Sei sulla pagina 1di 14

Proceedings of the Institution of

Civil Engineers
Engineering and Computational
Mechanics 161
March 2008 Issue EM1
Pages 316
doi: 10.1680/eacm.2008.161.1.3
Paper 700002
Received 12/10/2007
Accepted 22/04/2008
Keywords: concrete structures/
research & development/stress
analysis
Siew Chang Hee
Structural Engineer, School of
Engineering, Cardiff
University, UK
Anthony D. Jefferson
Senior Lecturer, School of
Engineering, Cardiff
University, UK
A new model for simulating cracks in cementitious composites
S. C. Hee and A. D. Jefferson
A new model for simulating cracks in cementitious
composites using embedded planes with local plasticity is
presented. In the model it is assumed that the
embedded planes of degradation (PODs) can not only
undergo separation but can also regain contact, with the
state of contact being controlled by a local contact
model. A local plasticity model is included to capture
permanent relative displacements associated with rough
contact between the surfaces of embedded cracks. This
model is a signicant development of the Craft model
developed by the second author, which allows shear
contact, or aggregate interlock, behaviour, as well as
crack opening/closing behaviour, to be simulated more
accurately. A fully consistent tangent stiffness matrix and
stress recovery algorithm is derived. The model has
been implemented in a constitutive driver program and
also in the nite-element program Lusas. The model is
assessed against a series of single-point stressstrain
paths, and against data from normal-shear and cyclic
tests. It is shown that the inclusion of local plasticity
signicantly improves the accuracy of the predictions of
the model when judged against experimental data.
1. INTRODUCTION
Over the years there have been many signicant contributions
in the area of constitutive modelling of cementitious composite
materials, and of concrete modelling in particular. These
include models developed using the plasticity theory,
13
continuum damage mechanics,
4,5
combined plasticity and
damage
69
and microplane theory.
1012
Problems encountered due to strain-softening in nite-element
material models are well documented and have motivated the
development of various regularisation techniques. These
include the crack band model
13
and non-local models.
14
The
latter may be subcategorised into integral-type models
15, 16
and
explicit and implicit gradient-type models.
17
It has been known for many years that aggregate interlock
provides a signicant contribution to the overall strength of
reinforced-concrete structures.
18
It is evident from
experimental tests that essentially the same mechanism of
rough crack contact or aggregate interlock occurs between the
surfaces of micro cracks, in the early stages of concrete
cracking, and macro cracks, in the latter stages of
cracking.
1922
Thus a single approach to modelling rough crack
contact throughout the cracking range is considered
reasonable. Shear transfer across cracks has been incorporated
into a number of interface models for concrete cracks, in which
the interaction between both normal and shear components of
crack planes is used to dene the constitutive
relationships.
2325
The Craft model
2628
employs embedded planes of degradation
(PODs), which are assumed to develop according to a damage
theory. Contact theory is then applied to the damaged part of
each POD. The model is able to describe certain characteristic
behaviour of concrete during cracking, including directional
softening (or cracking), and the loss and regaining of contact
with normal and shear movements of the crack faces. Since the
model employs a local damage model, the predicted unloading
behaviour of a crack follows a secant path, whereas
experiments show that unloading behaviour of cracked
concrete is neither purely elastic, as would be predicted by
plasticity theory, nor secant, as predicted by damage theory.
This inadequacy in the original model is addressed in the
present work by incorporating local plasticity on the embedded
planes of degradation.
29
The present paper describes the
development of the theory, its implementation, and
comparisons with experimental data. The derivation of a fully
consistent tangent stiffness matrix and stress recovery
algorithm is presented. A brief description of how consistency
is achieved is also demonstrated in this paper. The model has
been implemented in a 3D constitutive driver program as well
as in the nite-element program
30
via the user material model
interface feature of the program.
2. THEORETICAL FORMULATION
The global plasticity and local plastic-damage-contact
components of the model will be described in the following
subsections.
2.1. Global plasticity component
The global plasticity component of the model is essentially
similar to that of the original Craft model, and therefore only
essential details are given here.
The yield function is given by
31
F , Z

J
2
_
A
r


3
_ _
I
1
Z f
c
Z 1
1
Engineering and Computational Mechanics 161 Issue EM1 Simulating cracks in cementitious composites Hee Jefferson 3
in which the elliptic function A
r
is dened as
32
A
r
, r r
c
4 1 r
2
_ _
cos
2
2r 1
2
2 1 r
2
_ _
cos 2r 1
3

4 1 r
2
cos
2
5r
2
4r
_
_

_
_

_ 2
f
c
, I
1
and J
2
are the uniaxial compressive strength, the rst
stress invariant and the second deviatoric stress invariant
respectively; denotes the Lode angle, with range 0608;
r is the eccentricity parameter that controls the out-of-
roundness of the deviatoric section of the yield surface; and Z
is the friction hardening factor, which varies with the work-
hardening parameter k. The model constants , and r
c
are
computed as follows
31

b
r
1
2b
r
1
,
1 r
2r 1
and r
c

3
p

3
p
3
in which b
r
is the ratio between the biaxial and uniaxial
strengths.
The model assumes non-associated plastic ow rule, as follows

p

@G
@
4
in which is the plastic multiplier.
The form of the plastic potential is similar to that of the yield
surface, but with an additional parameter to control the
degree of dilatancy.
G , Z

J
2
_
A
r


3
_ _
I
1
Z f
c
Z 1
5
The rate of the work-hardening/softening parameter (k) is
dened as
_ kk X
T
_
p 6
where X is a ductility parameter that takes the form
X e

e
1
X
I 7
in which I
1
/(f
c
3 0
.
9) + 0
.
55 and X
I
0
.
0022.
k controls hardening/softening according to the function
Z Z
0

