Yield Stress and Elongation Prediction and Process Parameter Optimization of Two C-Mn Steel Based on XGBoost and Non-dominated Sorting Genetic Algorithm-IIization of C-Mn steel based on XGBoost algorithm

Abstract

Mechanical properties are the key guidance for controlling the product quality of steel and optimizing its process. In this work, four machine learning (ML) algorithms (XGBoost, SVR, BP, and RBF) are used to establish a correlation model between chemical composition, process parameters, and mechanical properties for performance optimization. The ML model showing high prediction accuracy is selected to analyze the importance of the model parameters. The XGBoost model has the highest prediction accuracy, with both yield strength (YS) and elongation-to-failure (EL) achieving a prediction accuracy of over 0.9 on the test set. Subsequently, the (YS) and (EL) of typical steel grades SPHC and Q235B are optimized by combining the high-precision XGBoost model and NSGA-II multi-objective optimization strategy. A comprehensive mechanical performance evaluation index (CP and EV) was proposed to screen the optimal Pareto frontiers, and two types of steel with excellent performance were successfully selected.

Keywords

• c-mn steel
• machine learning
• mechanical properties
• xgboost
• nsga-ii