Intelligent prediction model for water inrush risk in RF water-rich tunnel based on AHP improvement

Abstract To prevent large-scale water inrush accidents during the excavation process of the rich water tunnel, a method based on AHP improved RF for intelligent risk prediction of the water-rich tunnel is proposed. By analyzing the influence of geological and hydrological conditions, design factors, and construction factors, 13 factors were selected as evaluation indicators for the risk of water inrush in the water-rich tunnel, including stratum lithology, poor geology, rock inclination, negative topographic area ratio, surrounding rock grade, hydrodynamic zoning, tunnel length, tunnel burial depth, tunnel section area, advanced geological prediction, excavation method, advance support, and monitoring measurement; Through statistical analysis of a large number of accident cases, a dataset of water inrush accidents in the water-rich Tunnel was established and preprocessed. Using the RF model in machine learning, the weights of each indicator in the RF model are calculated through the application and parameter optimization of the dataset. Then, the weights are optimized through AHP and imported into the RF model to obtain the improved RF-AHP model. Compared with the test set prediction results of the RF model's RF-AHP model, the accuracy of the RF-AHP model reaches 98%, which is better than the RF model's 96%. This indicates that the performance of the improved RF model based on AHP has been improved, and it has good performance in predicting the risk of water inrush in the water-rich tunnel, providing a new means for predicting the risk of water inrush in the water-rich tunnel.