Research On Multi-channel Compressed Sensing Technology Of Rotating Machinery Mixed Vibration Signal

To solve the problem of data storage and transmission of rotating mechanical vibration signals in the context of big data,traditional compressed sensing technology is limited by the sparseness of vibration signals.A single-channel compressed sensing method based on fast block sparse Bayesian theory and a multi-channel compressed sensing method based on MMV model are applied to the processing of multi-source vibration signals in rotating machinery.The effectiveness of the proposed methods is verified by simulating the vibration signal and the laboratory test signals of rotating mechanical gears,bearings and rotor systems.The specific research work is as follows:(1)Summarize the research status of compressed sensing theory and algorithm;analyze the research results and existing problems of other scholars;study several key issues involved in compressed sensing and its development direction.(2)Establish a single channel vibration signal compression sensing method.Aiming at the characteristics of vibration signals,the effects of block diagonal LDPC observation matrix and fast block sparse Bayesian learning algorithm on vibration signal observation and reconstruction are studied.A single channel compressed sensing method based on diagonal block LDPC matrix+BSBL FM method is established.Compared with the traditional single channel compressed sensing method,the signal-to-noise ratio,correlation coefficient and time are taken as evaluation indexes.It lays the foundation for multi-channel compressed sensing method.(3)For the multi-sensor system of rotating machinery in practical application,the correlation model of multi-channel sampling signals of rotating machinery is studied.The multi-channel signals are uniformly observed by diagonal block LDPC matrix,and multi-channel signals are realized by STSBL-FM method.The reconstruction and comparison of the established multi-channel compressed sensing method based on STSBL_FM algorithm with the existing multi-channel compressed sensing method and single-channel compressed sensing method for the quality and efficiency of vibration signal reconstruction.The effectiveness of the algorithm is verified by the actual sampling of the mechanical signal.(4)For the characteristics that the sampling signal is mostly mixed signal,the blind source separation method is combined with the multi-channel compression sensing technology to establish a multi-channel source signal separation method based on CICA+BSBL FM algorithm to separate and reconstruct the joint observation signal.Using the mean correlation coefficient and time as the evaluation index,the separation quality and efficiency of the new method were verified by the actual mechanical sampling signal.(5)Based on the Matlab GUI tool,the several compression sensing methods proposed in this paper are systematically integrated to establish a rotating mechanical vibration signal compression sensing system that can be applied with the Windows platform,which reduces the difficulty of the method and is convenient for engineering applications.Through the above research,it is proved that the proposed method improves the compression ratio and reconstruction accuracy of the signal under the premise of ensuring the reconstructed signal quality and compression efficiency.This study also provides a reference for the storage and transmission processing of massive data brought by real-time monitoring and diagnosis of rotating machinery.