A Novel Hybrid Intelligent Approach to Assess Blasting-Induced Overbreak Incorporating Geological Conditions in Different Tunnel Sections

Overbreak induced by tunnel blasting is a harmful phenomenon. Accurate assessment of overbreak can effectively reduce investment and ensure operational safety. In this study, a hybrid intelligent model for assessing blasting-induced overbreak is proposed which can accurately predict overbreak and effectively evaluate the importance of feature parameters. To ensure accurate prediction of overbreak, hyperparameters of four machine learning algorithms are optimized using a whale optimization algorithm. Their performance is compared based on three regression metrics: R2, RMSE, and VAF. Given the limitations of traditional feature importance analysis methods, the Shapley Additive Explanation method is used in conjunction with the random forest algorithm. After accurately predicting overbreak caused by different sections of the tunnel, the impact of each input parameter on overbreak is analyzed, and recommendations for design values of certain significant parameters are provided. The research indicates that the proposed method can accurately predict overbreak caused by actual engineering blasts and provide insights into the selection of design parameter values.