Professional Documents
Culture Documents
.1
'
/'l..
,:5. l
~--
/1.
., .....
,//.( 11
v :._~ u:.::.~ e,
Jl
if
il
u,.
J._,....; .
1
t
....
Discontinu:um Mechanics
Using Finite and Discrete Elements
-----.-----
WITPRESS
------- --------, -
..
......
----------
WIT eLibrary
Making the latest research accessible, the WIT electronic-library foaturcs papers
presented at Wessex lnstitut~ ofTechnology's prestigious international conferences.
To access the library and view abstracts free of cbarge please visit www.witpress.com
..
-.
.,';
,~t!
~~~'!" . . .
. - -----
,,.,.
Discontinuum M echanics ,
Using Finite and Discrete Ele1nents
' ,
S. Mohammadi
University o/Tehran. Imn
(_',CJ
or>
/' -~,.~~i~E;~~}
WITPRESS
Southampton, Boston
Author: S. Mohammadi
~t'',
, r
Contents
\
.......
Published by
Chapter 1
Introduction
WITPress
Ashurst Lodge, Ashurst, Southampton, S040 7 AA, UK
Tel: 44 (O) 238 029 3223; Fax: 44 (O) 238 029 2853
E-Mail: witpress@witpress.com
http://www.witpress.com
1.1
1.2
Alternative Approaches
1.3
1.4
Notations
Chapter 2
2.1
11
12
Geomechanical Applications .
2.1.1
Rock BlRSting . . . .
12
2.L~
Mltti111~
13
2.1.3
14
2.1.4
16,
2.1.5
16
2.1.6
Falling
2.1.7
Porous Media .
Hock lhm1tM
2.2
Object~
17
,,
18
20
Granular Materials . . .
2.2.1
20
2.2.2
Superquadric Elements .
21
Ali rights reserved. No part of this publication may be reproduced, stored in a retrieval
systcm. or transmitted in any fonn or by any means, electronic, mechanical, photocopying,
rccording, or olhcrwisc, without thc prior writtcn pcrmission of the Publisher.
2.3
Penetration of a Missile
....
21
21;
2.11
2.5
2.3.2
Metal Cutting
22
4.2.4
73
2.3.3
Demolition of Structures .
24
4.2.5
Rate of Deforroation ..
74
2.3.4
Composites
24
4.2.6
Stress Measures . . . .
75 :' -
2.3.5
Masonry Structures
26
4.2. 7
77
2'.3.6
Repair Modellings
30
4.2.8
Multiplicative Decoroposition
79
80
32
4.3. l
Equilibrium Equation ..
80
34
4.3.2
Variational Forro . . . .
82
32
:~.4.1
Shockwa.ve' AnalyHiH
Other Applications .
Chapter 3
3.1
Initial/Bounda.ry Value...,Pr2blem
Particulate Flow
Introduction ..
35
35
35
3.3
Constraint Enforceroent
40
:J.a.1
1m ><nietrahility Constraiut
42
3.3.2
Penalty Method
..
43
3.3.3
Lagrangian Multiplier
47
:t.:\..l
tm
3.3.5
Variational Forros .
56
3.5
58
3.5.1
60
Contact Instability
N onlinear FE Discretization
4.4
85
4.5
87
4.5.1
87
4.5,2
88
4.5.3
89
Normal Contact
93
Chapter 5
5.1
93\. ''
Introduction . . . . . . .
52
3.4
Chapter 4
4.3
5.2.1
Original Algorithms .
5.2.2
DYNA 2D Formulation
'
'
'
95
9.9
;' :;;;;~i8~{~?\4k ;f:;
. 104
5.3.1
. 104
69
5.3.2
5.3.3
Contact Force . . . . ,, . . .
4.1
Introduction .
69
4.2
69
4.2.1
The Motion .
'1.2.2
4.2.3
5.3
..
'
10~
..... 1oi<:,}'
. 108 y~f,>:
70
5.4.1
Contact Geometry . . . .
Ddonnation Gradiont
71
5.4.2
Global Senrch
Polar Decoroposition .
72
5.4.3
5.4
"-.._
o:;.
Algorlth~
10~ ;,:~-0;
/,'}i_::,::
. 110 ','
5.4.4
. 111
G.4.G
. 115
r r
".,
Frictional Contact
121
6.1
Introduction . . . . . . . . .
. 121
6.2
. 121
6.4
181
Introduction . . . . . . . . . . .
. 181
7.1.1
Histmical Development
. 182
7.1.2
Applications
. 183 '
7.1.3
Specifications
. 183
. 185
6.2.l
. 121
7.2.1
Contact Geometry
. 187
6.2.2
Basic Kinematics . . . . .
. 122
7.2.2
. 188'
o.2.:3
. 126
7.2.3
Buffer Zone . . . . . .
. .)PQ
.. 194'.
.. 135
6.3.1
. 137
7.2.5
n.~t2
St.ifflWHH Contrilmtion
. . 138
7.2.()
Spaco
(i.a.:~
Cu11HiHl.c11t Li11eu.r:u~t.io11
. 142
7.2.7
. 147
6.4.1
Preliminaries . . . . . . . . .
. 147
().4,
. 147
6.4.3
Numerical Integration
. 150
6.4.4
FE Discretization . . .
. 152
6.6
7.1
... 118
Chapter 6
6.:3
Chapter 7
. 155
. 155
7.3
n(~COUlJ>OHitlon
Object Representation
.196
<'
203
.#t
. 22tc\ ...
~;:;_.t
f_; '(;" ..
"'
7.3.1
Circular Disks
222~,:..
7.3.2
Disk Clusters .
. 230<::
7.3.3
.. 233;>:.
7.3.4
Superquadric Objects .
. 239."
7.3.5
~,; .~t.fi' :.
.. 242 '
6.5.1
6.5.2
. 159
8.1
Intro<luction ..
. 245
6.5.3
. 163
8.2
Mesh Dependency
. 246
6.5.4
. 164
8.3
Fracture Mechanics .
. 247
. 166
8.4
. 249
6.6.1
. 166
8.5
Damage Mechanics . . .
. 253
6.6.2
. 170
8.5.1
. 257
Chapter 8
Remeshing Techniques
8.G
. 258
. 259
8.6.1
Tsai-Wu Model . . .
8.6.2
8.8
.. 260
... 267
8.7.l
Loss of Uniqueness . . . .
. 269
8.7.2
Discontinuous Bifurcation
. 272
Hmneshing Algorithm .
Appen<lix A
UTDem code
Preface
. 276
279
. 279
. 279
. 285
l1mylit
11111.
n smgle word'
tbl.;:::>
.)
."
-,11J1:
The book can be classified into three parts. The first part is dedica.te4;
to the basic concepts of constraint definitions and various methods foi
constraint enforcement as well as a short di~(11sfo11 1111 the finite
discretization ~d necessary finite deformation theory as the bases uf
combined finite/discrete element method.
element
the .
and
and
their
asimciated
111r1,1t1
1n1f'f'il11111ly u.1lrnow)t!tl}1.,i11R
To:
f.11t11tl
QJ
I would like to extend my best regards to Dr. S.S. Yazdi and to the
Farinkav Engineering Research Company for the software support. The
techn ical su pport of the Computer Centre, Department of Civil
EHgiueering of Univ<~rsity of Tehra11 for facilitating the preparation of the
nH1.1111script H also ndmowledged .
.F'inally, rny special thanks to my family for their never ending
support. I have spent many hours preparing this text; hours that had to
be spent for my wife and little 'sogol'. Without their understanding and
love this book may not ever have been written.
..
;i~ ;Mohammadi
Tehran
Mansoureh
&J
$ogol
Chapter 1
Introduction
1.1
Discontinuum Mechanics,
Why?
...~-
A mrnd hourglnHH, 1111~1 f(1r 11mtttrloH boeu ti Hl11qil rn1111111111 t it11 mmwurlng
tool as well as a relatively accurate reliable mt~asure fur f>cientific experiments. Nevertheless, until very recently, even at the age of supercomputers
and efficient numerical methods such as the finite element method, it has
not been possible to accurately model this simple phenomenon. A highly
varying geometry, and interacting objects (sands) were the main obsta.eles
to any efficient approach (see Figure 1.1).
An interesting set of problems which have recently attracted special at
tention, includes the general behaviour of granular materials. In this class of
problema, large number of interacting bcidies, usually simple rigid elementS, . ,
are interacting in a donmin which will govern tite p;t'11crnl re1-1ponse of the'.
medium through these individual interactious. 'l'l1t~ l11'sL 1'x11mple, me.y be,
the filling or emptying a silo with/from granular materials as depicted in
Figure 1.2 [1].
Discontinuurn Mephanics
T1me:23.1811
Chapter 1. Introduction
......
Fragmentation
Delamination
is
Even with the siguificant advances that aw 11ei11g 1111:1.de with these new ~..
materials, there is still a great deal unknown about their mechanical cha.rO:!.
teristics. These unknowns can be attributed to the large munber of different
ma.terials being used, to t.he contribution of 111inor11<wha11knl interaction1'
of difforout. compowmf.H t.o thu ovorall boluwi1111r 1111d t.lw vorlot;y of fallu~
mechanisms involvcd. lt is also well known t hnt tlw pn~::;euce or inhorri(;.i
geneities plays an important role in the behaviour of these materials.
One of the major problems that dominates the design procedure fr ;; ;,
structural applications is the strength of material to transverse impact which ;: .,
may cause substantial internal damage. Thes<~ impact. dama.ges are ofte~ ';\:Y
embedded inside the material, hardly detectable by the uaked eye (2, 3]. Ac.-. :\.::
cordingly, detection of impact damage is critically important in maintainfug :/'.'
the integrity of the structures n service. By examining the numerous cori- :..
tributions to this area of research [4, 5, 6, 7, 8}, it is evident that impact C'
loading represents a highly complex phenomenon cmnpused of several inter- ',
acting processcs.
Figure L3 represents the progressive fracturh1g and fragmentation ph~
nomena in a typical composite specimen subjected to impact loading. 'fh!s
schematic representation, is perhaps only related to the failure observed ..
in high velocity impact. For low velocity impnct., howPv<~r, while it is unlikely that extensive fragmentation will be observed, material fracture and
delamination will be the likely modes of failure that exist.
It has been shown that interna! cracking of a structure subj~cted to
low or high velocity impact loadings, are progressive phenomena which may .
rapidly propagate throughout the component. This rnight rc~mlt in the cnF
Discontinuum Mechanics
11
Chapter l. Introduction
Transltion Interface
Bondio Interlllce
n
LJ
ation of new totally separated parts, which interact with their surrounding
rngionH. Conseq11<~ntly; a powcrful scheme is required to be a.ble to monitor
the fracturing process and to effectively model both individual and interaction behaviour. Recent deyelopments of discrete element methods have_
prepared the ground for a new approach to modelling this behaviour based
on the concepts of the discontinuum mechanics.
The tradit.ioual a.pproach to thesimulation of i:>lress distributions in arbitrary shaped components under possible nonlinear geometric and material
conditions is by finite element techniques. However, the traditional finite element method (FEM) is rooted in the concepts of continuum mechanics and
is not suited to general fracture propagation problems since it necessitates
that discontinuities be propagated along the predefined element boundaries.
The.corresponding elasticity and fracture mechanics concepts are applicable oniy in situations dealing with a single crack or a low-fractured area
without any fragmentation [9]. In contrast, the discrete element method
(DEM) is specifically designed to solve problems that exhibit strong discontinuities in material and geometric behaviour (10]. The discrete element
rnetht)d idealizes the whole medium into an assemblage of individual bodies, which in addition to their own deformable response, interact with each
other (through a contact type interaction) to perform the same response as
the medium [11]. A far more natural and general approach is offered by a
combination of discrete element and finite element methods.
LJ
It should be emphasized that drawing such a concl usion <loes ndt necessarily require a fracture analysis. A relatively fast elastoplastic analysis can
be performed by simple 8hell or solid. elements t.o detmmine the different
rogl1111M 1H~cordi11g to HUlll<' offect.lv Htl'688 or nt 111i11 <rit.n1 iu. Tho prtt<llc!ted
fractured/delaminated regionR may then be exarnined i11 t.lw lnter stages of
the analysis through the combined finite/discrete element algorithm. The
regions' boundaries should be further extended if the cracking reaches
the boundaries of the fractured region [12, 13].
to
ti
Discontinuum Mechanics
Chapter 1. Introduction
Interactions between finite elements (not those which are used for DEM
discretization) and discrete elements are modeHed .by.tra,n.siti()n interfaces.
A transition interface is define<f as a bonding interface wi~b Ye'l'Y high 'bonding strengths which prevent debonding under all stre~s conditions.
Table 1.1: Comparison of FEM aiid BE!'v1 charn.ct.cristics.
11
1.2
FEM
Category
Mesh generation
entire domain
Solut.ion
entirc domain
sparse
usually symmetric
easier to evaluate
widely acceptable
ea.sy
Integrals
Nonlinearity
Implementation
..
BE1vI
l.>urii1tar.v cuii<litions
fully populated
non-symmetric .
far harder to evaluate
only linear problema
much more difficult
Alternative Approaches
. i.
E(x)
w-.a
w.a
=E0 exp ( Jl x )
E0 ~
V=
11
bou11Jaries
fi rst. on l,j 1<~ houndary
11'1
(w < 8)
0.3
.1 . ,J;;:;:..
,.
..
_,
...'.
Discontinuum Mechanics
-4,;;:
1.3
n
u
Cha1Jter 2 pr(Jvides sorne representative results and data on various applicatious thu.t: involve the general concepts of the contact mechanics and
related topics.
Chapter 3, reviews the available methods for enforcing a constraint relation onto a set of simultaneous equations. A detailed study of the penalty
method provides basic information for use in other chapters.
In Chapter 4, the main general computational aspects of the analysis will
be discussed which includi!s a comprehensive discussion on large deformation
theory, as is usually the case in the practica! simulation of impact loading
of structures. Then the basics of the finite element discretization will be
cxpla.incd. A short review of tho cxplicit algorithm as the timo integratton
scheme adopted will close the chapter.
,_,
'
Chapter 5 is devoted to providing historical and state of the art algorithms for normal contact problems. The fundamental aspects of classical
theory of normal contaet will he diHcRi-mci and practica.1 ambip;nitie~ will be
itJdret>s<.xl.
The relatively complicated theory of frictional contact will be explained
in Chapter 6. It includes discussions on various frictional contact phenomena and provides computational procedures for updating material state
variables for hardening/softening conditions and in general non-associated
formulation.
Chapter 7 is dedicated to discontinuum approaches of contact mechanics.
It includes a state of the art discussion on contact detection procedures for
problems with a large number of low or highly dense interacting objects.
Advanced object representation methods wiU also be explained which are
closely associated to contact detection procedures.
Chapter 8 summarizes the remeshing techniques and principies of crack
initiation and propagation phenomena and discussions on the problem of
size effect in fracture analysis of solids, softening and damage mechanics
methodologieH n.s well as lntroducing a three dimensional anlsotropic mar
terial model which consists of an 'stress return algorithm and deriving the
consistent tangent modulus.
Chapter 1. Introduction
9.
1.4
Notations
t, i.
4. Second order temmrs are represented us liold f11n~ upper case itallc
letters, such as D, N, A. A few Greek bo1dmath letters, however,
are used for stresses (u) and strains (e).
t =un
or
t = Tn
a=Ee
or
H=DN
axb
9. The tensor product of two vectors is a maJrix d<'li1H~d as
A=xy
Aij
XiYj
ternmrs i8
represented
by :
Chapter 2
Discontinuum Mechanics -
A Review
To attain a realistic ovorview of the extent a co11 t.acl. lin~a '( 1n.lgorithm can be
used for analysing variom; academic, engineeri11g awl iwlustriu.l problems,
a quick review of potential applications are provided. It is not intended,
primarily, to compare the results w.ith available data in the literature, as it
is usual in academic papers, but to illustrate to the reader the applicability
of the method to different applications that rnay be analysed by the use
of the computational cficoutinuum mechanics. It. is 1t!Ho airncd at sparklng
new ideas for further research and future challengcs in this subject.
This chapter reviews the following engineering applications, amongst
many others, which are currently being researclwd in many research instit.11tinm1 thro11p;hn11t. t.lw workl,
.
Geomechanical applications
Granular materia.Is
Impact analysis (progressfve fracturing)
Particulate flow
Computer graphics
11
I'..!
1
1
1
1
1
Discontl1iuu111 Mcd1auics
2.1
Geomechanical Applications
2.1.1
Rock Blasting
"
13
.,~'+-:/+:*,'
.
,....
1
'
' ....
'~
'
'
1("')
4.100.
~
,..._.owi
l:SOO(m)
Ul(m)
711n)
111("")
-11.00
\11' ......
! ,...........
0000
4.000.
4.100..
~
lllltlmlOO:'!
l'Mldll0004
1
1
1
1
1
...a-0.0JJ
,,
\.
'
2.1.2
D111..!1 hla.o.;ting is a very frcq11ent rock blasting operation. In this examilc~ a :lD prublem is modelled. The initial geometrical configuration and
pro~essivP fracture patterns in successive times from the initial blast are
rlepi2ted in Figure 2.1.
Discontinuum Mechanics
14
11
Figure 2.3: Exp<~rirnental failttre rcsults nnd numerical simulation of a simple rock bursting problem [20).
th~
The rock bursting process can well be simulated by the combined finite/discrete element technology [19]. Because of loss of many peoples Uves
after the mining rock b,umps, this field has remained a very important and
active part of mining 4ineering.
Figure 2.3 illustrates sorne experimental aml numerical simulation results reported by Hazzard et al. [20) based on [21, 22, 23].
2.1.3
....'
'
~-'
In this example, a wall of 2600 equi-size spheres betweeri two vertical plates
was rnodelled by O'Connor (24]. Gravity was first applied for 1000 iterations
ata time step of one milliseconds. At t = ls, the left boundary was suddenly
removed. The ensuing collapse configurations are shown in Figure 2.4. The
Hirnula,tion WClH l'Ull for a total Of 3000 iteratiOllS [24].
15
Discontinuum Mechanics
16
Figure 2.6: Remeshing procesa and the 45 shear band development in a'
tensile plate undergoing large lateral necking phenomenon.
Figure 2.5: DDA slope stability analysis for various discontinuity patterns
and friction angles [26].
2.1.4
Figure 2.7: Two shear band slope instability problrn1s [28, 29].
Figure 2.5 illustrates the results of slope stability analysis for various
discontinuity patterns and friction angles based on the discontinuous cieformation method as reported by MacLaughlin et al. [26].
Figure 2. 7 illustrates two different examples of slope instability simulations performed by Stead et al. [28] and Cramer et al. [29] in two and
three dimensions, respectively. In the 2D case, an h-adaptivc finite element
method has been adopted, whereas in the 3D example, only large deformation theory has heen considered.
2.1.5
2.1.6
Falling Objects
lJ
In contrast to the DDA method, a shear band slope stability analysis may
be performed by using a fully deformable nonlinear finite element simulation. An adaptive remeshing scheme has to be employed to avoid excess
distortions of the finite elements close to the highly deformed shearing band.
Figure 2.8 shows the application -of the discrete element approach to a full
three dimensional simulation involving randomly shaped grains falling under
gravity into a box. Two sides of the box on which sorne particles have come
to rest have been removed for rendering purposes [30, 24].
-------------------------------
Discontinuum Mechanics
18
------
2 .1. 7
LJ
Porous Media
19
r
20
Discontinuum Mechanics
21
:;{
{~':
2.2
Granular Materials
2.2.1
2.2.2
Superquadric Elements
11.111plo, t.lin predict.iou of pressuro nnd flow in siloa has been iuvstlga.tod
11tiliiing t.ho discrnle element nwthod. Silo and granular material are both
modelled in this approach as depicted in Figure 2.11 [33, 32). The results of
typical cond ucted analyses may be used to guide the silo design procedures - ~
by pointing out any u.Jl8onticipated loading conditions and pressure distributions which might altse during operation, as well as phenomena such as
arching, different filling/emptying regimes, seismic loading, etc.
Ttae:7.34281
Figure 2.12: Discrete element modelling ofa typical deposilion problem [3~j.
TtH:U.1111
2.3
41
)'
f
,b'
,:..
"':
~.
2.3.1
u
[33].
Penetration of a Missile
Structural design of a shelter, a.rmoured military Pqnipnwut and safety measures for bullet-proof vests may force a designer/a.nalyst: t.o check for impenetrability response of a structure subjected to a high velocity object.
A complete analysis of an object penetrating a structure and developing
22
Discontinuurn. Mecha.pies
23'
cxtensive damage in it, has only become possible by the use of combined
finitc/dii-;crnte clement techniques.
Figure 2.13, illustrates how the crack patterns are propagated within a
typical ceramic plate as a bullet penetrates the plate in different time steps.
Figure 2.14: Shear band development and chip creation in a metal cuti.i~s
application [34].
, Fig11rc~ 2.1 :~: Progrcssive fracturinp; in n structurn impa.cted by a high velocity h11lld.,
2.3.2
Metal Cutting
1'
i:
1'
l__.,J
\\
\\
L'
24
Discontinuum Mechanics
25.
Rimulat:ion of tlw whole cutting process from the first contact until the onset
of total separation [35].
'
2.3.3
Demolition of Structures
2.3.4
0.000
0.200
Composites
real time
=11.200
26
Discontinuum Mechanics
27
Figure 2.17: Different failure modes in composites subjected to impact loading [39].
11n Mesh
Interface
betweerr1a~rs
2 and :3.
2.3.5
Masonry Structures
h) T
0.00012 sec.
'28
Discontinuum Mecban.ics
29.'
FiKlll'U :J.20 llhwt.rn.t.nH t.hn ld11nll~ud prohJ0111 ol' f.110 f11il11n~ of 11 t.wo tept\ll
bridge studied and simulated by Petrlnic [1). Tliu prolilrnn was treated as
Figure 2.21: A two span masonry bridge and the finite element model [44).
Discontinuum Mechanics
30
..1 .....
2.3.6
Repair Modellings
31
---
- - - - - -------------------
Discontinuum Mechanics
Figure 2.26: Finite element simulation of anchor pull out test [43J.
2.4
Particulate Flow
2.4.1
Shockwave Analysis
A wall of 2GOO cqui-sizc spheres is constructed by placing the spheres between two vertical plates. Gravity was first applied for 1000 iterations at
a time step of one millisecond. A stream of slightly denser particles of the
same size- are dropped under gravity from above the wall. On impact, the
stream creates a series of stress waves which are shown in Figure 2.27. Later
in the simulation, the wavefront has been reflected off the boundary walls
and interferes with thc waves still originating at the source of disturbance
[24J.
33
Discontinuum Mechanics
2.5
Other Applications
1t
1
Chapter 3
Constraint Enforcing
Methods
1
1
3.1
./.;:'
Introduction
:,; . r
Many different methods have been developed for enforcing a constraint con,- :e:
dition on the governing equation of a well established physical behaviour. ~: .
In this chapter, the following four rnethods for enforcernent of constraints
within a finite element analysis are reviewed:
Penalty method
Lagrange multiplier method
Perturbed Lagrangian rnethod
Augmented Lagrangian method
1
The final part of the chapter describes the peualLy mcthod in more detall ,
and explains the stability of'method for contact analysis and the associated
contact damping techniques.
3. 2
Definition of a Constraint
LJ
35
36
Discontinuum Mechanics
37
Bodyl
Body2
-UB
UA
{ UA.-UO
Contact
Constraint
=o
=0
-1
{3.3);
o
Fiv;11re :3.1 repro:.;entH a. t~~cal impcnetrability comitraint between two contacting bodies. This cons~aint defines the necessary condition~ to prevent
the bodies from penetrating each other.
where
[~
Cu=Q
-1
(3.4)
Example 1
According to Figure 3.2, a series of nodes, i.e. A, B and C, are to be
constrained to stay in a straight line, e.g. for symmetry reasons. The
constraint equations may then be formed as:
Example 2
Figure 3.3 depicts a model combined of a continuum m<sh anda structural
mesh. To define the connection between a structural element to continuum
elements, i.e. to link the rotational freedom of the structural element, </>2
in Figure 3.4, to the displacement freedoms u 1 , u2 and u3 of the continuum
model, the rotational equilibrium about node 2 can be cowddered:
=O
(3.5)'
38
Discontinuum Mechanics
39
just in contact
u =+0.J
~1--------2~3-------4k1 =JO
k1 =JO
Continuum
Element
Structural
Element
l
;
.
[-l
-1)
U ]
~:
(3.6)
Example 3
Two straight bars which are just in contact are depicted in Figure 3.5.
Each node has a single degree of freedom along the bar directfon. A o.i
unidirectional rightward displacement is applied to node 1 of the left bar. ;
Each bar behaves as a linear spring, so
-10
10
(3.7)
l
The assembled system of equations will be
10
-10
_ _....... U3
[
:
o
o
-10
10
o
o
10
0
-10
-10
10
l[ l [ l
U U2 0.1
U;l
U4
=- ()
t
0
(3.8)
f.1
0.1
=0
(3.9)
40
Discontinuum Mechanics
41
0.1
1-----i
l-----..2
n
u
Figure 3.6 shows the deformed shapes of the bars for this analysis, which
clearly shows overlapping the elements.
