Cloud-edge collaborative hydraulic pump condition monitoring has received significant attention due to its superior real-time data processing capabilities. However, limitations in network bandwidth often lead to latency during data transmission from the edge to the cloud. Moreover, while this collaboration enhances system flexibility, it still struggles to fully adapt to the complex and dynamic operating conditions of hydraulic pumps. To address these issues, this study proposes a cloud-based intelligent reconstruction method for low-sampling-rate signals, aiming to reduce transmission latency. Additionally, a cloud-edge collaborative strategy is designed to improve system adaptability. Specifically, a Variational Autoencoder (VAE)-based novelty detection model is deployed at the edge to detect real-time changes in data distribution. The detection results subsequently trigger updates for the cloud-based models, ensuring the system remains responsive to variations in operating conditions. Experimental results demonstrate the effectiveness of the proposed novelty detection model on a piston pump loose slipper fault dataset, achieving a classification accuracy of 97%, an AUC of 0.9856, and an F1-score of 0.9109.
Xu et al. (Wed,) studied this question.