Вы находитесь на странице: 1из 21

http://ijd.sagepub.

com
Mechanics
International Journal of Damage
DOI: 10.1177/1056789509346693
2009; 18; 739 International Journal of Damage Mechanics
and Joseph M. Wells
Jie Shen, Jianghui Mao, German Reyes, Chi L. Chow, James Boileau, Xuming Su
A Multiresolution Transformation Rule of Material Defects
http://ijd.sagepub.com/cgi/content/abstract/18/8/739
The online version of this article can be found at:
Published by:
http://www.sagepublications.com
at:
can be found International Journal of Damage Mechanics Additional services and information for
http://ijd.sagepub.com/cgi/alerts Email Alerts:
http://ijd.sagepub.com/subscriptions Subscriptions:
http://www.sagepub.com/journalsReprints.nav Reprints:
http://www.sagepub.co.uk/journalsPermissions.nav Permissions:
http://ijd.sagepub.com/cgi/content/refs/18/8/739 Citations
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
A Multiresolution Transformation
Rule of Material Defects
JIE SHEN*
Department of Computer & Information Science, The University of
Michigan, Dearborn, MI 48128, USA
JIANGHUI MAO, GERMAN REYES AND CHI L. CHOW
Department of Mechanical Engineering, The University of Michigan,
Dearborn, MI 48128, USA
JAMES BOILEAU AND XUMING SU
Materials Research & Advanced Engineering Department, Ford
Motor Company, 2101 Village Road, Dearborn, MI 48121, USA
JOSEPH M. WELLS
JMA Associates, Mashpee, MA 02649, USA
ABSTRACT: The ability to quantify the material damage at different length scales is
critical in the multiscale analysis of material behavior from nanoscale to macroscale.
In this article, on the basis of the equivalence of complementary elastic energy we
propose a multiresolution rule that transforms different levels of material defects to
the equivalent degradation of material properties. It facilitates a sequential memory-
efficient processing of massive material defects in a multiresolution framework, and
also supports a functionality of partial damage conversion to serve different needs in
subsequent numerical analyses. Numerical simulation was conducted with different
settings of material defects. The analysis results indicate the efficacy of the proposed
method, offering a potential (i) to interface between multiscale material defects and
(ii) as an effective method of homogenization for the determination of the damage
variable in continuum damage mechanics.
KEY WORDS: material defect, multiscale analysis, microstructure, nanomechanics.
*Author to whom correspondence should be addressed. E-mail: shen@umich.edu
Figures 13, 58 and 10 appear in color online: http://ijd.sagepub.com
International Journal of DAMAGE MECHANICS, Vol. 18November 2009 739
1056-7895/09/08 073920 $10.00/0 DOI: 10.1177/1056789509346693
The Author(s), 2009. Reprints and permissions:
http://www.sagepub.co.uk/journalsPermissions.nav
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
INTRODUCTION
M
OST ENGINEERING MATERIALS contain defects of varying degrees and
sizes. These defects may range from nanoscale to macroscale. To con-
duct a multiresolution analysis of material damage, it is desirable to develop
an equivalence rule that can guide the transformation of different levels of
defects into their corresponding degrees of material degradation in the form
of effective mechanical properties or damage variables. The degraded
mechanical properties or damage variables can then be brought within the
concept of continuum damage mechanics that has been proven to be valu-
able for subsequent engineering design analysis and life prediction.
In the field of plasticity, McClintock model (McClintock, 1968),
RiceTracey model (Rice and Tracey, 1969), Gurson model (Gurson,
1977), and their modified versions (Budiansky et al., 1982; Duva and
Hutchinson, 1984; Tvergaard and Needleman, 1984; Tvergaard, 1989) pos-
tulated that all micro-defects in a material element were in the form of
spherical or cylindrical voids. This hypothesis leads to an unrealistic approx-
imation to real material defects of irregular shapes. Although the sizing
effect (Tvergaard, 1996; Fleck and Hutchinson, 1997; Liu et al., 2003;
Tvergaard and Niordson, 2004; Wen et al., 2005) and limited shape influ-
ence (Becker et al., 1989a,b; Lee and Mear, 1994; Gologanu et al., 1994;
Pardoen and Hutchinson, 2000; Pardoen and Hutchinson, 2003; Li and
Huang, 2005; Gao et al., 2005) of spherical or ellipsoidal voids were
addressed, the problem associated with the above approaches or similar
ones is the recognition that the arbitrary shape of real-life material defects
inevitably plays a crucial role in affecting micro/macro-mechanical proper-
ties, as illustrated in Figure 1 for large-size defects.
Gibson and Ashby (1982) linked the mechanical properties of cellular
materials with their relative density, a ratio of the apparent density of cel-
lular materials to the density of cell wall materials. In their mathematical
models, at least one constant needs to be determined from experimental data
by a statistical regression, which is thus not physics based. In addition, there
is no unified formula for different settings of cellular materials (e.g., open
cells vs. closed cells).
Scientists in nanostructured materials have made a phenomenal advance
in the reduction of grain or crystal size to 100 nm in the past three decades
(Gleiter, 1981). Many observations have been conducted on the quantum
effects, configuration entropy, thermal fluctuations, and discreteness of nano-
meter objects or systems with nanoscale components (particles, powders,
wires, rods, ribbons, or tubes). Different multiscale approaches have been
established to bridge between nanoscale and micro/macroscale (Tadmor
et al., 1996; Abraham et al., 1998; Rudd and Broughton, 1998; Wagner and
740 J. SHEN ET AL.
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
Liu, 2003; Shilkrot et al., 2004; Xiao and Belytschko, 2004). In the aspect of
analysis, massively parallel simulation with dislocation mechanics success-
fully produced a strain hardening effect in a 10 mm
3
volume of a single
crystal metal (Bulatov et al., 2004). Atomistic modeling approach normally
assumes, however, no microscale, mesoscale, and macroscale defects in its
analysis domain (Iesulauro et al., 2001). Although different types of hand-
shake regions were designed for a concurrent multiscale analysis with both
lattice grids and finite element (FE) meshes, an effective method of analysis
is yet to be emerged to incorporate a huge number of voids and cracks in
real-life materials into either lattice grids or FE meshes.
Traditional micromechanics provides another way to estimate the macro-
scopic mechanical properties of materials with microscale defects, including
ellipsoids and cracks (Mori and Tanaka, 1973; Budiansky and OConnell,
1976; Christensen and Lo, 1979; Hashin, 1988; Kachanov, 1992; Gelma
et al., 2009). However, multiresolution behavior of defects with different
sizes has less intensively been studied. The existing methods focus primarily
on microscale defects with little consideration on mesoscale and macro-
scopic defects, which may co-exist in real engineering structures.
The objective of this article is to establish a transformation rule that can
describe the equivalence of material damage from a multiresolution stand-
point. This rule provides a physical description of different sizes and degrees
of material defects. The rest of the article is organized as follows. In the
Nomenclature section, the notations of the problem are introduced, and in
the section following it the derivations of a new transformation rule of
material damage are provided. Later, numerical experiments and discussions
are given followed by some conclusions in the final section.
(a) (b)
Three
representative
volume elements
with different
shapes of voids
E
f
f
e
c
t
i
v
e

