Sie sind auf Seite 1von 22

A FINITE ELEMENT MODEL FOR ANALYZING THE DYNAMIC

CRACKING RESPONSE OF CONCRETE


Jeffrey W. Simons, Tarabay H. Antoun, and Donald R. Curran
SRI International, Menlo Park, CA 94025
Presented at 8th International Symposium on Interaction of the Effects of Munitions with
Structures, McClean, Virginia, April 22-25, 1997
SUMMARY
The physical basis for the softening of concrete (often with associated localization of strain and subsequent failure) is
the evolution of microcracking damage. A physically-based constitutive model is being developed that describes the
nucleation and growth of tensile and shear cracks in concrete. The constitutive relations are incorporated into a
multiplane cracking model that has been implemented in the finite element code DYNA2D. Microcracks nucleate
and grow on several prespecified planes in each element. Local stresses are introduced to account for local stress
variations caused by the aggregates. Time-dependence arises naturally from the finite propagation velocities of the
microcracks and helps ensure well-posedness and stability of the numerical solution. The model has been used to
simulate laboratory experiments including uniaxial tension and uniaxial compression at several different strain rates
ranging from static to impact loading conditions, and a spherical explosive test in marble. The results of the
simulations show promising results, including softening curves that result from the cracking process, the increase in
strength at high rates of loading, and calculated cracking patterns similar to those observed in experiments.
1. OBJECTIVE AND APPROACH
The main objective of this work, sponsored by the Defense Special Weapons Agency
*
, is to develop a physics-based
constitutive model for analyzing the deformation and failure of concrete including damage and failure due to
cracking. It is important that the model be thermomechanically consistent, numerically stable, and easy to
implement in a finite element code. We plan to demonstrate through numerical examples that the model reproduces
such macroscopic response features as softening, load-induced anisotropy, and rate effects. Our approach is to
explicitly model the nucleation and growth of microcracks in concrete. By modeling the damage processes directly,
the degradation of concrete properties and softening result from the accumulation of damage.
2. INTRODUCTION AND BACKGROUND
2.1 NEED
Much research has been conducted to develop constitutive models for concrete capable of simulating the static and
dynamic responses of concrete structures beyond peak load and into the softening region. A good review of some of
these models is given by Chen et al. (1991) in which descriptions are given for elasticity based [e.g., Saenz (1964),
Popovics (1973), Carriena and Chu (1985), Tsai (1988)], plasticity-based [Ohtani and Chen (1987), Chen and Chen

*
Review of this material does not imply Dept. of Defense endorsement of factual accuracy or opinion.
(1975), and Murray (1979)], elasto-plastic and fracturing [Maekawa and Okamura (1983), Dougill (1976), Bazant
and Kim (1979), and Han and Chen (1986)], and block models [Cundall (1971) and Kawai (1977)]. Although the
models described simulate much of the highly nonlinear response of concrete including softening, Chen et al.
conclude that a need still exists for realistic models that are rational, reliable and practical. Particular features of
such models include the capability to capture (1) the post-elastic non-linear deformation of brittle cracked concrete,
(2) the transition between brittle and ductile regimes; (3) the response of a concrete composite system in a non-
uniform and non-homogeneous stress field; and (4) the post-peak stress-deformation response of fractured concrete,
including localized deformation, strain softening, crack healing and ductile fracture.
2.2 FAILURE PROCESS IN CONCRETE
One recognized problem in the modeling of concrete beyond peak load is that the softening response is not a
continuum property of the material but the result of a structural response. As described by Read and Hegemier
(1984), strain softening observed in laboratory tests is the result of internal cracking of material caused by imperfect
boundary conditions between the specimen and the loading plates. The resulting strain distribution in a failed
specimen is often very nonuniform and contains regions of high local strains. The localization of deformation during
cracking in concrete has been measured experimentally under both tension (Gopalaratnam and Shah 1985) and
compression (Shah and Sankar 1987).
For concrete, microcracking has long been known to be the dominant source of nonlinear inelastic behavior. The
cracking response of concrete under uniaxial compression is well established (Mindess and Young 1981). Initially,
microcracks are present in concrete as a result of shrinkage, thermal movement, discontinuities at aggregate
interfaces, and voids due to incomplete compaction (Chen et al., 1991). At loads of about 30% of the peak load,
bond cracks (or interfacial cracks) begin to propagate, primarily under shear loading (Wittmann 1979). At loads of
about 50% of peak, cracks begin to grow into the matrix in the direction of the applied load. At about 75% of peak
load, matrix cracks increase rapidly in size and number and begin to coalesce. Ultimately the specimen fails by
splitting along planes parallel to the direction of the applied load. Although the response mechanism causing
softening is well established, the shape of the softening curves are strongly dependent on the compliance of the
testing machine (Kotsovos 1983).
2.3 PREVIOUS MODELS
Several classes of fracture models have been developed to analyze the response of material during microcrack
damage. Micromechanical models have been developed in which a brittle material like rock and concrete is modeled
as a homogeneous material containing a known distribution of pre-existing cracks (Kachanov 1958, Horii and
Nemat-Nasser 1985, Fanella and Kracjinovic 1988). Cracks then propagate in response to the stress state. Crack
activation is monitored using either a stress intensity factor formulation (Seaman et al. 1985) or a generalized
Griffith criterion (Margolin 1984). Espinoza (1995) has developed a micromechanically-based multiplane cracking
model to analyze dynamic impact experiments in ceramics. Margolin (1983) describes a model for an elastic
material with distributions of cracks in arbitrary orientations. A similar model has been presented by Krajcinovic
and Fonseka (1981) for application to concrete fracture. Other microplane models based on the slip theory of
plasticity have been developed and extended to model many complex response mechanisms of concrete including
cyclic loading (Bazant 1984, Bazant and Oh 1985 and 1983).
In smeared crack models, cracks are not treated explicitly but are used to generate prescribed stress-strain curves.
Such models have been developed by Rashid (1968), Hillerborg et al. (1976), Gopalaratnam and Shah (1985), Bazant
and Oh (1983) and Gran (1985). A useful property of the cracking models is damage-induced anisotropy in which
the stiffness properties of the concrete are reduced in the direction normal to cracking, a well-documented
characteristic of brittle material behavior. The dependence of crack propagation on the orientation of the applied
load introduces a directional property into the model that changes an initially isotropic response to anisotropic.
2.4 LOCAL STRESSES
Microcrack propagation in concrete is driven by local stresses which, in general, are different from the remotely
applied stresses. The microstructure of concrete includes aggregates, cement paste, and voids. Local stresses arise
as a result of the heterogeneous nature of concrete and magnitudes of local stresses are related to the mismatch in
material properties between the aggregate particles and the mortar matrix. In particular, the stresses around
aggregates will be higher than the average continuum stresses whereas stresses around voids will be lower. Local
strains up to four times the average strains have been recorded experimentally in concrete specimens loaded under
uniaxial stress conditions (Dantu, 1958). An indication of the local nature of cracking stresses is that the mode of
failure for concrete specimens loaded under unconfined compression is cracks aligned in a direction parallel to the
loading axis despite the fact that the applied stress is zero normal to the crack surface. Accepting that cracking is a
stress-driven process suggests that local tensile stresses are present that drive the cracking process.
Costin and Stone (1987) developed a model for microfractured brittle rock that includes local stress variations by
assuming that local tensile stresses are proportional to the deviatoric stress normal to the crack face and depend on
the average distance between cracks, and the size of the local tensile region around the crack tip. These local stress
variations are added to the continuum stresses and the combined stress is used as the driving stress for crack
propagation. A similar approach is taken in the present formulation to account for local stress fluctuations.
2.5 STABILITY AND UNIQUENESS
Softening models for concrete can have numerical problems associated with the stability and uniqueness of the
solution and may show severe mesh dependence of results particularly if softening is treated as a rate-independent
property of the material, as discussed by Sandler and Wright (1984), Belytschko and Bazant (1985) and Read and
Hegemier (1984). Theoretically, the reason for mesh dependency is loss of ellipticity in the governing equations
which causes strains to localize in regions of vanishing volume (Bicanic et al. 1991). Various methods have been
developed to help overcome this numerical sensitivity including: basing damage on nonlocal measures of strain
(Bazant, Belytchko and Chang, 1984), including rate effects (Needleman 1988; Sluys and de Borst 1990), or
specifying minimum mesh dimensions (Bazant and Oh, 1983).
In a broader sense, the framework presented by Coleman and Gurtin (1967) sets forth conditions that assure
uniqueness and stability for thermodynamic processes and can be straightforwardly applied to analyzing the
mechanical response of concrete. Within the Coleman and Gurtin irreversible thermodynamics framework, the
stress-strain relationship is derived from the Helmholtz free energy - a scalar valued function of the invariants of the
strain tensor and the vector
i
containing the set of internal state variables that represent the internal rearrangements
in the material during deformation. The energy function has the general form