1 Z
0
a
c
e
cc1
1 e
cc2

8
in which k/k
p
, k
p
being the value of k at the peak yield
surface position, which is given by
k
p
f
c
0
:
72
c

f
c
2E
_ _
9
where
c
denotes the uniaxial compressive strain at the peak
uniaxial compressive stress. The following expressions for the
coefcients in equation (8) ensure that the peak occurs at
Z 1.
c
c1

c
c2
e
cc2
1 e
cc2
and a
c
e
cc1
1 e
cc2

10
The actual values used are a
c
0
.
960137261 5, c
c1

0
.
033918274 5, and c
c2
5.
2.2. Local plastic-damage-contact component
The local plastic-damage-contact component of the model is
developed using effective crack planes, which have three non-
zero terms of stress (s) and strain (e). These terms denote the
normal and two orthogonal shear terms respectively: that is,
s (s
1
s
2
s
3
)
T
e (e
1
e
2
e
3
)
T
It is noted that local strains, rather than relative displacements,
are used in the constitutive relationships. The strain parameters
are associated to the relative displacements by a characteristic
length and the fracture energy, following the approach of
Bazant and Oh.
13
The local stress s comprises two parts: the undamaged part and
the damaged contact part. The local stressstrain relationship
is expressed as
s D
L
1 e e
p
H
f
g g
p
_ _
D
L
1 I H
f

d
e e
p
D
L
M
x
e e
p
D
ls
e e
p
11
in which D
L
and I are the local elastic constitutive matrix and
identity matrix respectively; and e, e
p
, g and g
p
are the local
effective strain, local plastic strain, embedment strain and
plastic embedment strain respectively. In the interlock state, the
direction of the embedment g is xed by the normal to the
interlock surface, as illustrated in Fig. 1.
d
is the
transformation matrix such that
g g
p

d
e e
p 12
where

d

1
1 m
2
g
@j
int
@e
_ _
@j
int
@e
_ _
T
j
int
@
2
j
int
@e
2
_ _
in the interlock region in which
4 Engineering and Computational Mechanics 161 Issue EM1 Simulating cracks in cementitious composites Hee Jefferson
j
int
e m
g
e
1

e
2
2
e
2
3
_
or
d
I in the closed region.
The proportion of undamaged material is controlled by the
damage variable , which depends upon a local strain
parameter . The contact proportion (H
f
) represents the
proportion of damaged material on a crack plane that is in
contact, and is given by
28
H
f
1 if e
1
, e
bg 13
H
f

1
1 f
h
r
f
e
rf1
2
0
1 r
f
e
rf21
f
h
e
rf10
_ _
14
where
f
h
1 r
f

r
f2
m
hi
r
f1
m
ful
,
0

e
1
e
bg
m
hi

0
and
1

e
1
e
bg
m
ful

t
and
0
denote the strain at peak tensile strength and the end
of softening curve respectively; r
f
0
.
9, r
f1
10, r
f2
2,
c
g
3, and e
bg
1
.
1
t
; m
hi
is the ratio between e
hi
(the
opening strain at which shear reduction becomes signicant)
and
0
; and m
ful
is the ratio between e
ful
(the opening strain
beyond which no shear contact can be attained) and
0
.
The proposed model uses the following damage evolution
function, which has been modied from that used in Craft to
account for local plastic straining by the introduction of the
parameter a
p
.
f
s


t
1 a
p
t
a
p
e
c1
t
0t
_ _
e
2
k
kp
_ _
E 15
where f
s
is fracture stress, E is Youngs modulus, and c
1
is a
constant set to a xed value of 5. The term a
p
relates the
effective local plastic strain (e
p
) to the effective strain
parameter (),
a
p

e
p

t
16
The local model in one dimension is illustrated in Fig. 2.
The damage function is unchanged from that of the Craft
model, and is given by
j e,
e
1
2
1

_ _
2
_ _

1
2r
2

r
2

_ _
2
e
2
1
4r
2

e
2
2
e
2
3
_ _
_

17
in which r

and

are the strain equivalents of the relative


shear stress intercept r

and the asymptotic friction factor

respectively.
The model uses a local yield function, as shown in Fig. 3,
which has the same general form as the damage function but
which is now dened in terms of local stresses rather than
local strains
f
e
s, f
s

r
2

_ _
f
t
2
2

s
1

r
4

f
t
2
2 r

f
t


2

4

f
t
4
2

s
2
2
s
2
3
_ _
_
2
2

f
s
18

Local shear strain e e


2
2
2
3

1
Local normal strain ( ) e
1
g
m
g
Open
Closed
Interlock
g
Fig. 1. Contact surface
e
1
e a
p p t
( )
s
1
f
s
f
s

t
Fig. 2. Damage evolution and plastic-damage unloading
reloading
s
1
f
s
Local yield surface
Elastic region
f s f
e s
( , ) 0
s s
2
2
2
3

Fig. 3. Local yield surface


Engineering and Computational Mechanics 161 Issue EM1 Simulating cracks in cementitious composites Hee Jefferson 5
The dependence of the fracture stress f
s
on the damage
parameter means that the size of the local plastic yield
surface in stress space and the damage surface in strain space
are both governed by . Jefferson
25
argued that the local
damage (and yield) function for a crack interface should
initially be compatible with that for the intact material, and
should then develop in a manner consistent with meso-level
experimental data. From a detailed consideration of available
data, and of the forms of local surfaces used by other authors,
it was shown that that local yield and damage functions should
have the general form shown in Fig. 3.
The evolution of the local plastic strain rate is determined by
the non-associated ow rule
de
p
d
@ g
e
@s
19
where is the local plastic multiplier. A spherical local plastic
potential is used, which has the form
g
e
s

s
2
1
s
2
2
s
2
3
_
20
It is noted that the form of plastic potential, as illustrated in
Fig. 4, is different from that normally employed in dilatant
interface models because it relates to the contact surface, and
simulates the plastic embedment of one surface into another.
The use of a spherical plastic potential is an alternative, but
equivalent, way of achieving plastic embedment of one surface
into another from that used in Jefferson.
24, 25
3. STRESS RECOVERY ALGORITHM
In the stress recovery procedure, the new stresses are updated
from those of the previous converged state. The subscripts k
and k+1 denote the values of quantities at the beginning and
end of the return mapping algorithm respectively. represents
the change of a quantity within the stress update iteration, and
is the total change of any quantity from the last converged
state.
In order to evaluate the new stresses
k1
for the incremental
strain from the previous stresses
k
, the following
relationships need to be satised.
(a) Total constitutive relationship