m
o
d
u
l
u
s

(
G
P
a
)
250
200
150
100
50
0
Volume fraction of material
0 0.2 0.4 0.6 0.8 1 1.2
Triangle
Rectangle
Circle
Figure 1. Influence of void shapes on the effective Youngs Modulus of representative
volume elements: (a) three representative volume elements with different shapes of voids
but the same material properties, (b) effect of volume fraction in the representative volume
element on effective modulus.
Multiresolution Transformation Rule of Material Defects 741
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
NOMENCLATURE
One way to represent material damage is the use of a symmetric second
order, D, as a damage tensor (Chow et al., 2007; Voyiadjis and Kattan, 2009).
The eigenvalues of D are written as D
k
(k =1,2,3), which denotes the damage
values in three orthogonal principal directions of D. The damage effect
tensor, M, is a fourth-order tensor for effective stress equation:
~ r = M(D) : r, (1)
where, ~ r and r are effective stress and conventional stress tensors, respec-
tively. Symbol (:) is a tensor product contracted on two indices. M(D) means
that M is dependent upon D.
At most, M(D) could have 21 independent components, and to another
extreme it may depend upon only one scalar value (D=D
k
, k =1,3) in
isotropic cases. For anisotropic damage, M(D) in its principal directions
could be simplified into the following form (Chow and Jie, 2009):
M(D) =
1
1D
1
0 0 0 0 0
0
1
1D
2
0 0 0 0
0 0
1
1D
3
0 0 0
0 0 0
1

(1D
1
)(1D
2
)
_
0 0
0 0 0 0
1

(1D
2
)(1D
3
)
_
0
0 0 0 0 0
1

(1D
3
)(1D
1
)
_
2
6
6
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
7
7
5
:
(2)
In this article, we focus on the simple form in Equation (2), because it
reflects the essential characteristics of damage effect just like eigenvalues
of a linear transformation matrix.
Each defect has an attribute of characteristic length, which is defined as
the maximum size among three principal directions. The principal compo-
nents analysis (Hoppe et al., 1993) is used to determine the directions at
vertex p, which is defined as the centroid of each defect. The covariance
matrix of the set of neighboring vertices is:
CV =
X
qcNbhd (p)
(q p) Q(q p),
where Nbhd(p) is the set of all the vertices that are associated with the defect
and Q is outer product operator of vectors. The vertices here could be either
surface or volume points within the analysis domain of each defect. A Jacobi
transformation (Press et al., 1992) can be used to determine eigenvectors
(v
1
, v
2
, v
3
) and eigenvalues (l
1
_l
2
_l
3
) of the CV. (v
1
, v
2
, v
3
) defines the
three principal directions of each defect.
742 J. SHEN ET AL.
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
All the defects are classified into n levels (n _1) on the basis of charac-
teristic length. Level 1 refers to the defects with characteristic length among
the smallest, while level n contains the defects with sizes being among the
greatest. When n equals 1, a multiresolution problem is degenerated into a
problem with only one level that includes all sizes of defects.
TRANSFORMATION RULE OF MATERIAL DAMAGE
Figure 2(a) demonstrates the concept of damage transformation of mate-
rial defects. Assume a domain that contains both a fine level and a coarse
level of defects. In one way, we calculate the material damage incrementally
by determining the effect of all the fine defects on the original Youngs
modulus, E
0
, first. The resulting effective modulus is E
1
due to these fine
defects alone. The influence of all the coarse defects is then determined on
the degraded material with a modified effective modulus, E
1
. The effect of
coarse defects leads to a new effective modulus, E
2
, as a final material deg-
radation sequentially from these two levels of defects. Alternatively, both
levels of defects could be considered simultaneously, leading to another
effective modulus, E
3
, which is degraded from E
0
.
E
2
and E
3
in the last paragraph reflect a damage estimation determined by
two different routes. E
3
is calculated by a regular way, i.e., by considering all
the defects simultaneously in an analysis, while E
2
is estimated in an incre-
mental way with respect to different levels of defects. Here, defects are clas-
sified into different levels on the basis of characteristic length defined in the
Nomenclature section. One effort in this article is to make the estimation of
E
2
close to E
3
, as illustrated in Figure 2(b). The validity of the damage
transformation offers a potential method of sequential homogenization, a
critical step in conducting multiscale damage mechanics.
(a) (b)
Y
o
u
n
g

