i v at i v e of g( X>t ) i ncl udes a f l u x
term:
69 39 dx dg d9
g = _ + = _ + _
v
(3.18)
dt dx dt dt dx
Numeri cal comput at i ons f or t he f l u x of scal ar f i e l d s t end t o smear
t hei r val ues, which wi l l not be defi ned as sharpl y as f or a Lagrangian
mesh. Mat er i al boundari es are d i f f i c u l t t o descr i be, as they move
r el at i ve t o the mesh. On the cr edi t si de, di st or t i on i s not a problem,
making Eul eri an meshes pr ef er r abl e t o Lagrangian meshes f or very l arge
deformati ons.
3. 4. 3 ARBITRARY LAGRANGIAN-EULERIAN
Ar bi t r ar y Lagr angi an- Eul er i an (ALE) des c r i pt i ons at t empt t o
combine t he advantages of Lagrangi an and Eul er i an meshes. The mesh
moves wi t h an a r b i t r a r i l y def i ned mot i on, X( X, t ) . dX/ dt =0 f or an
Eul eri an mesh, dX/ dt =v f or a Lagrangian mesh. X can be defi ned so as
t o f ol l ow the mat eri al i n the boundary, but wi t hout causing excessive
d i s t o r t i o n i n t he i n t e r i o r . ALE f or mul at i ons have been devel oped by
Noh (1964) and Hi r t et a l . (1974) i n FO f or mat s, and by Donea et a l .
(1977) and Belytschko and Kennedy (1978) i n FE.
Materi al der i vat i ves are given by
d9 dx 59
g = _ + (3. 19)
dt dt dx
46
Wi th an ALE des c r i pt i on pr oper t i es s t i l l need t o be f l uxed t hr ough
c el l s , and some smearing may occur as a r esul t .
A c r uc i al aspect i n ALE des c r i pt i ons i s t he d e f i n i t i o n of t he
a r b i t r a r y mot i on of t he mesh X f or i nt er nal poi nt s. Gener al l y a
complex rezoning al gor i t hm i s necessary f or opt i mi zi ng the new mesh
posi t i ons at each st ep. Such a general r ezoni ng al gor i t hm has been
proposed f or 2-D by Gi ul i ani (1982). Schreurs (1983) has proposed a
mesh opt i mi z i ng al gor i t hm based on t he def or mat i on of a f i c t i t i o u s
mat er i al f rom an " i d e a l " mesh. In f ac t ALE t echni ques woul d be
equi val ent t o Lagrangian descri pt i ons i n which rezoning i s performed
at every st ep. Some appl i c at i ons (e.g. metal f or mi ng) may not need
such f r equent r ezoni ng, and Lagrangi an t echni ques wi t h r ezoni ng at
wider i nt er val s could be pr ef er r abl e.
3. 5 LARGE DISPLACEMENT FORMULATIONS
Sever al f o r mu l a t i o n s are p o s s i b l e dependi ng on whi ch
c onf i gur at i ons t he st r ess and def or mat i on t ensor s are r ef er r ed t o.
Three al t er nat i ves wi del y used i n sol i d mechanics are presented below.
3.5.1 TOTAL LAGRANGIAN
The 2nd Pi ol a- Ki r c hhof f st r ess t ensor S and t he Green s t r a i n
t ensor E, both of whi ch r el at e t o t he r ef er ence c o n f i g u r a t i o n , are
used t o descr i be t he mat er i al behavi our . Hi b b i t , Marcal and Rice
(1970) proposed t hi s descr i pt i on i n the f i r s t published l ar ge- st r ai n,
l arge-di spl acement nonl i near f or mul at i on f or general purpose FE codes.
A const i t ut i ve r el at i on i s given by
S = S(E) (3.20)
47
and i n rate form by
S = D:E (3.21)
(
S
I J
= D
IJKL
E
KL components)
S and E bei ng mat er i al t ensor s, t h e i r mat er i al r at es are obj ec t i v e.
This f or mul at i on i s advantageous f or Hyper el ast i c mat er i al s , whose
behaviour i s described on the reference conf i gur at i on. In t hi s case,
cal l i ng W the st r ai n energy f unct i onal per uni t mass,
d
2
W
dE2 (3.22)
aw
s =
d
El a s t i c - p i a s t i c mat er i al behavi our i s best descr i bed on t he
cur r ent c o n f i g u r a t i o n , x ( H i l l , 1950). I t i s possi bl e t o t r ansf or m
such a law i nt o one of t he t ype (3.21) (e.g. Hi bbi t et a l . (1970),
Kr i eg and Key (1976)), but compl ex and comput at i onal l y expensi ve
t ransf ormat i ons are necessary. However, Simo and Or t i z (1985) suggest
t hat t ot al Lagrangian, Hyperel ast i c-t ype f ormul at i ons provide a more
r i gor ous approach f or i nc r ement al , non- l i near c a l c u l a t i o n s . Such
r i gour i s not j u s t i f i e d i n e x p l i c i t c al c ul at i ons wi t h very smal l
steps.
3.5.2 CAUCHY STRESS-VELOCITY STRAIN
A descr i pt i on based on the current conf i gur at i on may be used to
model the behaviour of mat eri al s wi t h smooth memory. In the si mpl est
case, t he Jaumann r at e of Cauchy st r ess ( a ) and t he r at e of
deformati on tensor ( vel oci t y st r ai n, d) are r el at ed by
ff= C:d ( 3. 23)
( ajj = Cj ^ i d k i i n component form)
48
where C i s t he c o n s t i t u t i v e t ensor . For an e l a s t i c - p l a s t i c mat er i a'
(hypoel asti c wi t h associated pl as t i c i t y ) C takes the form
C
i j k l
=
^ i j 5
k
i + 2G(5
i k
5
J1
- ^ i j n
k
i )
(3.24)
where rj >0 f or pl ast i c l oadi ng, =0 otherwise
n i s the uni t normal t o the yi el d surface
The Jaumann d e r i v a t i v e used i n eqn. ( 3. 23) pr ov i des t he
const i t ut i ve part of the stress r at e. To obtai n the t ot al stress rate
the r ot at i onal components must be added:
a = o- + w n + w- a
1J 1J
W
1P
CT
PJ
W
JP pi
( 3. 25)
I f t he mat er i al behavi our i s ani s ot r opi c , C must be updated wi t h a
si mi l ar obj ect i ve r at e:
c
i j k l
_ c
i j k l
+ w
i p
c
p j k l
+ w
j p
c
i p k l
+ w
k p
c
i j p l
+ w
l p
c
i j k p
(3.26)
An al t er nat i ve f or mul at i on resul t s from the use of the Truesdel l
rate i n eqn. ( 3. 23) :
(J = C:d :3.27
(3.23) and (3.27) are equi val ent i f one sets
Ci j kl =
C
i j k l
+ CT
ij
5
kl-^ik
5
jl
+or
il
5
jk
+(7
jk
5
il
+ CT
il
5
ik)/
2
(
3
-
2 8
)
The Tr uesdel l st r ess r at e i s t he f or war d Pi ol a t r ans f or mat i on
(eqns. 2.22, 2.25) of t he r at e of t he 2nd Pi ol a- Ki r c hhof f st r ess
t ensor:
O = 0
t
* ( J
- 1
S) [3. 29)
Pinsky, Or t i z and Pi st er (1983) have suggested t hat the Truesdel l
r a t e f o r mu l a t i o n i s t he n a t u r a l one t o use ( i n t he c u r r e n t
c onf i gur at i on) f or h y p e r e l a s t i c i t y . In t h i s case t he c o n s t i t u t i v e
tensor i s obtained di r ect l y, from the t ot al Lagrangian tensor D, as
49
= 0
t
* ( j - l D) .
Cl assi cal p l a s t i c i t y i s descr i bed on the current conf i gur at i on
( Hi l l , 1950). Jaumann Cauchy stress f ormul at i ons have been wi del y and
successf ul l y used f or e l a s t i c - p l a s t i c behavi our ( Wi l ki ns (1964),
Maenchen and Sack ( 1964) , Kr i eg and Key ( 1976) ) . Wi t h such
f o r mu l a t i o n s h y p e r e l a s t i c behavi our ( r el at ed t o t he o r i g i n a l
conf i gur at i on) may also be descri bed, al bei t i n a less convenient way,
as t he c o n s t i t u t i v e r el at i ons need t o be pushed f or war d i nt o t he
current conf i gur at i on.
Fi n a l l y , one probl em wi t h t h i s f or mul at i on i s t hat d i s not
i nt egr abl e ( i . e. i t i s not t he r at e of any v a l i d s t r a i n t ensor ) .
Addi t i onal st r ai n computations must be done i f a t ot al st r ai n measure
i s requi r ed.
3.5.3 UPDATED LAGRANGIAN
In t h i s f or mul at i on t he model i s descr i bed on a r ef er ence
conf i gur at i on, which i s updated at each increment t o coi nci de wi t h the
cur r ent c o n f i g u r a t i o n . From t h i s updated reference, the incremental
conf i gur at i on i s described wi t h a t ot al Lagrangian f or mul at i on. This
method was f i r s t proposed by Yaghmai and Popov (1971), and has been
wi del y used si nce f or i ncr ement al nonl i near anal y s i s : Osias and
Swedlow (1974), Bathe et al . (1975), Nagtegaal and de Jong (1981).
For t h i s d e s c r i p t i o n , F = I ( i d e n t i t y ) and J = 1. Hence, eqn.
(3.29) i mpl i es S = a. I t i s al so easy t o see f rom eqn. (2.19) t hat E =
d. In f ac t t h i s f or mul at i on r ever t s t o t he Tr uesdel l Cauchy st r ess
r at e f o r mu l a t i o n , eqn. (3.27). Thi s means t hat t he t ensor t o be used
f or the t angent i al st i f f ness i s .
3.6 TIME INTEGRATION
Using ei t her Fi ni t e Di f f erence or Fi ni t e Element Methods f or the
spat i al semi di scr et i zat i on, the par t i al di f f er ent i al eqns. of motion
(3.1) may be t r ansf or med i nt o a system of or di nar y d i f f e r e n t i a l
50
equations i n t i me:
Mii + Cu + P(u) = R (3.30)
These eqns. can be sol ved ei t her by modal anal ysi s or by d i r e c t
i n t e g r a t i o n . Modal anal ysi s methods (e.g. Bathe and Wi l son (1976),
chpt . 8) per f or m t r ansf or mat i ons of eqns. (3.30) whi ch are onl y
val i d f or l i near or quasi - l i near systems (i . e. P(u) = Ku)
For nonl i near anal y s i s , d i r e c t i n t e g r a t i o n ( t i me- mar chi ng)
methods must be used. For these the t i me domain i s di vi ded i nt o t i me-
steps {At), and an i ncremental anal ysi s i s performed f or each step.
Time i nt egr at i on procedures may be cl assi f i ed i nt o expl i ci t and
i mpl i c i t . Expl i ci t schemes compute the i ncremental displacements u
f rom t he equi l i br i um condi t i ons at t i me t . I mp l i c i t schemes, on t he
c ont r ar y , sol ve t he eqns. of mot i on (3.30) at t+hAt > t , pr oduci ng an
i mpl i ci t system of eqns. f or ut +At .
Two of t he most common and r epr esent at i ve t i me - i n t e g r a t i o n
schemes, one i n ei t her cl ass, are presented below.
3.6.1 CENTRAL DIFFERENCE (EXPLICIT)
Cent r al di f f er enc e methods are t he most wi del y used e x p l i c i t
schemes f or sol i d mechanics, being the opti mal from a very wide cl ass
(Key, 1978). The Fi ni t e Di f f erence expressions used f or vel oci t y and
accel erat i on are
a
n + l / 2
= ( l
/ i +l _j i
) M t ( 3 > 3 1 a )
ii
n
= ( u
n + 1 / 2
- u
n
"
1 / 2
) / At (3.31b)
Note t hat each d e r i v a t i v e l ags t he val ue by hal f a t i me- s t ep.
Par t i cul ar i zi ng the equations of motion (3.30) at time n,
Mu
n
+ Cu
n
+ P(u
n
) = R
n
(3.32)
51
Usi ng eqn. (3.31b) and l e t t i n g On = ( un- l / 2+
u
n + l / 2 ) / 2 , eqn. (3.32) may
be s ol v ed, y i e l d i n g :
u"
+
l / 2
=
(MMt -C/ 2)-
1
[MMt -C/ 2)u
n- 1/ 2
+R
n
-P(u
n
)] (3.33)
The new d i s p l a c e me n t s un + 1 ar e t hen f ound f r om eqn. ( 3. 31a) . I f t he
system has no damping (C = 0) and t he mass mat r i x M i s di agonal , eqns.
(3. 33) become uncoupl ed:
n+l / 2
=
- n- l / 2
+ M
- l
[ R
n_p( u%t (3.34)
These eqns. can t hen be s ol v ed i n d e p e n d e n t l y f o r each degr ee of
freedom I :
n + l / 2
= G
n - l / 2
+ A ( R
n _
p
n
) / l t ) i ( 3 > 3 5 )
The equat i ons al so become uncoupl ed i f t he damping i s assumed t o
be of t he Rayl ei gh t ype, as shown i n sect i on 4.6.
Thi s uncoupl i ng of t he equat i ons of mot i on i s t he maj or advantage
of e x p l i c i t i n t e g r a t i o n pr ocedur es. No mass or s t i f f n e s s mat r i ces need
be i nv er t ed or even assembl ed, as a l l t he i ncr ement al c al c ul at i ons f or
each degree of f reedom can be done i ndependent l y at t he l ocal l e v e l .
Thi s not onl y al l ows f or a si mpl er a r c h i t e c t u r e i n comput er codes, but
i t enabl es t he t r e a t me n t of n o n - l i n e a r i t i e s (be i t of C o n s t i t u t i v e ,
Geomet r i c or Boundar y t y pe) w i t h v i r t u a l l y no added c os t f r om t he
l i n e a r case. The number of o p e r a t i o n s per t i me - s t e p i s much s ma l l e r
t han f o r i m p l i c i t met hods ( s e c t i o n 3. 6. 2) , and s t or age r e q u i r e me n t s
grow onl y l i n e a r l y wi t h t he si ze of t he pr obl em.
The mai n di s adv ant age of t he c e n t r a l d i f f e r e n c e and o t h e r
e x p l i c i t methods i s t hat comput at i ons ar e onl y c o n d i t i o n a l l y s t a b l e
dependi ng on t he t i me- s t ep si ze. The t i me- s t ep must be smal l er t han a
c e r t a i n c r i t i c a l v al ue f o r n u me r i c a l e r r o r s not t o grow unbounded.
Thi s c o n s t i t u t e s a maj or o b s t a c l e f o r c e r t a i n pr obl ems wher e an
excessi ve number of t i me- st eps makes t he anal ysi s t oo c o s t l y .
The s t a b i l i t y of t he c ent r al di f f er enc e method i s consi der ed i n
s e c t i o n 4. 7. The t i me - s t e p i s l i m i t e d by t he Cour ant c r i t e r i o n , i . e.
52
the ti me i t takes the stress waves to t r avel across one element. This
l i mi t at i on i s consi st ent wi t h the l ocal , uncoupled i nt egr at i on of the
equat i ons of mot i on. I f t he t i me- st ep was l ar ger than t he Courant
c r i t i c a l val ue, stress waves would t r avel across an element wi t hi n one
t i me- s t ep, a f f e c t i n g t he sur r oundi ng el ement s. The i ncr ement al
behaviour of t hat element would no longer be independent from the rest
of the model.
Cent r al di f f er enc e schemes have been wi del y used i n nonl i near
numer i cal codes, f rom t he ear l y FD hydrocodes of Wi 1ki ns(1964) and
Maenchen and Sack(1964), t o the FE codes of Hal l qui st (1982a, 1982c),
Key (1974), and Bel yt schko and Tsay(1982). The accuracy and
s t abi l i t y of cent ral di f f erence methods has been studi ed and discussed
by var i ous aut hor s (e.g. Bel yt schko, Holmes and Mul l en (1975),
Belytschko(1978), Krieg and Key(1973)). The cent ral di f f erence method
i s considered as the most convenient wi t hi n the expl i ci t cl ass.
3.6.2. TRAPEZOIDAL RULE (IMPLICIT)
The so- cal l ed t r apezoi dal r ul e i s an example of i mp l i c i t
i nt egr at i on methods. In f act i t const i t ut es a par t i cul ar case of the
Newmark f ami l y, probably the most popular of the i mpl i c i t schemes. A
const ant average ac c el er at i on i s assumed f or each i ncr ement At. The
di f f erence equations are:
u
n + h
= (i i
n
+ u
n + 1
)/2 (0 s< h < 1)
u
n + 1
= u
n
+ ( i i
n
+ i i
n + 1
)At/Z (3.36)
u
n+l
= u
n
+ a
n
4 t + (
.jjn
+
yn+1
) A t
2
/ 4
For obt ai ni ng un+l the equations of motion are enforced f or ti me
t+At. In an undamped case,
Mii
n+1
+ P(u
n+1
) = R
n+1
(3.37)
Eqn. (3.37) i s an i mp l i c i t r e l a t i o n f or un+l . Su b s t i t u t i n g t he
di f f er enc e expr essi ons (3.36) i n (3.37) t he f o l l o wi n g system i s
obt ai ned:
53
L ( 4 Mt
2
) M+K
n + 1
j u
n + 1
= R
n + 1
+ M[ 4 Mt
2
) u
n
+ ( 4 Mt ) u
n
+i i
n
J ( 3. 38)
where Kn+1 i s t he secant s t i f f n e s s mat r i x ( pn+l = Kn+l un+1) .
For a l i n e a r system (Kn = const ant ) eqns. (3.38) may be sol ved by
i n v e r t i n g t he mo d i f i e d s t i f f n e s s ma t r i x K*, d e f i n e d as
K* = ( 4 / 4 t ^ M+K
n + 1
(3.39)
For n o n l i n e a r syst ems t he e q u a t i o n s ar e g e n e r a l l y s ol v ed by
d i r e c t e l i mi n a t i o n t echni ques, e.g. Newton-Raphson t ype methods (whi ch
r equi r e t r i a n g u l a r i z a t i o n of K*).
The t r a p e z o i d a l r u l e i s u n c o n d i t i o n a l l y s t a b l e (see e. g.
Bel yt schko and Schoeberl e, 1975), t he t i me- s t ep bei ng l i mi t e d onl y by
accur acy c o n s i d e r a t i o n s . Thi s i s t he mai n advant age of i m p l i c i t
schemes, whi ch makes them more appr opr i at e f or probl ems i n whi ch l ar ge
t i me- st eps can be used. I f t he t i me- st eps are l i mi t e d t o smal l val ues
f or reasons ot her t han s t a b i l i t y (e.g. st eep n o n l i n e a r i t i e s , accur acy,
st r ess- waves) i mp l i c i t methods l ose t h e i r advant age.
Anot her l i mi t a t i o n i s t he l ar ge st or age r equi r ed f or t he mat r i x
c o e f f i c i e n t s , whi ch r e s t r i c t s sever el y t he si ze of probl em t hat can be
s ol v ed i n - c o r e . Lar ger model s and 3-D anal y s es must o f t e n r e s o r t t o
out - of - c or e st or age wi t h f r equent and sl ow di sk I / O. Because of t h i s
pr obl em, a l t e r n a t i v e equat i on s o l v i n g met hods based on i t e r a t i v e or
r el axat i on procedures are becoming popul ar . Cl assi cal methods of t h i s
t y pe ar e t he Jacobi and Gauss- Sei del p r o c e d u r e s , a l t h o u g h t h e i r
ef f ect i veness f or f i n i t e Element codes i s l i mi t e d ( Bel yt schko, 1983).
More p r o mi s i n g f o r FE t o p o l o g i e s ar e c o n j u g a t e g r a d i e n t and q u a s i -
Newton methods. A new and seduct i ve El ement - by- El ement equat i on sol ver
has been pr oposed l a t e l y by Hughes, L e v i t and Wi nget ( 1983b) ;
b a s i c a l l y i t c o n s i s t s of an o p e r a t o r - s p l i t t i n g met hod t h a t t akes
advantage of t he FE data s t r u c t u r e (see a l s o s e c t . 3. 6. 3) .
I m p l i c i t met hods ar e l es s r e l i a b l e as t o c o mp l e t i n g t he
c o mp u t a t i o n s wi t h o u t c r a s h i n g and wi t h t he r e q u i s i t e accur acy t han
e x p l i c i t methods ( Bel yt schko, 1983) .
54
No n l i n e a r model s are g e n e r a l l y more d i f f i c u l t wi t h i m p l i c i t
schemes. F i r s t l y l i e s t he i n c r e a s e d c os t of t he Newt on- Raphson
i t e r a t i o n s . Then i s t h e f a c t t h a t a l l b o u n d a r y c o n d i t i o n s ,
n o n l i n e a r i t i e s , e t c . must be i n c l u d e d i m p l i c i t l y i n t he secant
s t i f f n e s s ma t r i x ( K* i n eqn. ( 3. 39) ) . Thi s g r e a t l y c o mp l i c a t e s t he
f o r mu l a t i o n and may r e s t r i c t t he r ange of c o n s t i t u t i v e model s and
boundary condi t i ons t hat may be used.
3. 6. 3 OPERATOR SPLIT METHODS
From t he di scussi on i n sect i ons 3.6.1 and 3.6.2, i t i s cl ear t hat
an oper at or combi ni ng t he advant ages of e x p l i c i t met hods ( uncoupl ed
equat i ons, si mpl e program a r c h i t e c t u r e , no st or age of gl obal mat r i x )
wi t h t he u n c o n d i t i o n a l s t a b i l i t y of i m p l i c i t met hods woul d be ver y
d e s i r a b l e . Op e r a t o r s p l i t met hods a t t e mp t t o c o mb i n e such
c h a r a c t e r i s t i c s .
An u n c o n d i t i o n a l l y s t a b l e e x p l i c i t pr ocedur e was pr oposed by
T r u j i l l o (1972) and r e c e i v e d v a r i o u s e x t e n s i o n s by Park ( 1982) . Thi s
c o n s i s t s i n s p l i t t i n g t he s t i f f n e s s K i n t o upper and l ower ma t r i c e s
and p e r f o r mi n g t wo passes i n a l t e r n a t e d i r e c t i o n s , each pass
cor r espondi ng t o hal f a t i me- s t ep. Al t hough t h i s procedure appears t o
wor k we l l f o r heat c o n d u c t i o n ( T r u j i l l o , 1975) , f o r s t r u c t u r a l
dynamics t he accuracy i s poor ( Mul l en and Bel yt schko, 1983).
A p r o mi s i n g avenue i s t he use of El e me n t - b y - El e me n t met hods,
whi ch p e r f o r m a f a c t o r i z a t i o n t a k i n g advant age of t he FE dat a
s t r u c t u r e . Such an a l g o r i t h m has been pr oposed f o r s o l i d and
s t r u c t u r a l mechani cs by Or t i z , Pi nsky and Ta y l o r ( 1983) . I n i t s
o r i g i n a l f or m t h i s met hod was pr oposed by Hughes, L e v i t and Wi nget
( 1983a) f o r heat c o n d u c t i o n . The same a u t h o r s ( 1983b) r e p o r t
u n s a t i s f a c t o r y accur acy i n s t r u c t u r a l dy nami c s , and suggest t h i s
t e c h n i q u e be used i n s t e a d as a l i n e a r e q u a t i o n s o l v e r w i t h i n an
i m p l i c i t scheme. The appr oach of Or t i z et al ( 1983) uses v e l o c i t i e s
and s t r e s s e s as unknowns. They p e r f o r m t wo passes i n a l t e r n a t i n g
di r ec t i ons wi t h i n a t r apezoi dal r ul e t o obt ai n second or der accuracy.
A quest i on mark s t i l l hangs over t he accuracy of t hese methods.
55
3.7 PRACTICAL CONSIDERATIONS FOR DISCRETE MESHES
A number of undesi r abl e ef f ec t s can be generat ed by t he nat ur e of
t he s pat i al s e mi d i s c r e t i z a t i o n . Unaccept abl e s ol ut i ons may be produced
by F i n i t e El ement meshes even i f t he el ement s s a t i s f y t he s t andar d
conver gence c r i t e r i a ( e. g. Zi enki e wi c z , 1977, c h p t s . Z&3). Two
pr obl ems , r e l a t e d t o t he degr ee of v o l u me t r i c c o n s t r a i n t , ar e
di scussed bel ow. "Locki ng up" occur s f o r ex c es s i v e c ons t r a i n t . wh i 1 e
" hour gl assi ng" modes appear i f vol umet r i c c ons t r ai nt s are t oo r el axed.
Anot her pr obl em not di s c us s ed her e, but t o whi ch r e f e r e n c e i s
made i n o t h e r s e c t i o n s ( 4. 2. 3, 7. 4. 4) i s t he e x c e s s i v e d i s t o r t i o n i n
Lagrangi an meshes f or models wi t h very l ar ge def or mat i ons, whi ch may
cause t angl i ng over of t he mesh.
3. 7. 1 "LOCKING-UP" FOR INCOMPRESSIBLE FLOW
Some meshes ( i n p a r t i c u l a r c o n s t a n t s t r a i n t r i a n g l e s and
t et r ahedr a) gi ve o v e r s t i f f s ol ut i ons f or i ncompr essi bl e p l a s t i c f l ow.
Al t hough t h i s i s a wel l known f a c t , some aut hor s have cont i nued t o use
t hem f o r c onv eni enc e ( e. g. Johnson, 1976, 1977) . The v a l i d i t y of
such p r a c t i c e has been s t r o n g l y q u e s t i o n e d ( H a l l q u i s t , Werne and
Wi l k i ns , 1977).
The r eason f o r " l o c k i n g - u p " was shown i n a c l a s s i c a l paper by
Nagt egaal , Par ks and Ri ce (1974) t o be an e x c e s s i v e number of
v o l u me t r i c i n c o mp r e s s i b i 1 i t y c o n s t r a i n t s i n t he d i s c r e t e meshes.
Accor di ng t o Nagt egaal , t he number of vol umet r i c c ons t r ai nt s must not
exceed t he number of Degrees Of Freedom (DOF) of t he model . I d e a l l y
t he degree of c ons t r ai nt shoul d be t he same as i n t he cont i nuum, wi t h
one i ncompr essi bi 1 i t y equat i on per mat er i al poi nt (v^ -j=(j). Thus t he
i deal r a t i o DOF/ const r ai nt s equal s 2 f or pl ane s t r a i n or axi symmet r i c
a n a l y s e s , and 3 f o r t h r e e - d i me n s i o n a l model s. For c onv er genc e, t he
r a t i o DOF/ const r ai nt s shoul d t end t o a val ue gr eat er t han 1 when t he
mesh i s i n f i n i t e l y r e f i n e d . Nagt egaal et al ( 1974) s t u d i e d a f ew
commonly used el ement s and f ound some of them t o be unsui t abl e. Some
of t h e i r r es ul t s are pr esent ed i n t abl e 3. 1.
56
PLANE STRAIN
Const ant St r ai n Tr i angl e
(CST) I
T j - n
i sopar amet r i c 4-node '
q u a d r i l a t e r a l j
Mi xed Tr i angl es- Quad.
(MTQ.MTQC)
A
/
t
y
AXISYMMETRIC
Constant St r ai n Tr i angl e A
(CST) (exact hoop s t r a i n )
Constant St r ai n Tr i angl e A
(CST) (reduced hoop s t r . ) **
Mixed Tr i angl es- Quad.
(MTQ.MTQC) wi t h reducec j
hoop s t r ai ns -~
3-DIMENSIONAL
Constant St r ai n Tet r ah.
(CST),5 per b r i c k ,
r egul ar l a t t i c e
Constant St r ai n Tet r ah.
(CST),6 per b r i c k ,
r egul ar l a t t i c e
8-node i sopar amet r i c '
br i ck
Mi xed Tet r ah. - Br i c k
(MTB.MTBC)
5 or 6 Tet r ah. / Br i c k
t
A
A
. _
" C 3
t
1
I
-
1-"
- ^
' / i V
' ' 1
Const r ai nt s
Element
1
3
1
3
1
1
1
1
7
1
Nodes
Element
1/2
1
1
1/2
1/2
1
1/5
1/6
1
1
DOF
Const r ai nt
1
2/3
2
1/3
1
2
3/ 5
3/ 6
3/7
3
Tabl e 3 . 1 : Ra t i o s of Degr ees of Fr eedom t o i n c o m p r e s s i b i l i t y
c o n s t r a i n t s , f or i n f i n i t e l y r ef i ned 2 and 3-D meshes.
57
To achieve convergence, the number of vol umet ri c const r ai nt s must
be r el axed. For t h i s , Nagtegaal et al (1974) proposed a modi f i ed
var i at i onal method which essent i al l y consi st s of performi ng a reduced
i nt egr at i on on t he vol umet r i c st r esses. Thi s can be appl i ed onl y t o
elements i n which f u l l numerical i nt egr at i on comprises more than one
quadrature poi nt . A one-poi nt vol umet ri c i nt egr at i on i s proposed f or
4-node q u a d r i l a t e r a l s and 8-node br i c k s . Thi s method i s c al l ed
" sel ect i ve" i nt egr at i on, and the elements produced possess the desi red
number of const r ai nt s. In pr act i ce, reduced i nt egr at i on i s
of t en extended f or conveni ence t o t he d e v i a t o r i c s t r e s s e s as
wel 1 (Hal 1 qui st,1982b). This excessive r el axat i on may create addi t i onal
problems ("Hourglassing") which must be t r eat ed appropri at el y (secti on
3.7.2). In a Fi n i t e Di f f er ence cont ext t he equi val ent t o one- poi nt
reduced i nt egr at i on i s obt ai ned by usi ng cont our i nt egr al f or mul ae
( Wi l k i ns , 1964).
An al t er nat i ve sol ut i on, st ar t i ng of f from the si mpl est t r i angl es
and t et rahedra, i s t o group several elements together and average the
vol umet r i c s t r ai ns between them. Such a s ol ut i on i s t he "Mi xed
Di scr et i zat i on" (MD) procedure proposed by Mar t i and Cundal l (1982).
This ref ers the devi at or i c st r ai ns to the basic t r i angl es/ t et r ahedr a
and averages t he vol umet r i c components wi t h i n one quadr i l at er al (2
t r i a n g l e s ) or one br i ck (5/ 6 t et r ahedr a) ( sect i on 4.2.2). Thi s
technique has been employed f or the present work. The r esul t of t hi s
MD may be i nt er pr et ed as new, l arger elements, which wi l l be termed i n
t he f o l l o wi n g MTQ (Mi xed Tr i angl es - Quadr i l at er al ) and MTB (Mi xed
Te t r a h e d r a - Sr i c k ) . These el ement s have t he desi r ed number of
const r ai nt s ( t abl e 3.1).
Johnson (1981) has also used a vol umet ri c averaging procedure f or
3-D si mi l ar t o the proposed MD. For 2-D Johnson (1981) employs meshes
of crossed t r i angl es, which were shown by Nagtegaal et al (1974) t o be
l ess s t i f f t han or di nar y t r i a n g l e l ayout s. The behavi our of t hese
crossed l ayout s i s not as good as t hat of MD, however (see sect i on
5. 7).
58
3. 7. 2 "HOURGLASSING"
Some commonl y used meshes admi t c e r t a i n unopposed def or mat i on
modes. These are c a l l e d " zer o- ener gy" or " hour gl assi ng" modes, because
of t he s i mi l i t u d e wi t h an hour gl as s p a t t e r n . I f not d e a l t w i t h ,
hour gl assi ng modes qui c k l y grow and domi nat e t he s o l u t i o n . Examples of
such meshes are q u a d r i l a t e r a l s or b r i c k s , wi t h one- poi nt quadr at ur e i n
FE, or c o n t o u r i n t e g r a l f o r mu l a e i n FO. I n bot h cases Ho u r g l a s s i n g
appear s as t he p r i c e f o r an ex c es s i v e r e d u c t i o n i n t he s t r e s s
i n t e g r a t i o n . However, ant i - hour gl as s i ng t r e a t me n t s do e x i s t and may
s o me t i me s be p r e f e r r a b l e t o t h e e x c e s s i v e c os t of s e l e c t i v e
i n t e g r a t i o n (Goudreau and Ha l l q u i s t , 1982).
The f i r s t a n t i - h o u r g l a s s i n g s o l u t i o n s wer e r e p o r t e d i n t he FO
hydr ocodes (Maenchen and Sack, 1964) , c o n s i s t i n g of a r t i f i c i a l
v i s c os i t y t er ms. A s i mi l a r procedure i n 3-D has been used by Wi l ki ns
et a l . ( 1975) . These t e c h n i q u e s i n v o l v e c o n s i d e r a b l e c o mp u t a t i o n and
are not c o mp l e t e l y i ndependent of r i g i d body modes. A more r i g o r o u s
t r eat ment of hour gl assi ng has been f ol l owed by Fl anagan and Bel yt schko
( 1981b) , who pur sue t he i dea of t he o r t h o g o n a l i t y of h o u r g l a s s i n g
modes t o t he f i r s t - o r d e r modes, f i r s t proposed by Kosl of f and Fr azi er
(1978). Flanagan and Bel yt schko use an addi t i onal c o n t r i b u t i o n t o t he
s t i f f n e s s mat r i x , c a l l e d s t a b i l i z a t i o n mat r i x .
From a more p r a c t i c a l v i e wp o i n t Goudreau and H a l l q u i s t (1982)
have c o n s i d e r e d Fl anagan and Be l y t s c h k o ' s a l g o r i t h ms t o i n v o l v e an
ex c es s i v e c o mp u t a t i o n a l c o s t . Fu r t h e r mo r e , exact t r e a t me n t of
hour gl assi ng i s not necessary, as hour gl assi ng modes are gl obal modes,
and el ement hour gl assi ng i s of t en a st abl e ki nemat i c component of a
gl obal energy mode. I n t he DYNA codes Ha l l q u i s t (1982b) uses an a n t i -
h o u r g l a s s i n g v i s c o s i t y based on F l a n a g a n ' s or t hogonal modes,
s i mp l i f i e d t o produce an oper at i on count 4 t o 5 t i mes l ower .
Reduced s e l e c t i v e i n t e g r a t i o n or " Mi xed Di s c r e t i z a t i o n " r es i s t
t he h o u r g l a s s i n g modes " n a t u r a l l y " w i t h d e v i a t o r i c s t r e s s e s , by -
passi ng t he need f or any speci al t r eat ment . A quest i on mark i s put by
Goudreau and H a l l q u i s t (1982) on whet her t he added c os t of t hese
t echni ques i s wor t hwhi l e i n t er ms of i mproved accuracy. Ad d i t i o n a l l y ,
f or some appl i c at i ons l i k e hi gh- pr essur e shock waves where st r engt h i s
59
not a f a c t o r , dev i at or i c st r esses may not be abl e t o r es t r ai n
hourgl assi ng.
3.8 CONCLUSIONS
Fi ni t e Di f f erence and Fi ni t e Element methods provi de al t er nat i ve
sol ut i ons f or spat i al semi di scr et i zat i on, wi t h t hei r own independent
school s and l i t e r a t u r e . However, recent t r ends i n Fi n i t e El ements
avoi d t he assemblage of gl obal c o e f f i c i e n t mat r i ces (e. g. e x p l i c i t
methods or i mpl i c i t methods wi t h r el axat i on techniques f or equation
sol vi ng) . In t hese cases FE and FD approaches are concept ual l y very
si mi l ar .
Lagr angi an, Eul er i an or ALE t echni ques pr ovi de a l t e r n a t i v e
descr i pt i ons f or t he mesh, each of them wi t h poi nt s i n f avour and
a g a i n s t . Lagr angi an meshes ar e gener al l y pr ef er r ed f or s o l i d
mechani cs. Vari ous nonl i near f or mul at i ons are pr esent ed, each best
sui t ed t o a di f f er ent cl ass of problems. Total Lagrangian f or mul at i ons
are advantageous f or hyper el ast i c- t ype mat eri al s (behaviour described
on the or i gi nal conf i gur at i on) , whi l e Cauchy stress f ormul at i ons are
pr ef er r abl e f or hypoel asti c and p l a s t i c mat er i al s ( descr i bed on t he
current conf i gur at i on) . Equi val ent al gori t hms f or a given mat eri al may
gener al l y be f ound i n bot h f or mul at i ons , a l b e i t at t he expense of
cost l y t ransf ormat i ons between current and or i gi nal conf i gur at i ons.
For ti me i nt egr at i on, the choice i s between i mpl i c i t and ex pl i c i t
schemes. Both possess very di f f er ent char act er i st i cs as to s t a b i l i t y ,
code ar chi t ect ur e and storage r equi r ement s. Ex p l i c i t t echni ques are
wel l sui t ed f or problems dominated by high frequency components (wave
propagati on), or which are steepl y non- l i near , and of short dur at i on.
I mpl i c i t techniques are t o be pref erred f or i nert i a-domi nat ed models
( l ow f r e q u e n c i e s ) , l ess n o n - l i n e a r , or of l onger d u r a t i o n .
The di r ect el i mi nat i on techniques used general l y f or equation sol vi ng
i n i mp l i c i t methods sever el y l i mi t t he si ze of probl ems t hat can be
modelled, speci al l y i n 3-D.
The new El ement -By-El ement t echni ques proposed r ecent l y are
pr omi si ng a l t e r n a t i v e s . They appear t o be most usef ul as l i near
60
equat i on sol ver s wi t h i n i mp l i c i t methods. As uncondi t i onal l y st abl e
expl i ci t ti me operat ors, accuracy problems s t i l l need t o be overcome
before they are used i n producti on codes.
Fi nal l y, "l ocki ng-up" and "hourgl assi ng" are i mportant pr act i cal
problems rel at ed t o the spat i al semi di scr et i zat i on, r equi r i ng speci al
at t ent i on; i f not provided f or , convergence may not occur i n otherwi se
t h e o r e t i c a l l y "sound" meshes. The method of t ackl i ng these problems
often condi t i ons st rongl y the overal l approach of the numerical model.
61
CHAPTER 4
EXPLICIT FINITE DIFFERENCE NUMERICAL MODEL
4.1 INTRODUCTION
4.1.1 General methodology
4.2 SPATIAL SEMIDISCRETIZATION
4.2.1 Constant strain triangles and tetrahedra (CST elements
4.2.2 Mixed Discretization (MTQ, MTB elements)
4.2.3 Prevention of negative volumes (MTQC, MTBC elements)
4.2.4 Mass lumping procedure
4.3 MOMENTUM BALANCE
4.4 CENTRAL DIFFERENCE TIME INTEGRATION
4.5 CONSTITUTIVE MODELS
4.5.1 Hypoelasticity
4. 5. 2 Pl as t i c i t y ; radi al ret urn al gori t hm
4.5.3 Hardening and uni axi al st r ess- st r ai n laws
4.5.4 Obj ecti ve stress rates
4.6 DAMPING
4.7 STABILITY OF TIME INTEGRATION
4.8 MODELLING OF CONTACTS
4.8.1 Contact interface laws
4.8.2 Contact detection algorithm
4.9 HEAT CONDUCTION
62
4.10 ENERGY COMPUTATIONS
4.11 INPLEMENTATION INTO FORTRAN PROGRAM
63
4. 1 INTRODUCTION
Numer i cal F i n i t e Di f f e r e n c e (FD) t e c h n i q u e s f o r t he n o n l i n e a r
anal ysi s of 2 and 3-Di mensi onal e l a s t i c - p l a s t i c cont i nua are descr i bed
her e. E x p l i c i t c e n t r a l d i f f e r e n c e i n t e r p o l a t i o n s ar e used bot h i n
space and t i me domai ns, wi t h t r i a n g u 1 a r / t e t r a h e d r a l c e l l s f o r t he
s pat i al s e mi d i s c r e t i z a t i o n . No r e s t r i c t i o n s are pl aced on t he t opol ogy
of t he mesh, whi ch i n t h i s aspect i s i d e n t i c a l t o t hose of F i n i t e
Element methods.
The method out l i ned i s a general purpose one and may be appl i ed
t o bot h l i n e a r and nonl i near , t i me- mar chi ng t hermomechani cal anal ysi s
of cont i nua; however, i t w i l l be most advantageous f or shor t dur at i on
hi ghl y nonl i near pr obl ems, such as ar i s e f r om i mpact . A gr eat freedom
i s av ai l abl e f or i mpl ement at i on of mat er i al c o n s t i t u t i v e l aws. For t he
present work ( mai nl y i n non- cycl i c l oadi ng of met al s) mat er i al l aws
have been r e s t r i c t e d t o el a s t i c - p i a s t i c Von Mi ses model s w i t h
k i n e ma t i c - i s o t r o p i c har deni ng.
Ma t e r i a l dampi ng i s i n c l u d e d , p e r mi t t i n g t he use of dynami c
r el ax at i on f or q u a s i - s t a t i c pr obl ems. A numer i c al c o n t a c t a l g o r i t h m
model s t he i n t e r f a c e bet ween c o n t i n u a , o f t e n pr es ent i n n o n l i n e a r
i mpact c a l c u l a t i o n s . Fi n a l l y , f or gener al i t y , t her mal ef f ec t s al l owi ng
heat gener at i on/ conduct i on and mat er i al dependence on t emper at ur e have
been i ncl uded, enabl i ng a f u l l y coupl ed t hermomechani cal anal ysi s.
The t h e o r y and numer i c al t ec hni ques d e s c r i b e d bel ow appl y i n
general f or 2 and 3-D model s. A 2-D code wi t h axi symmet r i c c a p a b i l i t y
was devel oped f or t h i s t h e s i s , i mpl ement i ng t he al gor i t hms descr i bed.
I n gener al , i mpl ement at i on d e t a i l s f o r t he al gor i t hms are gi ven onl y
f or 2-D. Some al gor i t hms (e.g. t her mal e f f e c t s , pr event i on of negat i ve
vol umes) have al so been i mpl ement ed by t he aut hor i n an e x i s t i n g 3-D
pr ogr am ( Ma r t i ( 1981, 1983) ) , ( Ma r t i , Go i c o l e a , K a l s i , and Macey
( 1984) ) , ( Ma r t i , K a l s i , and Last ( 1984) ) .
4.1.1 GENERAL METHODOLOGY
Before entering into details of the numerical techniques, it is
64
u s e f u l t o d e s c r i b e t h e p h i l o s o p h y of t i me - ma r c h i n g e x p l i c i t
c a l c u l a t i o n s . These pr ovi de a set of uncoupl ed di s c r et e equat i ons of
mot i on, whi ch may be sol ved e x p l i c i t l y wi t hout assembl i ng or i n v e r t i n g
gl obal s t i f f n e s s or mass mat r i ces. The program ar qui t ec t ur e i s qui t e
d i f f e r e n t f rom t he usual l i near ( s t a t i c or dynami c) or i mp l i c i t non-
l i near t echni ques, al l owi ng a gr eat er gener al i t y i n t he t r eat ment of
nonl i n e a r i t i e s ( l ar ge def or mat i ons, mat er i al behavi our , cont act s) .
I t i s assumed t hat t he f unc t i onal dependence of t he di spl acement s
on space and t i me can be s e p a r a t e d , g i v i n g r i s e t o i ndependent
semi d i s c r e t i z a t i o n s f o r bot h domai ns. The mesh f o r t he s p a t i a l
s e mi d i s c r e t i z a t i on i s Lagr angi an ( s e c t . 3. 4. 1) , i . e. t he g r i d nodes
r epr es ent ma t e r i a l p o i n t s and move wi t h t hem. The t i me domai n i s
di vi ded l i n e a r l y i nt o t i me- s t eps .
The comput at i onal cycl e per f or med t o advance t he probl em i n t i me
may be d e s c r i b e d as f o l l o ws ( f i g . 4. 12) . At a p a r t i c u l a r i n s t a n t ,
coor di nat es, v e l o c i t i e s and accel er at i ons are known. The accel er at i ons
are i nt egr at ed wi t h a cent r al di f f er enc e scheme f or each g r i d - p o i n t ,
t o f i n d t he new v e l o c i t i e s and di spl acement s. These are then used i n
s pat i al di f f er enc e equat i ons t o i n t e r p o l a t e t he def or mat i on gr adi ent s
and s t r a i n s f o r each c e l l . Ap p l y i n g t he ma t e r i a l c o n s t i t u t i v e
e q u a t i o n s , s t r a i n s gi v e r i s e t o s t r e s s e s , whi ch ar e t hen i n t e g r a t e d
around each g r i d - p o i n t t o pr ovi de nodal f or ces. Di v i di ng by t he mass
( l umped at nodes) t he new a c c e l e r a t i o n s may be f ound. Here t he
comput at i onal cycl e i s compl et e and may be recommenced f or t he next
t i me- s t ep. Thi s process cont i nues u n t i l t he t o t a l t i me f or anal ysi s i s
reached. Nonl i near boundary c ons t r ai nt s gi v i ng r i s e t o di spl acement or
f or ce condi t i ons are i nt r oduced i nt o t he cycl e e x p l i c i t l y .
For t he t her mal anal ysi s t he comput at i ons advance i n p a r a l l e l and
c oupl ed w i t h t he mechani cal c y c l e . I n f l u x of heat i n t o a c e l l f r om
heat sour ces, heat conduct i on, or produced by p l a s t i c wor k, gi ves r i s e
t o t e mp e r a t u r e i n c r e me n t s . The g r a d i e n t s of t he new t e mp e r a t u r e s
c r e a t e heat f l u x e s ( c a l c u l a t e d f r om F o u r i e r ' s c o n d u c t i o n l a w) ,
al l owi ng r e p e t i t i o n of t he cycl e f or t he next t i me- s t ep.
A p r i c e i s pai d f o r t he g e n e r a l i t y and s i m p l i c i t y of e x p l i c i t
al gor i t hms , i n t hat t he comput at i ons ar e onl y c o n d i t i o n a l l y s t a b l e ,
65
l i m i t i n g t he s i z e of t he t i me - s t e p t o s mal l v a l u e s . However , t he
o p e r a t i o n count per t i me - s t e p i s much s ma l l e r t han f o r i m p l i c i t
methods. The c r i t i c a l t i me- s t ep may be est i mat ed ac c ur at el y , and t he
r obus t nes s of t he c o mp u t a t i o n s does not s u f f e r as a r e s u l t of t he
c o n d i t i o n a l s t a b i l i t y . The mai n i mp l i c a t i o n of t he t i me - s t e p
r e s t r i c t i o n i s t he l ar ge number of comput at i onal cycl es necessary f or
medium or l ong dur at i on anal yses, whi ch of t en become uneconomi cal .
I n e x p l i c i t c a l c u l a t i o n s , t he l i m i t i n g f a c t o r i s g e n e r a l l y t he
CPU t i me , r a t h e r t han cor e s t o r a g e . When c o n s t r u c t i n g t he mesh, i t
must be t aken i nt o consi der at i on t hat t he t i me- s t ep i s pr opor t i onal t o
t he minimum el ement si ze. A l ocal r ef i nement of t he mesh, even i f i t
onl y adds a smal l number of el ement s, may i ncr ease gr eat l y t he number
of t i me- st eps necessary and hence t he comput at i onal cost .
Quas i - s t at i c m odels
By t h e i r nat ur e, e x p l i c i t comput at i ons are e s s e n t i a l l y dynami c,
r e l y i n g on numer i cal i n t e g r a t i o n of t he equat i ons of mot i on wi t h non-
zero masses f or t he t i me- mar chi ng s o l u t i o n . No s i mp l i f i c a t i o n of t he
c y c l e i s p o s s i b l e f o r q u a s i - s t a t i c a n a l y s i s . However , many such
probl ems can s t i l l be sol ved advant ageousl y, usi ng dynamic r el ax at i on
(damping out v i br at i ons u n t i l a st eady- st at e i s reached), or v e l o c i t y
scal i ng ( i ncr easi ng def or mat i on v e l o c i t y i n a c o n t r o l l e d manner).
Dynami c r e l a x a t i o n may be ar gued t o be a more n a t u r a l appr oach
than t he s t a t i c s t i f f n e s s equat i ons (3. 12), (3.13). As i n t he physi cal
wo r l d , wi t h dynami c r e l a x a t i o n s t a t i c e q u i l i b r i u m i s r eached as a
l i m i t i n g s t e a d y - s t a t e . I n g e n e r a l , s o l u t i o n w i l l r e q u i r e a g r e a t e r
comput at i onal e f f o r t , al t hough i f adapt at i ve t echni ques are used f or
mass and damping paramet ers (Underwood, 1983), t he di f f er enc e may be
s mal l . One d e f i n i t e advantage e x p l i c i t r el ax at i on methods have, whi ch
may be over whel mi ng i n l ar ge 3-D syst ems, i s t hat t hey do not need t o
st or e any l ar ge gl obal c o e f f i c i e n t mat r i ces. For non- l i near pr obl ems,
dynami c r e l a x a t i o n r e q u i r e s some car e so as not t o appl y t oo l a r g e
l o a d - s t e p s whi c h woul d cause an i r r e v e r s i b l e ov er s hoot i n t he
s ol ut i on. Ad d i t i o n a l l y , i f t he l oads are appl i ed gr adual l y, t he amount
of r e l a x a t i o n needed a f t e r r eac hi ng t he maxi mum l oad i s g r e a t l y
reduced.
66
Vel oci t y scal i ng ex pl oi t s t he f act t hat f or s u f f i c i e n t l y slow
l oadi ng, the r el axat i on process may be ommitted al t oget her , wi t hout
excessive er r or s. This al l ows model l i ng a sl ow, quasi - st at i c process
merel y by c ont r ac t i ng t he t i me- s c al e. As i n dynamic r e l a x a t i o n ,
l oadi ng must be sl ow enough not t o cause over shoot s. The er r or s
i ncurred i n t hi s process are i n the form of el ast i c stress waves l e f t
t r avel l i ng i n the model. For el ast i c analyses, vel oci t y scal i ng wi l l
not be appropri ate i n general , as the i nt ensi t y of these waves wi l l be
i mportant wi t h respect t o the overal l stresses. For pl ast i c analyses
however, t he er r or w i l l merel y be i n t he f orm of an e l a s t i c " noi se"
superposed on t he smoother q u a s i - s t a t i c sol ut i on (see also secti ons
6. 5. 1, 7. 3. 2. 2) .
4.2 SPATIAL SEMIDISCRETIZATION
A di scret e mesh of t r i angl es (2-D) or t et rahedra (3-D) i s af f i xed
on t he i n i t i a l ( r ef er ence) c onf i gur at i on of t he cont i nuum, and
p a r t i c l e s are f ol l owed t hr ough t i me (Lagrangi an mesh). The cur r ent
conf i gur at i on i s a f unct i on of the reference conf i gur at i on and t i me:
x = x( X, t ) (4. 1)
Displacements (u = x-X), vel oci t i es (v=u) and accel erat i ons (a=ii) are
defined at the nodes. Deformation gradi ents F, rates of deformati on d,
st r ai ns E, and st r esses crare r ef er r ed t o t he c ent r oi d of each c e l l
( f i g . 4.1).
The comput at i ons r el at ed t o t he d i s c r e t i z a t i o n f a l l i nt o t wo
areas: f i r s t l y the i nt er pol at i on of st r ai ns and deformati on gradi ents
from the nodal di spl acements, and secondly the i nt egr at i on of stresses
around each gr i d- poi nt to obtai n the nodal f orces.
Tri angl es and Tetrahedra are the only element types which provide
a one- t o- one c o r r e l a t i o n between t he set s of a l l possi bl e cor ner
movements and def or mat i on gr adi ent t ensor s. Thi s f ac t i nher ent l y
el i mi nat es zero-energy ("hourglassing") def or mat i on modes (unopposed
movements of the corner nodes), which occur f or other elements such as
67
q u a d r i l a t e r a l s or b r i c k s wi t h r educed i n t e g r a t i o n . I t al s o a l l o ws a
d i r e c t o b t e n t i o n of t he g r a d i e n t s f r om t he d i f f e r e n c e equat i ons
wi t hout usi ng cont our i nt egr al s ( eqn. 3. 4)
However, meshes of const ant s t r a i n t r i a n g l e s or t et r ahedr a of t en
" l ock- up" when model l i ng p l a s t i c f l o w, gi v i ng sol ut i ons whi ch are very
o v e r s t i f f (see sect . 3.7.1). The Mi xed Di s c r e t i z a t i o n (MO) procedures
pr oposed by Ma r t i and Cundal l ( 1982) ar e empl oyed her e t o over come
t h i s pr obl em. The new el ement s cr eat ed by Mi xed Di s c r e t i z a t i o n w i l l be
c al l ed MTQ (Mi xed Tr i angl es - Quadr i l at er al ) and MTB (Mi xed Tet r ahedr a-
Br i c k ) .
Mi x e d D i s c r e t i z a t i o n i s a t t r a c t i v e f o r e l a s t i c - p l a s t i c
c al c ul at i ons because i t al l ows t he use of si mpl e el ement s, whi ch avoi d
z er o- ener gy modes n a t u r a l l y , wi t h o u t t he need f o r a r t i f i c i a l a n t i -
h o u r g l a s s i n g t er ms (see s ec t . 3. 7. 2) . However , i t e l i mi n a t e s an
advantage of t r i a n g u l a r / t e t r a h e d r a l meshes, i . e. t he i mp o s s i b i l i t y of
t a n g l i n g over due t o a s y mp t o t i c a l l y i n f i n i t e r e s i s t a n c e t o zer o or
n e g a t i v e vol umes i n each c e l l . Thi s pr obl em need onl y cause concer n
f or meshes wi t h very l ar ge d i s t o r t i o n s . Never t hel ess, f or such cases
i t does c o n s t i t u t e a s e r i o u s dr awback, c aus i ng t he c o mp u t a t i o n s t o
cr ash unl ess t he mesh i s rezoned.
To over come t he pr obl em of t a n g l i n g over i n MD meshes, a new
f ami l y of cor r ect ed el ement s i s proposed i n sect i on 4.2.3: cor r ect ed
MTQ (MTQC) and cor r ect ed MTB (MTBC) f or 2 and 3-D r es pec t i v el y . These
el ements recover as y mpt ot i c al l y t he r esi st ance t o zero volumes of t he
basi c CST e l e me n t s , when one c e l l i n t he Mi xed D i s c r e t i z a t i o n gr oup
becomes much s ma l l e r t han t he r e s t and t ends t o zer o vol ume. The
a b i l i t y of model l i ng i ncompr essi bl e f l ow i s mai nt ai ned, t hus pr ovi di ng
a more robust a l t e r n a t i v e t o MD.
4. 2. 1 CONSTANT STRAIN TRIANGLES AND TETRAHEDRA (CST ELEMENTS)
Const ant s t r a i n t r i a n g l e s and t e t r a h e d r a (CST) may be used
d i r e c t l y as such f or e l a s t i c anal yses, or as t he basi c i ngr edi ent f or
MD ( sect . 4.2.2) or c o r r e c t e d MD ( s e c t . 4. 2. 3) meshes.
68
i n t e r p o l a t i o n of def or mat i on gr adi ent s
Vel oci t y gr adi ent s are i nt er pol at ed at t he c ent r oi d of t he c e l l
f r om d i r e c t use of d i f f e r e n c e e q u a t i o n s . Nodes ar e number ed N=0, l , 2
( t r i a n g l e ) or N=0, l , 2, 3 ( t e t r a h e d r o n ) ( f i g . 4. 1) . Node 0 i s t ak en as
an a r b i t r a r y r ef er ence poi nt . The di f f er enc e equat i ons t ake t he f or m:
V
P =
v
. .
( X
N _
X
Q)
( 4. 2,
where super scr i pt s r ef er t o nodes and subscr i pt s r ef er t o components.
Usi ng t he not at i on ( ) I J=( ) ! - ( ) J , ( 4. 1) may be wr i t t e n as f o l l o ws :
v
i
v
i , J
x
J
( 4. 3)
This may be put into matrix form and inverted, to obtain
[ v
i 5 J
J = [VR(i,N)JLXR(N,j)J
-1
(4.4;
wher e t he n o t a t i o n VR( i , NJ =v ^ , XR( N, j ) =x J has been used. Al l
mat r i ces i n eqn. ( 4. 4) are 2x2 f or 2-D or 3x3 f or 3-D.
The r at e of def or mat i on and spi n r at e t ensor s may be obt ai ned as:
d
i j
=
<
v
i , j
+ v
J , i
) / 2
w
i j
=
<
v
i , j "
v
j , i
) / 2
( 4. 5)
For 2-D addi t i onal consi der at i ons must be made t o det er mi ne t he
o u t - o f - p l a n e component s. Assumi ng ax i s no. 3 p o i n t i n g out of t he
pl ane, t hese condi t i ons are:
- Plane s t r a i n , d33=0;
- Plane s t r es s , 033=0; t h i s i s a st r ess c ondi t i on whi ch must be used
t oget her wi t h t he s t r e s s - s t r a i n law t o obt ai n t he appr opr i at e r at e of
d e f o r ma t i o n ( eqns. 4 . 4 1 , 4. 42) ;
- Axi symmet ri c ( axi s 2 of symmet r y) , d
33
= y\/^\ at each poi nt i n t he
cont i nuum; t h i s c ondi t i on may be r el axed by enf or ci ng i t onl y at t he
69
Finure 4.1: Trianaular and tetrahedral cells
Fioure 4.2: Path for stress inteqral around node 1
70
t r i angl e cent r oi d,
d
= (v
+
vj
+
vj ) / 3 (4. 6)
3 3
(x + x{ + xf ) / 3
Generalized natural st r ai n increments are defi ned as
Ae-ij = d^At ( 4. 7)
where At i s t he comput at i onal t i me- s t ep. These i ncr ement s may be
decomposed i nt o:
- Volumetric par t , Ae
v
= A6^
- Devi at ori c par t , A2-\j = ASjj - 5-jj/iC
kJ<
/3 (4. 8)
I nt egr at i on of stresses t o obtai n nodal forces
The i nt er nal f or ce appl i ed at each node 1 may be obt ai ned by
i n t e g r a t i n g t he st r esses i n a cl osed sur f ace cont ai ni ng t he node,
S ( D:
PJ
1}
= \ ^i j nj dS (4.9)
S( l )
where nj "is t he ( uni t ) out ward normal t o S( l ) . S( l ) i s const r uct ed i n
2-D ( f i g . 4.2) by f or mi ng a pol ygon j o i n i n g t he cent r oi ds of t he
t r i angl es and centres of the sides shari ng node 1, agbpcqdresfta. In
f act , t hi s polygon i s only a cross-sect i on of surface sC) ; f or plane
s t r a i n and pl ane st r ess S^ ) i s a sl ab of t hi ckness t , whi l e f or
axisymmetric models i t i s the surface of r evol ut i on generated r ot at i ng
t he pol ygon around t he axi s of symmet ry. A s i mi l a r sur f ace may be
const r uct ed i n 3-0 f rom a pol yhedron wi t h cor ner s i n cent r oi ds of
t et rahedra, cent roi ds of faces and centres of edges.
The c o n t r i b u t i o n of each c el l t o t he nodal f or ces i s det ai l ed
below f or 2-D. The not at i on of f i g . 4.2 wi l l be f ol l owed f or node and
cel l numbers. Di f f er ent f ormul at i ons are necessary f or plane st r ai n or
stress and axi symmetri c models.
71
Plane s t r a i n or pl ane st r ess
From eqn. ( 4. 9) t he c o n t r i b u t i o n of c e l l C t o t he f o r c e at 1 i s
obt ai ned as f o l l o ws :
pj
1 ) C
= t f CT^-rK-ds (4.10)
"IJ J
agb
The t h i c k n e s s t i s c o n s t a n t (=1) f o r pl ane s t r a i n but may var y f o r
pl ane st r ess. Consi der i ng Gauss' t heor em, t he i nt egr al i n eqn. (4.10)
may be expressed as f o l l o ws :
CT
ij
n
J
ds =
J
a
i j
n
j
d s +
J
a
i j
n
j
d s =
J
ff
ij,j
dA +
J * i j
n
j
d s ( 4
'
U )
agb agba ab ^agba
a b
where A
a
qt,
a
is the plane area enclosed by agba. As the divergence of a
constant f i e l d i s nul l , the i nt egr al over t hi s area i s zero. The nodal
f orce becomes:
p j
1
^ = t fffj f njds (4.12)
ab
(
CT
11
CT
12\
or e x p l i c i t l y i n mat r i x f or m:
LPJDC,
p
l)Cj
=
t
[ x
n
mj x
m
nj
[ " " |
( 4 > 1 3 )
r
12
CT
22
Axi symmet ri c
Cy l i n d r i c a l c o o r d i n a t e s ar e used f o r d e s c r i b i n g a x i s y mme t r i c
models ( xi r adi al coor di nat e, x2
a x i a l
coor di nat e) . The di ver gence of
a c o n s t a n t v e c t o r i s no l o n g e r z e r o as i t was f o r C a r t e s i a n
coor di nat es:
d i v L " i ! v
iZ
*
i 3
J = ffjn + ^ 3 3 + ^ i 2 , 2 /
x
l
+ o r
i l / x i = ^ i i / X i ( 4. 14)
72
The c o n t r i b u t i o n of a c e l l C t o t he i nt egr al i n eqn. ( 4. 9) becomes
/ ^ i j n j d S + j
CT
ijnjdS = / ( ^ i/x
l
)Znx.
l
dh + I
CTi
j ^ / r x ^ j d s =
s
agba
s
ab
A
agba
a b
; 4. 15)
= (7r/ 6)A
C
or
i l
+ 2 7T x } ^ - j |ab| n j
b
where: S
a q t ) a
, S
a b
denote sur f aces of r ev ol ut i on generat ed by agba & ab
AC t o t a l Cel l area
A
agba = AC/6 , area encl osed by t r i a n g l e agba
|ab| , n are t he l engt h and u n i t normal of segment ab
In addi t i on t o t he t erms f r om eqn. (4. 15), r adi al f or ces are produced
by t he hoop s t r e s s e s <7
33
. These f o r c e s ar e not i n c l u d e d i n ( 4. 15) , as
t he i n t e g r a l s cover t he c ompl et e 2 t o r o i d wher e o p p o s i t e r a d i a l
component s cancel out ( f i g . 4. 3) . However , i f one c o n s i d e r s o n l y a
sect or of angl e <p :
PJD = r
P
^ )
d =
^ D
=
J a
33dA
and t he c o n t r i b u t i o n of c e l l C f or t he compl et e 2 i s :
p{DC
=
1
A
Zo
33
2n= 2j a
3 3 A
C
( 4 < 1 6
whi ch added t o ( 4. 15) gi ves t he t o t a l nodal f o r c e c o n t r i b u t i o n s . I n
mat r i x f or m,
y
CT
12
a
2 2 / ( 4. 17)
+
6
A C
L
f f
l l
CT
12J
+
3 -
A C
C ^ 3 3 J
4. 2. 2 MIXED DISCRETIZATION (MTQ, MTB ELEMENTS)
Mi xed d i s c r e t i z a t i o n pr ocedur es ar e d e s c r i b e d i n t h i s s e c t i o n .
Thi s t ec hni ques av oi d t he ex c es s i v e s t i f f n e s s ot her wi se associ at ed
wi t h const ant s t r a i n t r i a n g l e s or t et r ahedr a f or i ncompr essi bl e f l ow
(see sect . 3. 7. 1) .
73
Ac c or di ng t o i Magtegaal , Par ks and Ri ce ( 1974) , f o r a c c u r a t e
model l i ng of i ncompr essi bl e p l a s t i c f l o w, t he r a t i o between t he number
of Degrees Of Freedom (DOF) i n t he mesh and t he number of c ons t r ai nt s
imposed by t he i nc ompr es s i bi 1i t y c ondi t i on must be > 1. For Const ant
S t r a i n T r i a n g l e s , The r a t i o DOF/ c o n s t r a i n t s f o r i n f i n i t e l y r e f i n e d
pl ane s t r a i n and a x i s y mme t r i c meshes i s equal t o 1 ( c o n s i d e r i n g f o r
t he a x i s y mme t r i c case t h a t t he r a d i a l l y v a r y i n g hoop s t r e s s e s are
reduced t o t he val ue at t he el ement c ent r oi d (eqn. 4. 6), ot her wi se t he
r a t i o woul d be 1/ 3) ( t a b l e 3. 1) . For t e t r a h e d r a f o r mi n g b r i c k s i n a
r egul ar l a t t i c e , t he r a t i o i s e i t h e r 3/5 or 3/6 dependi ng on whet her
each br i ck i s composed of 5 or 6 t et r ahedr a ( t abl e 3.1).
Nagt egaal et a l . (1974) s u g g e s t e d a m o d i f i e d v a r i a t i o n a l
p r i n c i p l e t o i mprove meshes whi ch woul d ot her wi se be uns ui t abl e. The
i dea behi nd t h i s modi f i c at i on i s t o make sure d i l a t a t i o n i s governed
by f ewer par amet er s t han f o r c o n v e n t i o n a l e l e me n t s . S t a r t i n g f r om
el ement s wi t h more t han one i n t e g r a t i o n p o i n t , t h i s i s ac hi ev ed by
a v e r a g i n g t he v o l u me t r i c s t r a i n s w i t h i n each el ement ( e f f e c t i v e l y
p e r f o r mi n g a r educed i n t e g r a t i o n on t hem) . Based on t he same
p r i n c i p l e , Mar t i and Cundal l (1982) have proposed Mi xed Di s c r e t i z a t i o n
(MD) pr ocedur es v a l i d f o r FE or FD meshes w i t h l o we r or der el ement s
(one i n t e g r a t i o n p o i n t ) , i n whi ch vol umet r i c s t r ai ns are averaged f or
groups of sever al el ement s.
The i d e a l val ue f o r t he r a t i o DOF/ c o n s t r a i n t s i s t h a t of t he
cont i nuum, i n whi ch f or each mat er i al poi nt t her e i s one vol umet r i c
equat i on and U ( no. of space di mens i ons ) DOF. Usi ng MD t hese i d e a l
r a t i o s are ac hi ev ed i f a v e r a g i n g i s per f or med w i t h i n t he f o l l o w i n g
MD groups ( t abl e 3. 1) :
- 2 Tr i angl es i n one Quadr i l at er al f or 2-D (MTQ el ement s) ;
- 5 or 6 Tet r ahedr a i n one Br i ck f or 3-D (MTB el ement s) .
I f Ci ( i = l t o M) st ands f o r t he i n d i v i d u a l c e l l s and f o r t he
t o t a l el ement (MTQ or MTB), t he Mi xed D i s c r e t i z a t i o n pr oc edur e
consi st s si mpl y of s u b s t i t u t i n g t he vol umet r i c s t r a i n s i n each c e l l by
t he vol ume-wei ghed average of t he group:
4.3: Radial forces from Hoop stresses (axisymmetric mooel)
.40
20
oo
80
60
40
20
00
80
60
40
20
00
CORRECTED M.D. ( MTQCELEMENTS)
ORDINARY M.D. (MTQ ELEMENTS)
NO M.D. (CST ELEMENTS)
E =
"V =
Y -.
= 67x10'
= 0. 3
-- 120x10
0.00 0.20 0.40 0.60 0.80
ADIMENSIONAL DISPLACEMENT. 2
w
d /a
Figure 4.4= RESISTANCE TO ZERO-VOLUME FOR DIFFERENT ELEMENT TYPES
75
M M
i = l i = l
where V^i are t he volumes of t he i ndi v i dual c e l l s . M=2 f or 2-0 and M=5
or 6 f o r 3- D. For axi s y mmet r i c model s, t he aver age i n eqn. ( 4. 18) i s
ar ea- wei ghed i n s t e a d , i n or der t o av oi d pr obl ems near t he ax i s of
symmetry.
Apart f r om havi ng t he i deal number of vol umet r i c c o n s t r a i n t s , MTQ
or MTB el ement s w i l l pr oduce non- z er o s t r a i n s and s t r e s s e s i n t he
component c e l l s f o r any set of c or ner movement s. No z e r o - e n e r g y
hour gl assi ng may occur , as i t i s opposed by dev i at or i c st r esses.
V a l i d a t i o n exampl es f o r MO pr ocedur es i n s i mp l e p l a s t i c i t y
probl ems are pr ovi ded i n s ec t i ons 5. 4. 1 and 5. 4. 2.
4. 2. 3 PREVENTION OF NEGATIVE VOLUMES (MTQC, MTBC ELEMENTS)
One i mpor t ant advantage of CST el ement s i s t hat no t angl i ng over
may occur wi t h o u t o c a s s i o n i n g n e g a t i v e vol umes i n t he basi c c e l l s
(Johnson, 1976). A hypoel ast i c mat er i al w i l l generat e r e s i s t i n g f or ces
t hat grow t owar ds i n f i n i t e as t he volume t ends t o zer o. Thi s advantage
i s l o s t by MO, whi ch accept s zer o vol ume i n one c e l l wi t h o u t
gener at i ng i n f i n i t e r e s i s t i n g pr essur es, as l ong as t he whol e MO group
has a non-zero vol ume. Thi s may be seen l e t t i n g t he volume of one c e l l
Cj t end t o zer o:
i f V
C j
* 0
CST: Z\P = K4J = K(/AV
C: i
/V
Cj
' ) * - CO
MO : 4P = K 4 6
v
= K( 4eJv
C: i
+y ^ C i
v
C l
) / v
E
- ^ >K( \ A6
1
V
C l
) / V
E
i t J i 3*j
( f i n i t e )
Thi s f a c t has a number of u n d e s i r a b l e s i d e - e f f e c t s . Lar ge
r educt i ons i n volume are usual l y associ at ed wi t h l ar ge d i s t o r t i o n s i n
76
s h a p e , c a u s i n g c o n s i d e r a b l e n u me r i c a l e r r o r s as t he el ement
i n t e r p o l a t i o n s become i l l - c o n d i t i o n e d . On t he ot her hand, t he c r i t i c a l
t i me- s t ep w i l l be reduced, i ncr easi ng t he cost s gr eat l y . F i n a l l y , i f
n e g a t i v e v o l u me s o c c u r f r o m t a n g l i n g o v e r i n t h e mes h, t h e
c a l c u l a t i o n s w i l l crash. I t must be not ed t h a t t hese pr obl ems occur
onl y f o r s p e c i a l cases w i t h h i g h l y d i s t o r t e d Lagr angi an meshes.
However t hey do det r act s er i ous l y f r om t he robust ness of MD procedures
f or general a p p l i c a t i o n s .
Johnson (1981) has r ecogni zed t h i s pr obl em, but hi s s ol ut i on has
been an i nc ons i s t ent use of aver agi ng pr ocedur es. For c er t ai n c r i t i c a l
por t i ons of t he c a l c u l a t i o n s , aver agi ng i s swi t ched o f f , i n or der not
t o i n c u r i n o v e r l a p p i n g . No e x p l a n a t i o n i s gi v en by Johnson f o r t he
i n e v i t a b l e l oss of accur acy, except t he conveni ence of t he anal yst .
A c o n s i s t e n t s o l u t i o n i s p r o p o s e d h e r e u s i n g a Mi x e d
Di s c r e t i z a t i o n mesh t hat recovers t he r esi st ance t o negat i ve volumes
of CST el ement s, when f or an i ndi v i dual c e l l t he volume tends t o zer o,
V
!
-' J->U. Thi s t akes t he shape of a c o r r e c t i o n t o t he aver aged
v o l u me t r i c s t r a i n i nc r ement (A
v
) f r om eqn. ( 4. 18) :
AeJ
1
:
= A( +ami n( AeJ , 0) i ; v
E
/ ( MV
Ci
) - 1J ( 4. 19a)
where M i s t he number of c e l l s i n t he MD gr oup, and a i s an empi r i c al
c o e f f i c i e n t (a=0.01 i s t he recommended def aul t val ue) . Thi s c or r ec t i on
onl y operat es when vol umet r i c s t r a i n i ncr ement s i n t he smal l est c e l l
of t he group are compr essi ve ( i . e. c e l l vol ume i s bei ng decreased).
I t can be seen eas i l y t hat t h i s c or r ec t i on (eqn. 4.19) conserves
t he volume wei ghed average, t hus mai nt ai ni ng t he consi st ency of t he
comput at i ons. Cal l i ng t he new average {Ae\j):
( ^ ^ ( J ^ e ^ v
0 1
) / v
E
= ^ e ^ + o ; mi n ( z \ e 5 J , 0 ) ( l / V
E
) ^ [ ( V
E
v
C i
) / ( MV
C i
) - V
C i
J
(4. 19b)
where t he second t er m on t he r i g h t hand si de i s n u l l , as V =) V .
On t he o t h e r hand, when t he c e l l vol ume t ends t o zer o t he
v o l u me t r i c s t r a i n i nc r ement gr ows a s y mp t o t i c a l l y , pr oduc i ng an
77
infinite resistance to,tangling over
f
0r
v
C j
0, dJ := ^ +Q! min(Z\e5J,0)LV
E
/(MV
C
J)-lJ -oo (4.20)
For axi symmet r i c pr obl ems, agai n t he c or r ec t i on f or mul a must be
ar ea wei ghed. I n or der t o pr es er v e t he v ol ume- wei ghed aver age ( eqn.
4.19b) t he axi symmet r i c c or r ec t i on t akes t he f or m
A^
:
= Ae + a
m
i n ( 4 e 5
J
, 0 ) [ A
E
/ ( MA
C i
) - l J r
E
/ r
C i
( 4. 21)
where A , A are areas of t he i ndi v i dual c e l l s and compl et e el ement
r , r are t he r adi al coor di nat es of t h e i r c e n t r o i d s .
The new el ement s c r e a t e d by t h i s c o r r e c t i o n a g a i n s t n e g a t i v e
volumes i n MO procedures w i l l be c a l l e d :
MTQC (MTQ cor r ect ed) f or 2- 0;
PlT'tlC (MTB cor r ect ed) f or 3- 0.
A s i mp l e t e s t wi t h a q u a d r a n g l e , i n whi ch one of t he component
t r i a n g l e s t ends t o zero vol ume, was done t o v e r i f y t he behavi our of
t hese el ement s ( f i g . 4. 4) . For Const ant S t r a i n T r i a n g l e s (CST), t he
r e s i s t i n g f or ce grows t owar ds i n f i n i t y when a node t r i e s t o cr oss over
t he o p p o s i t e s i d e . MTQ and MTQC el ement s d i s p l a y a s o f t e r r es pons e,
bot h bei ng near l y i d e n t i c a l up t o 90% r educt i on i n vol ume. For f u r t h e r
r educt i ons, t he r esi st ance of MTQC el ement s shoots up st eepl y t owar ds
i n f i n i t y , whi l e MTQ al l ows cr ossi ng over ( negat i ve vol ume).
Thi s i mpor t ant r es ul t means t hat t he new proposed el ement s are
capabl e of accur at el y model l i ng i ncompr essi bl e p l a s t i c f l o w, even f or
s i t u a t i o n s wi t h ver y l a r g e d i s t o r t i o n s , but r ec ov er t he d e s i r a b l e
r e s i s t a n c e of CST meshes t o o v e r l a p p i n g . Fu r t h e r r e s u l t s u s i n g MTQC
el ement s are present ed i n sect i ons 5.4, 5.7, and 7.3.4.4 whi ch conf i r m
t hi s p o i n t .
78
4.2.4 MASS LUMPING PROCEDURE
The mass i n the di scret e model i s lumped at the gr i d- poi nt s, by
i nt egr at i ng the continuum mass over a volume sur r oundi ng each node.
For a Lagrangian mesh, nodal masses need only be computed once, at the
beginning of the cal cul at i ons. Two condi t i ons must be met by the mass
lumping procedure: f i r s t l y the centre of gr avi t y must be preserved f or
dynamic c a l c u l a t i o n s , and secondl y t he volume of i n t e g r a t i o n , f or
consi stency, must correspond t o the surface used when i nt egr at i ng the
st r esses ( f i g . 4.2).
The c o n t r i b u t i o n of a c e l l C t o t he mass of node 1 i s ( f i g . 4. 2):
- Plane stress or plane st r ai n
P
V
al bg
=
Pt A
a
l bg
=
Pt Ai
mn
/ 3 (4.22)
(1/3 of the mass t o each node)
- Axisymmetric
P
v
al bg
=
P f
2 7 r r d A
=p(7T/54)A
l mn
(22r
1
+ 7r
m
+ 7r
n
) (4.23)
A
al bg
For 3-D, l / 4t h of t he mass of every t et r ahedr on i s lumped at each of
i t s nodes.
4.3 MOMENTUM BALANCE
Having i nt egrat ed the stresses (eqn. 4.9) and obtained the nodal
masses (eqns. 4.22, 4.23), bal ance of momentum i s appl i ed l o c a l l y ,
usi ng t he i nt egr al f orm of t he p r i n c i p l e (eqn. 2.28). Note t hat t he
cor r espondi ng f i e l d equat i on (Cauchy's eqn. of mot i on, (2.29))
requi res a stronger d i f f e r e n t i a b i l i t y , which i s not sat i sf i ed at the
element i nt er f aces.
I nt egr at i ng over a small surface s ( ^ around each node 1,
u
( 1 )
= [ / crndS + M^
1
) f + R^
1
) ] / M
( 1
) (4.24)
J
S
(D
where: R^ ' are intensive forces applied on node 1
IV ' is the mass lumped at 1.
79
4.4 CENTRAL DIFFERENCE TIME INTEGRATION
The accelerations are integrated in time with a centred scheme to
provide new values for velocities and displacements:
n+1/2
=
Q n-1/2
+
j j n ^
( 4#25 )
u
n+l
= u
n
+
^ n +1 / 2^
(4>26)
Note that all the variables representing rates or increments are
defined at mid-step. The central difference equations (4.25, 4.26)
together with the discretized momentum balance eqns. (4.24), form an
uncoupled system, which may be solved independently for each node in
the mesh.
The numerical dispersion (frequency distortion errors) associated
with a lumped mass idealization and a central difference scheme are of
different signs (Key, 1978), which makes this combin ation a natural
choice for explicit models.
4.5 CONSTITUTIVE MODELS
Explicit schemes allow a great generality in the constitutive
laws that can be modelled. In principle, any law that can be expressed
explicitly in the following rate form is acceptable:
8= C(E, d, ff.K.T) (4.27)
where: <j represents an obj ect i ve stress rate (secti on 2. 5. 1)
K i s a set of materi al parameters
T i s the temperature
E i s Green's st r ai n tensor
d i s the rate of deformation t ensor.
For t he present work, t he c o n s t i t u t i v e l aws consi der ed are
el ast i c- pl ast i c models wi t h Von Mises yi el d c r i t er i on. A hypoel asti c
pr edi c t i on i s computed f i r s t , and then cor r ect ed back t o t he y i e l d
surface i f the yi el d condi t i on i s exceeded (radi al r et ur n) . For a more
8Q
compl et e study of c o n s t i t u t i v e l aws i n e x p l i c i t Fi n i t e Di f er ence
Lagrangian codes, see Herrmann and Bert hol f (1983).
4 . 5 . 1 HYPOELASTICITY
The basic ( pr edi ct or ) const i t ut i ve behaviour i s a hypoel asti c law
of the type:
CT^ = Ad
k k
5
i
j + 2Gdij - a(3A+2G)f<$
i
j ( 4. 28)
where A, G are Lame's el ast i c constants
T i s the temperature rate
t r i s the coef f i ci ent of thermal expansion
Hypoel ast i c st r esses are computed al ways as pr edi c t or s , bei ng
l at er corrected f or other types of behaviour i f needed. Al l stress and
def or mat i on measures used i n eqn. (4.28) r ef er t o t he cur r ent
conf i gur at i on; i t produces a l i near r el at i on between t rue stress and
nat ur al s t r a i n i n a uni ax i al t e s t . For l ar ge s t r ai ns (4.28) does not
der i ve f rom an e l a s t i c pot ent i al ( hy per el as t i c ) . For smal l s t r a i n s ,
both concepts are equi val ent .
4. 5. 2 PLASTICITY; RADIAL RETURN ALGORITHM
The pl as t i c y i e l d c r i t e r i o n i s a st r ess c o n d i t i o n , def i ni ng a
y i e l d sur f ace i n st r ess space. The y i e l d c r i t e r i o n depends on t he
hi st or y of def ormat i on, through the pl ast i c hardening parameters, Q:
F((7, Q) = 0 ( 4. 29)
The t r i a l stress from the hypoel asti c pr edi ct i on (eqn. 4. 28) ,
C7[?
+ 1
=G
n
+
<j
+1/ 2
/ U (4.30)
i s t est ed wi t h t he y i e l d c r i t e r i o n , eqn. (4. 29). I f F(<7
+1
)<0 no
cor r ect i on i s necessary:
0*
+ 1
= G
n
E
+1
(4. 31)
81
I f F ( a ^
+ 1
) > 0 t he st r esses are r et ur ned nor mal l y t o t he y i e l d sur f ace
( f i g . 4 . 5 ) :
a
n +l
= a
n +l _
( / 5 / 3 )
^
{ 4 > 3 2 )
Ao
n+l/Z
- a
n+l
- a
n
where /3 i s a scal ar f a c t o r , used f or conveni ence of pr esent at i on. The
new st r esses must s a t i s f y F ( < j
n
) = 0.
Thi s r a d i a l r e t u r n met hod i s due t o Wi l k i n s ( 1964) . Ot her (more
s ophi s t i c at ed) methods f or c a l c u l a t i n g t he st r ess updat e e x i s t , such
as t he t angent s t i f f n e s s r a d i a l r e t u r n ( e. g. Kr i e g and Key, 1976) . A
v a r i a t i o n t o t he r a d i a l r e t u r n met hod, a p p l y i n g t he c o n s i s t e n c y
c o n d i t i o n t o t he end s t a t e r a t h e r t han i n a d i f f e r e n t i a l sense has
al s o been pr oposed by Kr i eg and Key ( 1976) . Re c e n t l y , Or t i z and Si mo
(1985) have pr oposed more gener al " r e t u r n mappi ng a l g o r i t h ms " , f o r
a r b i t r a r y p l a s t i c models and non- const ant t angent e l a s t i c i t y t ensor s.
Res ul t s o b t a i n e d her e ( e. g. s e c t . 5.6) i n d i c a t e t h a t f o r t he s mal l
e x p l i c i t st eps us ed, even i n h i g h l y n o n - l i n e a r c a l c u l a t i o n s , t he
perf ormance of t he si mpl est r adi al r et ur n method i s accept abl e.
The r at e of def or mat i on t ensor i s composed of e l a s t i c and p l a s t i c
par t s :
d = d
e
+ d
p
( 4. 33)
The elastic component may be expressed as
d
E
= l _ s + L l a s + 9KT a J (4.34)
ij 2G ij E kk ij
wh ere s-j j are t he dev i at or i c st r esses ( eqn. 2. 59) , and K= A+2G/3.
I n t he f o l l o w i n g , i t w i l l be assumed t h a t t he y i e l d s u r f a c e i s a Von
Mi ses c y l i n d e r , wi t h no dependence on h y d r o s t a t i c s t r e s s e s . The
p l a s t i c c o r r e c t i o n ( 4. 32) t ak es pl ac e i n t he d e v i a t o r i c hy per pl ane
( ^
k k
= c o n s t a n t ) , as W-
=
35- Hence, t he v o l u me t r i c s t r e s s - s t r ai n
behavi our i s pur el y e l a s t i c :
8?
n
S-c*
2 G\ r AS
E
2GA<
2GA^
s^i-w
c*
Figure 4.5: Radial return plastic stress correction
ff
kk
= <^
k
= 3K(d
kk
- 31a ) (4.35)
The r at e of def or mat i on may be expressed i nver t i ng eqn. (4.28):
d
i i
=
J _ s^j + I ( ^kk +9KTa) 5^- (4.36)
2G E
The p l a s t i c r a t e of d e f o r ma t i o n i s d e r i v e d c o mb i n i n g eqns.
( 4. 32) , ( 4. 33) , ( 4. 34) and ( 4. 36) :
d
P
= d - d
E
= l _(s
E
- s) = MldL (4.37)
2G 2G do
Hence t he r e t u r n a l g o r i t h m ( eqs. 4.30 - 4. 32) i m p l i c i t l y d e f i n e s a
nor mal i t y r ul e f or t he p l a s t i c f l ow ( a s s o c i a t i v i t y ) .
I n p a r t i c u l a r , f o r a Von Mi ses y i e l d c r i t e r i o n wi t h k i n e ma t i c -
i s ot r opi c hardeni ng of t he t ype
F = ( 3 / 2 ) (
S i j
-
i j
) ( s
i j
- a - j j ) - Y
2
( 4. 38)
if the plastic condition is reached (F(ajr ) > 0) , the deviatoric
stresses are simply scaled back along the radius of the Von M i s e s
ci rcle (fi g. 4.5):
s-
1
]
1
- ij = ( ( s ? }
1
^ -
i j
) / ( l +/ 3) (4.39)
and i f the hardening i s purel y i sot r opi c ( a ^ =0 ) :
s? ]
1
= ( s
r
j
1
j
1
)
E
/ ( l + /J ) (4.40)
The pl ane st r ess c ondi t i on ^33=0 requi res speci al treatment i n
the el ast i c- pl ast i c al gor i t hm. I f t he mat er i al behaves e l a s t i c a l l y ,
pl ane st r ess i s enf or ced i mp l i c i t l y by s e t t i n g ( i n t er ms of t he
incremental st r ai ns, eqns. ( 4. 7) , ( 4. 8) ) :
A6
33
=
V
{A6
n
+ A
Z2
) ( 4 . 4 1 )
l-v
If plastic d e f o r m a t i o n s occur within the step, an exact one-step
84
al gor i t hm enf or ci ng t he pl ane st r ess c ondi t i on i s not possi bl e (as t he
p l a s t i c c or r ec t i on i s st r ess- dependent ) . I t may be achi eved appl yi ng
t he f o l l o w i n g c o r r e c t i o n s a f t e r t he r a d i a l r e t u r n ( eqn. 4. 32) i s
per f or med:
'
a
m
: =
"^33
Ae
v
:= A(J
m
/(A+2G/3) (4.42)
Ae^ := Ae^ + (1/3)(A - A6
y
)8^
where: a
m
= ( 1/ 3) a ^ , mean st r ess
Ae = pr edi ct ed vol umet r i c s t r a i n
Thi s c or r ec t i on i mpl i es a s/ery s l i g h t vol umet r i c n o n - a s s o c i a t i v i t y ,
<
+l
= U
+ 1
)
E
/ U
+
^ ) (4-43)
whi ch has no e f f e c t f o r d e v i a t o r i c p l a s t i c i t y c r i t e r i a ( e. g. Von
Mi ses) .
4. 5. 3 HARDENING AND UNIAXIAL STRESS-STRAIN LAWS
Pl ast i c s t r ai ns gi ve r i s e t o mat er i al har deni ng. For a k i nemat i c -
i s o t r o p i c Von Mi ses model ( eq. 4. 38) t he har deni ng i s composed of a
k i n e ma t i c t r a n s l a t i o n ( - j j and an i s o t r o p i c expansi on (Y) of t he
y i e l d sur f ace:
0 9 p
3
( 4. 44;
AY = h
Y
A e
p
where ^ e
P
= - Ae^ Ae^A (effective plastic strain increment)
h
a
and hy are the kinematic and isotropic moduli respectively
The plastic hardening modulus in uniaxial tension is
85
h' = h
a
+ h
Y
(4.45)
I f h
a
= 0, hardeni ng i s pur el y i s o t r o p i c ; i f h
Y
= 0, i t i s pur el y
ki nemati c.
Uni axi al s t r e s s - s t r a i n l aws may be def i ned as the t r ue axi al
stress - nat ural axi al st r ai n curves i n uni axi al tensi on t est s (sect.
6. 2) :
o= f ( ) (4.46)
where t he not at i on a= ^ n , e= e n has been used. Decomposing t he
di f f er ent i al st r ai n increment i nt o el ast i c and pl ast i c components,
de = de
E
+ de
p
= da( l / E + 1/h") (4.47)
the t ot al st r ess- st r ai n modulus may be expressed as
h = ^L = I ( 4 . 4 8 )
de 1/ E+l / h'
Uni axi al s t r e s s - s t r a i n l aws are conveni ent f or speci f yi ng
i sot r opi c mat eri al behaviour. A few examples of laws used wi t hi n t hi s
work are given i n f i gur e 4. 6.
The yi el d condi t i on may depend also on the st r ai n- r at es and the
t emper at ur e, whi ch pr ovi de addi t i onal sources of hardeni ng (or
sof t eni ng, as the case may be):
Y = Y
0
( / ) f
R
( e
P
) f
T
( T) (4.49)
Functions f^ ( st r ai n- r at e dependence) and f j (temperature dependence)
may be defi ned as piecewise l i near laws. A useful anal yt i cal form f or
fR i s (Bodner and Symonds, 1960):
f
R
( e
P
) = 1 + (e
P
/ B)
m
(4.50)
where m and B are mat er i al const ant s. Eqn. (4.50) i s equi val ent t o a
h = 1 / (l/h'+l/E)
o = Ee (o<o)
o = o
0
+he (o>o
0
)
NATURAL STRAIN. .
A) BILINEAR ELASTIC-PLASTIC
NATURAL STRAIN.
B) PIECEWISE LINEAR ELASTI C- PLASTI C
o
0
e
0
n e
o = Ee (o<o)
o/ o = (e/ e) " (o>o
0
)
NATURAL STRAIN.
C) POWER LAW EL AST I C- PL AST I C
Fiqure 4. 6: Uni axi al st r ess- st r ai n laws implemented i n numerical model
'A = min. Area i
' rrnn i
of any element) |
NODE-NOD^
I
I pQSVAmi n
NODE-SIDE
NODE-NODE
K
Finure 4. 7: Node to node and node to side contacts - regions and normals
87
v i s c o p l a s t i c i d e a l i z a t i o n such as suggested by Perzyna (1963) or Owen
and Hi nt on (1980):
e
p
= 7<( F/ F
0
- i ) N> df_ ( 4. 51)
do
where: F
0
and F are the static and dynamic yield conditions resp.;
<A> = A for A>0,
<A> = 0 for A<0.
Equivalence between eqns. (4.50) and (4.51) is achieved by assigning
the values
m = 1/N
i
B = 7 1(2/3) &-:^
4.5.4 OBJECTIVE STRESS RATES
Rat e- t ype c o n s t i t u t i v e equat i ons (4.27) need t o be f or mul at ed i n
t erms of obj ec t i v e st r ess r at es ( sect i on 2.5.1). Jaumann and Tr uesdel l
obj ec t i v e r at es have been i mpl ement ed i nc r ement al l y i n t he f o l l o wi n g
way:
Jaumann r at e of Cauchy st r ess
SO + 1/2 = a
n + 1
/
2
+ a
n
+V + l/2
+ ff
n + l
w
n + l / 2
( 4 5
2 )
1J 1J i p
W
PJ JP
W
pi 1<+.D,|
Tr uesdel l r at e of Cauchy st r ess
Sn+1/ 2
=
^n+1/ 2 _ n + l n + 1 / 2 n+l . . n+l / 2 , n + l n + 1 / 2
i J i J i p
V
J, P
CT
JP
v
i , P U
V
P,P
= gn+1/ 2 _ n +l
d
n +l / 2 _ n +l
d
n +l / 2 _ n +l
d
n +l / 2
( 4 5 3 )
1J " i p
U
PJ
U
JP
U
pi
CT
1J
U
PP \*.03)
Other vectors or tensors (e.g. a , eqn. 4.44) linked to the
material frame must also be corrected for rigid body rotations. For a
vector T, a corotational rate equivalent to eqn. (4.52) is defined by
^n+1/2
= i n +
l / 2 _ n
+
l/2
T
n
( 4 > 5 4 )
88
Not e t h a t t h e r e i s h a l f a st ep l ag f o r t he v al ues of s t r e s s e s i n eqns.
( 4. 52) - ( 4. 54) . For e x p l i c i t i n t e g r a t i o n i n whi c h st eps ar e very
smal l er r or s i nt r oduced by t h i s w i l l not be i mpor t ant ( sect . 5.6). For
l ar ger i mp l i c i t st eps, more r i gor ous al gor i t hms have been proposed by
Hughes and Wi nget ( 1980) , Pi ns k y , Or t i z and P i s t e r ( 1983) , and
Rubi nst ei n and At l u r i ( 1983) .
Jaumann' s r a t e i s t he most commonl y used f o r mu l a t i o n , bei ng a
nat ur al choi ce f or removi ng r i g i d body r ot at i ons ( Pr ager , 1961). Some
i nconsi st enci es may ar i se however i n l ar ge def or mat i on si mpl e shear
wi t h k i n e ma t i c har deni ng ( Nagt egaal and de Jong ( 1982) , Da f a l i a s
(1983), Lee and Ma l l e t t (1983)). On t he ot her hand Tr uesdel l ' s r at e i s
t he f o r wa r d Pi o l a t r a n s f o r ma t i o n of t he r a t e of 2nd P i o l a - k i r c h o f f
st r ess and pr ovi des a " canoni cal " f or m f or hy per el as t i c r at e equat i ons
( Pi nsky, Or t i z and Pi s t er , 1983).
Bot h Tr u e s d e l l and Jaumann r at es can be made e q u i v a l e n t by
a p p r o p r i a t e d e f i n i t i o n of t he c o n s t i t u t i v e t e n s o r ( eqn. 3. 28) . The
c l a s s i c a l el a s t i c - p i a s t i c r e l a t i o n s ( Pr andt l - Reuss eqns. (2.64)) are
expr essed i n t er ms of t r u e s t r e s s and t r u e s t r a i n ( H i l l , 1950) . The
as s umpt i on i s made her e, as by H i b b i t , Mar cal and Ri ce ( 1970) , t h a t
t he Pr andt l - Reuss equat i ons are r ef er r ed t o Jaumann r at es of st r esses.
4. 6 DAMPING
The semi di scr et e equat i ons of mot i on (4.24) may be assembl ed i n
mat r i x f orm f or t he compl et e model , and gener al i zed t o i ncl ude vi scous
dampi ng:
Mii + Cii + P = R ( 4. 55)
where M i s t he di agonal mass mat r i x , C i s t he damping mat r i x , and P, R
r epr esent i nt er nal and ext er nal f or ces r es pec t i v el y .
A u s e f u l f or m f o r t he dampi ng ma t r i x i s t he s o - c a l l e d Ra y l e i g h
89
dampi ng:
C = AM + BK
t
( 4. 56)
where: Kt i s
t n e
t angent s t i f f n e s s mat r i x , !C
t
=dP/du
A, B are t he mass and s t i f f n e s s damping c o e f f i c i e n t s .
Ray l ei gh dampi ng a l l o ws eqns. ( 4. 55) t o r emai n unc oupl ed, and
t her ef or e i s eas i l y i mpl ement ed i n an e x p l i c i t code. Mass damping i s
i nt egr at ed i nt o t he t i me- mar chi ng scheme (eqns. (4. 25), (4.26)) by t he
f ol l owi ng oper at or :
u
n + 1 / 2
= [ u
n _ 1 / 2
( l - A / i t / 2 ) + M
_ 1
( R
n
- P
n
) 4t J / ( l +AAt / 2) ( 4. 57)
whi ch i mp l i e s a mo d i f i c a t i o n of t he t i me - i n t e g r a t i o n eqn. ( 4. 25) .
St i f f nes s damping i s consi der ed by addi ng vi scous f or ce t erms t o t he
i nt er nal f or c es :
p*n =
P
n + B6
n
dP
n
/du ( 4. 58)
approxi mat ed i n f i n i t e di f f er enc e f orm by
P
* n
= p
n
+ B A P
n + l / 2
M t
St i f f nes s damping i s consi der ed by s u b s t i t u t i n g P
n
i n eqn. (4.57) by
p*n.
The amount of dampi ng p r o v i d e d by ( 4. 56) i s a f u n c t i o n of t he
f r e q u e n c y . I f t he syst em of eqns. ( 4. 55) i s put i n modal f or m
( assumi ng t angent l i n e a r behav i our and s mal l d i s t u r b a n c e s ) , one
equat i on of t he f o l l o wi n g t ype i s obt ai ned per mode:
x + (A+B a
2
)x + <w
2
x = r ( 4. 58)
where x i s t he modal ampl i t ude and a> t he f r equency of v i b r a t i o n . The
amount of damping f or each f r equency co i s gi ven by
/5 = (A/w + Bft> )/ 2 ( 4. 59)
90
where ft =1 corresponds t o t he c r i t i c a l dampi ng. From eqn. (4.59) i t
may be seen t hat t he ef f ec t of mass damping i s smal l f or hi gh
frequenci es, where st i f f ness damping i s i mport ant , and vi ceversa. Mass
damping may be l i kened t o dashpots connecting each degree of freedom
to ground, and st i f f ness damping t o dashpots between connected nodes.
4. 7 STABILITY OF TIME INTEGRATION
Four i er methods (e.g. Ri cht meyer and Mor t on, 1967) may be used
f or studyi ng the s t abi l i t y of l i near systems and small per t ur bat i on
t angent behavi our i n nonl i near syst ems. The di f f er ent i al modal eqns.
(4.58) may be r ewr i t t en as
x + Zjicox + co
Z
x = 0 ( 4. 60)
f rom whi ch t he ext er nal f o r c i n g f unc t i on ( r ) has been dropped. The
harmonic sol ut i on t o t hi s equation i s
x
n
= e
>>nAt
= A
n
( 4 > 6 1 )
where V, A are compl ex numbers (not e t he super scr i pt on A i s an
exponent ). Subs t i t ut i ng eqn. (4.61) wi t h i n cent r al di f f er enc e eqns.
(4.25 - 4.26) one obtai ns
x
n
= (x
n+1
-2x
n
+x
n
'
1
) Mt
2
= A
n
-
l
(A
Z
-ZA
+
l)/At
2
n
=
(
X
n
+
l _
x
n- l
) / 2 4 t
= A
n
-
l
(A
2
-l)/ZAt
!4.62
which wi t h eqn. (4.60) y i el d
A
2
(l+Atw/5) + A( ^
2
At
2
- 2) + (1-AtwA) = 0 (4.63;
f or s t a b i l i t y , the sol ut i on A t o t hi s equation must be complex and of
modulus not greater than 1. These condi t i ons impose
fi >, 0
At < IJl-d
2
( 4. 64,
O)
91
The second condition must be satisfied for all the frequencies
present in the model. As damping is usually small, a sufficient
condition will be
At < 2.V
1_
/3
2
(4.65)
CO
max
Damping has an adverse ef f ec t on s t a b i l i t y , al t hough s t i f f n e s s
and mass damping behave di f f er ent l y i n t hi s respect. Considering t hat
the condi t i on (4.55) appl i es f or the hi ghest frequency of the model,
from eqn. (4.59) i t may be seen t hat mass damping wi l l v i r t ual l y have
no ef f ect , whi l e st i f f ness damping may be very det r i ment al .
For undamped motion ( /3 =0) the c r i t i c a l t i me-st ep reduces t o
At < 2/o>
max
(4.65)
An upper bound expression f or the c r i t i c a l t i me- st ep i s given by the
Courant cr i t er i on (Courant, Fri edri chs and Lewy, 1928), which l i mi t s
the t i me-st ep t o the i nt er val necessary f or the stress-waves t o t r avel
across one element,
At
<
n
mi n/ c (4.66)
where hmin ^
s t n e m
'
i i r
'
, T1um
element di mensi on, and c the maximum st r ess-
wave v e l o c i t y . For an e l a s t i c - p l a s t i c mat er i al t he f as t es t st r ess
waves are the compressional or "P" waves, f or which c = -J{A+2Gi)/p .
4.8 MODELLING OF CONTACTS
Contact between di f f er ent boundaries of one or more conti nua i s
frequent i n impact model l i ng and imposes speci al nonl i near boundary
c ondi t i ons . Two main aspect s are i nvol ved i n cont act model l i ng: t he
contact i nt er f ace laws and the det ect i on of i nt er act i ng surfaces.
92
4.8.1 CONTACT INTERFACE LAWS
A penal t y method i s used f or t he cont act l aws, c ons i s t i ng of
pl aci ng i nt er f ac e spr i ngs between " i n t r u d i n g " nodes and " t ar get "
surfaces. Forces generated by these spri ngs wi l l act on both sides of
the i nt er f ace. For complete symmetry of the l ogi c, both sides of the
i nt er f ac e may be t aken i n t ur n as " i n t r u d i n g " si des. The i nt er f ac e
laws are
Normal f or ce: F? = K
N
d
N
C
(4.68)
- Shear f or ce: F = min(K
s
df Fg/* )
where K , K^are the normal and shear st i f f ness values
d , d are the normal and shear penetrati ons
fj. i s the coef f i ci ent of f r i c t i o n
The shear law defines a ki nemati c sl i p c i r c l e of radius FQ ^ /K ^
Note t hat the penetrati on d i s a vector l i nked t o the Lagrangian frame
and t her ef or e must be updated wi t h a c or ot at i onal f or mul at i on (eqn.
4. 54).
An i mportant det ai l of the f or mul at i on i s the value used f or the
contact st i f f ness. Marti (1983) uses a constant user-def i ned value f or
the whole model, which should be of si mi l ar magnitude t o the element
s t i f f n e s s . Thi s method i s si mpl e and appr opr i at e f or uni f or m 3-D
meshes, but probl ems w i l l ar i se f or non- uni f or m or axi symmet r i c
model s. Hal l qui s t (1982) computes a d i f f e r e n t s t i f f n e s s f or each
cont act , proport i onal t o t hat of the t ar get element.
In t h i s work, t he cont act s t i f f n e s s var i es f or each c ont ac t ,
using a simple but ef f ect i ve method. The st i f f ness i s cal cul at ed f or
each contact as
K = M
l
/Atl
( 4 # 6 9 )
where M^ i s the mass of the i nt r udi ng node
At
c
i s the c r i t i c a l t i me-st ep f or the model (eqn. 4.66)
k i s a coef f i ci ent normal l y def aul t ed t o 0.1 (Larger values may
93
cause i ns t abi l i t i es i f a corresponding reducti on i s not used f or
the computati onal t i me- st ep)
Thi s f orm f or t he cont act s t i f f n e s s ensures a val ue below t he
maximum el ement s t i f f n e s s i n t he model (t hus not compr omi si ng
s t a b i l i t y ) and above t he s t i f f n e s s of t he el ement i n cont act (thus
avoi di ng unacceptable penet rat i ons):
- maximum element st i f f ness - K
max
= ( l / 4) V
c r
( A+2G) / h2
i n
- contact st i f f ness - KQ = 2kV
cr
(A+2G)/h
m
-j
n
- i nt er f ace ( t ar get ) element st i f f ness - K
T
= ( 1/ 4) V
T
( A +2Gvh^
i n
where Vcr
i s t n e v o
^
u me
f
t n e
element gi vi ng the c r i t i c a l t i me- st ep,
and VT t hat of the i nt er f ace t ar get element.
This der i vat i on of the contact st i f f ness al so produces a uni f orm
"push-back" recovery act i on al ong t he cont act i n t e r f a c e , whi ch i s
i mpor t ant s pec i al l y f or axi symmet r i c model s, where t he el ement
st i f f nesses and masses are r adi us dependent. Thi s may be shown by
consi deri ng the accel erat i ons at the i nt r udi ng si de (a
1
) and t ar get
side (aT) f or a given penet rat i on d:
a
I = FQ/ M
1
= Kd/At^
( 4. 70)
a
T
= F
C
/M
T
= (KdMt^)(M
I
/M
T
)
Assuming the mesh i s uni f orm on both sides of the i nt er f ace ( M* / MT=1) ,
f or symmetric (double) contacts the recovery of the penet rat i on i n one
t i me-st ep wi l l be
d
R
= 2 ( l / 2 ) ( a
I
+ a
T
) At
2
= Kv
2
d ( 4. 71)
where At=rjAt
c
i s the computational t i me-st ep ( * / <l ) . The mesh can be
non-uni form along the contact i nt er f ace, but should general l y not be
too di f f er ent accross i t ( M^ M
1
) . The frequency of the contact spri ng
connecting M^ and M^ wi l l be proport i onal t o
(K/At) M
J
/M
94
Large r at i os M^/ MT wi l l , requi re reducti ons i n the computational t i me-
st ep.
For hi gh v e l o c i t y i mpact i n whi ch l ar ge cont act pr essur es are
produced, the contact st i f f ness may have t o be scaled up by changing
t he def aul t val ue of k i n eqn. (4. 69), i n or der t o avoi d unaccept abi y
l arge penet rat i ons. This would again requi re a corresponding reducti on
of the computational t i me- st ep.
Appl i c at i on of eqns. (4.68) r equi r es d e f i n i t i o n of a normal t o
the cont act . For t hi s purpose, contacts are cl assi f i ed as node-node or
node-si de ( f i g . 4.7). For node- si de cont act s, t he normal i s d i r e c t l y
t hat of t he t ar get si de. For node-node c ont ac t s , t he normal i s
computed by averaging those from the two adj oi ni ng si des.
Fi nal l y l i e s t he quest i on of d i s t r i b u t i n g t he cont act f or ce on
t ar get and i nt r udi ng sur f aces. On t he i n t r u d i n g sur f ace t he sol e
r ec i pi ent i s al ways t he i n t r u d i n g node, but f or t he t ar get si de
force di st r i but i on depends on whether contact i s of node-node or node-
si de t ype. For node-node cont act s t he i n t e r f a c e f or ce i s assi gned
e n t i r e l y t o t he t ar get node. For node- si de cont act s t he i n t e r f a c e
f or ce i s d i s t r i b u t e d t o t he t wo nodes i n t he t ar get s i de, i n i nver se
proport i on t o t hei r di stance t o the i nt r udi ng node.
A f l owchar t of t he comput at i ons f or det er mi ni ng t he nor mal s,
penetrati on and contact forces i s given i n f i gur e 4.8.
4. 8. 2 CONTACT DETECTION ALGORITHM
Det ect i on of new cont act s and book-keepi ng f or t he l i s t s of
act i ve contacts must be performed i n an ef f i c i ent and r el i abl e way t o
provi de robust models f or impact and sl i di ng. In the present 2-D model
a contact i s establ i shed and kept i f the i nt r udi ng node ei t her
a) penetrates the t arget side (non-zero i nt er act i on f or ce)
b) proj ect s on the t ar get side wi t hout penet r at i ng, l yi ng i n ei t her of
the node-side or node-node areas ( f i g 4. 7) (zero i nt er act i on f or ce)
95
Intruding Node: P
Target Node: A
Normal & Tangential
vectors: n, t
(average adjoining sides)
Intruding Node: P
Target Nodes: A,B
Normal & Tangential
vectors: n, t
Old vectors: t
0
, n
Q
C = common Cel1
avge. vectors:
iv
=n
C> t av^C
C = new Cell
avge. vectors:
n
av
=n

