An AI-powered approach to improving tunnel blast performance considering geological conditions

In this study, a hybrid model was developed to improve blast performance during tunnel excavation, which is based on a multilayer perceptron (MLP) and incorporates meta-heuristic algorithms. Prior to optimization, precise prediction of four post-blasting indicators is important, including overbreak area, maximum linear overbreak, crown settlement, and maximum fragment size. To identify most suitable model with the highest prediction accuracy, a comparison was conducted among three meta-heuristic algorithms, namely Particle Swarm Optimization (PSO), Whale Optimization Algorithm (WOA), and Archimedean Optimization Algorithm (AOA), in terms of their capabilities to optimize MLP hyperparameters. Since relying solely on predictions for adjusting blasting parameters is not an efficient approach, and the optimization problem of blasting parameters is consistent with the characteristics of the meta-heuristic algorithm, the optimization of blasting parameters can be performed via the combination of the optimal prediction model and meta-heuristic algorithm. Accordingly, the improved PSO method for optimizing blasting parameters can effectively reduce the overbreak areas considering the infinite number of possible combinations of blasting parameters. Compared to the previous method, overbreak areas in different sections of the tunnel were ultimately reduced by 56.1% and 35.6%, respectively. Furthermore, an importance analysis of various parameters controlling overbreak was quantitatively conducted using the developed model in combination with machine learning interpretability analysis, confirming geological conditions as the most influential factor on overbreak. Finally, validation and recommendations regarding the perimeter borehole spacing to burden ratio, which typically falls within the range of 0.6 to 1.0, are proposed. The results demonstrate that the developed model effectively improves tunnel blast performance and serves as a valuable reference for blast design.

Automated summary

In this study, a hybrid model based on a multilayer perceptron (MLP) and meta-heuristic algorithms was developed to improve blast performance during tunnel excavation. Precise prediction of post-blasting indicators was important for optimization, and a comparison of meta-heuristic algorithms was conducted to find the most suitable model. The results showed that the developed model effectively reduces overbreak areas and quantitatively analyzes the influence of geological conditions.