s

m
o
d
u
l
u
s

(
G
P
a
)
70
40
60
50
30
10
20
E
3
E
1
E
2
0
E
0
E
1
E
2
E
3
Volume fraction of void
0.10 0.20 0.30 0.40
Figure 2. Transformation of material defects: (a) a simple case with two levels of defects
(blue and orange colors represent two different routes in the transformation of material
defects), (b) numerical equivalence of E
2
and E
3
in finite element analysis.
Multiresolution Transformation Rule of Material Defects 743
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
For the sake of illustration, the above example is limited to two levels of
defects. In reality, there are multiple levels of defects in materials. In the
following derivations, we propose a multiresolution transformation rule of
material defects. One weakness of Lemaitres hypothesis on strain equiva-
lence (Lemaitre and Chaboche, 1978) is that it leads to the asymmetry of a
stiffness matrix in the cases of anisotropic damage. By the inspiration from a
hypothesis on the equivalence of complementary elastic energy (Cordebois
and Sidoroff, 1979; Chowand Lu, 1989; Grammeneboudis et al., 2009a,b), we
assume that the complementary elastic energy at damage level i takes a form:
W
e
(r
D
i
, D
i
) =
1
2
r
T
D
i
:
~
C
1
D
i
: r
D
i
, i = 0, n, (3)
where, W
e
() denotes a complementary elastic energy, and D
i
is a damage
tensor that represents the extent of damage accumulated from level 1 to level
i. Superscript T represents the transpose of a matrix, and r
D
i
is a stress
tensor at damage D
i
.
~
C
1
D
i
refers to the inverse of an effective stiffness
tensor, and has the following special case:
~
C
1
D
0
= C
1
, (4)
in which C
1
is the inverse of a stiffness tensor at damage D
0
=0. A general
formula for
~
C
1
D
i
is given at a later part of this section. r
D
0
equals the con-
ventional stress tensor, r.
We also assume that the equivalence of complementary energy takes the
following incremental form:
W
e
(r
D
i1
, D
i1
) = W
e
( ~ r
D
i1
, D
i
) =
1
2
~ r
T
D
i1
:
~
C
1
D
i
: ~ r
D
i1
=
1
2
~ r
T
D
i
: (M(dD
i
)
T
:
~
C
1
D
i
: M(dD
i
)) : ~ r
D
i
, i = 0, n 1,
(5)
where ~ r
D
i1
is an effective stress tensor at damage D
i+1
and is given by:
~ r
D
i1
= M(dD
i
) : ~ r
D
i
,
dD
i
= D
i1
D
i
, i = 0, n 1:
(6)
Besides the above general formula, one special case for ~ r
D
i1
is:
~ r
D
1
= M(D
1
) : r
D
0
= M(D
1
) : r, (7)
where r is a conventional stress tensor at damage D
0
=0. From Equation (5),
the general formula for
~
C
1
D
i
is defined as:
~
C
1
D
i1
= M(dD
i
)
T
:
~
C
1
D
i
: M(dD
i
), i = 0, n 1: (8)
The above equation defines a multiresolution transformation rule for mate-
rial defects. The fundamental motivation for such an incremental form is to
744 J. SHEN ET AL.
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
develop an efficient means of sequential homogenization for varying sizes
and degrees of defects in a material element.
To take an example of using Equation (8), we consider a two-step incre-
mental transformation with the defects at levels i and i +1. Incremental
application of Equation (8) leads to:
~
C
1
D
i2
= M(dD
i1
)
T
: M(dD
i
)
T
:
~
C
1
D
i
: M(dD
i
) : M(dD
i1
): (9)
If
~
C
1
D
i
is represented by:
~
C
1
D
i
=
1
~
E
D
i
1 ~
D
i
~
D
i
0 0 0
~
D
i
1 ~
D
i
0 0 0
~
D
i
~
D
i
1 0 0 0
0 0 0 2(1 ~
D
i
) 0 0
0 0 0 0 2(1 ~
D
i
) 0
0 0 0 0 0 2(1 ~
D
i
)
2
6
6
6
6
6
6
4
3
7
7
7
7
7
7
5
,
where
~
E
D
i
and ~
D
i
are respectively the effective elastic modulus and
Poissons ratio, we then have:
^
C
1
D
i2
=
1
~
E
D
i

b
11
~
D
i
b
12
~
D
i
b
13
0 0 0
~
D
i
b
21
b
22
~
D
i
b
23
0 0 0
~
D
i
b
31
~
D
i
b
32
b
33
0 0 0
0 0 0 2(1 ~
D
i
)b
12
0 0
0 0 0 0 2(1 ~
D
i
)b
23
0
0 0 0 0 0 2(1 ~
D
i
)b
31
2
6
6
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
7
7
5
,
(10)
where
b
st
=
1
(1 dD
i
[s])(1 dD
i1
[s])(1 dD
i
[t])(1 dD
i1
[t])
, (where s, t =1, 2, 3)
in which, dD
i
[k] (k =1,3) and dD
i+1
[k] (k =1,3) are the primary components
of damage increments at D
i
and D
i+1
, respectively.
To check the accuracy of Equation (10), a single step transformation can
be envisioned as follows:
~
C
1
D
i2
= M(dD
i
dD
i1
)
T
:
~
C
1
D
i
: M(dD
i
dD
i1
), (11)
Multiresolution Transformation Rule of Material Defects 745
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
which then leads to:
~
C
1
D
i2
=
1
~
E
D
i