ij
,
i
( ) (1)
where is the Helmholtz free energy density,
ij
are the Cartesian components of the Green-St. Venant strain
tensor, and in the present model, the internal state variables are chosen to describe the cracking response. The stress-
strain relationship is derived from the Helmholtz free energy function as follows:
ij
ij

0
(2)
where
0
is the density and
ij
are the components of the second Piola Kirchhoff stress tensor.
A critical element of the Coleman and Gurtin approach is that the formulation is inherently time-dependent because
the dependence of the internal state variables on the strain field is specified using evolution equations of the form
( )
i ij i
f , & (3)
This is an important consideration in modeling the rate-dependent cracking response of concrete subjected to
dynamic loading.
The final element of the formulation is that the Helmholtz free energy function, and the state variable evolution
equations must satisfy the dissipation inequality
0

i
i

& (4)
which insures that the entropy production rate during irreversible thermodynamic processes is non-negative.
The concrete model presented in this paper can be cast in the form of Eqs. (1-3) and can be shown to unconditionally
satisfy the dissipation inequality represented by Eq. (4). The model therefore satisfies the conditions for uniqueness
and stability derived by Coleman and Gurtin (1967).
3. DESCRIPTION OF THE MODEL
In the model, concrete is assumed to be an elastic-cracking material. The mechanical behavior of concrete is
described using a multiplane cracking model similar in concept to some of the microplane models described above,
but is more explicit in describing the cracking process. Cracking can occur on a discrete number of predetermined
planes that may be randomly or preferentially oriented, depending on the nature and distribution of pre-existing flaws
in the material. The number of cracking planes is arbitrary, limited only by practical considerations related to the
speed and efficiency with which calculations can be performed, and physical considerations related to the ability of
the model to realistically capture the cracking patterns in the material for arbitrary loading conditions.
We assume that the cracks are non-interacting, penny-shaped, and uniformly distributed. On an arbitrarily chosen
plane with normal

r
n , the microcrack distribution can therefore be characterized by a microcrack density parameter,
N, and a characteristic microcrack radius, a . While the crack orientation,

r
n , and the crack density, N, are fixed
parameters, the crack size, a evolves as a function of applied loading in accordance with the laws of dynamic
fracture mechanics. Cracks on different planes evolve independently, but their combined effect on the overall
response of the material is accounted for within the framework of a compliance-based formulation. In describing
this formulation, we begin by presenting the model equations for a single cracking plane; we then generalize the
formulation for an arbitrary number of planes.
3.1 SINGLE CRACK PLANE LOADED IN TENSION
Consider an elastic body loaded in uniaxial tension with a single set of tension cracks oriented normal to the load.
We assume that the total strain can be decomposed into an elastic component and a cracking component. In a local
coordinate system, along an axis normal to the cracking plane we have,
e
t
e
e
+ e
c
(5)
where e
t
is the total strain, e
e
is the elastic strain and e
c
is the cracking strain. The cracking strain for a set of
noninteracting tensile cracks can be related to the remotely applied stress, s , as follows, (Sneddon and Lowengrub
1969, Kachanov 1993)
e
ct

16 1
2
( )
3E
Na
3
s s > 0

(6)
where E is the Youngs modulus and is the Poissons ratio for the elastic material and e
ct
identifies cracking
strain for tensile cracks. Here, we can define the compliance of the set of cracks in tension, C
ct
as,
C
ct

