Research on Fault Diagnosis of Scraper Conveyor Chains Based on Multi-source Data Fusion

As the core conveying equipment in coal mining faces, scraper conveyors have a direct impact on the production efficiency and safety of coal mines. With the development of coal mining faces towards long-distance transportation and high power, higher requirements are placed on the stability and reliability of scraper conveyors. The chain drive system is exposed to extreme mechanical environments for extended periods, which can easily lead to chain fatigue, fracture, and other failures.To grasp the real-time operating status of the scraper conveyor chain in the harsh underground environment and avoid the impact of chain failures on coal mine safety production, a multi-source heterogeneous data fusion method is adopted for monitoring and fault diagnosis of the scraper conveyor chain. This approach breaks through the limitations and one-sidedness of current single detection methods in terms of time and space for judging chain failures. The force conditions at various positions of the chain during the operation of the scraper conveyor are analyzed, and the force on a single chain link is studied using finite element analysis to identify the locations where the chain is prone to failure. Emerging technologies are combined with traditional technical means to collect data related to chain state detection, forming the data sources of the system. This includes four methods: AI video recognition of chain state, magnetic induction sensor detection of chain breakage, intelligent chain link force detection, and monitoring of motor torque mutations. The advantages and disadvantages of each detection method in chain state detection are analyzed based on their characteristics. Data analysis is performed using neural networks and big data technology, taking into account both local and long-term sequence dependencies of scraper conveyor chain monitoring data. The CNN module is used to extract local key features from the data, and the ABiLSTM module captures bidirectional dependencies in time-series data. The integration of the two modules achieves complementarity and enhancement of multi-dimensional features, improving the accuracy of chain state classification under complex working conditions. A scraper conveyor chain fault diagnosis and predictive maintenance system based on multi-source data fusion is constructed, effectively mining deep correlation features in multi-source monitoring data of the scraper conveyor to achieve high-precision identification of the chain's operating state. This verifies the effectiveness and feasibility of the algorithm in the engineering application of scraper conveyor chain fault diagnosis.