Caglar Oskay delivered a Distinguished Lecture on Computational Mechanics During ASME IMECE 2020 – 11/19/2010
Lecture Title: Prediction of Localized Events in Multiscale Simulations: Achieving Failure Initiation Predictions in Large Scale Structural Simulations
Over the past couple of decades, tremendous effort has been devoted to the development of multiscale computational modeling and simulation strategies for physics-based prediction of structural response. However in majority of the past studies, the focus has been placed on achieving an accurate response at the coarse (e.g., structural) scale. In a large scale simulation of a structure, fine-scale response fields are typically tracked either at very few specific locations within the structure (or more commonly, nowhere at all!) due to the prohibitive computational cost of retaining fine scale information. In an important class of problems, such as in characterization of failure initiation, the ability to capture fine-scale response fields across the structural domain becomes much more important and valuable than tracking coarse scale measures of structural response.
In this study, we propose and evaluate a reduced order multiscale modeling strategy to effectively track fine scale response fields everywhere within a structure throughout a multiscale simulation. The methodology employed for model order reduction is Eigenstrain-based Reduced Order Model coupled with the computational homogenization method employed in bridging the coarse and fine scales. employs the idea of precomputing certain information on the material microstructure such as the influence functions, localization operators and coefficient tensors through RVE scale simulations, prior to the macroscale analysis. The eigenstrain-based reduced order modeling employs the idea of precomputing certain information on the material microstructure (such as the influence functions, localization operators and coefficient tensors) using fine scale simulations, prior to the macroscale analysis. The key premise is to obtain efficient localization operators with low FLOP count as well as reasonable memory demands. We build and demonstrate the capabilities of the proposed strategy in the context of fatigue failure initiation and nonlinear mechanical response prediction in structures made of polycrystalline materials, where crystal plasticity finite element (CPFE) simulations at the scale of RVEs are concurrently coupled with a large scale structural analysis. Cycle-sensitive, dislocation density-based crystal plasticity models are used to idealize the nonlinear response of the material in the CPFE simulations at the RVE scale. We demonstrate the ability of the reduced order models in accurately capturing local, grain-scale features (grain level stress, strain, dislocation density evolution) as well as capturing failure initiation mechanisms in the context of a high-performance titanium alloy (Ti-6242S). In particular, the multiscale simulations quantify and characterize the spatial distribution and evolution of the dislocation pile-ups subjected to cyclic loading. The effect of non-uniform texture on the response of the structural component is also investigated.