Answers are grounded in this paper and cite their sources.
Demonstrates improved fault diagnosis accuracy in wind turbine systems, highlighting data privacy advantages.
Fault diagnosis algorithms for devices based on deep learning have flourished in recent years. However, there are limitations in the actual engineering application of the model. Because data privacy concerns may arise when using multi-party data to build models, and existing models lack interpretability. To address these issues, taking wind turbine bearing as an example, this paper proposes a federated graph isomorphism network model for fault diagnosis based on causal discovery using federated learning (FL). Firstly, for the non-stationary bearing vibration signal, we use the fusion wavelet convolution (FWConv) method to extract the fault features of the signal. Secondly, the causation between features is discovered by using federated directed acyclic graph structure learning (FedDAG) algorithm. Finally, a federated graph isomorphic network (FedGIN) fault diagnosis model is constructed and compared with the local model of client, GCN and other models. Experimental results show that the proposed model can protect data privacy while achieving the same training effect as centralized graph neural network models.
No takes yet. Share an insight, caveat, or question.
Pan et al. (2023) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: