Multiscale Materials Modelling through Machine Learning: Hydrogen-Steel Interaction during Deformation

TL;DR

This paper explores using physics-informed machine learning to predict hydrogen-steel interactions during deformation, aiming to improve prediction accuracy across multiple scales without complex constitutive models.

Contribution

It introduces a simplified physics-informed machine learning approach for multiscale materials modeling of hydrogen-steel interactions, reducing reliance on complex constitutive equations.

Findings

Improved prediction accuracy with less data needed

Model works across different length scales

Eliminates complex constitutive modeling requirements

Abstract

This short paper presents the potential of using machine learning to predict materials behaviour in the context of hydrogen interaction with steel. Effort has been made to understand the quality, and amount of data needed to get improved predictions. An approach known as physics informed machine learning has been adapted in a simplified way through data classification to show the improvement in predictions. Proposed model eliminates the requirement to solve complex materials constitutive models and can work for any length scale, in the present case it is used for single crystalline steel interacting with steel under different types of loading.