Abstract:

A prediction model of tunnel water inflow is established based on correlation coefficient method and extreme learning machine to achieve highprecision prediction of tunnel water inflow. Firstly, the influencing factors of tunnel water inflow are analyzed based on engineering practices, and the correlation coefficient method is used to analyze the correlation between each factor and water inflow so as to screen out the important influencing factors. Secondly, the selected important factors are used as the input layer of prediction model, the model parameters of extreme learning machine are optimized by trial algorithm and empirical formula, and then the prediction error is weakened by M estimation to construct RELM model for predicting tunnel water inflow. The study results show that: (1) There are many factors affecting the water inflow disaster in karst tunnels, including 5 primary factors and 12 secondary factors, and the influence degree of different factors on the water inflow disaster varies. (2) The average relative error of RELM model prediction results is only 1.12%, which shows that the model has high prediction accuracy and verifies the validity of model parameter optimization and M estimation optimization and the applicability of the model in the prediction of tunnel water inflow.

Key words: tunnel water inflow, correlation coefficient method, extreme learning machine, M estimation, RELM model, water inflow prediction