Body2
(3.10)
We will later use this simple example to verify the methods adopted as
constraint enforcing methods.
3.3
(3.12)
Constraint Enforcement
In an alternative approach, imposing the constraint equation is performed by means of variational principies [50]. In this approach, we can
introduce the constraint by forming a constraint functional, ncon' in addition to the functional of the system (51)
l
1
'
l_J
. (3.11)
(3.14). :;
'.,:'
'J~i~~,,k;
i,
Discontinuum Mechanics
which is valid providing TI = O and the constraint is imposed, simultaneously. Further development of the formulation requires a clear definition of
the constraint equation, e, and the constraint functional, rrcon' in equations
(8.13), (3.14).
e= sign(n)[l
42
43
- 1]
(3.17)
(3.18)
3.3.1
Impenetrability Constraint
This constraint defines the necessary conditions requirE;?d for preventing the
contacting bodies from entering into each others boundaries. In this section
only a frictionless contact is considered.
Figure 3.8 shows two bodies, 1 and 2, at two successive time steps. At
time t, the bodies are yet to make contact and there is no penetration.
However in the next time step, the deformation states that without any
additional constraint the bodies will penetrate each other.
deformation,
(1)
be written as
re= r 1 n r 2
3.3.2
(3.15).
where re is the common contact surface of two bodies, x 1 and x 2 are the
<Mormed positions of bodies 1 and 2, respectively, nis the normal to body
2 aml g is the gap function. Equation (3.15) specifies the gap in normal
direction.
In a one dimensional contact between two nodes of the bodies, the equations are sirnplified
g=Cu
(3.19)'
Penalty Method
-g-(2)
on
=o
tem and solving the resulting coupled set of equations. Hen\ four methods
whlch lmvo IHi<'n froq11011t.ly llf.16d ln vn.rlous co11t.1w1. 1w11l.vN1,~1 1m1 closcrlbad:
timet+dt
-.:::--..-----~--&
...~
~ay
=:u 2 -u 1
~
~.
(3.16)
Penalty Method
The Penalty method was probably the first approach adopted for a constraint enforcing method. It was developed by Hallquist and his colleagues
in Lawrence Livermore National Laboratory during the late seventies for
modelling impact/contact problema.
1
{3.21)
Discontinuum Mechanics
44
LJ
(3.22)
- [IT
-+gT a8gl
- fo
To maintain equilibrium,
8u
Bu
should be equal to
(3.23)
0.5E
The first term on the right hand side of (3.23) is the well known stiffness
equation
an
au
(3.24)
g=Cu-Q
(3.29)
An example
(3.25)
To illustrate how to use the penalty method for enforcing a constraint equation, Example 3 of Section 3.2 is re-considered (Figure 3.9).
(3.26)
(3.27}
[K + CTac] u= R+CTaQ
O.
-=Ku-R
45
il=
(3.28)
~
-1------2-3----4-:
ki =JO
k2 =JO
l
Discontinuum Mechanics
46
47
0.0333
(3.30)
with
(3.31)
0.0667
...
aCT C=a [ _1
1
{3.32)
the results show the existence of sorne penetrat.ion of bar 1 into the bar 2 ., ,
(See Figure 3.10).
10
-~10
Table 3.1 summarizes the results obtained for the same equation using
different penalty numbers. It is dearly seen that by increasing the penalty
number the solution converges to the exact solution.
-10
o
10 + a . -a
-a
lO+a
(3.33)
g_,-10
-a
u2
and
-a
10 + O:
l = l
u2 ]
U3
1
0
(3.34)
3.3.3
u3
Lagrangian Multiplier
U3
0.0333
0.0476
0.0497
0.0500
U2
0.0667
0.0523
0.0502
0.0500
100
1000
10000
a
10
?/,'},
[
'll3
l=
_1 [ 20 10
300 10 20
l [o l
1
(3.37)
= [ 0.0667 ]
0.0333
{3.-36)
48
Discontinuum Mechanics
the force required to enforce the constraint. In this method, new variables
L_)
are added to the system of equations and the constraints a.re satisfied if
g = O for which = II.
(53].
Lagrange multiplier method did not necessarily preserve a smooth
force distribution across interfaces. The lack of a smooth force field
has excited the zero energy modes in sorne of the prirnitive elements
-= an-au + TY]
au
.X -
ull
8.xfi = g.X
[54].
(3.38)
Example
(3.39)
()11,
(3.43)
wlt.h
(a.41)
(3.44}
-10
10
10
-10
-10
10
~l
-1
10
-10
l[~~ l [~ l
U
U4
0.1
= ()
f.1
(3.45)
'-'.
{3.46)
u2, U3
as
(3.47)
u2
and ua.
50
.
3.3.4
Discontinuum Mechanics
111111111111111
Cl1apter 3.
Constl'l:tint Enforcing
1888J'988674'2
[~ ~][~]=[~]
Methods
lJ
(3.48)
{3.49)
Cli
u and ,\
-= [Il
8gl 8u
-+,\
-Bu
Bu
1'
uIT
8,\fi
[g - A..\) 8,\
(3.50)
(3.55)
Ku-R+C T ..\=O
(3.52)
--
Example
Cu-Q-A,\ =0
w Itere
(3.53)
Taking up the Example 3 of Section 3.2 (Figure 3.9), the assembled system
of eq uations will be
an -K
au
u- R
[ -10
10
g=Cu-Q
!!JL
au =C
Expressing Equations (3.52, 3.53) in a matrix forro
o
o
o
-10
10
o
o
o
o
o
o
10
-10
0
-1
1
-10
10
-1
.=.!
a
l [ll
[U=
U2 0.1
U3
U4
=0
,\
(3.56)- '
Discontinuum Mechanics
~o
1o1
-1
~1 J [, ~~A J = . [ O~ ]
=-!.
(3.57)
. Table 3.2 summarizes the results obtained for the same equation using
chfferent penalty numbers. lt is clearly seen that by increasing the penalty
number the solution converges to the exact solution.
u
Q
10
100
1000
10000
U2
U3
0.0667
0.0523
0.0502
0.0500
0.0333
0.0476
0.0497
0.0500
A
0.0333
0.4762
0.4975
0.4998
(3.62) .
or
(K
a.:J.5
53 :
+ cT nC) u = R -
(3.63)
The objective. of this method, which was developed in the early nineties
is to minimize the disadvantages of the penalty and Lagrangian multiplie;
methods by combining the two methods. The method is independent of
the magnitude of penalty number and introduces no extra equation to the
system.
(3.64)
1
+ tXT g + ?,YT
ag
(3.58)
~A
Following the above strategy, the total potential energy of the system in
the Lagrange multiplier form can be written as
1 T
Il(u, .\) = Il(u) + -g ag
minimizing the
(3.59)
~3.66)
As a Lagrange method
1
(3.60)
(3.67)
Discontinuum MechMics
54
=O
1.0
Initialize Al
2.0
+ CT aC)u =
R-CT(aQ-Ak)
Example
fe= 6..X = ag
2.3
Working on the same Example 3 of Section 3.2 (Figure 3.9), the active
2.5
An iterative solution procedure is required. However, for contact problems an iterative procedure will usually be needrnl due to other con;:
siderations.
The algorithm can be used effectively for applications where the con~
tact pressures become very large in comparison with the material elastic parameters, i.e. with rubber materials.
(K
= Ak + 6A
10 + a
-a
GOTO 2.0
-a
10 + a
l[
u2 ]
u3
=[1]
(3.68)
- .\ [ -1 ]
1
20
[. -10
20
U2 ]
U3
= [ 1+A
-.\
2+,\
30
Iteration 1
.A1 == O
(3. 69 )
----+ {
(3.67)
----+
U2
U3
Iteration 2
= 0.06~7
= 0.0333
+ Li.A =O -
( 3.69 )
--+ {. U2
= 0.0556
U3
= 0.0444
(3.67}
----+
Iteration 3
( 3.69 )
Li,X
::::=
--().:333
0.3333 = -0.333
--0.1117
--+ {
U2
U3
= 0.05183
= 0.0482
(3.70)
1- ,\
(3.64).
(3.69)
[ u~-]:..!_ [ l
-10 ] [
=O
56
Discontinuum Mechanics
(3.67)
---4
Iteration 4
(3 . 69 )
A4
~{
= ,\3 +
U2.
U3
Iteration 5
=10(0.0482 -
~,\ === ng
As
(3.69) ~ {
:e-.::
= -0.03663
Penalty Method
= 0.05061
= .04939
ng = 10(0.04939 - 0.05061) = -0.01222
= ,\4 + 6,\ =
7.t2
0.05183)
(3.73) ..
= 0.0502
n:1 ~:=
0.0498
(3.74)
Lagrangian Multiplier
Here, the constraint term is introduced by the following functional
3.4
LJ
-Variational Forms
(3. 75)
(3.76}
= Il(u)
+ nc
n(c, u, ... )
(3.71)
(3.77)
where TI( u) is the functional of the unconstrained system. The variation of
the new functional leads to the general variational forrn of the constrained
system
8Il +mcon
(3.72)
Here, only a very brief review of the main functionals for each method
is provided.
(3.78)'.
58
Discontinuum Mechanics
59
CT(u)pC(u)dO
Oand
(3.79)
3.5
- U1
= 0
on
re
(3.81)
owcon
={
ro
agg(u) da
(3.82).
(3.80}
a) Be/ore contact
g = [a::2 - x 1] n 2: O
C :.= U2
Target segment
The general form of equation (3.80) for canta.et between two bodies may
tJwu be defi11ed by [5G, 52)
wcon = {
con
0
og u da
OU
Attention is now focused on a single boundary node in conta.ct to formulate the residual contribution of conta.ct coni;trnint., r'~. The component
form of the virtual work of the contact forces associated to the contact node:
is then given by [57]:
()U
Discontlnuum Mechanlcs
'' '
1.;'.:i',
t~.
satisfied by this method. This is clearly observed from Equatlon (3.21) '. {:
where a non-zero term is added to the total potential energy of the system, . ,
Without any additional treatment, the penalty method will cause the sy&- <
tem to gain energy artificially, although sometimes this extra energy is '
compensation for the loss of deformation energy dueto contact penetration..
//'
>o
otherwise
{3.88)
where g is the normal penetration. One may expcct tlw central dU,ference
scheme to be stable when applied with
J(
(3.86)
where a is the penalty term matrix, which can vary for normal and tangential gaps and even between single contact nodes. The corresponding
recovered residual force is then evaluated as:
~l !!llt2 < 4
m
(3.89)
(3.87)
The partial derivative part of equation (3.87) defines the direction and
distributiou of normal and tangential contact forces. The calculated contact
force has then to be distributed to the target and contactar nodes.
The potential normal and tangential gaps for each contacting couple are
evaluated by monitoring the coordinates of contacting couple nodes in ~ach
time step. Then by projecting the coordinates in the current and previous
time steps to a ref~ence con:figuration, the possible gaps are calculated
(Figure 3.12). Note that for a debonding problem, both penetration and
release normal gaps (positive and negative) are valid.
1
,,
3.5.1
Contact Instability
Figure 3.13: A material point entering and leaving a. ri;id wnll at successive
time steps [58].
Ata time step when the material point is about to enter the wall boundaries, for sorne time t 1 < ll.t no contact force exists to resist penetration.
62
Discontinuum Mechanics
Thn rns11lt.ing penetration genera.tes contact force and stiffness for subsequent time steps .. When the material is leaving the waH boundaries, the co~
tact force continues to be ptishing out the material point from the boundary
for sorne time t2 < .t, although there is no penetration any longer. Consequently, each time the hall enters and leaves the wall boundary, it gains
so me additional energy [58].
Figure 3.14 mustrates a simple contact problem in which a material
point of mass m = 0.lkg is moving between two parallel walls with an
initial velocity of v = lOm/ s. The walls are on the distance l = 7.5m from
each other and the penalty coefficient a: = le + 06.
63
45
35
+------.+-----
{ 30
25
i> 20
1
1
1
1
100
300
400
500
600
700
n(number of colllslons)
v=JO mis
----
200
/ .5m
1
Figure 3.15: Velocity of the material point after it passes n times from its
original, position.
Iu this method, at each Lime step the conta.ct 1<111<~ d11n t.o t.ho peuetra.tlon
is calculated from the modified version of (3.88) [10]
Figure 3.15 shows the velocity of the material point after it has passed
through the starting position n times. The results are obtained by unmodified central difference time integration scheme (3.88)[59].
o
if gt > O and gt+b.t < O
LJ
. It is clearly seen that as the Ll.t is increased, the velocity of the material
point is also rapidly increased, i.e. the system becomes faster and faster,
representing the system gaining extra energy during the time steps. Larger
values of D.t leads to larger values of ti < D.t for which larger energy values
are added to the system.
LJ
'
L_J
con=
(3.90)
To avoid instability, the evaluation of the contact force for the specified
domains has to be modified using a damping scheme. In the following, four
different sch<mrns whkh ha.ve been used to sta.bilize the conta.ct analysis, are
briefiy reviewed.
is negligible [lOJ.
64
Discontinuum Mechanics
65
where mi; = min( mi, mi) is the minimum mass of the contactirig bodies i
and j. (are input parameters controlling the penalty term, usually taken
0.04.
e,
The nonlinear contact force can then be written in terms of the g and
the field zone,
con= a
f ield -
(3.91)
{3.96).
In Equation (3.96), mmin and mma:i: are the mnimum and maximum
mass for all the bodies, respectively. However, if the penalty value or d&Dlping coefficient is too large, this acherne might not work properly.
Figure 3.16 compares the velocity of material point. of previous example
for a range of damping coefficient for two different time steps. The ~
sults have been obtained from Equation (3.93) and clearly show substantia.f
improvements relative to an undamped solution [59].
,
(3.92)
The stability of the acherne is controlled by parameter OS (J
< 1; proper
an+l
(3.97)
con = ag + c[v}
.where e is the contact damping coefficient. For normal
= (Rn + /1~) /m
(3.93)
Xn+l = Xn
c~ntact
+ Vn+1dt
then,
(3.94)
LJ
{3.98) .
with
an = emi;/ !l.t 2
{
(3.95)
e = (mii / l:i.t2
Discontlnuum Mechanlcs
()()
67'
~---
.... ~-- - - - -
~t=l x10' se
35
30
>.
20 ' - - - - - + - - - - - - ! - ...
25
......... '
-.---=-+----
--
. , C-0.5
........
,.
. <-.----~~------.;-.-
.......... - - --..,...., ...- - - - h..........~
15
..-t.~.-: - :,..: _.
, """'
10
100
.... -
300
200
400
i- -
C-0,8
C-0.95
500
600
700
n(number or colllslons)
---
-------
t=Sxl~'.
20
----
100
..- .
C-0
----+----l---+----1----,,,......:;.-+----l
..-
200
300
400
500
800
700
n(number of colllslons)
Chapter 4
N onlinear FE
Discretization
4.1
Introduction
4.2
69
70
Discontinuum Mechanics
L ___;
71
for example, a rotation of the given body in space. This kind of inva.riance
may, accordingly, be used to establish material isotropy [61 J.
time=t
4.2.2
time=t-+<ft
Deformation Gradient
Deformation gradient, the most basic definition in a. large deformation analysis, is defined as,
~,x
F=
ax.
ax
or
ax
F=-=l+Vu
(4.3)
ax
(4.4).
u=rn-X
4.2.1
The Motion
According to Figure 4.1, the relation between the refercnce and current
coordinates may then be expressed as
Consider the motion of a body during its deformation as depicted in Figure 4.1. The deformation of the continuum can be described by a mapping X
which gives the current coordinates, x, in terms of the material coordinates
in the undeformed configuration, X,
= X(X,t)
(4.1)
(4.5)
--
or,
= x(X,t)
(4.2)
--- -- --------------------
(4.6)
b= ppT
E=
He -1)
= ~ (J
(4.7}
- b-
In the literature, the Almansi strain tensor iH regar<led a.s. the push Jorward of the Green strain tens<;>r. Conyersely, E might be interpretad as
the pull back of e. These two strain tensors measure the total deformation.
According to equations (4.6) and (4.7), however, the relationship between
t.heso strains and displacemcnts is nonlinear, whirh ndcls t.o t.lm complexlty
of the numerical algorithm [63].
Discontlnuum Mechanics
72
4. 2.3
Polar Decomposition
73
where
\,'.
Ai = >..1,
i,l=l,2,3
i,l=l,2,3
F=RU
or
F=VR
(4.16)
(4.17)
(4.8)
be
which implies that the tensor R rotates the material vector triad N 1 into the
spatial triad n. Analogously, the deformation gradient and strain measure8
may be expressed in a similar way as,
3
U 2 =e
and
y2 =b
F =
(4.9)
L >-1n1N1
1=1
L 2(.X~ - l)N1N1
E=
l=l
,_j
1 . 2
E= -(U -I)
2
(4.10)
1(
-2)
e=-1-V
2
(4.11)
4.2.4
(4.12)
f=l
E<n) =
= L:>..~nini
1.1&cy"
.!.cu" - I)
n
(4.21)
(4.13)
(4.22)
i=l
and
3
U= L>..1N1N1
(4.14)
E<n> =
l=l
J=l
V= L..\inini
l=l
: .!:.c.x - I)N1N
(4.23)
(4.15)
e<n)
=E .!.(1 - xn}n10n1
'=l n
(4.24)
..
Discont.(nuum Mechanics
74
which implies
(4.25)
In the particular
sors,.
casen~
3
E(o)
= Lln..\1N1@N1
= lnU
(4.26)
75
current rate of deformation, but gives no information about the total deformation of the continuum. Furthermore, its integral in t.ime for a material
point does not yield a well defined, path dependent tensor so that infor-.
mation regarding phenomena such as total stretching (or uut..'<imum strain)
are not available. Therefore, to obtain a measure of total deformation, the
velocity strain has to be. transformed to some other strain rate that can be
integrated to yield a measure of the total deformation [63].
The spin tensor is then defined as the skew-symmetric part of the velocity
gradient matrix
1=1
l=w+d
(4.33) .
(4.27)
where
i=l
4.2.5
= 8v(x, t) = Vv
8v
F= =lF
8X
= ilRr + ~R (u- 1 - u- 1 u) nr
(4.35)
The second term vanishes in rigid body motion for which d =O.
(4.28)
8x
(4.34)
and
w = -(l - l )
Rate of Deformation
(4.29)
4.2.6
(4.30)
The Cauchy stress, which is also called the true stress, is defined as a spatial
tensor, u, which provides the nQcessary and sufficient condition for satisfaction of the momentum ha.lance la.w. This syrn111etrk tmHor relates the
normal vector n to the tracti~n vector t
The rate of deformation matrix (also called the velocity strain, or stretching) is defined according to the velocity gradient matrix, l,
Stress Measures
t(n) =un
1
T
d= -(l+l )
(4.31)
'"'
(4.32)
dln..\ 1
= L., --n1n1
1=1
dt
where d gives the rate of extension per unit current length of a line element
having a current direction defined by n. The velocity strain measures the
= piJ
,.,
,.
'i~~J~: ':
Discontinuum Mechanlcs
76
4.2. 7
An important aspect of the measure of stress i::; the o/Jjectivity (or jrame
invariance) of its rates. An objective rate must transfonn according to the
law for a second order tensor when the spatial refereuce frame is rotated.
This is important because constitutive equations are often expressed in re.te
form, so if the rates are not objective, the material behaviour will depend
on the choice of the spatial coordinate system, which is obviously physically
unrealistic and must be avoided.
'5Wint
u : ddv
(4.38)
77
T=J<r
= detF
Although the Cauchy stress tensor is an objective tensor, its rate does
not satisfy the objectivity condition. Therefore, for a general constitutive
material model 'Ye are bound to use other stress rntes. Only the m0$t
popular ones are briefly reviewed here.
The Truesdell Stress Rate is constructed directly from the terms.required
for the objectivity of Cauchy stress rate,
(4.39)
(4.44)
(4.40)
(4.45)
If the un-symmetric part,
RRT, is approximated by
'
U,
(4.41)
<Tv = t!r Another stress measure is the first Piola-Kirchhoffstress which is defined
at the reference configuration
p = Jup-1'
tTW -
wcr
(4.46)
the J aumann stress rate is obtained. This tensor has two important prop- . , ' .:
erties:
'
:<..
(4.42)
(4.47)
and is work conjugate to F. The second Piola-Kirchhoff stress tensor, another frequently used stress measure, is defined as
(4.48)
(4.43)
and is conjugate to the Aterial strain rate tensor, E, with respect to the
initial volume. Finally, the logarithmic strain measure (ln U), defined in
(4.26), is work conjugate to the rotated Kirchhoff stress, RTTR [65].
Discontinuum Mechanics
78
uv = Cd
(4.49)
The above form can be used for nonlinear materials in that C can be a
function of objective measures of stress or strain, such as u, u V d or other
'
Htat<~ va.ria.bles. Noto thn.t it can not be a function of ir or d because they
11w 1101. frnrno it1voriu.11t..
4.2.8
79
Multiplicative Decomposition
and
(4.54)
T=Cep:D
(4.51}
where T is the material rate of the rotated Cauchy stress tensor and D is
the rotated rate of deformation,
.
For strains that are too large to be treated as infinitesimal the decomposition still works, at least for longitudinal strain, when this is,taken as the
logn.rithmic Htrnin. In cmltH' to formula.te a tlwury of pl11HI i<'it.y for large deformation, we hav..e to establish the approprlate dccomposit.iuu fr arbitrary
states of deformation.
An ad.ditive decomposition of the logarithmic strain is equivalent to a
multiplicative decomposition of the deformation gradient F,
I'
; (4 56) : <>
F=FeFp
. . ,r;n:,'.~/
which was first explicitly formulated by Kroner [D], ami fnrther developed\
by Lee (70), Nemat-Nasser [71) and others on t.ht! hnsiH nf consideration of
the behaviour of crystals. The reasoning behind t.his, is comprehenslvely
presented, among others, by Lubliner (61].
.
The transformations involved in the decomposition are shown schemai.;
(:i:, x + dx) is usually sa.id to be
m an intermediate configuration.
where
T=RtFR
(4.52)
-1
-1
(4.57)
le= FeFe
D=RTdR
Then, the Cauchy stress tensor can be evaluated from the expression,
lp = FeFp
(4.58)
= Rn+l
( +l
O'n
tn+l
tn
Tdt R~+l
(4.53)
d =de+ (FelpF; 1 )
'(4.59)
.11,
Discontinuum Mechanics
80
= l+Vu
Fn+l
(F~+itial
= Fn+1 (F;rt =
IF (q,n+l) trial :S O
lnilia)(Rcfcrencc)
Conliguration
Elastics update
Set ( ) n+ 1 = ( () n+l) trial
Figure 4.2: Multiplicative decomposition of the deformation gradient.
n
i
LJ
ELSE
Plastic corrector phase
4.3
F;+i
(Fn+ir1 F~+i
= (Fn+1r1 R~ttiu~t+i
ENDIF
4.3.1
Equilibrium Equation
n.
CTn =
x=x
j8urf
on
on
ru
ru
(4.60)
(4.62)
Applying the divergence theorem to the first term iu the above equa.tlon ;
and using equation (4.60), the following is obtained
where u represents the Cauchy stress tensor and n representa the unit ohter
normal u.long
{ rur f da
(1'
Far this body to be in a state of static equilibrium, the following condition must be satisfied
. (4.61)
lr
= f O'nda = { d'ivCTdv
lr,,
Jn
fo
(divo- + body) dv
pdv
(4.63)
82
Discontinuum Mechanics
83
As this condition holds for any arbitrarily chosen subdomains, the integrands should be equal pointwise,
.
(l'tVO'
lJ
+ bml', : : : pu..
(4.65)
4.3.2
Variational Form
Here, a weak form of the equilibrium equation is derived, since this is utilized
as the basis of the Finite Element procedure. Thereby, the above constraints
are not applie<l pointwise, but rather in an average sense, by a weighted
residual scheme,
r (<huO" +
Jn
fbody -
p) wdv + lr~
r (rurf - un) wda =o'
(4.66)
un= O
un= con
where the weighting functions w satisfy the natural boundary condition over
r u, i.e. w = O. Applying Green's theorem and integrating by parts, results
in the following identity
V:= {5u:
+ fr unwda +Ir
(4.69)
=o
on
r uJ
(4.67)
the space of admissible variatiop.s, the variational (weak) form of the dy-.
namic initial/boundary valu~ problem may be Pxprn:-;sed as (73, 74)
which is simplified to
{a: Vwdv
Jn
Jn
pwdv =
lo
ldYwdv +
lr~
aurf.wda
(4.70)
(4.68)
where
Wint(u, tt) =
f O E(u) : u(u)dv
M(u, u)=
'Updv
0
Discontinuum Mechanics
84
85
L_.:
wext(u) =
uJbodydv
ururfda
4.4
(4.74)
r~
(4.75)
denote, respectively, the virtual work of interna! forces, the inertia forces
contribution, the virtual work of ex~ernal forces and the virtual work of
contact forces. Here u is the Cauchy stress tensor, e is the strain tensor,
u is t}:e displacement vector, while g representa the contact gap vector.
