An Asynchronous Quasi‐Cloud/Edge/Client Collaborative Federated Learning Mechanism for Fault Diagnosis

Although the federated learning method has ability to balance data and protect privacy by means of model aggregation, while existing methods are difficult achieve effectiveness centralized under sharing. The structure only a certain degree confidentiality for privacy, that is say, each client can reconstruct part information other clients based on parameters shared between server conditions. In order make mechanism more confidential, we breaks completely shared, establishes new asynchronous quasi-cloud/edge/client collaborative mechanism. We construct hierarchical multi-level confidential communication network, where network in way coordination without communication. cloud edges respectively use sequential Kalman filter algorithm perform an fusion uploaded their respective centers next round updates; proposed verified type rotating machinery