Minimization of overbreak in different tunnel sections through predictive modeling and optimization of blasting parameters

Engineering projects are confronted with many problems resulting from overbreak in tunnel blasting, necessitating the optimization of design parameters to minimize overbreak. In this study, an AI-based model for overbreak prediction and optimization is proposed, aiming to mitigate the hazards associated with overbreak. Firstly, the Extreme Gradient Boosting (XGBoost) model is integrated with three distinct metaheuristic algorithms, namely Particle Swarm Optimization (PSO), Whale Optimization Algorithm (WOA), and Sparrow Search Algorithm (SSA), respectively. Consequently, the hyperparameters are optimized, and the performance of predictions is enhanced. Meanwhile, to overcome the limitations of a small dataset and enhance the generalization ability of the three developed models, a 5-fold cross-validation is employed. Then, the performance of the different models with five distinct swarm sizes is evaluated via four metrics, including coefficient of determination (R2), mean square error (MSE), mean absolute error (MAE), and variance accounted for (VAF). Subsequently, by comparing the aforementioned developed models, the optimal prediction model with the highest accuracy can be obtained, which is then used for parameter optimization research. Finally, individual studies are conducted to address the issue of overbreak caused by the adoption of identical blasting parameters due to geological variations, aiming to minimize overbreak in different sections of the tunnel. By comparing the optimization abilities of PSO, WOA, and SSA, the objective of finding the minimum value of overbreak within a short timeframe is achieved. The results indicate that the model developed in this study accurately predicts overbreak, and effectively optimizes blast parameters for different sections of the tunnel.

Automated summary

In this study, an AI-based model was proposed for predicting and optimizing overbreak in engineering projects. By integrating the Extreme Gradient Boosting (XGBoost) model with three different metaheuristic algorithms, the performance of predictions was enhanced and hyperparameters were optimized. Cross-validation was employed to overcome the limitations of a small dataset and improve the generalization ability of the models. The optimal prediction model was then used to optimize blast parameters for different sections of the tunnel.