A Mathematical Model Based on Artificial Neural Networks to Predict the Stability of Rock Slopes Using the Generalized Hoek–Brown Failure Criterion

Stability of rock slopes is a critical issue in many mining and civil engineering projects. The current state of practice for slope stability analysis is based on obtaining the factor of safety (FOS). Stability charts are widely used by engineers to obtain a FOS for a quick assessment of the initial stability of slopes. The stability of rock slopes with vertical walls in urban areas adjacent to existing structures is another important issue in this regard. However, the stability of earth or rock slopes are usually analyzed ignoring surcharge loads. The effect of adjacent structures (as surcharge load) on slope stability (considering these loads) can be very useful in slope stability analyses in urban or even non-urban areas. In the present study, it is tried to investigate the effect of surcharge load on the stability of rock slopes based on the generalized Hoek–Brown failure criterion using a finite element numerical software and the related charts are presented. Since there is a stability chart for each slope angle, a comprehensive mathematical model utilizing artificial neural networks is proposed to predict the stability factor of rock slopes. The independent variables in this study were slope angles, slope height, the intensity of surcharge load, Geological Strength Index (GSI), and unconfined compressive strength of the intact rock. Sensitivity analysis showed that the changes in GSI, effect of surcharge and unconfined compressive strength of rock have the highest effect on the slope stability assurance coefficient.