Research On Fault Diagnosis Of Rolling Bearings Based On Bidirectional Long And Short Time Memory Network

As the basic key components of modern mechanical equipment,rolling bearings are of great significance for fault diagnosis for mechanical equipment.The traditional fault diagnosis method is cumbersome to extract features,which cannot meet the requirements of real-time and high efficiency of modern equipment fault monitoring.Based on this,this paper takes rolling bearings as the research object and uses the advantages of bidirectional long-short time memory network(BiLSTM)in processing one-dimensional timing signals to conduct bearing fault diagnosis based on bidirectional long-short-time memory network.The main tasks are as follows:(1)Aiming at the difficulty of feature extraction in traditional fault diagnosis,this paper establishes a bearing fault diagnosis model based on BiLSTM,which realizes endto-end bearing fault diagnosis and makes up for the shortcomings of traditional methods relying on manual extraction of features.The validity of the model was verified on the bearing dataset of Case Western Reserve University in the United States,and compared with the traditional fault diagnosis method of pre-extraction of wavelet packet energy characteristics and then using support vector machine classification,the accuracy of the model was higher.(2)Aiming at the problem that the bearing fault diagnosis model based on BiLSTM is slowly trained,this paper constructs a bearing fault diagnosis model based on CNNBiLSTM.Based on the rapid feature extraction of convolutional layers,the BiLSTM network layer is used to further analyze related features.Experiments have proved that on the same data set,the CNN-BiLSTM model not only greatly improves the training speed of the BiLSTM model,but also further improves the accuracy of the model.(3)Aiming at the problem of noise containing noise in the real working scenario of the bearing,in order to improve the noise immunity of the bearing fault diagnosis model,this paper introduces an attention mechanism on the basis of the CNN-BiLSTM model,and selects the ELU activation function with better noise adaptability in the convolutional network layer,and proposes a CNN-BiLSTM bearing fault diagnosis model with an attention mechanism.The bearing dataset of Case Western Reserve University in the United States and the bearing dataset of Paderborn University in Germany verified the validity of the model,and finally the model was visualized by t-SNE visualization technology,which enhanced the interpretability of the model.