Physics-Based Machine Learning Approach for Modeling the Temperature-Dependent Yield Strength of Superalloys

TL;DR

This paper introduces a physics-based machine learning model that predicts the temperature-dependent yield strength of superalloys, incorporating a break temperature parameter to ensure reliable performance predictions across temperature regimes.

Contribution

The work presents a novel bilinear log model with a physically meaningful break temperature parameter, enhancing the prediction of superalloy strength over traditional black-box ML methods.

Findings

Model accurately predicts yield strength across temperature ranges.

The bilinear log model extends to high-entropy alloys.

Global optimization improves parameter fitting.

Abstract

In the pursuit of developing high-temperature alloys with improved properties for meeting the performance requirements of next-generation energy and aerospace demands, integrated computational materials engineering (ICME) has played a crucial role. In this paper a machine learning (ML) approach is presented, capable of predicting the temperature-dependent yield strengths of superalloys, utilizing a bilinear log model. Importantly, the model introduces the parameter break temperature, $T_{break}$, which serves as an upper boundary for operating conditions, ensuring acceptable mechanical performance. In contrast to conventional black-box approaches, our model is based on the underlying fundamental physics, directly built into the model. We present a technique of global optimization, one allowing the concurrent optimization of model parameters over the low-temperature and high-temperature regimes. The results presented extend previous work on high-entropy alloys (HEAs) and offer further support for the bilinear log model and its applicability for modeling the temperature-dependent strength behavior of superalloys as well as HEAs.