Prediction of the Tricone Rotary Bits Wear by Combining Principal Component Analysis and Regression Methods in Blast Holes of Sarcheshmeh Copper Mine

The first production cycle in open-pit mines is the drilling of blast holes, which is a major part of the exploitation costs. Bits are considered as one of the most important parts of drilling in operational processes due to the type of application and high costs. Examining the parameters affecting the bits wear can reduce the effect of wear, prevent wasting time and additional costs of the drilling process. Therefore, in this study, first, by determining the effective parameters based on the results of univariate regression and Principal Component Analysis (PCA), the relationship between these factors and the tricone rotary bits wear has been determined through statistical methods. Then, 100 linear and nonlinear models were examined to predict the rate of the tricone rotary bits wears during drilling, of which only 21 models, including 6 linear models and 15 nonlinear models, were approved. The performance of these models is evaluated based on the root mean square error (RMSE), coefficient of determination (R2), Variance accounted for (VAF), and mean absolute percentage error (MAPE). then the priority of each method is determined based on performance evaluation and prioritization strategies. In this research, three strategies of prioritization, the Rank average method, the Borda method, the Copeland method, and finally the Aggregate of these methods have been used. The ranking results show that the best model for predicting the tricone rotary bits wear is the regression model with independent variables of Weight on Bits (WOB), Uniaxial Compressive Strength (UCS), and Geological Strength Index (GSI). According to the results of sensitivity analysis, the Rij value of the Weight on Bits (WOB) is 0.982, which is more than the other two parameters of the final model. Therefore, this parameter has the greatest effect on the tricone rotary bits wear and it is the most important parameter of the final model.