Abstract: This study aims to solve the key problems of poor real-time perception, insufficient connectivity and lack of predictive maintenance capability of the traditional operation and maintenance system for belt conveyors in mines. To this end, an intelligent O&M system based on the digital twin five-dimensional model is proposed and constructed. The method firstly constructs a multi-dimensional virtual model integrating geometry, physics, behavior and rules, and realizes high-fidelity real-time synchronization between physical data and virtual model based on EMQX server; then, an intelligent controller integrating BP neural network and PID is designed, and the parameters are dynamically adjusted on-line to improve the quality of control; finally, an intelligent platform integrating monitoring, diagnosis and decision-making is developed to form the “sensing-analysis-decision-making system”, and the “sensing-analysis-decision-making system” is designed. Finally, an intelligent platform integrating monitoring, diagnosis and decision-making is developed to form a closed loop of “sensing-analysis-decision-making-control”. The experimental results show that the system improves the response speed of deskew control by 33% on average and reduces the steady-state error by about 50%. This study realizes the new operation and maintenance paradigm of “accurate mapping-real-time synchronization-intelligent decision-making-closed-loop control”, which provides an effective solution for unmanned operation and maintenance of conveyor belt systems.