k1
D
e

k1

p k1

np
j1
N
T
j
e
a j k1
_
_
_
_
21
(b) Stress transformation
s
i k1
N
i k1

k1 22
(c) Local stressstrain relationships for all PODs i
s
i k1
D
ls i k1
e
i k1
e
pi k1 23
e
ai k1
e
i k1
C
L
s
i k1 24
(d) Flow rule and plastic parameters for both global and local
plastic yield

p

@G
@
k1
25
k X
k1

T
k1
@G
@
k1
26
e
pi

i
@ g
e i
@s
ik1
27
(e) Global and local yield functions
F
k1
, k
k1 0
28
f
e
s
i k1
, k
k1
,
i k1
0
29
( f ) total-local function
F
e i k1
N
i k1

k1
D
ls i k1
e
i k1
e
pi k1
0
30
D
e
and N are the elastic constitutive matrix and stress
transformation matrix respectively. In the following equations,
the overall iteration subscripts are dropped for clarity.
The errors in both global and local plastic strains, and the
plastic parameters, are considered in the following.
Specically, the plastic strain and hardening parameter errors
are dened as follows.
R


p

@G
@
31
R
k
k X
T
@G
@
32
Plastic
embedment
Fig. 4. Concept of plastic embedment
6 Engineering and Computational Mechanics 161 Issue EM1 Simulating cracks in cementitious composites Hee Jefferson
R
epi
e
pi

i
@ g
ei
@s
i
33
from which the iterative corrections are obtained as follows

p
R


@G
@

@
2
G
@
2

@
2
G
@@k
k 34
k a
k
R
k
k
T

c
k

_ _
35
e
pi
R
epi

i
@ g
ei
@s
i

i
@
2
g
ei
@s
2
i
s
i 36
where
a
k
1 X
T
@
2
G
@@k
_ _
1
,
c
k
X
T
@G
@
and
k


@X
@

T
@G
@
X
@
2
G
@
2
X
@G
@
The global yield function consistency condition is written as
F
@F
@
T

@F
@k
k 0 37
Using equation (35) in equation (37) and rearranging gives
F f
T
k
h
k

@F
@k
a
k
R
k
0 38
where
f
k

@F
@

@F
@k
a
k
k

and h
k

@F
@k
a
k
c
k
The local yield function consistency condition is expressed as
f
ei

@ f
T
ei
@s
i
s
i

@ f
ei
@k
k
@ f
e i
@
i
@
T
i
@e
i
e
i
0 39
Substituting for s using s N, utilising equation (35) in
equation (39) and rearranging yields
f
ei
f
T
eki

@ f
ei
@k
a
k
R
k

@ f
ei
@k
a
k
c
k

@ f
ei
@
i
@
T
i
@e
i
e
i
0
40
where
f
eki
N
i
@ f
ei
@s
f i

@ f
ei
@k
a
k
k

The so-called total-local function ensures that both local and


global relationships, as well as the stress transformation in
equation (22), are simultaneously satised for multiple crack
planes. The total-local consistency condition can be written as
F
ei
N
i

@N
i
@
s

@D
ls i
@e
i
s e
i
e
pi
e
i

@D
ls i
@ e
i
e
pi
s e
i
e
pi
e
i
e
pi
_ _

@D
ls i
@k
e
i
e
pi
k 0
41
in which s denotes a matrix contraction for the components of
a third-order matrix. The third term in the above equation is
null here, as the orientation of the POD remains xed during
the stress update iterations. Then, using equations (35) and (36)
in equation (41) and rearranging gives
F
e i
D
lpi
R
epi
D
lki
a
k
R
k
N
ei
D
lt i
e
i
D
lpi
@ g
ei
@s
f i

i
D
lki
a
k
c
k
0
42
where
N
ei
N
i
D
lpi

i
@
2
g
ei
@s
2
i
N
i
D
lki
a
k
k
T

D
lt i

@D
ls i
@e
i
s e
i
e
pi

@D
ls i
@ e
i
e
pi
s e
i
e
pi
D
ls i
D
lki

@D
ls i
@k
e
i
e
pi
D
lpi

@D
ls i
@ e
i
e
pi
s e
i
e
pi
D
ls i
The overall relationship between the total stresses and strains is
given by
D
e

p

np
j1
N
T
j
e
a j
_
_
_
_
43
The initial trial stress may be dened using equation (43), and
Engineering and Computational Mechanics 161 Issue EM1 Simulating cracks in cementitious composites Hee Jefferson 7
thereafter iterations are performed to satisfy equations (21) to
(29). As the total strain does not change throughout the
iterations, the iterative change in the stress is given by
D
e

p

np
j1
N
T
j
e
a j
_
_
_
_
44
The iterative change in the inelastic strain e
a
is expressed as
e
ai
I m9
xi
e
i
m9
pi
e
pi
m9
xi
k
45
where
m9
xi
M
x i

@ H
ci
@
i
e
i
e
pi
@
T
i
@e
i

@
i
@
i
H
f i
e
i
e
pi
@
T
i
@e
i

@ H
f i
@e
gi

i
e
i
e
pi
@e
T
gi
@e
i
_ _

di
m9
pi
M
x i

@ H
f i
@ g
i

i
e
i
e
pi
@ g
T
i
@ e
i
e
pi
_ _

di
m9
ki

@ H
c i
@k
e
i
e
pi

@
i
@k
H
f i

di
e
i
e
pi
Making use of equations (34), (35), (36) and (45), equation (44)
may be rearranged as
A
e
_
R
kp

_
@G
@

@
2
G
@@k
a
k
c
k

np
j1
N
T
j
m9
kj
a
k
c
k
_

np
j1
N
T
j
I m9
xj
_ _
e
j

np
j1
N
T
j
m9
p j
@ g
e j
@s
f j

j
_
46
in which
A
e

_
I
_
@
2
G
@
2

@
2
G
@@k
a
k
k
T

_ _

np
j1
N
T
j
m9
p j

j
@
2
g
e j
@s
2
f j
N
j
m9
kj
a
k
k
T

_ ___
1
D
e
R
kp
R


@
2
G
@@k
a
k
R
k

np
j1
N
T
j
m9
p j
R
ep j
m9
kj
a
k
R
k
_ _
Substituting for using equation (46) in equations (38), (40)
and (41) gives a set of coupled equations with ,
i
and e
i
as the unknowns. These can be written in the following
compact form, in which i and j are indices from 1 to n
p
, and in
which the summation of repeated indices is implied. It is noted
that i,j does not imply differentiation with respect to j
components.
F