t
a v
= t
C = ol d Cell
avge. vect ors:
' " C^ a
C = new Cell
avge. vect ors:
n+ n
C = new cel l
avge., ol d vect ors:
t a v - V *
Normal penetrati on
Ad| = (-d2n<?
<U.
"n "j
, C . ,.C
+
d
tj
..A
iu
C
Compute w\i, v
1
, <
o
t
o,,,C
fur - (v-+v- ( ~
L
v
1
vv, v
1 S J
^ . ->j
d + ( Ad\ + Adj ' j t ?
At
av
J At
di st r i but e forces
r
A
= AP-AB/(AB-AB)
F
A
= -F r
A
F
B
= - F( l - r
A
)
Apply force t o A:
F
A
= -F
J RETURN
FIGURE 4.8: SUBROUTINE CONFOR
(FLOWCHART)
CALCULATION OF CONTACT PENETRATION AND INTERFACE FORCES
96
A compl et el y general c a p a b i l i t y f or aut omat i c det ect i on of new
c o n t a c t s ( i . e . any node of any e x t e r n a l s u r f a c e wi t h any ot her
sur f ace) woul d gener al l y car r y an excessi ve comput i onal overhead. Two
i nt er medi at e procedures have been i mpl ement ed here f or t he det ect i on
of cont act s:
- New i n t e r a c t i o n s near an e x i s t i n g c o n t a c t w i l l be d e t e c t e d
a u t o ma t i c a l l y , by c h e c k i n g i n ever y st ep t he nei ghbour s t o a l l
e x i s t i n g c o n t a c t s . El ement adj acency l i s t s ar e kept i n t he pr ogr am
memory f or t h i s purpose. Al l cont act s wi t h i n one i nt er f ac e area w i l l
be pi c k ed up by g i v i n g j u s t one seed c o n t a c t . A c o n t a c t sear ch
al gor i t hm has been i ncl uded whi ch f i nds t he c or r ec t t ar get si de f r om
an i nc or r ec t guess.
- For Aut omat i c det ect i on of new cont act s, areas t hat may t ouch at an
unknown i ns t ant may be earmarked by s pec i f y i ng l i s t s of nodes, whi ch
w i l l be checked agai nst each ot her f or cont act at each t i me- s t ep. In
f a c t , t hese ar eas may encompass t he whol e e x t e r n a l boundar y of t he
bodi es i nvol ved, i f t he l oc at i on of cont act s i s t o t a l l y unpr edi ct abl e
bef or ehand, al t hough t h i s w i l l nor mal l y car r y a l ar ge overhead i n t he
comput at i onal cost s.
Sl i d i n g of e x i s t i n g cont act areas i s achi eved by checki ng, p r i o r
t o r e l i n q u i s h i n g a non- act i ve cont act , i f t he i n t r u d i n g node has s l i d
t o any of t he nei ghbour i ng t ar get si des.
Fl ow- char t s f or t he cont act c al c ul at i ons wi t h t he search and t he
book-keepi ng l ogi c are gi ven i n f i gur es 4. 9 and 4. 10.
4. 9 HEAT CONDUCTION
Heat f l ow r at es accross t he cont i nuum are computed f r om Four i er ' s
1 aw:
h
n + 1 / 2
= KT
n
.j ( 4. 72)
wher e h^ i s t he r a t e of heat f l o w i n d i r e c t i o n i per u n i t nor mal
sur f ace, K t he c o n d u c t i v i t y , and 1" t he t emper at ur e. An e x p l i c i t Eul er
97
START
existing contacts /CONZON)
- still valid? (search)'
- contact forces (CONFOR)
check neighbour Nodes
(NEIGHB)
check
neighbours
(NEIGHB)
check
new contacts from
lists (CONDEC)
check
neighbours
(NEIGHB)
RETURN
Figure 4.9: Subroutine CONTAC, driver for Contact calculations
(Flowchart)
START
98
Node P
intruder
pick side AB
on target Cell
Compute penetration
and interface forces
(CONFOR)
Figure 4.10: CONZON subroutine - check and search for contacts
(Flowchart)
99
forward i nt egr at i on scheme i s i mpl i ed i n eqn. (4.72).
For a smal l r egi on t he r at e of heat i nf l ow i nt o i t may be f ound
by i nt egr at i ng over i t s surface S,
H
n+l / 2
=
. T K T ^ n-jdS (4.73)
S
which may be par t i cul ar i zed f or a t r i angul ar or t et r ahedr al cel l as a
sum over the cel l f aces:
m
H = (1/ m)) K
j
S
j
( T
j
- T) / ( d
j
- d) (4.74)
j = l
where m = no. of sides (3 t r i angl e, 4 t et rahedron)
K
J
, S
J
are the conduct i vi t y and surface of side j
d
J
, T
J
are the height of and temperature at the cel l adj oi ni ng j
d, T are the hei ght and temperature at the current c e l l .
For boundary faces, the cont r i but i on t o the sum i n eqn. (4.74) i s
cal cul at ed according t o the appropri ate boundary condi t i on:
- Di r i chl et T
e
= Const., H
j
= mK(T
e
-T)S
j
/d (4.75)
- Neumann h
e
= Const . , H
j
= h
e
S
j
(4.76)
- Mixed T
e
/C
T
+ h
e
/C
n
= 1, H
j
= mKS
J
' (C
T
-T)/(d+KmC
T
/C
h
) (4.77)
Other sources of heat such as p l a s t i c work must be consi der ed
when f or mul at i ng t he energy bal ance (eqn. 2.32), t o compute t he
temperature increments i n each c e l l :
AT = [(HAt +(j:Ae
P
V
C
)//>V
C
+ AQJ/C
p
(4.78)
where: ^ def i nes t he amount of p l a s t i c work t r ansf or med t o heat ,
(def aul t ed t o 0.95)
\ r i s the cel l volume
pi s the mass densi ty
A0, i s heat generated from other sources per uni t mass
C
p
i s the speci f i c heat
f or an Eul er f or war d t i me i nt egr at i on scheme (eqn. 4. 72),
comput at i ons are c o n d i t i o n a l l y st abl e. For a r ect angul ar mesh, t he
c r i t i c a l t i me-st ep f or s t abi l i t y i s (Richtmyer and Morton, 1967):
At
c
= ( 1/ 2) ( 1/ Ax
2
+ l My
2
) / ( K/ pC
p
) ( 4. 79a)
For t r i angul ar and t et r ahedr al meshes, empi ri cal r esul t s obtai ned i n
t hi s work suggest the f ol l owi ng c r i t i c a l t i me- st eps.
- Tr i angl es: At
c
= ( l / 8 ) h
2
i n
/ (K/pC
p
) (4.79b)
- Tetrahedra: 4t
c
= ( l / 12) h
i n
/ ( W
p
) (4.79c)
4.10 ENERGY COMPUTATIONS
E x p l i c i t schemes such as c e n t r a l d i f f e r e n c e ar e onl y
c o n d i t i o n a l l y st abl e. In a l i near syst em, an i n s t a b i l i t y i s r eadi l y
detected, as i t wi l l cause overfl ow i n t he computations wi t hi n a few
cycl es. Unf ort unat el y, t hi s i s not the case general l y f or non-l i near
mat er i al s . I n s t a b i l i t i e s are associ at ed wi t h spur i ous r el ease of
energy. Pl as t i c mat er i al s may absorb t he spur i ous energy t hr ough
p l a s t i c def or mat i on, whereby i n s t a b i l i t i e s coul d go unnot i ced even
though l ar ge er r or s may be pr esent . An energy bal ance check (as
recommended by Belytschko (1978, 1983)) has been implemented i n order
t o detect " ar r est ed" i n s t a b i l i t i e s .
At every i nst ant , Energy balance di ct at es
Wlnt
+ u
K i n "
w
Ext = (
4
-
8 0
)
For checki ng eqn. (4. 80), W
I n t
( i nt er nal ener gy) , U|<in ( k i net i c
ener gy) and Wf r
x
t ( ex t er nal ener gy) are f o l l o we d d u r i n g t he
computations. The al gori t hms f or these energy computations are given
below. Note t hat some energy components are computed i ncr ement al l y,
W
n+1
= W
n
+AW
n + 1 / 2
whi l e f or others the t ot al value W" i s recomputed at each st ep.
101
I nt er nal Energy
I nt er nal energy i s composed of cont i nuum energy ( I ) , damping
energy ( I I ) , and energy i n the contacts ( I I I ) :
Wmt
= W
I
+ W
I I
+ W
I I I (
4
-
8 1
)
- Continuum energy i s computed i ncrement al l y as
NUMNOD
AW?
+1/2
= (4t/2)/(a
n
+0
n+1
):d
n+1/ 2
dV = (l/2)^^uS
+1/ 2
(PJ i+P|5
+1
) (4.82)
N=l
Where P are t he nodal f or c es , obt ai ned by i n t e g r a t i n g t he cont i nuum
st r esses (eqn. 4.9). For e l a s t i c - p l a s t i c mat er i al s t he cont i nuum
energy may be decomposed i nt o el ast i c and pl ast i c components:
Wj = W^-j + Wp-j
AWgf
1
/
2
= ( A t / 2 ) / ( < j
n
+ ( j
n + 1
) : ( d
P
)
n + 1 / 2
d V
( 4 . 8 3 )
AU l\
1
'
2
= Ut/2)J (a
n
+0-
n+1
):(d
E
)
n+1
/
2
dV
For an i s ot r opi c e l a s t i c Von Mises model t he expr essi on f or t he
pl ast i c work may be si mpl i f i ed t o
AWpf
1/2
= (At/2) / ( a[!
q
+ agj
1
) e
P
dV (4.84)
where a
eq
= y( 3/ 2) s: s (Von Mises equi val ent st r ess) .
- Damping energy increments are computed as
NUMNOD
A^
112
= ( 1/ 2) ) Aujj
+1/2
CAM
N
ujj
+1/2
+ (PR
+1
-PR)B/2Zitj (4.85)
N=l
- Energy at cont act s consi st s of pot ent i al energy st or ed pl us
di ssi pat ed f ri ct i on al en ergy:
NUMCON
W
III = d / 2 ) XZ CKj^(d^)
2
+ K ^ )
2
j + WP
M
(4.86)
J=l
^
n
F r i
= FJj ( d
0
- d
n
s
)
where d^
0
- d r epr esent s t he s l i p ( pr edi ct ed minus cor r ect ed shear
di spl acement ) . Subscr i pt N here i ndi cat es Normal ( penet r at i on or
st i f f ness)
NUMNOD
(1/ 8)]P] (ufi
+1/2
+ ^ ~
1 / Z
) \ (
4
-
87
)
N=l
NUMNOD
( 1 / 2 ) ) AU
N
(FJ) + F[j
+1
) (4.88)
N=l
For com putation s carried out on a CRAY-IS with around 10$ tim e-steps
and 500 cel l s , the errors in m atching energy balan ce (eqn. 4.80) have
n orm ally been lower than 0.005%W
E xt
. The same com putation s on a
VAX11/785 have had errors lower than 0.15%WE
X
f
4.11 IMPLEMENTATION INTO FORTRAN PROGRAM
The algorithm s described in thi s chapter have been im plem en ted
i n t o a Fortran 77 com puter program (G oi col ea, 1985a). The program
con tain s some addi t i on al f a c i l i t i e s n ot described here, such as rod
elem en ts (e.g. f or m odellin g bol t s , rei n f orced or pres tres s ed
con crete). Some ef f or t was taken to make the program us er - f r i en dl y,
al l owi n g a f l exi bl e s t ruct ure f or the i n put , f r ee form at command
i n terpretati on , some mesh gen eration f aci l i t i es , and sen sible defaults
for n on -specified param eters. User-frien dlin ess was im portan t for two
reasons:
Ki net i c energy
u
Ki n
External work
AW
Ext
- The program has been used as a pr oduct i on code f or non- l i near and
i mpact engi neer i ng anal yses (e.g. Pr i nc i pi a Mechanica Lt d. ( 1983) ,
Dost al , Phelan and Trboj evi c (1985))
- Ext ensi ve appl i c at i on of t he program has been done wi t h i n t h i s
t hesi s (Chapters 5, 6, 7) .
A r est ar t f a c i l i t y i s avai l abl e, which permi ts the computations
t o be recommenced from predetermined "saved" i nst ant s. I t al so al l ows
recovery from a system crash or forced hal t .
Output f i l e s wi t h h i s t o r i e s of moni t or ed v ar i abl es , deformed
geomet r i es, and s t r e s s / s t r a i n components f or cont our p l o t t i n g are
produced. These f i l es can be i nput t o post-processors such as PR2DPL
( Goi col ea, 1985b), CURVA ( Goi col ea, 1985c) and PRISM ( Pr i nc i pi a
Mechanica Lt d. , 1984). PRISM can al so act as a pr e- pr ocessor
generating meshes.
Program development was done i n i t i a l l y on a CDC7600 and l at er on
a CRAY-IS of t he Uni v er s i t y of London. Mi nor par t s of t he code have
been v e c t o r i z e d f o r t he CRAY, but a more t hor ough e f f o r t i n
vect or i zat i on was not consi dered necessary f or t he pr esent research
work. As i t i s , t he perf ormance i s about 15000 c el l s Xc y c l e per CPU
second. On a CDC7600 the program wi l l run approxi matel y 50% sl ower, on
a PRIME750 40 ti mes sl ower, and on a VAX11/785 20 ti mes sl ower.
De f i n i t i o n of t he pr obl em, t i me - c y c l e c o mp u t a t i o n s , and
r edef i ni t i on may be speci f i ed i n a f l ex i bl e way wi t hi n cer t ai n rul es
(e.g. mat er i al pr oper t i es and mesh must be def i ned bef or e t he t i me -
cycl e computati ons). A t ypi cal problem sequence i s i ndi cat ed i n f i g .
4.11. The f l ow- char t f or the t i me- cycl e computations i s given i n f i g .
4. 12.
START
104
- Geometry (Mesh)
- Material Properties
Define Problem - Boundary Conditions
- Contact properties
- Thermal properties
Define Monitoring Requests
Calculate time-step At, Masses, External Faces, ...