Observe that in the present formulation the contact terms correspond to a
penalty formulation of contact interaction.
u
1
DW(x, 11)u
ddv
(4.76)
(4.82}
'v'71 E V
(4.83)
j no r:
ddV
where wcon is the virtual work of the contact forces and w(~xt is the virtual
work of external loading
(4.77)
(4.84)
where ,,- = Ja is the Kirchhoff stress tensor and !lo is the initial (reference)
volume of the body, and
wint
wint
P :.FdV
(4.81)
wint
(4.80)
in the direction u,resulting in the linear part of the functional W(x, 17)
W(x,11f=0
(J':
W(:X, 71)
E=O
wint
= dd 1
(4.78)
Oo
Ne
no
S : EdV
(4.79)
x(O)
= U
e=l
0 if a i= o
where Ne is the number of finite elements. So tlmt over the typica.l element'
x(le) one has the interpolation
86
Discontinuum Mechanics
N .. ,.
uhlxcne)
= 2: N a(x)tta
(4.86)
F=
Fnodal
a=l
wlwre
N~n
87
Ne
Nw
e=l
n=l
A(/e)+ A(/~n)
Nen
vsymuhlX(2e)
=L
(4.93)
Batta
(4.87)
a=l
Ba = vsvmNa
{4.88)
The effects of the inertia term is directly considered in the ~X.P.!~cit time
integration scheme to evalute the
displacement, velocity and
celerati0"e1Cls:
. .......... ......
"-~~---.....
ciirreiit
"''~''
e.e:-:
':
, .. """ r, """"
:.~ . '
:! T.:!~:. ;,
: ''',_;./
4.5
{4.89)
wbcre
(4.95).
where t is the time, int(u, t) the interna.1 force vector, rxt(t) the externa!.
force vector, rn(t) the contact force vector while A1 denot~s the mass
matrix.
4.5.1
are H~ c~n.l.led elenwnt; intemal and extemal force vectors, respectively. The
contnbut1on of the contact forces is consi<lered directly through the associated contact nodes,
n = Acon
n . ( ng )
! con
(4.92)
(4.96).
Vn-1/2
'Un - Un-1
A
utn-1
(4.97)
Vn
= 2(Vn-1/2 + Vn+i/2)
,:__ Vn+l/2 - Vn-1/2
Vn -
.tn
(4.98)
( 4.99)
There are severa! ways of constructing the lumped mn.ss ma.trix, which
are briefly reviewed here: [75]
l. The same proportion of the masa is associat.c1l with rnHh node.
(4.100)
'1
~rth.er,
{4.101)
!J
LJ
89
Discontinuum Mechanics
88
:'here mi is the i-th.diagonal ter;n of M. Time incrementation is then readIly pe:formed by evaluating displacement, velocity and acceleration using
cquat1ons (4.96), (4.98) and (4.99), respectively.
. If a damping matrix pl'oportional to the mass matrix. is assumed the
solving of eql,l~tions is .still avoided and the equa~!ii ~(4\9if
is generahzed to
.___ . ., ...".
~1multane?us
2. Terms in each row of the conslstent m!Uis rnatrix un~ ::;ummed and
lumped on the diagonal. This technique has sometimes a tendency
to generate zero and negative masses, that is imacceptable in practice
and from a pl,!.ysical point of view.
3. Setting the terms of the lumped mass matrix proportional t the diagonal terms of the consistent mass matrix with the constant of proportionality selected to conserve the total element mass. [78}
matrix. [79}
The first and third schemes are the most frequently employed in practice
2~tnm'.':" 1
~,
and
ci
4.5.2
4.5.3
The stable time step for the central difference method is given by
At
~ min :n
( y'1
+e~ - en)
~tstab
(4.103)
where Wn are .the natural frequencies of the mesh ami ~n is the fractlon
of critica! damping in the n-th mode. Since strm:tural systems are usually
underda.mped, the mnimum of (4.103) is usually given by the maximU!i
of theao can
{4.102)
where /3i
111 1it.hN
frequency,
Wma.~
(4.104)
Dlscontlnuum. Mechanics
90
where
(4.105)
u:;,in
which is suitable for discrete element contact problems. The:m energy terms
may be expressed as,
.w.ne:i:t --
Therefore, the stable time step can be determined within the time integration loop, while the maximum frequency, w~lm is computed within the
element loop.
11
LJ
w.aur J
n
+ wbody
n
It is in fact usually notl~actical to compute even the ma.ximum eigenvalues of the element directly, which would undoubtedly increase the cost of
the computation considerably. !nstead, formulas which give the maximum
olmnont
or upper bounds on the maximum element eigenvalues ,
w eigenvalues
.
Amax, are used.
(4.110)
Jr Jreo
(4.111)
D.t < ~
- e
u:tr
(4.112)
u~r =
1. [t ud] dv
(4.113)
(4.106)
(4.114)
[f,
Energy Ealance
(4.115)
N umerical instabilities are normally associated with a large growth of energy. Therefore, monitoring the stability and accuracy of the solution can
be performed by continuously checking the energy balance of the system.
The energy balance equation at time tn can be expressed as [1],
wext
- ukin
n
n
'
ustr
n
wdam
<
l';r'
1
n
Yl'n
(4.107)
(4.116)
which has to be determined based on midpoints velocities
Discontinuum Mechanics
(4.117)
Chapter 5
a.ncl
N1p
Unstr
1 "' { i
}T
. .
= ustr
n-1 + 2 L_. O'n-1 +O'~
~e~ V'
(4.118)
i=l
{4.119)
.wl~en~ N nnd iV_p u.re t.he numbor of no<les an<l the nmuber of integration
~omts
the g1ve~ ~ody, respectively, C = cM is the linear viscous dampmg matnx, and V Is the volume associated to the integration point i.
fo:
n
LJ
5.1
Introduction
93
1
Discontinuum Mechanics
slave element s
slave segment s
slavenode s
slave line
95 .
5.2.1
Original Algorithms
(5.1)
where a~ und a:1+ 1 a.re the normal stress componentH in slave elements
and s + 1, respectively, ami (is o. dimensionlef-1. diHtnrnn foct.or,
l'
(5.2)
/"-~
1;,-
l,,,+r
slave element s+ 1
slave element s-
In this chapter, following the historical developments of the normal contact algorithrns, a brief review of the hydrocode based methods, as the first
developed contact method, is provided. Then, a comprehensive discussion
on the penalty based normal contact ~lgorithm is given which inclt.Idf.'s inaterials about the original DYNA approach and also the present state of the
art on the subject.
1
5.2
Hydrocode Formulation
master element m
96
97 ,
Discontinuum Mechanics
l~,s+t
.
+
phony element p
K.m
'"'nm
Mm
+ "ns
M i8
L--i=l
i
L--i=l
'"'~m M~n
L.Ji=l
slave element s
(5.3)
where nm and ns are the total number of master and slave elements along
the slideline, Mri and Mi are the total mass of the ith master and slave
elements, respectively.
1
1
master element m
1
r1
1
-----
(-----
~--- ---------
1
1
1
1
dvJ--- ---- --"". ___ ....1'1
\.---d~--i
1
---l
---1
"'m
To update the motion of the slave surface, the slave nodes are projected
onto the up<lated master surface. Velocities for the slave nodes are then
found by dividing the change in coordinates by the time step size.
To incorpora.te effects of oth-er slave elements for computation of interface pressure, TOODY finite difference code, developed in 1968, assumed
a phony element approach. A phony element is defined during ea.ch iteration, overlaying the slave ele.ments and matching the mesh density of the
corresponding master surfacl Figure 5.3 shows how such an elenent can
be defined for a typical slideline. The position of node q in Figure 5.3,
is determined in the same proportional distances of node p of the phony
element:
if . rm,s . tm > o
if rm,s tm $ O
Tm,s tm
1
""
,-
(5.6}
if rm ..~-t-1 tm < lm
if rm,s+l . tm 2: lm
.
_ { Tm,s+l tm
dm,s+l lm
(5.7)
(5.4)
LJ
= (dm,s+l -
CTB
dm,s) lm
(5.5)
noe-1
qP=
(5.8)
i=O
where noe is the number of overlapping elements. The procedure will then
be repeated for the next phony element.
98
Discontlnuum Mechanics
Chapter 5.
Avm-1,m+l
____
n _ __
m+l
m -
(time=t)
(5.12).
lm-1,m+l
n =
(time=t+M)
(!::' - !PCm)
(5.13)
Mm
(5.14)
m
(5.15)
Figure 5.4: Updating the sla.ve surfoco motion.
In updating the slave surface motion, TOODY projects the slave nodes
normally onto the updated master surface, and then moves them tangentially according to their tangential velocities and accelerations (see Figure
5.4),
This scheme, repeats the foregoing process of creating phony nodes and
elements to accelerate the slave surface. After updating the geometry, slave
nodes are projected normally to the surface, an<l the normal cornponents of
velocity are then computed by inter..Polating from the master side.
(5.9)
5.2.2
(5.10)
DYNA 2D Formulation
LJ
(5.16)
(5.11)
and
Wm
.. ;, < ..
connect~'.,;.:.
Discontinuum Mechanics
100
101
'1"
Mass Distribution
The mass of the slave elements a.long the master surfacc is attributed to
the appropriate master nodes. One half the mass of each slave element is
distributed to the corresponding master nodes if both slave nodes of the
slave segment are in contact with the eiement. If only oue slave node is
in contact, one half the nmsH lying between the ('()11tact.i11g 11ode and the
centre of the slave segment of the slave element is distril.Htt.cd. No masa is
distributed, if neither of the slave nodes are in contact.
The procedure of adding the contribution of the mass of slave node, M 8 ,
to the mass of master nodes can be summarized according to the terminology
of Figure 5.6,
m
Figure 5.6: Distribution of the slave element ma.ss to the master nodes.
(5.19}
_ d's,m+l -d'1,m ,_ 1
l,
i~ s
(5.20)
m+l -
s
M m+2
(5.21}
M:n
(5.17)
i
where
is the contribution of the mass of slave node s into the mass of
master node m. ls = ra,s+1 ta is the length of slave element s, t 8' is the
slave taugential vector, and
t = { r;,m
s,m
O
(5.18)
d'
If the slave node s lies very close to the master node m or m + 1, the
angle Bm-l,m+l or Bm,m+2 will be used to determine thenormal component
of the accelerations in (5.17).
t,
- { r;,m+l . t,
s,m+l -
la
if
if
if
i/
r~,m.
ts
>o
r~,m t 8 'SO
r~,m+l . ts < ls
r~,m+l . ts ~ ls
(5.22)
(5.23}
and vector r~,m. connects the node s to the middle of master segment m.
Discontinuum Mechanics
102
Moment um Balance
Ps+l -
103
(5.28)
ncsn
(5.25)
Pm =
t=O
Pz+i
(5.29).
~:. ' - . ' ' 1 ;
i=l
where ncs is the tqtal number of contributing slave segments. The pressure.
is then distributed to the master nodes m and m + 1.
Using_ the TENSOR algorithm, the normal interface force at master node
m is defined by:
Contact Pressure
From the state of stress in the slave elements that are in contact with
the master surface, a contact pressure and its equivalent nodal forces are
computed for each master segment, based on the approach given by TOODY
code.
. (5.30)
where ac is the coriolis acceleration, F![' == Mma:~b and f 11in is the normal:
internal force in master element m. The normai ncceleration of the master
node mis then computed:
(5.31)
m
Ps
(5.26)
where <J! is. the effective stress in slave element s perpendicular to master
1
segrnent m,
'
s
O'ef f
s s+l
f n'
(5.27)
l,,
where f~ and ~+ 1 are the internal normal forces in slave elements s and
s + 1 respectively.
pr:
Discontinuum Mechanics
104
105
(5.32)
(5.33)
Then, the new velocities are the combination of the tangential components from the slave side and normal components from the master side,
while the new accelerations can be computed from the dynamic equations
of relative motion.
5.3
The first practica! penalty based finite element conta.ct analysis was developed by the DYNA team in the eighties. It included both explicit and
implicit versions. In implicit code; a contact interface is treated like any
other element class, whereas in. the explicit code, the implementation of the
penalty method is straightforward. The main advantage of the method to
the earlier developed hydrocode algorithms was in maintaining the symmetry of the approach, eliminating any biaE. in choosing the slave and master
surfaces [54].
Figure 5.7: Node mis the closest master node to the slave node s.
For each slave node s, locate the closest master node m, and corre-
(5.34)
5.3.1
Figure 5.7 shows a typical contact problem in which four master segmenta
can harbour the slave node s, and node mis the closest master node.
If m and s do not coincide, s can usually be shown to lie in a master
segment i via the following tests:
(5.35)
where rm,i is the position vector along the common edge of the master
segments i and i - 1 and outward from node m, t 8 is the projection of the
g vector, connecting points m and s,
ta =g-(gn)n
n = rm,i
lrm,i
rm,i+l
X rm,1+il
(5.36)
(5.37)
since the sliding constraints keep s close but not necessarily on the master
surface, and since s may lie near or even on the intersection of two master
Discontinuum Mechanics
106
L_;
n
LJ
107
with,
/i(~, r) =
L N1x~i)
(5.40)
j=l
grm,i
lrm,il '
5.3.2
i = 1, 2, ...
(5.38)
(5.41)
where Ni are the standard shape functions, and xJl) is the nodal coordinate
of the jth node in the ith di~ection. The normal to the surface is non-zero,
(5.39)
(5.42}
The contact point coordinates (ec, TJc) must satisfy the following conditions:
(5.43)
(5.44)
slave node
In DYNA 3D, equations (5.43) and (5.44) are solved numerically to
compute the contact point coordinates (54).
5.3.3
Contact Force
If the slave node penetrates through its master segment, an interface torc~
is applied between the slave nade and its contact point, with a magnitude
proportional to the amount of penetration g,
g
= n [r, -
r(ec, TJc)]
(5.45)
u
!~ = -a:gn
Figure 5.8: Finding the closest master point e to the slave node s.
i/
<o
(5.46}
where J~ is the contact force acting on the slave node ::;. An equal and
opposite force is distributed over the master segment nodes,
1.08
Discontinuum Mechanics
if
<o
5.4.1
(5.47)
eKA 2
V
109
Contact Geometry
Figure 5.10 shows a typical slideline between two contacting bodies, defining
1.he contacting nodes nnd H(!gments.
(5.48)
a=--
where f. is a scalc factor for the interface stiffness, usually taken as 0.10.
Larger values reported to cause instabilities unless the time step size is
reduced.
11
LJ
lJ
5.4
In this section, a simple and logical approach for two dimensional nod~ .,to edge normal contact based on the penalty method is discussed. The approach was originally <leveloped in 1980 by Circa {83), and later documented
by Crook [52).
In this approach, the canta.et surfaces are explicitly defined. Two surfaces are then linked to forma slideline ora contact pair (or,contact couple).
A simple global search algorithm will be followed by an accurate local search
method to identify u.ll coi:itacting couples. A penalty based formulation is
adopted for contact inter.lon and evaluation of contact forces.and forming _
the non-consistent and consistent tangent matrices.
Both contact surfaces may be searched to give a symmetric treatment of
the contact surface, as opposed to earlier master-slave algorithms. To verify
the problem, in Figure 5.9, if point 1 moves up into edge e, then point 2
will penetrate edge e2. Therefore, we need to search for both surfaces at
each slideline.
2
o
segmentm
n (normal)
Discontinuum Mechanics
110
5.4.2
111
segment m-1
s
t., \
The first step is to find the closest node m on surface 2 (master) to node
s on surface 1 (slave) (see Figure 5.12). This is achieved by evaluating the
distance to each node, and finding the minimum value:
m-1
(2)
m - 1,nnode
(5.49)1
(5.50)
then there is a
pot~ntial
(5.51)
there will be a potential contact with segment m - l. Figure 5.13 depicts
the above mentioned cases based on the sign of the inner product of the
position and tangential vectors.
If only one potential contact is encountered for point. s we use the stan
dard case of contact with single segment, otherwiHe t.lie :;pedal case of contact near a corner has to be used.
m-1
segment m-1
5.4.4
The first step is to check for penetration. The normal vector is de:fined
by:
5.4.3
LJ
After identifying the closest point to a contact node s within the global
search algorithm, the associated contact segment on surface 2 is located
(Figure 5.12). This is done by verifying the sign of inner product of the
position vector r ms nd the segments tangential vectors tm and tm-1
If the following condition is satisfied:
n =ea x t
"(5.52)
(5.53)
112
Discontinuum Mechanics
lit 1d1~
113
.1
11.1u l
th(l two
(5.55)
t
1
where
(5.56)
r----1.
-1
The magnitude of the contact force in the local system is defined by:
e= Un
The second step is to locate the contact point. By defining a local
natural coor<linate Rystem E [O, 1], anda local segment numbering system
according to Figure 5.14, the !ocal contact point is defined as
{5.57)
(5.58)
(5.54)
The last step is to compute the contact force and its associated stiffness. The model is constructed by inserting the penalty spring between the
contactor node of body l and the target segment of body 2, as depicted in
(5.59)
Figure 5; 15.
e.
where Ni = 1 and N2 =
are the target segment shape function8 .
Figure 5.16 shows the geometric representation of the contact forces. It is
worth noting that in any case, the angular morne11t.11m of the body should .
not be changed.
.
The contact stiffness is then defined by:
Kc
' T
= aN,N,
(5.60)'
I
Figure
Discontinuum Mechanics
114
115
5.4.5
!/
lt luu; n.lrmtdy btmll pointod out thttt the 8lmplo w111dt:v 11wt.l1<1i I wlll le"d to
a stiffness matrix which is not in a consistent form. By usi11g thc prlnclple
of minimization of the total potential energy of the system, including the
contact constraint, a consistent form for the stiffness matrix can be derived.
As described earlier in Chapter 3, the equilibrium condition can be expressed as:
an T o:a;, =0
-+g
8u
8u
Body2
;f
Rody I
.,
....
(5.61)
JI
o
l)g
(5.62)
~-"00-
8u
Ke
pose numerical arnbiguities as well as numerical instability in particular
conditions.
For instance, there are. always oscillations between two comer nodes in
iterations. Actually, we may have. oscillations even without ;having any real
penetration. Figure 5.1 7 shows a typical situation where during the analysis,
poiat. 3 may soem to be. penetrated through its opposite segment, causing
the contact forces to be generated to return the node to an unpenetrated
position. At the same time another force with opposite direction is acting
011 node 1 to pm;h it out of the regiC>n occupied by body 3 near the comer
point 2. This causes oscillation of the systems around the corner points.
g)T + au
ag + gTo: fJua ( 8u,
og )]
= o: [( fJu
(5.63)
(5.64)
There exist remedies to treat this set of problems which may frequently
happen in progressive fracturing phenomena, for which several small parts
of a fractured and fragmented body may interact with each other simultaneously. One simple remedy is to slightly enlarge the end segments, so a
11ode to (~dge contad can be performed without any ambiguities. However,
x1)
(5.65}
Discon~inuum
116
Mechanics
117
1..............................................
and finally:
s
~
(5.74)
(5.75)
= e3 x t
(5.66)
Noting that
ec =
(5.67)
(5.76).
or,
(5.68)
(5.77)
and
(5.69)
To evaluate l, we have
(5.78)
(5.70)
uK c~u
so,
(5. 72)
uKc.u =
o! )
= a6u [( 8
1
'
a ] ~u
0~
(5.79)
(5.~l>:::
-\~::/('
Discontinuum Mechanics
118
aN
NT
. s s
(5.81)
(5.82) (
where
(5.83)
--
(5.84)
'. (5.85)
Note that for z~ro gap, the nonlinear term is zero. Also, as g becomes
smaller, the significance of the last term in (5.82) becomes smaller, however,
for the sake of convergence we need this term.
(5.86)
where Pn is the intensity of the normal surface pressure at the other end of
the contact surface, b is half the width of the contact surface and
5.5
Figure 5.19 illustrates a simple contact problem defined in terms of edge/ edge
contact couples. Although it is notas popular as the node to edge contact,
it has heen used by many research groups working on general crack propagation problems by discrete element technique.
Utilizing the Boussinesq solution for the concentrated force acting on an
elastic half space in combination with the Winkler elastic foundation the'ory
k1
k2
= ia~t
(;~
120
Discontinuum Mechanics
Chapter 6
Frictional Contact
Figure 5.20: Description of an edge to edge contact.
6.1
where G represents the shear modulus and v is the Poisson's ratio. The
normal contact stiffness can also be derived
(5.89)
and finally the normal contact force can be evaluated
F~"
Pn A =
= b (PnA + PnB)
(5.90)
In this chapter, the main features of computational frictional contact algorithms are discussed. The chapter is divided into five parts. The first
section, explains the basics of frictional contact problems including the kinematics and principies af the plasticity theory of friction. The second sootion, provides a detailed formulation for two dimensional frictional modela
without frictional hardening. The thh:d section generalizes the approach
to three dimensional models and the fourth section will introduce frictional
hardening within the 3D modela. Finally, a 2D node to edge model and a
3D node to quadrilateral face model for Coulomb frictional contact wi1Gho~ut
hardening e.re presented.
6.2
xconU Q
UnA
n
n UnA+UnB
Introduction
(5.91)
In this section, sorne of the fundamental aspects of a frictional contact. behaviour are discussed and sorne f the pri11cip~.l issues regarding the math-.
ematical formulation of the computational methods are addresse<l.
6.2.1
Because of microscopic nature of frictional coutuct, includiug the mlcroscopic geometrical conditions shown in Figure 6.1, friction can not b~ easily
modelled adequately in the macroscopic numerical analysis of many. applications such as metal forming processes.
c__.I
121
122
Discontinuum Mechanics
Tool
',
LJ
lnltlal
... ..
con,pgurutlon
>-.::~~r~~~~::::
.
' ..
~
123
'
'
Master
<
Current
Configuratlon
The frictional sliding is simulated by assuming mathematical relationships between contatt stresses and other related state variables, which describe the conta.ct conditions. These relationships can be detived from the
micromechanical models of friction and then incorporated into an appropriate constitutive theory [84). At first, only a quasi static type of friction is
considered, which occurs, for example, when two metallic surfaces are slowly
pressed together and relatively displaced in a quasi static equilibrium [85].
The most conspicuous mechanisms contributing to this type of friction
are:
plastic deformation ~f asperities.
In this figure:
Thermal coupling aspects of friction, which recently received considerable attention in the computational.literature, will not be considered. Their
inclusion, however, is possible without any major modification in the context of the general thermo-mechanical coupling strategy.
6.2.2
e
Yn
Yt
Basic Kinematics
are
l;'(I
ros=Xs+us
(6.l)J: .
l . . -'
-~;_ji~l~f-~"
124
125
Discontinuum Mechanics
(6.7)
where to is the time associated with the first contact of slave point So and
master node Mo.
Defining the tangential gap in the curvilinear coorclinate system (Equar
tion 6.6) constitutes one of the main features of described kinematics. It is
a nonlinear basis which is not in accordance with the main cartesian basis
in which the equilibriurri equations are to be stated.
The relative displacement of the slave node with respect to the master,
gr, can be obtained from the known master movement
1
LJ
gr=
where res and X s denote the current and initial positions of slave node
S, respectively, and us represents the deformation from initial to current
configuration.
1
t
to
'
dgt(t) =
1t
dg1(t)t(t)
(6.9)
to
In general [86],
(6.10)
(6.2)
Also,
The normal and tangential gaps are then defined as:
,r
-~
''
,.:.;
(6.11)+\
Yn = 9n n
9n
LJ
= (xs-XM) n
9t = 9t t
t<i
(6.3)
(6.4)
(6.5)
(6.6)
'-...
( ~Jmptdt 6. Mlct.Jmml
DJ..,contlnuum Muduudcs
121
C011t.11d
Stick-Slip Decomposition
(6.14)
n
LJ
6.2.3
.:,
(6.16)
'(6.17)
where e and P superscripts denote stick (elastic part of gaps) and slip (plasti~ : .
part of gaps), respectively. The stick gap is assumed to be reversible, while
the slip is irreversible, or permanent.
The objective of a computational theory of friction is to provide a computational description of the relationship between contact stress and stick/slip
phenomena at the interface of contacting objects.
In a pioneering work, Fredriksson (87) noticed that the theoretical deHcri ption of frictional contact behaviour can be derived from the classical
theory of plasticity [88, 86, 89). Table 6.1 indicates the analogy between
various frictional contact and plasticity features.
Un =g~
LJ
Stick-slip decomposition
Stick res.use
Slip criterion
Wear/tear rules
Slip rule
+----+
+----+
+----+
+----+
Elastoplastic decomposition
Elastic response
Yield criterion
Hardening/ softening rules
Flow rule
Stick Response
.,
Sticking is caused by the elastic (reversible) deformation of the asperities
on the contact surfaces. The sticking iaw C$ln be suitably described by
decoupling the normal and the tangential directions.