b
/
11
~
D
i
b
/
12
~
D
i
b
/
13
0 0 0
~
D
i
b
/
21
b
/
22
~
D
i
b
/
23
0 0 0
~
D
i
b
/
31
~
D
i
b
/
32
b
/
33
0 0 0
0 0 0 2(1 ~
D
i
)b
/
12
0 0
0 0 0 0 2(1 ~
D
i
)b
/
23
0
0 0 0 0 0 2(1 ~
D
i
)b
/
31
2
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
5
,
(12)
where
b
/
st
=
1
(1 dD
i
[s] dD
i1
[s])(1 dD
i
[t] dD
i1
[t])
:
^
C
1
D
i2
predicted by the incremental form in Equation (10) should be a close
approximation to
~
C
1
D
i2
calculated in Equation (12).
NUMERICAL ANALYSIS AND DISCUSSIONS
Numerical simulations are conducted by a FE package. To verify the pro-
posed method, a series of regular FE domains are constructed with different
settings of material defects embedded in each FE domain. In this article, we
consider these FE domains as representative volume elements (RVEs). When
our method is applied to the engineering analysis of real structures, RVEs will
be distributed as a local neighborhood within the domain of real structures.
One classical criticism to the RVE approach is that the RVE analysis gives
an upper bound and a lower bound of apparent stiffness with specified
displacement and traction conditions, respectively (Hollister and Kikuchi,
1992). In other words, the estimated stiffness varies with the boundary
conditions. In the context of regular FE domains, we found out a special
displacement boundary condition that almost (<1% error) produced an
exact solution of the apparent stiffness in the RVE analysis. The special
boundary condition includes the following specifications:
(1) All the RVE components are created in alignment with three coordinate
directions;
(2) A set of uniformly specified displacements in one of three coordinate
directions is applied only on two opposite boundary faces whose surface
normal is in the same direction as the tensile loading direction;
746 J. SHEN ET AL.
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
(3) Two nodes are constrained at one corner of the RVE to prevent any rigid
body movement.
The average strain and stress in the regular FE domain can be easily
computed on the basis of the FE analysis. The effective elasticity modulus
is then calculated by a ratio of average stress to average strain in one of three
coordinate directions.
To simplify the validation of Equation (10), 1D cases are considered.
Equations (10) and (12) are respectively degenerated into:
^
C
D
i2
=
~
E
D
i
(1 dD
i
[1])
2
(1 dD
i1
[1])
2
=
~
E
D
i
(1 dD
i
[1] dD
i1
[1] B)
(13)
~
C
D
i2
=
~
E
D
i
(1dD
i
[1] dD
i1
[1])
2
=
~
E
D
i
(1dD
i
[1] dD
i1
[1] A), (14)
where,
B = dD
i
[1]dD
i1
[1](1 (1 dD
i
[1])(1 dD
i1
[1]))
(dD
i
[1] dD
i1
[1])(1 dD
i
[1])(1 dD
i1
[1]),
A = (dD
i
[1] dD
i1
[1])(1 dD
i
[1] dD
i1
[1]): (15)
The difference between
~
C
D
i2
and
^
C
D
i2
is then:
~
C
D
i2

^
C
D
i2
=
~
E
D
i
(A B): (16)
On the basis of Equations (9) and (11), for 1D cases we have the following
three hypotheses:
I. FE prediction of
^
C
D
i1
can be represented by:
^
C
D
i1
=
^
E
D
i1
=
~
E
D
i
(1 dD
i
[1])
2
II. FE prediction of
^
C
D
i2
can be represented by:
^
C
D
i2
=
^
E
D
i2
=
~
E
D
i1
(1 dD
i1
[1])
2
III. FE prediction of
~
C
D
i2
can be represented by
~
C
D
i2
=
~
E
D
i2
=
~
E
D
i
(1 dD
i
[1] dD
i1
[1])
2
To ascertain the degree of approximation of
^
C
D
i2
to
~
C
D
i2
, a numerical
scheme is designed with the procedure described in Table 1.
The shape of real defects in materials has been observed to be random and
arbitrary, and may contain c1 discontinuity at their sharp edges and corners.
Here, c1 discontinuity means the discontinuity of the first-order derivative of
surface or boundary-line functions. Accordingly, numerical simulations of
the equivalency analysis become necessary to determine the change in effec-
tive elastic moduli and associated damage variables. In this article, we design
a series of numerical test examples with different levels of material defects.
Multiresolution Transformation Rule of Material Defects 747
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
Figure 3 depicts numerical models for different defect shapes, including
sphere, ellipsoid, triangular prism, rectangular prism, polygonal prism, and
arbitrarily shaped cracks. The numerical results of two-level damage trans-
formation in one dimension are given in Table 2. In the heading of this table,
i =0 refers to original or as-received state as a starting point for a config-
uration of two-level damage, and megapascal is the unit for all the moduli.
The material damage is measured in terms of effective modulus in the
vertical direction.
According to Table 2, the damage transformation estimated by the incre-
mental formula (
^
E
num
D
i2
) is quite close to the actual damage,
~
E
fem
D
i2
, for differ-
ent types of voids as well as the combination of voids and cracks. For cracks
alone, the difference between these two, however, becomes significant, as
indicated in row 6 for crack in Table 2.
One fundamental question to be addressed is the rationale behind the use of
incremental formula if the compound damage,
~
E
fem
D
i2
, can be calculated by a one-
step FE analysis. There are two main reasons for using the proposed method:
(1) The proposed method provides a sequential homogenization of material
defects. This can thus simplify what would otherwise be a time-consuming
and laborious computing in numerical modeling, meshing, and analysis of
prohibitive number of micro-defects. The simplification is achieved by
processing the defects level by level in sequence.
(2) The incremental formula facilitates partial damage transformation. This is
extremely useful in the cases where only fine defects need damage trans-
formation, while coarse defects remain in the subsequent analysis domain.
Take for example both voids and cracks as one type of inclusions in a
matrix-based material. The effective modulus of the inclusions may vary at
different levels with respect to that of the matrix material. Figure 4 is a test case
Table 1. Numerical Comparison of
^
C
D
i2
and
~
C
D
i2
.
Step 1. Conduct an FE analysis on defects at level i +1. On the basis of FE results, an
estimate of
^
E
D
i1
is denoted by
^
E
fem
D
i1
, from which dD
i
[1][
fem
= 1