Redefine
Problem
Time-cycle computations
1
Save Restart Files
f
Output: Printer Plots, Output Files,...
STOP
Figure 4.11: Typical sequence for explicit numerical modelling job
io;:
Recalculate A t
1
Save monitored Variables
I nt e gr a t e St r e s s e s and add t o nodal Forces P=/ c r
n
n d s
4
X
Find new Ve l oc i t i e s , Displacements,- updat e Coords. V
n41
, l/
1
"
1
', X
n+I
Compute St r ai n I ncr ement s, A
rn-l/2
New St r e s s e s , cr
n + i
^ JContacts, P
r
, f
nn
Compute Forces in<
n
, . C
D
n-fi V
a
dd to nodal forces
JRod elementSjHp 1
I
t
n
:= t"
+
A t
Figure 4.12: Flowchart for explicit computational cycle
CHAPTER 5
BENCHMARKS AND VALIDATION EXAMPLES
5.1 INTRODUCTION
5.2 WAVE PROPAGATION
5.2.1 Elastic waves in bars
5.2.2 Elastic-Plastic waves
5.2.3 Elastic waves in cone
5.3 VIBRATION OF A CANTILEVER
5.4 STATIC ELASTIC-PLASTIC PROBLEMS
5.4.1 Punch test
5.4.2 El ast i c- Pl ast i c sphere under i nt er nal pressure
5.5 HEAT CONDUCTION
5.5.1 Coupled thermomechanical analysis
5.5.2 Temperature redistribution in a slab
5.6 LARGE STRAINS AND ROTATIONS
5.7 IMPACT OF CYLINDER 5.8 CONCLUSIONS
107
5.1 INTRODUCTION
The p u r p o s e of t h i s c h a p t e r i s t o p r e s e n t a number of
r e p r e s e n t a t i v e benchmar k t e s t s , v a l i d a t i n g t he F i n i t e Di f f e r e n c e
t echni ques used i n t h i s work. Most of t he t es t s have been chosen wi t h
a v a i l a b l e a n a l y t i c a l s o l u t i o n s t o compar e w i t h , but f o r some t he
compari son can onl y be done wi t h ot her numer i cal r esul t s r epor t ed i n
t he l i t e r a t u r e .
The f e a t u r e s t e s t e d concer n p r i ma r i l y t he El a s t i c - P I a s t i c
mechani cal behavi our of cont i nua. This i ncl udes mai nl y dynamic t es t s
( bot h wa v e - p r o p a g a t i o n and i n e r t i a - d o mi n a t e d t ypes) , al t hough some
q u a s i - s t a t i c pr obl ems ar e al s o s ol v ed u s i n g dynami c r e l a x a t i o n . A
s i mp l e heat c o n d u c t i o n t e s t pr ov i des a check f o r t he t her mal
c a p a b i l i t i e s .
No t est s are present ed f or t her mo- mechani cal coupl ed pr obl ems, as
f u r t h e r a p p l i c a t i o n wor k on t he t o p i c was not done f o r t h i s t h e s i s .
Thi s l eaves some scope f or f u r t h e r work on t he mat t er .
5. 2 WAVE PROPAGATION
A few examples of wave pr opagat i on probl ems are gi ven her e. The
f i r s t t hr ee concern t he pr opagat i on of waves al ong a bar i n response
t o t he sudden a p p l i c a t i o n of a body f o r c e . An exampl e of t he
t r ansmi ssi on of waves i n an El a s t i c - Pl a s t i c mat er i al wi t h har deni ng i s
gi ven next . Fi n a l l y , t he behavi our of st r ess waves i n a coni cal bar i s
examined wi t h an i l l u s t r a t i v e exampl e.
5.2.1 ELASTIC WAVES IN BARS
Thr ee pr obl ems ar e pr es ent ed i n t h i s s e c t i o n c o n c e r n i n g t he
p r o p a g a t i o n of e l a s t i c waves i n bar s . The bar di mens i ons are i n a l l
cases l mxl mx20m, and t he common mesh u t i l i z e d i n t he anal yses consi st s
of 20 q u a d r i l a t e r a l s of l mxl m ( f i g . 5.1) (each q u a d r i l a t e r a l i s
composed of 2 t r i a n g l e s ) . The mat er i al pr oper t i es are as f o l l o ws :
108
P = 2000 Kg/ni
E = 13.2xl 0
8
Pa>
v = 0 . 3 1
lm x lm x 20m
gravity
force
X . 77777777^777
Figure 5.1: Mesh used for models of wave propagation in bars
CO
\
>-
LJ
O
_J
LU
2.50
2.00
1.50
1.00
-0.50
-1.00
-1. 50
-2. 00
- 2. 50
i r ~i r~
ll
1
1
II
1
II
If
1
ll
1
1
1
il MID-
i \
A
\\
\
\
\\
\
\
\
\\
\\
\\
\\
-POINT \
THEORY
CALCULATED
7
K
/ / \ \
/ / \\
7 ^\
// \\
/ / U
/ / \\
// \\
// \\
1
1
1
1
1
ll
\ 7
// \ //
FNH OF BAR \ //
\\ //
j i i i_ J L
0 20 40 60 80 100 120 1A0 160 180 200
T i me (msec)
Figure 5.2: CONSTRAINED LONGITUDINAL ELASTIC WAVES IN BAR
Density P = 2000 Kg/m
El ast i c modulus E = 1320 MPa
Poisson' s r at i o v = 0.31
The bars are i n i t i a l l y at r est , st r ess- f r ee and f i xed at one end.
A body f or ce ( gr av i t y ) of u n i t val ue i s t hen appl i ed suddenl y. Thi s
cr eat es a wave pat t er n as t he pul se rebounds and t r a v e l s back and
f or t h along the bar. The peri od of t hi s wave i s
T = 4L/c (5. 1)
where L i s t he l engt h of t he bar and c t he v e l o c i t y of pr opagat i on.
The maximum poi nt vel oci t y at t ai ned should be
v = gL/c (5. 2)
where g i s the gr avi t y force appl i ed.
Constrained P-waves
I f the nodes are constrai ned t o move only par al l el t o the axi s of
the bar and i n plane st r ai n, the vel oci t y of propagati on i s t hat of a
plane compressional wave i n an i n f i n i t e medium:
c
p
=V(A+2G)//3 (5. 3)
= 966.2 m/s
Theor et i cal and numer i cal r es ul t s showing the vel oci t y hi st or i es at
the f ree t i p and mi d-poi nt of the bar are presented i n f i g . 5. 2.
Unconstrained P-waves
In t h i s case t he bar i s al l owed t o expand l a t e r a l l y , bei ng
model l ed i n pl ane st r ess. The wave v e l o c i t y i s t hat of a one-
dimensional bar,
c
0 =VE/ P (5.4)
= 811.5 m/s
no
The resulting velocity histories are given in fig. 5.3
Shear waves
The body f or ce w i l l now be normal t o t he axi s of t he bar. To
prevent bendi ng f rom devel opi ng, t he nodes are const r ai ned i n t he
l ongi t udi nal di r ect i on; t hi s way, the di st or t i on i s resi st ed only by
shear st resses. The vel oci t y of wave propagation i s :
Cs =-JGJP~ (5. 5)
= 500 m/s
Results f or t hi s case are given i n f i gur e 5. 4.
5.2.2 ELASTIC-PLASTIC WAVES
Let an unconstrained rod of El ast i c- Pl ast i c mat er i al , wi t h l i near
st r ai n- har deni ng, i mpi nge nor mal l y upon a r i g i d f l a t a n v i l , gi v i ng
r i se t o a st r ess hi gher t han t he y i e l d val ue Y. An El as t i c and a
Pl ast i c wave are or i gi nat ed and propagate si mul taneousl y through the
mat er i al , the t heor et i cal sol ut i on t o t hi s problem has been discussed
by Johnson (1972).
The bar i s again of dimensions lmxlmx20m, and the same mesh from
f i g . 5.1 i s used. The remaining parameters are:
El ast i c modulus E = 6400 Pa
Yi el d st ress Y = 40 Pa
Mass densi ty P = 1 Kg/m3
El ast i c- Pl ast i c modulus H = 640 Pa
Impact vel oci t y v = 1 m/s
The el ast i c wave wi l l t r avel at a speed of
c
0 = -y]l/P = 80 m/s
whi l e the Pl ast i c wave lags behind wi t h a vel oci t y of
.50
2.00 -
1.50
1.00
0.50
0.00
0.50
1.00
1.50 (-
2.00
2.50
THEORY / w
CALCULATED /
\ /MID-POINT
\ / END OF BAR
i / i i i i
NV
i
v
. i i
\
i
0 20 40 60 80 100 120 H O 160 180 ~0(
T i me (msec)
Figure 5.3: UNCONSTRAINED LONGITUDINAL ELASTIC WAVES IN BAR
0.04
0.03
0.02
0.01
0.00
0.01
0.02
0.03
_l ! ! ! j , ! 1_
THEORY ,
CALCULATED
0.04
D-POINT
END OF BAR
0 20 40 60 80 100 120 140 160 180 200
T i me (msec)
Figure 5.4
S
ELASTIC SHEAR WAVES IN COLUMN
112
c
l =VH/P = 25.3 m/s
The El ast i c wave i s r ef l ect ed from the f ree end as an unl oadi ng wave.
As i t returns along the bar i t wi l l meet the oncoming pl ast i c f r ont at
a c er t ai n i ns t ant t i
a t a
poi nt def i ned by i t s di st ance f rom t he
impacting end, x-^
:
t]_ = 2L/(c
0
+c
1
) = 0.38s
X! = UC-L/ COJ / U+C-L/ CQ) = 9.61m
From t hi s poi nt el ast i c waves are r ef l ect ed back i nt o each part of the
bar ; t hese waves cont i nue t o t r av el and rebound succesi vel y. The
pl ast i c st r ai ns extend only t o the region x<x^.
In the cal cul at i ons, appr eci abl e p l a s t i c s t r ai ns occur r ed onl y
f or the f i r s t 10m of t he bar , as pr edi ct ed. A t i me hi s t or y of st r ess
at a di st ance x = 4.67m i s shown i n f i g . 5.5. Thi s poi nt sees t he
passage of the el ast i c f r ont , and then the pl ast i c f r ont . Af t er t hat ,
el ast i c rebounds create f ur t her f r ont s.
5.2.3 ELASTIC WAVES IN CONE
Experi ment s on c oni c al l y - t aper ed Hopki nson bars show t hat
compressive stress-waves create a tensi on t a i l , which makes the bar
move backwards as t he t i me- pi ece i s t hr own of f . Landon and Quinney
(1923) f i r s t analyzed t hi s phenomenon, assumi ng an exponent i al f orm
f or the appl i ed pressure pul se, t ypi cal of an expl osi ve charge. Kolsky
(1953) has presented the t heor et i cal anal ysi s i n a revi sed form. The
pulse i s given by a displacement law of the form
u = ( A/ r ) exp[ ( - r - co
t
) /
B
J " (
A
/
r
) t
5
-
6
)
where r i s t he di st ance f rom t he apex of t he cone and t t he t i me. t =0
corresponds t o t he a r r i v a l of t he head of t he pul se at t he apex, t
being negative whi l st the pulse i s t r av el l i ng towards the apex. Eqn.
(5.6) hol ds f or negat i ve val ues of t , where r > / c o t j . For r < l c o
t
l
the displacement i s zero. B i s the char act er i st i c l ength of the pul se,
113
30
20
10
-10
-20
-30
-40
-50
-60
0.00
STRESS HISTORY AT 4.67m FROM IMPACTING END
- - - THEORY
CALCULATED
IMPACT
0.20 0.80 0.40 0.60
T i me (sec)
Fi gur e 5.5s ELASTIC-PLASTIC WAVES IN BAR
1.00
114
whi l e A def i nes t he ampl i t ude, co
i s t h e b a r w a v e
v e l o c i t y (eqn.
( 5. 4) ) .
The stress associated wi t h t hi s pulse i s :
a= - ( AE/ Br ) exp[ ( - r - co
t
) /
B
- l
+
( AE/ r 2) j ;
1
_
e X
p[ ( _
r
_
c 0
t )/B J (5. 7)
The ri ght -hand si de of eqn. (5.7) consi st s of two opposing si gn t erms,
the f i r s t always negati ve (compressive f r ont ) and the second posi t i ve
(tension t a i l ) . The f ol l owi ng may be noted:
- t he peak compressi ve st r ess i ncr eases as t he wave approaches t he
apex (proport i onal t o -AE/Br);
- the l ength of the compressive por t i on decreases correspondi ngl y;
- a t ensi on t a i l i s l e f t , whi ch s u f f i c i e n t l y behi nd t he head of t he
pulse approaches AE/r2 i n magnitude.
A coni cal bar of end r adi i 6mm and 60mm, and l ength 2000mm, was
modelled numeri cal l y wi t h a graded mesh of 4x150 quadr i l at er al s ( f i g.
5.6). Mat er i al pr oper t i es were t y p i c a l of Al umi ni um (E = 67Gpa,
P=2700Kg/m3
s C
Q=4981m/s). The pul se i nput was gi ven as a v e l o c i t y
hi s t or y at t he wi de end, der i ved f r om eqn. (5. 6). Ampl i t ude of t he
pulse was A=1.3382xl0-4m2, and the l ength B=0.300m.
The numer i cal r es ul t s compare wel l wi t h t he a n a l y t i c a l st r ess
p r o f i l e s ( f i g . 5.6). Pr edi c t abl y , t he ext r emel y sharp peak at t he
wave f r ont i s bl unt ed somewhat, and some di sper si on i s i nt r oduced
behind i t . Part of t hi s i s at t r i but abl e t o the surface waves produced
i n the axisymmetric model.
5.3 VIBRATION OF A CANTILEVER
The mot i on of a v i b r a t i n g c a n t i l e v e r i s a probl em t hat may be
r e a d i l y checked wi t h e l a s t i c beam t h e o r y . A mesh of 4x21
quadr i l at er al s i s used t o model a ver t i cal cant i l ever clamped at the
t op end ( f i g . 5.7a). I t i s set i n mot i on by gi v i ng t he bot t om r i g h t
-20
-30
-40
-50
-60
-70
L I
t=.0.4
msec
ERICAL CALC
ANALYTICAL SOL
0.40 0.60 0.80 1.00 1.20 1.4
Distance from apex of c
F i g u r e 5 . 6
S
P R O P A G A T I O N OF S T R E S S
116
.
lOmm.
.
' 1 1
y s / / / / / / / S / S / / / S / /
e
e
IT)
PEAK LEFT DEFLECTION NDEFORMED ME C! -EAK RIGHT D'
F i g u r e 5.7-. E L A S T I C V I B R A T I O N OF C A NT I L E V E R ( D I S P L A C E M E NT S XI 0)
117
hand node a v e l o c i t y , of l Om/ s d u r i n g 0.05ms. I t i s t hen l e f t t o
v i br at e dur i ng 1ms. Mat er i al pr oper t i es are
Densi t y p = 7720 Kg/m3
El as t i c modulus E = 213.4 GPa
Poi sson ' s r a t i o '
;
= 0. 311
Fi gur es 5.7b and 5.7c show t he maxi mum p o s i t i v e and n e g a t i v e
di spl acement s, exagerat ed 10 t i mes. Di spl acement h i s t o r i e s of poi nt s
at t he t i p and mi d d l e of t he beam ( f i g . 5.8a) show t he domi nance of
t he f l e x u r a l f undament al mode of v i b r a t i o n i n t he r esponse. Thi s i s
al so not abl e i n t he f r equency spect rum at t he t i p , shown i n f i g . 5.8b,
where t he f i r s t t hr ee ( t h e o r e t i c a l ) nat ur al f r equenci es of a f l e x i o n -
onl y beam are shown as wel l f or compar i son. For t he f i r s t f r equency,
t he c o i n c i d e n c e wi t h t he t h e o r e t i c a l v al ue i s r e ma r k a b l e . Some
di s c r epanc y e x i s t s f o r t he second and t h i r d modes. Thi s i s ex pec t ed
f r om t he f a c t t h a t t he t h e o r e t i c a l f r e q u e n c i e s do not t ak e i n t o
account shear or f i n i t e d e f o r ma t i o n e f f e c t s . Thi s pr obl em has been
anal yzed p r e v i o u s l y by Wi1 k i ns ( 1969) w i t h s u b s t a n t i a l l y t he same
r es ul t s as here.
5.4 STATIC ELASTIC-PLASTIC PROBLEMS
Two c l a s s i c a l pr obl ems of s t a t i c p l a s t i c i t y are s ol v ed i n t h i s
sect i on. As di scussed i n sect i on 4. 1, t he e x p l i c i t F i n i t e Di f f er ence
code l acks t he a b i l i t y t o per f or m s t a t i c anal ysi s d i r e c t l y . The s t a t i c
s ol ut i ons are reached t hr ough dynamic r el ax at i on. The v i br at i ons are
damped out w i t h t he use of v i s c ous dampi ng, and t he f o r c e s ( or
d i s p l a c e me n t s ) ar e a p p l i e d g r a d u a l l y . Some " o v e r s h o o t " i n t he
s o l u t i o n s i s u n a v o i d a b l e f o r p l a s t i c ma t e r i a l s , a l t h o u g h i t can be
mi ni mi zed by proper use of damping and sl ow l oadi ng r at es .
5. 4. 1 PUNCH TEST
The case of a s e mi - i n f i n i t e body i ndent ed by a f r i c t i o n l e s s , f l a t
r i g i d punch, under condi t i ons of pl ane s t r a i n , was sol ved o r i g i n a l l y
by Pr andt l ( 1920) , u s i n g shear l i n e t h e o r y . The p l a s t i c c o l l a p s e
118
1.00
0.80
0.60
0.40
0.20
0.00
-0.20
-0.40
-0.60
-0.80
1.00
DISPLACEMENT AT TIP
DISPLACEMENT AT MID-POINT
0.00 0.80
0.20 0.40 0.60
T i m e (msec)
Figure 5.8a= ELASTIC VIBRATION OF CANTILEVER - DISPLACEMENT HISTORIES
00
0.50
0.45
0.40
0.35
0.30
0.25
0.20
0. 15
0.10
0.05
0.00
first
frequency
NATURAL FREQUENCIES FOR FLEXURAL BEAM (ANALYTICAL VALUES)
sedond
frequency
A,
third i
frequency
i i
10 20 60 70 30 40 50
f r e q u e n c y (KHz)
Fi gur e 5. 8bi ELASTIC VIBRATION OF CANTILEVER - FREQUENCY CONTENT
80
119
pressure i s const ant and equal t o
p = l(Z+n )/ V^J Y = 2.97Y ( 5. 8)
Numer i cal l y, t he hal f - space i s d i s c r e t i z e d r a t h e r c r u d e l y ( f i g .
5.9), model l i ng onl y a f i n i t e r ect angul ar p o r t i o n , wi t h di spl acement s
at t he boundary const r ai ned i n both d i r e c t i o n s . Pr escr i bed v e l o c i t i e s
are gi ven t o t he nodes under t he di e. The mat er i al pr oper t i es are
El as t i c modulus E = 10U Pa
Poi sson ' s r a t i o
v
= 0.3
Yi el d st r ess Y = 0. 01 Pa
For t h i s p r o b l e m, a mesh of c o n s t a n t s t r a i n t r i a n g l e s does not
gi ve a s a t i s f a c t o r y r e s u l t . Not onl y does i t over est i mat e t he col l apse
pr essur e, but t he l oad grows monot oni cal l y, f a i l i n g t o achi eve a l i mi t
v al ue ( f i g . 5. 9) , i n d i r e c t c o n t r a v e n t i o n of t he l i m i t t heor ems of
p l a s t i c i t y . The reason f or t h i s i s t he i nadequacy of t hese meshes t o
model i ncompr essi bl e p l a s t i c f l o w, as di scussed i n sect i on 3. 7. 1.
The Mi xed D i s c r e t i z a t i o n pr ocedur es advocat ed her e ( s e c t i o n s
4. 2. 2, 4. 2. 3) do p r o v i d e a s a t i s f a c t o r y s o l u t i o n ( f i g . 5. 9). The
s l i g h t o v e r e s t i mat i on i s due t o t he coar seness of t he mesh and t he
t o t a l c ons t r ai nt at t he boundar i es. Thi s probl em was sol ved usi ng t he
CMTl) q u a d r i l a t e r a l s proposed i n sect i on 4.2.3, t hus pr ovi ng t hat t he
modi f i c at i ons i nt r oduced i n t hese t o pr event negat i ve volumes do not
s t i f f e n t hese el ement s up f o r o r d i n a r y s i t u a t i o n s wi t h s mal l or
moderate s t r a i n s .
To achi eve t he s o l u t i o n , a r e l a x a t i o n pr ocess was used. Mass
densi t y ( f i c t i t i o u s ) was t aken as 0.1 Kg/m3, and f u l l Rayl ei gh damping
was used, wi t h 1% of c r i t i c a l damping at 1 Hz. The r el ax at i on process
was f ol l owed f o r 10 seconds (458 st eps) . The pr escr i bed v e l o c i t i e s at
t he di e were appl i ed gr adual l y dur i ng t he f i r s t second.
120
3.50
a. oo
2.50
2.00 -
1.50 -
1.00
0.50
.00
1
.
/f
_/ /
/ /
/r
/
r-l '
1 '
1 /
' ' '
r /
t '
f '
/ /
/ /
L /
1*' U T T
/ W i 1
/ - - WIT
i
zr ^z>
H M.D.
HqUT M
y
1
y
LI MI T LOAD (PRANDTL, 19;
^ ^ ^
/ :
v / / / / / / / / / / / / / / / / / / / / / / -
I n 1
D.
(CSJ ELEMENTS),
>0)
V.
/
"'/
/
/
-/
/
/
/
/
J -
- /
"
8 0 2 4 6
T i m e (see)
Figure 5.9, PUNCH TEST IN PERFECTLY PLASTIC. PLANE STRAIN HA L F -S P A C E
10
121
5.4.2 ELASTIC-PLASTIC SPHERE UNDER INTERNAL PRESSURE
When a t h i c k - wa l l e d spher i cal vessel i s subj ect ed t o gr owi ng
i nt er nal pressure, a pl ast i c zone st ar t s t o develop from the i nt er i or .
There i s a range of pr essur es f or whi ch cont ai ned p l a s t i c f l ow i s
achi eved, def i ni ng a p l a s t i c zone and an e l a s t i c out er zone. The
t heor et i c al s ol ut i on (e.g. _Hi l l , 1950) def i nes t he r adi us c of t he
pl ast i c zone from \A ^ f
p = 2YLn(c/a) + ( 2Y/ 3) ( l - c3/ b3) (5. 9)
where p i s t he pr essur e, a t he i nt er nal r adi us, and b t he ext er nal
r adi us.
The stresses i n the el ast i c zone (c<r<b) are given by
,5 / , 1
a
= -(b3/r3-l)2Yc3/3b
3
r - L^f * ( A. -i)
a
e
= (b3/2r3+l )2Yc3/3b
3
? V
c
i kl
+
A)
and i n the pl ast i c zone (a<r<c) by
" r = -2YLn(c/ r) - 2Y( l - c
3
/ b
3
) / 3
a
e = Y + ^
r
( 5. i o;
(5.11)
Numer i cal l y t h i s probl em was sol ved d i s c r e t i z i n g a 30 deg.
axi symmet r i c sect or , wi t h a mesh of 10x8 CMTQ q u a d r i l a t e r a l s ( f i g .
5.10a). The materi al parameters were
El ast i c modulus E = 1
Poisson' s r at i o v = 0.2
. Yi el d st rengt h Y = 0.83588xl 0"
3
Density p = 1
Wi th a=l , b=2 and an appl i ed pr essur e of p=10-3, t he p l a s t i c zone
extends t o c=1.5 exact l y (eqn. 5.9). The cal cul at ed pr of i l es of radi al
and hoop st r esses ( f i g . 5.10b) show an ex c el l ent agreement wi t h
t heor y. Only a smal l di f f er enc e i s pr esent i n t he peak of t he
E=l
w=0.2
Y=.000836
A) MESH USED AND MODEL PARAMETERS
CO
r
CD
L
CD
>-
\
CO
0)
CD
c_
e n
0. 60 -
0. 40 -
0.20
0.00
-0. 20
-0. 40
-0. 60
-0. 80
,40 1.60
R a d i a l c o o r d i n a t e
B) S T R E S S D I S T R I B U T I O N A C R O S S THE R A D I U
2.0
FIGURE 5.10: ELASTIC-PLASTIC SPHERE UNDER INTERNAL PRESSURE
123
ci r cumf er ent i al stresses at the pl ast i c boundary, due t o the crudeness
of the mesh and the stepwise represent at i on of stresses i n the model
(const ant s t r a i n el ement s) . As a check, c al c ul at ed c i r c u mf e r e n t i a l
stresses i n the sphere are provided i n both i n-pl ane and out - of - pl ane
di r ect i ons, showing excel l ent agreement.
To achi eve t he s o l u t i o n , a dynamic r el ax at i on process was
f ol l owed f or 40s, wi t h mass damping 50% of c r i t i c a l at 0.08 Hz. (782
st eps) . The l oad was i ncr eased gr adual l y t o i t s f i n a l val ue i n t he
f i r s t 30s.
5.5 HEAT CONDUCTION
The numer i cal procedures descr i bed i n chapt er
c a p a b i l i t y of per f or mi ng pur el y mechani cal , pur el y
simultaneous coupled thermo-mechanical cal cul at i ons.
5. 5. 1 COUPLED THERMOMECHANICAL ANALYSIS
The compl exi ty of the mathematical descr i pt i on of general coupled
t hermomechani cal syst ems, i n whi ch t he heat i s generat ed by pl aj t i c
f l ow, accounts f or the f act t hat anal yt i cal sol ut i ons are not r eadi l y
av ai l abl e i n t h i s f i e l d . As on t he ot her hand t her momechani cal
appl i cat i ons were not intended f or t hi s work, no val i dat i on example i s
presented here as such.
However, the mathematical model f or thermomechanical coupl i ng has
been der i ved and i mpl ement ed by t he aut hor , bot h i n t he 2-D program
devel oped f or t h i s t h e s i s , and i n an e x i s t i n g 3-D program ( Ma r t i ,
1981, 1983). A 3-D appl i c at i on of t h i s work has been publ i shed
el sewhere. (Mart i , Goicolea, Kal si and Macey, 1984), which concerns the
ext r usi on of an al umi ni um di sk wi t h an i r r e g u l a r di e. Very severe
pl ast i c di st or t i ons are achieved i n the process, producing increases
i n temperature.
4 have t he
t her mal , or
5.5.2 TEMPERATURE REDISTRIBUTION IN A SLAB
A t her mal - onl y anal ysi s i s per f or med f or an i n f i n i t e sl ab,
i n i t i a l l y at a uni form temperature T=l , and cooled down by keeping the
ends x=-L and x=L at T=0. The problem parameters are
half thickness
Conductivity
Density
specific heat
L = 20
K = 1
P = 1
C p - 1
Cal cul at ed t emper at ur e p r o f i l e s f or vari ous ti mes are compared
wi t h t heor et i cal values i n f i g . 5.11, showing excel l ent agreement. A
t i me- st ep of 56.25,ms) was used f or t he anal y s i s , wi t h a mesh of 40
t r i a n g l e s . The t h e o r e t i c a l s ol ut i on was obt ai ned by Four i er ser i es
anal ysi s ( e. g. Carslaw and Jaeger (1947), p83).
5.6 LARGE STRAINS AND ROTATIONS
A Quadr i l at er al subj ect ed si mul t aneousl y t o a f i n i t e r i g i d
r ot at i on and a si mpl e a x i a l e x t e n s i o n ( i . e . l a t e r a l movement
const r ai ned) i s consi der ed her e. The purpose of t h i s example i s
pr i nci pal l y t o check the obj ect i vi t y of the i nt egr at i on al gor i t hm.
The mat er i al behavi our i s descr i bed by a hypoel ast i c law (eqn.
4. 28) , i nvol vi ng the Jaumann rate of Cauchy st r ess:
0= C:d ( 5. 12)
Cal l i ng 0( t ) t he r i g i d r o t a t i o n at t i me t , and A( t ) t he axi al
st ret ch r at i o ( L( t ) / L) i n the rot at ed coordi nate frame, the two normal
st r ess components may be obt ai ned a n a l y t i c a l l y by i n t e g r a t i n g eqn.
( 5. 12) :
2
9 s i n
2
9 \ I I
)
E ( 1 - I > ) L I A / [ ( 1 + P ) ( 1 - 2 I ; ) ] (5.13;
2
9 cos
2
9i
v
M
1.00
0.80
0.60
0.40
0.20
0.00
-ee
ANALYTICAL SOLUTION (FOURIER SERIES)
> NUMERICAL CALCULATIONS
0.00 0.20 0.40 0.60
X/L
a=0.01
0.60 1.00
Figure 5.11, HEAT CONDUCTION IN SLAB - TRANSIENT TEMPERATURE PROFILES
CO
D
D
O

