Research On Fault Diagnosis Method Of Reciprocating Compressor Based On Adaptive Maximum Correlation Kurtosis Deconvolution

Reciprocating compressor is widely used in petrochemical enterprises.With numerous components and complex structure,reciprocating compressor usually operates in high temperature,high pressure,flammable and explosive working environment.A minor fault may lead to serious safety accident problems.The fault detection and diagnosis technology of reciprocating compressor has always been the focus of research by researchers.The minimization of downtime and cost is an important index in petrochemical production.Therefore,the research on fault diagnosis technology of reciprocating compressor has important economic significance to reduce the accident rate and increase the production capacity.Reciprocating compressor is a kind of high-noise equipment.In the production operation,the vibration,impact and background noise of each component are coupled with each other.It is difficult to extract useful fault information from complex signals based on traditional signal processing methods.In this paper,a fault diagnosis method of reciprocating compressor based on adaptive maximum correlation kurtosis deconvolution(MCKD)is proposed to realize the accurate fault diagnosis of reciprocating compressor.Firstly,the research background and significance of this paper are described,and the advantages of blind deconvolution fault diagnosis algorithms emerging in recent years are discussed.Several common intelligent pattern recognition methods are compared and analyzed,and the research conception of fault feature extraction and pattern recognition in this paper is put forward.Secondly,the MCKD deconvolution method is deeply studied.Aiming at the problem that the filter effect in the MCKD deconvolution method is affected by the filter length,fault period and other parameters,an improved MCKD method is proposed.The core of this method is to use the moth flame trap optimization algorithm,and the approximate entropy as the fitness function to optimize the MCKD filter length L,the failure period T,to get the optimal parameter combination,and realize the MCKD signal denoising adaptive.Then,the ensemble empirical mode decomposition(EEMD)method was used to decompose the denoising signals,and the intrinsic mode components(IMF)with rich fault information were selected according to the similarity criterion to reconstruct the signals.The results of simulation signals and measured signals show that this method can effectively extract the fault characteristic information of reciprocating compressor.The fine compound multi-scale fuzzy entropy of the reconstructed signal was calculated and the eigenmatrix was constructed.Finally,according to the characteristics of vibration signals of reciprocating compressor,such as multi-source impact coupling and strong background noise interference,a fault diagnosis method of reciprocating compressor based on adaptive maximum correlation kurtosis deconvolution was proposed by integrating MCKD,EEMD,refine compound multiscale fuzzy entropy and extreme learning machine.The process includes acquisition of vibration acceleration signals of reciprocating compressor,deconvolution processing of signals by adaptive maximum correlation kurtosis deconvolution,extraction of fault feature by EEMD and refine composite multiscale fuzzy entropy,and fault classification by extreme learning machine.The results show that the method can realize the diagnosis of different types of faults,and compared with a variety of feature extraction methods,which verifies the superiority of fault feature extraction based on the adaptive maximum correlation kurtosis deconvolution method.