Prediction of Peak Shear Strength of Rock Joints Based on Back-Propagation Neural Network

Abstract

The shear strength model, a predictive method for effectively characterizing the shear strength of joints, can be used to evaluate the stability of the rock mass. However, the traditional shear model is difficult to apply due to its complicated form. Considering the complicated mapping relationship between joint shear strength and influencing factors, this study combined the back-propagation (BP) neural network to propose a new model for predicting the shear strength of rock joints, which can comprehensively consider various influence factors, including external shear test conditions and surface morphology of joint itself. Direct shear tests of granite joints were carried out to verify the proposed model, and the results showed that the outputted peak strengths training by the BP neural network match well with the measured values. At last, a comparison of the proposed model with Grasselli’s model and Xia’s model showed that the overall prediction error based on the proposed model is smaller and more accurate. It is seen that the BP neural network prediction model has a reliable estimate of the peak shear strength for rock joints.