Model Discovery

Constitutive Model Discovery from Physics-Enforced Neural Network

This research project aims to solve two key research questions

1. Can we utilize the expressiveness of neural networks inside constitutive models while enforcing fulfillment of physical laws, such as the laws of thermodynamics?

2. For the calibrated (or trained) models, can we discover interpretable analytical expressions that model the material more accurate than existing models?

To do this, we use data from both physical and numerical experiments. See Meyer and Ekre (2023) JMPS,180 p.105416 (doi: 10.1016/j.jmps.2023.105416) for further details.