H. Horii
Department of Civil Engineering
University of Tokyo
Bunkyo-ku, Tokyo, Japan
First, a review of some theories of toughening mechanisms in quasi-brittle materials was presented by
Professor C. Atkinson. The spring model that includes the bridging mechanism was explained.
Micromechanical models for various bridging mechanisms are summarized. As an example of studies
on micromechanics of bridging, analyses of crack growth along the interface between short fiber and
matrix are presented. In the discussion, mechanisms other than bridging such as crack deflection,
crack branching, and microcracking are discussed.
Following the presentation by Professor C. Atkinson, a short presentation was given by Professor
S.J. Bennison. It was shown that a relationship between bridging stress a and crack opening
displacement w is obtained from the crack profile measured in situ and a Green function. Then, by
integrating it, a resistance curve, a relationship between J-value and COD, is obtained.
One of interesting issues on the bridging model is the generalization of the spring model to
nonplaner crack growth under general loading. The bridging property that is represented bya-w
curve may be different under pure mode I loading and mixed-mode loading. In fracture of concrete
it has been clarified that the bridging is the dominant mechanism and fracture of concrete has been
modeled by Dugdale-Barenblatt type model with a tension-softening curve. Crack growth phenomena
under pure mode I loading are reproduced by the model with reasonable accuracy. It is an interesting
question whether the analytical model developed for pure mode I loading is applicable to crack growth
under mixed-model loading. Even under mixed-mode loading a crack (or bridging zone) propagates
normal to the maximum tensile direction. The difference is that crack does not propagate straight but
kinks at the tip of an initial crack and curves as it propagates. The following is an example of a study
of crack growth under mixed-mode loading where deviation from pure mode I is considered to be
relatively small.
Pullout tests of a two dimensional anchor bolt are carried out on concrete specimen. Cracks are
initiated at the upper comer of the anchor and grow along curving path toward the pOSition of
support. Surface displacements are measured with laser speckle method, and distribution of normal
and tangential displacement gaps and the location of the tip of macrocrack (bridging zone) are
identified at different loading levels. Measured crack lengths at different pullout load are plotted in
Fig. 1.
A BEM code of Dugdale-Barenblatt type model with a linear tension-softening relation is
employed for the analysis of crack growth in the pullout test. It reproduces crack growth phenomena
under mode I loading with reasonable accuracy. The same value of critical opening displacement
w.,. = 25pm is used. The results are plotted in Fig. 1 and compared with experimental observation.
It is surprising that the observation and the prediction show very different tendencies. This
discrepancy seems to indicate a significant difference in fracture mechanisms between crack growth
under pure mode I loading and mix-mode loading even though the ratio of mode II component is
relatively small.
379
S. P. Shall (ed.), Toughening Mechanisms in Quasi-Brittle Materials, 379-382.
© 1991 Kluwer Academic Publishers.
380
intensity factor is local quantity. Corresponding to this point the relationship between the effective
compliance and energy release rate (and accordingly the stress intensity factor) was pointed out in the
discussion.
A method of solution for multicrack problems that is powerful for three dimensional problem was
also presented. Results of numerical simulation of crack-microcrack interaction with randomly
generated distribution of microcracks were shown. It was pointed out that no statistically stable effect
of shielding or amplification was observed. There was in the discussion one comment that the effect
of residual strain is much larger than that of reduction in stiffness in toughening due to residual strain
induced microcracking. Change in stress intensity factor of a stationary crack due to microcracks
cancels out if microcracks are induced only by stress concentration without residual strain.
One of interesting issues coming up in this session is on the method of approach for description
of mechanical phenomena: computer simulation versus micromechanics (or micro mechanics based
continuum approach). Both approaches are based on behavior of microstructures. However, ways of
bridging microscopic behavior and macroscopic behavior are different in two approaches.
In computer simulation relationship between micro and macro is treated through numerical
computation. Hence it is sometimes called numerical experiments. The arrangement of micro-
structures in each computation is important and sometimes the method is incorporated with statistical
technique.
On the other hand, micromechanics (or micromechanics based continuum approach) aims at, at
least as the final goal, establishing continuum model through homogenization. One of main tasks
of micromechanics approach is the analysis of behavior of microstructure. Many studies with
micromechanics approach are finished at this stage. However, our final goal in many cases is the use
of those results at microscopic scale for the prediction of macroscopic phenomena.
In general, it is not an easy task to bridge microscopic phenomena and macroscopic behavior of
materials by homogenization or averaging. One typical example of such process is the derivation of
overall moduli of cracked solids. Another example is found in the analysis of fracture of fiber-
reinforced materials shown by the first speaker of this session; see Fig. 2. Pullout behavior of a single
fiber with debonding and cracking is a microscopic behavior. By modelling the microscopic mechanism
one may reproduce microscopic phenomena. Then through averaging or homogenization process, the
relationship between bridging stress and crack opening displacement is obtained. It is a macroscopiC
behavior that is used as the constitutive equation in the bridging zone for the analysis of crack growth
in fiber-reinforced material.
Homogenization
Micromechanics Prediction
As a matter of course it is important to clarify the objective of study. What is the objective when
studies on material behaviors are carried out paying attention to microscopic phenomena? One
answer may be the clarification of the governing mechanism. It would be phrased as to develop an
improved understanding of process of deformation or fracture. Another answer may be the
establishment of analysis tools for prediction or reproduction of phenomena in actual problems. The
appropriate way to approach the problem should be different for different Objectives.
Computer simulation is suitable for clarification of the governing mechanism. If the governing
mechanism is modelled properly, the phenomena are simulated under general conditions. The
agreement of prediction with observation is a support of the modeled mechanism. However, one
should be careful since the simulation may not work if the most dominant mechanism is not included
or it is lost in the process of simulation. It is too optimistic to expect that real phenomena are
reproduced by introducing imperfections. The buckling of a column cannot be predicted by giving
initial imperfections unless geometrical nonlinearity is included. A typical example in solid mechanics
is localization phenomena. In localization of any type, the interaction between microstructures is
important. Unless it is treated properly, localization phenomena may not be simulated.
Computer simulation is not suitable for analysis of actual problems. In actual problems each
microstructure cannot be treated separately. We have to use continuum quantities and governing
equations for them. Micromechanics approach (or micromechanics based continuum approach) is
suitable not only for identification of the governing mechanism but also for development of analysis
tools. However, careful consideration is also necessary for micro mechanics approach because
important information may be lost in the process of homogenization.
For example, growth of two parallel cracks shown in Fig. 3 depends on the loading condition.
In the case (a) only one crack grows when stress reaches its critical value; localization. In case (b) two
cracks grow equally; delocalization. When we introduce homogenization and use a continuum model
as shown in Fig. 3c, the information that makes the distinction between (a) and (b) must be kept in
the homogenization process. Otherwise the prediction by the continuum model could be very different
from reality. It would be the future task to find the governing mechanism of localization phenomena
and to establish proper procedure of homogenization.
ffffIff+
+
{iiiili,
-
+
(a) Crack growth under uniform (b ) Crack growth under (c) Continuum model after
stress at infinity concentrated forces homogenization