Research On Rolling Bearing Fault Diagnosis Method Based On One-dimensional CNN＿LSTM

Bearing is one of the most used parts in modern equipment.Due to the complex operating environment of the equipment,the equipment fails.In view of the influence of noise on vibration signal,it will reduce the accuracy of fault diagnosis;Due to the non-stationary and complex characteristics of bearing vibration signal,the fault diagnosis method based on manual experience needs to formulate rules,and the diagnosis rate is low.Aiming at the above two problems,this paper studies the signal noise reduction method and builds a neural network model which can adaptively extract features.Firstly,the research of signal noise reduction method: because empirical mode decomposition(EMD)will cause the phenomenon of signal mode aliasing when decomposing the signal,and it is impossible to determine which of the decomposed natural mode components need noise reduction,this paper uses the improved decomposition method CEEMDAN to decompose the signal,eliminates the phenomenon of mode aliasing,and introduces the sample entropy algorithm,By calculating the sample entropy and average sample entropy of each decomposed component,the components higher than the average sample entropy are denoised by wavelet threshold,and the components lower than the sample entropy remain unchanged,so as to complete the signal reconstruction.Secondly,the construction of the network model: the convolutional neural network(CNN)model has a series of defects,such as slow convergence and large training samples,and then improves the model.This paper designs a method based on CNN＿LSTM network model structure and noise reduction of data samples before data input.The experimental results show that even when there are few training samples,the training accuracy of the model increases steadily,and the model tends to converge when iterating 40 times,which improves the above problems.The accuracy of the final test set is as high as 99.25%.Finally,through the experimental data collection and using the noise reduction method in this paper,the CNN＿LSTM network model built after signal processing is used for fault diagnosis.The result of fault diagnosis rate reaches 99%,which has a certain value in the field of bearing fault diagnosis.