The surface traction (force per unit area) can be represented by tfft,
and r as the work conjugates of the normal and tangential gaps, 9n ~el
128
Discontinuum Mechanics
(6.24)
(()~20)
{6.25)
(6.21)
129
= r(gn
Equation (6.20) which is relevant for the normal direction, can also be
regarded as the frictionless contact law.
(6.26)
= a. exp( blgn 1)
(6.22)
O'n
Cn
=Ct
Uf
(6.27}
(6.23)
It is worth noting that these nonlinear contact laws are primarily important for the accurate description of contact behaviour on the microscopic
level. However, the normal micr~displacements are practically negligible
compared with the overall macro displacements. Therefore, only a linear
sticking law in the normal contact direction will be considered in the present
model:
130
Discontinuum Mechanics
Slip Criterion
LJ
131
TSL
Tm
= sUs
[0, 1)
{6.34)
<P.(uc w) = { < O :
'
=O
st~ck
slip
(6.29)
'
{6.35):
where uc = { <Jn, T} contains all contact tractions and w indicates an interna! variable for describing the tear and wear phenomena (which will not be
included in this section). For 3D applications, tangential traction T can be
expressed in terms of two independent directions t 1 and h in the tangential
plane:
(6.30)
u a is the current shear strength of the deforming material with <1'1J yield
stress. This model is m01;e suited to cases of high contact pressures such a8
bulk forming operations.
3) Parabolic Law
In this model, the frictional resistance is constantly rising, but with a decreasing rate (92):
(6.36) .
(6.31)
(6.32)
Norm of the tangential traction, llrll, in a contact analysis is similar
to the equhalent stress function, f(u), in an elastoplastic analysis. In the
same way; the maximum allowable tangential traction (slipping stress), TsL,
is analogous to the yielding stress, u y.
= 1 and a + f3 =
4) Avltzar Model
A combination of Coulomb and constant shear laws has been derived by ari
upper bound analysis of the interaction between the rigid and deformable
asperities of two contacting surfaces [94):
t
for
Here, the most commfl:>. choices of friction law are reviewed. Any choice
for the friction law of a specific application has to be experimentally verified.
(6.37)
for
'>
5) Sinusoidal Modification
1) Coulomb Law
TSL
= lunl
E (0, 1)
(6.33)
where is the coefficient of friction. Studies have shown that this model
is adequately applicable to the frictional conditions with moderate normal
contact tractions, such as sheet forming problems.
1011:-;t.11111. slarn1.r
regime wlth
.
(6.38)
Discontinuum Mechanics
132
The way two contacting bdies slide against one another can be classified
to two distinguished classes of problems. In the first cla.ss 1 the bodies are, .
rubbed against each other in an arbitrary direction, creating a.n lsotroplc ,
wear phenomenon. While in the second class, the sliding is orientt3d along
a prefer~ntial direction. Thus, an anisotropic tearing phenomenon occurs~
Wear and tear phenomena perform as isotropic and anisotropic softening
processes, respectively and reduce the overall frid.ional strength of the con
tact surfaces.
6) Parabolic Approximation
Auother model with a smooth transition to the constant shear regime is :
f or . lanl < 2~
for
lanl ~ 2~
(6.39)
Any defined wear and tear model may be included through the appropriate interna! variables w in the general slip criterion
7) Trigonometrical approximation
nLJ
133
<l>(<Tc w) {
'
< O : sti~k
=O : slip
It should be noted that based on sorne theoretical considerations a.nd experimental observations, an internal scalar variable w = w( aw), describing
wear softening, is more work dependent
(6.40)
....
..-(1)
...
(6.43)
Slip Rule
1
1
p_.
aw
g - 'Y Bue
lJ
;.
(6.44)
where 'Y is the proportionality constant. The component form for a general
3D slip rule becomes:
.:J'
Ytl
ltt_
='Y'Hii
(6.45)
Discontinuum Mechanics
134
135
Associated slip rules has }ee.Jl used to a.void the non-symmetric tangent modulus despite its unjustified phy:rn.l chnrnct.l'l'Htics [87).
1t1
Here, a non-associated slip rule with isotropic frictional conditions is employed and consistently linearized. The slip potentials can then be defined
as:
(-a )
'1f(T)
= llrll
for 3D
\Jl(r)
= lrl
for 2D
(6.46)
exm~~s
P-'A
9t
- "'( llTll
gf ='Y sign(r)
where subscripts
and
' = 12
!
for 3D
for 2D
, (a.41Y:.
The third part of Equation (6.45), states that an irreversible normal gap
increment may exist for a frictional slip rule. In other words, even in a pure
frictional slip, an inelastic normal gap increment may be generated when
the slip criterion (6.45) is satisfied. This unacceptable physical behaviour
is in contradiction with the previously discussed basic assumption that no
slip occurs in the normal direction.
Therefore, the slip potential must not be a function of O'n As a result,
an analogy with the classical associated flow rule of metal plasticity is not
applicable, and a u n-independent non-associated slip rule must be adopted.
An associated slip rule would lead to uplifting in the normal direction under
frictional conditions where un is always compressive. Figure 6.5 compares
the associated and non-associated slip rules.
Thc non-associated slip rule will lead to generation of non-symmetric
tangent stiffness modulus. Two methods have been proposed to avoid creation of non-symmetric modulus:
The first appro~ is to utilize a two step algorithm, where in the
firnt step, the stress normal to the interface, O'n, is obtained from
equilibrimn conditions. In the second step, the slip rule is specified
only in the tangential plai:te (88). The linearization in this approach is
not consistent, and hence it <loes not provide quadratically convergent
solutions.
Equations (6.47) satisfy the condition that the direction of frictional slip '
is determined by tangential frictional contact traction and is independent
of the normal contact traction.
6.3
(6.64)
Discontinuum Mechanics
136
137
a)
a;M'l _ a;M1
-lla;_M_<J___a;_,,M-=-1""'-="11
(6.48)
(6.49)
9n =
(6.50)
Ut
(xs - a;M). n
= (xs - a;Mi). t
(6:51)
t=
= e3 x
Incremental displacement
uc = (T,un)T
,,..
= (gt, Un)T
= ( Uz, Uy) T
Contact force
6.3.1
b)
a;S -
(6.52)
= n x e3
(6.53)
[y'x --xcy]
(6.54)
t
M,
Mt
a=tan- 1
Un=
feo
= llres -zll
---
llx 8 - zll 9t
(6.55)
= R(a ~ a1)
(6.56)
(6.57)
R=Rb+Rc=
nel
nen
e=l
c=l
Are+ Are
{6.65)
e)
z 0 -z 8
n=
llz -msll
(6.58)
t=n
(6.59)
x eg
[y' - y]
M1
a=tan- 1 -
x' -xc
9n = R-
llzs -zll
9t = R(a1 - a)
6.gt == R(ao - a)
~j
(6.60)
where nel is the number of elements, nen is the number of boundary nades
which are in contact and R is th~ total global residual force vector.
The residual contribution for a single boundary nade in contact, S, be-.
comes
(6.61)
(6.62)
(6.63)
Figure 6 .6: Contact kinematics for a slave node S ~n contact with different
master segments; a) straight segment, b) convex cll'cular are segment, a.nd
e) concave circular are segment (84].
(6.66) ''
Discontlnuum Mechanlcs
138
[ 8u8~
ag
T2n
Equation (Je = Auc) represents a dependence of contact force to the associated contact area. According to Equation {6.64), the contact area itself
depends on the position of the slave node S and its neighboring boundary
points N 1 and N 2 As a result, variatins of the residual force vector (6.66)
has to be considered with respect to both u and uN.
,_.
Therefore, linearization of (r~ = TT e) yields not only to a (K") stiff11css matrix, but ah;o a (Kn~, i = 1, 2) stiffness contributiqn, which must all
be included at appropriate positions in an assembled global stiffness matrix .
[84]:
(6.68)
~ K
8r~
(6.69)
Kn' = 8r~
(6.70)
8u 8
{)uN
'
"'
"'
~
"'
11
~.
"'
a2g
!/
8u;a~~
''
(6.75).
and the stiffness contribution from two neighboring slave boundary nodes:
= 1,2
'(6.76)
Therefore, we would n~ed to find the first and second gap derivatives, as
well as area derivatives, and to define an appropriate constitutive law for a
slipping regime.
For .two deformable bodies in contact, stiffness terms resulting from
second arder gap variations are essential to achieve quadratic convergence
[96, 97], whereas for a rigid master/deformable slave contact, numerical tests
have indicated that the second order gap variations can practically be neglected. Therefore, the. second bracket in the contact node stiffness matrix
(6.73), which represents the variations of T, need not he computed, saving
computational time, while practically retaining tlie quadratic convergence
convergence rate [84).
It should be noted that Ni is not considered a contact no de ( at least while
considering the slave contact node S) and its associated contact stiffness
term, K~s, must be set to zero. As a result
(6.71)
1
(6.77)
(6.72)
(6.73)
Numerical investigations have shown that the stiffness terms arising from
the area variation, can be neglected in the present model without any significant influence on the quadratic convergence rate. Thercfore, as an acceptable approximation for K~, the second term in the first bracket of K:
(6.73) and the whole of K~n (6.76), can be negh~cted
Therefore,
where
(6.78)
(6.74)
Eflg
~
"'
11
The following sections provide closed form sol u tions originally provided
by Schonauer [84] for the first and second gap dt~rivativcs ns well as the
necessary area derivatives for ali three cases defined in Figure 6.6.
Discontinuum Mcchanics
141
(6.88)
Closed form solutions for the first gap derivatives of a straight master segment can be defined by
_a_2U_t_ = -2R..;...(x__-_x_c....;;.)...;.;(y'--_-__;:_Yc~)
8u~8u~
llx -
zc 114,
(6.89)
(6.79) .
(6.90}
(6.80)
. fJ2gn
{6.81)
'.
(y - yc)2
{6.91)
(6.82)
(6.92)
and for convex circular are segments,
llx -
(6.83)
(6.84)
8gn
8u~ =
X8 -
xc
llx - xcll
a:cll3
Again, for a concave circular are segment, equations (6.88-6.93) multiplied by -1 can be used.
Area derivatives
(6.85)
8A
au.
:::a
[A
8Al
au~ ' oui
For a concave circular are segment, the same equations {6.83-6.86) multiplied by -1 are used.
(6.93}.
(6.94)
(6.95).
which can be simplified for problema with constant thickness t,
= n, t and i, j = x, y),
(6.96)
()2gp
--=O
8uf8uj
t.
(6.87}
(6.97)
Discontinuum Mechanics
142
1 (x ~ xN)
N
2 llx - X ; 11
aA
lJ
--t
--.:- =
OUx;
= 1, 2
(6.98)
1 2
(6.99)
A =-~t (y"-yN')
OUyN i
11 X -
X N i 11 '
'
ve= [ t0 Ono ]
= 211"X
(6.104)
'fo derive au oxp1e::;::!lo11 for thc slip opera.tor, ~;i~\1rul 1111lhodH hu.ve boon
proposed. One approach considera the terms in (6.103) separately under
slipping conditions and then combines the final results to derive the slip
operator [92, 98). In an alternative approach, the slip operator may be
obtained by considering the whole matrix ~ in the context of return
mapping algorithm which arises from the integration algorithm employed
to up<late the nodal contact traction for the given gap iur.rements [84].
For axisymmetric problems, additional terms may arise due to the thickness variation. If the axis of symmetry coincides with axis y, an assumptibn
is made that the thickness associated.to the contact node is that at the slave
node itself [84]:
Here, the strategy of fully implicit (Euler backward) integration is utilized. For any iteration within the increment j, the elastic trial contact
traction vector is predicted as:
(6.100)
143
= uC.3-l
( uC.)trial
J
.
n
1
Consistent Linearization
Tn thir-; Rcction, the integrntion algorithm and the consistent linearization are
presented for the present frictinal contact constitutive law. The consistent
linearization will provide the consistent tangent slip operator needed for the
contact node stiffness matrix (6.73):
(6.107)
If the slip criterion is satisfied; i.e. the traction point lies in the sticking
domain, the correct traction is set to be equal to the trial traction
(6.102)
where the ~ term must be defined for both sticking and slipping conta.ct.
conditions, denoted by
and DP respectively,
ve
De
if
~<O:
(6.106)
LJ
6.3.3
+ (f1uC.)'1'ial
J
(6.108)
However, if the slip surface is violated, the contact traction return algorithm must be performed. According to the definition of slip criterion and
slip potential,
Slip criterion:
stick operator
<P(un, r) =
lrl -
rsL(an)
(6.109)
(6.103)
DP
if
~=o:
slip operator
Slip potential:
The stick operator is readily obtained from the linear penalty contact
constitutive law,
\ll(r) = lrl
(6.110)
Discontinuum
144
8\J!
[ 8\J! 8\J!
(bcf = 8uc = 8r' 8un
l= .
[sign(r), O)
Mech~ics
145
(6.111)
(6.120)
(6.112)
which vanishes for the present 2D case as it can be clearly seen from Equa
tion (6.U2).
(6.113)
for 2D
(6.121)
For 30 models, however, the slip potential \JI( r) = llrll includes tangential traction r defined as,
\
(6.114)
(6.122)
Linearization about the known solution (6uc, 6')') is performed by applying the Newton's method,,
(6.115)
with
(6.116)
(6.117)
The unknowns. are then pdated according to:
(6.118)
(6.125).
(6.119)
{6.126)
No wear and tear rule has been considered in deriving the Equation (6.119).
where i denotes local iteration on the level of each gauss point. The whole
iterative procedure is repeated starting from the residual calculation (6.113,
6.114) and terminated once the selected norms of the residuals rei,smaller
than a prescribed tolerance (84).
_ 8r2 = 0
86
The second term of the right hand side of Equation (6.116) is termed as
Bue
14
Discontinuum Mechanics
6.4
The corn1iRtent tangent Hlip operator can be derived from expression (6.124)
(6.127)
The simple forms of the expressions for ac, be and De, allows for an
explicit expression for the non-symmetric consistent tangent slip operator:
(6.128)
This is generally valid for any proposed frictional constitutive law. The
only unknown variable a~n can readily be derived from
(6.129)
Table 6.2, defines this variable for seven previously defined slip criteria
(Section 6.2.3)[57].
n
Law
(1)
(2)
(3)
(4)
(5)
(6)
---
(7)
acu ..
<1?
lrl - J.tlanl
lrl-Tm
lrl - alanln - ,Blanl
lrl - fLlanl
lrl-Tm
lrl - Tm sin ( ~lanl)
lrl-rm
lrl - {Lolanl + ~lanl 2
lrl-rm
lrl - Tm~ tan-
~~lanl)
nalan-n-i
In this ~ction, a general three dimensional frictional contact model using the
penalty method is presented, which is based on the work by Peric and Owen
[56]. By using the penalty method, successful applications are expected for
problems not involving high normal forces. For high normal force problema,.
other constraint enforcing methods shciuld be adopted.
In this model, kinematical nonlinearity associated with surface curvar
tures is avoided through discretization of the rigid body surface by a sufticient number of t~iangulai facet elements [56].
.
6.4.1
Preliminaries
Consider two bodies, ns and nm, which are in a state of unilateral contact
problem. The distance. between the bodies are determined by:
(6.130)
where 9n is the gap between the bodies, n is the normal vector on the
master surface, and x(0 8 ) and x(nm) are the configurnt.iou of the slave and.
master bodies, respectively.
The contact condition can be defined as:
J.b
147
+ ,8
9n ~O
/n ~O ,
fn9n =O
(6.131)
where fn is the normal contact force acting on the slave body, and the two
unilateral constraints are the'kinematic condition of no penetration and the
static condition of compressive normal force.
o cos ( ~lunl)
o - ..&un
21'm
/LO f1 +
(H;;1"4r
IJ
Tablo 6.2: Slip criteria and normal traction derivatives: (1) Coulomb law;
(2) coustant shear law; (3) parabolic law; (4) Avitzar model; (5) sinusoidal
trausitional approximation; (6) parabolic transitional approximation;(7)
trigonometric approximation for the whole range [57].
6.4.2
Following the standard formulation of the theory of plasticity, additive decomposition of the tangential gap velocity at the contact surface is adopted;
(6.132)
where
--
Discontinuum Mechanics
148
Yt =(1-nn)u
(6.133)
Assuming a perfect friction law, which states that the friction force is
proportional to the normal force and is independent of other state variables,
leads to the following slip criterion
cI>
149
<ll ~
(6.142)
A standard procedure of rate independent plasticity is followed. Therefore, under conditions of frictional slip without frict:ional hardening, the rate
form of the constitutive equation is obt.ined:
(6.134)
Jt
(6.143)
where
Consequently, simple classical elastoplasticity format of constitutive frictional contact equations can be derived,
(6.144)
g,.
ilt = 9 + iJf
(6.135)
= Dn'U~
(6.136)
fn
where Dep is the slip modulus tensor. Box 6.1 summarizes the
for constitutive model for the plasticity theory of friction [56].
proce<lti:~
Box 6.1 - Constitutive model for the plasticity theory of friction [56].
ft
P _
e_;
= Dtuf
,Yt - 'Y
8\J! (f, e)
fJf t
(6.137)
ilt = g~ + gf
Linear constitutive equation
(6.138)
Dn
= -annn
Dt = -at(I- nn)
e= i'h(gf, e)
(6.139)
with,
in= Dnu~
it = Dtf
Slip criterion and nonlinear adhesion hardening law
Dt
= -at(l-nn)
(6.140)
(6.141)
where Dt and Dn represent the tangential and normal parts of the elastic
modulus tensor, respectively, \J!(/, e) is the slip potential and h(gf, e) is the
hardening/softening law for the adhesion e (and not frictiori).
!
Loading and unloading conditions may then be formula.ted in the standard form,
'1!(f,c) =
11/cll -
gf = iT ,
T =
ff.tt
Loading/unloading condition&
q> ~
150
Discontinuum Mechanics
151
t.
6.4.3
N umerical Integration
+1
- f t,j+l
1
- 111tJ~ 1ll
(6.152) '
(6.145)
Applying the chain rule of differentiation,
(6.146}
(6.154)
LI
(6.147)
(6.148)
8t;+1 _
{}/trial t,j+l
{6.149)
11/trial
11
t,j+l
(I _ t.
.
3+1
t.
J+l
--'(;'
(6.155) ,if,:.
/_'
A two phase elastic predW;or/plastic corrector algorithm is utilized to update the frictional contacl9stress state according to slip potential and nonassociated slip rule.
(6.156)
1
where
If the radial return method is used, in the elastic predictor phase, incremctits of displacements in the contact region are calculated:
trial
= f j +D
f i+l
'Uj+l
(6.157)
(6.150)
Then, th'e plastic corrector phase begins with radially projecting the trial
elastic stress f ~~~t onto the slip surface
(6.151)
Comparing consistent modulus (6.156) with the continuum tange~t modulus (6.144), the stiffness in the tangential plane perpendicular to the slip
direction is reduced by the fMtor at,c For largc displacement increments,
the continuum tangent modulus (6.144) does not provide quadratic rate of
convergence due to the fact that Ot can become considerably larger than
Ot,c
Dlscontlnuum Mechanlcs
152
153
Update configuration
Xj+l
( i)
Xj+l
(i)
(6.159)
+ Uj+l
(6.160)
IF 4>}~1t
= llftJ+'1ll + ll/:::~~111 :5 O
and hep is the standard el~toplastic constitutive modulus. The terms on the
right-hand side of (6.158) are linear in u and provide the tangent stiffness.
Finally, the following expressions for the material, p;mrn1otric and frie .
tional contact tangent stiffness tensora, Km, K 9 and K e, respectively, can
be readily derived:
tH1
= nt!~m,,.,
f t,j+l
:::;;
11illftJt'1iltJ+1
(6.161)
_ a
t,c -
D~P =
11/ t,H1'1
t11J:~if1t11
K9
1.
Vu : (u 1) : V' r,dv
{6.162)
X(O)
ENDIF
(6.163)
where '\1 8 is the symmetric part of v. For the residual force we obtain,
(6.164)
6.4.4
FE Discretization
-r(x,17)
= f
lx(n)
\i'u: a: \i'rdv +
lx(roc)
(6.165)
u: D; rda
(6.158)
1
t is the surfacc
is
In a finite element simulation, the subspace vh e v is set as an apprax:imation to v space, where vh is J0D_!'rated through spatial finite element
discretization
154
Discontinuum Mechanics
155
(6.175)
nel
x(O)
= LJ x(Oe)
wth
x(Oa) n x(Ob) =O f a 1= b
(6.166)
where
e=l
(6.176)
nen
uhlx(Oe)
= 2: N aUa
(6.167)
a=l
(6.177)
wlwre ncl and nen are the number of clements, and the number of element
nodes, respectively. By defining the discrete gradient operators,
..... .-.
nen
V'uhlx(O.:)
==
L Baua
""'ne
(6.168)
F = F nodal +
a=l
nen
Ar + A
e=l
cn=l
c,cn
(6.178}
nen
"Vuhlx(Oe)
=L
GaUa
(6.169)
a=l
6.5
(6.170)
In this section, the formulation of three dimensional frictional contact models is extended to include frictional hardening or softening behaviour, based
on the work by Hashimoto et al. (72].
6. 5 .1
(6.171)
(6.172)
(6.173)
According to the classical Cou~omb friction law, the friction force is independent of the apparent area of the two contacting bodies, and is proportional
to the normal contact force between them. It is g<1wrn1ly accepted that the
kinetic friction is almost independent of the speed of sliding.
If A is the real contact area, and H is the hardness of the softer material
in contact, according to the adhesion theory, the normal contact force, /n
is defined by:
fn=AH
(6.174)
and the frictional force required for sh~aring of the junction of asperities,
and the element residual force vector,
''
156
Discontinuum Mech8Ilics
(6.182)
Table 6.3: General categories of dynamic frictiona.l effects (72].
Type
Category
Charactetistic
Initial stage of sliding
Small amount of sliding distance
Middle/late~,stages of sliding
..
Is =Ar
Figure 6.7 depicts various mechanisms of fransitional changes encountered on contact surface with regard to several surface coated sheets. The
main features of each type of frictional phenomena, may be summarized as:
Types (a): ,
. Progressive sliding leads to Hattening of asperities.
. Lower tangential force is required for asperity deformation.
(6.180)
where r is the shear stress of the softer material in contact. The coefficient
of friction may then he defined:
n
LJ
s
= n
T
= H(Rj
0.2
{6.181)
Type {b):
Softer surface coat than types (a) and (b).
. Larger asperity defrnation is expected than those of types
(a) and (d) with the same frictional work.
Larger frictional tangential force is required once the friction
condition is changed from the mixed lubricant condition to the
boundary lubricant condition
. Larger force is required for shearing the junctions
The main differ?nce of this section with Section 6.4, is that is no longer
a constant and has a hardening or softening. behaviour.
U nder very high .normal force, the real contact area no longer follows
fn. =.AH due to work hardening of materials. Other real industrial apphcat1o~s
may also exist that the conditions of the classica.l laws can not
be apphed. Table 6.3 compares different categories of frictiona.l effects in
dynamic friction problems (85, 72]:
Type 1 is related to quasi-static contacts with dry friction under relatively small sliding distance of an order of magnitude of t~n microns.
The adhesion theory may be effectively adopted for analysing this type
of problem.
Type 2. takes place in dynamic loadings with sliding friction. It shows
large rigid relative motions. The plasticity theo~y of friction based
o~ deco~position of displacement into elastic deformation and sliding
d1stance is used as the analytical approach.
1
Lhe required
Type (e):
Larger asperity deformation
. Rapid increase of ehe actual contact area
. Wear particles lmprove the lubricant condition
. Relatively stable frictiona.l behaviour unt.il occurrence of contact with the bulle metal.
Plowing of the zinc layer because of low resistance to shear.
Types (d):
. Progressive sliding leads to Hattening of coating layer ..
. The lubricant works 88 a weak layer t.hat reduces the required
shear force at contact junction.
158
Discontinuum MechMics
_.
159
LJ
tool
a) Cold rolling.
w=
b)
~nd
w,
= (Jn,w)
(6.184) .
f n and
(6.183)
:J
e)
tool 1
....._ _ _ _ _ _ _to_o_I_.1
~__, Zn;~~::c~:;oco~~=~]j~;:cog>''~
,__e_st_ee_
______
d)
(6.185)
6.5.2
Following the same procedure as the classical elu...'itoplmitic nnalysis and frictional contat behaviour without hardening, we begin with the additlve
decomposition rule of the displacement rate.
.
.. _--.1
(6.186)
Galvannealed coating.
(Zn-Fe alloy coat harder than the bulk material)
Ut
= (1-n n) u
(6.187)
o~ stick) and:
Discontinuum Mechanics
160
161
(6.188)
it =
--
Dt e = --at(I - n n) e
(6.i89)
,,
fn
~O
n 9n
=O
(6.190)
slip direction
b) Slip criteria.
