About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Characterization of Minerals, Metals and Materials 2026 - In-Situ Characterization Techniques
|
| Presentation Title |
Machine Learning Prediction and SHAP-Based Feature Analysis of Hydrogen Embrittlement Susceptibility in Pipeline Steels for Hydrogen Energy Systems |
| Author(s) |
Xin Fan, Hongchao Yang, Frank Cheng |
| On-Site Speaker (Planned) |
Xin Fan |
| Abstract Scope |
Hydrogen embrittlement (HE) remains a critical barrier to the safe deployment of hydrogen energy infrastructure, restricting the range of suitable structural materials. This study proposes an integrated machine learning (ML) framework that predicts HE susceptibility in pipeline steels by combining experimental results from slow strain rate tests, fatigue crack growth, and fracture toughness measurements in gaseous hydrogen environments. A curated dataset from peer-reviewed sources and open repositories covers API grades X42–X100 with diverse microstructures and test conditions. Regression models, including Random Forest, Gradient Boosting, and Neural Networks, were trained to estimate HE indices and threshold parameters. SHAP-based feature analysis quantified relative contributions of microstructure, hydrogen charging pressure, and strain rate to HE susceptibility. This approach highlights the capability of ML–assisted material screening to expedite the discovery and optimization of hydrogen-compatible steels for next-generation energy systems. |
| Proceedings Inclusion? |
Planned: |