Machine learning approach for prediction of hydrogen environment embrittlement in austenitic steels

Abstract

This study introduces a machine learning approach to predict the effect of alloying elements and test conditions on the hydrogen environment embrittlement (HEE) index of austenitic steels for the first time. The correlation between input features and the HEE index was analyzed with Pearson's correlation coefficient (PCC) and Maximum Information Coefficient (MIC) algorithms. The correlation analysis results identified Ni and Mo as dominant features influencing the HEE index of austenitic steels. Based on the analysis results, the performance of the four representative machine learning models as a function of the number of top-ranked features was evaluated: random forest (RF), linear regression (LR), Bayesian ridge (BR), and support vector machine (SVM). Regardless of the type and the number of top-ranking features, the RF model had the highest accuracy among various models. The machine learning-based approach is expected to be useful in designing new steels having mechanical properties required for hydrogen applications.

Keywords

• Austenitic steel
• Hydrogen environment embrittlement
• Correlation analysis
• Machine learning
• Random forest model