Abstract
Optimization has become a central concern in geotechnical engineering with increasing constraints on energy resources and the rising demand for cost-effective operations. Drilling, as a critical and energy-intensive component of mining and tunneling (particularly in transportation infrastructure), requires efficient and intelligent performance strategies. Monitoring While Drilling (MWD) provides a promising approach for real-time acquisition of drilling conditions. Recent advancements, including the integration of Acoustic Emission Technique (AET) with artificial intelligence (AI), enhance data-driven modeling and predictive analysis of drilling performance. In this study, vibroacoustic signals and drilling parameters were analyzed to predict penetration rate (PR) using three machine learning models: Artificial Neural Network (ANN), Random Forest (RF), and Support Vector Regression (SVR). Comparative evaluation showed that all three models achieved reliable predictive accuracy, with ANN reaching R(2) = 0.744, MAPE = 36.98%, RMSE = 0.161; RF yielding R(2) = 0.816, MAPE = 31.54%, RMSE = 0.142; and SVR attaining R(2) = 0.808, MAPE = 29.52%, RMSE = 0.141. The results demonstrate the feasibility of integrating vibroacoustic monitoring with AI-driven models for accurate PR prediction. This approach supports real-time decision-making, enhances drilling efficiency, and promotes sustainable practices in both underground and surface excavation projects.