Optimizing Tunnel Excavation: Intelligent Algorithms for Accurate Overbreak Prediction

Abstract

Excavating tunnels has become a widespread practice in the modern world, driven by the need for efficient transportation, subterranean storage, and mineral supply. One challenge encountered during tunnel excavation is the overbreak (OB) phenomenon, particularly prominent when utilizing drilling and blasting techniques. OB poses a risk by increasing operational expenses and compromising workplace safety. Therefore, accurately predicting the occurrence of OB during tunnel excavation is crucial. While various methods exist to forecast OB, traditional approaches like experimental, analytical, numerical, and regression methods face limitations due to uncertainties in geological and geotechnical parameters. In this paper, the use of Teaching–Learning-Based Optimization (TLBO) and Firefly (FF) algorithms is proposed to predict OB, aiming to fully comprehend the physical and mechanical characteristics of the rock mass while considering uncertainties and optimizing project completion in terms of cost and time. The model was constructed using data from three case studies: an Indian mine; the Azad tunnel on the Tehran-North route in Alborz, Iran; and the underground coal mine Tarzareh, comprising 217 data points. Parameters affecting the OB phenomenon in this study include rock mass rating (RMR), specific drilling (SD), perimeter holes powder factor (PPF), and spacing to burden ratio of contour holes (S/B). The dataset was divided into two groups: 80% for training the model and 20% for testing the relationship. To evaluate the model, statistical indices such as squared correlation coefficient (R²), root mean square error (RMSE), and mean square error (MSE) were used. The validation results indicated that the TLBO and FF algorithms performed satisfactorily, demonstrating high accuracy and low error. This suggests that engineers, scientists, and practitioners can benefit from employing intelligent approaches in mining and rock mechanics-related operations, utilizing the accurate model generated by these algorithms.