The Controlling of Longitudinal Cracks Defect Based on BO-XGBoost Model

This article delves into the application of machine learning models, particularly the BO-XGBoost algorithm, to predict and control longitudinal cracks in continuous casting slabs. The study identifies key features such as limestone addition, mold flux, and calcium carbide as critical factors influencing crack formation. Through feature importance ranking using Gini coefficient and SHAP values, the research provides a mechanistic analysis of these features and their impact on longitudinal cracks. The article also discusses the optimization of hyperparameters using Bayesian and Grey Wolf optimization algorithms, enhancing the model's predictive performance. Practical implementation of the model in industrial settings resulted in a significant improvement in product qualification rates, demonstrating the model's robustness and generalization capability. Readers will gain insights into the application of advanced machine learning techniques in metallurgical processes, the identification of key process parameters, and the practical benefits of optimizing these parameters to enhance product quality.