Accelerating Materials Performance Evaluation and Discovery: Prediction of Steel Structures for High-Throughput Design and Testing

Author(s):

Heather Murdoch, Michael Rupinen, and Krista Limmer

Abstract:

The Accelerating Materials Performance Evaluation and Discovery (AMPED) program was a 12-month cross-service program to build foundational elements for rapid, automated materials development and characterization. This report documents the ARL-led contributions to the AMPED program. Several modeling approaches to predict the categorical structure of a given steel based on composition and processing were evaluated with two datasets constructed from open literature. Significantly, these modeling approaches do not require commercial software and have low computational expense, enabling integration into accelerated design workflows including predictive modeling tools and guiding high-throughput experimental strategies. A new random forest classifier model was trained that was highly successful at predicting the structure of a steel, given composition and processing.