A NOVEL INTELLIGENT APPROACH FOR PREDICTING WEAR OF EXCAVATOR BUCKET TEETH BASED ON HYBRID ICA-XGBOOST MODEL

Abstract

The wear of the excavator bucket teeth is a significant operational issue in the mining industry, negatively impacting material production and increasing costs. Accurate prediction of this wear is crucial in mitigating its effects. This study introduces, for the first time, a hybrid intelligent approach that leverages four metaheuristic optimisation algorithms: Imperialistic Competitive Algorithm (ICA), Grasshopper Optimisation Algorithm (GOA), Antlion Optimiser (ALO), and Artificial Fish Swarm Algorithm (AFSA), to optimise the parametric weights of the Extreme Gradient Boosting (XGBoost) technique, enhancing its ability to predict excavator bucket tooth wear. The hybrid models, named ICA-XGBoost, GOA-XGBoost, ALO-XGBoost, and AFSA-XGBoost, were developed using a data set of 579 wear records from a surface mine in Ghana. Furthermore, the study implemented standalone models, such as XGBoost, Gradient Boosting Regressor (GBR), Random Forest (RF), and Categorical Boosting Regressor (CatBoost), to benchmark the performance of the various hybrid XGBoost models. To evaluate the predictive performance of these models, six statistical metrics were employed: Variance Accounted for (VAF), Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Square Error (MSE), Correlation Coefficient (R), and Coefficient of Determination (R²). In addition, the Akaike Information Criterion (AIC) was used to identify the most effective model. The results demonstrated that ICA-XGBoost outperformed the other models, achieving the lowest MAE, RMSE and MSE values (0.065536, 0.050359, and 0.002536, respectively) as well as the highest VAF, R, and R² values (99.9997, 0.999998, and 0.999996, respectively). Furthermore, ICA-XGBoost exhibited the lowest AIC value (-447.060), confirming its superior predictive capability for the wear of excavator bucket teeth. Finally, after evaluating the various models, a sensitivity analysis was performed to assess the influence of the input variables.