Drilling, as one of the key phases in mining and civil engineering activities, plays a decisive role in overall project productivity, cost reduction, and time optimization. Given the complexities involved in drilling operations, identifying and adjusting operational parameters to increase the drilling penetration rate is considered one of the primary challenges in the mining industry. In this research, a comprehensive framework based on machine learning models and optimization algorithms has been designed and implemented to improve the drilling penetration rate at the Shahid Hojaji mine with limited data. Initially, due to the limitation of mining data, related datasets from the Sarcheshmeh Copper Mine, which had similar geological and operational conditions, were used. Three prediction models, including the classification boosting algorithm, natural gradient boosting algorithm, and random forest algorithm, were developed to predict the drilling penetration rate based on three key input parameters: bit pressure, compressed air pressure, and geological strength index. Model eva‎luations indicated that the classification boosting model performed better than the other two models, achieving a high coefficient of determination and low error . In the next step, Bayesian optimization was utilized to optimize operational parameters and identify their optimal values to maximize the drilling penetration rate. This algorithm, using the classification boosting model as the objective function, identified optimal values for operational parameters that maximized operational productivity conditions. These values were then applied to the actual data from the Shahid Hojaji mine, and results showed that the model matched the new data with high accuracy. eva‎luation results indicated that the model achieved a high coefficient of determination and low error . One of the major challenges in this research was the lack of a geological strength index for the Shahid Hojaji mine data. To address this issue, mining data was classified into four geological strength ranges: low, medium, high, and very high. The Bayesian optimization algorithm was then applied to each of these ranges, and optimal operational conditions were extracted separately for each range. For low geological strength data, the best drilling penetration rate was 0/794 meters per minute when the bit pressure was 68 bar and the compressed air pressure was 7 bar. For medium geological strength data, the best drilling penetration rate was 0/856 meters per minute when the bit pressure was 68 bar and the compressed air pressure was 7 bar. For high geological strength data, the best drilling penetration rate was 0/611 meters per minute when the bit pressure was 70 bar and the compressed air pressure was 7 bar. For very high geological strength data, the best drilling penetration rate was 0/401 meters per minute when the bit pressure was 64 bar and the compressed air pressure was 7 bar. These classification and optimization results indicated that optimal operational conditions varied significantly in each range and had a direct impact on the drilling penetration rate. This research, by combining advanced machine learning models, optimization techniques, and uncertainty analysis, provided a comprehensive framework for optimizing drilling operations in mines. Besides improving the drilling penetration rate, this approach demonstrated that using intelligent methods to manage operational parameters can help reduce costs, increase safety, and enhance overall process productivity. This framework is not only applicable to similar mines but also holds significant potential for further development and application in other industrial processes. This research is a significant step toward leveraging intelligent tools to solve operational challenges in mines and can serve as a basis for future research in related fields.

راحب باقرپور

حميد كلهري

هادي حسيني , حسن طباطبائي