B
j
e
j
P
j

j 47
F
i
M
i
B
i, j
e
j
P
i, j

j 48
F
Ei
M
Ei
B
Ei, j
e
j
P
Ei, j

j 49
in which
F

F f
T
k
A
e
R
kp

@F
@k
a
k
R
k
F
i
f
ei
f
T
eki
A
e
R
kp

@ f
ei
@k
a
k
R
k
F
Ei
F
ei
N
ei
A
e
R
kp
D
lpi
R
epi
D
lki
a
k
R
k
M

f
T
k
A
e
@G
@

@
2
G
@@k
a
k
c
k

np
j1
N
T
k
m9
kk
a
k
c
k
_
_
_
_
h
k
M
i
f
T
eki
A
e
@G
@

@
2
G
@@k
a
k
c
k

np
k1
N
T
k
m9
kk
a
k
c
k
_ _

@ f
ei
@k
a
k
c
k
M
Ei
N
ei
A
e
@G
@

@
2
G
@@k
a
k
c
k

np
j1
N
T
k
m9
kk
a
k
c
k
_
_
_
_
D
lki
a
k
c
k
B
j
f
T
k
A
e
N
T
j
I m9
x j
_ _
B
i, j
f
T
eki
A
e
N
T
j
I m9
xj
_ _

@ f
ei, j
@
i, j
@
i, j
@e
i, j
@
i, j
B
Ei, j
N
e i
A
e
N
T
j
I m9
x j
_ _
D
l t i, j
@
i, j
P
j
f
T
k
A
e
N
T
j
m9
p j
@ g
e j
@s
j
8 Engineering and Computational Mechanics 161 Issue EM1 Simulating cracks in cementitious composites Hee Jefferson
P
i, j
f
T
eki
A
e
N
T
j
m9
p j
@ g
e j
@s
j
P
Ei, j
N
ei
A
e
N
T
j
m9
p j
@ g
e j
@s
j
D
lpi, j
@ g
ei, j
@s
i, j
@
i, j
where @
i,j
in the expression for B

, B
E
and P
E
is equivalent to
the Kronecker delta: that is, @
i,j
1 if i j and @
i, j
0 if i 6
j. Table 1 summarises the overall steps of the return mapping
algorithm for the case in which global plastic yield is active
and one or more PODs exist with local plastic yield.
The model was coded to allow for up to ve PODs in three
dimensions. In the two-dimensional nite-element cases this
number was generally reduced to two.
4. CONSISTENT TANGENT CONSTITUTIVE
RELATIONSHIP
The advantages of using a tangent stiffness matrix that is
consistent with the stress update algorithm are well known.
33
Recalling the total constitutive relationship in equation (21),
the differential form of the equation is given by
D
e

p

np
j1
N
T
j
e
a j

@N
T
j
@
I
s e
a j
D
I

_ _
_
_
_
_
50
in which D
I
is the secant elastic-damage constitutive matrix. In
the above equation, it is noted that the transformation matrix
is differentiated with respect to the trial stress
I
used for new
POD detection. Using equations (23) and (24), the iterative
change in e
a
is written as
e
ai
C
ltf i
s
i
M
pi
e
pi
M
ki
k
51
where
C
ltf
(m9
x
1
I)C
L
,
M
k
m9
x
1
m9
x
and M
p
m9
x
1
m9
p
The differential quantities of both global and local plastic
strains, plastic parameter and the local stress vector may be
written as

p

@G
@

@
2
G
@
2

@
2
G
@@k
k 52
k a
k
k
T

c
k

_ _
53
e
pi

i
@ g
ei
@s
i

i
@
2
g
ei
@s
2
i
s
i 54
s
i
N
i

@N
i
@
I
s
I
D
I
55
Using equations (51) to (55) in equation (50) and rearranging
gives
A
c
I
N
g
m

np
j1
N
T
j
M
p j
@ g
e j
@s
j

j
_ _
_
_
_
_
56
where
Step Description
1 Initialise 0, k 0 and
i
0
2 Compute trial stress using equation (43)
3 Evaluate F

, M

, B

, P

, F
E,
M
E
, B
E
, P
E
, F

, M

, B

and P

from equations (47) to (49) and solve for ,


i
and e
i
4 Compute using equation (46)
5 Update e
i
using e
i
e
i
+ e
i
and update
i
6 Compute
p
, k and e
pi
from equations (34), (35) and (36) respectively
7 Update plastic terms:

p

p

p
,
p

p

p
, , k k k, k k k, e
pi
e
pi
e
pi
, e
pi
e
pi
e
pi

i

i

i
8 Compute a new trial stress using
D
e
f
p

np
j1
N
T
j
[e
j
M
x j
(e
j
e
p j
)]g
9 Compute F, f
ei
, F
ei
, R

, R
k
and R
epi
from equations (28), (29), (30), (31), (32) and (33) respectively
10 Check for convergence
If F j j <
tol
, F
e
j j <
tol
, f
e
j j <
tol
, R

j j <
tol
, R
k
j j <
tol
and R
ep

<
tol
,
Exit iteration
Else
Return to Step 2*
End If
* Tolerance levels are
tol
f
t
3 10
6
and
tol

t
3 10
6
.
Table 1. Return mapping algorithm
Engineering and Computational Mechanics 161 Issue EM1 Simulating cracks in cementitious composites Hee Jefferson 9
A
c

_
I D
e
_

np
j1
N
T
j
_
C
ltf j
N
j
M
k j
a
k
k
T

M
p j

j
@
2
g
e j
@s
2
j
N
j
__

@
2
G
@
2

@
2
G
@@k
a
k
k
T

_
1
D
e
I
N
I

np
j1
_
N
T
j
_
C
ltf j
@N
j
@
I
s
I
D
I
M
p j

j
@
2
g
e j
@s
2
j
@N
j
@
I
s
I
D
I
_

@N
T
j
@
I
s e
a
D
I
_
g
m

@G
@

@
2
G
@@k
a
k
c
k

np
j1
N
T
j
M
k j
a
k
c
k
_ _
The consistency condition for the global yield surface may be
written as
@F
@
T

