The Research On Bearing Vibration Signal Detection And Reconstruction Method Based On Compressed Sensing Theory

In modern industrial production,rolling bearings play an extremely important role in rotating machinery.However,major mechanical equipment often faces complex conditions such as variable working loads,complicated working conditions,and high rotational speeds,which greatly increases the probability of failure of bearings in such environments.Since the bearing fault information will be reflected in the mechanical vibration signal,monitoring and analyzing the vibration signal generated during the operation of the equipment can maximize the potential of the equipment,locate and predict the fault in advance,and understand the fault.The mechanism and timely measures for mechanical failures to reduce or avoid as much as possible the unnecessary losses caused by the failure.The status information and fault information during the operation of the mechanical equipment can be expressed in the form of vibration signals.The traditional signal acquisition method mainly relies on the Shannon-Nyquist sampling theorem,but a fatal flaw of the theory is subject to the frequency of the signal being used.If the theorem is still used in the high-speed equipment,not only the sampling equipment is proposed.High requirements,and even under the conditions met by the sampling equipment,will inevitably produce a huge amount of data,which brings great pressure on the transmission,storage and processing of subsequent data.Since 2006,the theory of compressed sensing has subverted the limitations of the traditional Shannon-Nyquist sampling theory on sampling frequency requirements,not only achieving simultaneous sampling and compression,but also accurately reconstructing and recovering signals.Compressed sensing theory mainly consists of sparse representation,measurement matrix design and reconstruction method.Reconstruction method is one of the important links.For this reason,this paper takes bearing vibration signal as the object,reconstructs the algorithm and reconstructs it.The improvement of precision has been analyzed and studied,and the main research results obtained are as follows:(1)Research on the theory of compressed sensing and the theory of signal waveform detection.Compared with the traditional Shannon-Nyquist sampling theory,as the signal band becomes wider and wider,if the sampling of the signal continues to bring huge amounts of data and huge computational complexity,the theory of compressed sensing can break through this limitation.Reduce the amount of sampled data,creating favorable conditions for the processing and detection of vibrationsignals.Through the research of signal waveform detection theory,it is applied to the reconstruction process of signal reconstruction based on the reconstruction theory based on compressed sensing theory,so as to improve the signal reconstruction accuracy and reduce the reconstruction error.(2)Since the collected rolling bearing vibration signals generally contain a variety of interference information,and have the characteristics of randomness and non-stationaryness,it is not appropriate to directly extract features for subsequent analysis and processing.In view of the excellent performance of compressed sensing theory in signal processing,combined with local average method(LMD)and principal component analysis(PCA)in feature extraction,a local mean method and principal component analysis method are proposed.Fault feature extraction method for vibration signal compression domain.Firstly,LMD is used to decompose the compressed measurement value of the signal to obtain different PF components of the compressed measurement value of the signal.At the same time,the principal component analysis method is used to analyze the fault information of the signal compression measurement value.The theoretical analysis and simulation results show that under the same conditions,the above-mentioned method based on the compression measurement value of the signal can be accurately expressed.(3)For the traditional reconstruction algorithm,when reconstructing the vibration signal,the signal component with higher energy is always matched first,and the signal component with less energy for the mechanical device fault information may be ignored,thus making To reconstruct the important key information of signal loss,a method for detecting and reconstructing vibration signal compression measurements based on Neyman-Pearson criterion is proposed.In the first step,a linear Gaussian matrix is used to linearly compress the sample signal to obtain a low-dimensional compression measurement.In the second step,the compression measurement is detected based on the Neyman-Pearson criterion,and the best error in signal reconstruction error is obtained.Parameter;In the third step,the reconstruction algorithm uses a piecewise orthogonal matching pursuit algorithm to reconstruct and recover the signal based on the compressed measurement value.The experimental results show that compared with the traditional classical orthogonal matching pursuit algorithm(OMP)and the piecewise orthogonal matching pursuit algorithm,the proposed method can greatly reduce the signal reconstruction error and restore the original vibration signal more accurately.