O
CO
(U
CO
CO
CD
L
CO
!.00
0.80
0.60
0.40
0.20
0.00
ANALYTICAL SOLUTION
o o, NUMERICAL CALCULATION
o
a
, NUMERICAL CALCULATION
0.00 0.20 0.40 0.60 0.80
T i m e (t)
Figure 5.12s HOMOGENEOUS FINITE SIMPLE EXTENSION AND ROTATION
1.00
A f u l l 360 deg. r ot at i on wi t h a f i nal st r et ch r at i o of 2 was performed
( f i g . 5.12), t he numer i cal r es ul t s agr eei ng c l os el y wi t h t heor y. A
l ar ge number of t i me- st eps - 1824 - was used f or t he c a l c u l a t i o n ,
whi ch coul d have been reduced wi t hout a f f e c t i n g s i g n i f i c a n t l y t he
accuracy of t he r e s u l t s . However, t h i s i s a r e a l i s t i c s i t u a t i o n f or
ex pl i c i t cal cul at i ons, where t i me-st eps must be very small f or reasons
of s t abi l i t y . The si mpl e Jaumann rate al gor i t hm (eqn. 4.52) provides
excel l ent r e s u l t s , wi t hout needi ng t he more r i gor ous procedures
necessary f or l arger i mpl i c i t steps (Hughes and Winget (1980), Pinsky,
Or t i z and Pi st er (1983)).
5.7 IMPACT OF CYLINDER
The i mpact of a s o l i d c y l i nder agai nst a r i g i d st onewal l i s a
probl em i n whi ch t he f i n a l shape i s s ens i t i v e t o t he p l a s t i c
propert i es of the mat er i al . Wi l ki ns and Guinan (1973) used t est s and
numerical analyses of such impacts t o cal i br at e the pl ast i c mat eri al
behavi our . Here, t he i mpact of a copper c y l i nder i s descr i bed. Thi s
probl em has been anal yzed by Hal l qui s t wi t h NIKE2D (1979), DYNA2D
(1982a) and DYNA3D (1982c), and by Johnson (1981).
Impact v e l o c i t y i s 227m/s. A si mpl e El a s t i c - Pl a s t i c Von Mi ses
i deal i zat i on wi t h l i near i sot r opi c hardening i s used f or the mat er i al .
The dimensions and mat eri al parameters are given i n f i gur e 5.13a.
The mesh compri sed 50x5 CMTQ q u a d r i l a t e r a l s and was sever el y
d i s t o r t e d as a r es ul t of t he i mpact ( f i g s . 5.13b, 5.13c). The f i r s t
part of the deformati on (40 microsec.) ocassions a f l at t eni ng of the
i mpact end i nt o an el ephant ' s f oot . The l a t t e r par t ( t i l l 80
microsec.) produces a bar r el i ng upwards of the bar, as the bottom part
has become very hardened by then. The cyl i nder st ar t s t o rebound af t er
79 mi cr osec. ( v el oc i t y h i s t o r i e s i n f i g . 5.14), i n agreement wi t h
Hal l qui s t ' s r e s u l t s . Note t hat f or t h i s probl em t he nodes at t he
i mpacti ng end were f i xed i n the axi al di r ect i on. Although a stonewal l
boundary c ondi t i on ( per mi t t i ng separ at i on) woul d have been more
r e a l i s t i c , t he out er edges t endi ng t o l i f t s l i g h t l y , t h i s si mpl er
i deal i zat i on was adopted f or compat i bi l i t y wi t h Hal l qui st ' s r esul t s.
127
Ro
' "-P
**
A
4
Lo
R =3.2mm
L =32.4mm
o
impact velocity
227 m/s
I
P = 8930 Kg/rrT
E = 117 GPa
v = 0.35
Y = 400 MPa
H = 100 MPa
L
A) t =0usec x = AOpsec
r\
rtiffl
Zn R """
F i g u r e 5 . 1 3 : IMPACT OF CYLINDER - ONDEFORMED AND DEFORMED VIEWS
X
4->
l
O
C

>
-100
-120
- U0
-160
20
o
20
40
60
80
00
i i
-
MI D- P OI NT
-
-
i i i
ih)/
i - r
/
TOP POINT (B)
70 0 10 20 30 40 50 60
Ti me (ps)
Figure 5.14s IMPACT OF CYLINDER - VELOCITY HISTORIES
Resul t s f rom var i ous c al c ul at i ons are pr esent ed i n t abl e 5. 1,
gi vi ng t he deformed l engt h, bot t om r adi us, and maximum e f f e c t i v e
pl as t i c s t r a i n (achi eved al ways at t he bot t om cent r e1i ne eel 1). The
r esul t s obt ai ned here usi ng CMTQ or MTQ el ement s compare wel l wi t h
Hal l qui s t ' s r e s u l t s . Mesh d i s t o r t i o n , al t hough l ar ge, i s qui t e
uni f or m, and MTQ el ement s behave wel l here. As p r e d i c t a b l e , t he
response of CST el ement s i s poor. Al so pr esent ed i n t abl e 5.1 are
r esul t s from Johnson (1981); two di f f er ent types of crossed t r i angl e
l ayout s are seen t o pr ovi de bet t er r es ul t s t han CST el ement s, but
s t i l l not as good as the Mixed Di scr et i zat i on used here ( speci al l y as
t o maximum pl ast i c st r ai n) .
L
R
^max
P
Hallquist(1982a,1982c)
DYNA2D
21.47
7.127
3.05
DYNA3D
21.47
7.034
2.96
NIKE2D
21.47
7.068
2.97
This work
CMTQ
21.45
7.068
2.90
MTQ
21.44
7.155
2.94
CST
21.13
6.061
1.63
Johnson(1981)
XI
21.61
7.136
2.38
X2
21.61
7.040
2.70
CST
21.12
6.016
1.34
Table 5. 1: Comparison of r esul t s from vari ous cal cul at i ons f or impact
of c y l i nder (see f i g . 5.13)
5.8 CONCLUSIONS
The c a p a c i t y of t he pr oposed e x p l i c i t F i n i t e Di f f er ence
al gori t hms f or model l i ng El ast i c- Pl ast i c mat eri al behaviour i n plane
and axisymmetric models has been val i dat ed, wi t h a seri es of st at i c
and dynamic benchmark t est s. Special at t ent i on has been given t o the
wave propagation and f i n i t e deformati on capabi l i t i es.
129
CHAPTER 6
TENSION TESTS ON ALUMINIUM BARS
6.1 INTRODUCTION
6.1.1 Constitutive idealization
6.1.2 Tension tests - a review
6.2 THEORETICAL INTERPRETATION OF TENSION TESTS
6.2.1 Strain distribution at minimum neck section
6.2.2 Stress distribution
6.3 TENSION TESTS
6.3.1 Specimens and material
6.3.2 Programme
6.3.3 Procedure
6.3.4 Results
6.3.5 Microhardness measurements
6.4 MATERIAL HARDENING LAW
6.5 NUMERICAL CALCULATIONS FOR TENSION TESTS
6.5.1 Model
6.5.2 Analysis
6.5.3 Results
6.6 CONCLUSIONS
6.1 INTRODUCTION
Experimental and numerical studi es of tensi on t est s wi t h necking
i n bar s ar e r e p o r t e d i n t h i s c h a p t e r . The ai m i s t o d e r i v e
c o n s t i t u t i v e l aws f or t he al umi ni um mat er i al used i n t he t ube
crumpl i ng analyses (chapter 7). St r ess- st r ai n laws val i d f or up t o 1.5
st r ai n are necessary f or these analyses.
6 tensi on t est s were car r i ed out on HE30 annealed aluminium bars.
In t hese t e s t s , a f t e r an i n i t i a l st age of uni f or m ext ensi on, an
i ns t abi l i t y occurs, causing deformati ons t o become l ocal i zed i n a neck
wi t h very l arge st r ai ns. Reductions i n cross-sect i onal area of A/A
Q
_
0.26 ( f i g . 6. 1), cor r espondi ng t o a uni ax i al s t r a i n of 1.35, were
achieved before f r act ur e.
The st r ess and s t r a i n f i e l d s around t he neck are markedl y non-
uni f or m. Theor et i cal anal yses of t hese d i s t r i b u t i o n s wi t h some
empi ri cal basis have been proposed by Bridgman (1952) and Davidenkov
and Spi ri di nova (1946). These semi -empi ri cal di st r i but i ons were used
here t o provide a f i r s t i nt er pr et at i on of the t est r esul t s, obt ai ni ng
t ent at i ve hardening laws f or the mat er i al .
The tensi on t est s were then modelled wi t h the ex pl i c i t numerical
t echni ques proposed i n t h i s t hes i s ( chapt er 4). Cal cul at i ons were
perf ormed successf ul l y up t o t he l ar ge neck s t r ai ns observed i n
ex per i ment , t he s t r e s s / s t r a i n d i s t r i b u t i o n s o b t a i n e d i n t he
c al c ul at i ons conf i r med i n general t he v a l i d i t y of Bri dgman' s and
Davi denkov' s s ol ut i ons . Load and average st r ess ev ol ut i on curves
f i t t e d c l os el y t he exper i ment al r e s u l t s . Wi t h t he har deni ng l aws
obt ai ned usi ng Bri dgman' s semi - empi r i cal i n t e r p r e t a t i o n as a f i r s t
guess, the computations al l owed a f ur t her adjustment of the parameters
t o provide a bet t er f i t wi t h experiment.
The f ol l owi ng termi nol ogy i s employed i n t hi s chapt er:
- True (average) s t r es s , P/A, axi al l oad di vi ded by cur r ent c r os s -
secti on of bar;
- Natural (or l ogar i t hmi c) st r ai n, Ln(L/ L
o) >
l ogari t hm of current over
i n i t i a l l engt h;
131
Initial area of cross-sect
A = T T D V 4
o o
Deformed cross-section
area at neck
A - n\f/H
Figure fi.l: schematic view of neck region in tensile bar
50mm
h
1.5mm
13.2mm 375 mm
-v +- -
15 mm/
37,5mm 13.2mm 5 0 mm
- * - t " ' I
18.5mm
' 12.9mm
i / 16.2mm I 5. 9mm| i a9mm
I
Diameter taper
Figure 6.2: geometry of tension test specimen
- Effective plastic strain, e
p
(
e
qn. 2.60b)
- Von Mises equivalent (or effective) stress,
a
e q
= y ( 3/ 2) s :s (6.1)
where s is the deviatoric Cauchy stress tensor, eqn. (2.59).
6.1.1 CONSTITUTIVE IDEALIZATION
El ast i c- Pl ast i c theory has been very successful i n descri bi ng the
deformati on of metals and provides a good framework on which t o bui l d
models of mat er i al behavi our . A Von Mises y i e l d c r i t e r i o n wi t h
i sot r opi c hardening and associ at i ve pl ast i c f l ow provides a si mpl e and
robust model f or non- cycl i c l oadi ng, gi v i ng accept abl e answers f or
most cases. With t hi s choice of const i t ut i ve model, r el i abl e data f or
the el ast i c propert i es may be obtained from the suppl i er , handbooks,
or si mpl e t e s t s . The mat er i al paramet ers w i l l be compl et e wi t h t he
hardeni ng law 1(e
p
)
t
whi ch r el at es t he y i e l d st r engt h of t he
mat er i al , Y, t o a scal ar measure of the pl ast i c s t r ai n, the ef f ect i ve
pl as t i c st r ai n e
p#
The f l ow st r ess Y i s def i ned by t he val ue at y i e l d of t he Von
Mises equi val ent st r ess:
Y =
eq (6. 2)
In some speci al cases, a si mpl e i sot r opi c Von Mises model may not
r epr esent t he mat er i al s a t i s f a c t o r i l y . Speci al i zed cal cul at i ons f or
wel l def i ned l ocal st r ess st at es may r equi r e c ar ef ul consi der at i on
and/ or mat chi ng of t he l oadi ng c ondi t i on f or t he t es t i n whi ch t he
c o n s t i t u t i v e data were der i ved (e. g. pure t ens i on, pure shear ,
combined l oadi ngs).
I f i n such a problem pl as t i c i t y i s confi ned t o a par t i cul ar l ocal
st r ess s t a t e , a det er mi nat i on of t he y i e l d and f l ow paramet ers f or
t hat st at e wi t hi n an i sot r opi c pl as t i c i t y i deal i zat i on may suf f i ce and
give sat i sf act or y r esul t s. In a general si t uat i on more sophi st i cat ed
p l a s t i c i t y i d e a l i z a t i o n s (e.g. Mroz (1967, 1972), Or t i z and Popov
(1983)) wi l l be needed, determi ni ng the const i t ut i ve data by numerous
t est s f or each st r ess o r i e n t a t i o n , at each l evel of har deni ng. Thi s
process requi res extensi ve and d i f f i c u l t experi mental work, as wel l as
research t o assess the appropriateness of const i t ut i ve i deal i zat i ons.
Moreover, f or l ar ge s t r a i n anal y s i s , t he e f f o r t i nvol ved woul d be
great l y increased and compl i cat ed, wi t h uncert ai n r esul t s.
In the Concertina tube col l apse mechanism, l arge pl ast i c st r ai ns
are produced under a var i et y of stress st at es. For a global model l i ng
of t hi s phenomenon an i sot r opi c Von Mises i deal i zat i on i s a si mpl e and
r e l i a b l e i d e a l i z a t i o n , whi ch pr oduces s a t i s f a c t o r y r e s u l t s .
Addi t i onal l y, wi t h t hi s choice of model the const i t ut i ve data can be
determined easi l y from di r ect measurements i n si mpl e tensi on t est s.
Due t o t he q u a s i - s t a t i c nat ur e of t he probl em t o be model l ed,
mat er i al r at e ef f ec t s do not need t o be i ncl uded. Fr act ur e of t he
mat er i al i s not necessary i n t he model e i t h e r , as t he anneal i ng
process t o which the Aluminium was subj ected increased the d u c t i l i t y
suf f i ci ent l y t o remove f r act ur e from the range of rel evant phenomena.
6.1.2 TENSION TESTS - A REVIEW
Tension t est s have been wi del y used f or det ermi ni ng mechanical
propert i es of met al l i c mat er i al s. St andard s p e c i f i c a t i o n s e x i s t f or
c ar r y i ng out t hese t es t s : ASTM E8-82, ASTM E646-78, BS18. The
mat er i al behavi our can be eas i l y obt ai ned f rom t hese st andar d t e s t
procedures when s t r ai ns are smal l or moder at e, i n whi ch case t he
deformations i n the t ensi l e specimen may be assumed uni f orm (Over the
cross-sect i on and along the bar). Such i s not the case, however, when
s t r ai ns become l ar ge: t he specimen necks and def or mat i ons become
hi ghl y l ocal i zed, cr eat i ng as a r esul t non-uni form stress and st r ai n
di st r i but i ons.
In a t es t t o der i ve t he mechani cal pr oper t i es of a mat er i al
di scret e measurements are taken of forces and di spl acements, producing
average val ues f or st r esses and s t r a i n s . I f t he st r ess and s t r a i n
f i e l d s can be assumed uni f or m, t hese average measurements pr ovi de
di r ect l y the data f or the elementary const i t ut i ve l aws. However, al l
134
means of producing l arge pl ast i c deformations i nvol ve hi ghl y complex
non- uni f or m st r ess d i s t r i b u t i o n s . Appr opr i at e l ar ge s t r a i n t es t
procedures i ncl ude t or si on t est s on tubes or c y l i n d e r s , compressi on
t est s on cyl i nder s (i n which r esul t s must be correct ed f or bar r el l i ng
and f r i c t i o n ) , and tensi on t est s on cyl i ndr i cal bars or f l a t pl at es.
Tension t est s on cyl i ndr i cal bars were chosen here f or the si mpl i ci t y
i n measurements and s u i t a b i l i t y of avai l abl e equipment.
The anal ysi s and i n t e r p r e t a t i o n of t he necki ng def or mat i on i n
tensi on t est s i s a sol i d mechanics problem which has at t r act ed some
i nt er est . Bridgman (1944, 1945, 1952) and Oavidenkov and Spi ri di nova
(1946) were t he f i r s t t o st udy t h i s probl em i n some dept h. They
present ed a n a l y t i c a l i n t e r p r e t a t i o n s based on some ex per i ment al l y
confi rmed assumptions, which provi de st r ess/ st r ai n di st r i but i ons val i d
i n t he minimum sect i on of t he neck. The main assumpt i on i n both
Bri dgman' s and Davi denkov' s work i s t hat t he r adi al s t r ai ns are
constant across the minimum neck sect i on. This f act was corroborated
experi ment al l y by Davidenkov (1946) measuring the average grai n sizes
at di f f er ent posi t i ons and or i ent at i ons i n the neck sect i on. Bridgman
s i l v er - s ol der ed cores of var i ous di amet er s i nt o hol l ow speci mens,
obt ai ni ng a si mi l ar conf i r mat i on. Experimental evidence t o t hi s ef f ect
has been obt ai ned here f r om mi crohardness t es t s and i s pr esent ed
f ur t her down i n secti on 6.3.5.
Bridgman's and Davidenkov's earl y work i n the i nt er pr et at i on of
tensi on t est s has represented the st at e of the ar t f or many years. Not
u n t i l r ecent l y have mi nor ext ensi ons t o t h e i r a n a l y t i c a l work been
produced (Kaplan (1973), Jones (1979)).
On anot her f r o n t , t he devel opment of d i g i t a l comput ers has
enabled tensi on t est s t o be i nt er pr et ed i n a more general form through
the use of numerical models. The f i r s t general numerical sol ut i on was
obt ai ned by Wi l ki ns (1968) wi t h an ear l y ver si on of hi s HEMP f i n i t e
di f f er ence hydrocode, but r ecei ved l i mi t e d p u b l i c i t y . Fur t her work
al ong t h i s l i n e has been r epor t ed by Wi l ki ns (1978), Wi l ki ns et al
(1980), employing tensi on t est s t o obtai n mat eri al i deal i zat i ons f or
per f or at i on analyses and duct i l e f r act ur e t est models. Wi l ki ns car r i ed
out computer cal cul at i ons t o corroborate cer t ai n assumptions and gain
addi t i onal i nf or mat i on on c o n s t i t u t i v e par amet er s. However, i n t he
i n t e r p r e t a t i o n of t he t ensi on t es t s Wi l ki ns does not consi der t he
ef f ec t of t he non- uni f or mi t y of axi al st r esses i n t he neck. He
acknowledges the er r or , usi ng hi s si mpl e i nt er pr et at i on t o est abl i sh a
f i r s t crude guess f or t he mat er i al par amet er s, whi ch are t hen
corrected on the basis of the r esul t s of numerical cal cul at i ons.
Nor r i s et a l . (1978) have al so used HEMP t o si mul at e t ensi on
t e s t s . They f ol l owed e s s e n t i a l l y t he same method as Wi l k i n s ,
performi ng i t er at i ve computer analyses t o obt ai n uni versal f l ow stress
curves from the tensi on t est data. They reported t r i a x i a l tensi ons at
the specimen axis higher than those predi ct ed by Bridgman (1952).
A Kant or ovi ch numer i cal approach assumi ng gl obal pol ynomi al
di spl acement f unct i ons has been proposed by Chen (1971). He used an
ear l y l arge st r ai n f or mul at i on which contai ns some def i ci enci es such
as l ack of c o r r e c t i o n f o r r i g i d body r o t a t i o n s . The g l o b a l
di spl acement f unct i ons f a i l e d t o r epr esent adequat el y t he boundary
condi t i ons. A si mi l ar l arge st r ai n f or mul at i on was used by Needleman
(1972) wi t h i n a Fi n i t e Element cont ext . As Nor r i s (1978), Needleman
found Bridgman's (1952) sol ut i on t o underpredi ct the t r i a x i a l tensi ons
i n the specimen.
The work pr esent ed here uses both t h e o r e t i c a l s t r e s s / s t r a i n
d i s t r i b u t i o n s and numer i cal c al c ul at i ons f or t he i n t e r p r e t a t i o n of
t ensi on t e s t s . Anal y t i c al r es ul t s and empi r i cal cor r el at i ons due t o
Bridgman (1952) are f i r s t used t o pr ovi de an i n i t i a l guess f or a
uni ver sal f l ow st r ess curve f rom si mpl e t ensi on t es t dat a. Thi s
mat er i al law i s t hen f ed i nt o t he comput at i onal model , and by
compari ng t he r esul t s wi t h exper i ment , t he paramet ers are adj ust ed
more f i n e l y . The i n t e r p r e t a t i o n based on Bri dgman' s semi - empi r i cal
r esul t s pr ovi ded wi t h l i t t l e e f f o r t very reasonable guesses f or the
mat er i al l aws, and onl y mi nor changes were suggested by t he
c o mp u t a t i o n a l r e s u l t s . On t he ot her hand, t he s t r e s s / s t r a i n
di st r i but i ons obtai ned i n the cal cul at i ons conf i rm broadly Bridgman's
(1952) and Davidenkov' s (1946) assumptions.
Although the numerical approach used by Wi l ki ns (1978) i s broadly
s i mi l a r t o t he present one (both employ e x p l i c i t f i n i t e di f f er enc e
codes) nei t her he nor Nor r i s et al . ( 1978) used any i n t e r p r e t a t i o n of
t he t ensi on t es t data f ed i nt o t he computer model ; t he st r ess was
merel y averaged across t he neck sect i on. Adj ust ment of mat er i al
s t r e s s - s t r a i n law paramet ers was achi eved by i t e r a t i v e computer
anal yses. Here i t was f ound t hat t he di f f er enc e between t he d i r e c t
average data ( f i g . 6.4) and t he mat er i al uni ax i al s t r e s s - s t r a i n law
( f i g. 6.7) was consi derabl e f or the reducti ons i n area sustai ned (20%
di f f erence f or A/ A
o=0#2
6). An accurate i n i t i a l guess coul d be obtai ned
using a si mpl e semi -empi ri cal i nt er pr et at i on, thus saving consi derabl e
computer resources.
6.2 THEORETICAL INTERPRETATION OF TENSION TESTS
Tensi l e t est s on cyl i ndr i cal bars have been used ext ensi vel y t o
study the mechanical behaviour of metal s. For b r i t t l e or low d u c t i l i t y
met al s t he s t r a i n and st r ess f i e l d s are appr oxi mat el y uni f or m
t hr oughout t he t e s t , wi t h smal l r educt i ons i n ar ea. Fr act ur e occurs
before departi ng from t hi s uni f orm st at e. Uni f or mi t y i s also the case
f or a more d u c t i l e ma t e r i a l , when onl y t he det er mi nat i on of t he
el ast i c constants or the y i el d poi nt i s sought.
Duct i l e metals can sustai n l arge pl ast i c st r ai ns before f r act ur e.
The reducti on i n cross-sect i onal area i s i mpor t ant , and the or i gi nal
area (A
Q
)
a n d g a u
ge l engt h (Lo) no l onger pr ovi de a v a l i d basi s on
which t o measure stresses and st r ai ns. In a f i r s t phase of moderate
r educt i ons i n ar ea, def or mat i ons remai n u n i f o r m; t r ue (Cauchy)
stresses and natural ( l ogar i t hmi c) st r ai ns provi de a good descr i pt i on
of t he mat er i al behavi our . For a uni ax i al t e n s i l e t es t t hey are
defi ned thus:
where:
-zz
ffzz = P/A
dL/L = Ln( L/ L
0
;
Lo
!6.3)
6. 4'
a
z z
i s the t r ue axi al stress
P i s the appl i ed axi al load
A i s the current cross-sect i onal area
e
zz i s the natural axi al st r ai n
l
0
i s the or i gi nal gauge l ength
L i s the current gauge length
The natural st r ai n keeps a cumul ati ve record of the st r ai ni ng i n
each di r ect i on at a poi nt i n the mat er i al , ref erred at every i nst ant
t o the current conf i gur at i on. This descr i pt i on i s very convenient f or
t he t e n s i l e t e s t , f or as l ong as def or mat i ons remai n uni f or m, t he
axi al stress coi nci des wi t h the Von Mises equi val ent st r ess, and the
natural axi al st r ai n equals the ef f ect i ve pl ast i c st r ai n:
CTeq = zz
( 6
-
5 )
A proof of t hi s l ast assert i on i s given f ur t her down i n secti on 6. 2. 1.
A secon d phase i n t he t e n s i l e t es t begi ns when t he specimen
s t ar t s t o neck. At t h i s moment t he mat er i al har deni ng no l onger
compensates f or the reducti on i n area, occasioning a decrease i n the
t o t a l axi al l oad. The mat er i al t r i e s t o harden enough t o match t he
appl i ed l oad, producing l ocal i zed deformations i n a small neck regi on,
t o which subsequent pl ast i c f l ow i s conf i ned; the rest of the specimen
unl oads e l a s t i c a l l y . A non- uni f or m st at e of st r esses and s t r ai ns
exi st s, and formul ae!(6.4) alone no longer provi de a val i d descr i pt i on
of t he t e s t . ( ( 6. 3)
6. 2. 1 STRAIN DISTRIBUTION AT MINIMUM NECK SECTION
From the assumptions of axi al symmetry and uni f or mi t y of radi al
st r ai ns, and consi deri ng the symmetry condi t i on across the plane of
t he minimum neck s ec t i on, t he r at e of def or mat i on components (see
sect. 2.2.3) may be expressed as f ol l ows:
d
r r
= ^ r = D/D
r P
dr
4
ee
=
I + " r = 0+0/D = d
r r ( 6
.
6 )
r dd r
d
ro = dtf
Z
= drz = 0
where: r, 0 ,z st and f or r a d i a l , c i r c u mf e r e n t i a l and ax i al
di r ect i ons respect i vel y
^ i j are the physi cal components of the rate of def ormat i on.
u
.j are the displacement components
D i s the diameter
superposed dots represent materi al time der i vat i ves as usual
From (6.6) i t may be seeen t hat uni f or mi t y of dr r i mpl i es dr r
a
0O
as we l l . Thi s f ac t was checked i ndependent l y by Davi denkov(1946)
measuring grai n sizes i n both radi al and ci r cumf er ent i al di r ect i ons.
In the necking phase el ast i c st r ai ns are negl i gi bl e so i t may be
assumed t he def or mat i on i s produced e n t i r e l y by p l a s t i c f l ow. The
i ncompr essi bi l i t y char act er i st i c of pl as t i c i t y i n metals permi ts the
der i vat i on of the axi al rate of deformati on component:
]
zz = - 2 d
r r
= -2D/D ;6.7)
From eqns. (2. 60b), (6.6) and (6.7) t he e f f e c t i v e p l a s t i c s t r a i n may
be expressed as
.D
-2(D/D)dt = -2Ln(D/D
0
) =
e
zz
(6. 8)
Do
where Do
i s t n e o r i
9
i n a
l diameter of the bar.
Thus ep and consequent l y t he y i e l d st r engt h Y are const ant
across the minimum neck sect i on.
Thi s r esul t has been cor r obor at ed experi ment al l y here wi t h the
help of microhardness measurements, taken on secti ons of the deformed
t ensi l e specimens (sect . 6. 3. 5) .
6.2.2 STRESS DISTRIBUTION
From the Levy-Mises pl ast i c f l ow r ul e.
s
i j = Xdi j p ( 6 . 9 )
where X i s an a r b i t r a r y p o s i t i v e scal ar . Consi der i ng eqn. (6.6) one
may then wr i t e:
<7
rr
=G
ee
(6.10)
Thi s equat i on was der i ved i ndependent l y by Nadai (1946) f rom
sol el y t heor et i cal consi derat i ons of a Von Mises yi el d surface and a
minimum f or ce p r i n c i p l e . Consi der i ng t he const ancy of Y across t he
sect i on, one may also wr i t e:
O
zz
_ C
rr
=
Y
( 6. 11)
which i s val i d f or al l poi nt s i n the sect i on. This st ress di st r i but i on
may be i nt er pr et ed as a uni f orm axi al tensi on of value Y superposed t o
a var yi ng hydr ost at i c st r ess of val ue > r
T h e
hy dr os t at i c s t r es s ,
due t o the f ree boundary condi t i on, vanishes on the outer edge. In the
i n t e r i o r t he si gn i s t e n s i l e due t o t he concavi t y of t he c ur v at ur e,
and from the symmetry condi t i on around the axi s, a maximum value must
be reached on i t . Equat i on (6.11) i mpl i es t hen t hat ^zz
> Y
throughout the sect i on; hence the average value must also be gr eat er ,
a
zz
> Y. This expl ai ns the di f f erence reported by Wi l ki ns et al.(1980)
i n t h e i r c al c ul at i ons wi t h HEMP, where f or an axi al s t r a i n of 0.52,
O
zz
= 460 MPa whi l e Y = 430 MPa .
From the general equi l i br i um equations
a
i j , j = 0 ( 6. 12)
Bridgman(1952) and Davidenkov(1946) deri ved the di st r i but i on of axi al
st r esses, dependi ng on t he r adi al coor di nat e r and on t he r adi us of
curvature of the l ongi t udi nal pr i nci pal stress l i ne, p :
f
D/2
tfzz = Y(l +1 dr.) ( 6. 13)
J v P
Bridgman(1952) proposes the f ol l owi ng formul a f or the radius of
curvature p depending on the radi al coordi nate :
p = (D2/4 + OR - r 2 ) / 2 r
( 6. 14)
where R i s t he geomet r i cal r adi us of cur vat ur e at t he r oot of t he
neck. Subst i t ut i ng i n eqn. (6.13) and i nt egr at i ng t o f i nd the average
st r ess:
zz
= Y
^
+
Ln[(D
2
/4+DR-r
2
)/DRj )
(6.15)
a
zz
= Y( l + 4R/D)Ln(l + D/4R)
Thi s l as t equat i on gi ves t he val ue of t he mat er i al st r engt h Y
f rom t he measured average axi al st r ess ^zz* ^
d e
P
e n d s o n t n e
p o s s i b i l i t y of measuri ng R, t he r adi us of c ur v at ur e. Thi s can be a
l i t t l e bi t cumbersome i n a t est where mul t i pl e measurements are t o be
taken. Bridgman(1952) produced an empi ri cal cor r el at i on from mul t i pl e
obser vat i ons f or d i f f e r e n t sor t s of st eel and ot her met al s t hat
expresses D/R as a f unct i on of the reducti on of area A/A
0
/
o r w n a t
^
s
the same, the axi al natural st r ai n
e
zz'
:
D/2R = / L n A
o / A
.
QA
=^e
zz
- 0.1 (6.16)
This r esul t al l ows the cor r ect i on f act or ^ z z ^
t 0 t ) e e x
P
r e s s e d
di r ect l y as a f unct i on of the axi al st r ai ns:
a
z z
/ Y = [1 + 2^e
z z
- 0. 1] Ln[ l +^ e
z z
- 0 . 1 /2J (6.17)
Thi s f or mul a may be appl i ed onl y f or
e
zz > 01 For
e
zz
< u , i
the deformati on i s assumed uni f or m, and no cor r ect i on i s needed.
6.3 TENSION TESTS
6.3.1 SPECIMENS AND MATERIAL
The mat eri al used i n the t est s was HE30 (BS1474) Aluminium al l oy,
wi t h t he f o l l o wi n g composi t i on : 0.1% Cu, 0.4-1.5% Mg, 0.6-1.3% Si ,
0.6% Fe, 0.4-1.0% Mn, 0.1%Zn, 0.5%Cr, 0.2% ot her s, and t he r est
Aluminium. This mat eri al comes i n extruded cyl i ndr i cal form.
141
Mechanical propert i es i n the as-recei ved condi t i on were obtained
from the manufacturers:
0.1% proof stress 239 - 270 MPa
densi ty 2700 Kg/m
3
Young's modulus, E 67000 MPa
Poisson' s r a t i o , v 0.3
Ul ti mate t ensi l e st rengt h 278-293 MPa
El ongati on on 2 i n 7 - 10%
Thi s mat er i al was cut and machined t o f or m speci mens of t he
geometry descr i bed i n f i g u r e 6.2 . A s l i g h t t aper was gi ven i n or der
t o cont rol the necking posi t i on, the smal l est diameter being l ocated
at the mi d-sect i on.
The composi ti on and treatment of HE30 Aluminium i s i dent i cal t o
the HT30 Aluminium tube mat eri al employed i n the tube crumpl i ng t est s.
The onl y di f f er enc e l i e s i n t he f or mi ng processes ( ext r usi on or
dr awi ng) . As a r e s u l t , mechani cal pr oper t i es f or bot h mat er i al s are
very s i mi l a r . The smal l di f f er enc e was assumed t o vani sh i n t he
annealing process t o which both specimen types were subj ected (sect.
7. 3. 2. 1) .
6 . 3 . 2 PROGRAMME
The specimens were annealed at 350oc pr i or t o t est i ng. Six t est s
i n al l were performed wi t h the f ol l owi ng annealing t i mes:
Test An n ealin g ti m e, hours
CTl
CT2
CT3
CT4
CT5
CT6
5
7
none
3
5
5
142
After annealing, all specimens were oven cooled.
6.3.3 PROCEDURE
The t est s were car r i ed out using an INSTRON Model 1195 machine,
wi t h a lOOkN l oad c e l l . The specimens were secured wi t h t r ansver sal
12.9 mm di amet er st eel pi ns t o a f i x ed crosshead at one end and a
moving crosshead at the opposite end. The load was appl i ed wi t h stroke
cont r ol , at crosshead vel oci t i es of between 0.5 and 1.0 mm/min , up t o
specimen f r a c t u r e . Frequent measurements of t he neck di amet er were
taken wi t h a Verni er cal i per (preci si on ;o,o5 mm).
6.3.4 RESULTS
The r es ul t s of t he t est s are summari zed i n f i gur es 6.3 and 6.4.
Figure 6.3 shows the axi al load agai nst the axi al st r ai n
e
zz
a t t h e
minimum neck sect i on. In a l l t he t es t s a peak val ue of t he l oad i s
reached when the necking i ns t abi l i t y ar i ses. Al l specimens show f a i r l y
close r esul t s except CT3, which was not annealed. CT3 shows an ear l i er
onset of necki ng, at around e
z z
=0. 07,
a n d m u c h n i
9
h e r l e v e l s o f
l oads, wi t h peak value 57kN. This corresponds t o a st rengt h of 285MPa
ref erred t o the or i gi nal cross-sect i onal area, i n accordance wi t h the
manufacturer' s speci f i cat i ons (sect. 6.3.1). The non-annealed CT3 t est
showed al so consi der abl y l es s d u c t i l i t y (max. n a t u r a l s t r a i n
e
zz=0. 70).
Results for the remaining tests all lie within a narrow band;
peak loads range between 21 and 25kN, onset of necking between
e
zz=0.15 - 0.20 , and m a xi m u m strains before fracture between
e
z z
= 1.15 - 1.25 . Differences in annealing time clearly did not
influence the behaviour beyond experimental uncertainty. For instance,
specimens CT5 and CT6 were annealed for longer time (5 hrs) than CT4
(3 hrs), yet they yield results which are slightly closer to the non
annealed CT3. These differences must be attributed to experimental
scatter. The annealing process may be assumed complete for all
specimens except CT3.
00
00
00
00
00
00
9
00
0
0
0
0
0
0
0
0
- f r act ur e
1 s v,
'A V,
A ^
^
A
v
AV
A ^
CT3 (non-annealed)
o CTA
CT5
* CT6
A
V
J_
0.00 0.20 0.40 0.60 0.80 1.00 1.20
N a t u r a l S t r a i n , e
zz
Fi g u r e 6. 3s TENSION TESTS - AXIAL LOAD VS. STRA
In f i gur e 6.4 t he average axi al st r ess at t he neck i s pl ot t ed
agai nst t he axi al s t r a i n f or t he anneal ed speci mens. Al t hough t he
t o t a l l oad drops af t er t he onset of necki ng due t o t he geomet r i cal
r educt i on i n ar ea, i t may be seen t hat at t he el ement ar y l evel t he
materi al does not cease t o harden.
Fr act ur e occur r ed wi t h d u c t i l e cup-and-cone geometries f or al l
t he anneal ed speci mens. A r epr esent at i ve p i c t u r e of one of t he
f r ac t ur ed specimens i s gi ven i n f i g u r e 6.5 . For t he non-anneal ed
specimen CT3 f r a c t u r e was more b r i t t l e and occur r ed e a r l i e r wi t h a
bang, producing a much less wel l developed neck ( f i g. 6.6).
Judging from these r esul t s i t was concluded t hat compl eti on of
t he anneal i ng process was reached f or CTl , CT2, CT4, CT5 and CT6
specimens. The mechanical propert i es deri ved from each of the t est s
may be averaged i n order t o obt ai n r epr esent at i ve pr oper t i es f or
annealed HE30 aluminium al l oy.
6.3.5 MICROHARDNESS MEASUREMENTS
The necked bars were employed t o st udy t he d i s t r i b u t i o n of
p l a s t i c s t r ai ns across t r ansver sal sect i ons. Thi s was done t hr ough
microhardness measurements performed i n a sect i on normal t o the axi s
of one of the t ensi l e specimens af t er f r act ur e. This secti on was the
nearest t o t he f r a c t u r e zone whi ch di d not show v i s i b l e voi ds or
change t o i t s s t r u c t u r e f r om t he f r a c t u r e . A f t e r p o l i s h i n g ,
microhardness i ndent at i ons were performed at several l ocat i ons along
t wo normal r a d i i . Fur t her d e t a i l s of t he mi cr ohar dness t e s t i n g
procedure and equi pment are gi ven i n sect i on 7. 3. 1. 3. 1. The r es ul t s
are shown i n t abl e 6. 1; t hey show no s i g n i f i c a n t v a r i a t i o n of t he
hardness accross the sect i on.
A
veraqe ax
> g <
ial stress, P/A (MPa)
o
o
o
o
CD
o
4>
O
O
O
o
o
o
o
CO
o
o
o
o
o
CD
o
o
o
CD
o
o
o
o
o
o
o
CO
o
CD
CD
2
0
0
.

CD
o
2
2
0
.