16 1
2
( )
3E
Na
3
(7)
From Hooks law, the elastic response of the material can be expressed as,
e
e

s
E
C
e
s (8)
where C
e
is the elastic compliance. We can combine Eqs. (5-8) to obtain,
e
t
C
e
+ C
ct
( )s (9)
Expressing Eq. (9) in rate form, and realizing that the cracking compliance is a function of crack radius (i.e., Eq. (7))
gives,
( ) s C s C C e
ct ct e t
&
& & + + (10)
Combining this with Eq. (7) and rearranging gives,
( )

,
_

s
a
a C
e C C s
ct
t ct e
&
& &
3 1
(11)
This equation is very instructive because it reveals the mechanisms for strain softening and rate dependence in the
model. The second strain rate term, ( )s a a C
ct
& 3 , is the cracking strain rate associated with an increase in crack
radii at a given stress level. Because the crack growth rate is always positive, the effect of this term is to reduce the
stress whenever the crack growth rate is non-zero. If this term is greater than the applied strain rate,
t
e& , the stress
rate will become negative for a positive applied strain rate, i.e., softening will occur. Thus, the mechanism for
softening is the transfer of elastic strain into cracking strain.
Rate dependence is a result of the competition between the rate of applied strain,
t
e& , and the crack growth rate, a& .
At low loading rates, the increase in stress due to the applied loading will be small compared to the stress drop due to
increased cracking. At high rates of loading, the overall stress will continue to rise after cracking begins as long as
the applied strain rate is greater than the rate of cracking strain due to crack growth. Thus, at high rates the concrete
will appear stronger and more ductile than at low rates.
3.2 SINGLE CRACK PLANE UNDER GENERAL LOADING
The above relations apply to the case of crack opening under tension, where the only cracking strain component
occurs in the direction normal to the crack surface. Similar relations have been developed for shear loading of
tension cracks and for compression and shear loading of shear cracks. The shearing strain due to shear loading of an
open tension crack, e
cs
, is related to the applied shear stress, by an equations similar to Eq. (6),
e
cs

16 1
2
( )
3E
Na
3
(12)
For shear cracks, assuming a frictional relationship on the crack surface, the applied shear stress in Eq. (12) is
replaced by the effective shear stress,
eff
, given by

eff
+ s

s < 0 (13)
where is the coefficient of friction.
3.3 CRACK EVOLUTION
In accordance with experimental observations (e.g., Mindess and Young, 1981), concrete is viewed as an initially
elastic and isotropic material permeated by an array of pre-existing defects. Experimental evidence also suggests
that under the action of applied loads, the pre-existing defects grow into crack-like features that primarily populate
the aggregate-matrix interfaces (e.g. Hsu et al., 1963 and Gopalaratnam and Shah, 1985). In our model, this
interfacial crack formation phase of the damage evolution process is termed the crack nucleation phase. Penny-
shaped cracks with a size proportional to the maximum aggregate size are assumed to nucleate once the applied
stress reaches a threshold value. Cracks nucleate under both shear and tension. A tension crack nucleates if the
applied stress normal to the crack surface reaches
th
nuc
, the threshold stress for tensile crack nucleation. Likewise,
a shear crack nucleates if the shear stress acting on the crack surface reaches
th
nuc
, the threshold stress for shear
crack nucleation. Both
th
nuc
and
th
nuc
are model parameters determined based on experimental data.
Increasing the applied stress beyond the crack nucleation threshold causes crack growth. Cracks can grow either in
mode I (tension) or mode II (shear). Two different crack size variables are used to distinguish between the two
different modes of crack propagation. Thus a
t
designates the radius of a tension crack and a
s
designates the radius
of a shear crack.
Crack activation is monitored using a fracture-mechanics-based stress intensity factor approach. With this approach,
a tension crack with radius a
t
is activated if the tensile stress normal to the crack plane reaches a threshold value
given by the relation

th
grow

4a
t
K
Ic
(14)
where
th
grow
is the threshold tensile stress for crack growth and K
Ic
is the critical stress intensity factor (or fracture
toughness) for crack growth under mode I. Similarly, a shear crack with radius a
s
is activated if the shear stress
tangent to the crack plane reaches a threshold value given by the relation

th
grow

4a
s
K
IIc
(15)
where
th
grow
is the threshold shear stress for crack growth and K
IIc
is the critical stress intensity factor (or fracture
toughness) for crack growth under mode II.
Once a crack is activated, its propagation is governed by a stress- and crack-size-dependent viscous growth law
derived based on the principles of dynamic fracture mechanics. Freund (1973) analyzed the response of a semi-
infinite crack to an incident tensile stress wave and found that the instantaneous value of the mode I stress intensity
factor for a moving crack is equal to the product of the stress intensity factor for an equivalent stationary crack times
a universal function of the crack tip velocity. This is represented by the relation

,
_

ten
t
s
I
d
I
c
a
g
K
K
max
&
(16)
where K
I
s
and K
I
d
are the static and dynamic mode I stress intensity factors, respectively, and c
max
ten
is the
maximum crack velocity in mode I (equal to a fraction of the Rayleigh wave velocity). Tsai (1973) derived the
universal function for a penny-shaped crack moving at a constant velocity in an infinite elastic solid. This function is
shown in Figure 1 along with the second order polynomial approximation used by Gran and Seaman (1986). A more
useful form of Eq. (16), and of the functions plotted in Figure 1 can be obtained by inverting the function g and
realizing that the stress intensity factor for a propagating crack should always be equal to the fracture toughness.
This yields,

,
_

s
Ic
d
Ic ten
t
K
K
g c a
1
max
& (17)
Using the stress intensity factor equation for a penny-shaped crack, one may express Eq. (17) in terms of the driving
stress for tensile cracking,
dr
, and the threshold stress for crack growth,
th
grow
as follows,

,
_

dr
grow
th ten
t
f c a

max
& (18)
The crack growth law as expressed by Eq. (18) is used in the present model. Numerically, this growth law is
implemented using the polynomial approximation of Gran and Seaman (1986).
An analytic solution similar to the one expressed by Eq. (18) is not yet available for cracking in mode II. However,
existing solutions for the crack velocity function for semi-infinite cracks propagating in mode I and mode III
(antiplane shear) exhibit the same trend of decreasing stress intensity factor with increasing crack speed (Sih and
Chen, 1977). Extrapolating this observation to mode II, it is assumed that, to a good approximation, the same
functional form used to account for the rate dependence of mode I crack growth under tension can be used to account
for the rate dependence of mode II crack growth under shear. Mathematically, the crack growth equation for shear
cracks can therefore be expressed as follows

