关键词: 轴承故障诊断, 注意力机制, 卷积神经网络, 双向长短时记忆网络

Abstract: Aiming at the problems that traditional fault diagnosis methods cannot adaptively select features and are difficult to cope with load changes and noise interference,an end-to-end fault diagnosis method based on attention mechanism was proposed.The spatial features of the original vibration signal were extracted through Convolutional Neural Network (CNN),and the temporal features were extracted based on the Bidirectional Long Short-Term Memory Network (BiLSTM).The attention mechanism was used to judge the importance of the hidden layer state at each time of BiLSTM and give the corresponding weight.The hidden layer state at all times was weighted and summed,and the Softmax layer was used as the classifier for fault diagnosis.The data collected by VALENIAN-PT500 and public data were used for experimental verification.The results showed that the proposed method had high diagnostic accuracy and strong generalization,and could maintain good fault diagnosis performance under variable load and noise interference.

Key words: bearing fault diagnosis, attention mechanism, convolutional neural network, bidirectional long short-term memory network