Bayesian Model Reduction for Flow Fields

Abstract

Model reduction for numerical simulation is a key topic in the area of computational mechanics and system identification. In this study we propose a Bayesian model reduction method with the use of Gaussian process and Stein variational gradient method. As an application of the proposed method for flow field, we estimate a reduced model of one-dimensional advection diffusion equation. In this procedure, the parameter of the Gaussian process is estimated by posterior distribution by Bayes’ sense.