,
_

grow
th
dr shear
s
f c a

max
& (19)
where c
max
shear
is the maximum mode II crack velocity (equal to a fraction of the shear wave velocity) and
dr
is the
driving stress for shear cracking.
The driving stresses for tensile and shear cracking,
dr
and
dr
, eluded to in Eqs. (18) and (19) are local stresses
which, in general, are different from the remotely applied continuum stresses. These local stresses arise as a result of
inhomogeneities in the material and mismatches in material properties between the aggregates and the cement paste.
The effect of local stresses on the cracking response is included in our present formulation in a manner similar to that
used by Costin (1987) in his model for rock fracture. To graphically illustrate the local stress concept, let us begin by
examining the Mohr diagram of Figure 2. The large circle represents the continuum stress state, and the four smaller
circles represent the local stresses on each of four cracking planes labeled 1 through 4 in the figure. Mathematically,
the local stress is derived from the continuum stress using the simple relation

loc

ij

ij
(20)
where
loc
is the local stress and is a material constant that reflects the effect of microstructure on the local
stress field. As evident from Eq. (20) the local stress at a point (or in a finite element) has the same magnitude on all
the cracking planes. This is also illustrated in Figure 2 where the Mohr circles representing the local stresses on the
various cracking planes all have the same size.
Cracking on a plane occurs if the driving stress on the plane exceeds the threshold for either tensile or shear cracking.
The thresholds are shown as straight lines in Figure 2 where it is also shown that the local stress field exceeds the
1.0
0.8
0.6
0.4
0.2
0.0
K
I
d
/
K
I
s
1.0 0.8 0.6 0.4 0.2 0.0
CRACK VELOCITY/MAXIMUM CRACK VELOCITY
Exact Solution (Tsai, 1973)
Polynomial fit (Gran and Seaman, 1986)
Figure 1.Dependence of the ratio of the dynamic to static stress intensity factor on crack tip velocity
for a propagating mode I penny-shaped crack.
thresholds (i.e., the cracking criteria are satisfied) on two of the four cracking planes. The local stress state on
cracking plane No. 3 satisfies the criterion for crack growth under tension; whereas the local stress state on cracking
plane No. 2 satisfies the criterion for crack growth under shear. On these two cracking planes, cracks will grow
according to the crack growth laws (Eqs. 18 and 19) described earlier. Crack growth modifies both the stress state
and the cracking thresholds, and the cracks will continue to propagate until the local stress states on all the cracking
planes cease to satisfy the crack growth criteria.
Mathematically, the cracking process described above is implemented as follows. The driving stress for tension
cracks is calculated as the algebraic sum of the applied stress normal to the cracking plane,
n
, and the local stress,

loc
,

dr

n
+
loc
(21)
In Figure 2, the resulting stress state is represented by Point A on cracking plane No. 3.
The driving stress for shear cracks is calculated as the sum of the applied stress tangent to the cracking plane,
n
,
and the projection of the local stress,
loc
, in a direction normal to the shear cracking threshold surface. When the
normal component of the stress is compressive, the frictional resistance of the material on the cracking plane is also
taken into account to yield the following relation for the driving stress:

dr

n
+
loc
cos ( )+
n

loc
sin ( ) for
n
< 0

n
+
loc
cos
( )
for
n
0

'



(22)
where the friction angle, , and the friction coefficient, , are related by the standard relation,
tan (23)
In Figure 2, the stress state computed based on Eq. (22) is represented by Point B on cracking plane No. 2.

Mohrs circle representing


applied stress state
Mohrs circle representing
local stress fl uctuations
1
2
3
4
1



1, 2, 3 and 4 represent the stress
states on the desi gnated planes.

Tensi le
Threshold
Shear Threshold
(Coulomb fri cti on)
A
B
Figure 2. Mohr diagram illustrating the local stress concept and the
methodology used to incorporate local stresses into the
crack growth equations.
3.4 ASSEMBLING THE GLOBAL EQUATIONS
The development of the equations in three dimensions follows along similar lines to the one-dimensional
development. In global coordinates, the total strain vector is decomposed into the elastic and the cracking strain,

t

e
+
c
(24)
where
t
is the total strain,
e
is the elastic strain and
c
is the cracking strain vector. In the multiplane model,
cracking occurs on a set of planes of prespecified orientations. In the two-dimensional case, we specify a set of five
discrete planes oriented as shown in Figure 3. The relative orientation of one plane relative to another within the set
of five planes is fixed; however, the set of planes is assigned an initial global orientation, , that can be either a
random number between +22.5 and - 22.5, or any prespecified angle. This feature makes it possible to represent
randomly oriented as well as preferentially oriented cracking patterns. On each plane, k, the stresses acting across
the planes are related to the global stresses through the transformation matrix Q
k
,
Q
k
s
k
(25)
where,
Q
k

cos
2

k
sin
k
cos
k
sin
2

k
sin
k
cos
k
2sin
k
cos
k
cos
2

k
sin
2

k




1
]
1
1
(26)
and
k
is the angle between the normal to the crack and
the global x-axis. The cracking strains on the planes are
assumed to be independent of one another, and they are
all added together as they are transformed to the global
coordinate system.

c
Q
k
e
k
c
k1
nplanes

(27)
4
5
1
2
3
Figure 3. Orientation of the cracking
planes in two dimensions
The cracking compliance from the planes can also be assembled into the global coordinate system as,
C
c
Q
k
T
C
k
c
Q
k
k1
nplanes

(28)
Combining Eqs. (24) and (29) and using the elastic compliance matrix, C
e
, gives,

t
C
e
+ C
c
(29)
The rate form of this equation is
s s s e
c c t
C C C
e
&
& & & + + (30)
where
c
C
&
can be assembled from the derivatives of the cracking compliance matrices on the planes and are explicit
functions of the crack tip velocities on the planes. We can integrate this equation with respect to time to make it
suitable for implementation into a finite element code. The equations are solved numerically for the stress increment
using the semi-implicit algorithm described by Espinoza (1993) in which the crack tip velocity is updated explicitly
and held constant during the time step, and the stress increment is solved for implicitly. The solution for the stress
increment is then given by,
( ) { } ( )
0
1
t
c t c c e
C C C C
& &
+ +

