The Research On Fault Diagnosis Method For Rolling Bearing Based On Compressed Sensing

Mechanical fault diagnosis plays an important role in improving the working efficiency of equipment,ensuring the subsequent maintenance work and reducing maintenance cost.At present,most of mechanical fault diagnosis work is carried out by analysing the vibration signals sampled from equipment.This dissertation focus on rolling bearing and combines the theory of compressed sensing with the mechanical fault diagnosis,at the same time,the fault diagnosis method of rolling bearing is studied in detail.The achievements are summarized as follows:?1?Aiming at that sample sets used for training over-complete dictionary will face problems with insufficient signal type,which causes difficulties of the precision of subsequent analysis,classification and recognition.Therefore,a method of rolling bearing vibration signal compression reconstruction method based on complete training samples of over-complete dictionary is proposed in this research.Firstly,the sample sets used for dictionary learning is constructed,which makes it contain various signal components.Then,the sparse representation of signal is obtained in K-SVD over-complete dictionary based on the theory of compressed sensing.Finally,the signal is reconstructed based on a small amount of compressive measurement data.The test results of simulated data demonstrated that the proposed method has higher reconstruction accuracy and shorter reconstruction time without losing the main information of the original vibration signal.?2?The vibration signal of rolling bearing sampled in real-time is random,non-stationary and contains multiple interference signals,which cannot be directly used as fault feature.Therefore,a fault diagnosis method for rolling bearing based on EMD decomposition and neural network is proposed.Firstly,a series of intrinsic mode function?IMF?components with different characteristics is obtained via empirical mode decomposition for the reconstructed signal.Then select the first six IMF component containing the main fault features and extract the energy characteristics.Eventually,the energy characteristic parameters of each IMF component are regarded as input parameters of neural network for fault identification and classification.The simulation results show that the proposed method has the advantages of fast speed and high accuracy in identifying the fault type of rolling bearing.?3?Traditional rolling bearing fault diagnosis methods based on the Nyquist sampling theorem will generate vast amounts of data in the process of signal processing,which causes great burden for subsequent storage and transmission.The recently introduced theory of Compressed Sensing alleviates this problem,which needs to recovery the original signal from a small amount of compressed measurements data based on l₁ norm.However,it is unnecessary to recovery full signal from the compressive measurements and then extract fault feature in the process of fault diagnosis,only needs to extract the feature associated with the fault.Therefore,a method of rolling bearing vibration signal feature extraction based on compressed domain is proposed in this research.Firstly,the signal was acquired by solving the minimum l₂ norm according to the compressive measurements.Then,its spectrum was obtained via Fast Fourier Transform.Finally,compared with the amplitude of original signal,the amplitude of the signal based on the proposed method decreased.So it is necessary to correct the amplitude of the signal,and fault feature was directly extracted in the compressed domain.The test results of simulation and experiment data demonstrated that the proposed method not only reduces the data of subsequent processing and the computational complexity,but also achieves feature extraction with less amount of data in the case of without losing accuracy.