Abstract: In view of the complex operation conditions and high failure rate of mining belt conveyor, the existing fault diagnosis method has the shortcomings of low efficiency and poor accuracy. In order to effectively improve the fault diagnosis efficiency of belt conveyor, based on the main structure of belt conveyor and combined with its common fault phenomena, the application of AI technology is designed to design the fault diagnosis system of belt conveyor. This paper specifically carried out vibration spectrum feature extraction based on multi-source sensing data, built a fault classification model integrating convolutional neural network and time series analysis, designed a dynamic belt tear detection algorithm based on machine vision, built an intelligent diagnosis architecture based on edge computing and cloud collaboration, verified the adaptability of multi-modal data fusion to complex working conditions, and optimized the solver of fault location Interpretation model visualization module, not only that, but also combined with specific application cases, from the belt conveyor abnormal vibration, abnormal image diagnosis effect analysis of the application effect, the final results show: The application of AI technology based belt conveyor fault diagnosis system can diagnose the belt conveyor fault in a timely and efficient manner, quickly locate the fault part, shorten the fault handling time of the belt conveyor, and ensure the safe, stable and efficient operation of the belt conveyor.