(31)
where t is the time step.
4. SIMULATIONS OF LABORATORY EXPERIMENTS
Standard materials properties tests, including uniaxial tension, uniaxial compression, and confined compression tests,
are commonly used to measure the key mechanical properties of concrete. The response mechanisms for these
standard experiments are fairly well understood, and the cracking patterns and associated failure modes are well
documented. To illustrate the ability of our newly developed model to capture the important features of the response
of concrete, we performed calculations of the standard tests mentioned.
4.1 UNIAXIAL TENSION TEST
We modeled uniaxial tension tests performed by Gopalaratnem and Shah (1985). In these experiments the concrete
specimen was a notched rectangular prism 30.5 cm (12 in.) tall, 7.6 cm (3 in) wide and 1.9 cm (3/4 in) thick. The
loading was applied by shear grips on either end of the specimen. The specimen was notched to predetermine the
location of failure. The notch depth resulted in a 5.1 cm (2 in.) effective loaded dimension across the width. Local
strains were measured at several locations on the specimen with strain gages over 10 mm (0.25 in.) gage lengths and
overall specimen displacement was measured with displacement transducers over a 8.2 cm (3.25 in.) gage length.
The specimens were tested quasistatically to failure.
The finite element model for the tension specimen is shown in Figure 4a. Material properties for the model are listed
in Table 1. The elastic properties were chosen to match the initial response of the specimen. The simulated load-
displacement curve for the specimen loaded in uniaxial tension is shown in Figure 4b along with records from two
experiments. The applied stress as reported is the applied load averaged over the notched cross section. The
displacement is that measured over a 8.2 cm gage length. The specimen reaches a peak average stress of about 3.6
MPa, at a displacement of about 400 in.
The resulting cracking patterns are shown in Figure 4c at two strain levels. Here the damage in the rock is
represented graphically by a line oriented along each cracking plane in each element. For each plane orientation, the
4
3
2
1
0
S
t
r
e
s
s

(
M
P
a
)
1200 1000 800 600 400 200 0
Displacement (in)
MON*
MSN
Model simulation
* Golpalaratnam and Shah (1985)
Applied displacement
0.3
10.2
5.1
7.6
dimensions in cm
(a) Finite element mesh
Damage at A
Damage at B
(c) Calculated damage patterns
Figure 4. Calculated response of notched tension test.
(b) Load-displacement curve
A
B
length of the line is proportional to the amount of damage, defined as a function of the crack density and crack
length. A damage of one is displayed with a line that reaches the element. Thus, the graphics displays tensile and
shear "cracks" that represent damage on specific orientations. At point A, at peak load, small cracks have nucleated
throughout the specimen, above and below the material inside the notches, and cracks have started to grow from the
notches into the specimen. At point B, after the load has dropped to about 1/3 of the peak, tension cracks have
developed over much of the material between the notches.
TABLE 1. Tension Properties for concrete model simulation.
Youngs Modulus 19.5 GPa
Poissons ratio 0.20
Tensile strength 3.8 MPa
Nucleated crack length 1. mm
Crack density 1/cc
Maximum crack growth rate 5% Rayleigh wave velocity
Density 2.18 g/cc
Localized stress factor 0.12
4.2 RATE EFFECT IN TENSION
Rate effects in tension are shown in Figure 5. Figure 5a shows measured effects on peak stress of strain rate. The
measured points are the results of several experimental studies. The general trend is that there is a small effect of
strain rate up to rates of about 1/s. At the lowest strain rates effects are probably due to creep. At strain rates greater
than 1/s the strength increases significantly with strain rate. Between strain rates of 10 and 100/s the strength
increases dramatically. There is significant scatter in the measured data, some of which is attributable to different
testing techniques (Antoun 1991).
Also shown in Figure 5a is the calculated effect of finite crack velocity on peak strength as a function of strain rate.
The rate effect was calculated using a single element pulled in uniaxial tension. The maximum crack tip velocity
was taken as 5% of the Rayleigh wave velocity. No rate effect is seen for strain rates up to 0.1/s. The strength
increases at rates above 1/s. At 10/s the strength is about twice the static strength and at 50/s the strength is over 5
times the static strength. The calculated strain rate effect on strength agrees well with the measured data.
The calculated effect of strain rate on the stress-strain curve and the damage distribution is shown in Figures 5(b) and
(c). For these simulations a planar concrete specimen 40 cm (16 in) tall and 20 cm (8 in.) wide was pulled in tension.
The specimen was loaded by steel plates above and below the specimen moving at a constant velocity. These
