Abstract: In view of complex operating conditions and high failure rate of mine belt conveyors, and the low efficiency and insufficient accuracy of existing fault diagnosis methods, this paper designs an intelligent fault diagnosis system based on the main structure and common fault types of belt conveyors by applying artificial intelligence technology. Specifically, vibration spectrum feature extraction is carried out based on multi-source sensor data, a fault classification model integrating convolutional neural networks and time series analysis is constructed, a dynamic detection algorithm for belt tearing based on machine vision is designed, and an intelligent diagnosis architecture combining edge computing and cloud collaboration is established. Meanwhile, adaptability of multi-modal data fusion to complex working conditions is verified, and the interpretable visualization module for fault location is optimized. Combined with practical application cases, system performance is analyzed from abnormal vibration recognition and image diagnosis effects. The results show that the proposed AI-based fault diagnosis system can realize rapid and accurate identification and location of belt conveyor faults, significantly shorten fault handling time, and thus effectively ensure safe, stable and efficient operation of the equipment.