Slip Criteria
Figure 6.8 illustrates the associated and non-associated slip rules for linear
and nonlinear friction laws [72]. It provides sorne general understanding
of the expansion or shrinkage in the radial direction of the siip surface
according to the amount of frictional work parameter w. The slip criterion
rnay be defined by:
l?(/' w) =
11/tll -
(J n111)ll/ nll
~ ~ :!!~k
(6.191)
non-auociated
Assuming the slip criterion (6.191), the associated slip rule generates a
normal force, causing normal separation of the surfaces. In order to avoid
this slave body separation from the contacting surface, a non-associated slip
rule is typically adopted [86].
The slip potential '11 can be defined as. a cylinder with radlus
isotropic frictional contact:
\J!(f)
,,
= 111t11
11/t11
linear slip
surface
for
(6.192)
'
and the slip direction is then defined as the outward normal to the slip
potential '1!:
Discontinuum Mechanics
162
1
p
Ut
{J'iI!
= -"(8ft
(6.193)
or,
f= -At
(6.194)
= ~ +f
= (1 -
n n) 'l.t
in
= Dn e = Dn Yn
= Dt e
Dn = -an(n n)
it
(6.195)
To obtain the continuum. tangent modulus Dep,
consistency condition:
i'
Dt =--at(l-nn)
Slip criterion
f)p) .
( oft
ft+
( f)p ) .
fJfn
fn+
(f)p)
8w
w=O
(6.196)
\J!(f) = llftll
itf = -'Yt
and then substituted into the constitutive law, defining the following linearized forra:
(6.197)
where
Dep
+ (! n,w)llJnll
8.:V -~
- 8Tt
- llJ1ll
t-
Hardening law
w= - f t itf = i'll f t 11
-Ot
[ ~ll 11:1(!!.i!.)l
t
(3 "
Dw
'Y2:0,
i'4>=0
t t) .
(6.198)
sliding direction on the contact surface. The last two remaining terma,
indicate contribution of hardenittg/Sbftening phenomena to the stick and
slip stiffness matrices in the sliding direction, respectively.
Box 6.3 summarizes the constitutive model for the pla...,ticity theory of
friction with nonlinear evolution [72].
with
f3=1+(fn,w)llf
Ot
11 2 (
"
8
8w
(6.199)
The first term of nev indicates the stiffness in the normal direction,
while the second term, denotes the adhesion stiffness perpendicular to the
6.5.3
Among severa} available numerical integration schemes [105, 102, 106], the
implicit backward Euler method is a.dopted in this section. Similar pro~
dures may be followed for alternative numerical integration schemes. Box
Discontinuum Mecluwics
1()4
6.5.4
1.1)
1.2)
1.3)
,(/~~H 1 ,w;), ~
n,j+l
= f n,j + Dn AUj+l
and
anf.. n
= Jt,j + D t ''-\Uj+l
A
IF
ci>trial
Hl -_ llftrial
t,J+l
or
SET
11 - (/(i)
n,j+l' W_ )11/ 11,.J 11 l
~O THEN
EXIT
il!
ELSE
LJ.
df ]+1 = -~--------. au
<I>t.rial
fiTe
J+l
(6.202)
j+l
8w j+l
resulting in
df ;+t = Dnd>u+1
A"'(=
+ D,dt.u;+1 .
11/CriG&
11 trial ll / n,;+111
C.11 -i+l
ac+,.triG&(.?fu.)tric.111/
. 112
,..;+1 l1W 1+1
n,3+1
A
t
t,j+l - '-\"'(C<t
l trial
Wj+l'::::: Wj + llft,J+lllArj+l
Jt,j+l
=i
<l>j+l
--
(6.205)._.
SET
ep
GOTO 3.1
ENDIF
(6.206)
i;.
,::;;; . >
l
f(f i' Wj, AU;+t) = fj +A/ n,j+l + Aft,j+l
186.
ENDIF
Discontinuum Mechanics
166
nep=
e
167
11
-an(n n) - t.llf !;.{U (I - n n - t t)
timestep j+ I
t,;+111
11
timestep j
n
(6.207)
Ht.
.
On
(-4-)
llf
811Jn11 . .
__ ,,.,,,"'"'
------
11] (t n)
!'j
L_j
= 1+ (H1ll/n,j+ill
Ot
6.6
6.6.1
_b/y) (aw)
.
llfn,;+111
(6.208)
j+l
8 -
X1) . t
= Yt
+ (eJ+1 - eJ)l;
(6.213)
.~'
k
1
K=Kn+Kt
+1
=
O!n
.
r({}gn.)T
8.'9n + gT !_ (Yn)]
8u
8u
nau 8u
(6.214)
(6.215)
L ,
(6.209)
)]
K -. t .[(8gt)T
-8gt + 9t -8 (Yt
8u
8u
8u -.
aut -
Yt,j+l
(6.212)
Kn
T
1 (X
= O:n9n
_,., g
t
l int
t
- v.t t.au ,
Based on the work by Crook [52], the necessary formula.e for a two dimensional nade to straight edge frictional contact problem are provided in this
section, though the general p.umerical formulation has previously been discussed in detail. Figure 6.9 shows a typical node to edge contact problem
in a two dimensional space.
ec =
8g '
{}g
= a g __!!,
f int
n
n "au'
'''
;e
(6.210)
(6.216)
(6.211)
lj {T + r9nN m }
f tint = t9t y
(6.217)
(6.218)
i'
169
Discontlnuum Mecha.nlcs
l8
>
,,
where
langlllllal
lilllplltlol
pe. 1,
/Ol'OI
T; = [t
- (1-e)t
(6.219)
-etJ
11uvd111um
1hear1trus
T~ =[O
N! = [n
(6.220)
t]
-t
-~n]
- (1-e)n
(6.221)
Figure 6.10: Coulomb friction law.
N~ =(O
(6.222)
-n n)
(6.223)
;~ = sign(ft)
""
(6.230)
The stress update procedure may then be.followed as (see Fig1_1re 6.U)~
D
B = {
<eJ+i - eiW
(6.231)
I + O!tU.9t
A
t,j+l = j
f trial
(6.232)
!:l.gt =
(6.224)
~:~J+'1 =
111:1+'1 ll - fn ::; O
~t [111:1+'1 ll -
fn,j+l]
!:l.gf = A-;sign(ft,J+l)
The classical assumption of Coulomb friction, as stated in Figure 6.10, is
that the body is at rest until the maximum shear force is reached. However,
in reality, bodies exhibit sorne relative movement within the contact area
prior to sliding.
slip rule:
slip potential:
(6.226)
l/tl -
(6.227)
,fn
8\1!
1.:f = !:l."( 8f t
'I!(ft)
(6.236) ..
(6.225)
!t = atgr
~ =
slip condition:
(6.235)
(6.237)
gf + gf
elastic region:
'(6.234)
(6.233)
(6.238)
resulting in:
(6.228)
1
(6.229)
(6.239)
Discontinuum Mechanics
170
171
f,
a) Geometrical description.
where
C={Ts+ 9Nm}N';
(6.240)
b} Basic vectors.
3
6.6.2
x,1
)-11,
x,
011
Frictionless Contact
e) Initial and current
configurations.
Using the standard shape functions Ni(e, 77), associated with node i, the
position vector r may be expressed by the nodal position, W,
.
Discontlnuum Mecha.nlcs
172
173'
(6.241)
(6.248)
(e, 77).
N2 = l{l - e)(1 + r)
N3 =
N4
(6.242)
Hi - eH1 - ,,,)
= i{l + e)(1 -
where,
77)
(6.250)
and
U
(6.251)
(6.243)
o
o
(6.252)
Nf = (t/.)
r), (1 - r)}
NT=
7J
(ii-)
NfTJ=
(,N)
8(8rj
(6.253)
where
aml the vector connecting s to its estimated projection onto the quadrilateral surface, e, g,
(6.245)
(6.254)
Once the correct value of (e, i]) is known, the vector g will be orthogonal
to the two vectors r~ and r,, which lie on the master surface (See Figure
6.12b), i.e. g 11
{6.255)
(6.256)
. aE=rEg=
or
ae
{6.~~6)
';.~~ .~\:
w~~:.1?i~~t"::,ic:f:~{
174
175
Discontinuum Mechanics
(6.267)
(6.259)
so 8HT and n are to be determined. For determining n we have,
(6.260)
n =
fgjj
(6.268)
so,
(6.269)
(Ao1dl
+ DJI{, =O
~ ~]
(6.261)
9ni
(6.270)
whcre
calculated:
(6.262)
(6.263)
[ Ge ]
G,,,
'
G~ = H~n
G.,,=H.,,n
(6.271)
Therefore,
(6.272)
g,. =
(;n g)
70
(6.264)
g = n [Hu-A6(]
where nis the normal vector to the surface at the contact point. Knowing
ATn=O,
= {H - An- 1YT]u
with
(6.274)
(6.265)
= ~,
resulta
(6.273)
in
(6.275)
(6.266)
'
Discontinuum Mechanics
176
177
(6.285)
(6.276)
where
(6.277)
(6.286)
K e =K1e+K2n
e
(6.278)
(6.287)
where
(6.288)
(6.279)
e= [
t
K~n =
~ [HT H - HT nnT H
ff1
{ ff2
= Ot9tl
= Ot9t2
(6.289)
ft ]
ff2
'
(6.280)
(6.290)
with
Sticking Friction
Referring to Figures 6.12b,c, which depict the initial and current configurations, the tangential gap vector, Yt, can be defined by:
(6.291)
(6.281)
(6.292)
where
(6.282)
(6.293)
(6.283)
(6.2&4)
i
=E Ni(~, TJ)Xi
~ ~]
(6.294)
and
Discontinuum Mechanics
178
F"' = [
~ ~
179
.,...-jc,trial
K,-
{6.295)
] /"'
(6.296)
11J~ 1 trial11
(6.306)
BT/cHl
r e_
-
(6.307)
g = Bu
(6.308)
(6.297)
Q= [
~f]'
(6.298)
(6.309)
G~ 11
=n
7'H
(6.299)
~r1
(6.310)
(6.301)
with,
Sliding Friction
(6.311)
Finally, the nurnerical formulation is extended to incorporate sliding .frictional behaviour. Assuming a Coulomb type of sliding friction,
<P
where
UDT = (ffi,f{2 ).
For (j
= 11/fll + f~
(6.302)
(l _
O:t:l.'Y ) c,trial
llf~'trialll
= K./c,trial
t
(6.304)
(6.305)
Far non-associated slip rule, /:l.gP is normal to the cylinder
Constant, hence, ~Y~+l = /:l.gP,trial. Frorri <I> =O, we get,
(6.312)
(6.303)
e .
J t,J+l
f n,j+l = O:n9n
llnll
L_,
,----!
Chapter 7
Discontinuum Contact
Mechanics
7.1
Introduction
The pioneering work by Cundall and his colleagues, who completed the original work by Goodman in 1968 (108} on jointed rocks, marks the beginning
of modelling of discontinuum media [109}. They developed an algorithm for
modelling the behaviour of jointed rigid rocks, soon termed as the Distinct
Element Method.
By advancing the capabilities of the finite element method, and increasing power of computing facilities, fully defo~mable blocks replaced the original rigid bodies, with the new Discrete Element Method terminology.
Nowadays, the discrete element method has reached t.o a.u ever increasing
popularity for modelling ali potential discontinuum media. Nevertheless, it
is mainly used for two classes of problems:
181
182
Discontinuum Mechanics
7.1.2
The essential point is that the finite element method is rooted in the
concepts of continuum mechanics, thus not suited to general fracture propagation and fragmentation problems. The firiite element method may only
effectively deal with a single crack ora low fractured area without any fragmentation, whereas the discrete element method is specifically desigried to
.solve problems that exhibit strong material and geometrical discontinuities.
'
;,
Applications .
The discrete element method has found extensive attraction within various
fields of the industrial and scientific applications. The main categories are:
l. Geomechanical applications
........ ~
Before dealing with the main issues, a quick review of historical qevel:
opments and present industrial/scientific applications is provided.
7 .1.1
183
Historical Development
3. Granular nm.teritilH
4. Industry
- 3D printing
5. Impact analysis (progressive fracturing)
- Composites [39].
[119].
7. Computer Graphics .
7 .1.3
L 1
I t should be noted that for each ca.se, earlier less sophisticated modela
can also be found in the literature and the mentioned years show the time
of major advancements of the method.
--~-------------ft - - - -
Speciflcations
Table 7.1 briefiy provides compaxative information about the principie speclifications of continuum (finite element) analysis and discontinuum (discrete
element) analysis.
-!
~'."; :.,
:.':;
i:
:~~.
t{'.j,
Discontinuum Mechanics
184
185
l. Object representation
2. Contact detection
continuum
- only finite elements
- explicit definition of contact surfaces
geomctry
3. Contact interaction
4. Physics of objects
discontinuum
- finite elements (FE) and rigid"'bodies (RB)
- automatic evaluation of contact surfaces
5. Visualization
This chapter mainly deals with the contact detection procedures closely
coupled with object representation. The contact interaction algorithms have
already been discussed in previous four chapters. Physical behaviour of
objects may essentially aff~ct the material modelling (plasticity) within the
finite element procedure, and may also cause cracking/fracturing which will
be dealt with in the next cha.pter. Finally, visualization is a necessary tool
for a successful discrete element analysis. Changing geometry and mesh,
creation of new bodies, and checking the interaction stresses/forces require
a powerful visualization tool, without which, the huge bulk of result data
may not ever be verified and discussed. However, it is a mu.tter of co111n:tl:\r
graphics and CAD /CAE systems and is out of scope of this book.
continuum
- simple algorithm
global search
discontinuum
- must be efficient for large numbers of bodies
- must be efficient for dense/loose packs
. ; _ must be applicable to FE a.nd RB
.....
continuum
- no9e to node contact search
- simple surface node to element edge contact
local search
contact
interaction
discontinuum
- complex algorithm (arbitrary shapecl bodies)
- different algorithms for different types of RB
7.2
continuum
- interface elements
- n<?de/node, node/face and face/face contacta
- penalty and various Lagrangian formulationa
In this section, the contact detection procedures are briefly revi~wecJ. and
their main advantage/disadvantage points are discussed [125). Then, the
alternating digital tree, as one of the fastest geometric intersection sea.rch
algorithms, are explained in detail and its application to general contact
detection problema will be reviewed by providing samplc probl~ms.
discontinuum
- node/node, node/face and face/fac,e contacta
- penalty formulation for FE bodies
- closed form solutions. for rigid elements
- energy conserving/dissipating algorithms
The problem of detecting the interacting bodies in multi-body analyses, also known as the geometric.. intersection search, has become a major
computational research topic ..
- - continuum
- explicit or implicit dynamic solution
solution
discontinuum
- only explicit dynamic solution
1---
continuum
- simple deformed geometry visualization
visualization
discontinuum
- advancecl graphical postprocessing
- continuously changing geometries
Contact Detection
1
1
N-1
N-2
(7.1)
Discontinuum Mechanics
186
2
3
4
2
-3
4
0
D
[[]
187
(7.2)
The existing detection methods have so far laid in between the two
extremes .
The alternating digital tree (ADT) algorithm, which developed initially
to solve the problem of mesh generation, reduces the number of operations ,
required to deterniine the contacts between bodies by creation of short lista
of potential contactors for each target body. Figure 7.2 shows a sample part
of the created short list for a set of N contacting bodies. In this case, a ,
direct checking is undertaken for the number of relevant bodics of a ta.rget,
and the procedure is repeated for other target objects.
Figure 7.1: All to all check, the simplest contact detection procedure.
7.2.l
Contact Geometry
The contact geometry is then either computed from the input definition
of rigid bodies, e.g. a circular disk is defined by a centre point and radius, or
188
189
Dlscontlnuum Mechanlcs
F.
.E
.,,, ...
___/
a 1
cient for both loose and tight packs of elements. A single approach might not
achieve all the mentioned goals, and different approaches may be adopted
for different applications.
Contact Geometry
1 3
3 2
2 1
5 6
6 7
7 8
8 5
1 2
6
2 3
4
2 4
5
5 6
2
Direct Checking
Assume there is a system of N interacting bodies; all may happen to come
into contact with any other body (Figure 7.6).
The simplest approach for detecting the potential contact8, will check for
contact between each body and every other body within the i;ystmn. Figure
7 .1 shows how such a simple approach requires a checking link between each
(target) body and the remaining (cont~tor) bodies.
7.2.2
N-1
N--.
2
(7.3)
Discontinuum Mechanics
190
---
F.
,,..---l. . .
' , . 1/
1 ',
\ /
\
\'
......
.""
'
'w""
..... __ _,,_,
191
'
/~
~
/\
,._-'1
\/ G
,,,,---1. . . .
y , He1
/\
/\
9-\-~v~~-
Figure 7.7: Defining a potential contact zone for a system of moving bodies
( disks ).
In this method, the mtinber of contact node searches for ea.ch target ~ur
face is reduced by forming a reduced set of nearby contact nodes, thereby
reducing the total number of contact searches. This reduced set (short list)
has to be periodically updated. Figure 7. 7 shows a simple way of defining
a buffer zone for creating a reduced set of ,nearby nodes.
Therefore, the first step in such a search algorithm is to define the buffer
zone for each target, and then finding the disks iuside the specified zone
(Figure 7.8).
By knowing the velocities at each time step, we may approximately
evaluate the distance in which a disk can travel in next n time steps. Accordingly, the need for a new glooal iearch for fiuding nodes inside each
new zone can be decide<l from the leaving/entering nuy nodPs from/into the
zone.
Afterward, only the associated reduced set will be searched for potential
contacts with any node.
Almost all of the developed search algorithms differ mainly on how to
create such a short list in a faster systematic way for specific or general
applications.
7.2.3
Buffer Zone
As indicated in Figure 7.7, a buffer zone may be defined for ea.ch target. In
L
cai;;t~ of rnoving disks, the buffer zone may be defined by a circle with radius
n,,.
(7.4)
Therefore, no contact is expected to happen before N t::.t time steps and
the global search to locate all contactor objects within a target buffer zoe
is performed every N At time steps.
The sarne concept is also used to enable direct evidence checking in association with meshed polygons and more complex object shapes. Different
rules for forming a buffer zone can be utilized depending ou the object shape
nml type of application:
193
Discontinuum. Mechanics
192
r--
..,,
l~-r--------,.-"":1
/Buffer ZOne
,,
LL
1
1
1
1
1
1
1
.~
1'
11
1
.,
'
"'
-r--------ll..1 ',
1
1
1
1
1
1
1
',
'
'
'
'
',
'
'
',
'
'
,---------------.....-'
1
',
',
'
'
~'l
}'
~---~-------------------../~'
,c.":'-1--------~---------,
1
11
lt1
'1
1
1
1
1
1
Figure 7.9: Definition of the buffer zone; uniform expanslon of the mesh
boundary (left), and a rectangular box (right).
11,,
+-~------.,1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
~---------,-------1
-h,
I~...:---+- - - - - - - --!:-----..:---..:-..:---..:-..:-..:-J-'
Figure 7.11: Forming the buffer zone based on each segment geometry.
In order to minimize the total time spent on up<lc.i.Ling Lht: shcrt Hst
of contactors, the size of the buffer zone should be defined with respect to
the actual size of the bodies under consideration. Sorne reports proposed an
optima! size equal to the average size of ali bodies, although its applicabllity
to general meshed polygon systems ls in doubt.
If the relative motion of the bodies is small, a small buffer zone will be
most efficient. The faster the IJ\Otion of elements, the larger wll be the
necessary size of buffer zone.
Flying Through
j___
Based on the dimension of the buffer zone, db, t.he tinw stcp increment
Discontinuum Mechanlcs
194
195
Bufferzone
contactor
flying
through
the target
t'~' tf':-~
................
I
"",
}
I .,
n
'
\
.\
\\
>
///
Figure 7.12: Flying of a contactor object through the buffer zone of a target.
lJ
must satisfy,
(7.5)
where
7 .2.4
Vma:t
For example, consider Figure 7.13, where after the impact, point D has
penetrated the body B AC. However, by only monitoring the displacements,
the program can not simply decide whether the point D has penetrated
though the segment B A or CA. The !fme problem may arise near the
corners, where sveral paths seem valid for the penetration of the point lnto
the body.
1
Discontlnuum Mechanlcs
196
l::i..t <
-
d2 +d ( 1- d7
d2)
-;
b
V~
(7.6)
where d is the current distance between bodies, d is the field (contact) zone
thickness and v~ is the relative normal velocity between interacting bodies.
The condition 7.6 ensures a gradual change of,tfie time step increment
during contact, leaving a :QJ.inimum number of time steps for treating a
general impulse/contact problem. As a result, the accuracy of the solution
is also preserved.
..;""'
iR
(j
A binary tree structure is a specific method of sorting data that allows new
data to be ea.sily added (inserted) or removed (deleted). Binary trees a.re
one of the most important non-sequential types of data structures [126]. At
each node, the information, stored consista of data and two pointers known
s the left and right links that point to further data. Each added link can
either be equal to zero or equal to the position in memory where another
node of the tree is placed.
1 -------- _,
rcm ~
rn:;i
kwlD
7.2.5
------------------
l.@~1,,
~ ~
----lrle/2
,,,..,
. \
""' ......
1111
,,..
rlght
notk
m3
,4
x-4
-B
x.
"' "-,;;-\
"'s"',,........ . . ,,61........
' ',
... ....,
\
'"'e
-o
~- .. B
\
'F
,,,
Xc
'
Xo
7.,
"
'
x,
le
Discontinuum Mechanics
198
199
----------------~ level O
(011)
------- level 3
(110)
a) Degenerated tree
A nade in the tree can only be visited or examined if all its ancestors are
visited first. However, an efficient systematic retrieva.l of data requires that
each node is only visited once. This can be achieved with a recursive scheme
of the so called stay left rule,
l. Retrieve data for the node at iev;i i.
+ 1.
3. Else if the right link exista visit the pointed node at level i
+ 1.
Figure 7.19 illustrates this simple rule for a typical tree. The order of
traversa! is: A-+ C -+ D -+ E-+ F-+ B -+ G --+ H. Tlw procedure can
be summarized as Box 7.1.
Discontinuum Mechanics
200
201
Programming languages such as C++ and F90 support automatic recursive algorithms, thus a relatively easy implementation of the traversing
technique ca.11 be performed. However, the more common F77 language does
not provide this capability and morti complkatcd implementation, such as
the stack technique is required.
In a so ca.lled stack technique the items are inserted onc by one, and then
extracted in the reverse order one by one. During the insertion procesa, the
value of the right link is stored in order to enable the subsequent traversal of
the right subtree. In fact, a list containing the address of all right subtrees
encountered along the way which are yet to be traversed must be kept and
continuously updated [126] 'j
The traversa! continues until a zero left link is encountered. Then, the
last right link inserted in the stack list is retrieved. The subtree rooted a.t
this address is traversed and the corresponding right link is removed from
the stack.
Box 7.2 - A traversa! algorithm
\uu;~
l. root_address
2. size =O
2.1 - visit the node stored at root_address
= size +1
leftJink
- GOTO 2.1
2.5 - ELSE
- Visit B
- Take the left child (G) and traverse its subtree
- Visit G
- size = size - 1
- GOTO 2.1
- ENDIF
2.6 - ENDIF
202
Discontlnuwn Mech8f1lcs
203
7. 2.6
The last remaining point about the binary tree structure is to define
a methodology for removing a node. If the undesired -TIOde. is a terminal
node, it is sufficient to change to zero the corresponding link of its father.
Therefore, the occupied memory will be available for new node insertion.
Space cell decomposition methods, in general, comprise finding the suitable shape of a cell, space subdivision into cells, and cell enurneration.
In the following, various methods of space decomposition are .briefly
reviewed.
il
i
Space Decomposition
simple modification of the left and right Jinks of the father of the unwanted
node will be sufficient. Figure 7.20 illustrates the deletion process for a
simple binary tree. The deleted node C has been replaced by one of itJs
descendants terminal node H.
In this method, each coordinate axis is divided into finite intervals and ~a.ch
finite interval is given a number. Ea.ch cell is defined as a set of points that
have coordinates that belong to the same interval (Figure 7.21).
If a particular application demands frequent deletion/insertion processes, a memory rnanagernent system is necessary for the efficient implementation
of tree str~ctures so that new nodes can be placed in the memory space
released by the deletion of previous nodes. With programming languages
featuring dynamic allocati9n, this is automatically provided, F77, however,
<loes not offer this capability, and the problem should be handled by using a
linked list structure to record all the available memory spaces. For further
details see [126].
Each cell can be enumerated by two (for 20) or three (for :m) numbers.
Instead, a single cell number may also be constructed using a binary number
structure.
(6,7)
8
Cis removed
7
6
C '1
-......_
(7,2)
l_
Discontinuum Mecha.nics
204
205
o
2
? 8 9 10
11 11 JJ 14
10
11
13
11
To better explain the way the tree is constructed, assume the problem
stated in Figure 7.24, which consists of eleven disks moving in a domain.
To form the tree an optional order of. points should be assumed, e.g.
B - t C -+ D -+ E ~ F -+ G -+ H -+ 1 -+ J ~ K. If the root
(here A} or the order of the points is changed, another tree structure will
be formed.