simulations thus include effects of both finite wave velocity and also some wave propagation effects. To minimize
initial dynamic transients, the rock was given an initial velocity field consistent with the elastic response. The
interface between the rock and the plates was assumed to be frictionless. The nominal stress-stress curves at 10
-2
/s
and 10/s are shown in Figure 5(b). At the low rate, the response is relatively brittle; the stress reaches a peak
7
6
5
4
3
2
1
0
N
O
R
M
A
LI
Z
E
D
P
E
A
K
T
E
N
SI
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
1
10
2
STRAIN RATE (s
-1
)
Antoun, 1991
Cowell, 1966a
Cowell, 1966b
Gran et al., 1987
Johns et al., 1991a
Johns et al., 1991b
Mellinger, 1966
Ross et al., 1989, concrete, splitting tension
Ross et al., 1989, concrete, direct tension
Model Simulations
(a) Rate effect on tensile strength
c) Calculated rate effect on damage mechanism
(b) Calculated rate effect on stress-strain curve
Figure 5. Calculated rate effects on concrete response in tension.
strain rate = 10
-2
/s
strain rate = 10/s
8
6
4
2
0
N
O
MI
NA
L
ST
RE
SS
(M
Pa
)
0.020 0.015 0.010 0.005 0.000
NOMINAL STRAIN
10
-2
/s
10/s
of about 3.2 MPa at a strain of about 0.02% and drops quickly after the peak. At the higher rate the response is
significantly stronger and more ductile; a peak stress above 8 MPa is reached at a strain of about 0.06%.
As shown is Figure 5(c) the calculated cracking patterns at the two rates show a very different mechanism. At the
low rate the damage is localized, a single crack develops across the specimen. At 10/s the damage is distributed
throughout the specimen and cracking occurs on more than just a single plane. The wide distribution in damage is
because the stress relief from the cracks happens slowly compared to the applied loads.
4.3 UNCONFINED COMPRESSION TEST
We performed simulations of unconfined compression tests of concrete presented by Shah and Sanker (1987). The
specimens tested were nominally 6 in tall, 3 in diameter cylindrical specimens. To minimize the effects of the end
conditions, through cuts were made in the specimens spaced at 3.25 in. The specimens were tested quasistatically in
unconfined compression well beyond the peak load. Specimens were tested to various levels of axial strain then
unloaded. These specimens were sliced and prepared to allow mapping of the cracks for both radial and longitudinal
slices.
The finite element configuration for our simulations are shown in Figure 6(a). We modeled the specimen including
the through cuts, the capping compound (quick drying cement) and the steel end caps. An axisymmetric analysis was
performed. The through cuts were modeled with slidelines with no friction. The specimen was loaded by specifying
a constant vertical velocity to the steel end caps. The calculated load-strain curve is shown in Figure 6(b) along with
two representative measured curves from the tests. The load is normalized by the compressive strength of the
concrete (4200 psi). The nominal strain is the normalized displacement in the specimen measured over a 3.25 in
gage length. The calculated load curve shows a much more brittle response than the measured curves. For the
measured curves the peak load occurs at a strain value of about 30 x 10
-4
and 80% of the peak load post-peak occurs
at about 65 x 10
-4
. The peak calculated load occurs at a strain value of about 17 x 10
-4
and 80% of the peak load post
peak occurs at about 25 x 10
-4
.
Measured and calculated cracking damage is shown in Figure 6(c). At peak load the tested specimen shows some
cracking around aggregates and a few vertical cracks connecting aggregates are observed near the outside edge of the
specimen. The calculated damage shows similar patterns, cracks have nucleated over much of the specimen and a
few vertical cracks appear near the outside edges of the specimen. At 80% of peak load in the post-peak range, the
tested specimen shows much cracking, many of the aggregates have cracks around them and there are several long
vertical cracks connecting aggregates. The calculated damage shows similar patterns, vertical cracks have grown
throughout the specimen.
Spherical Wave Test
The model was used to simulate the internal cracking for a spherical explosive test in marble. Qualitatively, the
response of marble is similar to that of concrete; it is a frictional material that fails by cracking in tension and shear.
In the experiment, a small spherical charge of explosive is detonated inside a block of marble (Antoun and Curran,
1996). The marble specimen also contains a concentric set of wire loops that measure radial particle
Figure 6. Unconfined compression test
(b) Nominal stress-strain curves
(c) Cracking patterns
CALCULATION
EXPERIMENT
AT PEAK LOAD AT 81% LOAD POST-PEAK
Steel cap
Concrete
specimen
Capping
compound
Through
cut
8.3
7.6
(a) Finite element mesh
dimensions
in cm
3.8
1.2
1.0
0.8
0.6
0.4
0.2
0.0
N
O
R
M
A
L
I
Z
E
D