^
E
fem
Di1
=
^
E
Di
q
.
Step 2. Conduct an FE analysis on defects at level i +2. On the basis of FE results, an
estimate of
^
E
D
i2
is denoted by
^
E
fem
D
i2
, from which dD
i1
[1][
fem
= 1

^
E
fem
Di2
=
^
E
fem
Di1
q
.
Step 3. Conduct an FE analysis on defects at the combination of level i +1 and level i +2.
On the basis of FE results, an estimate of
~
E
D
i2
is denoted by
~
E
fem
D
i2
.
Step 4. Calculate
^
E
num
D
i2
by using Equation (13):
^
E
num
D
i2
=
~
E
D
i
(1 dD
i
[1][
fem
dD
i1
[1][
fem
A), where A is calculated by Equation (15) with
dD
i
[1][
fem
and dD
i1
[1][
fem
replacing dD
i
[1] and dD
i1
[1], respectively. Note that for cracks,
^
E
num
D
i2
should be further corrected to
^
E
cor
D
i2
(see the later part of this section for details.)
Step 5. Compare
^
E
num
D
i2
and
~
E
fem
D
i2
as a numerical comparison of
^
C
D
i2
and
~
C
D
i2
.
748 J. SHEN ET AL.
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
for studying the effect of the effective modulus of inclusions on the damage
transformation. Results of numerical analyses are given in Table 3, indicating
that the smaller the ratio of inclusions modulus to matrixs modulus is, the
more remarkable the difference between
^
E
num
D
i2
and
~
E
fem
D
i2
becomes. When the
(a) (b)
(c)
(d)
2
1
3
2
1
3
2
1
3
2
1
3
(e) (f)
2
1
3
2
1 3
(g)
2
1 3
Figure 3. Numerical models of material defects with different shapes: (a) sphere, (b) ellip-
soid, (c) triangular prism, (d) rectangular prism, (e) polygonal prism, (f) crack,
(g) crack +void.
Multiresolution Transformation Rule of Material Defects 749
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
ratio is zero, inclusions are solely consisted of holes, and the difference reaches
its maximum if no strong inclusion is considered. Although the difference
between
^
E
num
D
i2
and
~
E
fem
D
i2
in Table 3 is not significant for void-type inclusions
(Figure 5), this is not true when the increment of damage becomes large, as
demonstrated by the next example in Figure 6 and Table 4.
Figure 6 is a typical model with two-level circular holes: one large hole
and four small holes. It is used to investigate the effect of damage increment
on the transformation rule of material defects. Here, the damage increment
means the extent of extra damage induced at certain level of defects. Let dD
i
(i =0, n 1) be defined an incremental damage tensor. In 1D cases, dD
i
is
degenerated into one principal component of the damage tensor.
The first column of Table 4 represents the radii of these two sizes of holes
in each test case. A general tendency is that the bigger the holes are, the
2
1
3
Figure 4. Numerical model with weak inclusions.
Table 2. Shape effect of material defects on the transformation
of material damage (i ^0; unit: MPa).
Defect shape
^
E
fem
D
i1
^
E
fem
D
i2
^
E
num
D
i2
~
E
fem
D
i2
Sphere 66134.2 58412.2 58196.8 58022
Ellipsoid 69048.4 62109.6 62063.2 62000.2
Triangular prism 66211.4 51661.2 51276.9 51181.6
Rectangular prism 62877 41520.8 40471.9 39283.4
Polygonal prism 52308 38642.4 36686.7 34647.8
Crack 43609 25921 21972.8 29610.8
Crack +Void 63098.8 52734.6 52208.6 52082.8
750 J. SHEN ET AL.
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
more significant the difference between
^
E
num
D
i2
and
~
E
fem
D
i2
becomes. It is inter-
esting to note that even for void-type defects the difference between
^
E
num
D
i2
and
~
E
fem
D
i2
could increase, as illustrated in the last row of Table 4.
In the table, the larger the radii are, the bigger the damage increment is.
This is analogous to the error caused by a Taylor series expansion. When the
increment becomes large, a linear approximation becomes insufficient to
maintain a low approximation error. One solution to this is to try to use
a reasonable number of multiple levels in the multiresolution damage trans-
formation such that each damage increment is not too large at each level.
Figure 7 is a typical model with two-level cracks: one large crack and four
small cracks. The first column of Table 5 represents the lengths of these two
sizes of cracks in each test case. A general tendency is that the bigger the
cracks are, the more significant the difference between
^
E
num
D
i2
and
~
E
fem
D
i2
E
i
: inclusion E
0
: matrix
50000
55000
60000
65000
70000
75000
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
E
i
/E
0
E