@F
@k
k 0 57
Substituting for and k in equation (57) using equations
(56) and (53), and rearranging, gives
f
T
k
A
c
g
m
h
k

1
3 f
T
k
A
c
I
N
f
T
k
A
c

np
j1
N
T
j
M
p j
@ g
e j
@s
j

j
_
_
_
_
58
The consistency condition for the local yield function is given
by
@ f
T
ei
@s
f i
s
i

@ f
e i
@k
k
@ f
e i
@
i
@
T
i
@e
i
e
i
0 59
Differentiating equation (23) and rearranging gives
e
i
m9
xi
1
C
L
s
f i
M
pi
e
pi
M
ki
k 60
Using equations (53), (55), (56), (58) and (60) in equation (59),
and collecting terms, yields

j

1
i, j

i

61
where

i
N
i
A
c
I
N
f
ni
@N
i
@
I
s
I
D
I
f
kei
(f
T
k
A
c
g
m
h
k
)
1
f
T
k
A
c
I
N

i, j
f
kei
(f
T
k
A
c
g
m
h
k
)
1
f
T
k
N
i
_ _
A
c
N
T
j
M
p j
@ g
e j
@s
j

@ f
ei, j
@
i, j
@
T
i, j
@e
i, j
M
pi, j
@ f
ei, j
@s
i, j
@
i, j
N
i
f
ni
N
i

@ f
ei
@k
a
k
k
T


@ f
ei
@
i
@
T
i
@e
i
M
ki
a
k
k
T

f
ni

@ f
ei
@s
i

@ f
ei
@
i
@
T
i
@e
i
m9
xi
1
C
L
M
pi

i
@
2
g
ei
@s
2
i
_ _
f
kei

@ f
ei
@k
a
k
c
k

@ f
ei
@
i
@
T
i
@e
i
M
ki
a
k
c
k
N
i
A
c
g
m
Substituting for in equation (58) using equation (61) gives

62
in which
(f
T
k
A
c
g
m
h
k
)
1
3 f
T
k
A
c
I
N
f
T
k
A
c

np
j1
N
T
j
M
p j
@ g
e j
@s
j

1
i, j

j
_ _
_
_
_
_
Using equations (61) and (62) in equation (56) gives the
consistent tangent relationship
A
c
(I
N
g
m

1
) D
ep
63
where D
ep
is the consistent tangent operator, and

np
j1
N
T
j
M
p j
@ g
e j
@s
j
5. NUMERICAL EXAMPLES
The following examples were chosen to assess the key
component of the modelthat is, the local plastic-damage-
contact componentin simulating the characteristics of cracks
in cementitious composite materials, for example crack
opening and closing, and shear contact. The rst example
involves a series of stressstrain paths, which form the basis of
a benchmark. The model was then assessed against data from
tests in which cracks were formed and then loaded in shear.
Table 2 shows the material properties used for the respective
examples.
10 Engineering and Computational Mechanics 161 Issue EM1 Simulating cracks in cementitious composites Hee Jefferson
For the examples based upon experimental datathat is,
examples 2 to 5the material parameters were based upon the
available experimental data. The crack contact parameters m
g
,
m
h
and m
ful
, as dened in section 2.2, have the same meaning
as in the original Craft model, and the guidance provided by
Jefferson
26
was followed here. It is noted, however, that there
was a degree of calibration of the material parameters once the
initial numerical response had been calculated.
All of the examples are essentially two-dimensional, but the
model has been developed and coded in three dimensions.
Specialisation to two dimensions is achieved in a conventional
manner by using only the required stressstrain components
and by using an appropriate elastic D-matrix. The examples
chosen are considered sufcient to demonstrate the key
features of the model.
The use of a consistent algorithmic approach ensures quadratic
convergence of a solution, but only if there is not a state
change within a time step. State changes include a new crack
forming at a Gauss point, a crack closing or a crack moving
from elastic to plastic regions. The fact that local plasticity has
been added to the model means that more state changes occur
during the solution process, and therefore convergence of this
model is not, overall, as good as for the original Craft model.
However, as would be expected, convergence is quadratic when
no state changes occur, and therefore the use of a consistent
algorithmic approach is still more efcient than its non-
consistent counterparts. In the single-point examples (1 to 3)
convergence tolerances of the initial fracture stressstrain 3
10
6
were used. The aim in the nite-element examples was to
use tolerances of 0
.
1% for the force and displacement norms,
but in Example 5, in particular, this did not prove possible, and
to overcome a limit point a tolerance of 1% for the
displacement norm and 5% for the force norm was adopted.
This is not regarded as good practice, and the solution to the
problem is considered to be the use of a non-local or gradient
implementation of the model.
As mentioned in section 2.2, the limiting strain in softening
depends upon the characteristic length according to the crack
band model of Bazant and Oh,
13
which applies fracture energy
regularisation to the nite element solutions.
5.1. Example 1: Single-point stressstrain path
benchmark
In this example, a precompression of 3 N/mm
2
in the yy
direction was rst applied, after which a crack was opened in
the xx direction (stage 1) to ve different opening strains:
0
.
0002 (BM1), 0
.
001 (BM2), 0
.
002 (BM3), 0
.
003 (BM4) and
0
.
007 (BM5). The crack was then loaded in shear (stage 2),
while the xx strain and yy stress were kept constant. Fig. 5
shows the tensile softening response for the same set of
properties, and also the normal opening strains for the ve
cases considered. A schematic illustrating the loading stages is
also included in Fig. 5. Figs 6 and 7 show the overall results
obtained from the simulations.
The results have shown that the model is able to describe the
characteristic behaviour of cracked cementitious materials
under normal-shear loading. The model correctly predicts the
gradual build-up of shear and compressive stresses at early
stage of loading, which is consistent with that generally
observed in experiments.
21, 34
These stresses do tend to a
plateau with increasing shear strains in the later stages of
loading, owing to the inclusion of local plasticity, as was
Stage 2
Shear
0
1
2
3
4
Strain xx
S
t
r
e
s
s
:

N
/
m
m
x
x
2
BM5
BM4 BM3
BM2
BM1
y
Stage 1
Opening
x
Precompressed
In - direction y y
0 0002 0004 0006 0008
Fig. 5. Tensile softening curve, showing opening points
(stage 1)
6
4
2
0
2
4
Strain xy
S
t
r
e
s
s
:

N
/
m
m
x
x
2
BM1 BM2 BM3
BM4 BM5
0020 0015 0010 0005 0
Fig. 6. Variation of normal stress with shear strain (stage 2)
Example 1 Example 2 Example 3 Example 4 Example 5
E: N/mm
2
30 000 35 000 30 000 40 000 30 000
0
.
2 0
.
15 0
.
15 0
.
15 0
.
18
f
t
: N/mm
2
3
.
0 3
.
2 2
.
5 3
.
2 3
.
2

0
0
.
002 0
.
06 0
.
002 0
.
0022
G
f
: N/mm 0
.
1
m
g
0
.
3 0
.
4 0
.
425 0
.
3 0
.
5
m
hi
1
.
0 2
.
0 2
.
0 1
.
0 0
.
4
m
ful
3
.
0 10
.
0 10
.
0 4
.
0 3
.
0
Table 2. Material properties
Engineering and Computational Mechanics 161 Issue EM1 Simulating cracks in cementitious composites Hee Jefferson 11
shown in the experiments of Walraven and Reinhardt.
20
This
plateau cannot be simulated with the original version of the
model. Another key experimental characteristic simulated is
that the wider the crack, the larger the shear strain needed for
both normal and shear stresses to start to increase.
5.2. Example 2: Reinhardt
The model was used to simulate the direct tension
compression cyclic tests carried out by Reinhardt
35
on notched
fracture specimens. The test specimens, as illustrated in Fig. 8,
had an effective area at the notch of 50 3 50 mm
2
. The
characteristic crack dimension was assumed to be equal to the
gauge length (35 mm), which was then used to obtain the
limiting strain parameter
0
that is, by dividing the opening
displacement at the end of the softening curve (0
.
21 mm) by
the gauge length.
Figure 9 shows the results from both experimental and
numerical tests. The inclusion of local plasticity on embedded
crack planes enables the model to capture the experimentally
observed permanent strains developed upon unloading
reloading of the specimen, which the original model was
unable to simulate. This unloadingreloading behaviour is
typical of that observed experimentallythat is, neither secant
nor elastic as predicted by the damage and plasticity models
respectively, but rather somewhere between these two
extremes. Fig. 10 shows the inuence of the parameter a
p
on
the unloadingreloading process.
5.3. Example 3: Walraven and Reinhardt
The third example used the data obtained from a pair of
normal-shear tests undertaken by Walraven and Reinhardt.
20
The test specimens, as depicted in Fig. 11, had a shear plane of
300 3 120 mm
2
, and were tested in a stiff testing frame with
external restraint bars used to control the crack opening. The
tests were carried out with specied initial crack opening
displacements. Once a crack had been formed to the required
opening, the specimens were then loaded in shear while the
normal and shear displacements were monitored.
Figures 12 and 13 show comparisons of the results with two
experimental tests with initial opening displacements of
0
.
2 mm. It may be seen that the numerical results obtained
differ a little from the experimental data. The response is an
50 mm
250 mm
Thickness 50 mm
Fig. 8. Reinhardts test specimen
200 160 120 80 40 0
200 160 120 80 40 0
8
6
4
2
0
2
4
40
Deflection: m
(a)

S
t
r
e
s
s
:

N
/
m
m
2
12
8
4
0
4
40
Displacement: m
(b)

S
t
r
e
s
s
:

N
/
m
m
2
Fig. 9. Uniaxial tension with crack opening and closing: (a)
experimental; (b) numerical
0020 0015 0010 0005
0
2
4
6
0
Strain xy
S
t
r
e
s
s
:

N
/
m
m
x
y
2
BM1 BM2 BM3
BM4 BM5
Fig. 7. Variation of shear stress with shear strain (stage 2)
70 60 50 40 30 20 10 0
15
10
5
0
5
10
Displacement: mm
S
t
r
e
s
s
:

N
/
m
m
2
a
p
01
a
p
04
Fig. 10. Inuence of parameter a
p
on crack closure
12 Engineering and Computational Mechanics 161 Issue EM1 Simulating cracks in cementitious composites Hee Jefferson
improvement over that produced with the original model, and
does represent the experimental response to a reasonable
degree of accuracy.
5.4. Example 4: Hassanzadeh
This example was taken from the normal-shear tests carried out
by Hassanzadeh.
21
The test specimen had an effective cross-
section at the notch level of 40 3 40 mm
2
, as illustrated in Fig.
14. During the tests, a tensile load was rst applied up to the
point of rst fracture, after which shear was applied according
to a parabolic relation between the displacements, u

v
p
,
where is a variable that varies in the range 0
.
4 , , 0
.
9.
Only the tests with 0
.
4 are considered here.
Figures 15 and 16 give the stressdisplacement responses, and
provide comparisons between the numerical and experimental
results. As can be seen in these gures, the model correctly
predicts the gradual increase in stresses with both normal and
shear displacements. At the early stage of loading, there is a
Thickness 120 mm
600 mm
400 mm
Fig. 11. Walraven and Reinhardts test specimen
12 10 08 06 04
5
4
3
2
1
0
02
Opening displacement: mm
N
o
r
m
a
l

s
t
r
e
s
s
:

N
/
m
m
2
Exper. 1/02/04
Exper. 1/02/14
Numerical
Fig. 12. Variation of normal stress with opening displacement
20 15 10 05
0
1
2
3
4
5
00
Shear displacement: mm
S
h
e
a
r

s
t
r
e
s
s
:

N
/
m
m
2
Exper. 1/02/04
Exper. 1/02/14
Numerical
Fig. 13. Variation of shear stress with shear displacement
60 mm
70 mm
40 mm 40 mm
unnotched
area
2
Thickness 70 mm
u
v
Fig. 14. Hassanzadehs test specimen
030 025 020 015 010 005
4
2
0
2
4
0
Normal displacement: mm
N
o
r
m
a
l

s
t
r
e
s
s
:

N
/
m
m
2
Experimental
Numerical
Fig. 15. Normal stressdisplacement response
08 06 04 02
0
1
2
3
4
5
0
Shear displacement: mm
S
h
e
a
r

s
t
r
e
s
s
:

N
/
m
m
2
Experimental
Numerical
Fig. 16. Shear stressdisplacement response
Engineering and Computational Mechanics 161 Issue EM1 Simulating cracks in cementitious composites Hee Jefferson 13
slow build-up of stresses, after which it accelerates, and then
plateaus towards the later stage. The plateaux in both the shear
and normal stress responses occur because of the inclusion of
the local plasticity. While the resulting behaviour is idealised, it
does provide a reasonable representation of the experimental
response.
5.5. Example 5: Arrea and Ingraffea
Example 5 involves a nite-element simulation of the
experimental tests undertaken by Arrea and Ingraffea
36
on a
single-notch concrete beam. Fig. 17 shows the geometry of the
test specimen and the corresponding mesh employed for the
analysis. In the analysis, the main load was applied in a patch
on the upper edge of the beam. Criselds arc-length method
was used to allow for the simulation of material softening.
Fifty steps were used in the solution, with the maximum
number of iterations per increment set to ten. During the
analysis the step size was automatically reduced twice when
convergence was not achieved in the maximum allowable
number of iterations.
Convergence problems were encountered in this analysis, as
discussed above, and therefore the number of PODs was
reduced to one per Gauss point. However, the crack direction
was not xed until the local crack strain had reached a value
of 4
t
, and this ensured that the fracture stress was not
exceeded in any direction at the end of each time step.
Figure 18 shows the variation of load with the crack mouth
sliding displacement (CMSD). As shown in this gure, the
initial response is a little less stiff than that observed in the
experiment, as also found by Rots
37
and Jefferson.
27
Figs 19
and 20 show the deformed mesh and crack plot at the nal
step of loading respectively. The numerical crack path closely
matches that from the experiment, and is a signicant
014 012 010 008 006 004 002
0
20
40
60
80
100
120
140
160
0
CMSD: mm
L
o
a
d
,
:

k
N
F
Exp. upper
Exp. lower
Numerical
Fig. 18. Loaddisplacement responses
Fig. 19. Deformed mesh plot (310)
Fig. 20. Crack plot
61 397
P
397 61
61
61
013P
156
Front elevation
A
A
Section AA
Fig. 17. Test arrangement and nite-element mesh (all dimensions in mm)
14 Engineering and Computational Mechanics 161 Issue EM1 Simulating cracks in cementitious composites Hee Jefferson
improvement on that given by Jefferson.
27
The major principal
strain distribution is depicted in Fig. 21.
6. CONCLUSIONS
The proposed plastic-damage-contact model, with embedded
local plasticity, has been developed to capture the local plastic
deformations associated with the opening/closing and shearing
of cracks in cementitious composite materials. The model
employs an implicit stress recovery algorithm that fully couples
plastic ow to directional damage, and ensures that local and
global damage constitutive relationships and the plastic yield
condition are all simultaneously satised. A tangent
constitutive matrix has also been derived that is consistent
with the return mapping algorithm.
From the numerical examples it has been shown that the model
simulates well the chief characteristics observed in experiments
in which cracks are rst opened and then loaded in shear. The
advantage of using embedded planes with local plasticity is
apparent: for cyclic tests, permanent strains are captured upon
unloading, and these gradually increase as the unloading
reloading process continues; for cracks loaded in shear, the
local yield surface limits the stresses that develop between the
opposing faces of cracks in a realistic manner.
The inclusion of local plasticity considerably complicates the
formulation relative to that of the original Craft model;
however, the new model improves the accuracy of simulations
and, most importantly, simulates directly an important physical
mechanism that is known to strongly inuence the post-
cracked behaviour of cementitious composites.
REFERENCES
1. ESTE G. and WILLAM K. Fracture energy formulation for
inelastic behavior of plain concrete. Journal of Engineering
Mechanics, 1994, 120, No. 9, 19832011.
2. GRASSL P., LUNDGREN K. and GYLLTOFT K. Concrete in
compression: a plasticity theory with a novel hardening
law. International Journal of Solids and Structures, 2002,
39, No. 20, 52055223.
3. PARK H. and KIM J. Y. Plasticity model using multiple
failure criteria for concrete in compression. International
Journal of Solids and Structures, 2005, 42, No. 8, 2303
2322.
4. LABADI Y. and HANNACHI N. E. Numerical simulation of
brittle damage in concrete specimens. Strength of
Materials, 2005, 37, No. 3, 268281.
5. TAO X. and PHILLIPS D. V. A simplied isotropic damage
model for concrete under bi-axial stress states. Cement and
Concrete Composites, 2005, 27, No. 6, 716726.
6. HANSEN E., WILLAM K. and CAROL I. A two-surface
anisotropic damage/plasticity model for plain concrete.
Proceedings of the 4th International Conference on Fracture
Mechanics of Concrete Structures, Cachan, 2001, pp. 549
556.
7. SALARI M. R., SAEB S., WILLAM K. J., PATCHET S. J. and
CARRASCO R. C. A coupled elastoplastic damage model for
geomaterials. Computer Methods in Applied Mechanical
Engineering, 2004, 193, No. 2729, 26252643.
8. BIELSKI J., SKRZYPEK J. J. and KUNA-CISKAL H.
Implementation of a model of coupled elastic-plastic
unilateral damage material to nite element code.
International Journal of Damage Mechanics, 2006, 15,
No. 1, 540.
9. WU J. Y., LI J. and FARIA R. An energy release rate-based
plastic-damage model for concrete. International Journal of
Solids and Structures, 2006, 43, No. 34, 583612.
10. OZ

BOLT J., LI Y.-J. and KOZ

AR I. Micro-plane model for


concrete with relaxed kinematic constraint. International
Journal of Solids and Structures, 2001, 38, No. 3, 2683
2711.
11. BAZ

