Risk evaluation of railway tunnel water inrush based on PCA-improved RBF neural network model

In order to accurately predict the risk level of water inrush in railway tunnel and reduce the risk of water inrush accident in the process of tunnel construction, combined with the relevant specifications, 13 factors were selected to build the evaluation index system based on the investigation and analysis of the risk factor set of water inrush disaster in tunnel. The principal component analysis was used to extract and reduce the dimension of water inrush risk evaluation index. Fuzzy c-means clustering algorithm was used to calculate the center of RBF neural network. Gradient descent method was used to modify the weight and variance. The principal component obtained after analysis was used as the input vector of the improved RBF neural network evaluation model. The PCA- improved RBF neural network based railway tunnel water inrush risk evaluation model was established. Finally, the prediction effect of the model is verified by combining with Tianxiushan tunnel. And the evaluation result was consistent with the actual situation. The case study shows that the model is reasonable and operable. Compared with other methods, it has higher accuracy, faster training and smaller mean square error. It provides anew way and reference for similar railway tunnels to prevent water inrush accidents. © 2021, Central South University Press. All rights reserved.