(
M
P
a
)
E
2
E
3
Figure 5. Damage transformation with weak inclusions.
Table 3. Effect of weak inclusions on the transformation of
material damage (i ^0; unit: MPa).
Ratio of inclusions modulus
to matrixs modulus
^
E
fem
D
i1
^
E
fem
D
i2
^
E
num
D
i2
~
E
fem
D
i2
0 68681.8 52955.8 52815.5 53120.4
0.25 69279.4 61436.8 61397.4 61444.8
0.5 69618.8 65641.6 65630.9 65627.4
0.75 69841.4 68229.2 68227.4 68219
1 70000 70000 70000 70000
Multiresolution Transformation Rule of Material Defects 751
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
becomes. As observed by other researchers (Kachanov, 1992), cracks behave
differently from void-type defects. The damage transformation of Equation
(14) may be modified specifically for the cases of cracks as:
^
C
cor
D
i2
=
^
E
fem
D
i2
dD
i
[1] 50:15 or dD
i1
[1] 50:15
~
E
D
i
(1 dD
i
[1] dD
i1
[1] rA otherwise
8
<
:
, (17)
where dD
i
[1] is a scalar value to represent one principal component of dD
i
,
an incremental damage tensor at level i. When it is smaller than 0.15
(a threshold that was found to be reasonable in our study), we use the
same formula as that in Equation (14). r is a geometric correction factor,
and is defined as:
r =
L
defect
L
circle
, (18)
3
1
2
Figure 6. A typical two-level hole model.
Table 4. Influence of holes on the transformation
of material damage (i ^0; unit: MPa).
Radii of holes
^
E
fem
D
i1
^
E
fem
D
i2
^
E
num
D
i2
~
E
fem
D
i2
0.025/0.05 68387.6 66805 66786.6 66761.6
0.05/0.1 63910 58290 58034.6 57711.4
0.07/0.14 58836 49285.8 48460.9 47476.2
0.1/0.2 49952.6 35275.2 32854.8 30568.8
752 J. SHEN ET AL.
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
where L
defect
and L
circle
are respectively the perimeter of each defect and a
corresponding circular hole whose diameter is the same as the maximum
length of the defect. For a flat crack, r = 4=(2) - 0:64. The above formula
is suitable to crack-type defects, while Equation (14) is for void-type defects.
The last column of Table 5 refers to the results using Equation (17).
Figure 8 demonstrates the effect of damage increment on the error in
damage transformation. The horizontal axis represents the amount of mod-
ulus change caused by material defects. The larger the modulus changes, the
bigger the damage increment becomes. The vertical axis in Figure 8 refers to
relative error, which is evaluated by (
^
E
fem
D
i2

~
E
fem
D
i2
)/
~
E
fem
D
i2
and
(
^
E
cor
D
i2

~
E
fem
D
i2
)/
~
E
fem
D
i2
respectively for (crack, hole) and (crack_cor, hole_cor)
in the legend of Figure 8. Here, crack_cor and hole_cor represent the
damage transformation with a correction term by using
^
E
cor
D
i2
, while
(crack, hole) in the legend refers to the computation without a correction
2
3
1
Figure 7. A typical two-level crack model.
Table 5. Influence of cracks on the transformation of material
damage (i ^0; unit: MPa).
Lengths of cracks
^
E
fem
D
i1
^
E
fem
D
i2
^
E
num
D
i2
~
E
fem
D
i2
^
E
cor
D
i2
0.05/0.1 68906.2 67839.2 67830.8 67834.2 67839.2
0.1/0.2 65801 61846.4 61724.2 61765.6 61846.4
0.15/0.3 61084.6 53209.4 52674.8 53058.4 53209.4
0.2/0.4 55224.2 43345.6 41952.4 44379.2 43345.6
0.25/0.5 48751.4 33505 30820.7 37493.6 36446.7
Multiresolution Transformation Rule of Material Defects 753
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
term developed in this article. From this figure, it is evident that
^
E
cor
D
i2
offers
a more accurate damage transformation than
^
E
fem
D
i2
for hole-type voids. As
to cracks,
^
E
cor
D
i2
is effective in reducing the transformation error when the
damage increment is very large.
The computation of
^
E
num
D
i2
in Table 1 and
^
E
cor
D
i2
in Equation (17) is con-
ducted with the two-level damage configuration. For multiple-level defects,
we could easily apply it in a recursive way. To demonstrate this, we consider
a typical three-level case shown in Figure 9. For three-level defects, we need
two recursive steps to compute the final material damage, as illustrated in
Table 6. In step 1, defects at levels 1 and 2 are considered, while in step 2,
defects at levels 2 and 3 are taken into account. The result of step 1,
^
E
cor
D
i2
, is
fed into step 2 as the degraded principal component caused by the defects at
level 2. In this particular case, both
^
E
num
D
i2
and
^
E
cor
D
i2
are close to
~
E
fem
D
i2
. There
is no need to apply the geometric correction factor r in the computation,
because the damage increments are smaller than the threshold.
The incremental formula in Table 1 and Equation (17) would be especially
suited for a significant number of defects in real materials. Figure 10 is a
typical case of multiresolution visualization of defects in a specimen of alu-
minum alloy. It is noted that the damage field in each specimen is not
expected to be uniformly distributed. The damage transformation rule
0.00
2.00
4.00
6.00
8.00
10.00
12.00
14.00
16.00
18.00
0
Modulus change (MPa)
R
e
l
a
t
i
v
e

