Explainable Machine Learning Model for Predicting Molten Steel Temperature in LF Refining Process

The accurate prediction of molten steel temperature in the ladle furnace (LF) refining process has an important influence on the quality of molten steel and the control of steelmaking cost.. Extensive research has been conducted on establishing models to predict molten steel temperature. However, most researchers focus solely on improving the accuracy of the model, neglecting its explainability. This study aimed to develop a high-precision and explainable model and improve reliability and transparency of model. The eXtreme Gradient Boosting (XGBoost) and Light Gradient Boosting Machine (LGBM) were utilized, along with Bayesian optimization and Grey Wolf Optimization (GWO), to establish the prediction model. Different performance evaluation metrics and graphical representations were used to compare the optimal XGBoost and LGBM models obtained through different hyperparameter optimization methods with the other models. The findings indicated that the GWO-LGBM model outperformed other methods in predicting molten steel temperature, with a high prediction accuracy of 89.35% within the error range of ±5℃. Based on the tree structure visualization and SHapley Additive exPlanations analysis, the model's learning/decision process was revealed and the degree of influence of different variables on the molten steel temperature was clarified, which enhanced the explainability of the optimal GWO-LGBM model and provided reliable support for prediction results.