Research On Intelligent Diagnosis Method Of Rolling Bearing Fault Based On Deep Learning

Rolling bearings are widely used in a variety of rotating machinery,called mechanical joints,which are vital components in rotating machinery and one of the most prone to failure.Whether the operating state is normal or not directly affects the safety of personnel and equipment,so finding an effective method for fault diagnosis of rolling bearing is of great significance.This paper mainly proposes two new intelligent diagnosis methods for the complex and indistinguishable structure of rolling bearing vibration signals.The first is the intelligent diagnosis methods of rolling bearing fault based on the combination of Variational Mode Decomposition(VMD)and Deep Belief Network(DBN).The second is the intelligent diagnosis of rolling bearing fault based on improved S transform combined with Sparse Auto-Encoder(SAE).The effectiveness of the new method in fault diagnosis of rolling bearings is verified by experiments.First,the basic structure of rolling bearings and common types of failures were studied.The analysis method of rolling bearing vibration signal is studied in detail.The methods of feature extraction of rolling bearing vibration signal are studied by time domain analysis,frequency domain analysis and time-frequency domain analysis.Secondly,two deep learning neural network models are studied,which are Sparse Auto-Encoder and Deep Belief Network.The structure and principle of the Auto-Encoder are studied.The cost function of the Auto-Encoder is changed by adding the sparse penalty term to make the encoder get sparse.The principles of Dropout and anti-noise coding are also studied.At the same time,the network structure and training process of the deep belief network are studied in detail.Thirdly,for the problem that the vibration signal of rolling bearing is complex and difficult to extract fault features,a fault diagnosis method based on VMD and Deep Belief Network is proposed.Firstly,the vibration signal is decomposed into several modal components by using VMD,and then the components are reconstructed by sample entropy and cross-correlation.After the signal is reconstructed,the useless information such as noise is effectively filtered out,and the useful features of the signal are more obvious.The spectrum of the reconstructed signal is input into the Deep Belief Network,which can quickly and effectively classify various rolling bearing faults.The effectiveness of the proposed method is verified by actual signal experiments.Then,for the problem that the bearing vibration signal is nonlinear and non-stationary,it is proposed to solve the fault diagnosis method of rolling bearing based on the combination of improved S-transformation and Sparse Auto-Encoder.Compared with the fast Fourier transform,the S-transform has the advantage that the window function window width can be changed.By adding the window width adjustment factor to the S-transform,the window function' window width can be flexibly adjusted to obtain better time-frequency resolution to have a good effect on handling bearing vibration signals.Combined with the Sparse Auto-Encoder,it can effectively diagnose faults of various faults in different parts of the rolling bearing and different fault scales.And the effectiveness of the method is proved by actual signal experiments.Finally,a rolling bearing condition monitoring and fault diagnosis system based on LabVIEW and MATLAB is designed.Combining LabVIEW's graphical programming with MATLAB's powerful computing capabilities,it can realize the functions of rolling bearing vibration signal acquisition,condition monitoring and fault diagnosis.