A high-performance computational framework for fast crystal plasticity simulations

We present a new computational framework aimed at dramatically reducing time needed for crystal plasticity simulations. The framework is based on a combination of the recently developed numerical implementations of crystal plasticity in the spectral representation for obtaining the response of single crystals and specialized computer hardware that integrates a graphics-processing unit (GPU). Following a divide and conquer approach adapted here from a fast GPU8 method for matrix operations, we describe a new GPU based implementation of the spectral crystal plasticity and demonstrate its performances through a few example case studies involving a Taylor-type polycrystalline model. Using a single GPU card, the novel framework described herein provides speedup factors exceeding three orders of magnitude over the conventional crystal plasticity numerical schemes. (C) 2013 Elsevier B. V. All rights reserved.