Rolling bearing fault diagnosis based on optimized A-BiLSTM

Abstract

In order to improve the efficiency of hyperparameter setting and its adaptability to the model, and to reduce the high cost and low efficiency of manually setting model parameters, a rolling bearing fault diagnosis method based on Honey Badger Algorithm (HBA) optimized attention bidirectional long short-term memory network (HBA-A-BiLSTM) is proposed. Firstly, the optimal hyperparameter combination of A-BiLSTM model is searched by HBA, and then the fault diagnosis performance of A-BiLSTM model under the optimal hyperparameter is tested. Finally, the generalization ability of the model is tested based on data sets under different working conditions. The fault diagnosis effect of the proposed method is verified by using CWRU data set, and the diagnosis accuracy and confusion matrix are used for evaluation. The experimental results show that compared with other swarm intelligence optimization algorithms, the Honey Badger algorithm has good global search performance and fast convergence speed. The fault diagnosis accuracy of the optimized final model reaches 99.5%, which has good effects. It can also achieve stable and accurate fault diagnosis performance under different working conditions and has strong generalization ability.