CD
o
o
o
o
o
o
IV)
o
4>
o
o
o
CD
O
CO
o
o
o
<a
o
t>

> <c o
*
9
1
3

o <a t>
n n o o
i i i i
ui -i> rv)
>

Q *
<
> o
><
c
0
Q
t>
<1
O
< Q
iv)
o
>
O
#o
[>
1 2 -f
1 1 -|
-10g
9 -g
8 -g
7 -g
6 -H
4 -=:
1 -=
M o d e m
: ng land c
O
igure 6.5: CT4 tensile specimen after fracture Figure 6.6: CT3
147
Point
1
2
3
4
5
6
7
Indentation Diagonal
(Microns)
53
53
54
52
52
53
53
Microhardness
H
m(Kg/mm2)
64
64
63
66
66
64
65
I ' 5 6 7
*r
Current diameter
D = 9.2mm
Ori gi nal diameter
D
o = 15.9mm
Microhardness load
P = 99g
e
zz
=
-2Ln(D/Do)
= 1
-
n
Tab! e 6. 1: Mi crohardness measurementes i n neck sect i on of t e n s i l e
specimen CT4.
From si mpl e consi der at i ons of p l a s t i c i t y t heor y ( Hi l l (1950),
p.254), the hardness value H can be r el at ed l i near l y t o the mat eri al
st r engt h Y, i t s e l f f unct i on of the pl ast i c st r ai n e
H = CY( e
p
)
C bei ng a const ant whose val ue l i e s between 2.5 and 3.0 . Hence t hese
microhardness resul t s support the assumption of constancy of Y and e
across the minimum neck secti on (sect. 6.2.1).
6.4 MATERIAL HARDENING LAW
For aluminium power laws of the type
a= A e' (6.18)
where: a uni axi al t rue stress
e uni axi al natural st r ai n
A,n materi al constants
148
f i t t he mat er i al behavi our very we l l . The paramet ers A,n have a
physical i nt er pr et at i on: A equals the stress f or uni t st r ai n, and i t
can be shown t hat n corresponds t o t he s t r a i n at t he peak l oad i n a
t ensi l e t es t .
Given the equivalence between the uni axi al stress and equi val ent
st r ess poi nt ed out i n eqn. ( 6. 5) , a conveni ent and general f orm of
expressing the mat eri al behaviour i s
Y = A( e
z z
)
n
= A(
p
+Y/ E)
n
? Ae (6.19)
Y/E represents the el ast i c part of the st r ai ns; these are usual l y very
small compared t o the pl ast i c par t , e
p>
The average axi al st r ess data pr esent ed i n f i g u r e 6.4 were
cor r ect ed wi t h f or mul a (6.17) t o obt ai n t he val ues f or t he mat er i al
s t r engt h, Y. A power law Y = A( e
z z
) n was t hen f i t t e d t o t he data
from each t est , by performi ng a l i near regressi on i n the l ogar i t hmi c
values (t abl e 6.2). The parameters A,n so obtai ned f or each t est were
aver aged y i e l d i n g mean val ues of A = 181.7 MPa, n=0.182; t hese
paramet ers pr ovi de a reasonabl e f i r s t est i mat e f or t he mat er i al
hardening law.
Test
CTl
CT2
CT4
CT5
CT6
Average
A (MPa)
169.1
172.5
184.7
191.5
190.6
181.7
n
0.190
0.177
0.196
0.170
0.178
0.182
Table 6.2: Results of power law f i t s t o tensi on t est r esul t s
149
The r esul t s of t h i s average f i t , t oget her wi t h t he equi val ent
stress - st r ai n data deri ved from experi ment, are pl ot t ed i n f i gur e
6.7 . Comparing wi t h t he average axi al st r esses i n f i g u r e 6.4 t he
i nf l uence of t he c or r ec t i on f or mul a (6.17) i s pat ent , l ower i ng
consi der abl y t he st r ess val ues f or t he l ar ger s t r a i n s . For a u n i t
st r ai n, the average axi al st ress i s approxi matel y 215 MPa, whi l e the
equi val ent stress i s lowered t o 182 MPa , a reducti on of 18%.
Al t hough t he "average" power law obt ai ned r epr esent s a good
over al l f i t , i t can be seen i n f i g u r e 6.7 t hat some di scr epanci es
s t i l l ex i s t i n t he r egi on of low s t r ai ns (
e
zz
<
0
>2
) > w h i l e
t n e
f i t
i s very good f or t he hi gher s t r a i n s . I t can al so be seen t hat t he
di sconti nuous appl i cat i on of the cor r ect i on formul a (6.17) from
e
zz
>
0.1 has produced a d i s c o n t i n u i t y i n t he sl ope of t he st r ess - s t r a i n
dat a.
I t i s reasonabl e t o suspect t hat t he c r i t i c a l r egi on f or
appl i c at i on of t he c or r ec t i on f or mul a (6.17) i s i n t he l ow s t r a i n
regi on, before the necking or when the neck curvat ure i s smal l . I t i s
here t hat the empi ri cal r el at i onshi p (6.16) between the neck curvat ure
and t he axi al s t r a i n may be more i n er r or . Thi s r e l a t i o n s h i p has a
bi as t owar ds s t e e l , whi ch usual l y necks e a r l i e r t han al umi ni um.
Great er conf i dence can be pl aced i n eqn. (6.16) f or t he hi gher
st r ai ns, where the necking i s wel l developed.
Consequent l y t wo a d d i t i o n a l cur ve f i t s t o t he cor r ect ed
exper i ment al data have been t r i e d . These mai nt ai n r oughl y t he same
st r ess val ues f or t he hi gher s t r a i n s , but d i f f e r s l i g h t l y f rom t he
average curve f i t f or the lower s t a i n regi on. A si mpl e way f or t hi s
v a r i a t i o n i s , mai nt ai ni ng t he power law approach, t o keep t he same
value f or parameter A (which represents the st ress f or uni t st r ai n) ,
and vary the exponent n. Two addi t i onal values of n were t r i ed :
n l
= 0.159 whi ch gi ves a cur ve i n t he mi ddl e of t he st r ess
band, f or e
zz
< Q. 2
n
2 = 0.133 whi ch gi ves a curve near t he t op of t he st r ess
band, f or
e
zz
<
^2
appears that n
1 S
hou1d give the best fit, but this will be
150
I M I
II
-
e* l
# 1 <
1 D>
s
0
<
\ 0
1 ^ c\i <r un -o
#, a i h- h- i i
H ] <g U U U U U
I
S ^ <a > O *
P <|
* \P ><
V- >
<
\i
b ><3
1
\ t o t >
' i *
G
>
i i i i i i i ^ ^
i
S
d c>
CO Ci)

0 0 CO
II
o
1
1
1
1
II
0
1 1
o

00
II
o
1
1
1
_
-
-
-
1
o
o
o
-3"
o
o
o
CM
CM
o
o
o
o
CM
O
O
O
QO
O
O
O
SO
O
O
o
o
o
o
CM
o
o
o
o
o
o
o
00
o
o
o
SO
o
o
o
-4"
o
o
o
CM
o
CM
o
o
o
00
o
SO
o
o
o
o
CM
o
o
o
o
o
o
e^^j) s s e j u g aue] BA ! nbg
151
deci ded f rom t he r es ul t s of numer i cal c al c ul at i ons usi ng t he t hr ee
suggested curve f i t s , and t h e i r compari son wi t h exper i ment al
observat i on.
6. 5 NUMERICAL CALCULATIONS FOR TENSION TESTS
The tensi on t est s were si mul at ed numeri cal l y. From the r esul t s of
the cal cul at i ons i t was possi bl e t o decide on the power law parameters
t hat provi de the best f i t f or the experimental r esul t s. Addi t i onal l y,
the cal cul at i ons provided a check f or the v al i di t y of the assumptions
t hat were made i n t he i n t e r p r e t a t i o n of t he t e n s i l e t e s t f or t he
stress and st r ai n di st r i but i ons. Fi nal l y t hi s exerci se was usef ul as a
val i dat i on of the Expl i ci t Fi ni t e Di f f erence model proposed i n t hi s
t h e s i s ( c hapt er 4) , when a p p l i e d t o l a r g e s c al e n o n - l i n e a r
computations f or pr act i cal engi neeri ng problems.
6. 5. 1 MODEL
The cal cul at i ons were performed usi ng axi symmetri c anal ysi s, wi t h
a 2-D d i s c r e t i z a t i o n i n t he r-z pl ane ( f i g . 6.8). The symmetry
condi t i on around the plane of the minimum cross-sect i on (z=0) al l owed
a f ur t her r educt i on t o a quar t er of t he r - z s ec t i on. The mesh
consi sted of 875 t r i angul ar c el l s , grouped i nt o 438 MTQ elements (see
sect. 4.2.2) and 390 nodes. The model di d not comprise the end f i xi ngs
as t hey l i e f ar f rom t he area of i n t e r e s t and i t was consi der ed t hat
they behave i n pract i ce as r i gi d l oad- t r ansmi t t i ng devi ces. Only the
cent r al t aper ed por t i on of h a l f - l e n g t h 37.5 mm was model l ed. Shear-
f r ee boundary condi t i ons were assumed at t he t op and bot t om ends of
the specimen modelled. The mesh becomes progressi vel y f i ner near the
neck r egi on, whi ch i s t he area of i n t e r e s t , and where t he st r ess and
s t r a i n gr adi ent s wi l l be st eepest . Ad d i t i o n a l l y , a hi gh r:z aspect
r a t i o was gi ven t o t he el ement s i n t h i s area i n or der t o avoi d
excessi vel y di st or t ed shapes when the neck i s elongated i n the axi al
di r ect i on.
152
? * *
S.I mm
A x i s o f symmetry.,
to p s ec ti o n -
(p r es c r i b ed m o ti o n ] '
b o tto m s ectio n_
(fixed)
1
,
37.5 mm
795m m .
Figure 6.8. MESH USED FOR TENSION TEST CALCULATIONS
12
10
6 -
T E N S 5
T E N S 7 T E N S 3
0.00 0.20 0.40 .0.60 0.80 1.00 1.20 1.40 1.60
T I me (msec)
Figure 6.9, HISTORIES OF VELOCITY APPLIED AT END OF NUMERICAL MODEL
153
6.5.2 ANALYSIS
The t est s were performed under quasi - st at i c condi t i ons at speeds
of between 0.5 - 1.0 mm/min. The numerical al gori t hm used solves the
equat i ons of cont i nuum mechanics i n t i me wi t h an e x p l i c i t t i me -
marchi ng scheme; t h i s l eaves no opt i on f or s t a t i c anal y s i s , as
c a l c u l a t i o n s r epr es ent al ways a f u l l dynami c model . As t he
computational t i me- st ep i s l i mi t ed f or numerical s t a b i l i t y t o a small
val ue ( sect . 4. 7), i t woul d be t oo expensi ve i n computer t i me t o
perform the dynamic anal ysi s i n real t i me.
To be abl e t o per f or m t he c al c ul at i ons v e l o c i t y scal i ng was
i nt r oduced (see sect i ons 4. 1, 5.4). The v e l o c i t y of def or mat i on was
increased i n order t o bri ng down the number of computati onal cycles
necessary f or t he anal y s i s , f rom t he 2x1010 t hat woul d be needed i f
the real vel oci t y of deformati on were t o be used, t o around 4x104.
Vel oci t y scal i ng usual l y works wel l f or quasi - st at i c problems, as
l ong as t he def or mat i on i s appl i ed sl owl y enough so as t o l e t t he
st r ess- waves t r a v e l back and f o r t h al ong t he model sever al t i mes ,
al l owi ng r edi st r i but i on of stresses. In other words, the model must
not depar t t oo much f rom t he s t a t i c e q u i l i b r i u m whi ch must be
si mul ated at each step i n t i me. Two ways of quant i f yi ng t hi s departure
from equi l i br i um are:
- Ensuri ng t hat t he l oad h i s t o r i e s at opposi t e ends of t he specimen
are approxi matel y equal (the out of balance forces must be small when
compared t o the t ot al f or ces) ;
- Checking t hat t he k i n e t i c energy i s smal l compared t o t he ov er al l
energy i nvol ved i n the deformati on process.
Three cal cul at i ons are reported here, f or each of the power law
curve f i t s f or t he mat er i al c o n s t i t u t i v e behavi our suggested i n
sect i on 6.4 (eqn. 6.18):
Y = A e
n
Analysis
TENS3
TENS5
TENS7
A (MPa)
181.7
181.7
181.7
n
0.182
0.159
0.133
I n t he c a l c u l a t i o n s t he t y p i c a l a x i a l v e l o c i t y was 10 m/s and
t o t a l t i mes of anal ysi s were between 1.25 msec and 1.6 msec.
The neck sect i on was kept f i x ed i n i t s pl ane whi l e a pr escr i bed
v e l o c i t y was a p p l i e d t o t he upper end ( f i g . 6. 8) . The h i s t o r i e s of
appl i ed v e l o c i t i e s are gi ven i n f i g u r e 6.9 . Ramps were i nt r oduced t o
a v o i d sudden changes c a u s i n g a l t e r a t i o n s t o t h e e q u i l i b r i u m .
Consi der i ng t he st r ess- wave pr opagat i on v el oc i t y i n Al umi ni um of 6000
m/s and an i n i t i a l l engt h of 37.5 mm , t he l ar gest st r ess- wave per i od
i s ( i n i t i a l l y ) 0.0125 msec . Thus t he number of st r ess- wave per i ods i n
t he c a l c u l a t i o n s , or i n o t h e r wor ds , t he number of t i me s t he s t r e s s
waves t r a v e l l e d back and f o r t h r e d i s t r i b u t i n g st r esses, i s between 100
and 128. We s h a l l see l a t e r ( s e c t . 6. 5. 3) t h a t t he d e p a r t u r e f r om
e q u i l i b r i u m i n t he f or ce h i s t o r i e s t h i s amount of v e l o c i t y scal i ng
ocassi oned was s mal l .
The i mpor t ance of t h i s c o s t - c u t t i n g compromi se wi l l be c l ear by
ment i oni ng t hat CPU t i me f or each of t he c al c ul at i ons i n a CRAY-IS was
bet ween 30 and 45 mi n u t e s . Thi s r e p r e s e n t s c o n s i d e r a b l e comput er
r e s o u r c e s . I n a model wi t h o u t v e l o c i t y s c a l i n g t he c o s t s woul d be
5x105 t i mes hi gher (tens of year s CPU t i me) , way beyond the f eas i bl e
range wi t h t he cur r ent devel opment i n d i g i t a l comput ers.
6. 5. 3 RESULTS
The average st r es versus s t r a i n r es ul t s ( f i g . 6.10) l i e pr ec i s el y
wher e i t was pr esumed f r om t he c h o i c e of c ur v e f i t s done i n s e c t i o n
6.4 . I n a l l t h r e e cases f o r t he l a r g e s t r a i n r egi on ( e
z z
> Q. 2) t he
f i t i s very good, i n t he mi ddl e of t he exper i ment al band. For t he low
s t r a i n r egi on ( e
z z
< Q. 2 ) t he r e s u l t s f r om TENS3 ar e i n t he l ower
p a r t of t he band, TENS5 i n t he mi d d l e , and TENS7 i n t he upper p a r t .
(V.
OD
r\j -<i- i_n o
I I t I H-
u u u u u
i i i i i
z -z. z z z
LU LU LU LU LU
cz cz cz cz cz
LU LU LU LU LU
CL CL Q_ Q_ CL
X X X X X
LU LU LU LU LU
<\ t> O LU
o
o
2
4
0

o
CD
O
o
o
o
o
o
o
o
CO
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
CO
o
o
o
o
o
o
o
o
o
f \ l
o
o
o
;
dU) V/d "ssajug ] ejxe ebejaAy
26. 00
24. 00
22. 00
20. 00
18.00
16.00
14.00
12.00
10.00
8.00
6. 00
4. 00
2. 00
0. 00
3
\-
CALCULATION (TENS3)
CALCULATION (TENS5)
- - CALCULATION (TENS7)
A EXPERIMENT (CT1)
v EXPERIMENT (CT2)
G EXPERIMENT (CT4)
0 EXPERIMENT (CT5)
* EXPERIMENT (CT6)
0.00 0.20 0. 40 0. 60
N a t u r a l a x i a l S
Fi gur e 6.11s TENSION TEST CALCULATI
157
Thi s suggest s t h a t t he cur ve f i t used f o r TENS5 i s t he most
appropri at e, as had been ant i ci pat ed. Hence, the hardening law f or the
HE30 aluminium mat eri al wi l l be taken as:
Y = 181.7 e
0
*
1 5 9
(MPa) (6.20)
A view of the deformed mesh for TENS5 with all symmetries plotted
is given in figure 6.12. At this point the axial strain at the neck
was
e
zz
=
*" *
The curves presented i n f i gur es 6.10 and 6.11 were taken from the
cal cul at ed t i me- hi st or i es, af t er aver agi ng t he l oad between t op and
bot t om of t he model , and havi ng appl i ed a numer i cal f i l t e r wi t h a
centred moving average techni que, t o f a c i l i t a t e the vi sual i zat i on and
i nt er pr et at i on. The exact t i me- hi st or i es obtai ned i n the cal cul at i ons
f or t op and bot t om l oads are gi ven i n f i g u r e 6.13 . I t can be seen
t hat a st at e of st at i c equi l i br i um exi st ed f or most of the anal ysi s,
t he t wo l oad h i s t o r i e s bei ng appr oxi mat el y equal . At about 1.1 msec
however, some o s c i l l a t i o n s occur i n t he h i s t o r i e s c r eat i ng c e r t a i n
depar t ur e f r om s t a t i c e q u i l i b r i u m. These o s c i l l a t i o n s have
approxi matel y the peri od of the l ongi t udi nal stress waves t r av el l i ng
along the specimen:
T = 2 L. = 2 45xl0-3m = 0.015 msec
c 6000 m/s
L bei ng t he cur r ent l engt h of t he speci men and c t he v e l o c i t y of
st r ess waves. I t can be appr eci at ed t hat onl y a smal l number of
st r ess- wave per i ods occur i n an i n t e r v a l of t i me at whi ch l ar ge
var i at i ons i n the neck si ze (see f i gur e 6.14) and mat eri al behaviour
t ake pl ace. The model does not have t i me t o di gest t hese changes
qui ckl y enough and the i nf or mat i on i s t r ansmi t t ed i n jumps or i gi nat i ng
t he consequent e l a s t i c o s c i l l a t i o n s . I t i s i n t e r e s t i n g t o not e t he
l ar ger ampl i t ude of t he o s c i l l a t i o n s at t he t op end, whi ch i s
unloading el as t i c al l y , than at the neck, where the pl as t i c i t y of the
mat eri al behaviour dampens out the osci l l at i ons. At no moment however
do the osci l l at i ons i n load depart from the equi l i br i um value by more
than 10%, and the expected t rend of the load curve i s mai ntai ned.
Ibtf
For details of
initial geometry
see Figure 6.8
_L L
K 1 /\ f\ /\ r\ I A
^gHIHN^g
V-HLLLl LI j IH
pHrtm::"::inhhW
mm
l w
: :
i | |
iBBk
| S E E E H E E E | 1
p
'
strain at neck
e= 1.22
Figure 6.12. DEFORMED MESH FOR A/A
o
=0.29
HE30 ANNEALED ALUMINIUM BAR TENSION TEST CALCULATION (TENS5)
26.00
24.00
22.00
20.00
18.00
16'. 00
14.00
12.00
10.00
8.00
6.00
A.00
2.00
0.00
^ } (' I.
I T
" :;.
h i i yn
' ( i
LOAD AT MINIMUM NECK SECTION
LOAD AT TOP END OF SPECIMEN
NUMERICAL CALCULATION FOR.TENSION TEST (TENS5)
HE30 ANNEALED ALUMINIUM BARS
7 ' ,i
0.00 0.20 0.40 1.00 1.20 0.60 0.80
T 1 me (msec)
F i g u r e 6.13= C A L C U L A T E D L O A D H I S T O R I E S W I T H O U T F I L T E R I NG
8.00
7.50
7.00
6.50
6.00
5.50
5.00
4.50
0.60 0.80
T i me (msec)
1.20
Figure 6.14: HISTORY OF NECK RADIUS
For f i gur es 6.10 and 6.11 t he l oad h i s t o r i e s were averaged and
f i l t e r e d , removing osci l l at i ons of frequenci es hi gher than 50000 Hz,
which corresponds roughly t o the stress-wave peri od. In t hi s way the
over al l t r end of t he r es ul t s can be seen more c l e a r l y , wi t hout
unwanted numerical osci l l at i ons, due mainly t o the compromise made i n
the vel oci t y scal i ng.
Some cont our pl ot s of d i f f e r e n t st r ess components f or t he neck
region are presented i n f i gur e 6.15; the f ol l owi ng observati ons may be
made about these:
a) The r a d i a l ( f i g . 6. 15b) and hoop ( f i g . 6.15c ) s t r e s s e s ar e
approxi matel y equal i n the neck regi on, t endi ng t o zero at the outsi de
edge, and reaching maximum values near the axi s of about 60 MPa;
b) The ax i al st r ess ( f i g . 6.15d) i s c l e a r l y non- uni f or m across t he
neck s ec t i on, decr easi ng f rom approx. 270 Mpa i n t he axi s t o 200 MPa
on the out si de;
c) The y i e l d st r engt h Y (eqn. 6.2) i s r oughl y uni f or m across t he neck
secti on ( f i g. 6.15a), wi t h an approximate value of 189 MPa;
d) The axi al st r ai n at the minimum sect i on i s e
z z =
i
#
22 which from
eqn. (6.20) corresponds t o a f l ow stress of
Y = 181. 7xl. 22
0
'
159
= 188 MPa
and an average axi al stress (from eqn. (6.17)) of
<7
ZZ
= Y(l+2//T7l2)Ln(l+lfr7l2' /2) = 1.23Y = 231 Mpa
These val ues correspond very c l os el y t o t hose expressed above i n c)
and b) respect i vel y.
The y i el d st rengt h contours f or the complete specimen are pl ot t ed
i n f i g u r e 6.16, showi ng t he much l ower l ev el s of s t r ai n- har deni ng
whi ch occur r ed i n t he upper par t of t he model . Thi s har deni ng
happenned i n the f i r s t stage of uni f orm deformati on of the specimen;
af t er the necking t hat region has been unl oadi ng el as t i c al l y .
:"155:"MPlV"-
T^
"165-fiPa-:-....
17.5 MPa'"7""
1 8 5 lip'a";
- i - ^
WPT' -yr-
1 ^ J:
: 40 MPa
a) Yield Strength Y
b) Radial Stress (j^
.MPa
feO...MJ
4'0-MPa
" " ' [ " ( ' ) " ! ' "[' " V ' - ; \ - - - : - ;.........
/ 12tfMPa/ / " '"''.-/lBOMPa
c) Circumferential Stress On
d) Axial Stress (jr.
Figure 6.15: Tension test calculations - Stress contours in neck region
: r
16:
H -
:/.140(JMEa.:.
\.
V.
. l i . 1. 30. . ^. .
: ( :
\ :
.\l4Q:MEa.
: ~ T " 1 4 0 JTP&
!^}.'lSQ.
:
.Mp.a....
:
..
;i-50.;MP^:
:
r6
;
Q
;
-^p'a
: ^ 7 P^ PMP

Fi gur e 6.16: Tension t es t c al c ul at i ons - y i el d st rengt h (Y) contours
(general view)
163
Al l t hese r e s u l t s mat ch ver y c l o s e l y t he t h e o r e t i c a l
i n t e r p r e t a t i o n of t he t e n s i l e t es t det ai l ed i n 6.2, suggest i ng t he
general v al i di t y of t hat i nt er pr et at i on and of the numerical model.
6.6 CONCLUSIONS
1. Tensi on t es t s on c y l i n d r i c a l bars c o n s t i t u t e a si mpl e and
r e l i a b l e means of obt ai ni ng mat er i al s t r e s s / s t r a i n data at l ar ge
st r ai ns, when viewed as a 3-D pl as t i c i t y problem.
2. The i n t e r p l a y between t he t heor y ( sect . 6.2) and numer i cal
anal ysi s ( sect . 6. 5), coupl ed wi t h exper i ment al f ac t ( sect . 6. 3),
al l owed a si mpl e and r epr esent at i ve c o n s t i t u t i v e law f or t he HE30
Al umi ni um t o be est abl i shed (eqns. 6.18, 6. 20), f or use i n f ut ur e
l ar ge s t r a i n p l a s t i c i t y anal yses ( chapt er 7). Thi s c o n s t i t u t i v e law
assumes Von Mises pl as t i c i t y wi t h power law i sot r opi c hardeni ng.
3. Excel l ent agreement was achieved between numerical pr edi ct i ons,
experimental f act , and t heor et i cal understandi ng, of the behaviour of
a pl ast i cal l y deforming necked cyl i ndr i cal bar (secti ons 6.3.5, 6.5.3,
f i g s . 6.10, 6. 11, 6.15). The semi -empi r i c a l t heor i es of Bri dgman
(1952) and Davidenkov and Spi ri di nova (1946) were broadly confi rmed by
the numerical r esul t s.
CHAPTER 7
CONCERTINA TUBE COLLAPSE ANALYSIS
7.1 INTRODUCTION
7.1.1 Scope
7.2 OVERVIEW OF ENERGY DISSIPATING DEVICES
7.2.1 Definition and criteria
7.2.2 Types of energy dissipating devices
7.2.3 Tubes as energy absorbers
7.2.3.1 Lateral compression
7.2.3.2 Axisymmetric axial crumpling
7.2.3.3 Diamond-fold axial crumpling
7.2.3.4 Tube inversion
7.2.4 Methods of analysis
7.3 QUASI-STATIC CONCERTINA TUBE COLLAPSE MECHANISMS
7.3.1 Related experimental work
7.3.1.1 Experimental Programme and method
7.3.1.2 Typical experimental results
7.3.1.3 Microhardness tests
7.3.1.3.1 Equipment and procedure
7.3.1.3.2 Derivation of material strength, Y
7.3.2 Numerical model
7.3.2.1 Discretization and material
7.3.2.2 Velocity scaling
7.3.2.3 Interpretation of output
7.3.3 Results from numerical calculations and experiment
7.3.3.1 Tube geometry A: ID=19.05mm, t=1.64mm, L=50.8mm
7.3.3.2 Tube geometry B: ID=19.05mm, t=1.17mm, L=50.8mm
7.3.3.3 Tube geometry C: 0D=38.1mm, t=1.65mm, L=50.8mm
7.3.3.4 Tube geometry D: 0D=25.4mm, t=0.95mm, L=25.4mm
7.3.4 Parametric studies in numerical analyses
7.3.4.1 Influence of friction
7.3.4.2 Influence of velocity scaling
7.3.4.3 Influence of mesh refinement
7.3.4.4 Influence of element type
7.3.5 Discussion
7.4 MEDIUM VELOCITY (176m/s) TUBE IMPACT ANALYSIS
7.4.1 Description of problem
7.4.2 Numerical idealization
7.4.3 Numerical results
7.4.4 Discussion
7.5 CONCLUSIONS
7.1 INTRODUCTION
The numerical methods described i n chapter 4 are appl i ed here t o
t he anal ysi s of t ube col l apse mechani sms, wi t h a vi ew t o st udyi ng
t hei r performance as energy di ssi pat i ng devi ces.
In accordance wi t h the char act er i st i cs of the computer program,
t he at t ent i on was r e s t r i c t e d t o mechanisms whi ch coul d be st udi ed
through axi symmetri c or 2-dimensional models. In par t i cul ar , most of
t he work r el at es t o axi al col l apse of t ubes t hr ough axi symmet r i c
sequenti al f ol di ng. This mode of col l apse plays an i mport ant rol e i n
energy di ssi pat i on and i t s numerical model l i ng poses some i nt er est i ng
chal l enges. For example:
a) Large st r ai ns (wi t h values higher than 1.3) and gross deformati ons
(compl et e cr umpl i ng of t ube wal l s ) need a c ar ef ul and consi st ent
mathematical treatment (chapter 2);
b) The need t o pr ovi de c o n s t i t u t i v e l aws v a l i d f or l ar ge s t r ai ns
(chapter 6) ;
c) Cont act s wi t h t he pl at ens and between f ol ds make necessary a
general l ogi c f or i nt er act i on between cont i nua;
d) Dynamic anal ysi s i s necessary, as energy di ssi pat i on occurs mai nl y
i n impact si t uat i ons, where stress-wave propagation or i ner t i a ef f ect s
are often i mport ant ;
e) L a s t l y , a r obust numer i c al a l g o r i t h m i s necessary whi ch
accomodates t he s ol ut i on t o t he above probl ems and works wel l i n
pract i ce f or r eal , engi neeri ng scale problems (chapter 4).
7.1.1 SCOPE
This chapter rel at es t o the f ol l owi ng:
1- Fi r s t l y a general overview of energy di ssi pat i on, and the rol e of
tubes i n i t , i s pr esent ed. Thi s i s f ol l owed by a des c r i pt i on of some
167
experimental work on tube col l apse; most of t hi s work was car r i ed out
by Ghani (1982), and i t w i l l be used t o compare wi t h t he numer i cal
predi ct i ons f or axi symmetri c tube crumpl i ng t hat f ol l ow.
2- Resul t s f r om numer i cal c al c ul at i ons of q u a s i - s t a t i c axi al
col l apse are gi ven f or 4 d i f f e r e n t t ube geomet r i es ( t abl e 7.3). Over
20 computer analyses (t abl e 7.4) were performed i n order t o assess the
a p p l i c a b i l i t y of t he method, and eval uat e t he i nf l uence of var i ous
model l i ng choi ces. Opt i mal models were t hen sel ect ed f or t he f i n a l
analyses. The r esul t s are compared wi t h experimental data.
3- In the medium vel oci t y range, a f ul l - s c al e model of the impact of
a t hi n- wal l ed vessel at 176 m/s was performed. This problem t est ed the
perf ormance of t he computer code f or l ar ge scal e engi neer i ng
computati ons, i nvol vi ng several thousand degrees of freedom and over a
hundred thousand t i me- st eps. The r esul t s compared wel l wi t h avai l abl e
exper i ment al dat a. The cr ushi ng f or ce obt ai ned was f ound t o be
consi derabl y l arger than f or l ow- vel oci t y col l apse.
7. 2 OVERVIEW OF ENERGY DISSIPATING DEVICES
Energy di ssi pat i ng devices are reviewed br i ef l y i n t hi s sect i on.
The i nt ent i on i s t o i ntroduce pr act i cal pr oj ect i ons f or the numerical
appl i c at i ons present ed l a t e r . The i mpor t ance of t he var i ous tube
col l apse mechanisms i s hi ghl i ght ed.
7. 2. 1 DEFINITION AND CRITERIA
Energy di ssi pat i ng devices are mechanisms which can di ssi pat e the
ki net i c energy from the impact of c ol l i di ng bodies i n an i r r ever si bl e
manner.
El a s t i c , r ecover abl e systems are t her ef or e excl uded f rom t h i s
def i ni t i on. Usual l y, energy i s di ssi pat ed through pl ast i c work, being
conver t ed i nt o heat event ual l y. To f u l f i l l t h e i r f unc t i on wi t hout
excessive decel erat i on and consequent damage, energy absorbers must
provide a reasonably constant operati ng force over as long a stroke as
168
possi bl e. This must occur i n a cont r ol l ed ( i . e. wi t hout cat ast rophi c
f ai l ur e) and r el i abl e way.
I deal l y energy absorbers shoul d be made of low cost and wi del y
avai l abl e components, wi t h high values f or the f ol l owi ng parameters:
- Speci fi c energy, i . e. Energy di ssi pat ed / Mass of device
- Stroke ef f i ci ency, i . e. Absorbing l engt h / Total l engt h
- Energy d i s s i p a t i n g d e n s i t y , i . e . Energy di s s i pat ed / Volume of
devi ce.
7.2.2 TYPES OF ENERGY DISSIPATING DEVICES
Some of the main types of energy absorbers are summarized below,
concent r at i ng on des t r uc t i v e (one shot ) devi ces, mainly of met al l i c
mat er i al s. No attempt i s made t o provi de a comprehensive review of the
t opi c . Excel l ent revi ews have been gi ven by Johnson and Reid (1978)
and Ghani (1982).
Schemati cal l y one may cl assi f y the vari ous mechanisms according
t o t hei r mode of behaviour, as f ol l ows.
a) Ext ensi on. St eel rods or cabl es under t ensi on have been used
ext ensi vel y i n pi pe- whi p r e s t r a i n t systems (e. g. Her nal st een and
Lebl oi s , 1976). A di sadvant age i s t hat t e n s i l e def or mat i ons t end t o
become l ocal i zed i n necks and may produce overal l f ai l ur e.
b) Compressi on. For i nst ance t he cr ushi ng of l i g h t we i g h t c e l l u l a r
bodies or of sof t copper bumpers (Hernalsteen and Lebl oi s, 1976).
c) Bending. Beams and pl at es may di ssi pat e energy through r ot at i on i n
pl ast i c hi.nges. Lat er al l y crushed tubes or assemblies t hereof provide
some very usef ul devi ces (Thomas, Reid and Johnson (1976), Shr i ve,
Andrews and England (1984)).
d) Compression and bendi ng. Systems i n whi ch bendi ng modes are
produced as a r esul t of a st r uct ur al compression may be i ncl uded here.
These compri se some of t he most e f f i c i e n t energy absor ber s, such as
the axi al crumpl i ng of ci r cul ar tubes (axisymmetric or diamond modes)
and t ube i nv er s i on. Tubes wi t h non- c i r c ul ar sect i ons have been used
advantageously by Kukkola (1976). Honeycomb panels provi de also a very
popular and r el i abl e mechanism (McFarland, 1963).
e) Cycl i c bendi ng, Here one may i ncl ude systems such as t he r o l l i n g
t or us of Johnson, Reid and Singh (1975). These t end t o be associ at ed
wi t h l i ght er operati ng l oads.
7 . 2 . 3 TUBES AS ENERGY ABSORBERS
Tubes have a par t i cul ar i nt er est as energy absorbers, because of
a l l t he e x i s t i n g devi ces t hey pr esent t he wi dest range of possi bl e
uses; These i ncl ude some of t he most e f f i c i e n t and r e l i a b l e energy
di ssi pat i on mechanisms.
Tubes are si mpl e s t r uc t ur es and t her ef or e cheap and wi del y
avai l abl e. Addi t i onal l y, the basic st r uct ur e of many moving vehi cl es
and ai r c r af t i s a tube, which wi l l double up i n f unct i on as an energy
absorber i n the event of a c o l l i s i o n .
The main tube col l apse mechanisms are reviewed below. Some simple
cl assi cal anal yt i cal formulae f or col l apse l oads are i ncl uded where
per t i nent .
7.2.3.1 LATERAL COMPRESSION
Tubes crushed l at er al l y provi de a very r el i abl e energy absorber
i n whi ch t he p l a s t i c def or mat i on i s concent r at ed around l o c a l i z e d
hi nges . DeRuntz and Hodge (1963) s t u d i e d e x p e r i me n t a l l y and
t h e o r e t i c a l l y t h e i r behavi our , der i v i ng a f or mul a f or t he col l apse
load based on a r i gi d- pl as t i c , t hi n shel l model:

g
(4//F) Y t\ (7. 1)
D^l -(5/ D)*
where 6 i s t he di spl acement , t
0 a n
d
D
t he t hi ckness and Di amet er of
the tube r espect i vel y, and Y the Yi el d stress f or a per f ect l y pl ast i c
materi al model. Due t o the absence of mat eri al hardening i n (7.1) some
di scr epanci es wi t h exper i ment s were not abl e; Rei d and Reddy (1978)
have proposed more recent l y sol ut i ons i n which l i near st r ai n hardening
i s consi dered, achi evi ng bet t er r esul t s.
7. 2. 3. 2 AXISYMMETRIC AXIAL CRUMPLING
When subj ect ed t o axi al compr essi on, t ubes may buckl e i n an
axi symmet r i c f as hi on, f or mi ng concer t i na- t ype f ol ds whi ch p i l e up
sequent i al l y on top of each other (Figure 7. 1a) .
This col l apse mode const i t ut es a very ef f i c i ent energy absorber,
by reason of t he gross p l a s t i c def or mat i ons whi ch occur i n a l ar ge
proport i on of the tube. The sequenti al nature of the f ol di ng process
accounts f or a reasonably constant operat i ng f orce over a long st roke.
For gi ven t ube di mensi ons t h i s mechanism i s r epr oduci bl e and very
r e l i a b l e (Ghani, 1982).
Alexander (1960) f i r s t deri ved an approximate sol ut i on f or t hi s
mechanism f rom r i g i d - p l a s t i c t h i n shel l assumpt i ons. The average
col l apse load and f ol d l ength according t o t hi s theory are
P = 6t
0
YyDt
0
(7.2)
h = / *" ^ 0 (7.3)
Alexander's f ormul ae, however si mpl e, give sensi bl e predi ct i ons
i n gener al ; t hey are of t en r ef er r ed t o (e.g. Johnson, 1972) and
conti nue t o be used today by many workers (e.g. Hurl ey, 1983, Mamalis
and Johnson, 1983). However, eqn. (7.2) assumes a per f ect l y pl ast i c
mat er i al (no har deni ng) ; f or good pr edi c t i ons care must be t aken t o
use a val ue f or Y r epr esent at i ve of t he act ual average s t r a i n -
171
hardening of the mat er i al . Experimental and numerical studi es car r i ed
out here show ef f ect i ve pl ast i c st r ai ns accross the f ol d secti ons of
t he crumpl ed t ube between 0.3 and 1.3 ( sect i on 7.3.3). Thi s suggest s
an appr opr i at e average val ue of 0.8 s t r a i n t o be used f or obt ai ni ng
t he val ue of Y.
Numerical analyses of these mechanisms are undertaken i n secti ons
7.3 and 7.4, const i t ut i ng the core of t hi s chapter.
7.2.3.3 DIAMOND FOLD AXIAL CRUMPLING
When subj ect ed t o axi al compr essi on, another possi bl e col l apse
mechanism i s t he f or mat i on of non- axi symmet r i c f ol ds i n a diamond
pat t er n, accompanied by a change i n cross secti on of the tube ( f i gur e
7.1b). Thi s mechanism i s al so r e l i a b l e and e f f i c i e n t as an energy
absorber, although the speci f i c energy i s s l i ght l y lower than f or the
concerti na mode (Ghani, 1982).
Diamond f o l d cr umpl i ng was st udi ed by Pugsl ey and Macaulay
(1960), who proposed formulae based on r i gi d- pl as t i c anal ysi s. Further
st udy has been made by Johnson, Soden and Al - Hassani (1977) who
assumed ' i nextensi onal col l apse modes'. Thornton and Magee (1977) have
also used these devices as energy absorbers; they prepared the tubes
pr evi ousl y by making 3 c i r c u mf e r e n t i a l i ndent at i ons at one end, i n
order t o avoi d t he i n i t i a l peak l oad and t r i g g e r smoot hl y a 3- l obe
diamond pat t er n.
7.2.3.4 TUBE INVERSION
Tube i nver si on modes may be achi eved ei t her by pushi ng a t ube
a x i a l l y agai nst a r adi used di e or by pr ef or mi ng and cl ampi ng
appropri at el y one end. Inversi on may be external or i nt er nal . Pl ast i c
deformations extend t o v i r t ual l y the complete l engt h of tube, as each
secti on i s f i r s t bent, and then st rai ght ened out. An ef f i c i ent energy
absorber wi t h a nearl y f l a t load response i s thereby obtai ned.
Gui st and Marbl e (1966) have consi der ed i nver si on tubes f or
172
i mpact absor pt i on i n l andi ng of a i r c r a f t and space v ehi c l es ; t hey
proposed a si mpl e r i gi d- pl ast i c equation f or the col l apse l oad:
P = 7Tto\/8Dtcr (
7
-
4 )
Al -Hassani , Johnson and Lowe (1972) have presented a det ai l ed study of
these devi ces, developing f ur t her t he anal ysi s t o i ncl ude power - l aw
s t r ai n- har deni ng, achi evi ng t hus a bet t er match t o exper i ment al
r esul t s.
7.3 QUASI-STATIC CONCERTINA TUBE COLLAPSE ANALYSIS
7. 3. 1 RELATED EXPERIMENTAL WORK
A des c r i pt i on i s gi ven here of t he exper i ment al work on quas i -
st at i c col l apse of Aluminium tubes t o which reference wi l l be made i n
t he f o l l o wi n g sect i ons, where t he exper i ment al r es ul t s w i l l be
compared wi t h numer i cal pr edi c t i ons . Most of t he exper i ment al data
menti oned here are der i ved f rom work done by Ghani (1982) who
undert ook an ext ensi ve programme of st udy and c l a s s i f i c a t i o n of
col l apse modes, spanni ng 189 d i f f e r e n t t ube geomet r i es. A good
synopsis of these t est s and t hei r r esul t s has been given by Andrews,
England and Ghani (1983).
In addi t i on t o Ghani's work, 5 ext ra tubes have been crumpled f or
t hi s i nvest i gat i on i n order t o pr ovi de addi t i onal exper i ment al data
whi ch were needed (Tabl e 7.1). Thi s was done wi t h t he same mat er i al
and under i dent i cal condi t i ons as Ghani (1982).
Fi n a l l y , i n or der t o pr ovi de data f or t he st r ess d i s t r i b u t i o n s
around the f ol d hinges, some crumpled tubes were cut and microhardness
t es t s perf ormed on t he wal l sect i ons. The mi crohardness data were
cal i br at ed and cor r el at ed t o the mat eri al st rengt h Y.
173
Geometry
t o(mm)
1.27
1.28
1.67
1.65
1.67
L(mm)
50.8
88.9
50.8
50.8
88.9
013 (mm)
38. 1
38. 1
38. 1
38. 1
38. 1
Col l apse mechanism
Concerti na, 3 1/2 f ol ds
Concerti na, 6 f ol ds
Concerti na, 3 f ol ds
Concert i na, 3 f ol ds
Concert i na, 5 f ol ds
Table 7. 1: Addi t i onal t est s f or axi al col l apse of tubes
(HT30 annealed aluminium)
7.3.1.1 EXPERIMENTAL PROGRAMME AND METHOD
The t est s were performed on HT30 (BS1471) Aluminium al l oy t ubi ng.
Thi s mat er i al has i dent i c al composi t i on and s i mi l a r mechani cal
propert i es as the HE30 al l oy used f or the bar tensi on t est s reported
i n chapt er 6. The as- r ecei ved mechani cal pr oper t i es of HT30 are as
f ol l ow:
0.1% proof stress
densi ty
Young's modulus, E
Poisson's ratio
Ultimate tensile strength
Elongation on 2 in.
228 -243 MPa
2700 Kg/m
3
67000 MPa
0.3
304 MPa
7 - 9%
The sl i ght var i at i on i n mechanical propert i es from those of HE30
i s due t o the di f f er ent col d- f or mi ng processes: HT30 tubes are formed
by drawing, whi l e HE30 bars are ext ruded.
The specimens were annealed pr i or t o t est i ng at a temperature of
350o c f or t i mes r angi ng between 3 and 7 hour s, dependi ng on wal l
t hi ckness; al l specimens were oven cool ed.
The t ubes were crumpl ed under st r oke cont r ol usi ng an I nst r on
Model 1195 pr ess, wi t h a 100 kN l oad c e l l . The few speci mens whi ch
r e q u i r e d l a r g e r l oads wer e t e s t e d wi t h a Losenhausen pr es s .
Deformation rat e was approxi matel y 5 mm/min. Undeformed tube lengths
ranged between 6.35 and 533.4 mm.
7. 3. 1. 2 TYPICAL EXPERIMENTAL RESULTS
The col l apse modes obtai ned by Ghani (1982) may be grouped i nt o
the f ol l owi ng broad cat egor i es:
- Concertina (49 t ubes); Axi symmetri c, sequenti al f ol di ng st ar t i ng at
one end of tube
- Pi amond (5 t u b e s ) ; Non- axi sy mmet r i c but sequent i al f o l d i n g
accompanying a change i n cross sect i on shape
- Mi xed concer t i na- di amond (44 t ubes) ; s t a r t i n g as concer t i na and
changing over t o the diamond mode af t er some f ol ds
- Euler (31 t ubes) ; buckl i ng of tube as a st r ut
- Vari ous cr ushi ng modes (60 t ubes) ; si mul t aneous col l apse of t he
whole tube due t o crushi ng, t i l t i n g , 2-lobe diamond, or simultaneous
concerti na f ai l u r e .
Fi gur e 7.1 shows a s el ec t i on of t e s t r e s u l t s , f eat ur i ng an
example of each of t he col l apse modes ment i oned above. Accor di ng t o
t he l engt h/ di amet er and wal l t hi ckness/ di amet er r at i os of t he
undeformed t ubes, a c l a s s i f i c a t i o n char t was drawn up ( f i gur e 7.2).
This diagram i ndi cat es di st i nct regions i n which a cer t ai n col l apse
mode can be expected.
Curves i n d i c a t i n g t he v a r i a t i o n of ax i al l oad wi t h axi al
shorteni ng of the tubes were pl ot t ed di r ect l y f or al l the t est s. From
t he anal ysi s of t hese cur ves, t he energy absorpti on char act er i st i cs
were obtained f or the vari ous modes of col l apse.
Usi ng t he col l apse mode r egi ons def i ned i n f i gur e 7.2, t ube
geomet r i es can be sel ect ed f or whi ch one may expect concer t i na
sequent i al f a i l u r e . The 5 ext r a tubes ( t abl e 7.1) t est ed f or t h i s
i nv es t i gat i on were sel ect ed wi t h such c h a r a c t e r i s t i c s ; t hey a l l
col l apsed i n concer t i na mode as expect ed. A t y p i c a l r es ul t t oget her
wi t h the corresponding load-compression curve i s shown i n f i gur e 7.3.
FIGURE 7.1:COLLAPSE MODES FOR ALUMINIUM TUBES UNDER AXIAL .COMPRESSION (GHAN'I, 1952)
A - Concertina, B - 3 lobe diarrmd, C - Euler buckling, D - Mixed Concertina-
Diamnd, E - Crushing, F - 2 lobe dicrond crushing, G - sinultaneous concertina
9. 00
. 00
7. 00
6.00
5.00
A. 00 -
3.00
2. 00
1.00
0. 00
-
-
-
-
-
1
ffl \
a
a
cm
B
m
ffl
ffl O
ffl
o
1
V
a
G9
*B
B
^B
a
o J>
Mixed
0 m ode

O O

o
0 /
/
0 <S>


O
-r-"~
u
a
BW
/ G O
Z
0
o
o
0

1
V

o
i
V
V
V
^
f\
Q
o \
o \
o
1
V
c onc er t i na


0 <>
0

1
1?

