2020/10/07 by Anh Tran, Tran, Anh, Timothy Wildey +1 · 3 citations
Computer Science · Decision Sciences · Materials Science · #Advanced Multi-Objective Optimization Algorithms #Computational Drug Discovery Methods #Computational Engineering #FOS: Computer and information sciences #Finance #Machine Learning in Materials Science #Probabilistic and Robust Engineering Design #and Science (cs.CE)
paper · pdf · doi:10.48550/arxiv.2010.03603
openalex publication_date 2020/10/07 · openalex created_date 2022/07/18 · openalex updated_date 2026/07/28
Determining process-structure-property linkages is one of the key objectives\nin material science, and uncertainty quantification plays a critical role in\nunderstanding both process-structure and structure-property linkages. In this\nwork, we seek to learn a distribution of microstructure parameters that are\nconsistent in the sense that the forward propagation of this distribution\nthrough a crystal plasticity finite element model (CPFEM) matches a target\ndistribution on materials properties. This stochastic inversion formulation\ninfers a distribution of acceptable/consistent microstructures, as opposed to a\ndeterministic solution, which expands the range of feasible designs in a\nprobabilistic manner. To solve this stochastic inverse problem, we employ a\nrecently developed uncertainty quantification (UQ) framework based on\npush-forward probability measures, which combines techniques from measure\ntheory and Bayes rule to define a unique and numerically stable solution. This\napproach requires making an initial prediction using an initial guess for the\ndistribution on model inputs and solving a stochastic forward problem. To\nreduce the computational burden in solving both stochastic forward and\nstochastic inverse problems, we combine this approach with a machine learning\n(ML) Bayesian regression model based on Gaussian processes and demonstrate the\nproposed methodology on two representative case studies in structure-property\nlinkages.\n