Machine learning-based stability assessment of unlined circular tunnels under surcharge loading

This study investigates the stability of an unlined circular tunnel subjected to surcharge loading over the rock mass surface. A plane strain analysis has been performed following the Generalized Hoek–Brown (GHB) failure criterion by adaptive finite element limit analysis (AFELA). An extensive parametric study was conducted to investigate the impact of cover depth (C), diameter of the tunnel (D_t), geological strength index (GSI), GHB material constant for intact rock (m_i), and unit weight (γ) on the stability of the tunnel. A dimensionless stability number (N) has been evaluated and presented that the geotechnical engineers in the field can utilize for preliminary investigation. Furthermore, the influence of the abovementioned parameters, along with the location of the tunnel, has been investigated on the development of the critical failure planes. The numerical analysis was performed using a finite element method-based computational tool, and then a machine learning-based technique, Support Vector Machine (SVM), was utilized to obtain a predictive model for stability number (N). The R² values of the training and testing sets were found to be 0.996 and 0.949, respectively. The SVM model was validated with a tenfold cross-validation method. The sensitivity analysis of the input parameters showed that the GSI of rock mass highly influences the N; whereas the C/D_t ratio has the least influence.