Some aspects of strain localization modeling in large deformation finite-element analyses

Chen, Jin (2022) Some aspects of strain localization modeling in large deformation finite-element analyses. Doctoral (PhD) thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (13MB)

Abstract

Shear strength of soils might decrease with an increase of shear strain, which is called “strain-softening.” The degradation of strength occurs in different types of soils (e.g., loose sand and sensitive clay) and under different loading conditions (e.g., drained or undrained). Strain-softening plays a major role in a wide range of geotechnical problems, including catastrophic failures (e.g., large-scale landslides) and laboratory tests (e.g., biaxial tests). While physical modeling and laboratory testing provide an overall or macro-scale response, numerical analysis could contribute insights into the mechanisms for a wide range of conditions. Unfortunately, numerical simulation of problems involving strain-softening materials is equally challenging, and the challenges include strain localizations, mesh sensitivity and large deformation. Several methods have been proposed and implemented in traditional Lagrangian-based finite element (FE) programs to overcome mesh sensitivity and localization issues. However, the suitability of these methods for large deformation problems has been verified in limited studies. While the soil types and loading conditions might be different, the geotechnical problems involved in softening could be analyzed developing a common framework. Therefore, the aim of the present study is to develop large-deformation FE modeling techniques for minimizing mesh sensitivity and simulating large-scale landslides. Nonlocal regularization methods were previously proposed to solve the mesh dependency issues in modeling strain localization. These methods relate the strength reduction of the local element to the strains of neighbouring elements by using weight functions. However, their applications for a large-deformation FE are limited. In the present study, several numerical algorithms are developed, successfully implemented in a Eulerian-based FE program, and validated against previous numerical studies with three nonlocal regularization methods. Simplification and optimization are applied to reduce the computational costs. The performance of the three nonlocal methods is evaluated by simulating a biaxial test, and one of them is found to be relatively more effective and thus used for further in-depth study of strain regularization. Shear bands in local and nonlocal models are usually much thicker than in real soil, and their thickness increases with an increase in shear displacement. Softening scaling is effective to model the macroscopic force−displacement behaviour without modeling the thin shear band that develops in real soil. Softening scaling in a local model can also function as a regularization tool, which is usually called “element size scaling.” The element size scaling method and the nonlocal method show comparable effects in strain regularization in a biaxial compression test simulation and a slope failure analysis with Eulerian-based FE program. However, a significantly larger computational time is required for a nonlocal analysis and some interaction between two neighbouring shear bands might occur. Therefore, the element size scaling method is chosen in the analysis of a large-scale slope failure. Finally, the large-scale flow slide failure of the Lower San Fernando Dam is modeled using the techniques developed in a Eulerian-based FE program. Simplified constitutive models are developed to simulate the strain-softening behaviour of different soil layers. The seepage is modeled by developing a new technique based on the thermal−fluid analogy, which is incorporated in the in-situ stress calculations. Eulerian-based FE simulations show a similar failure pattern observed in post-slide investigations and also explain potential failure processes.

Actions (login required)

View Item