Estimation of tunnel axial orientation in the interlayered rock mass using a comprehensive algorithm

Abstract

The axial selection of tunnels constructed in the interlayered soft-hard rock mass affects the stability and safety during construction. Previous optimization is primarily based on experience or comparison and selection of alternative values under specific geological conditions. In this work, an intelligent optimization framework has been proposed by combining numerical analysis, machine learning (ML) and optimization algorithm. An automatic and intelligent numerical analysis process was proposed and coded to reduce redundant manual intervention. The conventional optimization algorithm was developed from two aspects and applied to the hyperparameters estimation of the support vector machine (SVM) model and the axial orientation optimization of the tunnel. Finally, the comprehensive framework was applied to a numerical case study, and the results were compared with those of other studies. The results of this study indicate that the determination coefficients between the predicted and the numerical stability evaluation indices (STIs) on the training and testing datasets are 0.998 and 0.997, respectively. For a given geological condition, the STI that changes with the axial orientation shows the trend of first decreasing and then increasing, and the optimal tunnel axial orientation is estimated to be 87°. This method provides an alternative and quick approach to the overall design of the tunnels.