Three-invariant non-coaxial elastoplastic constitutive modeling and its implications on the localization properties of rocks

Principal Investigator: Ronaldo I. Borja
Project Sponsor: Division of Chemical Sciences, Geosciences and Biosciences, Office of Science, Department of Energy


Project Description

Strain localization is a ubiquitous feature of geomaterials undergoing nonhomogeneous deformation. In soils and rocks, the zone of localized deformation is generally referred to as either a fault, shear band, rupture zone, or simply a failure plane, and is well explained by either fracture mechanics or bifurcation theory.  Localized deformation is typically followed by a reduction in the overall strength of the material as the loading proceeds. It is thus of  considerable interest and importance to be able to predict when a shear band forms, how this narrow zone of discontinuity is oriented within the material, and how the  propagation of the shear band is influenced by the post-localization constitutive responses.

Much experimental work has been conducted to understand the inception of localized deformation in rocks. It is important to recognize that the overall material response observed in the laboratory is a result of many different micromechanical processes such as microcracking in brittle rocks, mineral particle rolling and sliding in granular soils, and mineral particle rotation and translation in the cement matrix of soft rocks. Ideally, any localization/bifurcation model for geomaterials must capture all of these important micromechanical processes. However, current limitations in the laboratory testing capabilities and mathematical modeling techniques inhibit the use of a micromechanical description of their behavior, and a macromechanical approach, such as that employing theory of plasticity, is still heavily favored by the geomechanics modeling community.

Strain localization is often viewed as an instability process that can be predicted in terms of the pre-localization constitutive relations. The material is assumed to deform homogeneously until its constitutive relations allow a bifurcation from a smoothly varying deformation field into a highly concentrated shear band mode. The bifurcation point is detected by a stability analysis. For modeling purposes the bifurcation point signals the onset of a localization mode. Therefore, an accurate prediction of the bifurcation point is very crucial in the simulation of the mechanical behavior of geomaterials. Equally critical is an accurate representation of the mechanical response following localization.

It is well known that predictions of faulting as a bifurcation from homogeneous deformation are strongly dependent on the constitutive description of homogeneous deformation. Unfortunately, standard plasticity models are inadequate for predicting the bifurcation point. The absence of the third stress invariant and the inability of a plasticity model to capture the vertex-like structure exhibited by many brittle materials may have contributed to the model's poor prediction of the bifurcation point.

In brittle rock masses the existence of a vertex-like structure makes the constitutive response a function not only of the state of stress but also of the direction of the stress increment. This may be illustrated by considering a stress path tangent to a yield surface (loading to the side), which is predicted by any conventional plasticity model with a smooth yield surface to result in an elastic response, whereas the presence of a vertex structure at the same stress point would otherwise predict a plastic response. Vertex effects play no role for proportional loading, and are insignificant or non-existent for non-proportional loading with fixed principal stress axes. However, they are very important for highly non-proportional loading with rotating principal stress axes and are known to enhance strain localization. Unfortunately, the vertex effect creates enormous difficulties with the classical tensorial models of plasticity, and to date no tensorial models exhibiting the vertex effect are currently available for large-scale finite element analysis.

In this project, we investigate the effects of the third stress invariant and vertex-like responses on the prediction of the onset of strain localization in geomaterials in general, and in rocks in particular. Our analysis will build upon a recently developed stress-point integration algorithm for a class of three-invariant elastoplastic constitutive models for geomaterials. Our rationale for proposing to carry out this investigation contingent upon the availability of a robust stress-point integration algorithm is that in a majority of cases a robust algorithm is necessary to describe the discrete evolution of the stresses and plastic internal variables, and that without this algorithm it may not be possible to accurately capture the plastic loading paths leading to strain localization.

Some of the three-invariant plasticity models we will consider include the smooth versions of the Mohr-Coulomb plasticity model, i.e., the Matsuoka-Nakai and Lade-Duncan  models, recently implemented within the context of return mapping algorithm by the Stanford Geomechanics group. Widely used for representing the yield and failure behavior of cohesive-frictional materials, these models are most conveniently expressed in terms of principal stresses and predict higher yield/failure strength in compression than in tension. In addition, we will also introduce enhancements to these models to capture the curved shape of the yield surface on the meridian plane. By utilizing constitutive models truly appropriate for geomaterials, we hope to represent their mechanical responses more realistically and predict the onset of strain localization more accurately.

In addition to choosing a more accurate characterization of the overall behavior through the use of more realistic plasticity models, we also plan to investigate the effect of vertex-like yielding on the localization properties of brittle rock masses. Our investigation will build upon a non-coaxial plasticity model formulation that circumvents the assumed coaxiality between the principal stresses and the principal plastic rate of deformation. We assume that the non-coaxial part of the deformation rate is related to the rate of relative stress, which in turn translates the yield surface, thus alleviating the highly nonlinear consistency condition that results from the classical vertex plasticity formulations.

To complete the strain localization modeling, we will put the above bifurcation theory in the framework of the strong discontinuity concept for modeling post-localization responses. This final step allows us to track the propagation and evolution of the shear band using an enhanced finite element formulation that alleviates if not completely circumvents the mesh dependence issues in finite element modeling. Modeling the post-localization responses is a critical component of strain localization analysis since shear bands do not appear instantaneously in the material, but rather, they develop progressively. The occurrence of plastic slip at the tip of the shear band destroys the symmetry of the finite element problem, and thus alters the stress state in the next localizing elements. With a non-coaxial plasticity formulation, we hope to develop a more realistic finite element description of the localization and post-localization responses of brittle rock masses particularly at the tip of the shear band where the influence of vertex yielding is expected to be more pronounced.