A computational framework for microstructural crack propagation

• Published in: International journal of fatigue 2021-11, Vol.152, p.106397, Article 106397
• Main Authors: Brockman, Robert A., Hoffman, Rebecca M., Golden, Patrick J., Musinski, William D., Jha, Sushant K., John, Reji
• Format: Article
• Language: English
• Subjects: Abaqus; Computational mechanics; Constraint modelling; Crack propagation; Crystal plasticity; Cyclic loads; Decision analysis; Decision making; Load fluctuation; Material model; Materials fatigue; Metal fatigue; Microstructural crack propagation; Nonlinear response; Propagation; User subroutines
• ISSN: 0142-1123; 1879-3452
• DOI: 10.1016/j.ijfatigue.2021.106397

•Describes methods for implementation of complex user-defined models.•Demo combines user materials, XFEM control, load cycling, and re-equilibration.•Crystal plasticity and microstructural crack propagation used as example.•Methods are suitable for fully parallel computation.•Shared memory utilities used for implementing non-local computations. This paper describes a computational framework for solving, in the context of large-scale commercial mechanics codes, complex problems in which specialized models are required for phenomena that are larger in scope than pointwise material models, or for loading and constraints that vary by position. Production analysis codes typically include interfaces for user-supplied submodels, but supply information only at a single point of interest, such as a model node or integration point. The particular example addressed herein is that of crack propagation on the microstructural scale, in which communications are required not only between the submodels and the analysis code, but between individual submodels to allow decision-making about nonlinear material response, crack propagation criteria, material interface behavior, time-dependent load variation, and convergence of nonlinear cyclic forced response. While the specific models discussed are of interest in metal plasticity and fatigue analysis, the methodology described is applicable to numerous other complex problems in computational mechanics where communication between user-written submodels and the analysis code require more than pointwise response information.