Intelligent Fault Detection for Tunnel Drainage Pumps: A Case Study Based on Random Forest

This study focuses on fault detection and maintenance planning for water pumps, aiming to minimize the adverse effects of pump failures on tunnel operational status. The proposed approach leverages high-and-low-frequency data from pump systems to diagnose faults from multiple perspectives. Unlike traditional monitoring methods that may miss subtle fault indicators, this study integrates high-and-lowfrequency data to construct feature subsets, enabling comprehensive monitoring and analysis of pump operational states. Practical attempts were carried out on the Changjiang Road Tunnel, where the fault probability is calculated, and maintenance planning recommendations for the pumps are provided. The research results offer a reliable maintenance strategy for tunnel management, promoting intelligent and personalized maintenance of water pumps. Recommendations include regularly using the model for fault detection, strategically adjusting pump rotation, and establishing a dynamic maintenance management system to ensure the reliability and safety of tunnel facilities.