ANT Z. P. and CANER F. C. Microplane model M5 with


kinematic and static constraints for concrete fracture and
anelasticity. I: Theory. Journal of Engineering Mechanics,
2005, 131, No. 1, 3140.
12. LEUKART M. and RAMM E. Identication and interpretation
of microplane material laws. Journal of Engineering
Mechanics, 2006, 132, No. 3, 295305.
13. BAZ

ANT Z. P. and OH B. H. Crack band theory for fracture


of concrete. Materials and Structures, 1983, 16, No. 3,
155177.
14. DI LUZIO G. and BAZ

ANT Z. P. Spectral analysis of


Fig. 21. Major principal strain distribution
Engineering and Computational Mechanics 161 Issue EM1 Simulating cracks in cementitious composites Hee Jefferson 15
localization in nonlocal and over-nonlocal materials with
softening plasticity or damage. International Journal of
Solids and Structures, 2005, 42, No. 23, 60716100.
15. RODRIGUEZ-FERRAR A., MORATA I. and HUERTA A. A new
damage model based on non-local displacements.
International Journal for Numerical and Analytical
Methods in Geomechanics, 2005, 29, No. 5, 473494.
16. GRASSL P. and JIRA

SEK M. Plastic model with non-local


damage applied to concrete. International Journal for
Numerical and Analytical Methods in Geomechanics, 2006,
30, No. 1, 7190.
17. ASKES H. and METRIKINE A. V. Higher-order continua
derived from discrete media: continualisation aspects and
boundary conditions. International Journal of Solids and
Structures, 2005, 42, No. 1, 187202.
18. ACIASCE Committee 426. The shear strength of
reinforced concrete members. Journal of the Structural
Division, 1973, 99, No. 6, 10911187.
19. PAULAY T. and LOEBER P. J. Shear transfer by aggregate
interlock. In Shear in Reinforced Concrete. ACI Publication
SP-42, American Concrete Institute, 1974, pp. 115.
20. WALRAVEN J. C. and REINHARDT H. W. Theory and
experiments on the mechanical behaviour of cracks in
plain and reinforced concrete subjected to shear loading.
Heron, 1981, 26, No. 1A.
21. HASSANZADEH M. Behaviour of Fracture Process Zones in
Concrete Inuenced by Simultaneously Applied Normal and
Shear Displacements. PhD thesis, Lund Institute of
Technology, Sweden, 1991.
22. VAN MIER J. G. M. Fracture Process of Concrete. CRC Press,
Boca Raton, Florida, 1997.
23. CAROL I., PRAT P. C. and LO

PEZ C. M. A normal-shear
cracking model: application to discrete crack analysis.
Journal of Engineering Mechanics, ASCE, 1997, 123, No. 8,
765773.
24. JEFFERSON A. D. Constitutive modelling of aggregate
interlock in concrete. International Journal for Numerical
and Analytical Methods in Geomechanics, 2002, 26, No. 5,
515535.
25. JEFFERSON A. D. Tripartite cohesive crack model. Journal of
Engineering Mechanics, ASCE, 2002, 128, No. 6, 644653.
26. JEFFERSON A. D. Craft, a plastic-damage-contact model for
concrete. II. Model implementation with implicit return
mapping algorithm and consistent tangent matrix.
International Journal of Solids and Structures, 2003, 40,
No. 22, 60016002.
27. JEFFERSON A. D. Craft, a plastic-damage-contact model for
concrete. I. Model theory and thermodynamic
considerations. International Journal of Solids and
Structures, 2003, 40, No. 22, 59735999.
28. JEFFERSON A. D., BARR B. I. G., BENNETT T. and HEE S. C.
Three dimensional nite element simulations of fracture
tests using the Craft concrete model. Computers and
Concrete, 2004, 1, No. 3, 261284.
29. HEE S. C. Numerical Simulation of Cracking Using
Embedded Surfaces in a Three Dimensional Constitutive
Model for Concrete. PhD thesis, Cardiff University, UK,
2006.
30. LUSAS. User Reference Manual. FEA Ltd, Kingston upon
Thames 2006.
31. LUBLINER J., OLIVER J., OLLER S. and ONATE E. A plastic-
damage model for concrete. International Journal of Solids
and Structures, 1989, 25, No. 3, 299326.
32. WILLIAM K. J. and WARNKE E. P. Constitutive model for
triaxial behaviour of concrete. Proceedings of International
Association for Bridges and Structural Engineering, Seminar
on Concrete Structures Subjected to Triaxial Stresses,
Bergamo, Italy, 1974. Proceedings of the International
Association for Bridge and Structural Engineering (IABSE),
Report 19, section 111, Zurich, pp. 130.
33. SIMO J. C. and TAYLOR R. L. A return mapping algorithm for
plane stress elastoplasticity. International Journal for
Numerical Methods in Engineering, 1986, 22, No. 3, 649
670.
34. NOORU-MOHAMED M. B. Mixed-mode Fracture of Concrete:
An Experimental Approach. PhD thesis, Delft University of
Technology, Delft, the Netherlands, 1992.
35. REINHARDT H. W. Fracture mechanics of an elastic softening
material like concrete. Heron, 1984, 29, No. 2, 142.
36. ARREA M. and INGRAFFEA A. R. Mixed-mode Crack
Propagation in Mortar and Concrete. Department of
Structural Engineering, Cornell University, USA, 1982,
Report No. 8113.
37. ROTS J. G. Computational Modelling of Concrete Fracture.
PhD thesis, Delft University, the Netherlands, 1988.
38. WU J. Y., LI J. and FARIA R. An energy release rate-based
plastic-damage model for concrete. International Journal of
Solids and Structures, 2006, 43, No. 34, 583612.
What do you think?
To comment on this paper, please email up to 500 words to the editor at journals@ice.org.uk
Proceedings journals rely entirely on contributions sent in by civil engineers and related professionals, academics and students. Papers
should be 20005000 words long, with adequate illustrations and references. Please visit www.thomastelford.com/journals for author
guidelines and further details.
16 Engineering and Computational Mechanics 161 Issue EM1 Simulating cracks in cementitious composites Hee Jefferson

Potrebbero piacerti anche