Bayesian model class selection using quoFEM

Dr. Aakash Bangalore Satish | January 12, 2024

"Live Expert Tips" with the NHERI SimCenter is a virtual event that touches on diverse aspects of simulation and modeling in Natural Hazards Engineering. In each session, a technical example will demonstrate features of SimCenter Tools relevant to research in natural hazards and the built environment. After the example, time is allocated for questions, answers, and discussion on the presented topic. This session offers an introduction to Bayesian model class selection using the SimCenter's quoFEM tool. Bayesian model class selection involves the application of Bayes' rule to identify the most plausible model class for the data from among various candidate model classes. After a brief introduction to the quoFEM tool, a demonstration of its use for model class selection will be provided. This demonstration will be through an example focusing on the selection of a material constitutive model for the experimentally measured hysteretic response of a steel coupon.

About the Presenter:
Dr. Aakash Bangalore Satish is a Postdoctoral Scholar at UC Berkeley and works as a Software Developer at the NHERI SimCenter specializing in uncertainty quantification and Bayesian inference. He earned his Ph.D. from Johns Hopkins University, where he developed a novel method to quantify and propagate model probability uncertainty. His work at the SimCenter has focused on the implementation of features for Bayesian model calibration and Bayesian model class selection. Aakash's current research interests are in the application of hierarchical Bayesian modeling for uncertainty quantification, and in leveraging surrogate models for efficient approximation and updating of expensive computational models.