L
O
A
D

60 50 40 30 20 10 0
STRAIN (e-4)
Measured gage 1*
Measured gage 2
Model calculation
* Shah and Sanker 1987
Circumferential
cracks
Explosive
charge cavity
(b) Model Simulation
Figure 7. Spherical explosive charge in marble.
(a) Experiment
Radial
cracks
velocity histories as the loops expand in an imposed magnetic field. In addition to velocity histories, dynamic strain
paths can also be obtained from these records.
The test was modeled including the detonation of the explosive. A small sphere of compactable material was
included around the explosive to more accurately model the velocity histories for the material close in to the
explosive. Figure 7 compares the calculated and observed fracture patterns from a test. The model calculates both
the radial cracking pattern emanating from the blast hole and the circumferential cracks caused by reflection of the
shock wave from the outer free surface. Combined with the measured particle velocity records, the observed
cracking patterns can be used to help calibrate the dynamic parameters of the model for marble.
CONCLUSIONS
The model being developed for analyzing the response of concrete including damage and failure appears promising.
The model is physically-based, in that the cracking process is modeled explicitly and the major input parameters are
all physical, measurable quantities. Cracks nucleate and grow along prespecified orientations according to fracture
mechanics laws. Softening is a result of crack growth, as cracks grow elastic strain feeds into cracking strain. Model
simulations of notched tensile tests produce nominal stress-strain curves that agree to within about 20% of
experimentally measured curves . Rate effects due to the finite propagation speed of crack growth give large
strength enhancements at rates between 10 and 100/s, results that are consistent with those observed experimentally.
At low rates the calculated cracking damage is localized to a single crack across the specimen, at high rates the
damage is distributed throughout the specimen.
Simulations of unconfined compression tests show damage patterns that are consistent with those observed
experimentally, vertical cracks that near the peak load, nucleate along the outside of the specimen, then in the post
peak region grow throughout the specimen. The calculated concrete response in unconfined compression is much
more brittle than that measured. We believe that the laws used for crack growth, taken from linear elastic fracture
mechanics, need to be modified especially for cracking in compression, to be more consistent with many cracks
nucleating and coelescing around aggregates rather than cracks growing in an uncontrolled manner. Simulations of
explosive tests in marble show patterns and extent of cracking can be modeled. Future development plans for the
model include futher comparison with experiments and also implementing a three-dimensional version of the model
into DYNA3D.
ACKNOWLEDGMENTS
This work was sponsored by the Defense Special Weapons Agency (DSWA) under the technical direction of Dr.
Paul Senseny and Dr. Michael Giltrud. The authors gratefully acknowledge their support.
REFERENCES
Antoun, T. H., Constitutive/Failure Model for the Static and Dynamic Behaviors of Concrete Incorporating Effects
of Damage and Isotropy, Ph.D. dissertation, Dept. of Mechanical Engineering, University of Dayton (1991).
Antoun, T. H., and A M. Rajendran, "Constitutive Modeling of Concrete Under Impact Loading," Proceedings of the
APS 1991 Topical Conference on Shock Compression of Condensed Matter, Williamsburg, VA, June 17-20,
1991, edited by S.C. Schmidt, R.D. Dick, J.W. Forbes, and D.G. Tasker, pp. 497-500 (1991).
Bazant, Z. P., Imbricate Continuum and Its variational Derivation," J. Engineering Mechanics Div., ASCE, 110,
1693-1712 (1984).
Bazant, Z. P., T. B. Belytschko, and T. -P. Chang, Continuum Model for Strain Softening, J. Engineering
Mechanics Div., ASCE, 110, 1666-1692 (1984).
Bazant, Z.P., and B. H. Oh, "Microplane Model for Progressive Fracture of Concrete and Rock," J. Engineering
Mechanics Div., ASCE, 111(4), 559-582 (1985).
Bazant, Z.P., and B. H. Oh, Crack Band Theory for Fracture of Concrete, Materials and Structures (RILEM,
Paris), 16, 155-177 (1983).
Bazant, Z.P., and S. Kim, "Plastic-Fracturing Theory for Concrete," J. Engineering Mechanics Div., ASCE,
105, 429-446 (1979).
Belytchko, T. B. and Z. P. Bazant, Wave propagation in a Strain Softening Bar: Exact Solution, Journal of
Engineering Mechanics, ASCE, 111, 381-389 (1985).
Bicanic, N., R. de Borst, W. Gerstle, D. W. Murray, G. Pijauder-Cabot, V. Saouma, K. Willam, and J.
Yamazaki,Chapter 7: Computational Aspects, in Finite Element Analysis of Reinforced Concrete Structures
II, Proceedings of the International Workshop, ed. J. I. Isenberg, ASCE/ACI 447, Columbia University, New
York, June 2-5, (1991).
Carreira, D.J., and K. Chu, "Stress-Strain Relationship for Plain Concrete in Compression," ACI Journal,
Proceedings, 82, 797-804 (1985).
Chen, A.C.T., and W. F. Chen, "Constitutive Relations for Concrete," J. Engineering Mechanics Div.,
ASCE, 101(4), 465-481 (1975).
Chen, W. F., E. Yamaguchi, M. D. Kotsovos, and A. D. Pan,Chapter 2: Constitutive Models, in Finite Element
Analysis of Reinforced Concrete Structures II, Proceedings of the International Workshop, ed. J. I. Isenberg,
ASCE/ACI 447, Columbia University, New York, June 2-5, (1991).
Coleman, B. D., and M. E. Gurtin, Thermodynamics with Internal State Variables, Journal of Chemical Physics,
47(2), 597-613 (1967)
Costin, L. S., and C. N. Stone, A Finite Element Model for Microfracture-Damaged Brittle Rock, SANDIA
Report, SAND87-1227UC-13, (June 1987).
Cowell, W., "Dynamic Properties of Plain Portland Cement Concrete," U.S. Naval Engineering Laboratory,
Technical Report no. R447, Port Hueneme, CA (1966).
Cundall, P. A., A Computer Model for Simulating Progressive, Large-Scale Movements in Blocky Rock Systems,
Proceedings of the International Symposium on Rock Fracture, II-8, Nancy, France (1971).
Curran, D. R., L. Seaman and D. A. Shockey, Dynamic Failure of Solids, Physics Reports 147(5&6), 253-388
(March 1987).
Dantu, P., Annal de LInstitut Technique du Batiment et des Traveaux Publiques, 11(121), 55-77 (1958).
Dougill, J. W., Some Remarks on Path Independence in the Small Plasticity, Quarterly of Applied Mathematics,
32, 233-243 (1975).
Espinosa, H. D., "On the Dynamic Shear Resistance of Ceramic Composites and it's Dependenceon Applied
Multiaxial Deformation," Int. Journal of Solids and Structures, 32(21), 3105-3128 (1995).
Fanella, D., and D. Kracjinovic, "A Micromechanical Model for Concrete in Compression," Engineering Fracture
Mechanics, 29(1), 49-66 (1988).
T. H. Antoun, and D. R. Curran, Wave Propagation in Intact and Jointed Calcium Carbonate Rock, Final Report to
Defense Nuclear Angency DNA-TR-95-47, from SRI International, Menlo Park, CA (1996)