-t
1
1
- - ' - - - t
---1--------<b----
. .-tr--'.! o_J!______ _
t
'
1
1
---+:: -
1
1
'
levelO
level 1
leve/ 2
level 3
level 4
- - farther levels
1 .G
K
J.
n
LJ
DiscontlziUum Mechanics
206
207
"
--------------~-------'
.,---------'----"-----
..
"
--------=,------'-----------.
.9.f '
.E
2-
'
:~,)
--~-
" e
--------I. _ _
J_-'----------------~
'
2 -- ------ -A --:---+--1
C
1 " e
Figure 7.25 shows a step by step procedure of constructing a tree associated to binary space decomposition of problem stated in Figure 7.24.
The main advantage of the geometric ~earch based on the binary tree is
that the area associated with descendants of a no<le A.t. l'lny leveJ ii:; always
contained within the area associated with "the node [52]. For example in
Figure 7.23, the area associated with nodes 3 and 4 are contained within
the area associated with node 1 (and so node O). Therefore, in a contact
detection analysis, if a target does not intersect. with tlw area. associated
with an object, the object and ali its descendants will not be considered as
potential contactors for the target; i.e. they will not be included in the target
short list of contactors. Figure 7.26 illustrates this key feature. Objects A
and B and ali their descendants can not be a candidate for contact with the
target (as it is clear!y observed from the figure) .
Hextree Decomposition
Je
_____J__
Discontinuum Mechanics
208
209
third level
aggregate 1
leve/3
- - level2
- level 1
Fip;urn 7.27: First lovel aggregate for a hextree space decomposition [127).
lJ
secondlevel---....._
aggregate 1
level2
Figure 7 .29: Third level aggregate for a hextree space decomposition (127). .
- - levell
23
04
4
Figure 7.28: Second level .~egate for a hextree space decomposition [127].
Discontinuum Mechanics
210
211
Adaptive: Grid
In this method of space decomposition an adaptive grid of cells is used. In
this method, a variable fine cell positioning system is used to define the
space which is occupied by the objects. Elsewhere, coarse grid of cells will
be used. Figure 7.30 illustrates the way an adaptive grid is constructed for
a 2D problem.
'~ ,
1
Y1
-"
L.
....
.... '
~-
____....,.______--4________.....,___________
--
'' -e'
~
"'
...
,.
i~
be sorted, and the second array (rank a.rray) is uscd to rnnk the order llf
the objects when they are srted.
A hoap is gen~rally defined as a trcc with the property that a.11 of the nodes
below a given nade contain a key value that is less than that of the parent
[24].
Figure 7.31 illustrates the heapsort algorithm for a collection of 20 objects. The lower bound extents of the object boundaries are projected onto
the. X and Y roces. To perform the spatial. heapsort, .two arrays of integer
fields are require<l for each dimension of the problem doma.in. The first
array (id array) of each pair is used to store the set of object identifiers to
1. At the first step, an object from the id array is selecte< 1 (called pivot),
which is the lower bound search index. Then using a binary search, the
2. The same procedure is followed for the remaiuiug axes to identify both
an upper-bound anda lower-bound index for the pivot.
Discontinuum Mecha.nics
212
213
.;-"'
I
I\
<> \""'"'.
'
'-...-/
....,....,
..... .....,
..........
o
/
--'
/'
/
~'7'
body
based
cells
'O'
,_/
\
7.2.7
...., . -...,,
(7.7)
214
Discontinuum Mechanlcs
215
Basic specifications of binary trees have been previously rliscussed. The data
items stored at the nodes of a binary tree, whkh is cwated with respect to
the geoIJletric intersection search, must contain iuformntion on geometric
properties of the bodies under consideration.
By performing the location mapping (7.7) for all bodies from the Rn into
the R 2n space, the creation of a binary tree data structure for a geometric
intersection problem can start. It begins by selecting one body from a given
set as the root and continues by inserting new nodes.
The node insertion criterion for the ADT is based 011 decomposition
of the space region assigned to the body at a particular node by a space
bisection algorithm. The bisection is performed with respect to the body
location coordinate component which is used for checking the relative position during the tree traversa} and in turn is associated with the binary tree
level containing the traversed node'[l]. By defining
L_J
_ [, 1
.
n
.
1
. n
Yi,mint ... , Yi,min1 Yi,maxi ... , Yi,ma:c
Yi -
]1'
(7.8)
Uf,min ~ X. ~ Y,ma:c
f or
j = 1, 2, ... , 2n
(7.9)
..
,.
bodyi
Intersection Search
~
1'
x,..,,,,
...
The recursive nature of the space bisection process for creating the binary
tree data structure provides an appropriate basis for nn efficient automatic.
geometric intersection algorithm
In additibn to the information of the mtt.pped points in the R 2n space,
each tree node contains the information of its as1mciated bounding h,y-pc.rcube region. A hypercube stored at any node of the biuary tree supersedes
all bounding hypercubes stored at the descending nodes, providing a key
feature to the fast geometric intersection search algorithm.
Discontinuum Mechanics
216
217
'~
.....
x',._
- ...
x'
l~IO
.'
40
JS
AS S)
JO ..___
(J~ 4)
CIF
IS
_..__...__.....__._..__....__...__.-........_ XI
IS
:JO :IS
tU
..
10
P( 4,11
10
CJIB
10
l'IW
:4
:JS
level 1 .
Jllil
30 JS 41
o
I
10
IS
21 JJ JI
--
JS 40 X!.,,.
AD
cCJ
level O
Y,";.r.
level 1
L.J
BD
levd2
levelO
~
q.1
.iL
D[J
cm
D&'.]
cm
LI
,......,
!~h.._
level o
______________.
kvell
[~
o
lnll2
levelO
AD
EO
level O
levell
'b
""'' 1
~,.,,..
1...,.:1k+-______.._______...
,.,
Figure 7.35: Creation of an alternating digital tree data structure and ddn~
ing the coordinate limits of the bounding hypercube region.
218
Discontlnuum Mechanics
ai. -
.
-
n
1
n
Yr,min1 ... , Yr,min' xi,min' ... , xi,min
[ 1
]T
(7.10)
bi
n+l
2n
.....
'
(7.11)
JI
'
'
.1
"
1l
"
'
,.
"
,_
H
n "
n "
JO
--1--+--r--i--r--r--i,-0-1-
J4 --- -- -- -- --
---+--+---t----+--+--r--
..
:14
(7.12)
- x'
"
JI
40
;
j == 1, 2, ... , 2n
e-
Furthermore, the bounding boxes of bodies i and k are found to be overlapping if the mapped location of a body k satisfies the following condition
f or
lf+ITTH
-l
1
-e
IJ
b=11=~~1
-~--+---+--t--r--r--;----i--1-. -~
r---:
:.10
,.
.....
o.~
-.... .
-1
.. -- . ...
~o
11 ,__,__ ~- - - --4--H-:J.+--+--+--+--+-
This condition is checked while traversing the binary tree in order to generate the short lists of potential contactors k for each target body i [1], leading
to creation of short lists of all potential contact couples.
For a selected target body f, the binary' tree is traversed until the targets'
search region intersects with the bounding hypercube of a body k ata visited
node. If the body k satisfies the condition (7.12), it becomes a contactar
to the target body i and is added to the targets short list of contactprs.
Otherwise, the whole branch of the body k descendants can be exclu:ded
from frther search [1].
The process of creating short lists of contactors for each target body
from a given one dimensional set of objects is illustrated in Figure 7.36.
It is worth noting that a further reduction in the number of required operations to accomplish the intersection search may be achieved by comparing
ouly those coordinate components which are associated with the binary tree
bisection level at which the traversed node is. inserted. For example, according to Figure 7.36 only the first coordinate components of nodes A, E, F, D,
N and Pare required to be compared with the first coordina.te components
of a and b search region, while the second coordina.te componerits.of nodes ...
n, C, I, G, K, L, M and O will be compared with the second coordinate
components of a and b search region.
16
,___,f----+- - .. . it ~
14 ---- .
..'
, 1 l~ -tl
--+---+---+--_,___+--.--..-
...
... .
6L-J....-1.-J"--1.-Ji---.,j......L......i.--i.--i.-i.--i.-'--1--'---'--'-~_.~xmm
'
l
'
11 1 10 1:1 u 11 11 11 u " 26 28 Jo n 34
tree
level
search
component
Discontinuum Mechanics
220
221
X"""'
'
JI
36
-----------b.
(2)
-f2....i2-,~~ -- - ~~
... -
----
--r--
--1--+--+--t--+-- - - -
-~
The global search method based on the ADT approach uses a bounding box
search to a.void numerous difficulties that me.y u.riHe iu dealing with complex
geometi:ies of real objects. As a result, the resulting short lists of potentlal
contacta have to be re-examined within an accurate local contact search
procedure to determine the intersecting objects. The contact interaction
algorithms may then be utilized to define the surface tractions arising from
contacts between intersecting bodies.
-- ---::ir
u1---l-+-t--~-1--1----t--1-+-41-+-+--t--+tM:H-'-f
31 -+-+-+--l--+--+--+--1---1-+---+--t--+--+--t--1
JO
i--r---
-+--+--+-'-+----t-+-t--l-f-t-IJ-t-+--f--1---+--t
u1--r~-+--1--t--t--+--+-al+l-+----+--t-+--t---ir-t
u1--1-1-+-+-+--t--+--+-+-i-lt-t-+-t--t---t--1
This accurate local contact search, also called local contact resolution
phase, requires a detailed.geometric description of the contacting couples in
order to accurately deterfuine the bodies really intersect.
n ----1--~-+-~-+--+--+----+-+-+--1--<1-+-+--1
/ll--l-IC-Q-1--+-+-l-i~~IH-+-i-lt-+-+-t--t---t--I
00
r=(O,U) ce-+..............-+'""'!l!l!-J-+--+--+-~f-li-t-t--r--r-i
10 - - - ... -- -
'
- - - --- ---+--+--+--+-1--1
---
Rigid discrete elements have simple shapes with closed form surface definitions. Therefore, an accurate local contact search can be performed based .i}~
on ma.thema.tical intersection solution of potential contact couples.
111--1-l-+-+-+--l---l--+-+-~-+-+-+--t---t--1
~-t--+--<f--t-+--t----t
6.__._.__.,_.__.__.__._......_..._......._._..._.__.__._...__l.~x
'
'
10 IJ "
16
11 "
'1
u " "
JI J:l
In contrast, if the surfaces of the bodies are approximated using meshed : \::.
polygons/polyhedrons as a result of the finite element discretiza.tion, the
geometric intersection search procedure can be subsequently rearranged to
use the edges/facets rather than the whole bodies in order to simplify the
accurate local contact search.
..
[1).
The conta.ct detection procedures, rely heavily on indire?t addressing. As a result, it is virtually impossible to vectorize them to any
significant degree.
'
Nevertheless, they offer possibilities for parallel compute11-3 as work on
disjoint subtrees can, in principie, be carried out independently.
7.3
Object Representation
In this chapter the main classes of object representation methods are discussed. It includes circular disks, ellipse shaped disks, and the general
superquadric forms. Additionally, the meshed polygon systems, which frequently encountered in general finlte element c.ontact. analyses are also
among the object representation techniques. They usually provide speclfic
Discontinuum Mechanics
222
223
7. 3 .1
Circular Disks
Circular disks/ spheres a.rflprobably the most frequently used type of element in modelling of granular ftow by the discrete element metbod. They
have been used as rigid body objects interacting ea.ch other in a granular
fiow simulation. Both penalty and continuum mechanics based methods
have been used for contact interaction formulation.
From the object representation point of view, they consist of the simplest
forros for two and three dimensional mbclelling. Their geometric representation includes the coordinates of the centre and the magnitude of radius.
The motion of particles can be readily calculated from the equations of rigid
'
body dynamics.
Figure 7.37 shows the geometrical description of a system of two interacting 2D disks.
Rele.tive velocities (i = 1, 2)
Xi= (xia -
ni =
.Xini ,
Relative displacements
.6.n = n.6.t,
At = i.6.t
AFn = OnAn,
Total forces at time i
+ IJ.F11 ,
F~ = F~- 1
Ft
Ma=L:FaRa,
Assume constant force and moment from ti-t to ti+t
Accolorntion
.. j_~
x, -
'
Velocity
--
= x! + .6.txi+i
'
'
oJ+i = Oi + IJ.tOi+i
a;1+i
At the end of sequence, the time is incremented and the whole procedure
is repeated. A more sophisticated approach is presented here to clarify the
main specifications of a disk ha.sed discrete element technique as described.
by Petrinic [1].
Dlscontlnuum Mechanics
224
225
~.
Figure 7.38: A model for normal contact force between two circular disks.
K1
_
-
l-111
G1
(7.15)
(7.13)
For small overlaps, the nonlinear spring behaves linearly which can be
expressed as
'
The spring force Pn is defined using the elasticity solution for two disks
in contact,
(7.16)
--
K _
n -
dPnl
dun
(7.17)
Un=O
Discontinuum Mechanics
226
227
(7.18)
(7.19)
Sliding Friction
,,
r1
e,
LJ
F,,
ur
K,
(7.22)
Figure 7.40: A model for sliding friction force between two circular disks.
(7.23)
(7.20)
L
where Pt
228
Discontinuum Mechanics
(7.24)
where Un is the size of normal. overlap, tl.t is the length of the time step
during which the contact occurred and v is the disk velocity vector, while
229
Rolling Frlction
Consider the situation where the disk is set to roll on a rough horlzontl
plane (Figure 7.43),
(7.25)
Figure 7 .42 illustrates the relation between the relative tangential displacement and thefrictional spring force.
..._-----------------Ur
n
(7.29)
Figure 7.42: Relation between the relative tangential displacement and the
frictional spring force.
where
(7.30)
(7.26)
(7.27)
(7.31)
: (7.28)
--
(7.32)
230
Discontinuum Mechanics
231
(7.~3)
Fr = sign(v[c) lFn -
IFtll
(7.34)
Condition (7.32) is often sa.tisfled automatically since the critica! damping (7.27) depends on the sliding friction stiffness (7.30), which decreases
when approaching the maximum allowed friction force Fn (Figure 7.43).
This is why the rolling part of the tangential component of the contact force
is chosen to be applied in a form of damping. It ensures good cooperation
of sliding and rolling friction.
l
1
Applying the force at the centroids of the disks also implies adding a
resisting moment in the direction opposite to the direction of rolling as
(7.35)
1
7.3.2
Disk Clusters
(7.36)
where T is a displacement function of disk i
=[1
O -(y-yo) ]
O 1 (x-: xo)
(7.37)
Applying the principie of mnimum potential energy, the following system of simultaneous equations is derived for a syst.em consh;t.ing of n disks,
Figure 7.44 illustrates five simple basic disk clusters. They can be utilized to generate various particle shapes.
KD=F
(7.38)
Ku
K21
K12
K22
Kn1
U?
Kn2
..
K2n
Kin
...
Kn.n
l[ l[ l
D2
D1
F2
F1
Dn
Fn
.
..
.
..
(7.39)
232
Discontinuum Mechanics
!l
1
LJ
7 .3.3
233
With more widespread use of disk and sphere based numerical codea, and
the recent development of sphere based constitutive models for granular
assemblages, it is particularly important to assess the degree to which these
models are applicable for real non-spherical materials.
One common problem when using disk and sphere ha.sed discrete element
modelling of soil is the low aggregate friction angle inherent in these systems,
regardless of the angle of inter pa.rticle sliding friction, which is used.
Particle shape has the 1largest effect on mechanical behaviour, with reported increases I\ peak interna! friction angle up to 10 for systems consisting of angular particles compa.red with round particles.
With the realization that disk based discrete element model has serious
deficiencies when used for modelling real granular materials, it has recently
become popular to use the clHpse as th~ barde parl.ldt1 r>lmpo. The elllpee
shape has the advantage of having a unique and c0ntin11rn1s 011tward normal
and no singularities along its surface.
Solution for ellipse-ellipse contact detection requires solution of fourth
degree algebraic equations, which can be done analytically rnther than with
iterative procedures. For these objects, normal c:ontnct forces actlng eccentrically on a particle can generate applied moments w hich potentlally
inhibit particle rotation. As a result, this shape is well suited ph)i'sically
and numerically to modelling granular soilS, powders and grains. '
Contact Decomposition
.............. ~
.........
Figure 7.47 indicates the nomenClature for two ellipses in contact. Points
A and B, which can be usecA as a measure of the total normal overlap
(penetration) between the two objects, a.re determined from the current
ellipse-ellipse intersection algorithm.
1
1
For particle i, .the vector from the centroid in the direction of the presumed point of contact is:
1.
rci =:te - X1
(7.40)
234
Discontinuum Mechanics
235
dvcn
(7.46)
(7.47).
ellipsej
dvct
noting that
(7.48)
where nis the unit .outwards normal at the contact for ellipse i and t is
the unit vector perpendicular to the normal, defined clockwise positive tO
particle i.
Figure 7.47: Contact between two ellipses.
(7.41)
Ata given instant, the individual terma in (7.46, 7.47) may be separated
lnto the incremental net contact deformation, nct normal contact deformation, or net tangential contact deformation due to particle translation or .
particle rotation.
(7.42)
(7.43)
(7.49)
dv~n =
+ [(eirci cosa) -
(7.50)
(7.44)
dvc
(7.45)
where the terms with vare attributed to individual particle translation, and
O to particle rotation. This relative velocity may be resolved parallel and
(7.51)
(7.52)
Discontinuum Mechanics
236
237
Numerical tests have shown that the particle rotation accounts for twice
as much contact motion for round particles as does particle translations
[118].
Ei =
3D Ellipsoid Elements
Three dimensional ellipsoid elements have not been adopted until very re- .
cently due to their difficult contact detection procedure.
E2
a1x
2
+ biy 2 + c1z2 + 2fiyz + 2g1zx
+2p1x + 2q1y + 2r1z +di= O
+ 2h1xy
+ 2h2xy
~(7.53)
(7.54)
(7.55)
with
""
~ = O -+
(7.56)
= O -+
(7.57)
Pj! =
O -+
(7.58)
simplified to
a1x
(7.60)
238
Discontinuum Mechanics
239
(7.63)
(7.64)
_ __..8....
z __ = O
8h(:i:hu'1
(7.65)
,.i1 >
- - - - - - =0
(7.66)
with
,-1
:
LJ
'(7.67)
Equations (7.59, 7.60) pro.vide four equations for determining four unknowns (x, y, z, ..\). The remaining part of the algorithm is similar to 20
approach.
L.\i
(7.68)
L.\i =
(7.61)
(7.62)
The common normal concept will lead to the following conditions:
8f2(x~~1,z1) + (8h(xi:i.1)
f+ (~M:~:1,z1))2
(7.69)
7.3.4
Superquadric Objects
Discontinuum Mechanics
240
241
n.
LJ
D
n
lJ
where the parameters A and 'Y define the strengU1 and the degree of decay
of the isopotential field :F surrounding the potential source I.
M(x,y,z) =
(7.70)
where (xo, yo, zo) is the origin of the function, (ai, a2, aa) are the dimensions of the superquadric semi-major axes extents, and a and {3 are the
roundnes&-squareness. parame~ers of fnction in. two .perp~9~~war. 4.irections, respectively.
<
:F(x, y, z)
= -ye-AI(x,y,z)
(7.71)
(7.72).
:.ri
-1
..... &:,
242
Discontinuum Mechanics
(7.73)
This check is exactly the one needed when determining if one body is1 in
contact with another in. a contact detection procedure.
7 .3.5
n
lJ
Figure 7. 52 shows a typical porous medium where some set of the cavities
in the bulk material have formed channels through the region.
243'
which form networks that cross the boundarie8 of the bulk material. This
is achieved by taking a solid cast of the pore spaces in a densely packed
assemblage of particles.
The process can be visualized as injecting wax into the pore spaces and
dissolving the solid particles in some manner. The wax which rema.ins is a
cast of the original pore space. Dueto the conncctivity of tho initial particle
assemblage, the pore spaces of the wax cast can be guaranteed to connect
across the boundaries, ensuring a permeablemodel [24).
In order to obtain valid void geometries, the modelling process mllst first
generate a matrix that is in mechanical equilibrium. Only then we can take
a wax cast of a void spacE\ that is geometrically valid and exists as a stable
mechanical system.
The numerical model, therefore, must support multibody contact mechanics. To model more complex media, the superquadric representation tech..
niques can be used to describe the particle geometry and the mecha.nisms
by which they interact.
LJ
While classical methods propase algorithms for mechanical or computational mo<lelling of void spaces and their mechanical behaviour within a
geometrical void-less model, the present approach has been designed to create a geometric modelling of permeable solid. The model, can then be used
as the input model for numerical experiments such as fluid flow simulations.
Geornetric modelling of permeable materials requires the creation of a
geometric representation of the solid matrix enclosing the interstitial spaces
A(x, y, z)
= L :Fi
i=l
(7.74)
244
Discontinuum Mecbanlcs
..-
Chapter 8
o
Figure 7.54: Typical porous models [24}.
P(x,y, z)
= C -.A
(7.75)
Figure 7.54 shows typical porous blocks created by this procedure (24].
Finally, the process of generating a porous model can be summarized
.
into three distinct stages:
l. Object representation. Generating the geometric description of the
Introduction
Creation and propagation of eraclcs are frequently encountered in many engineering and industrial problema. Forming a crack is followed by releas.ing
energy and redistributing the forces which ca.used the initiation of the crack.
If this procedure happens immediately after occurrence of a crack~ lt wlll
lead to inaccurate solutions, and more lmportautly, to rnimlts that strongly
depend on the size of the finite elements used in the analysis (14].
To remedy this, strain softening plasticity procedures have been pro- '
posed. In concept, normal hardening plasticity procedures can be applled
to softening materials, by simply introduclng a Boft.ening mther than a hard
ening equivalent stress - equivalent plastic strain relationship. In practlce,
however, there are many difficulties; a umber of which are still unresolved.
The most important one is the mesh dependency of the finite element analysis, in which by continually r~fining the mesh, we can make the crack
propagate at lower and lower loads (107].
Having recognized the serious limitation of stress-based softening failure
criteria, we could move to methods directly involving fracture mechanics.
An alternative procedure, is to introduce a softening stress-strain relation,;_;,
ship which is related to the fracture energy. In the following, both methods
for considering the load transfer in cra.cked regions will be reviewed.
In this chapter, only some of the main computational issues are addressed. Firstly, the phenomenon of mesh dependency is <liRc.nssed. It has
246
Discontinuum Mechanics
8.2
Mesh Dependency
Bazant, among others, has studied the problem of mesh dependency and
has explained many associated diffi.culties that there exist with stress based
failure criteria for purely brittle and quasibrittle materials such as concrete
f132).
The concept does also apply to other types of tinitt~ nlements, i.e. Gauss
point cracking in higher order elements. lt is worth noting that there is a
class of problems that does not seiverely su:ffer from these difficulties. In some
analyses, integration points through the depth of beam or shell elements may
be used. In these circumstances, even by degrading the properties to zero
and while there would be a stress concentration in a full 3D or 20 ana.lysis,
no such concentration occurs in relation to the adopted strain assumption
that 'plape sections remain plane'.
8.3
Fracture Mechanics
A classical paper published by Inglis in 1913 {134] ma.rks the fil'.st theoretical description of stress concentration at the tip of a crack. He obtained
the elastic solution for stresses at the vertex of an ellipsoidal cavity in an
infinite solid. He then noticed that as the ellipse approaches a line crack,
the stress at the vertex of the ellipse tends to infinity.
Figure 8.1: Finite element modelling of a tensile ple.te.
247
The extensive need for methods which quantify the effects of the presence of
cracks on material performance has led to the evolution and development of
the theory of fracture mechanics. With various types of cracks being always
present in all structures, to sorne degree, a fundamental requirement for a
fracture mechanics theory is some means of asseHsing Uw sta.bility of such
cracks. In this respect, the most significant advance has been the introduCtion of the stress intensity fa~tor. as a single parameter for categorizilig the.
onset of crack propagation {133).
'--'
sen ;_'.(.';: .
248
Discontlnuum Mecha.nics
The next major achievement was the introduction of the stress intensity
factor K by Irwin in 1957 [138) as a single parameterfor categorizing the
onset of crack propagation and relating the intensity of stresses close to the
crack tip to the energy release rate [133].
.
Categorizing the types of cracking, there exist three pure modes of fracture as depicted in Figure 8.2; Mode /, which is a pure opening mode Mode
JI which represents in-plane shearing mode, and Mode 111 for an~iplane
shearing mode.
8.4
Remeshing Tedmiques
249
Model
Modell
Modelll
A displacement controlled solution, based on the longitudinal displacement A is adopted (Figure 8.3b). At the maximum load P, the stress in
the element would each have reached the point B in Figure 8.3a, i.e. Ut
Following this stage, a possible equilibrium path in the load/defiection
space is given by the falling curve 1 in Figure 8.3c. For this path, it is
888umed that each of the elements follows the Hoftening Ht.rcss-strain llne
BC.
