Predicting the rock cutting performance indices using gene expression modeling

In this study, a machine learning approach was employed to optimize the rock cutting process for construction practices and geotechnical engineering applications. The primary objective was to develop predictive models for granite rock cutting performance indices, specifically cutting depth and cutting width, using an Abrasive Water Jet (AWJ) system. Initially, an extensive dataset was generated through multiple rock cutting experiments on granite specimens using AWJ. Subsequently, the Gene Expression Programming (GEP) technique was utilized to create new mathematical models for predicting rock cutting depth and width. These models incorporated various influencing input variables, including water pressure, flow rate, orifice diameter, abrasive feed rate, traverse speed, and standoff distance. The accuracy of the predictive models was evaluated using several error metrics: R-square (R²), root mean square error (RMSE), mean absolute error (MAE), and relative standard error (RSE). The proposed models demonstrated high R² values of 0.94 for both training and testing datasets when predicting cutting depth, and 0.95 (training) and 0.92 (testing) for predicting cutting width. Additionally, the RMSE, MAE, and RSE values were consistently low, underscoring the models' strong performance. Furthermore, sensitivity and monotonicity analyses were conducted to evaluate the robustness of the proposed models, and the impact of individual parameters on cutting depth and width was analyzed. These robust models hold significant potential for practical application in real-time engineering scenarios, ultimately enhancing rock cutting performance.