e
r
r
o
r

(
%
)
Crack
Crack_cor
Hole
Hole_cor
10,000 20,000 30,000 40,000
Figure 8. The effect of damage increment and a correction term on the damage transfor-
mation of material defects.
754 J. SHEN ET AL.
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
should be applied only to each local neighborhood inside the material
domain. For the previous cases in Figures 19, we treat each analysis
domain as one RVE, which should be viewed as being equivalent to a
local neighborhood in a material specimen or structure.
In the numerical simulations of this article, we do not particularly con-
sider anisotropic damage, even though the proposed damage transformation
scheme is equally applicable. The effect of Poissons ratio is also neglected
but will be a topic of our future investigation.
(a) (b)
Figure 9. Numerical models with multiple-level defects: (a) holes, (b) cracks.
Table 6. Influence of multiple-level defects on the transformation
of material damage (i ^0; unit: MPa).
Sizes of defects
^
E
fem
D
i1
^
E
fem
D
i2
^
E
num
D
i2
~
E
fem
D
i2
^
E
cor
D
i2
Recursive Level 1
Hole 0.1/0.05/0.025 64,149.6 58,646.4 58,406.7 58,406.7
Crack 0.2/0.1/0.05 55,371.4 38,595.2 36,719.9 38,595.2
Recursive level 2
Hole 0.1/0.05/0.025 58,406.7 53,544 53,315.8 51,451 53,315.8
Crack 0.2/0.1/0.05 38,595.2 26,737 24,648.8 25,855 26,737
All defects
Multiresolution visualization of material defects of aluminum alloy
Large defects only
Figure 10. Multiresolution visualization of material defects.
Multiresolution Transformation Rule of Material Defects 755
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
CONCLUSIONS
In this article, on the basis of the equivalence of complementary elastic
energy, we establish a multiresolution transformation rule for material
damage due to defects. Numerical simulations show the efficacy of the pro-
posed formula in dealing with various types of defects, including micro-voids
and micro-cracks. The transformation rule performs reasonably well in dif-
ferent testing cases, including various defect shapes, weak inclusions, different
damage increments, and two-level and three-level configurations. It would
facilitate the multiresolution analysis of material damage, especially for the
conversion of partial damage at a certain level of defects and for numerical
expediency in handling millions of defects in real materials in a sequential
manner. Our numerical strategy is considered as an effective tool for sequen-
tial damage homogenization and multiscale damage characterization.
ACKNOWLEDGMENTS
This work was supported in part by U.S. National Science Foundation
Grant CMMI-0721625, a University of Michigan OVPR grant, and a
University of Michigan Dearborn CEEP grant.
REFERENCES
Abraham, F.F., Broughton, J., Bernstein, N. and Kaxiras, E. (1998). Spanning the Continuum
to Quantum Length Scales in a Dynamic Simulation of Brittle Fracture, Europhysics
Letters, 44: 783787.
Becker, R., Smelser, R.E. and Richmond, O. (1989a). The Effect of Void Shape on the
Development of Damage and Fracture in Plane-strain Tension, Journal of the
Mechanics and Physics of Solids, 37: 111129.
Becker, R., Smelser, R.E., Richmond, O. and Appleby, E.J. (1989b). The Effect of Void Shape
on Void Growth and Ductility in Axisymmetric Tension Tests, Metallugical Transactions
A, 20: 853861.
Budiansky, B., Hutchinson, J.W. and Slutsky, S. (1982). Void Growth and Collapse in Viscous
Solids, In: Hopkins, H.G. and Sewell, M.J. (eds), Mechanics of Solids, The Rodney Hill
60th Anniversary Volume, Oxford, UK, Pergamon Press, pp. 1345.
Budiansky, B. and OConnell, R.J. (1976). Elastic Moduli of a Cracked Solid, International
Journal of Solids and Structures, 12: 8197.
Bulatov, V.V., Cai, W., Fier, J., Hiratani, M., Pierce, T., Tang, M., et al. (2004). Scalable Line
Dynamics in ParaDiS, In: Proceedings of the ACM/IEEE Super Computing 2004
Conference, p. 19.
Chow, C.L. and Jie, M. (2009). Anisotropic Damage Coupled Sheet Metal Forming Limit
Analysis, International Journal of Damage Mechanics, 18: 371392.
Chow, C.L., Jie, M. and Wu, X. (2007). A Damage-coupled Criterion of Localized Necking
Based on Acoustic Tensor, International Journal of Damage Mechanics, 16: 265282.
756 J. SHEN ET AL.
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
Chow, C.L. and Lu, T.L. (1989). On Evolution Laws of Anisotropic Damage, Engineering
Fracture Mechanics, 34: 679701.
Christensen, R. and Lo, K.H. (1979). Solutions for Effective Shear Properties in Three Phase
Sphere and Cylinder Models, Journal of the Mechanics and Physics of Solids, 27: 315330.
Cordebois, J. and Sidoroff, F. (1979). Damage Induced Elastic Anisotropy, In: Mechanical
Behaviour of Anisotropic Solids, London, Paris, Martinus Nijhoff Publishers, pp.761774.
Duva, J.M. and Hutchinson, J.W. (1984). Constitutive Potentials for Dilutely Voided
Nonlinear Materials, Mechanics of Materials, 3: 4154.
Fleck, N.A. and Hutchinson, J.W. (1997). Strain Gradient Plasticity, In: Hutchinson, J.W. and
Wu, T.Y. (eds), Advance in Applied Mechanics, New York, Academia Press, pp. 295361.
Gao, X., Wang, T. and Kim, J. (2005). On Ductile Fracture Initiation Toughness: Effect of
Void Volume Fraction, Void Shape and Void Distribution, International Journal of Solids
and Structures, 42: 50975117.
Gibson, L.J. and Ashby, M.F. (1982). The Mechanics of Three-dimensional Cellular Materials,
Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences,
382: 4359.
Gleiter, H. (1981). Deformation of Polycrystals, In: Proceedings of 2nd Riso International Sym-
posium on Metallurgy and Materials Science, Rosekilde, Denmark, Riso Nat. Lab., p. 4.
