2025 International Conference on Equipment Intelligent Operation and Maintenance (ICEIOM2025)

302 / 2025-06-19 12:01:21

Hidden Markov Model-Based Weighted Graph Representation for Bearing Fault Diagnosis with Graph Attention Networks

Fault diagnosis,Bearing,Path graph,Hidden Markov Model,Graph Attention Network

Special Sessions > SSL 09 AI Methods and Applications in Condition Monitoring, Diagnosis, and Maintenance of Structure, Power, Equipment, etc

Rolling bearings serve as critical components in rotating machinery, where their operational condition significantly impacts overall equipment safety and reliability, making accurate fault diagnosis crucial. While deep learning approaches have advanced bearing fault diagnosis, current methods still face limitations in feature extraction comprehensiveness and interpretability. To overcome these challenges, this study proposes a graph attention network incorporating Hidden Markov Model (HMM)-weighted path graph for end-to-end bearing fault diagnosis. Specifically, bearing vibration signals are segmented and transformed into path graphs, where edge weights are assigned using HMM modelling. These weighted graphs are then processed by the graph attention network for diagnostic classification. Experimental results demonstrate that the proposed HMM-weighted path graph construction method outperforms alternative approaches, significantly enhancing bearing fault diagnosis accuracy.