On the other hand, the falling line 2 in Figure 8.3c ca.u be obtained if
three of the elements softened dGwn BC in Figure 8.3a, while one of them
unloaded elastically down BA.
In relation to the adopted model, the only stable path would be pa.th 4,
for which only one element softened while the others all unloaded elastically. :;}::1,,~
If the mesh was refined, this. ata.ble path would be a line doubling' back;
directly on the elastic loading line.
'
. ~;~' ~
~,.
'
250
Discontlnuum Mecbanlcs
251
CJ
Fracture lndicator
a,
l'
.
p
' b) simple
bar.
While with a material imperfection, there would strictly bP. only one attainable equilibrium curve, and so resulting in a mesh independent solutio.t:!,
with finite steps, a finite element solution procedure can easily converge on
to a higher unstable state. It is true that these problems are at their most
severa for constant stress states such as depicted in Figure 8.3, nevertheless, equivalent problems can plague finite element solutions even in other
circumstances with a varying stress gradient.
''
One model that provides a simple approach to localization zone simulation is the Rankine softening plasticity model and in this work a bilinear
local soaenlng model is adopted (10] (See Figure 8.4).
The fracture energy relea.se, is de.fined as the integral of the area under
the softening branch of the ~tress-strain curve
e) structural response.
G =
[~ft (fu -
fu
Et)] le
(8.1)
/,
-11
252
Discontlnuum Mechanlcs
253
1
1
for 2D
(8.6}
le=
-fwl-
vi
/or 3D
EP =/lle
2G
(8.7)
The position of the stress point on the softening branch, or the value
of the fracture indicator (Figure 8.4), could be used as a measure, being
compared to a predefined maximum value, to quantify the level of material
dama.ge for different regions.
Figure 8.5: Discrete and smeared crack models.
1
1
GJ =
Undw
8.5
(8.2)
UJ = / Undecr
(8.3)
Damage Mechanics
The concept of damage as a load bearing area reduction was first introduced
by Kachanov in 1958 [142]. Since then, the terminology damage mechanica
has been used in many different _ways with the common theme of reductio[l
in the secant stiffness. In the following, a brief outline of the !'~lativelY; '
simple scalar damage is discussed.
the .
u = (1 - d)Eoe = (1 - d)a
(8.8)
G =le g
(8.5)
which applies with loading phase (e= fmoa:), ur unloading phase (e< frncl#)
where fma:i: is the maximum strain experienced so far. Eo is the initial
elastic stiffness and d is the scalar damage para.meter,
(8.9)
- 1
254
Discontinuum Mecha.nics
255 .
..
Differentiating Equation (8.8),
a
&
a,
= (1 -
d)Eof. - dEoe
(8.14)
with
Q
Et.
. dd.
d =-f=---
de
a-1 e2
(B.15)
a.e,
(8.16)
u
d
= --.
(t - :!)
a-1
e
~ (t -
a-1
Tt)
r
= (1 - d)E0 e -
--~ -Et E
a-le
1
1-a:
oe = - - E oe = E tE
(8.17)
(8.10)
u.
= (1 -
(S.18)
- -
f =E - max{Eold, Et}~ O
(8.12)
(8.19)
with,
E< Et
{
E>
Et
~ O:Et
no damage; f < O
material damages
aged material; d = O
(8.13)
where the constant A is chosen based on the fracture energy. Clearly, Equations (8.10) and (8.19) can not be directly used with algorithm given in Box
8.2, because the complete material would then degrade in this tensile soft..;
ening manner. A simple solution can be to define two different damage'.
parameters for the tensile and compressive behaviour.
Discontlnuum Mechanics
256
= (1 -
7"=
~E
f =O
= 'T ,
d = a~l (1
(!) + (!I)
-1:")
G1
=d
dold
The simplest mbced mode of fracture is a linear combination of the separate modes,
IF/~OTH~
'Told
8.5.1
dozd)Eoe
Et = (1 - d)Eo
e
= (1 -
257
(8.20)
= 1
G 11
oc
d)Eor:.
Et= (1-d)Eo -
--L
(i;.
-E0
2 Eo = a- 1 ~
T V"2
1-a
1
as [148]
:!,
ENDIF
= ( 1 - dozd)DoE = (1 - dotd)iT
O'
Dt = (1 - d)Do
. T
= (~e 7'DoE) t
= r - max( Told, Tt)
IF
j~O
Told
U'
G 11
= (I - D]
~oE =
T,
= 1
(8.21)
_!!_p]
Eo
l+K
[1 -
THEN
(8.22)
where it and denote stress and strain (or traction and relative displacement) with opening mode 1 and shearing mode 1I, re:;pectively,
d = d~r,rt)
dold
G1
T = ( !o-T DQ" 1) !
f
(!)' + (!)'
= d
(8.23)
= (1 - d)DoE
= (1 .
d)Do _ l
'T
dd(-r,Tt) T
d-r
E=
ENDIF
E
-E
(8.24)
258
Discontlnuum Mechanics
Eo = Diag
[~:]
in equation
(8.27)
where ft, ev.. and t are illustrated in Figure 8.4. The scalar
(8.22) is defined as
(8.26)
K
It has been analytically shown that the linear mixed mode formu~ation
is satisfied if
i
lJ
tJ
tJ[
(8.28)
f.11 = pl
8.6
259
the mecha.nical response of single crystals and metals. Hill (1950)r'[l49] generalized von Mises' formulation to include anisotropy. A variation of maximum stress theory for unidirectional composites wa.s proposed by Stowell
and Liu [150] in which the failure stress of the fibres was taken as the
limiting .strength of the lamina in the fibre direction, while the limiting
transverse and shear stresses are those of Lhe rr.atrix. In 1965, Hill's theory
was adopted by Azzi and Tsai [151) as a strength criterion for composites.
Hoffman {152], in 1967, altered Hill's criterion to provide for unequal tensile and compressive strengths by introducing linear terrns in the functiona.l
form.
(8.25)
F=Diag[~]
fu - t
for th1s type of behav1our m wh1ch a layered configuration ca.uses difrerent behaviours in and through the layers directions. Another source of
anisotropy is the creation of cracks in which different behaviours are expected along and normal to the crack direction.
Severa! of the anisotropic strength or failure theories are extensions of
isotropic yield criteria. Much of the early workon the strength of anisotropic
materials was conducted independently in the areas of wood technology.a.nd
8.6.1
Tsai-Wu Model
where F and Fi3 are the strength parameters. The linear term in O'' talces .
into account interna! stresses which can describe the difference between
positive and negative stress induced failures. The quadratic terms CT'CTJ
260
Discontlnuum Mechanics
define an cllipsoid in the stress space. The main features of this criterion
are as follows:
261
are
deflned
where the projection matrix P, and the project.iou vector p
based on the nine material yield strengths ami a. normalized yield strength
; (see Schellekens et al. (159]),
(8.31)
2 (0:31
0:12)
-20:12
P=
-2as1
-20:12
-20:31
-20:23
2 (0:23
0:12)
-2a23
2 (0:31
o
o
o
o
o
+ 0:23)
o
o
o
60:44
o
o
o
o
o
6a55
where
au
= -2
(1
(uc _
11T)
O'UCO'U T
The criterion predicts the imminence of failure but provides no information 'l;l.bout the failure mode.
The study by Gregory [158] has confirmed the discrepancies found by
other researchers regarding this criterion. The inability of this theory and
other similar criteria to predict the correct direction of crack growth is not
d ue to an inaccurate finite element analysis, but due to limitations in the
theories themselves.
o
o
(8.34)
(8.35)
0:33
-2
(1
(liase - iiaar)
-
(8.36)
0'33C0'33T
(8.37)
8.6.2
j2
0:55
= 32
(8.38)
31
Htructed from three tensile yield strengths ~"T' three compresslve yield
strengths
as:
aiiC,
1= j.
It may be defined
(8.39)
(8.30)
(8.40)
262
Discontinuum Mechanics
263
{8.41)
1
(8.42)
Only one shear strength is required for implementation in thin shell
elements, and the size of the projection matrix is reduced to 4 (160].
The Tsai..:Hill yield criterion is .a special ca.se of this criterion, in which
the projection vector Poc is reduced to a null vector. The projection matrix
P a is determined by (8.32) under the restriction that O'iiT = O'c.
For the jth iteration of the ith load step, the additivity postulate is
adopted,
Then, the final stress at the end of the iteration is solved from this
relatiou
(8.43)
(8.47)
(8.48)
which can be re-written in terma of the D matrix
(8.44)
; (8.49)
where A>.J is the plastic multipUer. The derivative of ~ (8 ..30) with respect
to O'j is the flow vector a,
'
8~
a= = P at.Tj + P
8u;
(8.45)
(8.50)
Re-writing equation (8.44) with the aid of Cui-1 - Ef~ 1 =O, will eventually result in
(8.46)
_,.-:.~
Hence, the derivative of <I> with respect to AAJ is now needed. For hardening or softening materials, ~is also a function of the hardening/softeni11..g
parameter, K. Therefore,
Discontinuum Mechanlcs
264
(8.51)
The first part of the first term is the ftow vector a, and equation (8.49)
is used for the second part
,,,
ffii
(8.59)
(8.60)
1
(8.52)
(8.61)
Box 8.3 summarizes the stress return algorithm for the Hoffman yield
surface with isotropic hardening/softening.
(8.53)
Consistent Ta11ger1t Matrix
(8.54)
8<JJ
-{}jj = -2cr
(8.55)
(8.62)
(8.56)
With the relations for incremental elaatic nnd plast.ic strnins, the streeestrain relationship can be written as
(8.57)
Therefore, from (8.51), one gets
(8.63)
Taking the time derivative of(S.63) results in
(8.58)
{}2<JJ
e= Cu3 +A>.-D'j
3
'
80'J
. 84!
+ ..\3
(8.64).
OCT j
(8.65) .
Discontlnuum Mechanics
266
267
The last term in {8.65) v~ishes for infinitesimal increments, which results in the classical elastoplastic tangent stiffness matrix. For finite loading
increments, however, the contribution of this term becomes significant. Rearranging {8.65) with the aid of
=a and ~ = P co results in
IJ:
= u i-1
+ D.6.e
.. -
IF
~ tr
u!1 - 2
sO
(8.30)
(8.67)
{Elastic update)
Gj :::: <Ttr
(8.68)
(Elastoplastic updte)
ELSE
n
LJ
{8.4~}
'.
. {8.69)
<Pe
= u~J -
IF 1~c
(8.66)
{c) 2
>. may
{8.30)
/ {c) 1~ TOLER
2
(8.70)
<Tj =<Te
ELSE
A= (aTa) i, e~}=
8CT
lM.>..
E {i,~;>.A
FIR$T = aT
\k+l -
lfx
SECOND = 2fi,H
u.Aj
(8.71)
(8.53)
\/e.
~/\i
Next iteration
- FIRST-SEdNb
Dct
=Q
aaTQ ]
[1- B.+aTQa
(8.72)
{850}
.
8. 7
ENDIF
ENDIF
near
Discontlnuum Mechahics
268
269
section, the conditions for localized failure are examined on the grounds of
bifurca.tion analysis, following the methodology reported by Pearce [164).
Beginning with the classical elastoplasticity, the rate of stress tensor can
be described by the rate of conjugate strain tensor,
(a)
(b)
(8.73)
w here D
ftow,
el b i<:>.. a . Del
D = net - D . . 'OI
H +ti: Del: b
(e)
{d)
(8.74)
1
1
111'
where a = ~ is the gradient pf the loading surface, bis the plastic fiow
direction and H is the hardening/softening modulus. In general, D will
not be symmetric, Dijkl :/:- Dklij, although m.inor symmetry is assumed:
Di;kl
D;ikl
!.........- ONIMof
1
llltlp-bock
1
1
bdavlolll'
8
Di;lk
A = _H_+_a_:_D_e..,...,-:-b
(8.75)
The numerator of (8.75) can be regarded as a. scalar product of an elastically predicted trial stress rate and the normal vector to the loading surface. Therefore, the numerator is always positive, and according to (8.75)
the dominator must also be positive,
H
+a
: Del : b > O
(8.76)
E"= El( .
E+H
(8.78)
!\
or
8.7.1
H >-a: Del: b
l_j
Loss of Uniqueness
(8.77)
= -a: Dei: b
Figure 8.8 illustrates the same concept for an idealized uniaxial problem,
where the elastoplastic modulus is defined by:
The losa of uniqueness of the stress rate, also called the onset of material
branching or simply the limit point is associated with a null stress rate
tensor,
il'=D:=O
(8.79)
Discontlnuum Mecbanlcs
'270
ecr, we have
D:
ve
ecr =O
(8.80)
The stationary value of ;, , can be related to the elastic strain rate using
'
(8.81)
(8.82)
271
Equation (8.86} is true only for zero critical hardening/softening modulus corresponding to perfect plasticity (H = O). In other words, when
H = O, the tangent modulus D ~~()!11-eB singular, and for any strai.ning
in the critica! eigendirection equivalent to b the stress rate remams a. ni:ill
tensor. The same concept is valid in associated plasticity, except for the critica! eigendirection which becomes proportional to the normal to the loading
surface a.
It should be noted that H =O represents a necessary but not sufficient
condition for the loss of uniqueness. Under certain conditions such as kinematic constraints it is possible that the critica! eigendirection will not be
activated.
To verify the stability condition, severa! postulates ha.ve been propase(!.
According to the Drucker's stability condition, the external work in
of a cycle of application and removal of the externa! loading, must not be
negative. In Hill's postulate, an equivalent stability condition is defined as:
case
(8.83)
(8.88).
De is positive definite, thus
(8.84)
(8.85)
Therefore, the direction associated to singular D or the critica! eigendirection is proportional to the plastic fiow direction b. From Equations (8.85,
8.80, R74),
De - De : b a : De) : b = O
H+a:De:b
O< H <Her
(8.89)
By substituting Equation (8.73) in (8.89), tho onsd of material lnstability can be defined:
(8.86)
E:D:e=O
(8.90)
--
(8.87)
D"m:
=O
(8.91)
(8.92)
Dlscontlnuum Meche.nics
272
Equation (8.91) states that the eigenvalues of the tangent operator are
bound by those ofits symmetric counterpart. Thus, the loss of material
stability occurs when the symmetrized tangent operator is no longer positive
definite, characterized by at least one zero or negative eigenvalue [164]. For
a symmetric tangent operator, the instability indicator coincides the loss of
positive definiteness of D.
8. 7.2
Strong Discontinuity
In a strong discontinuity, a jump in the displacement rate is also admitted.
Therefore, preservation of the displacement rate jump, (), along the plane
is required. The directional deriva.tive Qf ((u) in a tangential direction ze
to the plane must vanish:
D())t = (Vu) mt
Discontinuous Bifurcation
(Vu) = mn
(e)= 4(mn+nm)
while the
= mn
__
...-
(8.97)
1 (
=2
2m1n1
m2n1 + m1n2
m3n1 + m1 n3
m1n2+ m2n1
2m.2n2
m3n2 +m2n3
(8.98)
(8.93)
To further investigate the problem, a plane stress case, w hcre the disconti
nuity plane normal is aligned with the x 2 axis is considered. Therefore, the
discontinuity normal vector n becomes
(8.96)
or in component form,
()
(Vu)
(8.95)
Weak Discontinuity
According to the Max:well's compatibility condition, only the normal derivative of u across the plane of discontinuity can be discontinuous
=o
and the kinematic restriction resulted from the Max:well's compatibility condition
273 . t,
(8.94)
n = (0, l)T
(8.99)
(e)=! ( o
2
m1
m1 )
2m2
(8.100}
.,
274
Discontinuum Mechanics
275
Localization Condition
For a strain jump problem, the corresponding discontinuous stress rate eld
can be expressed as,
n
(&)
mlln
= D:
(e)
(8.104)
spUttnode
e+-s-=F-o
a-s-=F-o
i/-u=o
;,+.,;,.,.o
(&n) b n -(O-)= n D : () =O
mln
{8.105)
sh1armo
(8.106)
Figure 8. 9: Weak and strong discontinuities for splitting and shear failure
modes [164].
or,
(0-n) =(nDn)m=
Two limiting cases for the vector m are parallel and. perpendicular to
the vector nas illustrated in Figure 8.9,
parahel to n
perpendicular to
m = (0, l)T
= (l,O)T
(8.101}
m.Ln
l_J
() = (
()
= !2 (
oO m2
o )
O
m1
m1 )
det Q
= det (n
D n) = O
(8.108)
mlln
(~.107)
(8.102)
(8.103)
where mlln represents the splitting localization mode, and m .L n represents the shearing localization mode (see Figure 8.9). A variety of mixed Iocalization mode can be potentially initiated in between these two extremes.
dened
276
Discontinuum Mechanics
277
F"'
e
Figure 8.1 O: Localiza.Uon failure indicator for different discontinuity orientations.
q.
Fzoc =
detQ
det Qe
(8.109)
where qe .is the acou'stic tensor based on. an elastic tangent operator De,
and :fioc is a localization failure indicator. For :Fioc = 1 the response is
fully elastic, while ':Fioc =O represents the 1 imminent localization. In practice, for different discontinuity directions, the localization failure indicator ..
is evaluated. Figure 8.10 illustrates a typical curve obtained fr 'different
orientation angles. The maximum failure indicator will identify the discontinuity direction.
8.8
Remeshing Algorithm
Material fracture may result in the creation of new discreta bodies which
are in contact by friction interaction with neighbouring bodies. Therefore,
remeshing techniqu~s have to be developed to maintain compatibiiity conditions in newly fractured regions. Here, a simple, though efficient approach
is introduced which can be used for both 2D and 3D applications.
The failure indicator and the crack direction for ea.ch individual element are evaluated from the material model routines. A weighted averaging
scheme is then used to evaluate both the failure indicator and the average
Figure 8.11: Weighted averaging of the failure indicator and the crack direction for a. failed point.
crack direction of each node. Figure 8.11 illustrates this scheme for t.) t-r1 dimensional problem. The same a.pproach may be used for three dimensional
problems with an appropriate weighting. factor for the third dimension.
The next step is to geometrically represent the crack and perform the
necessary split, separation and thet remeshing processes. Figure 8.12 representa the two dimensional remeshing algorithm which comprises four steps:
splitting the element, separating the failed nodes, creating new remeshing
nades, and dividing uncracked elements to enforce compatibility a.t new
nades.
Adopting this local remeshing algorithm will provide a relatively finer
mesh in the fractured region and prevents the distortion of the elements in
this region, improving the finite element approximation of the analysis.
Alternative remeshing techniques include adaptive remeshing algorithme
(as a.n example see Figure 2.6 which illustrates an adaptive shear band sim
ulation). They are a class of e:ffective techniques for improving the finite
element approximation which can be applied to almost any proolem. Unfortunately, they usually require a. large a.mount of computational time and
278
n
LJ
Discontinuum Mechanics
crackdirection~ ..,.
....::::-
node separation
Appendix A
~7
crack
~'"
. :;7
.~ing
--
newnode
Figure 8.12: Remeshing technique for ~odelling a crack ata failed point.
A.1
memory every time a remeshing update is undertaken. Due to the continuous changing geometry and behaviour of discontinuum problems, severa!
remeshing updates are required, causing massive computational expenses.
Therefore, using an adaptive approach should be carefully studied and then
Le adopted.
Introduction
First, necessary blocks of data are explained and a sample data file is
provided. Several samples have been inch~ded in accompanying CD. For
a more comprehensive data file description manual or accessing the source
files and full unlimited version of the program, please contact the author.
A.2
In this section a brief mamutl for preparlng a U'l'< lnm dnt.a filo iH provlded.
The preprocessor creates this da.ta file based on the information giwn by
the user, and it is not expected to prepare this file by hand. Bowever, in
certain circumstances the user may wish to change parts of the data file for
a new run.
The model is defined as a set of different groups. Each group of elements
may have different element types and material properties. Each block of
data begins by a { and ends by a }. The position of each block is trivial
279
Discontlnuum Mechanics
within the data file. The srune holds for the positlon of each subblock withln
any block. In the following data file description, capital words represent the
beginning of a new block/subblock of data while lower.case words represent
numbers defining the dimension of the corresponding array. Words in capital
letters must remain unchanged. Any irrelevant block may be discarded from
the data file.
280
'rhis is an interna! control 'Iine for the program and must be kept
exactly the same.
2. GROUPDATA
ngrcode =O
NUMBERS : block specifies the numl?ers associated to each group
ngroups : total number of groups
ACTIVITY: defines the active/inactive state of ea.ch group
4= 3 nodedtlane stress
5= 4 noded plane stress
'l
6. NODES
nncode =O
NODES: node numbers
npoints : total number of points
COORDINATES: coordinates of each node
ndim = 2
7. MATERIAL : one block for each material data set
'
Ea;, Ey, Ez, V~"' Vyz, Vzll:I Gxy' Gyz1 Gzx1 ax, ay, z, p,u.,P
8= 3 noded axisymmetric
9= 4 noded axisymmetric
MATERIAL : material number
EOPTIONS: group options
nopts = 50
SURFACE : surface contact specification numbers
nsurfp = 2
281
nsp = 1
=2
282
Discontinuum Mechanics
13. GLOADS : global loads. one set for each load case
ndim
283
=2
=O
=2
11
[I
1= impenetrability contact
11
Hd.8
ndim = 2
VPOINTERS : pointers of prescribed velocity sets to
corresponding nodes
nvnodes : number of points subjected to VELOCITY
7= delaminatio'n
PROPERTIES :
ncsp = 4
1
nlcode =O
LOADCASES : load case numbers
nlcases : number of load cases
nsetl = 2
ACCELERATION : prescribed acceleration
na.sets : number of prescribed acceleration sets
ndim
=2
=2
12. LOAD CURVES : load curve definition. one set for ea.ch load case
ndim = 2
rl
11
11
f:
284
=2
=2
A.3
ACTIVITY { ngroups .
nsetl = 2
1 1
}
FRACTURING { ngroups
1 1
}
DEFORMABILITY { ngroups
1 1
}
ELEMEN11'YPES { igroup
TYPE {
4
MATERIAL {
1
EOPTIONS { nopts
F F F F F F
F F F F F F
F F F F F F
}
SURFACE { nsurfp
1 1
L.J
Sample
Data
File
.
1
'""
285
Discontlnuum Mechanlcs
F
F
F
SOLU'fION : always 1
ELEMENTTYPES { igroup
F F F F F F
F F F F F F
F F ""F -
F
F
F
F
F
F
F
F
F
F
F
F
Discontinuum Mecbanics
286
287
L_;
TYPE {
}
}
THICKNESS { igroup
THICKNESS { ngelements nnode
MATERIAL {
10*1
2
}
ANGLEFLAG {
EOPTIONS { nopts
F
F
F
F
F
F
F
F
F
F
F F
F
F
}
F F F F
F F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F F F
F
F
F
NODES { nncode
NODES { npoints
591
592
601
602
593
603
'
10
594
604
695
605
696
606
597
598
599
600
SURFACE { nsurfp
2
}
ELEMENTS { igroup
CONNECTIVITY { nnodes ngelements
601
108
605
-50
-48.7772
3.79429
600
107
NUMBERS { ngelements
1
2
3
4
MATERIAL { imaterial
991 . 992
993
994
995
~96
997
998
10
999 1000
ELASTICFLAGS { nef
o1oo
}
ELASTICPROPERTIES { nep
1000000 O O O O O O O O O OO 1000
ELEMENTS {igroup
CONNECTIVITY { nnodes
141
ngelements
NSTATES {
12
315
222
521
147
SURFACEFLAG {
212
}
}
NUMBERS { ngelements
SURFACEPROPERTIES { nsp
1001 1002 1003 1004 1005 1006 1007 1008 1009 1010
0.2
}
}
'
THICKNESS { igroup
THICKNESS { ngelements nnode
3000*1
MATERIAL { imaterial
ELASTICFLAGS { nef
o1 oo
ANGLEFLAG {
ELASTICPROPERTIES { nep
oo
288
10e+5 O O O O O O O O O O O 1000
Discontlnuum Mechanics
oo
PLASTICFLAGS { nef
o8
PROPERTIES { ncsp
-PLASTICPROPERTIES { nep
1000 50
FRICTIONTYPE {
1e+006
NSTATES {
O -10e-10
12
FRICTIONPROPERTIES {.ncsffp
SURFACEFLAG {
0.2
}
1
}
SURFACEPROPERTIES { nsp
0.2
SURFACE { isurface
TYPE {
1
}
}
DISCRETEPROPERTIES {
DAMPING {
PROPERTIES { ncsp
1e+006
O -10e~10
FIELD {
LOADINGS { nlcode
LOADCASES { nlcases
0.2
ZONE {
ACTIVITY{ nlcases
1
}
MINSIZE {
0.2
LOADCURVES { iload
TIME { ntpoints
CONTACT {
o.o
1
}
}
}
LOADFACTOR { ntpoints
DEBOUNDARIES { ndecode
DEBOUNDARY { nnodseg ndeseg
45
1
1.0
44
24
GLOADS { igload
POINTLOAD { ndim npsets
-10.1
DEPOINTERS { ndeseg
2
}
}
18.2
12
13
289
r
L_.;
290
;___)
""
\1
-10.1
18.2
12
1
13
1
14
1
124
2
135
2
VELOCITY { ndim
-0.02 o
0.02 o
2
1
209 210
1
1
1
1
210
2
1
410
408
403
404
10
1
104
105
106
107
207
8
2
208
}
HISTORY { ihistory
NODES { nhnodes
156
157 45
413 412
NVALUES { nvalues
"x_DISP"
23
1
GLOBALVALUES { ngvalues
"E_KIN" 11 E_EL 11
CONSTRAINTS { nccode
DISPLACEMENT { ndim ndsets
STEPS
o
1
}
{
1
1
CONTROL {
TITLE {
11
10
Sampledata 11
196
197
198
199
200
201
202
203
204
206
1
2
3
1
1
195
205
L...
