Operation state prediction model training method and equipment operation state detection method

The invention discloses an operation state prediction model training method and an equipment operation state detection method. The prediction model is trained by using the vibration time-frequency data of the plurality of devices corresponding to the target device category at the client to obtain the first model parameter, and the first model parameter and the target category identifier corresponding to the target device category are sent to the server, so that the pertinence of the prediction model can be improved, and the prediction efficiency is improved. The influence of the training data distribution difference on the model training result can be reduced, and the detection accuracy of the operation state prediction model on the equipment operation state can be improved. Further, the server performs classification aggregation processing on the first model parameters sent by the plurality of clients according to the target category identifier to obtain a second model parameter corresponding to the target c