High-Speed Elevator Fault Generation and Diagnosis Method: A Scalable Model-Driven Approach With Variational-Mode-Decomposition Graph Convolution Network

Elevator faults can cause significant harm, especially in high-speed elevators. However, obtaining fault samples from in-service high-speed elevators by monitoring sensors is challenging due to high simulation costs and limited efficiency, which hinder the development of fault diagnosis. To address this, a scalable fault simulation model is proposed, offering a cost-effective and efficient approach to generate fault samples. Leveraging this model, an innovative fault methodology that combines variational-mode-decomposition and graph convolutional network (VGCN) is proposed to enhance the extraction and representation of deep features from the fault data. This approach consists of two core components: a variational-mode-decomposition layer (V-Layer) that decomposes and reconstructs vibration signals, effectively mitigating noise interference, and a graph-convolution layer (GC-Layer) that uses Hermite polynomial to adaptively fit the graph convolution kernel, ensuring robustness in high-noise conditions. Faults are then classified through a fully connected layer with a softmax classifier. The proposed method is validated using both a simulated fault dataset and a real-world high-speed elevator dataset, demonstrating superior performance in fault diagnosis.