}
INITIALTIMESTEP {
0.1
}
TIMESTEPCHANGE {
100
}
MINTIMESTEP {
.'-,
20~
414
1
1
207
1 o
o1
-0.2
107
106
104 106
j
l
l
nvsets
291
Discontinuwn Mechanlcs
- --
292
Dlscontinuum Mechanics
MAXTIMESTEP
____;
1
}
References
CRITICALFACTOR {
0.9
}
SOLUTION {
1
} NUMBEROFTIMESTEPS
[1] Petrinic, N., Aspects of Discrete Element ModeUing Involving Facetto-Facet Contact Detection and Jnteraction. Ph.D. thesis, Department
of Civil Engineering, University of Wales Swansea, UK, 1996.
20
}
STOPTIME {
graphite/~poxy
100
}
FRESULTS {
[3] Roh, Y. & Chang, F.K., Effect ofimpact damage on lambwave prop&gation in laminated composites. Dynamic Response and Behamourof
Composites, eds. C. Sulk B.,,.Sankar & Y. Rajapakse, ASME AD-Vol.
46, pp. 127-138, 1995.
50
}
OUTPUTFREQUENCY {
50
""
MESSAGESTEPS
1
TOUTPUT {
o
}
(4] Abrate, S., Impact resista.nce of composite materials - a review. Applied Mechanics, 44, pp. 155-190, 1991.
293
..
- . . :~-~it,1y~--- '
W4
lteferences
Discontinuum Mechanics
{22] Chandler, N., Dixon, D., Gray, M., Hara, K., Courant, A. & Tillerson,
(11] Bicanic, N., Munjiza, A., Owen, D. & Petrinic, N., From continua to
?isc?~tinua. - a ~ombined finite element / discrete element modelling
m c1v1l engmeermg. Developments in Computational Techniques for
Structural Engineering, ed. B. Topping, Civil-Comp Press pp. 1-13
1995.
'
'
J., The tunnel sealing. experiment: An in situ demonstration of technologies for vault sealing. 19th Annual Conference of the Canadian Nuclear Society, pp. 4o-44, 1998. Toronto, Canada.
delamination and fracture analysis of composites. Thin Walled Structures, 40, pp. 595-609, 2002.
(
- Algorithms, Jmplementation and Applications. Ph.D. thesis, Department of Civil and Environmental Engineering, MIT., 1996.
[14] Mi,
y,
[26] MacLaughlin, M., Sitar, N., Doolin, D. & Abbot, T., Investigation of
struc~ures.
ed.
M.
i !
[15] Zhang, X., Hitchings, D. & Davies, G., Modelling composite failure
[27] Lak, M., Adaptive Shear Band.Slope Stability Analysis. Master's the. sis, Department of Civil Engineeri~g, University of Tehran, 2002.
and delamination during low velocity impact. Computational Mechanics in UK - 5th ACME Conference, ed. M. Crisfield, pp. 44-47, 1997.
London, UK.
[28] Stead, D., Eberhardt, E., Coggan, J. & Benko, B., Advanced numerical techniques in rock slope stability analysis - applications and
limitations. Landslides - causes, impacts and countermeasures, PP
615-624, 2001. Davos, Switzerland.
[17] Kirn, J., Mixed mode fracture 'analysis via coupling the finite ele-
P;., ~
proach to the adaptive finite element analys1s m associated and nonassociated plasticity considering localization phenom:ena. Oomputer
Methods in Applied Mechanics and Engineering, 176, pp. 187-202,
1999.
[30] Masuya, H. & Kajikawa, Y., Numerical analysis of the colllsion b&.
[19] Malagaris, G., Finite/discrete elements in blasting and mining applications. Master's thesis, Un~versity of Wales Swansea, 1994.
&
[32]
[21.J Martin, C., Strength o/ Massive Lac du Bonnet granite around un-
.:':
'.1 ~--~
Collins, D., Examining unstable shear fracture using dynamic numerical modela ~d IJlicroscopic
monitoring. Technical report, University of Liverpool,':2002.
-!
A
296
Discontinuum Mechanics'
[33) Petrinic, N., Owen, D., Munjiza, A. & Bicanic, N., Rolli~g resistance
of disks in contact. Extended Abstracts for the 3rd ACME UK Conference, pp. 105-110, 1995. Oxford, UK.
[34] Vaz-Jr., M., Comm.tational approaches to simulation of metal cutting
process. Ph.D. thesis, Department of Civil Engineering, University of
Wales Swansea, 1998.
2001.
'
297
References
[46] Rockfield Software Ltd., ELFEN helps save our historie structures.
Newsletter 4, 200L
[47] Mak, K.W., Some aspects o/ combined discrete/finite element simulation of shot peening problems. Master's thesis, University of Wales
Swansea, 1998.
[48] Connel, M., Virtual reality simulation for finite/discrete element problema. Ph.D. thesis, University of Wales Swansea, 1998.
1
Computational Techniques for Contact, Impact, Penetration and Perforation of Solids, eds: L. Schwer, N. Salamon & W. Liu,
ASME AMD-Vol. 103, 1989.
e~plosives.
1997.
element algorithm for delamination a.nalysis of composites. Finite Elements in Analyss and Design, 28, pp. 321-3~6, 1.998.
[38] Forouzan-sepehr, S., Analysis of lnitiation and Propagation o/ Interlarninar Cracks in Composite Shells Subjected to lmpact Loading. Master 's thesis, Department of Civil Engineering, University of Tehran,
[51} Zienkiewicz, O. & Taylor, R., The Finite Element Method. McGraw
2001.
[39] Mohammadi, S., Combined Finite/Discrete Element Analysis of Impact Loading of Composite Shells. Ph.D. thesis, Department of Civil
Engineering, University of1Wales Swatlsea, 1998.
[40) Worswick, M., Strazincky, P. & Majeed, O., Dynamic fracture of fiber
reinforced composite coupons. Dynamic Response and Behaviour of
Composites, eds. C. Sun, B. Sankar & Y. Rajapakse, ASME AD-Vol.
'1G, pp. W-'11, 1n!J5.
[41] Price, D., Discrete element modelling of masonry walls. Master's thesis, University of Wales
~ansea,
1997.
[53] Belytschko, T. & Neal, M., Contact-lmpact by the. pinball algorithm with penalty, projectlon, and lagrangian methods. Oomputational Techniques /or Oontact, Impact, Penetration and Per/oration
& Benson,
[56] Peric, D. & Owen, D., Computational model for 3-D contact problems
[44] Marefat, M., Ghahremani-Gargari, E. & Naderi, R., Full scale test of
of
1
1
~.:
;
!'.
;.!':'.
!
1
;
~
11
'
298
Discontinuum Mechanics
[58] Munjiza, A., Bicanic, N., Owen, D. & Ren, Z., The central difference
time integration scheme in contact-impact problems. 4th Jntemaonal
Conference on Nonlinear Engineering Computations - NEC-91, eds.
N. Bicanic, P. Marovic, D. Owen, V. Jovic& A. Mihanovic, pp. 569575, 1991.
[60] Yu, J., Report on discreta element 3D contact analysis algorithm implementation. Interna! report, Rockfield Software Ltd., Swansea,
UK, 1996.
[68] Nagtegaal, J., On the implementation of inelastic constitutive equations with special reference to large deformation problema. Computer
Methods in Applied Mechanics and Engineering, 33, pp. 469-484,
1982.
(69] Peric, D., Owen, D. & Honnor, M., A model for finite strain elastoplasticity ha.sed on logarithmic strains: Computational issues. Computer Methods in Appled Mechanics and Engineering, 94, pp. 35-61,
1992.
[70] Lee, E., ElaHtic-plostic deforma.tion at finite strains. Joumal o/ Applied Mechanics - ASME, 36, pp. 1-6, 1969.
299
References
[71] Nemat-Nasser, S., On finite defor~ation elasto-plasticity. International Joumal of Solids and Structures, 18(10), pp. 857-872, 1982.
[72} Hashimoto, K., Neto, E.S., Peric, D. & Owen, D., A study on dynamic
frictional behaviour of coated steel sheets: experiments, forrnulation
and finite element simulations. Technical report, U niversity College. of
Swansea, 1993.
(73] Mohammadi, S., Owen, D. & Peric, D., Delamination analysis of composites by discrete element method. Computational Plasticity, COMPLAS V, eds. D. Owen, E. Onate & E. Hinton, pp. 1206-1213, 1997.
Barcelona, Spain.
'l'
'
[74] Mohammadi, S", Owen, D. & Peric, D., Discontinuum approach for
damage analysis of composites. Computational M echanics in UK - 5th
ACME Con/erence, ed. M. Crisfield, pp. 40-44, 1997. London, UK.
[75] Hughes, T., The Finite Element Method. Prentice-Hall International,
Inc., 1987.
[76] Kole, J., Figge, M. & Raedt, H.D., Unconditionally atable algorithms
to solve the time dependent maxwell equation. Phys Rev E, 64, pp.
66-75, 1999.
[77] Cook, R., Malkus, D. & Plesha, M., Goncepts and Applications of
Finite Element Analysis. John Wiley & Sons, 3rd edition, 1Yts9.
(78] Hinton, E., Rock, T. & Zienkiewicz, O.,. A note on mass lumping and
related processes in the finite element method. Earthrwike Engineer
ing and Structural Dynamics, 4, pp. 245-27U, 1!>76.
[79] Bicanic, N. & Maca, J., An improved mass lumping technique by
rnatching element spectral characteristics. Technical report, University College of Swansea, 1991.,
[80) Wilkins, M., Calculation of elastic-plastic flow. Methods in computational physics, ed. B. Alder, Academic Press, volume Vol. 3, pp.
211-263, 1964.
[81] Maenchen, G. & Sack, S., The TENSOR code. Methods in computational physics, ed. B. Alder, Academic Press, volume Vol. 3, pp.
211-263, 1964.
[82] Bertholf, L. & Benzley, S., TOODY II, a cornputer program for two
dimensional wave propagation. Sc-rr-68-41, Sandia N ational Laboratories, USA, 1968.
'
fu(~H.
ACME, pp.
Discontinuum Mechanics
300
References
301
[84) Schonauer, M., Unified numerical analysia of cold and hot metal forrning processes. Ph.D. thesis, University of Wales Swansea., 1993.
[85} Oden, J. & Martins, J., Modela a.nd computa.tional methods for dy-
[99) Peric, D. & Owen, D., A model for 3-D contact problema with frlction. Proceedings o/ the European Conference on New Advances in
Computational Structural Mechanics, 1991. Giens, France.
._.,
(100) Simo, J. & Taylor, R., Conaistent tangent operators for rate indepen-.
dent elastoplasticity. Computer Methods in Applied Mechanics antl
Engineeri,ng, 48, pp. 101-118, 1985.
[101) Simo, J. & Hughes, T., General return mapping algorithms for rate
independent pla.stici~y. Constitutive Laws for Engineering Materials:
Theory and Applications, ed. C.D. et al., pp. 221-231, 1987.
(102) Mitchell, G. & Owen, D., Numerical solutions for elastic-plastic pro~
lems. Engineering Compu.tations, 5, pp. 277-284, 1988.
(90] Back, N., Burdekin, M. & Cowley, A., Review of the research on fixed
and sliding joints. Proceedings o/ the 19th Intemational MTDR Conference, eds. s. Tobias & F. Koenigsberger, pp. 87-97,: 1973. London,
UK.
[91] Connoly, R. & Thornley, R., Deterclining the normal stiffness of joint
faces. ACME, pp. 67-6, 1967.
[104] Suh, N. & Sin, H.C., GenesiS of friction. Wear, 69, pp. 91-114, 1981.
[105] Ortiz, M. & Simo, J., An a.nalysis of a new class of integration algorithms for elastoplastic conatitutive relations. Intemational Joumal
jor Numerical Methods in Engineering, 23, pp. 353--366, 1986.
(106] Crisfield, M., Non-linear Finite Element Analysis of Solids and Struc-.
E.,
(93} Kragelsky, I., Dobychin, M. & Kombalov, V., Priction and Wear Calculation Methods. Pergamon Press, Oxford, 1982.
[94] Avitzur, B. & Zhu, Y., A friction model base<i on the upper bound
approach to the ridge and sublayer deformations - update. Proceedings
oj the 19th North American Metalworking Res. Conference, 1-IESL
Publication Edition, pp. 103-109, 1985. Berkeley, USA.
[95] Kobayashi, S., Oh, S. & Altan, T., Metal Forming and the Finite
Element Method. Oxford University Press, NewYork, 1989.
!
nonline~
[96] Wriggers, P. & Simo, J., A note on tangent stiffnesa for fully
contact problems. Communications in Applied Numerical Methods, 1,
pp. 199-203, 1985.
[97] Parish, H., A consistent ta.ngent stiffness ma.trix for three dimensional
nonlinear contact analysis. Intemational Joumal for Numerical Methods in Engineering, 28, pp. 1803-1812, 1989.
[107) Crisfield, M., Non-linear Finite Element Analysis of Solids and Structures. Volume 2: Advanced Topics. John Wiley & Sons Ltd., 1997.
[108] Goodman, R., Taylor, R. & Brekke, T., A model for mechanics of
jointed rock. Joumal of Solid Mechanics and Foundation, ASCE,
SM3, p. 94, 1968.
(109] Cundall, P., A computer model for slmulat.ing progrn:;::ve, large scale
movements in blocky rock system~ Proceedings of Intemational Syrn,-:
positt.m on Rock Structures, 1971. Na.ncy, France.
[110] Ghaboussi, J., Fully deformable discrete element analysis using a finite
element approach. Computefi"s nd Geotechnics, 5, pp. -175-195, 1988.
[111] Foster, N. & Metaxas, D., Visuallzation of dynamic fluid simula.tions:
waves, splashing, vorticity, boundaries, buoyancy. Engineering. Computations, 12, pp. 109-124, 1995.
[112] Hustrulid, A. & Hall, B., Parallel implementation of the discrete element method. Technical report, Colorado School of Mines, 1996.
Discontinuum Mechanics
302
(113] Xian, L., Bicanic, N., Owen, D. & Munjiza, A., Rock blasting simulation by rigid body dynamic analysis and rigid-brittle fracturing
model. 4th Intemational Conference on Nonlinear Engineering Computations - NEC-91, eds. N. Bicanic, P. Marovic, D. Owen, V. Jovic
& A. Mihanovic, Pineridge Press, pp. 587-597, 1991.
References
303
[126] Bonet, J. & Peraire, J., An alternating digital tree (ADT) algorithm
far 3d geometric searching and intersection problems. International
Journalfor Numerical Methods in Engineering, 31, pp. 1-17, 1991.
[127) Pande, G., Beer, G. & Williams, J., Numerical Methods in Rock Mechanics. John Wiley and Sons Ltd., 1990.
[115] Dutton, A. & Meek, J., Distinct element modelling of longwall gate
road roof support. Computer Methods and Advances in Geomechanics,
pp. 1757-1764, 1991.
.. "'
[116] Sawada, S., Pradhan, T. & Iwas.ki, Y., Simple shear simulations of
sand using d.e.m. Computer M ethods and Advances in Geomechanics,
pp. 511-516, 1991.
{129] Holst, J., Rotter, J., Ooi, J. & Rong,'G., Numerical modelling of silo
filling, ii: discrete element analysis. Journal o/ engineering mechanics,
125(1), pp. 104-110, 1999..
[117] Lin, X. & Ng, T., Contact detection algorithms for three-dimensional
(130) Lin, X. & Ng, T., A three-dimensional discrete element model using
arrays of ellipsoids. Geotechnique, 47(2), pp. 319-329, 1995.
ellipsoids in discrete element modelling. Intemational Journal for Numerical and Analytical Methods in Geomechanics - Short Communication, 19, pp. 653-659, 1995.
[118] Ting, J., Meachum, L. & Rowell, J., Effect of particle shape on the
strength and deformati&" mechanisrns of ellipse-sha.ped granular assemblages. Engineering Computations, 12, pp. 99-108, 1995.
[122] Herzen, B., Barr, A. & Zatz, H., Geometric collision for time dependent parametric surfaces. ACM SIGGRAPH 90, 24, pp. 39-48, 1990.
{131] Williams, J., Contact analysis of large numbers of interacting bodies using discrete modal methods for simulating material failure on
the microscopic scale. Engineering Computations, 5(3), pp. 150-161,
1988.
[132] Bazant, Z. & Planas, J., .Fracture an4 Size l!J/Ject in Concrete .nd
Other Quasibrittle Materials. CRC Press, 1998.
[133) Owen, D. & Fawkes, A., Engineering Fracture Mechanics. Pineridge
Press, 1983.
(134] Inglis, C., Stresses in a plate due to the presence of cracks and sharp
cornera. Journal o/ Institutio'h o/ Naval Architects, 55, pp. 219-~41,
1913.
[135] Griffith, A., The phenomena of rupture and fiow in solids. Joumal o/
Philos Royal Society Architects, 221, pp. 163--197, 1921.
(136) Griffith, A., The theory of rupture. The first intemational conference
o/ applied mechanics, pp. 55-63, 1924.
[137] Rice, J., A path independent integral and the approximate analysis
of strain concentration by notches and cracks. Joumal o/ Applied Mechanics - ASME, 35, pp. 379-386, 1968.
[124] Pentland, A., Computational complexity versus simulated environments. ACM SIGGRAPH Computer Graphics, 24(2), pp. 185-192,
1991.
.
[138] Irwin, C., Analysis of stresses and strains near the end of a crack
traversing a plate. Joumal of Applied Mechanics - ASME, 24, pp.
361-364, 1957.
304
Discontinuum Mechanlcs
305
L___;
English translation.
[141] Hinton, E., Numerical Methoda and Software for Dynamic Analysis
of Plates and Shells. Pineridge Press, 1987.
[155) Tsai, S. & Wu, E., A general theory of strength for anisotropic materials. Composite Materials, 5, 1971.
[142] Kachanov, M., Time of rupture process under creep conditions. Izv
Akad Nauk SSR, 8, pp. 26-31, 1958. In Russian.
[156] Whitworth, H. & Yin, 18.W., A strength theory for failure of composite laminates. Failure Analysis: Techniques and Applications, ed.
J. Dickson, 1991.
[143] Banks-Sills, L. & Schur, D., On the influence of crack plane orientation
in fatigue crack propagation and catBBtrophic failure. Fracture Mechanics: Perspectives and Directions, ASTM STP 1020, eds. R. Wei
& R. Gangloff, pp. 497-513, 1989.
[157) Rowlands, R., Strength (failure) theories and their experimental correlation. Handbook o/ Composites, Vol. 9 - Failure Mechanics of Composites, eds. G. Sih &? A. Skudra, Elsevier Science Publishers B.V.,
chapter2, pp. 71-125, 1985.
[144] Russell, A. & Street, K., On the development of a mixed mode tensileshear interlaminar fracture criterion for graphite-epoxy composites.
Modelling Problems in Crack Tip Mechanics, ed. J. Pindera, Martin
Nijhoff Publishers, 1983.
[159) Schellekens, J. & de Borst, R., The use of Hoffman yield criterion
j
'
o/ Plasticity.
Oxford University
[160) Koh, C., Owen, D. & Peric, D., Explicit dynamic analysis of elasto-
'j
[161j Peric, D., On a claBs of coqstitutive equations in viscoplasticity: Formulation and computational issues. lnternational Joumal o/ Numerical Methods in Engineering, 36, pp. 1365-1393, 1993.
composite laminate behaviour. Delamination and Debonding of Materials, ASTM STP 876, ed. W. Johnson, pp. 282-297, 1985.
[162] Simo, J. & Taylor, R., A return mapping algorithm for plane stress
elastoplasticity. Intemational Journal of Numerical Methods in Engineering, 22, pp. 649-670, 1986~
[163] Schellekens, J., Computational Strate,gies for Composite Structures.
Ph.D. thesis, Delft University of Technology, 1992.
242-260, 1961.
307
lndex
Index
acoustic, 275, 276
adaptive, 182, 210, 277, 278
ADT, 187, 212, 213, 215
alternating digital tree, 185, 187,
212, 213, 217
anisotropic, 8, 78, 133, 182, 258260
associated flow rule, 134, 262
associated slip rule, 134, 160, 161
augmented Lagrangian, 35, 43, 5254,57, 94
'-_J
...
~\
BEM, 6, 7
binary tree, 196-200, 202, 207, 215,
218
bisection, 204-207, 215
boundary element method, 6
bounding box, 213, 218
buffer zone, 190-194, 282 ., ,
279
composite, 2-5, 24, 26, 182, 183,
257-259
consistent tangent, 8, 94, 87-89,
108, 113, 115, 117, 118,
135, 142, 146, 151, 152,
164, 169, 174, 176, 177,
265, 267
constitutive, 77, 78, 122, 135, 139,
142, 146, 148-150, 153,
160, 162, 163, 233, 249,
272
306
EFM, 247
elastoplastic, 5, 122, 126, 130, 148,
150, 153, 159, 253, 266268
ellipse, 183, 221, 232-238, 240, 247,
260
Euler, 70, 71, 73, 79, 143, 144,
150, 163, 262
')'
215
'
ti
178
non-consistent, 108
271
heapsort, 210, 211
HEMP, 95
hexagon, 207, 209
hextree, 207-209
Hoffnian,259, 260, 265, 266
hydrocode, 94, 95, 99, 104
hypercube, 215, 217, 218
281
stick, 126, 127-130, 133, 142; 143,
" .., .
1, 2,
2~
181, 182
170
slideline, 93-97, 108, 109
sliding, 93, 105, 122, 129, 133, 155157, 163, 168, 178, 226,
229, 230, 233
slip, 14, 125-127, 130, 132-135,
139, 142, 143, 145, 146,
148-152,159-161,163,165,
168, 178, 223
softening, 8, 126, 132, 133, 148,
155, 156, 163, 245, 246,
249-253, 255, 263-266, 268,.
269, 271
trial function, 40 .
uncoupled, 39-41, 89
uncracked, 277
undamaged, 254
uniqueness, 269, 271
UTdem, 9, 279, 280, 285
83,87,115, 118,125,129,
137-139, 142, 153, 173, 1
174,259
virtual reality, 34
viscous damping, 92, 224-228
visualization, 184, 185
WITPRESS
Structuras Under Shock
and lmpact VII
Editors: N. IONES, University o/
Liverpool, UK, C.A. BREBBIA, Wessex
lnstitute o/ Technology, UK, and
A.M. RAJENDRAN, US Army Research
Laboratory, USA
The shock and impact behaviour of structures
is an extremely challenging area. It is
therefore important to ..recognize and utilize
the contributions emerging from theoretical,
numerical and experimental studies on
structures, as well as investigations into
material properties under dynamic loading
conditions.
Reflecting the broad range of research work
currently being carried out in academic
institutions and industry, this book contains
the papers presented at the Seventh
Intemational Conference on Structures Under
Shock and Impact. The topics highlighted
are: Protection of Structures from Blast
Loads; Surface Structures Under Explosive
Shock; Missile Penetration and Bxplosion;
Interaction Between Analytical and
Experimental Research:
Structural
Crashworthiness and New Software for
Shock and Impact; Bnergy Absorbing
Systems; Seismic Engineering Applications;
Material Characterization; Material Response
to High Rate Loading; Behaviour of Concrete
Structural Components; Behaviour of Steel
and Composite Structures; Fluid Solid
Interaction; and Analysis of Unbound
Domains.
440pp
Series:
Structures and Materials, Vol 11
Fracture of Rock
E-mail: marketing@witpress.com
612pp
L._
We~sex
Institute
Thermomechanical
Fatigue and Fractura
Series:
. lmpact and Damage on Structures, Vol 1
ISBN: 1-85312-928-3
2002
304pp
Series:
Advances in Fracture Mechanics, Vol 7
ISBN: 1-85312-549-0 2002
U 15.00/US$173.00/172.50
Series:
Computational and Experimental
Methods, Vol 8
ISBN: 1-85312-963-1
2003
!89.00/US$142.00/133.SO
232pp
344pp
Find us at
~
http://www.witpress.com
Ali prices correcl al time of going to press hui
subject lo change.
WIT Press books are available through your
booksel/er or direcl from the publisher.
Series:
l:omputational and Experimental Methods.
Vol JO
ISBN: 1-85312-987-9
2003
apx 300pp
apx '9UO/US$153.00/161.12
WITPress
,.