Fault Feature Extraction And Diagnosis Of Reciprocating Compressor Based On Optimized VMD And IMFSE

Reciprocating compressors are widely used in essential fields such as the petrochemical industry due to their advantages of low power consumption per unit,the vast application range of pressure and wide exhaust range.As an essential connecting part in the reciprocating compressor,the sliding bearing often has too large a bearing clearance fault due to collision friction and assembly error,which makes the reciprocating compressor stop due to excessive vibration and noise,affecting the routine work,and even causing safety accidents.Therefore,the sliding bearing of the reciprocating compressor was selected as the research object,and the feature extraction method and pattern recognition method were studied according to the non-stationary,non-linear and feature coupling characteristics of the bearing clearance fault vibration signal,to reduce the accident rate and ensure the safe and efficient operation of enterprise production.Aiming at the local strong non-stationary characteristics of vibration signals of reciprocating compressors,the difficulty in selecting mode numberand penalty factorof the variational mode decomposition algorithm was analyzed,and the sparrow search algorithm optimized variational mode decomposition method（SSA-VMD）for local strong non-stationary characteristics of vibration signals of reciprocating compressors was proposed.The optimal parameter combination of VMD algorithm[K₀,α₀]was obtained by using SSA to optimize the parameters of the mode number and penalty factor of VMD algorithm.The VMD signal decomposition method optimized by SSA was applied to the vibration signals of the bearing clearance fault of the reciprocating compressor for signal adaptive decomposition processing.The results show that this method can improve the efficiency of signal decomposition,reduce the error of decomposition and reconstruction,and has the advantage of better signal decomposition.Based on the low efficiency of sample entropy calculation in multi-scale sample entropy algorithm（MSE）and the problem of“neutralizing”the dynamic mutation behavior of the original signal caused by the mean coarse-grained method,an improved multi-scale fast sample entropy（IMFSE）feature extraction method was proposed.By constructing fast sample entropy based on a symbolic variable matrix,the computational efficiency of the algorithm is improved,and the standard deviation coarse-grained is used instead of mean coarse-grained for multi-scale processing,which makes the entropy analysis method more accurate.The simulation signals and the measured signals show that compared with the MSE algorithm,the IMFSE algorithm has higher entropy stability and can better distinguish the characteristic curves of different states.The computational efficiency of IMFSE algorithm is much higher than MSE algorithm.To further improve the recognition accuracy of the kernel extreme learning machine model,a competitive learning particle swarm optimization multi-kernel extreme learning machine model（CLPSO-MKELM）was proposed.By combining the improved gaussian kernel function,polynomial kernel function and sigmoid kernel function with linear weighting,the multi-kernel extreme learning machine model was obtained,and the kernel parameters of the model were optimized to form the CLPSO-MKELM intelligent pattern recognition method.The example verification analysis results show that the model has better classification accuracy and generalization performance,and can meet the needs of classification and recognition of different types of samples.Finally,the established fault feature extraction method based on SSA-VMD and IMFSE and CLPSO-MKELM intelligent pattern recognition methods are combined,and applied to reciprocating compressor sliding bearing vibration signals for fault diagnosis research.The results show that the proposed fault diagnosis method can accurately diagnose different sliding bearing faults in reciprocating compressors.