Scraper conveyor gearbox fault diagnosis based on multi-source heterogeneous data fusion

Gearboxes are critical components of scraper conveyors, and owing to their complex working environment and fluctuating loads, they are prone to faults, such as gear wear, tooth breakage, and insufficient lubrication. Accurate fault diagnosis is crucial for preventing fault escalation, reducing downtime, and minimizing maintenance costs, as it allows for intervention before faults develop, thereby improving system reliability and efficiency. Existing fault diagnosis methods for scraper conveyor gearboxes include signal-processing techniques, model- based methods, machine-learning approaches, and deep-learning methods. This study proposes a fault diagnosis method for a scraper conveyor gearbox based on multi-source heterogeneous data fusion. First, variational mode decomposition (VMD) and modal selection criteria were applied to denoise the vibration and electrostatic signals of gears with different fault modes. The Gramian angular field (GAF) is subsequently used to convert onedimensional vibration and electrostatic signals into two-dimensional images, which are combined with infrared images of the gearbox to construct a dataset that reveals subtle anomalies, particularly in the early stages of faults. Finally, a convolutional neural network (CNN) with a convolutional block attention module (CBAM) was employed to extract multiscale features and enhance sensitivity to early fault patterns. The experimental results show that the method achieves an accuracy of 99.4% under complex conditions, significantly improving the accuracy and robustness of gearbox fault diagnosis and outperforming traditional methods in fault classification.