A Study On Processing Technology Of Machinery Vibration Signal Based On Compressed Sensing

The machinery fault diagnosis is achieved by analysis of vibration signal of mechanical equipment.Vibration signal acquisition under the sampling theorem make a lot of data and the big amount of data cause a serious burden of storage and transmission.Compressed sensing is a new framework of signal acquisition.In compressed sensing,with simple linear projection,the signal acquisition and compression is done simultaneously.Therefore the hardware consumption is decreased.But it is hard to reconstruction the signal.That is,the compressed sensing transfer cost from sampling end to calculation.The model of acquisition and reconstruction of mechanical vibration signal is established based on the principle of compressed sensing.The model is composed of three parts.The first part is vibration signal acquisition.The compressed signal that istransmitted and stored is measured with Gaussian random matrix.The second part is reconstruction of vibration signal.The mechanical vibration signal is not sparse in time domain and is approximate sparse in frequency domain.Therefore,the mechanical vibrationsignal belongs to compressible signal.The signal reconstruction comes down to optimization question that is solved by Orthogonal Matching Pursuit.The reconstruction matrix used in signal construction is a matrix product of Gaussian random matrix and Discrete Fourier matrix.The third part is evaluation target of construction.To judge the success of signal reconstruction,the evaluation target is the feature frequency in Hilbert demodulation spectrum.The effectiveness of the model is verified through simulation signal and gear vibration signal.In the last,the fault diagnosis model is proposed,that is Sparse Representation Diagnosis Based on Random Compression of Frequency Domain(SpaRCFD).In this model,Fast Fourier Transform is implemented to transfer vibration signal from time domain to frequency domain.Then the compression feature in frequency domain that is as feature vector is made by the product of amplitude sequence and Random matrix.The library of fault feature vector is established with the feature vector of various faults.To diagnose one feature vector,the combination coefficients that is the solution of representation question of that feature vector under feature library is used to determine which kind of fault the feature vector is.The effectiveness of the diagnosis model is verified by the experiment of gear and bearing.