Research on prediction of PPV in open-pit mine used RUN-XGBoost model

Mingzhi Sun; Jiamian Yang; Chengye Yang; Weiping Wang; Xiaobing Wang; Hongfei Li

doi:10.1016/j.heliyon.2024.e28246

Research on prediction of PPV in open-pit mine used RUN-XGBoost model

Heliyon. 2024 Mar 20;10(7):e28246. doi: 10.1016/j.heliyon.2024.e28246. eCollection 2024 Apr 15.

Authors

Mingzhi Sun^{1

2}, Jiamian Yang^{1

2}, Chengye Yang³, Weiping Wang^{1

2}, Xiaobing Wang^{1

2}, Hongfei Li^{1

2}

Affiliations

¹ State Key Laboratory of Safety and Health for Metal Mines, Sinosteel Maanshan General Institute of Mining Research Co., Ltd., Maanshan, 243000, China.
² National Engineering Research Center of Huawei High Efficiency Cyclic Utilization of Metal Mineral Resources Co., Ltd., Maanshan, 243000, China.
³ School of Resources and Safety Engineering, Central South University, Changsha, Hunan, 410083, China.

Abstract

The drill-blasting method is a commonly used mining technique in open-pit mines, and the peak particle velocity (PPV) caused by blasting vibrations is an important indicator for evaluating the rationality of blasting mining design parameters. To develop an effective PPV prediction model, a parameter self-optimizing RUN-XGBoost prediction model is implemented using the Runge-Kutta optimization algorithm (RUN) combined with extreme gradient boosting (XGBoost). The factors affecting the prediction of PPV, including maximum explosive (ME), total explosive (TE), blast center distance (BCD), blast hole depth (BHD), and height difference between the measurement location and the blast location (DH), are selected as the influencing indicators. 188 pieces of blasting operation data were measured at the RK open pit copper-cobalt mine. Then, the RUN-XGBoost prediction model for PPV is studied and compared with the Sadovsky empirical formula, traditional XGBoost model, PSO-XGBoost model, and some traditional machine learning models (Ridge, LASSO, SVM, and SVR) using R², RMSE, VAF, MAE, and MBE as evaluation indicators for model prediction results. Finally, the Shapley Additive Explanations (SHAP) method is used to evaluate the contribution of different influencing indicators to the PPV prediction results. The results show that the RUN-XGBoost prediction model is significantly better than other machine learning models and the Sadovsky empirical formula in the prediction of PPV, further demonstrating that the RUN-XGBoost prediction model can handle the nonlinear features of multiple factors and provide a reliable, simple, and effective PPV prediction model, forming a rapid prediction and evaluation method for blasting vibrations in open-pit mining.

Keywords: Blasting works; Machine learning; Open-pit mines; PPV; XGBoost.