Q
O



_ * -
1
1
V
V
V
B |

B


--
1
Eul er
V
V
~^~~~^*
x
2-lobe
1
V
~^
N -
^di am on d/ ^
axi s ym m et ri c
crus hi n g '
* * l
1


*

*
i
1
1
1
- C
I

#

i
V
V
V
# _

A xi s ym m

4
- r"

cr us n
_--

" "
1
V

o
ffl
B



etri
n g
1
1 1 1
KEY
EULER 1 AN/ STRUT FAI LURE
CONCERT I NA
CONCERTINA . 2- LOBE OI AWJMT
CONCERTINA . 3 LOBE DUMP*
CONCERTINA . 2-L08E AW 5-L0BE 0IAH0W
S-L0BE D1ANCM
CONCERTINA (AXISTItCIRICI CRUSHING
2-L0BE 0IANDW CRUSHING
CRUSHING TILTING OF I L A*T5
V
V
V
& *-
c mul t i pl e
( ol ds
*
single
* f ol ds
^ .
* si mpl e
c ompr es s i on
i i i
i
V
V
V
+

"
"V '

i
_
-
-
_
0.00 .02 .04 .06 .08 .10 .12 .16 .20 .22 .24 .26
THICKNESS/DIAMETER RATIO . T/D
FIGURE 7.2: CLASSIFICATION CHART OF AXIAL COLLAPSE MODES IN CIRCULAR TUBES
(FROM GHANI, 19S2)
17
m v*i
cm.
O
A) view of crumpled tube
CO
o
S t r oke (mm)
B) LOAD - COMPRESSION CURVE
Figure 7. 3: Typical r esul t s of quasi - st at i c tube crumpl i ng experiment
wi t h HT30 annealed al umi ni um a l l o y ( Concer t i na mode of
col lapse)
177
7.3.1.3 MICROHARDNESS TESTS
L o n g i t u d i n a l s e c t i o n s of some t ubes have been c ut and
microhardness t est s performed on them, i n order t o deri ve st r ai n and
st r ess d i s t r i b u t i o n s whi ch coul d be checked agai nst numer i cal
pr edi ct i ons.
7. 3. 1. 3. 1 Equipment and procedure
A Rei cher t mi crohardness t e s t e r was used f or t hese t e s t s . I t
consi st s i n essence of a 1360 apex angl e square pyrammi d diamond
i ndent or, which can be pressed agai nst the specimen wi t h a speci f i ed
l oad, and t hen moved out of t he way f or measuri ng wi t h a mi croscope
lens the si ze of the i mpressi on.
The microhardness H
m
- j
s
defi ned as the mean pressure exerted on
the surface of the i ndent or. Simple geometri cal consi derat i ons permi t
the expression of H
m
(Kg/mm2) as a f unct i on of the appl i ed load P (g)
and the i ndent at i on di agonal , d (mi crons):
H
m = 1854.4 P/d (7. 5)
Although there i s geometri cal si mi l i t ude between i ndent at i ons of
d i f f e r e n t si zes, and cont r ar y t o what happens f or (macro)hardness
(VPN) at hi gher l oads, t he mi crohardness i s not i ndependent of t he
l oad appl i ed. Hence an a r b i t r a r y const ant l oad of 99 g was sel ect ed
f or al l the measurements.
The sur f aces of t he specimens were l apped p r i o r t o t e s t i n g , i n
order t o obt ai n smoot h, f l a t sur f aces i n whi ch t he mi cr oscopi c
i ndent at i on measurements coul d be taken wi t h suf f i ci ent pr eci si on.
Three crumpl ed t ube specimens were sel ect ed f or t e s t i n g . The
or i gi nal dimensions of these tubes were
- ID = 19.05mm t o
= 1

64r nm
> L = 50.8mm (geometry A, sect . 7. 3. 3. 1);
- ID = 19.05mm, t
Q =
i . i 7mm, [_ = 50.8mm (geometry B, sect . 7. 3. 3. 2);
= 1.65mm, L = 50.8mm (geometry C, sect . 7.3.3.3).
In each case, the microhardness measurements were done only i n an
area around one of t he f o l d s . The si ze of t he i ndent at i on di agonal s
vari ed between 50 and 70 mi crons; a separati on of 200 t o 300 microns
was l e f t between di f f er ent i ndent at i ons i n order t o el i mi nat e cr oss-
i nf l uences. Typi cal r es ul t s f or one of t he above t ubes are shown i n
f i gur e 7.4.
7. 3. 1. 3. 2 Derivation of m aterial stren gth, Y
The strai n-hardened f l ow st ress Y i s r el at ed t o the hardness. For
a per f ect l y pl ast i c mat eri al t hi s r el at i onshi p i s l i near ( Hi l l (1950),
p260):
H = CY (7. 6)
where C i s a constant f act or depending on the geometry of the i ndent or
and t he angl e of f r i c t i o n , whi ch nor mal l y l i e s between 2.5 and 3.0.
For a smooth, f l a t di e Pr andt l ' s sol ut i on ( sect . 5. 4. 1) yi el ds
C = ( 2+7T) / f T= 2.97
In the event of real mat eri al s wi t h st r ai n- har deni ng, eqn. (7.6)
i s not t r ue i n a s t r i c t sense. The p l a s t i c d i s t o r t i o n around t he
i ndent or i s non- uni f or m and d i f f e r e n t el ement s harden by d i f f e r e n t
ammounts. Mean equi val ent val ues of Y need t o be used i n eqn. ( 7. 6) ;
d e t e r mi n a t i o n of t hese mean val ues i n v o l v e s n e c e s s a r i l y an
approximation and a cer t ai n degree of er r or . Nevertheless f or metals
which are heavi l y pr e- st r ai ned the er r or wi l l be smal l , as:
a) the slope of hardening has di mi ni shed consi derabl y;
b) t he i ndent at i on s t r ai ns are smal l wi t h r e l a t i o n t o t he e x i s t i n g
st r ai ns.
The area around t he f ol ds i n whi ch mi crohardness t es t s were
per f or med has ver y hi gh s t r a i n s , i n gener al upwar ds of 0. 5.
Cal i br at i on data f or the der i vat i on of a l i near r el at i onshi p between
1/9
B) fllCROHARDNESS CONTOUR MAP ( K G / M M 2)
FIGURE 7-1: TYPICAL RESULTS OF fllCROHARDNESS TESTS ON CRUMPLED TUBE SECTIOM
(0D=38-lmn) 1b=l-67mm, L=50-8m^ (TUBE GEOMETRY C)
H
m
and Y were obt ai ned f rom mi crohardness t es t s i n t r ansver sal
sect i ons of an anneal ed HE30 necked t ensi on bars ( chapt er 6). Thi s
r el at i onshi p i s val i d as wel l f or HT30, which af t er annealing has the
same mechanical propert i es as HT30 (see secti on 7.3.2.1).
For each sect i on of t he necked bar t he s t r a i n can be computed
from eqn. (6.8); from the st r ess- st r ai n r el at i onshi p of the mat er i al ,
eqn. (6. 20), a val ue f or t he st r engt h Y was f ound ( t abl e 7.2). A
l i near r egr essi on was t hen per f or med t hr ough t he (H
m Y
) data poi nt s
( f i gur e 7.5). For the l east squares f i t the poi nt corresponding t o the
unstrai ned mat eri al has been disregarded as the region of i nt er est i s
at high st r ai ns. The r el at i onshi p obt ai ned,
Y = -38.5 + 0.35H
m ( MP a ) ( 7 # 7 )
was t hen used t o der i ve cont our maps of t he st r ai n- har dened f l ow
st r ess Y ( f i gur es 7.13, 7.18 and 7.23), whi ch coul d be compared
di r ect l y t o those obtained from the computational model.
Section
0
1
2
3
4
5
6
Microhardness(*)
H
m(MPa)
384
493
522
509
540
565
638
Original
Diameter
Do(mm)
31.7
16.2
16.2
16.1
16.1
16.0
16.0
Deformed
Diameter
D(mra)
31.7
15.0
14.8
14.3
14.0
12.4
9.2
Strain
e=-2Ln(D/D
0
)
0.00
0.15
0.18
0.24
0.28
0.51
1.11
Yield
Strength
Y=181.7
e
.159
59.4(**)
134.4
138.3
144.8
148.4
163.2
184.7
(*) Mi crohardness val ues averaged f r om a minimum of 6 measurements
along two perpendi cul ar radiuses of each sect i on
(**) Strength f or unstrai ned mat eri al corresponds t o i n i t i a l y i el d
Tabl e 7.2: Mi cr ohar dness, s t r a i n and st r ess val ues on t r ansver sal
sect i ons of necked HE30 t ensi on bar f or c a l i b r a t i o n of
Mi crohardness-strength r el at i onshi p
200
IBO -
160 -
140 -
120 -
100 -
80
1
-
-
- o
1
1 1
G ) ^ ^
_ ^ ^ o
i i
I I I -~ p,
O^""^
HE30 ANNEALED ALUMINIUM
o (Hm.Y) Data points
T= -oU.L) + U.oL)llm
i i i
400 650 450 500 550 600
MIcrohardneee, Hm (MPs)
Figure 7.5, YIELD STRENGTH - MICROHARDNESS CORRELATION
/7^77~77 77777V 777 77777
v=0
77~7WyV 7~77~
ur e 7 . 6 : For ces i n v o l v e d i n dynami c c o n c e r t i n a f o l d i n g ;
P= t o t a l f o r c e , P ^ c r u s h i n g f o r c e , v = c r u s h i n g v e l o c i t y
7.3.2 NUMERICAL MODEL
The axi al col l apse of t ubes was si mul at ed usi ng t he 2-D
numer i cal code descr i bed i n chapt er 4. Axi symmet r i c cont i nuum MTQC
el ement s ( sect . 4.2.3) were used under an e x p l i c i t t i me- mar chi ng
procedure.
The t ubes were crushed between t wo pl at ens ; t he bot t om one was
f i x e d , whi l e t he t op pl at en moved downwards, causi ng t he t ube t o
col l apse. The pl at ens, as t he r est of t he t ube, were model l ed wi t h
axi symmetric continuum elements. I nt er act i on between pl atens and tube
as wel l as between t ube f ol ds t hemsel ves was consi der ed t hr ough
numerical contacts ( sect i on 4.8).
In or der f or t he c al c ul at i ons t o be meani ngf ul , i t must be
checked t hat axi symmet r i c def or mat i ons are t o be expect ed f or t he
p a r t i c u l a r t ube geomet ry, e.g. by l oc at i ng t he appr opr i at e poi nt i n
the col l apse cl assi f i cat i on chart ( f i gur e 7.2). Non-axisymmetric modes
of col l apse such as t he diamond mode cannot be s i mul at ed, as axi al
symmetry of def or mat i on i s imposed by t he model . The anal yses
performed here al l corresponded t o experi ment al l y observed concert i na
f ai l ur es .
Apart from the axi al symmetry, no f ur t her numerical const r ai nt s
were imposed upon t he model , bei ng compl et el y f r ee t o devel op i t s
pref erred form of col l apse (e.g. si mul taneous, sequent i al ) as wel l as
t he number and l engt h of f o l d s . No geomet r i cal i mper f ect i ons were
i ntroduced t o t r i gger of f buckl i ng modes. ' Nat ural ' i mperf ect i ons were
provided by the random numerical roundoff er r or s.
The axi al col l apse of tubes i s a f a i r l y r el i abl e and repeatabl e
mechanism (e.g. Ghani , 1982), i n d i c a t i n g l i t t l e s e n s i t i v i t y t o
i mperf ect i ons. In the numerical model, the i mperf ect i ons depend on the
preci si on of the f l oat i ng poi nt ar i t hmet i c. A check was done, running
the same tube col l apse model (TUBE7, see t abl e 7.4), on a CRAY-IS (64-
bi t f l oat i ng poi nt words) and on a VAX 11/785 ( 32- bi t f l oat i ng poi nt
words). Results obtained were subst ant i al l y i dent i cal . The devi at i on
from the energy balance condi t i on was l arger on the VAX (0.15% error
versus 0.005% er r or on t he CRAY), but s t i l l wi t h i n very accept abl e
l i mi t s .
7. 3. 2. 1 DISCRETIZATION AND MATERIAL
Di scr et i zat i ons wi t h varyi ng fi neness were used f or the analyses.
The meshes employed may be grouped i nt o the f ol l owi ng cat egor i es:
" Coarse mesh, 3 q u a d r i l a t e r a l MTQC elements through the t hi ckness,
approximate aspect r at i o 2:1 (l argest dimension ax i al l y ) ;
" Medium mesh, 3 quadr i l at er al MTQC elements through the t hi ckness,
approximate aspect r at i o 1: 1;
- Fi ne mesh, 4 quadr i l at er al MTQC elements through the t hi ckness,
approximate aspect r at i o 1:1.
The pl at ens were di s c r et i z ed wi t h t wo CST el ement s each i n a l l
cases. They were assumed i n f i n i t e l y r i g i d , i .e. no deformati ons were
al l owed.
Symmet ri c cont act s ( sect . 4.8) were used f or t he i n t e r a c t i o n
between opposi t e wal l s i n t he t ube f o l d s . Each i n t e r a c t i n g node i s
i ncl uded i n t wo si mul t aneous numer i cal c ont ac t s : f i r s t l y as an
i nt r uder on t he opposi t e si de, and secondl y as par t of a t ar get si de
being penetrated by the node opposi te. This ensures f u l l y symmetri cal
behaviour i n the i nt er f ace al gori t hms.
For t he cont act s between t ube and pl at ens onl y t he t ube nodes
were consi der ed as i n t r u d e r s ; t he l ack of i nt er medi at e nodes i n t he
pl at ens woul d have pr event ed ot her wi se. However, t h i s i s not
undesi rabl e, as the contact between deformable tube and r i gi d pl aten
i s nat ur al l y asymmetric.
For the mat eri al model, an el ast i c- pl ast i c i sot r opi c Von Mises
i deal i zat i on was used. El ast i c paramet ers were
El ast i c modulus, E 67000 MPa
Poisson' s r at i o v 0.3
Strai n-Hardeni ng was considered wi t h the power law obtained i n chapter
6 (eqn. 6. 20):
Y = 181.7 e
u
-
i o y
(MPa)
The above hardening law was deri ved f or HE30 annealed Aluminium bars.
HE30 and HT30 mat eri al s are i dent i cal aluminium al l oys, except f or the
f act t hat HE30 comes i n bars (extruded), whereas HT30 comes i n tubes
(drawn). Af t er bei ng anneal ed bot h mat er i al s can be assumed t o have
the same mechanical behaviour.
In ef f ect , di f f erences due t o col d-worki ng may be banished i n the
anneal i ng pr ocess. Accor di ng t o Co t t r e l l (1975), " wor k- har deni ng i s
caused by t he mut ual o b s t r u c t i o n of d i s l o c a t i o n s g l i d i n g on
i nt er act i ng systems". The anneal i ng t r eat ment produces r ecover y and
r ecr yst al l i zat i on processes. The recovery al l ows di sl ocat i ons t o move
out of the sl i p planes, di sent angl i ng and t i dyi ng- up the col d-worked
st r uct ur e, whi l e the r ecr yst al l i zat i on replaces the col d-worked grai ns
by a new set of more perf ect gr ai ns, gi vi ng complete sof t eni ng. As on
the other hand the di f f erences due t o col d-worki ng were only mi nor, i t
i s j u s t i f i e d t o assume the const i t ut i ve laws deri ved f or annealed HE30
val i d f or HT30 tubes as wel l .
7. 3. 2. 2 VELOCITY SCALING (see al so sect i on 4. 1. 1)
Ex p l i c i t t i me- mar chi ng procedures by t h e i r nat ur e can onl y
perform dynamic analyses. Quasi - st at i c or slow l oadi ng problems coul d
be sol ved i n r eal t i me , but usual l y t h i s becomes t oo expensi ve. For
reasons of numerical s t a b i l i t y the maximum t i me- st ep i s l i mi t ed by the
Courant c r i t er i on, which i s independent of the vel oci t y of l oadi ng. In
quas i - s t at i c anal yses such as t hese t he c a l c u l a t i o n s need t o be
speeded up, which may be done by a r b i t r a r i l y i ncreasi ng the vel oci t y
of def or mat i on i n t he numer i cal model ; t h i s i s c al l ed " v el oc i t y
scal i ng" .
For these cal cul at i ons the crushi ng vel oci t y employed was 20 m/s,
r e p r e s e n t i n g an i nc r eas e of 5 or der s of magni t ude f r om t he
exper i ment al v e l o c i t y . I t w i l l be seen ( sect i on 7.3.4.2) t hat t he
overshoot s and d i s t o r t i o n i nt r oduced by t h i s amount of v e l o c i t y
s c a l i n g wer e s ma l l . As t h e r e was no s t r a i n - r a t e har deni ng i n t he
ma t e r i a l , t he onl y a d d i t i o n a l s af eguar d needed was t o ensur e t h a t
dynami c e f f e c t s c o n t i n u e d t o pl ay a s mal l r o l e i n t he mechani cal
def or mat i on process. Some c r i t e r i a t o quant i f y t h i s r ol e were gi ven i n
s e c t i o n 6. 5. 1.
The dynamic f or ces i nvol ved i n t he concer t i na cr umpl i ng may be
assessed wi t h t he f o l l o wi n g si mpl e consi der at i ons. Fi gur e 7.6 shows a
wal l sect i on i n a t ube whi ch i s bei ng cr umpl ed wi t h an axi al v el oc i t y
v. The p o r t i o n of t he t ube a l r e a d y c r umpl ed ( bel ow A) i s s t a t i o n a r y ,
t he undeformed por t i on above C moves wi t h v e l o c i t y v, and t he f o l d i n g
p o r t i o n bet ween A and C i s bei ng d e c e l e r a t e d f r om v t o 0. Thi s
d e c e l e r a t i o n occur s i n a t i me ( h - 2 t
0
) /
v > h b e i n g t n e f o
i
d
l e n g t h and
t o
t he wa l l t h i c k n e s s ; t he aver age i n e r t i a f o r c e necessar y i s , by
v i r t u e of Newton's second l aw,
P-i = 7 T Dt
0
p h v
2
/ ( h - 2 t
0
) ( 7. 8)
where D i s t he mean t ube Di amet er and p i s t he mass densi t y. Hence t he
average dynamic f or ces grow wi t h t he square of t he cr ushi ng v e l o c i t y ,
a l l ot her f act or s bei ng const ant f or a p a r t i c u l a r t ube.
The t o t a l f or ce F exer t ed at t he bot t om sur f ace equal s t he f or ce
necessar y f o r c r u mp l i n g f e l t at D, P
Cj p l u s t h a t
necessary f o r
dec el er at i ng t he f o l d ABC, P i
:
p = p
c
+ P
i
= P
c
+
7TDt
0
pv
2
h/ ( h- 2t
0
) ( 7. 9)
The def or mat i on v e l o c i t y v used i n t he c al c ul at i ons was chosen so
as t o produce accept abl y smal l dynamic d i s t o r t i o n s (Pi < 5%Pc'*
T n e s e
dynamic ef f ect s are q u a n t i f i e d i n sect i on 7.3.4.2, where t he i nf l uence
of t he v e l o c i t y s c al i ng i s di scussed.
7. 3. 2. 3 INTERPRETATION OF OUTPUT
Load- c ompr es s i on cur ves o b t a i n e d f r om numer i c al a n a l y s i s
needed some p o s t - p r o c e s s i n g ( av er agi ng and f i l t e r i n g ) i n or der t o
f a c i l i t a t e t h e i r i n t e r p r e t a t i o n .
186
F i r s t l y , a r epr esent at i ve val ue of t he l oad was taken as t he
average between t he l oads measured at t he t op and bot t om pl at ens at
each i nst ant . In a quasi - st at i c si t uat i on the di f f er ence between both
measurements would be negl i gi bl e. Here, due t o the vel oci t y scal i ng,
t her e were f i n i t e , al t hough s ma l l , di f f er ences between t he t wo end
l oads.
Secondl y, hi gh- f r equency o s c i l l a t i o n s i n t he l oad h i s t o r i e s
ar i si ng from stress-wave propagation between the two extremes of the
specimen were f i l t e r e d out , usi ng a centred moving average f i l t e r (see
bel ow) . Agai n i n a t r ue q u a s i - s t a t i c model t h i s woul d not be
necessary; f or each load i ncrement, the new displacements are those
t hat f u l f i l l st at i c equi l i br i um. I t i s assumed t hat the stress waves
have t r a v e l l e d back and f o r t h al ong t he specimen many t i mes ,
r edi st r i but i ng stresses, and event ual l y damped out.
The v el oc i t y of def or mat i on has been gr eat l y i ncr eased her e,
whi l e the i nf ormat i on (stress waves) conti nues t o be t r ansmi t t ed at
the same speed. On the other hand, t here are steep changes i n load due
t o t he buckl i ng i n s t a b i l i t i e s and r api d v a r i a t i o n s i n boundary
condi t i ons ( cont act - i mpact ) . Whereas i n t he r eal - sl ow l oaded -
speci men t her e i s s t i l l t i me f or numerous st r ess waves t o t r ans mi t
these changes smoothly enough, here they are t r ansmi t t ed more abrupt l y
as f i n i t e amplitude dynamic pul ses, occasi oni ng vi si bl e osci l l at i ons
i n t he l oad cur ves. As l ong as t he ampl i t ude of t hese o s c i l l a t i o n s
remains small r el at i ve t o the t ot al l oads, t hei r di st or t i ve i nf l uence
i s not i mport ant , and the underl yi ng t rend i n the load curves may be
recovered by f i l t e r i n g them out.
High frequen cy f i l t e r (cen tred moving average)
Centred moving average t echni ques are si mpl e and e f f e c t i v e f or
high frequency f i l t e r i n g . From a given t i me seri es x^
t
-j=i t o N, wi t h
constant t i me i nt er val At, a new f i l t e r e d t i me seri es i s defi ned as
i+M
Yi = 1 / xj M<i<N-M ( 7 . 1 0 )
2M+1 i-M
This f i l t e r removes frequenci es higher than l / (2MAt ). The values
18
of M were chosen so as t o remove o s c i l l a t i o n s occasi oned by s t r e s s -
wave propagati on.
7.3.3 RESULTS FROM NUMERICAL CALCULATIONS AND EXPERIMENT
Numer i cal r e s u l t s ar e pr es ent ed her e and compar ed wi t h
experi ment, f or the axi al col l apse of 4 di f f er ent tube geometri es:
- Tube geometry A, ID=19.05mm, t
0
=1.64mm, L=50.8mm;
- Tube geometry 6, ID=19.05mm, t
0
=1.17mm, L=50.8mm;
- Tube geometry C, 0D=38.10mm, t
0
=1.65mm, L=50.8mm;
- Tube geometry D, 0D=25.40mm, t
o
=0.95mm, L=25.4mm.
Other a u x i l i a r y c al c ul at i ons were necessary f i r s t i n or der t o
t es t t he i nf l uence of var i ous paramet ers and choi ces:
- velocity of deformation;
- mesh refinement;
- friction;
- element type.
Resul t s f rom t hese a u x i l i a r y c al c ul at i ons are di scussed i n sect i on
7.3.4. In t o t a l over 20 c al c ul at i ons of t he compl et e t ube cr umpl i ng
process f or 6 di f f er ent tube geometri-es (t abl e 7.4) were car r i ed out
successf ul l y.
Mesh r ef i nement was found t o have an i mpor t ant ef f ec t upon t he
r es ul t s ( sect i on 7.3.4.3). However, due t o t he hi gh comput er cost s
i nvol ved i n these analyses, some of the meshes used coul d not have the
maximum desi rabl e fi neness. Two of the models presented below have a
" f i ne" mesh, whi l e the other two have "medium" meshes.
The numer i cal pr edi c t i ons are pr esent ed t oget her wi t h t he
cor r espondi ng exper i ment al r e s u l t s . The t es t s were descr i bed i n
sect i on 7. 3. 1.
The computer used f or t he numer i cal work was a CRAY-IS of t he
Uni versi t y of London. Central Processing Uni t (CPU) ti mes quoted ref er
t o t hi s machine.
7. 3. 3. 1 TUBE GEOMETRY A: ID=19.05mm, to =1.64mm, L=50.8mm
A " f i n e " mesh was used f or t h i s anal y s i s , compr i si ng 633 nodes
and 996 t r i a n g u l a r c e l l s ( f i g . 7.7.a). The c a l c u l a t i o n proceeded f or
95789 t i me- s t eps , up t o an axi al compressi on of 37.5mm, f or mi ng 3
concerti na f ol ds: f i r s t l y at the bottom end, then at the t op, f i n a l l y
i n t he mi ddl e ( f i g . 7.7b). CPU t i me f or t h i s c a l c u l a t i o n was 8641
seconds.
In the experi ment, the tube col l apsed f ormi ng 3 concert i na f ol ds
as we l l , wi t h a shape very s i mi l a r t o t he numer i cal p r e d i c t i o n
( f i gur es 7.8, 7.9). Fi gur e 7.10 cont ai ns t he l oad- compr essi on curves
( numer i cal and exper i ment al ) , whi ch show a r emar kabl e agreement .
Average col l apse l oads were 12.2 kN f or t he c a l c u l a t i o n s and 11.9 kN
f or t he exper i ment . The successi on of peaks and v al l ey s i n t he l oad
curves i s due t o t he al t er nancy between axi al compressi on ( s t i f f e r
peaks) and bending (less s t i f f val l eys) mechanisms. The snapshot views
of the mesh at di f f er ent poi nts i ncl uded i n f i gur e 7.10 conf i rm t hi s
expl anat i on, pr ovi di ng a det ai l ed des c r i pt i on of a phenomenon
pr evi ousl y observed exper i ment al l y i n a f or cef ul l y more qual i t at i ve
manner (Ghani, 1982).
The numerical load-compression curve i n f i gur e 7.10 was obtained
by averagi ng and f i l t e r i n g t he raw data as expl ai ned i n sect i on
7.3.2.3. To show t he e f f e c t of t h i s a v e r a g i n g - f i l t e r i n g pr ocess,
i ndi vi dual load curves obtai ned di r ect l y at the top and bottom pl atens
are presented i n f i gur e 7.11. These curves show some small t r ansi ent
imbalances due t o the vel oci t y s c al i ng. The under l yi ng q u a s i - s t a t i c
t r end i s seen more c l e a r l y when t h i s noi se i s f i l t e r e d out ( f i g .
7. 10).
The vari ous energy components were monitored (sect. 4.10) duri ng
t he c al c ul at i ons ( f i g . 7.12). t he energy of def or mat i on i n f i g . 7.12
compri ses p l a s t i c work, r ecover abl e e l a s t i c energy, and energy
di ssi pat ed through f r i c t i o n (some 1.4% of the pl ast i c work at the end
of the anal ysi s). In order t o provide a check f or the s t abi l i t y of the
189
6
6
CO
ID--19. 05mm
HT30 ANNEALED ALUMINIUM
TUBE
:-l
tosl.64mm
Ri gi d Platens
b) mesh a f t e r 37.5mm a x i a l c ompr es s i on
gur e 7 . 7 : Ca l c u l a t i o n s f o r a x i a l c o l l a p s e o f t ube
I D=l 9. 05mm, t =l . 64mm, L=50.8mm ( t u b e geomet r y A)
A X I A L C O L L A P S E
I D =1 9 . 05 m m , t
0
= 1 . 6 4m m , L = 5 0. 8 m m T U B E
- . ^ * *
1
^ v * *
F i g u r e 7 . 8 : V i e w o f c r u m p l e d t u b e ( g e o m e t r y A )
A X I A L C O L L A P S E
I D =1 9 . 05 m m , t
0
= i . 6 4m m , L = 5 0. 8 m m T U B E
,i n.
J
...L , J i. J .1
F i g u r e 7.9: S e c t i o n o f c r u m p l e d t u b e ( g e o m e t r y A )
JSZ
CD
O
20.00
18.00
16.00
AXIAL COLLAPSE OF TUBE
ID=l 9.05mm, t =l.64mm, L=50.8mm
o
(Gecmetry A)
EXPERIMEN
CALCULATI
o.oo
o.oo 5.00 10.00 15.00 20.00
St r o k e (mm)
FIGURE 7.10: CALCULATIONS FOR AXIAL COLLAPSE OF TUBE; LOAD - COMPRESSION CURVE
20.00
CD
o
Bot t om p l a t e n
Top p Lat en
A x i a l c o l l apse
0 5 mm, t = 1 . 6' 4mm, L = 5 0 . 8 mm T u b e
15 20
S t r o k e (mm)
Fi gur e 7 . 1 1 . INDIVIDUAL LOAD CURVES FOR NUMERICAL MODEL AT BOTH PLATENS
500
450
400
350
D
O
~ >
"
>^
cn
0)
c
LU
300
250
200
150
100
50
Tube geometry A
A x i a l co I l a p s e
I D=19. 05mm, t
0
=1 . 64mm, L = 50. 8mm Tube
E x t e r n a l wo r k
T o t a l d e f o r m a t i o n e n e r g y
P l a s t i c wor k
K i n e t i c e n e r g y
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80
Ti me (msec)
Figure 7.12s ENERGY HISTORIES FOR NUMERICAL MODEL
comput at i ons, t he work done by t he ext er nal f or ces was computed
independently and compared t o the sum of the deformati on and ki net i c
energies. Wi t hi n the preci si on of the f i gur es (5 di gi t s) no di f f er ence
was perceived between both val ues, i ndi cat i ng absence of i n s t a b i l i t i e s
and soundness of numerical i nt egr at i on al gori t hms.
Contour maps of the Von Mises f l ow stress (Y), from the numerical
model and experi ments, are presented i n f i gur e 7.13. The experi mental
data were der i ved f r om mi crohardness t e s t s , as l a i d out i n sect i on
7.3.1.3. Si mi l ar pat t er ns and l ev el s of st r ess can be observed f or
both cases, although somewhat steeper gradi ents and a wi der range of
values i s appreci abl e f or the experiment. This i s due on the one hand
t o the higher curvature f or the experi mental f ol d sel ect ed, and on the
other hand t o the r el at i ve coarseness of the numerical mesh. Agreement
f or a l l t hese l ocal ef f ec t s woul d need an i nor di nat e degree of mesh
r ef i nement . Consi der i ng t he unc er t ai nt i es of t he mi cr ohar dness
measurements and of t h e i r c o r r e l a t i o n t o t he mat er i al s t r engt h, Y,
t hi s l evel of agreement was deemed very good.
7. 3. 3. 2 TUBE GEOMETRY B: ID=19.05mm, t
0
=1.17mm, L=50.8mm
Thi s probl em was d i s c r e t i z e d wi t h a mesh of medium f i neness,
compr i si ng 532 nodes and 784 t r i a n g u l a r c e l l s ( f i gur e 7.14a). 99503
computational t i me-st eps were needed f or an axi al compression of 40
mm. 3 f u l l concerti na f ol ds plus a f our t h i ncompl ete f ol d were formed
( f i gur e 7.14b). The or der of f or mat i on was: bot t om, t op, second f r om
bottom, second from t op. CPU time consumed was 7176 seconds.
In the corresponding experiment 4 f u l l f ol ds were formed i n the
tube ( f i gur es 7.15, 7.16). The di scr epancy between exper i ment and
c a l c u l a t i o n i n t he f our t h f o l d arose f rom an ov er pr edi c t i on of t he
f o l d l engt h i n t he numer i cal model . The f i r s t 3 f ol ds spanned an
excessive tube l engt h, not l eavi ng enough f or the natural f ormat i on of
a f our t h f o l d . The reason f or t h i s ov er pr edi c t i on i s an excess
st i f f ness of the model, caused by the coarseness of the di scr et i zat i on
employed. Convergence towards the experimental behaviour was observed
f rom a pr evi ous anal ysi s of t h i s same probl em wi t h a coarse mesh
( sect i on 7.3.4.3). No f u r t h e r r ef i nement s of t he mesh were perf ormed
194
'X*
y - 4 "X
'/r
>v
. /
<
J:
-\ . v
'.; . . < o
z:r " ^
:
i
\
r\
r-.rs
* X
^
.V
t
xlOMPa
a) f rom numeri cal c al c ul at i ons
b) f rom mi crohardness t es t s
MPa
Fi gur e 7. 13: Contours of y i e l d s t r engt h Y i n crumpl ed t ube s ec t i on
(10=19.05mm, t =I.64mm, L=50.8mm Tube) (geomet ry A)
195
Rigid Platens
b) Mesh after 40mm axial compression
(iqure 7.14: Calculations for axial collapse of tube
ID=19.05mm, t
o
=l.l7mm, L=50.8mm, tube geometry B
AXIAL COLLAPSE
ID=19.05mm, t
0
=l.17mm, L=50.8mm HIRE
Figure 7.15: View of crumpled tube (geometry B)
AXIAL COLLAPSE
m=19.05mm, t
0
=1.17mm, L=50.8mm TUBE
._; i ; .
i ... 1... -L. 1.
Figure 7.16: Section of crumpled tube (geometry B)
i n order t o avoi d excessive computer cost s.
Fur t her i ns i ght i nt o t he behavi our of t he model may be gai ned
f rom t he l oad- compr essi on curves ( f i gur e 7.17). Si mi l a r l oad l ev el s
and pat t er ns of r i ses and f a l l s i n t he curves were obt ai ned i n
experiment and cal cul at i on. However, the separati on between successive
peaks i s greater f or the numerical model, i ndi cat i ng longer f ol ds. For
the l ast f ol d the l ength of tube remaining i s too short . A very s t i f f
f i nal f ol d i s then formed, wi t h a l arge degree of shear def ormat i on.
Computed load l evel s become here much greater than experimental ones.
Nei t her t he i ndi v i dual u n f i l t e r e d l oad- compr essi on curves f or
both pl at ens nor t he energy bal ance gr aph, gi ven f or t he pr evi ous
anal ysi s (tube geometry A) i n f i g s . 7.11 and 7.12, are gi ven f or t h i s
or the f ol l owi ng tube analyses (geometries B, C, D). The st ory t ol d by
these graphs i s much the same as f or tube geometry A.
Numerical and experimental contour maps of Von Mises st r engt h, Y,
are gi ven i n f i gur e 7.18. As f or t ube geometry A, s i mi l a r l ev el s and
pat t er ns of st r ess may be appr eci at ed i n bot h maps. Agai n, t he
gradi ents f or the experimental contours are steeper.
7.3.3.3 TUBE GEOMETRY C: 0D=38.1mm, t
0
=1.65mm, L=50.8mm
A " f i ne" mesh was used f or t h i s anal y s i s , wi t h 633 nodes and 996
t r i angul ar cel l s. The i n i t i a l and f i nal conf i gur at i ons of the mesh may
be seen i n f i gur e 7.19. One concer t i na f o l d was f ormed at ei t her end
of the tube. As f or the previous anal ysi s, the l engt h of the f ol ds was
over pr edi ct ed, l eavi ng t oo l i t t l e t ube l engt h l e f t f or t he proper
f ormat i on of a t hi r d f ol d i n the mi ddl e. Although at one poi nt i t di d
seem t hat a t h i r d shor t er f o l d woul d be f or med, event ual l y t h i s
mechanism proved too s t i f f and the model f ai l ed through shear near the
t op.
The experimental r esul t s di d not show t hi s behaviour, producing 3
compl et e f ol ds ( f i gur es 7.20 and 7.21). The l oad- compr essi on curves
( f i gur e 7.22) show t hat , although somewhat t r ansl at ed t o the r i ght f or
the cal cul at i ons, a great s i mi l ar i t y exi st s between the experimental
\
r,
I BS^
X> /
\
J *~
?
v
V
AXIAL COLLAPSE OF TUBE
ID=l 9.05mm, t = I . I 7mm, L=50.8mm
o
(Geometry B)
0.00 5.00 10.00
15.00 20.00 25.00
- St r oke (mm)
FIGURE 7.17: CALCULATIONS FOR AXIAL COLLAPSE OF TUBE; LOAD - COMPRESS
199
xlOMPa
a) f rom numer i cal c a l c u l a t i o n s
b) from mi crohardness t es t s
Fi gur e 7. 18: Cont ours of y i e l d st r engt h i n crumpl ed t ube sect i on
(ID=19.05mm, t
Q
=i . | 7mm, L=50.8mm Tube) (geom etry B)
200
OD = 38.lmm
b
E
00
O
\i=
V= 20m/s
:::a t o-l 65mm
ES
3
HT 30 ANNEALED ALUMINIUM TUBE
Fine Mesh solution
a) Undeformed mesh
^Rigid Platens
I
b) Mesh after 38mm axial compression
Fi gure 7.19: Calculations for axiaI col lapse of tube
0D=38.lmm, t =l.65mm, L=50.8mm , tube geometry C
AXIAL COLLAPSE
00=38.l mm, t
0
=1. 65mm, L=50.8mm TUBE
/.. . - *n
w
cm.
201
O 3
4 5 6
Figure 7.20: View of crumpled tube (geometry C)
AXIAL COLLAPSE
00=38.lmm, t
0
=1.65mm, L=50.8mm TUBE
Figure 7.21: Section of crumpled tube (geometry C)
EXPERIMENT
CALCULATION
.AXIAL COLLAPSE OF TUBE - 00=38.111111, t
0
=l , 65na L^O.Snm
0.00 5. 00 10.00 15.00 20.00 25.00
- St r oke (mm)
Fi gur e 7. 22: CALCULATIONS FOR AXIAL COLLAPSE OF TLBE - LOAD VS COMPRE
and numer i cal l oad pat t er ns f or t he f i r s t f o l d . The def or mat i on
process can be cl ear l y understood by examining the snapshot views of
t he mesh i ncl uded i n f i g u r e 7.22. At poi nt A i n t he f i g u r e , a f t e r t he
b ending and sl i di ng of the top edge, contact has been made between the
out er wal l of t he t ube and t he t op pl at en, occasi oni ng a sharp
increase i n l oad. The edge then separates s l i ght l y from the top pl at en
l eavi ng only the outer f ol d surface i n contact wi t h the pl at en. With
f ur t her compression the f ol d conti nues t o bend unt i l the bottom part
of t he f o l d h i t s t he t op edge. Thi s produces a r i p p l e c l e a r l y
not i ceabl e i n both exper i ment al and numer i cal l oad curves (B). The
bottom part of the f ol d then pushes the edge up unt i l contact wi t h the
t op pl at en i s renewed. The l oad i ncr eases st eepl y f or a moment when
t he tube deforms i n axi al compr essi on, u n t i l t he bot t om edge s t a r t s
s l i p p i n g (C) and bendi ng t o f orm anot her f o l d (D). The l as t f o l d i s
ready t o be formed at E, but the remaining tube l ength has become too
shor t . A yery hi gh l oad i s necessar y, whi ch event ual l y l ev el s of f as
shear f ai l ur e occurs.
The numer i cal and exper i ment al maps of Von Mises st r engt h (Y),
gi ven i n f i g u r e 7.23, i ndi c at e agai n s i mi l a r st r ess d i s t r i b u t i o n s
around a f o l d . They al so show t he same pat t er n as t he t wo pr evi ous
analyses.
7. 3. 3. 4 TUBE GEOMETRY D: 0D=25.4mm, to =0.95mm, L=25.4mm
The mesh employed f or t hi s problem was of medium f i neness, wi t h
332 nodes and 484 t r i a n g u l a r c e l l s ( f i gur e 7.24a). A dr awi ng of t he
deformed mesh af t er an axi al compression of 20.2 mm i s given i n f i gur e
7.24b, showi ng 2 concer t i na f o l d s . 53554 comput at i onal cycl es were
performed, using 2664 CPU seconds.
the experimental r esul t s, shown i n f i gur es 7.25 and 7.26, proved
again t o be s l i ght l y less s t i f f than the computati onal model. In the
exper i ment , a f t e r t he f i r s t 2 f o l d s , some ext r a l engt h remai ned t o
f orm anot her hal f f o l d . Load compressi on curves f or exper i ment and
c al c ul at i ons are gi ven i n f i g u r e 7.27. In a way t he compar at i ve
behavi our here i s t he i nver se of t hat r epor t ed f or t he pr evi ous t wo
analyses. Here i t i s i n the experiment where a shorter l ength of tube
204
xlOMPa
a) from numeri cal c al c ul at i ons
b) f rom mi crohardness t es t s
Fi gur e 7. 23: Cont ours of y i e l d st r engt h Y i n crumpl ed t ube sect i on
(0D=38.lmm, t =I.64mm, L=50.8mm Tube)(geom etry C)
205
OD= 25.4 mr
1
A.
e
E
HT30 ANNEALED ALUMINIUM TUBE
Medium Mesh s ol ut i on
to=0.95mm.
a) Undef or med mesh
Ri gi d Platen s
b) Mesh a f t e r 20.2mm a x i a l c ompr es s i on
Fi gur e 7. 24: Ca l c u l a t i o n s f o r a x i a l c o l l a p s e of t ube
0D=25.4mm, t =0.95mm, L=50.8mm, tube geom etry D
o
AXIAL COLLAPSE
00=25.4mm, t
o
=0.95mm, L=25.4mm TUBE
Figure 7.25: View of crumpled tube (geoinetry D)
AXIAL COLLAPSE
00=25.4mm, t
o
=0.95mm, L=25.4mm HIRE
.1 I !.
_J 1_ ..: _
4 3
Figure 7.26: Section of crumpled tube (geometry D)
10.00
8.00
6.00
4.00
2.00 -
0.00
EXPERIMENT
CALCULATION
AXIAL COLLAPSE OF TUBE - 0D=25.%ni t
0
=0.95rmb L=25.4mii
0.00 _ 2.00 _. 4.00 6.00 8.00 10.00 12.00 1
St r o k e (mm)
Figure 7.27: calculations for axial collapse of tube; load - conpression
remains at the end, occasioning s t i f f e r behaviour. The st i f f eni ng i s
not so t r agi c though, because the extra l ength occurs at one end; t hi s
end event ual l y sl i pped towards the outsi de rounding of f i t s inner edge
t hr ough f r i c t i o n wi t h t he pl at en. The numer i cal model does not have
enough def i ni t i on t o represent these edge ef f ect s. I t set t l es f or two
s l i ght l y longer f ol ds than what would be pr ef er r ed.
t he ev ol ut i on of t he l oad curves ( f i g . 7.27) f or t he f i r s t f o l d
show again the same s i mi l ar i t i es between experiment and cal cul at i ons
as f or t ube geometry C ( sect . 7.3.3.3), and t hey may be expl ai ned i n
i d e n t i c a l f ashi on ( poi nt s A,B,C,D,E,F i n f i g . 7.27).
7. 3. 4 PARAMETRIC STUDIES I N NUMERICAL ANALYSES
The i nf l uence of t he f o l l o wi n g paramet ers and
model l i ng choices upon cal cul at i on r esul t s i s st udi ed:
- Coef f i ci ent of f r i c t i o n f or tube-tube and t ube-pl at en cont act s;
- Vel oci t y scal i ng;
- Mesh refi nement;
- Element t ype.
The aim i s t o obt ai n numer i cal f eedback, i n or der t o sel ect t he
c h a r a c t e r i s t i c s of t he models t o be empl oyed i n t he numer i cal
anal ysi s.
Some of the analyses presented here correspond t o tube geometries
already mentioned i n sect i on 7.3.3 (A,8,C,D); t here the f i nal analyses
were presented. Here the r esul t s of pr el i mi nar y analyses (t abl e 7.4)
are consi dered.
7. 3. 4. 1 INFLUENCE OF FRICTION
A Coulomb l aw of f r i c t i o n was used i n t he numer i cal model
( sect i on 4.8.1). The t angent i al f or ce Ft
a t t n e
i n t e r f a c e i s l i mi t e d
by t he normal f or ce between t he sur f aces i n c ont ac t , F
n> anc
(
t n e
coef f i ci ent of f r i ct i on, fi:
F
t
4 ^ F
n
( 7. 11)
An experimental det ermi nat i on of the f r act i onal behaviour of the
sur f aces and mat er i al s i nvol ved was not at t empt ed wi t h i n t h i s
i n v e s t i g a t i o n . In a s t r i c t sense, t he per f ect f r i c t i o n gi ven by t he
Coulomb law i s not appl i cabl e f or al l types of surfaces and ar bi t r ar y
act i ng normal pressures (Curni er, 1984). This i deal i zat i on represents
an approxi mati on t o the f r i c t i o n phenomenon, consi st ent wi t h our lack
of knowledge and experimental data about i t .
The i nf l uence of the var i at i on of the coef f i ci ent of f r i c t i o n was
st udi ed by performi ng several tube col l apse analyses on the same model
(Tube geometry E, f i g . 7.28). Values f or \i ranged between 0.0
( f r i ct i onl ess) and 0.3. The r esul t s are summarized i n f i gur es 7.29 and
7.30. For comparison, the corresponding experi mental r esul t s are al so
included i n f i gur e 7.29; the col l apse geometry obtai ned experi ment al l y
consi sted of 3 1/2 concert i na f ol ds.
The col l apse of the f r i c t i onl es s model was consi derabl y di f f er ent
f rom t he ot her models and f r om t he exper i ment al r e s u l t s , pr oduci ng
only two l arge concerti na f ol ds. The other models al l produced three
f ol ds and load-compression curves si mi l ar t o each other. The r esul t s
f or ju. = 0.2 and [i= 0.3 were p a r t i c u l a r l y cl ose. A tendency f or
shor t er f o l d l engt hs i s observed as f r i c t i o n decreases. Thi s can be
a t t r i b u t e d t o edge e f f e c t s , consi der i ng t hat f or t h i s p a r t i c u l a r
probl em two out of t he t hr ee f ol ds occur at t he t ube ends, where t he
const r ai nt i s smal l er the l ess f r i c t i o n between tube and pl at en.
The reason f or the di f f er ence between experimental and numerical
r es ul t s i s t hat t he numer i cal model was i n s u f f i c i e n t l y r ef i ned i n
t er ms of t he "coar se" mesh used and v e l o c i t y s c al i ng (40m/s pl at en
v e l o c i t y ) . As wi l l be seen i n f u r t h e r sect i ons, more r ef i ned models
are necessary f or good pr edi c t i ons . However, f or t he par amet r i c
st udi es a l ar ge number of computer anal yses was necessary, and cost
consi derati ons forced compromises i n the model refi nement.
210