Freund, L.B., "Crack Propagation in an Elastic Solid Subjected to General Loading-III. Stress Wave
Loading," J. Mech. Phys. Solids, 21, 47-61 (1973).
Gopalaratnam, V. S., and S. P. Shah, Softening response of plain concrete in direct tension. J. American Concrete
Institute, 82(3), 310-323 (1985).
Gran, J. K., Development of an Experimental Technique and Related Analyses to Study the Dynamic Tensile
Failure of Concrete, Air Force Office of Scientific Research Technical Report, AFOSR-TR-85-1240, Bolling
Air Force Base, Washington D. C. (1985).
Gran, J. K., Y. M. Gupta, and A. L. Florence, "An Experimental Method to Study the Dynamic Tensile Failure of
Brittle Geologic Material," Mechanics of Materials, 6, 113-125 (1987).
Gran, J. K., and L. Seaman, "Observations and Analyses of Microcracks Produced in Dynamic Tension
Tests of Concrete," SRI International Technical Report, prepared for the AFOSR under Contract No.
F49620-82-K-0021 (1986).
Han, D. J., and W. F. Chen, Strain-Space Plasticity Formulation for Hardening-Softening Materials with Elasto-
Plastic Coupling, Int. J. Solids Structures, 22
Hillerborg, A., M. Modeer, and P.-E. Peterson, Analysis of crack formation and crack growth in concrete by means
of fracture mechanics and finite elements, Cem. Concr. Res., 6(6), 773-782 (1976).
Horii, H., and S. Nemat-Nasser, "Compression-Induced Microcrack Growth in Brittle Solids: Axial Splitting and
Shear Failure," J. of Geophysical Research, 40(B4), 3105-3125 (1985).
Hsu, T. C., F. O. Slate, G. M. Sturman, and G. Winter, "Microcracking of Plane Concrete and the Shape of the
Stress-Strain Curve," Journal of the American Concrete Institute, 60(2), 209-224 (1963).
John, R., T. H. Antoun, and A. M. Rajendran, "Effect of Strain Rate and Size on Tensile Strength of Concrete,"
Proceedings of the APS 1991 Topical Conference on Shock Compression of Condensed Matter,
Williamsburg, VA, June 17-20, 1991, edited by S.C. Schmidt, R.D. Dick, J.W. Forbes, and D.G. Tasker, 501-
504 (1991).
Kachanov, L. M., On the Creep Fracture Time, Izv. AN SSR, Otd. Tekhn. Nauk, No. 8, 26-31 (1958).
Kachanov, M., Elastic Solids with Many Cracks and Related Problems, Advances in Applied Mechanics, v. 30,
editors: J. Hutchinson and T. Wu, Academic Press (1993).
Kawai, T., New Discrete Structural Models and Generalization of the Method of Limit Analysis, Finite Elements
in Nonlinear Mechanics, 2, Norwegian Institute of Technology, Tronheim, 885-906 (1977).
Kotsovos, M. D., Effect of Testing Technique on the Post-Ultimate Behavior of Concrete in Compression.
Materials & Structures, RILEM, 16(92), 3-12 (1983).
Krajcinovic, D., and G. Fonseka, "The Continuous Damage Theory of Brittle Materials - Part 1: General Theory," J.
Appl. Mech., ASME, 48, 809-815 (1981).
Maekawa, K., and H. Okamura, The Deformational Behavior and Constitutive Equation of Concrete Using the
Elasto-Plastic and Fracture Model, Journal of Faculty of Engineering, University of Tokyo (B), XXXVII(2),
253-328 (1983).
Margolin, L. G., Elastic Moduli of a Cracked Body, Int. Jour. of Fracture, 22, 65-79 (1983).
Margolin, L.G., "A Generalized Griffith Criterion for Crack Propagation," Engng. Fract. Mech., 19(3), 539-543
(1984).
Mellinger, F. M., and D. L. Birkimer, "Measurements of Stress and Strain on Cylindrical Test Specimens of Rock
and Concrete Under Impact Loading," Department of Army, Ohio River Division Laboratory Technical
Report no. 4-46 (1966).
Mindess, S., and J. Young, Concrete, Prentice Hall, Englewood Cliffs, NJ (1981).
Murray, D. W., et al., Concrete Plasticity Theory for Biaxial Stress Analysis, J. Engineering Mechanics Div.,
ASCE, 105(6), 989-1006 (1979).
Needleman, A., Material Rate Dependence and Mesh Sensitivity in Localization Problems, Comp. Methods Appl.
Mech. Eng., 67, 69-85 (1988).
Ohtani, Y., and W. F. Chen, Hypoelastic-Perfectly Plastic Model for Concrete Materials, J. Engineering
Mechanics Div., ASCE, 113(12), 1840-1860 (1987).
Popovics, S., A Numerical Approach to the Complete Stress-Strain Curve of Concrete, Cement Concrete
Research, 3(5), 583-599 (1973).
Rashid, Y. R., Analysis of prestressed concrete vessels, Nuclear Engrg. and Design, 7(4), 334-344 (1968).
Read, H. E., and G. A. Hegemier, Strain Softening of Rock, Soil, and Concrete- A Review Article, Mechanics of
Materials, 3, 271-294 (1984).
Ross, C.A., P. Thompson, and J. Tedesco, "Split-Hopkinson Pressure-Bar Tests on Concrete and Mortar in Tension
and Compression," ACI Materials Journal, 86(5), 475-481 (1989).
Saenz, L. P., Discussion of Equations for the Stress-Strain Curve of Concrete, by P. Desai and S.
Krishman, J. American Concrete Institute, 61(9), 1229-1235 (1964).
Sandler, I., and J. Wright, Summary of Strain-Softening, in Theoretical Foundations for Large-Scale
Computationsof Nonlinear Material Behavior, DARPA-NSF Workshop, ed. by Namat-Nasser, Northwestern
University, (1984).
Seaman, L., J. K. Gran, and D. R. Curran, "A Microstructural Approach to Fracture of Concrete," Application of
Fracture Mechanics to Cementitious Composites, edited by S.P. Shah, Martinus Nijhoff Publishers, Dordrecht,
the Netherland, 481-506 (1985).
Shah, S. P., and R. Sankar, Internal Cracking and Strain-Softening Response of Concrete under Uniaxial
Compression, ACI Materials Journal, 84(3), 200-212 (1987).
Sih, G.C., and E. P. Chen, "Cracks Moving at Constant Velocity and Acceleration," in Mechanics of
Fracture, Vol. 4: Elastodynamic Crack Problems, edited by G.C. Sih, Noordhoff International
Publishing, Leyden, the Netherlands (1977).
Sluys, L. J., and R. de Borst, A N6umerical Study of Concrete Fracture Under Impact Loading, Micromechanics of
Failure of Quasi-Brittle Materials, S. P. Shah et al., eds., Elsevier Applied Science, London, 524-535 (1990).
Sneddon, I. N., and M. Lowengrub, Crack Problems in the Classical Theory of Elasticity, John Wiley & Sons,
New York, (1969).
Tsai, Y.M., "Exact Stress Distribution, Crack Shape and Energy for a Running Penny-Shaped Crack in an
Infinite Elastic Solid," International Journal of Fracture, 9(2), 157-169 (1973).
Tsai, W. T., "Uniaxial Compressional Stress-Strain Relation of Concrete, J. Structural Div., ASCE, 114(9),
2133-2136 (1988).
Whirley, R. G., and B. E. Engelmann, DYNA2D a Nonlinear, Explicit, Two-Dimensional Finite Element Code for
Solid Mechanics, User Manual, Lawrence Livermore National Laboratory Technical Report No. UCRL-MA-
110630 (April 1992).
Wittman, F. H. Influence of time on crack formation and failure of concrete. Application of fracture Mechanics to
Cementitious Composites, NATO Advanced Research Workshop, ed. by S. P. Shah, Sept 4-7 (1984).
Wittmann, F.H., "Micromechanics of Achieving High Strength and Other Superior Properties," High Strength
Concrete, Proceedings of a Workshop held at the University of Illinois at Chicago Circle, Dec. 2-4 (1979).

Das könnte Ihnen auch gefallen