Bearing Vibration Signal Analysis Based On Compressed Sensing Technology

Bearing is one of the important basic components of modern industry,and the monitoring of bearing working conditions is of great significance to en sure the safe and stable operation of equipment.In the process of online vibration monitoring of the bearing condition,a large amount of data is required to be collected,transmitted,stored,and processed,and the software and hardware of the analysis s ystem are expensive.The compressive sensing theory emerging in recent years provides new ideas and methods for low-cost signal acquisition and transmission.Aiming at the application of monitoring the running state of the bearing,the paper studies the analysis and application of the compressive sensing theory in the bearing vibration signal.In the aspect of signal sparse decomposition,the method of determining the initial dictionary in the process of bearing vibration signal sparse decomposition is given.The method is based on the kurtosis value analysis which is sensitive to fault impact,selects signal segments with significant fault information to form an initial dictionary,and determines the dimension of the initial dictionary based on the frequency analysis of fault features to reduce calculation time and enhance the ability to express signal fault features.On the standard data set,the simulation analyzes the influence of different initial dictionaries and different initial dictionary dimensions on the signal sparse decomposition.In terms of signal compression,a measurement matrix is constructed based on Hadamard matrix analysis,which is used for bearing vibration signal compression and dimension reduction.One row vector of the Hadamard matrix is selected,and the other row vectors are shifted to the right of the previous row vector to obtain the measurement matrix.Compression and classification experiments are carried out on the standard data set,and the bearing vibration signal is compressed with the proposed measurement matrix,and then its normalized wavelet is extracted.Information entropy features,and finally complete the fault diagnosis with four classification algorithms.The experimental results show that the classification and recognition accuracy of the compressed vibration signal is high,which can achieve the purpose of identifying the running state of the bearing.In terms of feature extraction of bearing vibration signal,wavelet packet decomposition technology is applied,the optimal decomposition level of compressed signal is determined according to the characteristic frequency of bearing fault,the normalized wavelet information entropy of each sub-band is calculated as bearing feature parameter,and four classification algorithms are applied.It is verified that the extraction of bearing features as feature parameters by this method can achieve better signal classification results.By comparing the effects of different decomposition levels on vibration signal classification,the optimal decomposition level of wavelet packet calculated from fault characteristic frequency is still applicable to compressed signals.