O D ^ 38.1 m m

E
E
CO
O
;; HT30 ANNEALED |ALUMINIUM TUBE
t o = l. 22 mm
Figure 7.28: Tube model used for friction and velocity scaling
parametric analyses
0.00 5. 00 0. 00 15.00 20. 00 25. 00 3
S t r o k e (mm)
Figure 7.29, INFLUENCE OF FRICTION ON N
212
a) u=0.0 (TUBED
b) p=0.1 (TUBE2)
c)
M
=0.2 (TUBE3)
d) u=0.3 (TUBE4)
TUBE GEOMETRY E
Figure 7.30, FINAL DEFORMED MESHES FOR DIFFERENT VALUES OF FRICTION (u)
0D=38.Imm, t,= 1.22mm, L=50.8mm TUBE COLLAPSE ANALYSIS
213
7. 3. 4. 2 INFLUENCE OF VELOCITY SCALING
The c al c ul at i ons present ed i n sect i on 7.3.3 a l l had an axi al
v el oc i t y of def or mat i on of 20 m/s, r epr esent i ng a scal i ng up of t he
experimental vel oci t i es of f i ve orders of magnitude. Some r esul t s are
presented here studyi ng the i nf l uence of varyi ng the crushi ng vel oci t y
i n t he model , j u s t i f y i n g t he choi ce made f or t he f i n a l anal yses i n
secti on 7. 3. 3.
Three analyses were performed wi t h vel oci t i es of 10, 20, and 40
m/s on t he same model as sect . 7.3.4.1 (t ube geometry E, f i g . 7.28).
The val ue of t he f r i c t i o n c o e f f i c i e n t was sel ect ed as [i= 0. 2. The
r esul t s are summarized i n f i gur es 7.31 and 7.32. A ramp was i ntroduced
i n the appl i ed vel oci t y hi st or y f or the beginning of the def ormat i on,
t he v el oc i t y i ncr easi ng f r om 0 t o i t s f i n a l val ue i n 0.1 msec. Thi s
was found t o produce a smoother s t a r t f or t he def or mat i on and f ewer
i n i t i a l os c i l l at i ons .
Al l t hr ee anal yses produced 3 concer t i na f ol ds i n t he t ube. The
t hi r d cent ral f ol d was s l i ght l y short er f or 20 m/s and 10 m/s than f or
40 m/s. In the load-compression curves ( f i g. 7.31), some di f f er ence i s
noti ceabl e when the vel oci t y i s decreased from 40 m/s t o 20 m/s, but
f ur t her r educt i on t o 10 m/s produced very l i t t l e d i f f e r e n c e . The
average col l apse l oads show a monot oni c decrease t owar ds t he
experimental val ue, again the r esul t s f or 20 m/s and 10 m/s being very
cl ose.
The c r i t e r i o n proposed i n sect i on 7.3.2.2 f or eval uat i ng t he
dynamic forces i nvol ved i n the wal l crumpl i ng, Pi >
ma
Y
De u s e d
here.
From equation (7.8), the f ol l owi ng r esul t s were obtai ned:
v=40 m/s Pi=0.737 kN (8.2% of Pc*
Pt
)
v=20m/s Pi=0.184 kN (2.0% of Pc*
p t
)
v=10m/s Pi=0.046 kN (0.5% of Pc*
p t
)
Judging from these r esul t s, a vel oci t y of 20 m/s was sel ected f or
t he f i n a l anal yses. The i n e r t i a f or ces i nt r oduced by t h i s v e l o c i t y
scal i ng were smal l . Only marginal improvements are produced when the
vel oci t y i s f ur t her reduced, whi l e computati onal costs grow r api dl y
22.00
_V
CD
O
0.00
0.00 5.00
p-j^ure 7.31
10.00 15.00 20.00 25.00 30.00
Stroke (mm)
INFLUENCE OF DEFORMATION VELOCITY ON
215
a) v=40m/s. MTQC ELEMENTS (TUBES)
b) v=20m/s. MTOC ELEMENTS (TUBE?)
c) v=10m/s, MTQC ELEMENTS (TUBE8)
d) v=40m/s. MTQ ELEMENTS (TUBE6)
(TUBE GEOMETRY E)
Figure 7.52, FINAL DEFORMED MESHES FOR VARIOUS CRUSHING VELOCITIES AND
ELEMENT TYPES - 0D=38.1mm. t.= 1.22mm. L=50.8mm TUBE COLLAPSE
216
(1758 CPU sees, for 20m/s, 3474 CPU seconds for 10 m/s).
7.3.4.3 INFLUENCE OF MESH REFINEMENT
The i nf l uence of mesh r ef i nement i n t hr ee d i f f e r e n t cases was
studi ed (tube geometries A, B, and C). These models have already been
discussed i n secti on 7.3.3, where r esul t s f or the more r ef i ned meshes
considered here were gi ven.
For t ube geometry B ( f i g . 7.14a, sect . 7. 3. 3. 2), t wo anal yses
wi t h "coarse" and "medium" meshes were done; some r es ul t s are
pr esent ed i n f i g u r e s 7.33 and 7. 34. As can be seen, t he mesh
refinement had a dramatic i nf l uence, i ncreasi ng the number of f ol ds
f rom 3 (coarse mesh) t o 4 (medium mesh), more i n l i n e wi t h t he
exper i ment al r e s u l t s . The l oad- compr essi on cur ve shows c l e a r l y t he
shor t er l engt h of f ol ds f or t he more r ef i ned model . S t i l l t he f o l d
l ength was s l i ght l y overpredi cted f or the medium mesh, the 4th f ol d
bei ng a very shor t and s t i f f one whi ch event ual l y f a i l s by shear.
Further refi nement of the mesh would probably achieve bet t er r esul t s.
For t ube geometry A ( sect . 7. 3. 3. 1, f i g . 7.7a) t hr ee d i f f e r e n t
mesh grades were used: " coar se" , "medi um" and " f i n e " . The r e s u l t s ,
present ed i n f i gur es 7.35 and 7.36, are very i l l u s t r a t i v e ; mesh
r ef i nement produces a c l ear and consi st ent convergence t owar ds t he
less s t i f f experimental val ues. The f ol d l engths are i ndi cat ed by the
pos i t i on of t he val l eys i n t he l oad- compr essi on cur ve; t hese are
c ons i s t ent l y s h i f t e d t owards t he l e f t as t he mesh i s r e f i n e d ,
i ncr easi ng t he av ai l abl e l engt h f or t he l a s t f o l d . As t h i s l as t f o l d
grows nearer t o i t s " nat ur al " unconstrai ned l engt h, the energy - and
thereby the f orce - i nvol ved i n i t s f ormat i on decreases. A monotonical
approxi mati on of the load l evel s produced i n the l ast f ol d towards the
experimental values i s apparent as the mesh i s r ef i ned.
Las t l y , f or t ube geometry C ( sect . 7.3.3.3, f i g . 7.19a), " coar se"
and " f i ne" mesh analyses were performed. The r esul t s ( f i gs. 7.37,7.38)
f ol l ow a si mi l ar pat t ern t o those out l i ned f or the previous two cases:
a decrease i n t he over pr edi ct ed s t i f f n e s s and f o l d l engt h wi t h mesh
refi nement. As f or tube geometry B ( f i gs. 7.33, 7.34), the coarse mesh
14.00
12.00 -
10.00 -
8.00 -
6.00
4.00
2.00 -
- EXPERIMENT
Coarse Mesh
Medium Mesh
(see t abl e 7.4)
0.00
ID=19. 05
0.00 5.00 10.00 15.00 20.00
S t r o k e (mm)
Figure 7.33, INFLUENCE OF MESH REFINEMENT
218
a) COARSE MESH, MTQC ELEMENTS (TUBE10)
b) MEDIUM MESH, MTQC ELEMENTS (TUBE12)
I
c) COARSE MESH, MTQ ELEMENTS (TUBE11)
Figure 7.34, FINAL DEFORMED VIEWS FOR VARIOUS MESHES AND ELEMENT TYPES
ID=i9.05mm, t,=l.i7mm, L=50.8mm TUBE (geometry B)
16.00 -
14.00 -
10.00 -
0.00 5.00
10.00 15.00 20.00
S t r o k e (mm)
Figure 7.35* INFLUENCE OF MESH REFINEMEN
a) COARSE MESH. MTOC ELEMENTS (TUBE17) b) MEDIUM MESH. MTQC ELEMENTS (TUBE13)
c) FINE MESH. MTQC ELEMENTS (TUBE15) d) MEDIUM MESH, MTQ ELEMENTS (TUBE18)
e) MEDIUM MESH. CST ELEMENTS (TUBE16)
Figure 7.36, FINAL DEFORMED VIEWS FOR VARIOUS MESHES AND ELEMENT TYPES
D=19.05mm, t,=1.64mm, L=50.8mm TUBE(geometry A)
CD
O
24.00
22.00
20.00 \-
4.00
2.00 F
0.00
EXPERIMENT
Coarse Mesh (TUBE19))
- Fine Mesh (TUBE21) j
0D=38.1mm
0.00 5.00 10.00 15.00 20.00 25
S t r o k e (mm)
Figure 7.37, INFLUENCE OF MESH REFINEMENT
222
a) COARSE MESH (TUBE19)
b) FINE MESH (TUBE21)
Figure 7.38: FINAL DEFORMED VIEWS FOR VARIOUS DEGREES OF
MESH REFINEMENT - 0D=38. 1mm, t.= 1.65mm. L=50.8<r
(tube geometry C)
model missed one f ol d al t oget her. With the f i ne mesh some more l engt h
was avai l abl e f or t he t h i r d f o l d , and i t s f or mat i on was at t empt ed.
However t he model was s t i l l somewhat o v e r s t i f f ; t he l engt h of t ube
l e f t over f or t he t h i r d f o l d was t oo s hor t , and a shear f a i l u r e
mechanism was pref erred f i n a l l y .
7. 3. 4. 4 INFLUENCE OF ELEMENT TYPE
Al l t he anal yses r epor t ed above used t he Mixed Tr i a n g l e -
Quadr i l at er al wi t h Cor r ect i on (MTQC) el ement s, proposed i n sect i on
4.2.3. Here an at t empt i s made t o assess t he behavi our of t hese new
elements when compared t o the previous Mi xed Tr i angl e- Quadr i l at er al
(MTQ) el ement s ( sect i on 4.2.2) and Const ant St r ai n Tr i angl es (CST)
(sect . 4. 2. 1) .
I t was ment i oned i n sect i on 3.7.1 t hat CST meshes impose an
excessi ve number of vol umet r i c c ons t r ai nt s and t her ef or e become i n
pr act i ce a r t i f i c i a l l y s t i f f f or i ncompr essi bl e p l a s t i c f l ow. Mesh
r ef i nement does not sol ve t he pr obl em, f a i l i n g t o converge t owar ds
t r ue sol ut i ons. MTQ elements solve t hi s problem by reducing the number
of vol umet ri c const r ai nt s, but f or ^ery l ar ge def or mat i ons t angl i ng
over of the mesh can occur, cr eat i ng unacceptable negati ve volumes. A
cor r ect i on t o prevent t hi s t angl i ng over was proposed i n sect i on 4.2.3
wi t h the MTQC elements.
The convergence of MTQC meshes has been proved i n t he pr evi ous
sect i on 7.3.4.3, where successi vel y f i n e r meshes were shown t o
approach monot oni cal l y t he exper i ment al r e s u l t s . Never t hel ess, i t
seems i nevi t abl e t hat some st i f f eni ng be i ntroduced by MTQC elements.
The aim here i s t o check, f or pr act i cal l arge deformati on pl ast i c f l ow
problems, t hat t hi s st i f f eni ng i s smal l . I f so, i t wi l l be proved t hat
MTQC elements r et ai n the advantages of MTQ elements whi l e at the same
time provi di ng a more robust model.
I t must be ment i oned t hat a l l t he low v e l o c i t y t ube col l apse
problems considered coul d be anal yzed successf ul l y wi t h MTQ meshes.
Although i n other cases negative volumes di d occur (secti on 7.4.4), no
probl ems arose here due t o t he l ack of boundary c ons t r ai nt s and t he
low deformati on vel oci t i es.
The "coar se" mesh t ube model (t ube geometry E, f i g . 7.28a) was
analyzed wi t h MTQ and MTQC elements. Load-compression curves f or both
c al c ul at i ons are shown t oget her wi t h t he exper i ment al r es ul t s i n
f i gur e 7.39, where l i t t l e di f f er enc e can be appr eci at ed between t he
two cal cul at i ons. The lower value of the minimum loads at the val l eys
f or the MTQ mesh denotes a l ower resi stance t o high curvat ure bending
at f o l d s . Thi s i s cor r obor at ed by i nspect i on of t he deformed shapes
( f i gur e 7.32a, 7.32d)
The same compari son as above was done f or t ube geometry B ( f i g .
7.14a). Resul t s f or coarse meshes wi t h MTQ and MTQC el ement s are
present ed i n f i gur es 7.40, 7.34a and 7.34c. I t was seen i n t he
pr evi ous sect i on 7.3.4.3 t hat t he MTQC coarse mesh f a i l e d t o devel op
t he f our t h f o l d whi ch was observed i n exper i ment . A s l i g h t l y bet t er
behaviour i n t hi s respect was obtained from the MTQ coarse mesh. For
the f i r s t three f ol ds the l ength i s very s l i ght l y short er than f or the
MTQC mesh. Some ext ra l engt h i s l e f t over af t er the f i r s t t hree f ol ds,
but s t i l l too short f or the f our t h f ol d, evant ual l y f a i l i n g i n a high
cur vat ur e hi nge. The i mprovement however i s f ar f r om t hat obt ai ned
wi t h a f i n e r mesh ( f i g . 7.12b) where a f our t h f o l d was f or med. Agai n
sof t er bending behaviour i s appreci ated f or MTQ.
Fi nal l y t he ef f ec t of d i f f e r e n t el ement t ypes was st udi ed f or
t ube geometry A ( f i g . 7.7a). Cal cul at i ons were c a r r i e d out wi t h CST,
MTQ, and MTQC medium meshes, the r esul t s being shown i n f i gur es 7.41,
7.36b, 7.36d, and 7.36e. The gr eat ov er s t i f f nes s of t he CST mesh i s
i mmmediately apparent. The model i s compl etel y unable t o represent the
concerti na mechanism and buckles i nt o one l arge f ol d ( f i g. 7.36e) wi t h
subst ant i al l y higher loads ( f i g. 7.41). For the Mixed di scr et i zat i ons
MTQ and MTQC, no advantage i s apparent f or MTQ as t o f o l d l engt hs i n
t hi s case." An excessi vel y sof t behaviour i n high curvat ure bending i s
exhi bi t ed by MTQ elements.
One pr ac t i c al di sadvant age of MTQ over MTQC el ement s i s t h a t ,
even wi t hout developing negati ve volumes, some t r i angul ar cel l s may
become very smal l i n si ze. As i t i s c e l l si ze whi ch governs t he
comput at i onal t i me- s t ep, t h i s means i ncr eased cost of anal ysi s. In
22.00
2.00
0.00
0.00
0D=38.1mm, t
5.00 10.00 15.00 20.00 25.00
S t r o k e (mm)
f i g u r e 7.39.- INFLUENCE OF ELEMENT TYPE ON
EXPERIMENT
- MTQC elements (TUB
MTQ elements (TUBE
\
\
W
\ V i
\ ^\
\
XT
\/ \ ' v -
M
c
\
W 1
V I
\ '/
\ \
I /
/\
ID=19.05m
0-00 5.00 10.00 15.00 20.00 2
S t r o k e (mm)
.Figure 7.40: INFLUENCE OF ELEMENT TYPE ON
CD
O
28.00
26.00
24.00
22.00
20.00
18.00
16.00
14.00
12.00
10.00
8.00
6.00
4.00
2.00
0.00
< V
ID=19. 05mm,
0.00 5.00 10.00 15.00 20.00
S t r o k e (mm)
Picure 7.41 INFLUENCE OF ELEMENT TYPE O
228
general i t was found t hat r ef i ni ng MTQC meshes was a more consi st ent ,
robust and cheaper method of i mprovi ng the r esul t s than using sof t er
MTQ meshes. Thi s i s c l e a r l y ex empl i f i ed i n f i g u r e 7.42, where t he
l oad- compr essi on r es ul t s f or a MTQC f i ne mesh (8641 CPU sec) are
cl ear l y f ar bet t er than those f or a MTQ medium mesh (12882 CPU sec) at
approxi matel y 2/3 CPU cost .
7.3.5 DISCUSSION
The q u a s i - s t a t i c axi symmet r i c cr umpl i ng of t ubes has been
model l ed successf ul l y. Resul t s f or f our t ubes are gi ven i n sect i on
7.3.3 and compared wi t h exper i ment s. Gener al l y l oad- compr essi on
cur ves, deformed geomet r i es, and st r ess d i s t r i b u t i o n s showed good
matching between pr edi ct i ons and experimental r esul t s.
Nevert hel ess a consi st ent tendency f or o v e r s t i f f s ol ut i ons i s
present t o a greater or l esser degree i n most of the cal cul at i ons. The
load l evel s predi ct ed i n the load-compression curves are very cl ose t o
the experimental val ues, but the predi ct ed f ol d l engths are of t en too
great. Two f act or s may be the cause of these di f f er ences: coarseness
of the meshes and edge ef f ect s.
F i n i t e el ement s o l u t i o n s gi v e al ways upper bound energy
appr oxi mat i ons ( Zi enki ewi cz, 1977), appr oachi ng gr adual l y t he t r ue
sol ut i ons upon mesh r ef i nement . In sect i on 7.3.4.3 i t was shown t hat
MTQC meshes di d converge towards the experi mental r esul t s. Some of the
cal cul at i ons would have benef i t ed from f i ner meshes, although t hi s was
not attempted f or reasons of cost .
Edge ef f ec t s pl ayed an i mpor t ant r ol e i n most of t he probl ems
analyzed, where the number of f ol ds produced vari ed between 2 and 4.
Out of t hese, 2 f ol ds occur al ways at t he ends and may be heavi l y
i nf l uenced by t he i n t e r a c t i o n between pl at en and edge of t ube. A
compl ex p l a s t i c def or mat i on process occurs at t he t ube edges when
sl i di ng agai nst the pl at ens, cr eat i ng a char act er i st i c rounded shape
( f i gur e 7.43a), which i s not modelled properl y by the meshes employed
and t he di scr et e cont act s wi t h Coulomb f r i c t i o n ( f i gur e 7.43b).
Accurate representati on of t hi s phenomenon together wi t h the overal l
CD
O
20.00
18.00 -
16.00
H.00
12.00
10.00
8.00
6.00
4.00 -
2.00 -
EXPERIMENT
Fine Mesh, MTQC el
Medium Mesh, MTQ e
0.00
0.00
ID=19.05mm.
j _
5.00 10.00 15.00 20.00
S t r o k e (mm)
.Figure 7Ji2-. INFLUENCE OF ELEMENT TYPE ON
230
->-S~9C*5?
.
V:
'i-->*
t
--l:.
1
----
Sss' &?rC^i-.i
;
- _ .r-.V.---._-'-r-. c - - j - . - . J i ^ - - - - *
/ . , r .. , f i - , , | . I , - - - -
. - i ' ^ f ^ ^ ' " . ^ - '
:
^ ^ ^ S J V ^
* - - " T "
e - = & ^
' "*^*. -' ^v *?
i
^*' ~"
a) e x p e r i me n t a l shape
b) CaLcuLat ed shape (TUBE21)
Fi gur e 7. 43: Local d e f o r ma t i o n f r om s l i d i n g a t t ube edges
( 0D=38. l mm, t =l . 65mm, L=50.8mm Tube) ( g
e o m e
t r y C)
crumpl i ng anal ysi s needs such l ocal l y f i ne meshes t hat computations
would become uneconomic. This edge mi srepresent at i on i s bound t o have
a st i f f eni ng ef f ect upon the end f ol ds, t her ef or e i ncreasi ng the f ol d
l engt h.
The number of f ol ds and the average col l apse forces obtai ned i n
t he c al c ul at i ons are compar ed wi t h e x p e r i me n t a l r e s u l t s and
pr edi c t i ons f r om Al exander (1960) (eqs. 7.2, 7.3) i n t abl e 7.3.
Al exander' s t heor y gi ves poor pr edi c t i ons as t o t he f o l d l engt h f or
t he t hi c k t ubes wi t h hi gh t
0
/ o r a t i o s . Thi s i s pr obabl y due t o t he
inadequacy f or t hi cker t ubes of Al exander ' s hi nge/ t hi n shel l model .
The average l oad pr edi c t i ons are cl oser t o t he exper i ment al val ues.
One must bear i n mind however t hat a f a i r l y a r b i t r a r y average val ue
f or Y was sel ected f or equation (7.2).
Tube geometry
(A) I D=19. 05mm, t o
=1
-
64mm
'
L=50
-
8mr r
(B) I D=19. 05mm, t
o=1#17mni j L=
5
0#8mr i l
(C) 0D=38.1
mm
, t o
= 1
-
6 5 mm
'
L = 5 0
-
8 mi T
(D) 0D=25.4 mm, t
o =0
.
9
5
mm5 L =2 5 >4 mn i
no. of f ol ds
eq(7.3)
4.8
5.6
3.4
2.7
expt
3
4
3
2.5
cal c
3
3.5
2.25
2
av. col l apse load
eq(7.2)
(*)
9.6
5.8
13.6
4.8
expt
11.9
6.5
13.8
4.5
cal c
12.2
7.7
16.1
4.8
(*) f or Y=175MPa, corresponding t o 0.8 st r ai n
Table 7.3:' Summary of r es ul t s f or axi symmet r i c ax i al cr umpl i ng of
tubes f rom t heor y by Al exander ( 1960) , exper i ment s, and
numerical cal cul at i ons.
Anal.
TUBE1
TUBE2
TUBE3
TUBE4
TUBE5
TUBE6
TUBE7
TUBE8
TUBE9
TUBE10
TUBE11
TUBE12
TUBE 13
TUBE14
TUBE15
TUBE16
TUBE17
TUBE18
TUBE19
TUBE20
Geom.
( + )
E
E
E
E
E
E
E
E
F
B
B
B
A
D
A
A
A
A
C
C
Mesh
(++)'
Coarse,60x3
Coarse,60x3
Coarse,60x3
Coarse,60x3
Coarse,60x3
Coarse,60x3
Coarse,60x3
Coarse,60x3
Coarse,100x3
Coarse,65x3
Coarse,65x3
Medium,130x3
Medium,90x3
Medium,80x3
Fine, 124X3
Medium,90x3
Coarse,45x3
Medium,90x3
Coarse,60x3
Fine, 124x4
Element
type
(&*)
MTQC
MTQC
MTQC
MTQC
MTQC
MTQ
MTQC
MTQC
MTQC
MTQC
MTQ
MTQC
MTQC
MTQC
MTQC
CST
MTQC
MTQ
MTQC
MTQC
Frict.
(W
0.0
0.1
0.2
0.3
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
Platen
veloc.
(m/s)
40(A)
40(4)
40 U )
40(4)
40
40
20
10
20
20
20
20
20
20
20
20
20
20
20
20
no of
folds
2
3
3
3
3
3
3
3
5
3
3
3 1/2
3
2
3
1/2
2 1/2
3
2
2 1/2
Avge.
Load P
[kN) (*)
10.2
11.7
11.4
11.4
10.5
9.9
10.0
9.8
10.2
7.2
7.3
7.7
12.9
4.8
12.2
22.2
14.3
11.5
14.8
16.1
CPU
(sec)
(**)
1040
1414
1364
962
1079
1196
1758
3474
5333
2263
3734
7167
3352
2664
8641
2346
1035
12882
1530
7732
(+) Geometry A: ID=19.05mm, t
0
=1.64mm, L=50.8mm
Geometry B: ID=19.05mm, t
0
=1.17mm, L=50.8mm
Geometry C: 00=38.10mm, t
0
=1.65mm, L=50.8mm
Geometry D: 00=25.40mm, t
Q
=0.95mm, L=25.4mm
Geometry E: 0D=38.10mm, t
0
=1.22mm, L=50.8mm
Geometry F: 0D=38.10mm, t
0
=1.22mm. L=88.9mm
(++) For cl assi f i cat i on of meshes see sect i on 7. 3. 2. 1
{&) Vel oci t y was appl i ed wi thout i n i t i a l ramp f or these cases
{&&) See secti on 4.2
(*) P = [JPdxj/L
(**) CPU times on a CRAY-IS
Table 7.4: Det ai l s of numerical analyses f or axi al tube col l apse
The numer i cal model i s not const r ai ned by beam or t h i n shel l
assumptions, and t her ef or e f i nds no problems i n model l i ng the t hi ck
tubes, where i n f act i t gives bet t er pr edi ct i ons. The load r esul t s are
o v e r s t i f f i n a l l cases; however, f or t he second and t h i r d t ubes i n
t abl e 7.3 t he r e l a t i v e l y l ar ge di f f er enc e wi t h exper i ment i s due t o
the short er, s t i f f e r i ncompl ete f ol ds at the end of the anal ysi s.
7.4 MEDIUM VELOCITY (176m /s) TUBE IMPACT ANALYSIS
The i mpact of medium v e l o c i t y hard and s of t mi s s i l es on
st ruct ures has been the obj ect of some i nt er est i n the past few years.
Extensive programs of experi mental research have been under t aken at
Meppen, FDG (e. g. Rudi ger and Ri ech, 1983) and Wi n f r i t h , UK (e. g.
Barr, 1983a, 1983b). The research i s concerned wi t h "medium" and "low"
vel oci t y impact, f or which st r ai n- r at e ef f ect s are moderate.
A crashi ng ai r c r af t consi st s pr i nci pal l y of a sof t t ubul ar body
wi t h some smal l er hard or semi hard par t s . I t has been model l ed at
Wi n f r i t h and Meppen by t ubul ar capped st eel mi s s i l es wi t h added
masses, i mpact i ng at v e l o c i t i e s between 150 and 300 m/s. In general
the impact of such mi ssi l es produced concert i na type crumpl i ng. The
axi al di r ect i on of the impact and the extensi ve crumpl i ng produced has
prompted some one-di mensi onal empi r i c al - numer i c al s ol ut i ons (e. g.
Bignon and Ri era (1979), Hurl ey (1983)). These s o l u t i o n s , however,
depend heavi l y on the i nt er pr et at i on of empi ri cal data from crushi ng
l oads and s t r a i n - r a t e ef f ec t s on col l apse mechanisms, not provi di ng
t heref ore t rue general pr edi ct i ons.
The numerical procedures described i n t hi s t hesi s (chapter 4) are
sui t abl e f or model l i ng the extensi ve axi symmetri c crumpl i ng produced
by the impact of a t ubul ar steel mi ssi l e agai nst a r i gi d surface. Such
an anal y s i s , cor r espondi ng t o Wi n f r i t h exper i ment no. M i l l , i s
descri bed here.
7.4.1 DESCRIPTION OF PROBLEM
The missile (fig. 7.44) consists of a thin spherical cap followed
234
makewei qht
4. Kcj
V
=
!76m/s
L
2.5mm "> !'> i c k e r pi pc-
l"Ot-cj| mass |SK<
1,6mm t h i n [if; \ r, j p
t
)
Sbmrn 740 mm 625 mm
1 A
S'!;S ' ry
Node 3890 Cell 5693
Node 400 Cell 5671
FW
Figure 7.45s MESH FOR TUBULAR MISSILE CALCULATIONS ( G L N L R A L VIEW>
by a pi pe and, ev ent ual l y , an end mass. The pi pe i s composed of t wo
par t s, each wi t h a di f f er ent thi ckness and s l i ght l y di f f er ent mat eri al
pr oper t i es.
The mi s s i l e was const r uct ed of col d r o l l e d mi l d st eel sheet ;
apart from t hi s f act , other ex pl i c i t mat eri al data were scarce, being
l i mi t ed t o the values of the y i el d stress and el ongat i on at f a i l u r e :
thi ckness Yi el d st ress El ongati on at f ai l ur e
1.6 mm 308 MPa 29%
2 . 5 mm 270 MPa 24%
The mi s s i l e i mpact s t he t ar get pl ane nor mal l y, wi t h an axi al
v e l o c i t y of 176 m/s. The t ar get pl ane i s assumed bot h r i g i d and
f r i c t i onl es s .
7. 4. 2 NUMERICAL IDEALIZATION
Because of the axi symmetri c character of the probl em, MTQC t or oi dal
el ement s ( sect . 4.2.3) can be used t o r epr esent t he body. The mesh
consi st ed of 3901 nodes and 5771 t r i a n g u l a r pl ane c e l l s (2887 MTQC
el ement s) ( f i gur es 7.45 and 7.46). The mesh was prepared usi ng t wo
quadr i l at er al s through the thi ckness of the spheri cal cap. In the tube
sect i on, three quadr i l at er al s through the thi ckness wi t h approximate
shape r a t i o 2:1 were t aken, i n or der t o r epr esent accur at el y t he
concerti na f ol di ng expected. For a smal l f urt hermost part of the tube
where bending was not expected, the mesh was coarsened t o two and one
quadr i l at er al s through the t hi ckness. The makeweight was represented
wi t h l arger sol i d elements having the appropri at e mass. The t ar get was
modelled as a r i g i d , f r i ct i onl ess st onewal l .
The steel mat eri al was assumed t o have a densi ty of 7800 Kg/m3,
and an e l a s t i c - p l a s t i c s t r e s s - s t r a i n behavi our . The e l a s t i c par t i s
charact eri zed by a Young's modulus of 210 GPa and a Poisson's r at i o of
0. 3.
The pl as t i c i t y was described by a Von Mises Yi el d surface and an
236
^'"/'//////////////////////////////ST-
Figure 7.46, MESH FOR TUBULAR MISSILE CALCULATIONS (DETAIL OF TIP)
////////////"/////////
Figure 7.47, TUBULAR MISSILE IMPACT - VIEW OF TIP AT 0.1msec
associ at i ve law. However, the radius of the Von Mises cyl i nder i s not
constant but changes wi t h pl ast i c st r ai n and wi t h st r ai n- r at e. At each
t i me, the ef f ect i ve stress at y i el d i s af f ect ed by a power law s t r ai n-
hardeni ng and a power law c or r ec t i on f or s t r a i n - r a t e s e n s i t i v i t y ,
according t o the expressi on:
Y = A e
n
( l +e
m
/ B)
where A and n descri be the hardening behaviour
f or t
0
=i.6mm (t hi nner pi pe) A=495 MPa, n=0.17
f or t
0
= 2
-
5 mm
( t hi cker pi pe) A=470 MPa, n=0.17
8 and m represent the st r ai n- r at e sensi t i vi t y
f or both cases B=40.4 sec-m, m=0.2
7. 4. 3 NUMERICAL RESULTS
The anal ysi s was perf ormed on a PRIME 750 comput er . I t was
pursued up t o a t i me of 2.5 msec, bei ng i nt er r upt ed t hen because of
two reasons:
- the computer costs had become excessi ve, over 39 CPU days;
- f ur t her deformati on of the mi ssi l e would only i nvol ve r epet i t i on of
the concerti na f ol di ng mechanism already observed; not much addi t i onal
i nf ormat i on would be gained from the cont i nuat i on of the anal ysi s.
The number of computational cycl es was 111336. Successive views
of the deformed mesh are presented i n f i gur es 7.47 - 7.53 ( det ai l s of
mi s s i l e t i p ) and 7.54 (general vi ews) . I t can be seen t hat a f t e r an
i n i t i a l stage i n which the spheri cal cap buckled i nwardl y ( f i g. 7.47),
t he r est of t he def or mat i on consi st s of t he p i l i n g up of successi ve
concer t i na f ol ds (18 i n t o t a l ) . Al l t he f ol ds were produced i n t he
t hi nner tube sect i on, the t hi cker part not having being reached at the
ti me the anal ysi s was stopped.
The r esul t ant of the contact forces between mi ssi l e and stonewal l
i s pl ot t ed i n f i g u r e 7.55 as a hi s t or y of t i me. The i n i t i a l par t , up
t o 0.2 msec, corresponds t o the buckl i ng of the spheri cal cap, and i s
char act er i zed by bei ng wery " noi sy" . The remai nder consi st s of a
238
i n i i i n i T i i m i m=^s
V777777777V77777777777777777777777777777777777777777777777
Fi g u r e 7 . 4 8 : TUBULAR MI SSI LE IMPACT - VIEW OF TI P AT 0. 2 msec
'y//////77/7/7/7777777//77/////////777/////777/7777777///
F i q u r e 7 . 4 9 : TUBULAR MI SSI LE IMPACT - VIEW OF TI P AT 0. 25 msec
239
/
777777777777777777777777777777777777777777777777,
Fi g u r e 7 . 5 0 : TUBULAR MI SSI LE IMPACT - VIEW OF TI P AT 0. 6 msec
h : 1
^ ^ 3^ _L_LJ_LLi j -

^7777777777777777777777777777777777777777777777777777777777
F i g u r e 7. 51= TUBULAR MI SSI LE IMPACT - VIEW OF TI P AT 1. 0 msec
240
Figure 7.52: TUBULAR MISSIVE IMPACT - VIEW OF TIP AT 2.0 msec
Figure 7.55, TUBULAR MISSILE IMPACT - VIEW OF TIP AT 2.5 msec
241
+=0. 00msec t =0. 20msec t =I . 0 0 ms e c t =1. 80msec
t-=0. 04msec t =0. 40msec t = 1.20msec +=2. 00msec
t =O. I Omsec t =0. 60msec +=I . 40msec
l t
t - - 2. 20msec
+=0. 15msec +=0. 80msec +=I . 60msec
lI
" =2. 50msec
Fi gur e 7 . 5 4 : T u b u l a r mi s s i l e i mpact - s u c c e s i v e g l o b a l vi ews
of def or med mesh
1400
1200
1000
800
600
400
200
EXPERIMENT
CALCULATION
Figure 7.55: TUBULAR MISSILE IMPACT - LOAD HISTORY ON TARGET
1000
800
600
400
200
~i 1 r
i 1 1 r
EXPERIMENT
CALCULATION

