Fault detection algorithm and application for belt conveyors based on DAS system

Identifying faults in belt conveyors within high-noise industrial environments remains challenging due to the limitations of traditional methods, which are labor-intensive, sensitive to noise, and unable to provide real-time early detection. To address these challenges, this study proposes a novel hybrid framework that combines multimodal signal decomposition [short-time Fourier transform (STFT), continuous wavelet transform (CWT), empirical mode decomposition (EMD), variational mode decomposition (VMD)] with convolutional neural networks (1D/2D-CNN), marking the first application of this integration in distributed acoustic sensing (DAS)-based fault diagnosis of belt conveyors. The main innovation lies in the adaptive fusion of VMD and CNN, where VMD effectively isolates fault-related modes from high-noise backgrounds, whereas the CNN architecture simultaneously extracts temporal and spectral features from both raw signals and time-frequency representations. Experimental results show that our method achieves a classification accuracy of 99.75% under high-noise conditions, surpassing traditional single-model approaches by 11.5%, and enabling real-time detection. This advancement offers a robust solution for intelligent monitoring in complex industrial environments.