Abstract: Drums and idlers are core components that bear the main load of belt conveyors and continuously perform rotational motion, and their health condition directly determines the operational efficiency and reliability of the entire belt conveyor system in coal mines. Focusing on key rotating mechanical components such as drums and idlers of coal mine belt conveyors, this paper systematically explains the typical fault types that are prone to occur under harsh roadway working conditions in coal mines and the corresponding fault monitoring methods, and analyzes the monitoring principles and technical routes based on vibration and sound signals. From three core aspects including vibration and sound signal preprocessing, feature extraction, and fault identification for rotating components of coal mine belt conveyors, the research progress in fault diagnosis is compared and reviewed. Research on vibration and sound signal preprocessing shows a development trend toward adaptive optimization of fixed parameters and the integration of multiple methods. The feature extraction methods show a trend from traditional methods to adaptive learning and from single methods to the integration of multiple methods. The fault identification methods show a trend from traditional machine learning models with simple structures to deep learning models. The main challenges encountered in the field of health monitoring and fault diagnosis of underground coal mine belt conveyors are summarized, including poor preprocessing performance of vibration and sound signals in harsh environments, insufficient feature extraction capability of single-signal perception methods under complex working conditions, the scarcity of underground coal mine fault samples, and insufficient generalization ability of fault diagnosis models. Future research and application of fault diagnosis technology for underground coal mine belt conveyors should focus on intelligent adaptive preprocessing methods for harsh underground environments, the development of multi-source monitoring and information fusion technologies based on in-depth understanding of fault mechanisms, and the exploration of new intelligent fault identification methods incorporating small-sample learning and enhanced generalization.