<
- -"I i_ j i_ _ i i i i _
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00 2.20 2.40
T I me (msec)
Figure 7.56, TUBULAR MISSILE IMPACT - IMPULSE HISTORY
number of peaks, t wo per f u l l concer t i na f o l d . For each f o l d t wo
hinges are formed: one external and one i nt er nal . The bending i n these
hinges corresponds t o the less s t i f f periods ( val l eys) , whi l e f or the
per i ods i n bet ween, t he l oads are t r ans mi t t ed by t he much s t i f f e r
axi al compression mechanism (peaks).
The f or ce was i nt egr at ed wi t h r espect t o t i me , obt ai ni ng t he
i mpul se hi s t or y ( f i gur e 7.56). Thi s c o n s t i t u t e s a much smoother and
nearl y l i near curve. I t i ndi cat es a f a i r l y constant col l apse f or ce.
Ve l o c i t y h i s t o r i e s ar e pr es ent ed i n f i g u r e 7.57 f o r t wo
represent at i ve poi nt s. One corresponds t o the back end of the mi ssi l e,
showi ng an appr oxi mat el y const ant average dec el er at i on. The second
poi nt corresponds t o the f ol ded sect i on, showing an i n i t i a l stage of
slow decel erat i on f ol l owed by a sudden stoppage i n the f ormat i on of
the f ol d, cont i nui ng at zero vel oci t y t her eaf t er .
Fi nal l y, hi st or i es of ef f ect i ve stresses are given i n f i gur e 7.58
f or a poi nt near t o t he nose whi ch has suf f er ed subst ant i al p l a s t i c
deformations i n the f ol di ng, and f or another poi nt near the back which
has behaved el as t i c al l y , wi t hout reaching y i el d.
7.4.4 DISCUSSION
The concer t i na axi symmet r i c cr umpl i ng has been model l ed on a
l arge scale wi t hout any computati onal problems (apart from the cost ) ,
provi ng the robustness of the numerical al gori t hms and computer code.
The sequent i al f o l d i n g process i s appar ent l y yery s i mi l a r t o t he
mechanisms al r eady di scussed f or q u a s i - s t a t i c anal ysi s. However,
consi der abl y l ar ger cr ushi ng l oad l ev el s were observed f or medium
vel oci t y crumpl i ng than f or quasi - st at i c si t uat i ons.
The experiment no. Mi l l at Wi nf r i t h produced 23 f ol ds i n the t hi n
sect i on of t he t ube and 6 f u r t h e r 4- l obed f ol ds i n t he t h i c k e r
s ec t i on, l eavi ng an undi s t or t ed t ube l engt h of 550 mm; cont act t i me
was 9.5 msec (data from Hurl ey, 1983).
The average f o l d l engt h obt ai ned i n t he exper i ment f rom t he
244
20
-20
-40
-60
-80
-100
-120
-140
-160
-180
i 1 1 r
Node 400 (in crumpled
portion)
Node 3890
(back end of projectilejL
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00 2.20 2.40
T i me (msec)
Figure 7.57
s
TUBULAR MISSILE IMPACT - VELOCITY HISTORIES
500
450 -
400
350 -
300
250
200
150
100 -
50
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00 2.20 2.40
T I me (msec)
Figure 7.58, TUBULAR MISSILE IMPACT - STRESS HISTORIES
t hi nner pipe f r ont sect i on was
n"exp = 27.2 mm
This compares with an average value (based on 18 folds), obtained in
the calculations of
^calc = 25.5 mm
while the value obtained from equation (7.3) by Alexander (1960) is
fi
Alex = 29.3 mm
I n i t i a l l y i t was at t empt ed t o sol ve t he probl em wi t h a mesh of
MTQ el ement s ( s e c t . 4. 2. 2) . Al t hough t hese el ement s wor ked
sat i sf act or i l y f or low vel oci t y cr umpl i ng anal yses ( sect i on 7. 3. 4. 4),
t hey proved unsui t abl e f or t h i s probl em i n whi ch t he v e l o c i t y of
def or mat i on i s s u b s t a n t i a l l y gr eat er . Tangl i ng over of t he mesh
occur r ed i n t he f or mat i on of t he f i r s t f o l d , pr oduci ng a negat i ve
volume i n one of t he t r i a n g u l a r c e l l s on t he i nner si de of t he f o l d
( f i gs. 7.59a, 7.59b). This caused the computations t o become unstabl e
immediately ( f i g . 7. 59c).
Previ ousl y, si mi l ar problems had been of t en observed when usi ng
Mixed Di s c r e t i z a t i o n procedures (MTQ i n 2-D, MTB i n 3-D) f or very
l ar ge def or mat i on anal yses, c o n s t i t u t i n g a drawback f or t hese
procedures. I t was t hi s par t i cul ar case which moti vated the present
research t o f i nd a correct ed element; t hi s element should preserve the
capabi l i t y f or model l i ng i ncompressi bl e pl ast i c f l ow of the exi st i ng
Mixed Di scr et i zat i on elements, but generati ng a greater resi st ance to
t angl i ng over. The r esul t was the correct ed mixed elements (MTQC i n 2-
D, MTPC i n 3-D) pr esent ed i n sect i on 4. 2. 3, whi ch have been used f or
t hi s anal ysi s.
The h i s t o r y of l oads o b t a i n e d f r om t he c al c ul at i ons was
subst ant i al l y " noi si er " than t he exper i ment al hi s t or y ( f i gur e 7.55).
Two f act or s are the cause f or t hi s di f f er ence. Fi r s t l y the f i n i t e r i se
t i me f or the l oad- cel l used i n the experiments wi l l have probably made
i t miss some of t he very f as t t r a n s i e n t s , al t hough unf or t unat el y no
/C\
\ X
/ \
T^-\Ax
\ \ Ar / W A
-A/ X>
\
A
*f~.
3
!
A
/
- -
a) j u s t p r i o r t o t a n g l i n g ov er -
b) j u s t a f t e r t a n g l i n g o v e r '
Fi gur e 7. 59: Tu b u l a r mi s s i l e i mpact c a l c u l a t i o n s - t a n g l
data are av ai l abl e t o quant i f y t h i s hypot hesi s. Secondl y, t he gr eat
number of a r t i f i c i a l contact spri ngs i n the numerical model may have
i ntroduced some addi t i onal noise i nt o the el ast i c f ol ded regi on, which
wor ki ng i n a l i near non- di s s i pat i v e r egi me, i s n a t u r a l l y " noi sy" on
i t s own.
However, the ti me-average values of numerical and experimental
loads proved t o be remarkably cl ose:
^expt = 338.4 kN
Pcalc
= 3 2 9
-
6 k N
These averages were taken f or the peri od comprised between 0.2 and 2.5
msec, whi ch corresponds t o t he axi symmet r i c cr umpl i ng of t he t ube
af t er the i n i t i a l buckl i ng of the spheri cal cap.
The dynamic i ner t i a f orce from the decel erat i on of the f ol ds was
computed according t o eqn. (7.8); the vel oci t y v( t ) i ntroduced i n t hi s
equat i on was t he hi s t or y of v e l o c i t i e s at t he t op node (node 3890,
f i g u r e 7.57). The t i me average of t h i s i n e r t i a f o r c e , agai n r ef er r ed
t o the peri od 0.2-2.5 msec, was
Pi = 179.3 kN
whi ch, subst r act ed f r om t he val ue of t he t o t a l f or c e, P=329.6 kN,
according t o eqn. (7.5) leaves an average crushi ng f orce of
p
c
= 150.3 kN
From t hese val ues i t i s apparent t hat t he dynamic ef f ec t s were
i mportant f or t hi s problem, the i ner t i a forces being approxi matel y 1/2
of t he over al l t o t a l f or ces, and of t he same or der as t he f or ces
or i gi nat ed by the crumpl i ng process i t s e l f .
I t i s not possi bl e t o compute pr eci sel y the mean crushi ng f orce
r esul t i ng from experiment as no data exi st f or vel oci t y hi st or i es of
the mi ssi l e duri ng the impact. However, the f ol l owi ng consi derat i ons
support t he assumpt i on of t he exper i ment al and numer i cal cr ushi ng
forces being very si mi l ar . According t o eqn. (7.9) the crushi ng f or ce,
p
c
, may be expressed as the di f f er ence between the t ot al f or ce, P, and
the i ner t i a f or ce, P^
#
i t was seen above t hat the di f f erence between
experimental and numerical average t ot al forces was small (2.6%). On
t he ot her hand, f or t he per i od anal yzed, t he decel er at i on has been
small (from 176m/s t o 152m/s, see f i g . 7.57). The cl ose match between
numerical and experimental impulse hi st or i es ( f i g. 7.56) suggests t hat
a si mi l ar l y small decel erat i on would have happenned i n experiment. As
t he i n e r t i a f or c e, P^ depends on t he v e l o c i t y (eqn. 7. 8), i t may be
presumed t hat exper i ment al and numer i cal cr ushi ng f or ces w i l l be
si mi l ar wi t hi n an er r or of approxi matel y 10%.
This "medium" vel oci t y crushi ng f orce i s consi derabl y l ar ger than
t hat obtained from st at i c t est s performed on i dent i cal tubes (Hurley,
1983), which yi el ded a value of
pstat
= 61 kN
This st at i c value i s consi st ent wi t h the general l y good quasi - st at i c
l oad pr edi ct i ons f r om eqn. (7.2) ( Al exander , 1960). Consi der i ng an
average value of Y = 476 MPa, which corresponds t o 0.8 s t r ai n,
pAlex
= 7 1 k N
Of t hi s di f f er ence (150.3kN medium vel oci t y, 61kN quasi - st at i c) ,
the st r ai n- r at e mat eri al hardening accounts only f or a minor par t . The
st r ai n- r at e enhancement may be approximated as f ol l ows.
- time f or the formati on of one complete f ol d (two hi nges):
t = (h-2t
Q
)/ v = (27.2-3.2)10~
3
m/(170m/s) = 1.41xl0"
4
(sec. )
- average st r ai n- r at e:
A6= 2x0.8 = 1.6
e=
A
- = l . l xl O
4
(sec-1)
At
- st r ai n- r at e enhancement f act or :
1+ e
m
/ 8 = l +( l . l xl 0
4
) -
2
/ 40. 4 = 1.16,
which corresponds t o an increase of only 16%.
Even al l owi ng f or errors i n these si mpl i f i ed consi derat i ons, the
di f f erence between quasi - st ast i c and medium vel oci t y crushi ng loads i s
t oo l ar ge (approx. 100%) t o be expl ai ned by s t r a i n - r a t e mat er i al
hardening al one. This r esul t i s extremel y i nt er est i ng and suggests an
i nher ent l y s u b s t a n t i a l l y gr eat er energy d i s s i p a t i o n f or medium
v el oc i t y t ube cr umpl i ng, even f or r at e- i nsensi t i ve mat er i al s. I t i s
a l l t he more not abl e si nce e x t e r n a l l y no apparent di f f er ences are
appr eci at ed between t he crumpl ed geomet r i es f or l ow and medium
vel oci t y i mpact.
An expl anat i on f or t h i s may be f ound i n t he pronounced non-
l i n e a r i t y of t he pr ocess. A f i n a l geometry whi ch i s s u b s t a n t i a l l y
t he same may be ar r i v ed at by d i s t i n c t pat hs, associ at ed wi t h
di f f er ent amounts of di ssi pat ed energy.
7.5 CONCLUSIONS
1- Tubes subjected t o axi al col l apse are ef f i c i ent energy absorbers
f or impact si t uat i ons. Experimental knowledge from the bahaviour of
Aluminium tubes under quasi - st at i c axi al col l apse was obtained from
t es t s c ar r i ed out mai nl y by Ghani (1982). Mi cr ohar dness t es t s i n
sect i ons of t he crumpl ed tubes al l owed t he det er mi nat i on of y i e l d
st r ess (Y) d i s t r i b u t i o n s around t he f o l d s . These d i s t r i b u t i o n s were
useful as a check of the numerical pr edi ct i ons.
2- The numer i cal model l i ng of q u a s i - s t a t i c , axi symmet r i c axi al
col l apse of tubes was approached wi t h the non-l i near ex pl i c i t Fi ni t e
Di f f er ence code ( chapt er 4). The numer i cal pr edi c t i ons ( l oad-
di spl acement cur ves, deformed geomet r i es, st r ess d i s t r i b u t i o n s )
compared wel l wi t h experiment. The numerical code used const i t ut es a
r obust s ol ut i on method f or model l i ng what i s a hi ghl y compl ex non-
l i n e a r pr ocess, wi t h very l ar ge s t r ai ns and a r b i t r a r y t ube- t ube and
t ube-pl at en cont act s. The si mpl e, sound mat eri al model employed, based
on Von Mises i s ot r opi c p l a s t i c i t y ( chapt er 6) , al so per f or med
excel l ent l y.
3- An e x p l i c i t (dynami c) s ol ut i on method was appl i ed successf ul l y
f or model l i ng the quasi - st at i c col l apse, usi ng vel oci t y scal i ng i n a
cont r ol l ed manner.
4- The numer i cal model proved s ens i t i v e t o mesh r ef i nement ,
speci al l y i n short tubes wi t h few f ol ds i n the crumpl i ng mechanism,
f or which end ef f ect s are i mport ant . An over al l tendency f or over st i f f
predi ct i ons ( f ol d l engths too l arge) was detected. Convergence towards
experi mental l y observed behaviour was achieved wi t h mesh ref i nement .
5- The choice of appropri at e elements f or the numerical model was
essent i al f or successful pr edi ct i ons. CST elements (secti on 4.2.1) are
t o t a l l y uns ui t abl e. CMTQ el ement s ( sect i on 4.2.3) pr ovi ded t he best
resul t s and the most robust meshes.
6- A medium vel oci t y (176m/s) impact of a long t ubul ar pr oj ec t i l e,
ocassioning l arge-scal e crumpl i ng, was modelled successf ul l y, al bei t
at a hi gh cost i n computer r esour ces. Resul t s compare wel l wi t h
exper i ment , and i ndi c at e s u b s t a n t i a l l y gr eat er l ev el s of energy
di ssi pat i on i n the crumpl i ng process than f or quasi - st at i c col l apse.
CHAPTER 8
CONCLUSIONS AND SUGGESTIONS FOR FURTHER RESEARCH
8.1 CONCLUSIONS
8.2 SUGGESTIONS FOR FURTHER RESEARCH
8.2.1 Theoretical and numerical developments
8.2.2 Additional applications
8.1 CONCLUSIONS
1. Expl i ci t Fi ni t e Di f f erence techni ques, as used i n t hi s work, are
advantageous f or non- l i near , short durat i on t r ansi ent anal ysi s (wave
propagation t ype). They can be also useful i n slow l oadi ng but st eepl y
non- l i near pr obl ems, i n or der t o t ake advantage of t h e i r r obust ness
and ease of i mpl ement at i on. In general , t heor et i cal sol i d mechanics
concepts and equations can be implemented st r ai ght f or war dl y. Many of
t he more r i gor ous st r ess update procedures necessary f or i mp l i c i t
methods need not be used, as t i me-st eps are r est r i ct ed t o very smal l
values f or s t a b i l i t y . A Lagrangian mesh together wi t h a Cauchy st ress
- v el oc i t y s t r a i n f or mul at i on are appr opr i at e f o r l ow - medium
vel oci t y, l arge st r ai n el ast i c- pl ast i c problems wi t h cont act s.
2. Meshes based on const ant s t r a i n t r i a n g l e s or t et r ahedr a (CST)
pr ovi de unaccept abl e r es ul t s f or i ncompressi bl e pl ast i c f l ow. Mixed
Di scr et i zat i on procedures based on aver agi ng t he vol umet r i c s t r ai ns
f rom several t r i a n g l es/ t et r ahedr a ( Mar t i and Cundal l , 1982) provi de
accurate predi ct i ons and are easi l y implemented i nt o computer codes;
however, t angl i ng over of t he mesh can occur i n probl ems wi t h very
l ar ge d i s t o r t i o n s . A c or r ec t i on method i s proposed here ( sect i on
4.2.3) f or Mi xed Di s c r e t i z a t i o n pr ocedur es, whi ch mai nt ai ns t h e i r
a b i l i t y f or accurate pl ast i c f l ow model l i ng, recoveri ng the resi st ance
t o t angl i ng over of CST meshes. Hence, a more r obust and equal l y
accurate al t er nat i ve t o Mixed Di scr et i zat i on i s obt ai ned.
3. Tension t est s are a simple and r el i abl e means f or obt ai ni ng l arge
s t r a i n mat er i al s t r e s s - s t r a i n behavi our ( chapt er 6). Exper i ment al
dat a, t oget her wi t h some t h e o r e t i c a l i n t e r p r e t a t i o n and r es ul t s of
numerical si mul at i on t est s, al l owed a represent at i ve const i t ut i ve law
f or Aluminium t o be est abl i shed. A si mpl e Von Mises pl as t i c i t y model
wi t h power law hardening was employed. Results f or more complex tube
cr umpl i ng si mul at i ons showed t he appr opr i at eness of t h i s mat er i al
model.
4. Numeri cal model l i ng of t ube col l apse mechanisms t hr ough
axisymmetric sequenti al crumpl i ng was successf ul l y achieved (chapter
7) wi t h the Expl i ci t Fi ni t e Di f f erence code developed (chapter 4). The
contact al gori ht ms employed were successful and ef f i c i ent i n model l i ng
t he i nt er f ac e between t ube f ol ds and between t ube and pl at ens. The
r esul t i ng l oad-di spl acement curves and deformed geomet r i es compare
wel l wi t h exper i ment al data f or q u a s i - s t a t i c t ube c ol l aps e. St r ess
d i s t r i b u t i o n s i n c r umpl ed t ube s e c t i o n s wer e compar ed wi t h
experimental data deri ved from microhardness t est s, wi t h good r esul t s.
The i nf l uence of some i mpor t ant aspect s of t he numer i cal model was
ascert ai ned:
- the model behaviour i s over st i f f (f ol ds too l ong) i f the mesh i s not
suf f i ci ent l y r ef i ned;
- t he choi ce of appr opr i at e el ement t ypes (CMTQ, sect . 4.2.3) was
essent i al f or ef f i c i ent and successful model l i ng;
- The vel oci t y scal i ng used ( wi t hi n the low vel oci t y regime) al l owed
model l i ng of a q u a s i - s t a t i c phenomenon wi t h an e x p l i c i t , dynami c
procedure;
- end ef f ec t s ( def or mat i on and f o l d i n g at t ube edges) may have a
s i g n i f i c a n t i nf l uence i n some cases, bei ng d i f f i c u l t t o r epr esent
appropri at el y wi t hi n a l ar ger , global model.
5. Anal ysi s of a st eel tube medium vel oci t y (176 m/s) impact (sect.
7.4) provided r esul t s which compare wel l wi t h experi mental data. The
deformed geometry was s i mi l a r t o t he low v e l o c i t y case, but t he
crushi ng force was found t o be s i gni f i c ant l y hi gher.
6. Comput at i onal cost s f or t ube col l apse s i mul at i ons are hi gh,
although af f ordabl e f or speci al i zed safety or impact cal cul at i ons.
8.2 SUGGESTIONS FOR FURTHER RESEARCH
A number of opti ons are avai l abl e f or researchers i nt er est ed i n
the t opi cs dwel t upon i n t hi s t hesi s. Theoret i cal work along some of
the l i nes suggested below would enhance the appl i cabi l i t y of numerical
s i mul at i on procedures t o non- l i near processes. Other suggest i ons
i nvol ve t he use of s i mi l a r numer i cal methods as proposed her e,
widening the f i e l d of appl i cat i ons t o other non-l i near problems.
8. 2. 1 THEORETICAL AND NUMERICAL DEVELOPMENTS
The choi ce of an obj ec t i v e st r ess r at e f or l ar ge s t r a i n anal ysi s
i s not uni que ( s e c t i o n 2. 5. 1) . Ge n e r a l l y , t he use of a p a r t i c u l a r
st r ess r at e i s post ul at ed and appl i ed t o t he e x i s t i n g mat er i al dat a.
I t woul d be u s e f u l t o st udy t he i n f l u e n c e of t he d i f f e r e n t c hoi c es
a v a i l a b l e , b o t h i n t h e o r e t i c a l t e r ms and f o r t h e p r a c t i c a l
i mpl i c at i ons i n l ar ge s t r a i n engi neer i ng anal yses.
The mat er i al models empl oyed here are r e s t r i c t e d t o si mpl e Von
Mises p l a s t i c mat er i al s wi t h har deni ng. Thi s coul d be extended t o more
s o p h i s t i c a t e d ma t e r i a l model s ( s e c t . 2. 2. 7) a p p r o p r i a t e f o r met al s
( d i s t o r t i o n of y i e l d s u r f a c e , y i e l d and f a i l u r e s u r f a c e s ) , s o i l s
( d i l a t a t i o n , n o n - a s s o c i a t i v i t y , n o n - c y l i n d r i c a l y i e l d s u r f a c e s ,
mu l t i p l e s u r f a c e i d e a l i z a t i o n s ) or c o n c r e t e ( c r a c k i n g and f a i l u r e
c r i t e r i a , n o n - c y l i n d r i c a l y i e l d s u r f a c e s ) .
More r i g o r o u s a l t e r n a t i v e s e x i s t t o t he s i mp l e p l a s t i c r a d i a l
r e t u r n and s t r e s s updat e a l g o r i t h ms empl oyed her e ( s e c t i o n s 4. 5. 2,
4. 5. 4) . I t woul d be u s e f u l t o s t udy t he a p p l i c a b i l i t y and p r a c t i c a l
i nf l uence of t hese a l t e r n a t i v e s wi t h i n e x p l i c i t and i mp l i c i t codes.
El ement by El ement met hods (Hughes, L e v i t and Wi nget ( 1983b) ,
Or t i z , Pi nsky and Tay l or ( 1983) ) ar e p r o mi s i n g new a l t e r n a t i v e s f o r
non- l i near t r ans i ent anal ysi s. The main probl em seems t o be accur acy,
and publ i shed appl i c at i ons are r e s t r i c t e d t o smal l academic exampl es.
Such a l g o r i t h ms can be i mpl ement ed wi t h o u t much e f f o r t i n t o t he
ar c hi t ec t ur e of an e x p l i c i t code, i n or der t o st udy t h e i r behavi our i n
pr ac t i c al terms f or l ar ge scal e engi neer i ng comput at i ons.
For p e n a l t y met hod c o n t a c t model s , f u r t h e r r es ear c h woul d be
u s e f u l i n or der t o i n c r e a s e t he r obus t nes s and e f f i c i e n c y of t he
a l g o r i t h ms . Thi s c o u l d i n c l u d e a d a p t a t i v e met hods f o r d e t e r mi n i n g
c o n t a c t s t i f f n e s s e s , and i mpr oved sear ch a l g o r i t h ms f o r a u t o ma t i c
det ect i on of c ont ac t s .
Ex p l i c i t t i me i nt egr at i on procedures coul d be used much more
e f f i c i e n t l y f or s t a t i c or q u a s i - s t a t i c probl ems t hr ough t he use of
adapt at i ve dynamic r el ax at i on techni ques (Underwood, 1983). In f act ,
r el axat i on t echni ques are al so bei ng proposed l a t e l y f or use i n
equat i on- sol vi ng wi t h i n i mp l i c i t methods (e.g. Bel ytschko, 1983). I t
i s possi bl e t hat a r el ax at i on al gor i t hm coul d be f ound t hat was
appr opr i at e bot h f or i mp l i c i t equat i on- s ol v i ng and f or e x p l i c i t
dynamic r e l a x a t i o n . Thi s woul d make possi bl e t he hi ghl y desi r abl e
combi nat i on of e f f i c i e n t e x p l i c i t and i mp l i c i t sol ut i on techniques
wi t hi n the same al gori t hm and computer code.
8.2.2 ADDITIONAL APPLICATIONS
The thermomechanical coupl i ng al gor i t hm developed can be r eadi l y
appl i ed t o study coupled phenomena, such as metal f ormi ng processes
wi t h generation of heat from pl ast i c work and the consequent thermal
stresses and mat eri al sof t eni ng.
Further use can be made of tensi on t est s, coupled wi t h numerical
si mul at i ons, f or more det ai l ed i nvest i gat i ons i nt o mat eri al behaviour
and f a i l u r e modes i n met al s, and ef f ec t s such as t he i nf l uenc e of
pores and mechanisms f or voi d growth.
The numerical techniques employed here coul d also be used f or the
st udy of l a t e r a l compressi on of t ubes or t ube assembl i es ( sect .
7.2.3.1), t ube i nver si on ( sect . 7. 2. 3. 4), or diamond f o l d col l apse ( 3-
D model) (sect . 7. 2. 3. 3) .
The cr ushi ng f or ce f or medium v e l o c i t y i mpact ( sect . 7.4) was
s i g n i f i c a n t l y hi gher t han t hat expect ed f or low v e l o c i t y or quas i -
s t a t i c c ol l aps e, both f rom t h e o r e t i c a l and experi mental r esul t s. I t
would be very i nt er est i ng t o st udy, exper i ment al l y and numer i c al l y ,
t he i nf l uence of i mpact v e l o c i t y on cr ushi ng mechani sms, both f or
r at e- i nsensi t i ve and rate-dependent mat er i al s.
REFERENCES
ALEXANDER, J.M. (1960): 'An appr oxi mat e anal ysi s of t he col l apse of
t h i n c y l i n d r i c a l shel l s under axi al l oadi ng' , Quart . J. Mech. Appl .
Mat h. , Vol XI I I ( 1) , pp 10-15.
AL- HASSANI , S. T. S, JOHNSON, W., and LOWE, W.T. ( 1 9 7 2 ) :
' Char ac t er i s t i c s of i nver si on t ubes under axi al l oadi ng' , J. Mech.
Eng. Sci . , Vol 14, pp 370- 381.
ANDREWS, K.R.F., ENGLAND, G.L. and GHANI, E. (1983): ' Cl a s s i f i c a t i o n
of t he axi al col l apse of c y l i n d r i c a l tubes under q u a s i - s t a t i c
l oadi ng' , I nt . J . Mech. Sc i . , Vol 25(9-10), pp 687-696.
ARGYRIS, J.H. and KELSEY, S. (1954): ' Energy t horems and s t r u c t u r a l
anal ysi s' , ai r cr af t eng. , Oct. 1954.
BARR, P. (1983a): ' St udi es of t he ef f ec t s of mi s s i l e i mpact s on
st r uct ur es' , Atom (UKAEA), Vol 318( Apr i l ) .
BARR, P. (1983b): ' Col l apsi ng st eel mi s s i l e i mpact on concr et e
t ar get ' , pr i vat e communication, UKAEE Wi nf r i t h (UK).
BATHE, K.J., RAMM, E. and WILSON, E.L. (1975): ' Fi n i t e El ement
f ormul at i ons f or l arge deformati on dynamic anal ysi s' , I nt . J. Numer.
Meth. Eng., Vol 9, pp 353-386.
BATHE, K.J. and WILSON, E.L. (1976): ' Numeri cal methods i n Fi n i t e
Element anal ysi s' , Pr ent i ce- Hal l , New Jersey.
BELYTSCHKO, T. (1978): ' Ex p l i c i t t i me - i n t e g r a t i o n of s t r u c t u r e -
mechanical systems' , i n 'Advanced St r uct ur al Dynamics', J. Donea ed.,
Applied Science publ i sher s, London, pp 97-122.
BELYTSCHKO, T. (1983): 'An over vi ew of s e mi d i s c r e t i z a t i o n and t i me
i nt egr at i on pr ocedur es' , i n ' Comput at i onal methods f or t r ans i ent
anal y s i s ' , eds. T. Bel yt schko and T.J.R. Hughes, Nor t h- Hol l and,
Amsterdam.
BELYTSCHKO, T. , HOLMES, N. and MULLEN, R. ( 1975) : ' E x p l i c i t
i nt egr at i on - s t a b i l i t y , s o l u t i o n , pr oper t i es , c os t ' , i n ' Fi n i t e
Element Analysis of Nonlinear St r uct ur al Behaviour' , T. Belytschko et
a l . eds. , ASME AMD-Vol 14, N.Y., pp 1-23.
BELYTSCHKO, T. and KENNEDY, J.M. (1978): ' Computer models f or
subassembly si mul at i on' , Nucl . Eng. Design, Vol 49, pp 17-38.
BELYTSCHKO, T. and SCHOEBERLE (1975): 'On the uncondi t i onal s t a b i l i t y
of an i mpl i c i t al gori t hm f or nonl i near st r uct ur al dynamics' , J. Appl .
Mech., Vol 42, pp 865-869.
BELYTSCHKO, T. and TSAY, C.S. (1982): 'WHAMSE - A program f or t h r e e -
di mensi onal nonl i near s t r u c t u r a l dynami cs' , Nor t hwest er n U., 111. ,
r ept . NP-2250.
6ERTH0LF, L.D. and BENZLEY, S.E. (1968): 'TOODY I I - A comput er
program f or t wo- di mensonal wave pr opagat i on' , SC-RR-68-41, Sandia
Laborat ori es, Albuquerque, New Mexico.
BIGNON, P. and RIERA, J.D. (1979): ' Ve r i f i c a t i o n of methods of
anal ysi s f or sof t mi s s i l e i mpact ' , I nf or me de l a b o r a t o r i o CT- 6/ 79,
Curso de posgraduacao em Engenheri a c i v i l , U. Ri o Grande do Sul ,
Br azi l .
BILLINGTON, E.W. and TATE, A. (1981): 'The physi cs of def or mat i on and
f l ow' , McGraw-Hill, New York.
BODNER, S.R. and SYMONDS, P.S. (1960): ' Pl as t i c def or mat i on i n i mpact
and i mpul si ve l oadi ng of beams' , i n ' P l a s t i c i t y ' , E.H. Lee and P.S.
Symonds eds. , Pergamon press, N.Y., pp 488-500.
BRIDGMAN, P.W. (1944): 'The st r ess d i s t r i b u t i o n at t he neck of a
t ensi l e speci men' , Trans. Amer. Soc. Metal s, Vol 32, pp 553-574.
BRIDGMAN, P.M. (1945): ' Ef f ect s of hi gh hydr ost at i c pr essur e on t he
pl as t i c pr oper t i es of met al s' , Rev. of Modern Phys., Vol 17(1), pp 3-
14.
BRIDGMAN, P.W. (1952): ' St udi es i n l ar ge p l a s t i c f l ow and f r a c t u r e ' ,
McGraw-Hill, New York.
CARSLAW, W.S. and JAEGER, J.C. (1947): ' Conduct i on of heat i n s o l i d s ' ,
Clarendon press, Oxford.
CHEN, W.H. (1971): ' Necki ng of a bar ' , I n t . J. Sol i ds & St r uc t . , Vol
7( 7) , pp 685-717.
CLOUGH, R.W. (1960): 'The f i n i t e el ement i n pl ane st r ess anal y s i s ' ,
Proc. 2nd ASCE Conf. El ect roni c Computation, Pi t t sburgh Pa. , Sept.
CLOUGH, R.W. and PENZIEN, J. (1975): 'Dynamics of st r uct ur es' , McGraw-
Hi l l , Tokyo.
COTTRELL, A. (1975): 'An i nt r oduc t i on t o met al l ur gy , 2nd ed. ' , Edward
Arnol d, London.
COURANT, R. (1943): ' Var i at i onal methods f or the sol ut i on of problems
of equi l i br i um and v i br at i ons ' , Bul l . Amer. Math. Soc. 49, pp 1-23.
COURANT, R., FRIEDRICHS, K. and LEWY, H. (1928): 'On t he p a r t i a l
di f f er ent i al equations of mathematical physi cs' , Math. Ann., Vol 100,
pp 32-74.
CRUSE, T.A. ( 1969) : ' Numer i cal s o l u t i o n s i n t hr ee- di mensi onal
el ast ost at i cs' , I nt . J . Sol i ds & St r uct . , Vol 5, pp 1259-1274.
CUNDALL, P.A. and MARTI, J. (1979): 'Some new developments i n di scr et e
numerical methods f or dynamic model l i ng of j oi nt ed reck masses', 1979
RETC proceedings, chpt . 84, pp 1466-1484.
CURNIER, A. (1984): 'A t heor y of f r i c t i o n ' , I nt . J. Sol i ds & St r uc t . ,
Vol 20(7), pp 637-647.
DAFALIAS, Y.F. (1983): ' Corotati onal rates f or ki nemati c hardening at
l arge pl ast i c s t r ai ns ' , J . Appl . Mech., Vol 50, pp 561-565.
DAVIDENKOV, N.N. and SPIRIDINOVA, N. I . (1946): ' Anal ysi s of t he st at e
of st r ess i n t he neck of a t ensi on speci men' , Proc. Amer. Soc. f or
t est i ng Mater., Vol 46, pp 1147-1158.
DE ROUVRAY.A., ARNADEAU,F., DUBOIS,J., CHEDMAIL,J.F. and HAUG,E.
(1984): ' Numeri cal t echni ques and exper i ment al v a l i d a t i o n s f or
i ndust r i al appl i cat i ons' , i n ' St r uct ur al Impact and Cr ashwor t hi ness' ,
Vol 1, G.A.O. Davies ed. , pp 193-242.
DERUNTZ, J.A. and HODGE, P.G. (1963): ' Cr ushi ng of a t ube between
r i gi d pl at ens' , J . Appl . Mech., Vol 30, pp 391-395.
DIETER, G.E. (1976): ' Mechani cal Met al l ur gy , 2nd ed. ' , McGr aw- Hi l l ,
New York.
DONEA, J. , FASOLI-STELLA, P. and GIULIANI, S. (1977): 'Lagrangian and
Eul er i an Fi n i t e Element t echni ques f or t r a n s i e n t f l u i d - s t r u c t u r e
i n t e r a c t i o n pr obl ems' , Trans. 4t h SMIRT conf . , paper 61/ 2, San
Franci sco, Aug. 1977.
DOSTAL, M., PHELAN, P., and TRBOJEVIC, V.M. (1985): ' Anal ysi s of an
acci dent al drop of LMFBR i nt er medi at e heat exchanger
1
, I nt . Mt g. on
f ast breeder reactor saf et y, Knoxvi l l e (Tennessee), Apr. 1985.
DRUCKER, D.C. (1951): 'A more fundamental approach t o pl ast i c st r ess-
s t r a i n r e l a t i o n s ' , Proc. 1st U.S. Nat. Congress Appl . Mech. (Chi cago
1951), pp 487-491.
FLANAGAN,. D.P. and BELYTSCHKO, T. (1981a): 'Simultaneous r el axat i on i n
s t r u c t u r a l dynami cs' , J. Eng. Mech. Di v. (ASCE), Vol 107(Dec.),pp
1039-1055.
FLANAGAN, D.P. and BELYTSCHKO, T. ( 1981b) : 'A u n i f o r m s t r a i n
hexahedron and quadr i l at er al wi t h orthogonal hourglass cont r ol ' , I nt .
J. Numer. Meth. Eng., Vol 17, pp 679-706.
hUNG, Y.C. (1965): ' Foundat i ons of s o l i d mechani cs' , Pr e n t i c e - Ha l l ,
New Jersey.
GARCIA, F. ( 1981) : ' For mu1aci on del metodo de l as ecuaci ones
i nt egral es de contorno en el ast opl ast i ci dad t r i di mensi onal ' , Doctoral
t hesi s, ETSI I ndust r i al es, U. Pol i t ecni ca Madri d.
GHANI, E. ( 1982) : ' St a t i c and dynami c c o l l a p s e behav i our of
cyl i ndr i cal tubes and tube assembl i es' , Ph.D. Thesi s, Dept. Ci vi l Eng.
Ki ng' s Co l l . , U. of London.
GIULIANI, S. (1982): 'An al gor i t hm f or cont i nuous r ezoni ng of t he
hydrodynamic gr i d i n Ar bi t r ar y Lagr angi an- Eul er i an comput er codes' ,
Nucl . Eng. Design, Vol 72, pp 205-212.
GOICOLEA, J.M. (1985a): 'PR2D - An Ex p l i c i t Fi n i t e Di f f er ence Non-
l i near Thermo-mechani cal code. User' s gui de' , Dept. of Ci v i l Eng.,
Ki ng' s Co l l . , U. of London, r ept . CE-85-01
GOICOLEA, J.M. (1985b): 'PR2DPL - A program f or mesh pl ot t i ng (user's
gui de) ' , Dept. Ci v i l Eng., Ki ng' s Co l l . , U. of London, r ept . CE-85-02.
GOICOLEA, J.M. (1985c): 'CURVA - A program f or graph dr awi ng ( user ' s
gui de) ' , Dept. Ci v i l Eng., Ki ng' s Co l l . , U. of London, r ept . CE-85-03.
GOUDREAU, G.L. and HALLQUIST, J.O. (1982): ' Recent devel opment s i n
l arge-scal e Lagrangian Hydrocode technol ogy' , Comp. Meth. Appl . Mech.
& Eng., Vol 33, pp 725-757.
GREEN, E.H. and NAGHDI, P.M. (1965): 'A general t heor y of an e l a s t i c -
pl ast i c conti nuum' , Arch. Rat. Mech. Anal . 18.
GUIST, L.R. and MARBLE, D.P. (1966): ' Predi ct i on of the i nversi on load
of a ci r cul ar t ube' , US NASA t echni cal note TN D-3622.
HALLQUIST, J.O. (1979): 'NIKE2D: An i mp l i c i t , f i n i t e def or mat i on
Fi ni t e Element code f or anal yzi ng the st at i c and dynamic response of
t wo- di mensi onal s o l i d s ' , Lawrence Li ver mor e l abor at or y , U. of
Cal i f or ni a, r ept . UCRL-52678.
HALLQUIST, J.O. ( 1982a) : ' User ' s manual f o r DYNA2D', Lawrence
Livermore l abor at or y, U. of Cal i f or ni a, rept . UCID-18756.
HALLQUIST, J.O. (1982b): ' Theor et i cal manual f or DYNA2D', Lawrence
Livermore l abor at or y, U. of Cal i f or ni a, r ept . UCID-19401.
HALLQUIST, J.O. (1982c): 'DYNA3D user ' s manual ( nonl i near dynamic
anal ysi s of sol i ds i n t hr ee d i me n s i o n s ) ' , Lawr ence L i v e r mo r e
l abor at or y, U. of Cal i f or ni a, r ept . UCID-19592.
HALLQUIST, J.O., WERNE, R.W. and WILKINS, M.L. (1977): di scussi on t o
" Hi gh- vel oci t y i mpact c al c ul at i ons i n t hr ee di mensi ons" by G.R.
Johnson (1977), J . Appl . Mech., Vol 99(4), pp 793-794.
HANCOCK, S.L. (1976): ' Fi n i t e Di f f er ence equations f or PISCES-2DELK',
Physics I nt er nat i onal company report TCAM-76-2
HERNALSTEEN, P. and LEBLOIS, L.C. (1976): 'The use of energy absorbers
t o prot ect st ruct ures agai nst impact l oadi ng' , Nucl. Eng. Design, Vol
37, pp 373-406.
HERRMANN, W. and BERTHOLF, L.D. (1983): ' Ex p l i c i t Lagrangi an Fi n i t e
Di f f er ence met hods' , i n ' Comput at i onal Met hods f o r Tr a n s i e n t
Anal ysi s' , T. Bel yt schko and T.J.R. Hughes eds. , Nor t h- Hol l and,
Amsterdam.
HIBBIT, H.D., MARCAL, P.V. and RICE, J.R. (1970): 'A Fi n i t e El ement
f or mul at i on f or problems of l arge st r ai ns and di spl acements' , I nt . J.
Sol i ds & St r uct . , Vol 6, pp 1069-1086.
HILL, R. (1950): 'The mat hemat i cal t heor y of p l a s t i c i t y ' , Cl arendon
press, Oxford.
HIRT, C.W., AMSDEN, A'.A. and COOK, J. L. ( 1974) : 'An A r b i t r a r y
Lagrangi an-Eul eri an computing method f or al l f l ow speeds', J. comput.
phys., Vol 14, p 227.
HUGHES, T.J.R. ( 1983) : ' An a l y s i s of t r a n s i e n t al gor i t hms wi t h
p a r t i c u l a r r ef er ence t o s t a b i l i t y behavi our ' , i n ' Comput at i onal
methods f or t r ansi ent anal ysi s' , eds. T. Belytschko and T.J.R. Hughes,
North-Hol l and, Amsterdam.
HUGHES, T.J.R., LEVIT, I . and WINGET, J. (1983a): ' El ement - By- El ement
i mp l i c i t al gor i t hms f or heat conduct i on' , J. Eng. Mech. ASCE, Vol
109(2), pp 576-585.
HUGHES, T.J.R., LEVIT, I . and WINGET, J. (1983b): 'An El ement - By-
Element s ol ut i on al gor i t hm f or probl ems of s t r u c t u r a l and s o l i d
mechanics' , Comp. Meth. Appl . Mech. & Eng., Vol 36, pp 241-254.
HUGHES, T.J.R. and WINGET, J. (1980): ' Fi n i t e r o t a t i o n ef f ec t s i n
numerical i nt egr at i on of rat e const i t ut i ve equations ar i si ng i n l arge
deformation anal ysi s' , I nt . J. Numer. Meth. Eng., Vol 15(12), pp 1862-
1867.
HURLEY, S. (1983): 'Dynamic cr ushi ng l oads of t h i n - wa l l e d st eel
t ubes' , UKAEE Wi nf r i t h, pr i vat e communication.
JOHNSON, G.R. (1976): ' Anal ysi s of e l a s t i c - p l a s t i c i mpact i nv ol v i ng
severe di s t or t i ons ' , J . Appl . Mech., Vol 98(3), pp 439-444.
JOHNSON, G.R. (1977): ' Hi gh- v el oc i t y i mpact c a l c u l a t i o n s i n t hr ee
di mensi ons' , J . Appl . Mech., Vol 99(1), pp 95-100.
JOHNSON, G.R. (1981): 'Recent devel opments and anal yses associ at ed
wi t h t he. EPI C-2 and EPIC-3 codes' , i n ' 1981 advances i n aerospace
St r uc t , and Mat er . ' , S. I . Wang and W.J. Renton eds. , wi nt er mt g. ASME,
Washington D.C., Nov 1981, pp 142-147.
JOHNSON, W. (1972): 'Impact st rengt h of mat er i al s' , Arnol d, London.
JOHNSON, W. and REID, S.R. (1978): ' Me t a l l i c energy d i s s i p a t i n g
systems' , Appl . Mech. Rev., Vol 31( 3) , pp 277-288.
JOHNSON, W., REID, S.R. and REDDY, T.Y. (1977): 'The compressi on of
crossed l ayers of t hi n t ubes' , I nt . J . Mech. Sc i . , Vol 19, pp 423-437.
JOHNSON, W., REID, S.R. and SINGH, L.B. (1975): ' Exper i ment al st udy of
t he r o l l i n g t or us l oad d e l i mi t e r ' , I n t . J. Mech. Sc i . , Vol 17, pp 603-
615.
JOHNSON, W., SODEN, P.D. and AL-HASSANI, S.T.S (1977): ' I next ensi onal
col l apse of t h i n - wa l l e d t ubes under axi al compr essi on' , J. St r ai n
Anal . , Vol 12(4), pp 317-330.
JONES, S.E., GI LLI S, P.P. and SHALA6Y, A.H. ( 1979) : ' St r es s
d i s t r i b u t i o n s i n t he v i c i n i t y of a neck' , J. Appl . Phys., Vol 50( 5) ,
pp 3168-3173.
KALSI, G.S. and MARTI, J. (1985): ' Thr ee- di mensi onal r ezoni ng of
Lagrangian meshes i n el ast o- pl ast i c t her mal l y coupl ed pr obl ems' , t o
appear i n Comp. Meth. Appl . Mech. & Eng.
KAPLAN, M.A. (1973): 'The st ress and deformati on i n mi l d st eel duri ng
axisymmetric necki ng' , J . Appl . Mech., Vol 40, pp 271-276.
KEY, S.W. (1969): 'A v a r i a t i o n a l p r i n c i p l e f or i ncompr essi bl e and
ear l y- i ncompr essi bl e ani sot r opi c e l a s t i c i t y ' , I n t . J. Sol i ds &
St r uct . , Vol 5, pp 951-964.
KEY, S.W. (1974): 'HONDO - A Fi n i t e Element comput er program f or t he
l ar ge def or mat i on dynamic response of axi symmet r i c s o l i d s ' , Sandia
Nat. l abor at or i es, Albuquerque N.M., r ept . 74-0039.
KEY, S.W. (1978): ' Tr ansi ent response by t i me i n t e g r a t i o n : r evi ew of
i mp l i c i t and e x p l i c i t oper at or s' , i n 'Advanced St r uct ur al Dynamics',
J. Donea ed. , Appl i ed Science publ i sher s, London, pp 15-41.
K0LSKY, H. (1953): ' Stress waves i n s o l i d s ' , Clarendon press, Oxford.
KOSLOFF, D. and FRAZIER, G.A. (1978): 'Treatment of hourglass pat t erns
i n low or der Fi n i t e Element codes' , I n t . J. Numer. Anal . Met h.
Geomech., Vol 2, pp 57-72.
KRIEG, R.D. (1975): 'A pr act i cal t wo-surf ace pl ast i c t heor y' , J. Appl.
Mech., Vol 97(E3), pp 641-646
KRIEG, R.D. and KEY, S.W. (1973): ' Tr ansi ent shel l response by
numer i cal t i me i n t e g r a t i o n ' , I nt . J. Num. Meth. Eng., Vol 7, pp 273-
286.
KRIEG, R.D. and KEY, S.W. ( 1976) : ' I mp l e me n t a t i o n of a t i me -
independent pl as t i c i t y theory i nt o st r uct ur al computer programs' , i n
' Cons t i t ut i v e equat i ons i n v i s c o p l a s t i c i t y ' , J.A. S t r i c k l i n and
K.J. Saczalski eds. , ASME wi nt er mt g. , New York, Dec. 1976.
KUKKOLA, T. (1976): ' Energy absorbers used agai nst i mpact l oadi ng' ,
Nuc-1. Eng. & Design, Vol 37, pp 407-412.
KUNAR, R. and MINOWA, N. (1981): 'A comparison between Fi ni t e Element
and Fi n i t e Di f f er ence e x p l i c i t f or mul at i ons f or t r i a n g u l a r and
quadr i l at er al pl ane s t r a i n el ement s' , I n t . J. Num. Anal . Met h.
Geomech., Vol 5( 2) , pp 217-224.
LANDON, J.W. and QUINNEY, H. (1923): ' Exper i ment s wi t h t he Hopki nson
pressure bar ' , Proc. Royal Soc , A103, p622.
LEE, E.H. (1969): ' El a s t i c - p i a s t i c def or mat i on at f i n i t e s t r a i n s ' , J.
Appl . Mech., Vol 36, pp 1-6.
LEE, E.H. and MALLETT, R.L. (1983): ' St r ess anal ysi s f or ani s ot r opi c
hardeni ng i n f i n i t e - d e f o r ma t i o n p l a s t i c i t y ' , J. Appl . Mech., Vol 50,
pp 554-560.
LICHNEROWICZ, A. (1958): ' El ement s de cal cul t e n s o r i e l ' , Li b. Armand
Col i n, Par i s.
McFARLAND, R.K. (1963): 'Hexagonal cel l st ruct ures under post buckl i ng
axi al l oad' , AIAA Jour nal , Vol 1( 6) , pp 1380-1385.
MAENCHEN, G. and SACK, S. (1964): 'The TENSOR code' , i n ' Methods of
Comput at i onal Physi cs' , 6. Al der et a l . eds. , Academic pr ess, New
York, pp 181-210.
MALVERN, L.E. (1969): ' I nt r oduc t i on t o t he mechanics of a cont i nuous
medium', Pr ent i ce- Hal l , New Jersey.
MAMALIS, A.G. and JOHNSON, W. (1983): 'The q u a s i - s t a t i c cr umpl i ng of
t hi n- wal l ed ci r cul ar cyl i nder s and f r us t r a under ax i al compr essi on' ,
I nt . J . Mech. Sc i . , Vol 25, p 713.
MAMALIS, A.G., JOHNSON, W. and VIEGELAHN, G.L. (1984): 'The cr umpl i ng
of steel t hi n- wal l ed tubes under axi al compression at el evated st r ai n
r at es: some exper i ment al r e s u l t s ' , I n t . J. Mech. Sci . , Vol 26( 11- 12) ,
pp 537-547.
MARSDEN, J.E. and HUGHES, T.J.R. (1978): ' Topi cs i n t he mat hemat i cal
f oundat i ons of e l a s t i c i t y ' , i n ' Non- l i near anal ysi s and mechani cs:
Heri ot-Watt symposium', Vol I I , ed. R.J. Knops, Pitman, London.
MARSDEN, J.E. and HUGHES, T.J.R. (1983): ' Mat hemat i cal f oundat i ons of
e l a s t i c i t y ' , Pr ent i ce- Hal l , New York.
MARTI, J. (1981): 'PR3D - Theoreti cal basis (versi on 1P-1)' , Pr i nci pi a
Mechanica Lt d. , London, r ept . PR-TN-6.
MARTI, J. (1983): 'PR3D - Programmer' s gui de' , Pr i n c i p i a Mechani ca
Lt d. , r ept . PR-TN-4.32.
MARTI, J. and CUNDALL, P.A. (1980): ' Cons t i t ut i v e l aws f or dynamic
model l i ng of s o i l s ' , r epor t t o USAF weapons l abor at or y , Dames and
Moore, London.
MARTI, J. and CUNDALL, P.A. (1982): ' Mi xed Di s c r e t i z a t i o n procedure
f or accur at e model l i ng of p l a s t i c c ol l aps e' , I nt . J. Numer. Anal .
Meth. Geomech., Vol 6( 1) , pp 129-139.
MARTI, J . , GOICOLEA, J.M., KALSI, G.S. and MACEY, G.E. (1984):
' Tr ansi ent t hermomechani cal anal ysi s wi t h e x p l i c i t schemes' , i n
'Numer. Meth. f or Tr ansi ent and Coupled Pr obl ems' , R.W. Lewi s et a l .
eds. , Pineridge press, Swansea, UK.
MARTI, J. , KALSI, G.S. and LAST, N. (1984): ' Thr ee- di mensi onal
c a l c u l a t i o n s of mi s s i l e i mp a c t ' , i n ' S t r u c t u r a l I mpact and
Crashworthi ness' , J.Morton ed. , Vol 2, pp 310-319.
MROZ, Z. (1967): 'On t he des c r i pt i on of ani s ot r opi c wor k- har deni ng' ,
J . Mech. and Phys. Sol i ds, Vol 15, pp 163-175.
MROZ, Z. (1972): 'A des c r i pt i on of work har deni ng of met al s wi t h
appl i c at i on t o v ar i abl e l oadi ng' , Proc. I n t . Symp. on f oundat i ons of
pl as t i c i t y , A. Sawczuk ed. , Noordhoff i nt er nat i onal .
MULLEN, R.L. and BELYTSCHKO, T. ( 1983) : 'An a n a l y s i s of an
uncondi t i onal l y st abl e ex pl i c i t method', Computers & St r uct ur es, Vol
16, pp 691-696.
NADAI, A. (1946): di scussi on t o ' Anal ysi s of t he st at e of st r ess i n
t he neck of a t e n s i o n speci men' by N.N. Davi denkov and N. I .
Spi r i di nova (1946), Proc. Amer. Soc. f or t e s t i n g Mat er . , Vol 46, pp
1175-1178.
NAGTEGAAL, J.C. and DEJONG, J.E. (1981): 'Some comput at i onal aspect s
of e l a s t i c - p l a s t i c l ar ge s t r a i n anal y s i s ' , I nt . J. Numer. Met h. Eng.,
Vol 17, pp 15-41.
NAGTEGAAL, J.C. and DEJONG, J. E. ( 1982) : 'Some aspect s of non
i sot r opi c work hardening i n f i n i t e st r ai n p l a s t i c i t y ' , i n ' Pl ast i ci t y
of metals at f i n i t e st r ai n: t heory, experiment and comput at i on' , E.H.
Lee and R.L,. Mal l et t eds. , Di v. Appl . Mech., Stanford U. , Cal i f or ni a.
NAGTEGAAL, J. C, PARKS, D.M. and RICE, J.R. (1974): 'On numer i c al l y
accurate Fi ni t e Element sol ut i ons i n the f u l l y pl ast i c range' , Comp.
Meth. Appl . Mech. & Eng., Vol 4, pp 153-177.
NEEDLEMAN, A. (1972): 'A numer i cal st udy of necki ng i n c i r c u l a r
cyl i ndr i cal bar s' , J . Mech. & Phys. Sol i ds, Vol 20(2), pp 111-127.
NEMAT-NASSER, S. (1982): 'On f i n i t e def or mat i on el ast o- pl a s t i c i t y ' ,
I nt . J . Sol i ds & St r uct . , Vol 18(10), pp 857-872.
NEWMARK, N.M. (1959): 'A method of comput at i on f or s t r u c t u r a l
dynami cs' , J . Eng. Mech. Di vi si on ASCE, Vol 85(EM3), pp 67-94.
NICHOLS, B.D., HIRT, C.W., and HOTCHKISS, R.S. (1980): 'SOLA-VOF: A
s ol ut i on Al gor i t hm f or t r ans i ent f l u i d f l ow wi t h mu l t i p l e f r ee
boundaries' , Los Alamos s c i ent i f i c l abor at or y, U. of Cal i f or ni a, rept .
LA-8355.
NOH, W.F. (1964): 'CEL: A t i me- dependent , t wo- space di mensi onal ,
coupl ed Eul er i an- Lagr ange code' , i n ' Met hods of Comput at i onal
Physi cs' , Vol 3, B. Al der et a l . eds.,Academi c pr ess, New York, pp
117-179.
NORRIS, D.M., MORAN, B., SCUDDER, J.K., and QUINONES, D.F. (1978): 'A
computer si mul at i on of the tensi on t est ' , J. Mech. & Phys. Sol i ds, Vol
26, pp 1-19.
ORTIZ, M., PINSKY, P.M. and TAYLOR, R.L. (1983): ' Unc ondi t i onal l y
st abl e El ement -By-El ement al gor i t hms f or dynamic pr obl ems' , Comp.
Meth. Appl . Mech. & Eng., Vol 36, pp 223-239.
ORTIZ, M. and POPOV, E.P. (1983): ' Di s t o r t i o n a l har deni ng r ul es f or
metal p l a s t i c i t y ' , J . Eng. Mech. (ASCE), Vol 109(4), pp 1042-1057.
ORTIZ, M. and SIMO, J.C. (1985): 'An anal ysi s of a new cl ass of
i nt egr at i on al gori t hms f or el ast o- pl ast i c const i t ut i ve r el at i ons' , t o
appear i n . I n t . J . Numer. Meth. Eng.
OSIAS, J.R. and SWEDLOW, J. L. ( 1974) : ' F i n i t e e l a s t o - p l a s t i c
d e f o r ma t i o n - I ; t heor y and numer i cal exampl es' , I nt . J. Sol i ds &
St r uct . , Vol 10, pp 321-339.
OWEN, O.R.J, and HINT.ON, E. (1980): ' Fi n i t e El ements i n p l a s t i c i t y :
theory and pr act i ce' , Pi neri dge press, Swansea (UK).
PARK, K.C. (1982): 'An i mproved semi - i mpl i c i t method f or s t r u c t u r a l
dynamic anal ysi s' , J . Appl . Mech., Vol 49(3), pp 589-593.
PERZYNA, P. (1963): 'The c o n s t i t u t i v e equqt i ons f or r at e s ens i t i v e
pl ast i c mat er i al s' , Quart. Appl . Mat h. , Vol 20, pp 321-332.
PINSKY, P.M., ORTIZ, M. and PISTER, K.S. ( 1983) : ' Numer i c al
i nt egr at i on of r at e c o n s t i t u t i v e equat i ons i n f i n i t e def or mat i on
anal ysi s' , Comp. Meth. Appl . Mech. & Eng., Vol 40, pp 137-158.
PRAGER, W. (1956): 'A new method of anal yzi ng st r ess and s t r a i n i n
work-hardening pl ast i c s o l i d s ' , J . Appl . Mech., Vol 23, pp 493-496.
PRAGER, W. (1961): ' I nt r oduc t i on t o t he mechani cs of c ont i nua' , Ginn
and company, Boston.
PRANDTL, L. (1920): 'Ueber di e Haerte pl ast i cher Koerper' , Goetti nger
Nachr., Math.-Phys. Kl . 1920, pp 74-85.
PRANOTL, L. (1924): ' Spannungsvert ei 1ung i n pl ast i chen Koer per n' ,
Proc. 1 I nt . Congr. Appl . Mech. ( Del f t , 1924) , pp 43-54.
PREVOST, J.H. (1978): ' Pl a s t i c i t y t heor y f or s oi l s t r e s s - s t r a i n
behavi our ' , J. Eng. Mech. (ASCE), Vol 104(EM5), pp 1177-1194.
PRINCIPIA MECHANIGA LTD. (1984): 'PRISM - User' s qui de ( ver si on 1C-
2 ) ' , Pr i nci pi a Mechanica Lt d. (London), r ept . PR-TN-29.11.
PRINCIPIA MECHANICA LTD. (1983): 'Impact anal ysi s of the W.V.P. glass
cont ai ner ' , Pr i nci pi a Mechanica Lt d. (London), rept s. 205-83, 217-83,
221-83.
PUGSLEY, A. and MACAULAY, M. (1960): 'The l ar ge- s c al e cr umpl i ng of
t h i n c y l i n d r i c a l col umns' , Quart . J. Mech. Appl . Mat h. , Vol X I I I ( l ) ,
pp 1-9.
RE I D, S.R. and REDDY, T. (1978): ' Ef f ect of s t r a i n hardeni ng on t he
l a t e r a l compressi on of t ubes between r i g i d p l a t e s ' , I nt . J. Sol i ds &
St r uct . , Vol 14, pp 213-225.
REUSS, E. (1930): ' Ber uecksi cht i ngung der el ast i chen Formaenderungen
i n der pl ast i zi t at st heor i e' , Zei t s. Angew. Math. Mech. 10, pp 266-274.
RICHTMEYER, D. and MORTON, K.W. (1967): ' Di f f er ence methods f or
i n i t i a l value problems (2nd e d . ) ' , I nt er sci ence, New York.
RIZZO, F.J. (1967): 'An i nt egr al equat i on approach t o boundary val ue
probl ems of c l as s i c al e l a s t o s t a t i c s ' , Quart . Appl . Mat h. , Vol 25( 1) ,
pp 83-95.
RUBINSTEIN, R. and ATLURI, S.N. (1983): ' Ob j e c t i v i t y of i ncr ement al
const i t ut i ve r el at i ons over f i n i t e t i me-st eps i n computati onal f i n i t e
def or mat i on pr ocesses' , Comp. Meth. Appl . Mech. and Eng., Vol 36, pp
277-290.
RUDIGGER, E. and RIECH, H. (1983): ' Exper i ment al and t h e o r e t i c a l
i nvest i gat i ons on the impact of deformable mi ssi l es onto r ei nf or ced
concrete sl abs' , 7th SMIRT Conf. (Chicago Aug. 1983), paper J8/ 3.
SCHOFIELD, A.H. and WROTH, C.P. ( 1968) : ' C r i t i c a l s t a t e s o i l
mechanics' , McGraw-Hi l l .
SCHREURS, P. (1983): 'Numerical si mul at i on of formi ng processes; The
use of Ar bi t r ar y Lagrangi an-Eul eri an (AEL) f or mul at i on and the Fi ni t e
Element met hod' , Doct or al t h e s i s , Techni sche Hogeschool Ei ndhoven,
Netherlands.
SHRIVE, N.G., ANDREWS, K.R.F. and ENGLAND, G.L. (1984): 'The i mpact
energy di ssi pat i on of cyl i ndr i cal systems' , i n ' St r uct ur al Impact and
Crashworthi ness' , Vol 2, J . Morton ed. , pp 544-554.
SIMO, O.C. and ORTIZ, M. (1985): 'A u n i f i e d approach t o f i n i t e
deformation el ast o- pl ast i c anal ysi s based on the use of Hyperel asti c
const i t ut i ve equat i ons' , t o appear i n Comp. Meth. Appl . Mech. & Eng.
SOUTHWELL, R.V. (194Q): ' Rel axat i on methods i n t h e o r e t i c a l physi cs' ,
Oxford Uni ver si t y press.
THOMAS, S.G., REID, S.R. and JOHNSON, W. (1976): ' Large def or mat i ons
of t h i n - wa l l e d c i r c u l a r t ubes under t r ansver sal l oadi ng' , I n t . J.
Mech. Sci . , Vol 18, pp 325-333.
THORNTON, P.H. and MAGEE, C.L. (1977): 'The i nt er pl ay of geometric and
mat er i al var i abl es i n energy abs or pt i on' , J. Eng. Mat er , and Tech.
(t ransact i ons ASME), Vol 99( Apr i l ) , pp 114-120.
TRUESDELL, C. and TOUPIN, P.A. (1960): 'The cl assi cal f i e l d t heor i es' ,
i n 'Handbuch der Physi k' , S. Fl ugge ( ed. ) , Vol I I I ( l ) , pp 226-858,
Spri nger-Verl ag, Ber l i n
TRUJILLO, D.H. (1972): 'An unc ondi t i onal l y st abl e ex pl i c i t al gor i t hm
f or s t r uc t ur al dynami cs' , I nt . J. Numer. Met h. Eng., Vol 11, pp 1579-
1592.
TRUJILLO, D.H. (1975): 'An unc ondi t i onal l y st abl e ex pl i c i t al gor i t hm
f or Fi ni t e Element heat conduction anal ysi s' , Nucl. Eng. & Design, Vol
32, pp 110-120.
UNDERWOOD, P. (1983): 'Dynamic r el axat i on' , i n 'Computational Methods
f or Tr ansi ent Anal ysi s' , T. Bel yt schko and T.J.R. Hughes eds. , Nor t h-
Hol l and, Amsterdam, pp 245-266.
VON MISES, R. (1913): ' Mechani k der f est en Koerper im p l a s t i c h
deformablen Zustand' , Goetti nger Nachr., math-phys. Kl . 1913, pp 582-
592.
WILKINS, M.L. (1964): ' Cal cul at i on of El a s t i c - p l a s t i c f l o w' , i n
'Methods .of Comput at i onal Physi cs' , Vol 3, B. Al der et a l . eds. ,
Academic press, New York, pp 211-263.
WILKINS, M.L. (1968): ' Thi rd progress report of l i ght armour program' ,
Lawrence Livermore l abor at or y, U. of Cal i f or ni a, r ept . UCRL-50460.
WILKINS, M.L. (1969): ' Cal cul at i on of El ast i c- pl ast i c f l ow' , Lawrence
Livermore l abor at or y, U. of Cal i f or ni a, r ept . UCRL-7322.
WILKINS, M.L. (1978): 'Mechanics of penet rat i on and per f or at i on' , I nt .
J. Eng. Sci . , Vol 16, pp 793-807.
WILKINS, M.L., BLUM, R.E., CRONSHAGEN, E. and GRANTHAM, P. (1975): 'A
method f or computer si mul at i on of problems i s sol i d mechanics and gas
dynamics i n t hree dimensions and t i me' , Lawrence Livermore Laboratory,
U. of Cal i f or ni a, r ept . UCRL-51574.
WILKINS, M.L. and GUINAN, M.W. (1973): 'Impact of cyl i nder s on a r i gi d
boundary' , J. Appl . Phys., Vol 44(3), pp 1200-1206.
WILKINS, M.L., STREIT, R.D., and REAUGH, J.E. (1980): ' Cumul at i ve
strain-damage model of duct i l e f r act ur e: si mul at i on and pr edi ct i on of
engi neer i ng f r a c t u r e t e s t s ' , Lawrence Li ver mor e l abor at or y , U. of
Cal i f or ni a, r ept . UCRL-53058.
YAGHMAI, S. and POPOV, E.P. (1971): ' I ncr ement al anal ysi s of l ar ge
def l ect i ons of shel l s of r evol ut i on' , I nt . J. Sol i ds & St r uct . , Vol
7, pp 1375-1393.
ZIEGLER, H. (1959): 'A modi f i c at i on t o Prager' s har deni ng r u l e ' ,
Quart. Appl . Math. 17, pp. 55-65.
ZIENKIEWICZ, O.C. (1977): 'The Fi n i t e El ement method (3rd e d i t i o n ) ' ,
McGraw-Hi l l , Berkshi re (UK).