Research On Fault Diagnosis Method Of Reciprocating Compressor Based On MEMD And MSE

At present,the vibration signal of reciprocating compressor has non-stationary and nonlinear characteristics.When the traditional feature analysis method is used for feature extraction,only the single channel vibration signal after the multi-sensor measurement is processed,and the multi-channel vibration signal is later characterized.The level of data fusion does not consider the existence of correlation between sensor channels,which is not conducive to feature analysis of the vibration signals generated by the same physical system.Therefore,in order to analyze the characteristics of the multi-channel vibration signals in reciprocating compressors,this paper introduces the Multivariate Empirical Mode Decomposition(MEMD)method and combines with Multi-scale Sample Entropy(MSE)analysis.A multi-channel vibration signal of the reciprocating compressor is jointly analyzed to quantitatively describe the nonlinear characteristics of the reciprocating compressor fault,and is classified and identified by a support vector machine,which provides a new method for fault diagnosis of a reciprocating compressor.First,the research background and significance of this paper are expounded.The development history of fault diagnosis technology for reciprocating compressors and the research status of common fault diagnosis methods are described.This paper summarizes,researches and contrasts the fault diagnosis technology of reciprocating compressors,and proposes the content of the fault diagnosis technology of reciprocating compressor.Then,it outlines the structure and principle of the reciprocating compressor,common failure mechanisms,summarizes the vibration signal characteristics of the reciprocating compressor,designs fault test layouts,and briefly describes the fault experimental content of the 2D12 reciprocating compressor,providing fault diagnosis analysis for reciprocating compressors.Data foundation.Secondly,the traditional methods such as EMD and LMD cannot handle the fault feature extraction of multi-channel vibration signals of reciprocating compressors,and deeply study the basic principles and algorithms of MEMD.A MEMD-based fault diagnosis method was proposed.Through simulation signal and vibration signal of reciprocating compressor fault,the MEMD and EMD methods were compared and analyzed.The results show that the MEMD method has superiority and effectiveness in the accuracy and robustness of multi-channel signal decomposition,and is a multi-channel vibration for reciprocating compressors.Information troubleshooting provides new ideas and tools.Furthermore,the coarse-grained time series calculation for multi-scale entropy method produces inaccurate entropy estimation or undefined entropy,and the problem of endpoint "flying wing" phenomenon arises.A new improved coarse graining method is proposed to replace the common average roughness.Granulation method.Based on the analysis of multi-channel signals by the MEMD method,combined with the improved MSE method,a feature extraction method based on MEMD and MSE is formed.The method firstly uses MEMD to decompose multi-channel signals to form a series of multivariate IMF components,then uses the cross-correlation coefficients to screen the IMF components,and then uses the improved multi-scale entropy to quantify the IMF components to form a feature matrix.Finally,a singular value decomposition optimization matrix is ??used to construct the IMF component.Separable eigenvectors.Finally,taking multi-channel test signals of common faults of reciprocating compressors as the research object,using feature extraction methods based on MEMD and MSE to analyze,extract fault feature vectors and use support vector machine(SVM)as classifier to achieve accurate fault diagnosis.And compared with a variety of feature extraction methods to verify the superiority of the method.