DATA-PHYSICS DRIVEN REDUCED ORDER HOMOGENIZATION FOR CONTINUUM DAMAGE MECHANICS AT MULTIPLE SCALES

Abstract

The paper introduces a novel approach, referred to as data-physics driven reduced order homogenization, for continuum damage mechanics. The proposed method combines the benefits of the physics-based reduced order homogenization and data-driven surrogate modeling by striking a balance between accuracy, computational efficiency, and physical interpretability. The primary objective of this hybrid approach is to minimize computational cost associated with online predictions at the macroscopic scale while preserving accuracy and physical interpretability. This is achieved by leveraging a surrogate-based Bayesian inference to extract crucial information at a representative volume element (RVE) level. With the inferred data, online predictions are performed using a data-enhanced reduced order homogenization. Consequently, the computational time required at a macroscopic scale is significantly reduced compared to both the conventional direct numerical simulation and the computational homogenization approach.