Glema, A., Lodyowski, T., Sumelka, W. and Perzyna, P. (2009). The Numerical Analysis of the
Intrinsic Anisotrpic Microdamage Evolution in Elastic-plastic Solids, International
Journal of Damage Mechanics, 18: 205232.
Gologanu, M., Leblond, J. and Devaux, J. (1994). Approximate Models for Ductile Metals
Containing Nonspherical Voids Case of Axisymmetric Oblate Ellipsoidal Cavities,
Journal of Engineering Materials and Technology, 116: 290297.
Grammeneboudis, P., Reckwerth, D. and Tsakmakis, C.H. (2009a). Continuum Damage
Models Based on Energy Equivalence: Part IIsotropic Material Response,
International Journal of Damage Mechanics, 18: 3164.
Grammeneboudis, P., Reckwerth, D. and Tsakmakis, C.H. (2009b). Continuum Damage
Models Based on Energy Equivalence: Part IIAnisotropic Material Response,
International Journal of Damage Mechanics, 18: 6591.
Gurson, A.L. (1977). Continuum Theory of Ductile Rupture by Void Nucleation and Growth,
1. Yield Criteria and Flow Rules for Porous Ductile Media, Journal of Engineering
Materials and Technology, 99: 215.
Hashin, Z. (1988). The Differential Scheme and its Application to Cracked Materials, Journal
of the Mechanics and Physics of Solids, 36: 719734.
Hollister, S.J. and Kikuchi, N. (1992). A Comparison of Homogenization and Standard
Mechanics Analyses for Periodic Porous Composites, Computational Mechanics, 10:
7395.
Hoppe, H., De Rose, T., Duchamp, T., MaDonald, J. and Stuetzle, W. (1993). Mesh
Optimization, In: Proceedings of the 20th Annual Conference on Computer Graphics and
Interactive Techniques, Anaheim, California, USA, 26 August, Vol. 27, pp. 1926.
Iesulauro, E., Dodhia, K., Cretegny, T., Chen, C.S., Myers, C. and Ingraffea, A.R. (2001).
Continuum-atomistic Modeling for Crack Initiation and Propagation in Polycrystals, In:
Proceedings of the 10th International Conference on Fracture, Hawaii, Elsevier Science
Publishers, 26 December, p. 6 CD-ROM.
Kachanov, M. (1992). Effective Elastic Properties of Cracked Solids: Critical Review of Some
Basic Concepts, Applied Mechanics Reviews, 45: 304335.
Lee, B.J. and Mear, M.E. (1994). Studies of the Growth and Collapse of Voids in Viscous
Solids, Journal of Engineering Materials and Technology, 116: 348358.
Lemaitre, J. and Chaboche, J.L. (1978). Phenomenological Approach of Damage Rupture,
Journal de Mecanique Appliquee, 2: 317365.
Multiresolution Transformation Rule of Material Defects 757
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from
Li, Z. and Huang, M. (2005). Combined Effects of Void Shape and Void Size Oblate
Spheroidal Microvoid Embedded in Infinite Non-linear Solid, International Journal of
Plasticity, 21: 625650.
Liu, B., Qiu, X., Huang, Y., Hwang, K.C., Li, M. and Liu, C. (2003). The Size Effect on Void
Growth in Ductile Materials, Journal of the Mechanics and Physics of Solids, 51:
11711181.
McClintock, F.A. (1968). A Criterion for Ductile Fracture by the Growth of Holes, Journal of
Applied Mechanics, 35: 363371.
Mori, T. and Tanaka, K. (1973). Average Stress in Matrix and Average Elastic Energy of
Materials with Misfitting Inclusions, Acta Metallurgica, 21: 571574.
Pardoen, T. and Hutchinson, J.W. (2000). An Extended Model for Void Growth and
Coalescence, Journal of the Mechanics and Physics of Solids, 48: 24672512.
Pardoen, T. and Hutchinson, J.W. (2003). Micromechanics-based Model for Trends in
Toughness of Ductile Metals, Acta Materialia, 51: 133148.
Press, W.H., Teukolsky, S.A., Vetterling, W.T. and Flannery, B.P. (1992). Numerical Recipes in
C: The Art of Scientific Computing, Cambridge, Cambridge University Press, p. 994.
Rice, J.R. and Tracey, D.M. (1969). On the Ductile Enlargement of Voids in Triaxial Stress
Fields, Journal of the Mechanics and Physics of Solids, 17: 201217.
Rudd, R.E. and Broughton, J.Q. (1998). Coarse-grained Molecular Dynamics and the Atomic
Limit of Finite Elements, Physics Review B, 58: 58935896.
Shilkrot, L.E., Curtin, W.A. and Miller, R.E. (2004). Multiscale Plasticity Modeling: Coupled
Atomistics and Discrete Dislocation Mechanics, Journal of the Mechanics and Physics of
Solids, 52: 755787.
Tadmor, E., Ortiz, M. and Phillips, R. (1996). Quasicontinuum Analysis of Defects in Solids,
Philosophical Magazine A, 73: 15291563.
Tvergaard, V. (1989). Material Failure by Void Growth to Coalescence, Advances in Applied
Mechanics, 27: 83151.
Tvergaard, V. (1996). Effect of Void Size Difference on Growth and Cavitation Instabilities,
Journal of the Mechanics and Physics of Solids, 44: 12371253.
Tvergaard, V. and Needleman, A. (1984). Analysis of the Cup-cone Fracture in a Round
Tensile Bar, Acta Metallurgica, 32: 157169.
Tvergaard, V. and Niordson, C. (2004). Nonlocal Plasticity Effects on Interaction of Different
Size Voids, International Journal of Plasticity, 20: 107120.
Voyiadjis, G.Z. and Kattan, P.I. (2009). A Comparative Study of Damage Variable in
Continuum Damage Mechanics, International Journal of Damage Mechanics, 18: 315340.
Wagner, G.J. and Liu, W.K. (2003). Coupling of Atomistic and Continuum Simulations Using
a Bridging Scale Decomposition, Journal of Computational Physics, 190: 249274.
Wen, J., Huang, Y., Hwang, K.C., Liu, C. and Li, M. (2005). The Modified Gurson Model
Accounting for the Void Size Effect, International Journal of Plasticity, 21: 381395.
Xiao, S.P. and Belytschko, T. (2004). A Bridging Domain Method for Coupling Continua with
Molecular Dynamics, Computer Methods in Applied Mechanics and Engineering, 193:
16451669.
758 J. SHEN ET AL.
at UNIV OF MICHIGAN DEARBORN on October 27, 2009 http://ijd.sagepub.com Downloaded from

Вам также может понравиться