Classification of Rocks Surrounding a Tunnel Based on Factor Analysis and Fisher Discriminant Analysis

Abstract

To forecast the tunnel surrounding rock category quickly and effectively and to enhance the stability of underground engineering and security, we apply theory of factor analysis and Fisher discriminant analysis. In addition, six indicators, namely, rock quality, integrity, saturated uniaxial compressive strength, longitudinal wave velocity, elastic resistance coefficient, and structure surface friction factor, were selected as discriminant factors in Fisher’s discriminant analysis. A Fisher prediction model based on factor analysis was built to predict the tunnel surrounding rock category. Thirty groups of tunnel surrounding rock data in the survey site were used as learning samples for the training. The resubstitution method was used to test the model, which yielded a 96.7% accuracy. The established discriminant model was used in an engineering application and used six sets of engineering data as test samples to forecast the classification of tunnel surrounding rock. We also compared this model simultaneously with the neural network and Bayes methods. The factor analysis can effectively extract the surrounding rock classification index and remove the redundant factors. Fisher’s discriminant model based on factor analysis can effectively predict the tunnel surrounding rock category with 100% prediction accuracy.