Model Order Reduction for Strongly Coupled Multiphysics Problems: A POD-ANN Framework and a Physics-Driven Surrogate Model

Abstract

This work presents and evaluates two complementary model order-reduction strategies: POD-ANN and a physics-driven surrogate model, for strongly coupled thermo-mass-mechanical systems. Using high-fidelity cooking simulations as a case study, both approaches achieve substantial computational speed-ups while preserving dominant thermal, transport, and deformation mechanisms, enabling fast prediction and parametric exploration.