Rolling bearing fault diagnosis method based on sequential memory enhanced network and terminal

The invention provides a rolling bearing fault diagnosis method based on a time sequence memory enhanced network and a terminal, relates to the technical field of rolling bearing fault detection, and aims to solve the problem that diagnosis errors are large when a CNN-LSTM network processes small sample noisy data. Secondly, recursive average filtering is carried out on input data of the LSTM, the processing capacity of time sequence noise-containing data is enhanced, a forgetting gate and an input gate of the LSTM unit are coupled, the forgetting gate and the input gate are connected with the memory unit, the memory capacity of time sequence data is improved, and the method is more suitable for learning of small sample data; and finally, the Softmax function after the full connection layer is utilized to realize multi-bearing fault state identification. Experiments based on a rolling bearing data set of the Kaisixi University show that the model is high in average accuracy of standard data and noise-added da