An Optimized Nasnet-Mobile Network Based on the BOA Algorithm for Bearing Health Monitoring

In the situation that the poor adaptive ability of traditional feature extraction methods and the weak generalization ability of a shallow neural network, a Nasnet-Mobile network framework method rely on the butterfly optimization algorithm (BOA) is proposed to optimize the internal parameters. Firstly, to process the bearing data, the bearing data is shifted by two-dimensional time-frequency transformation to obtain some normalized time-frequency spectrum. And the uppermost accuracy of the test set in the neural network is regraded as a fitness function, where BOA is used to search for the best parameter combination of Nasnet-Mobile. Then, the learning rate and the amount of batch learning that play the vital roles in the BOA will be selected appropriately, and at the same time, the Nasnet-Mobile will be given by comparison Finally, combined with the powerful adaptive feature extraction deep learning, the obtained samples were input into Nasnet Mobile to train, and a bearing fault diagnosis model using Nasnet-Mobile is established. After several tests on the test samples, the diagnostic rate of the model can be more than 99%, which is much better than the traditional fault diagnosis methods. Experimental results show that the proposed method can extract effective features better and has higher classification and detection accuracy.