Research On Dispersion Entropy In Fault Diagnosis Of Reciprocating Compressor

Reciprocating compressors are the most widely used compressor types in industrial applications and are key equipment in gas pipelines,petrochemicals,oil refineries,ethylene chemicals,coal chemicals,and other industries.Monitoring and fault diagnosis of reciprocating compressors can help the machine continue to operate normally and is of great importance.Due to the complex structure and numerous components of the reciprocating compressor,collision and impact vibration will occur in the production and operation,which makes the vibration signal has obvious nonlinear and non-stationary characteristics.It is difficult to effectively extract fault information from complex vibration signals by traditional signal processing methods.The dispersion entropy is a new fault feature extraction method proposed in recent years,which has higher classification accuracy compared with traditional methods and has been widely studied in the fault diagnosis of rotating machinery,but it is still rare in the research of reciprocating machinery.To address this problem,this paper introduces the nonlinear dispersion entropy method to the fault diagnosis of reciprocating compressors and carries out a series of research works on it.Firstly,a brief overview of the current research status and trends in fault diagnosis of reciprocating compressors is given,highlighting the rationality of the choice of the scattered entropy method,while the structure,operating principle,and mechanism of reciprocating compressors are also analyzed and introduced.Secondly,based on the dispersion entropy of the nonlinear analysis method,because of the problem that the current dispersion entropy method has little research on the reciprocating compressor,a dispersion entropy influencing parameter suitable for the nonlinear and nonstationary vibration signals of the reciprocating compressor is proposed.The state cumulative distribution map and the parameter combination of m=2,c=8,and d=1 are more suitable for the application research of reciprocating machinery.Moreover,the properties of the dispersion entropy method are studied in detail through the measured vibration signals of reciprocating compressors,and the advantages of the dispersion entropy method in the application of reciprocating compressors are highlighted in terms of nonlinear dynamic change detection,robustness,stability,and timeliness.Then,the feature extraction method of the dispersion entropy is used to extract the feature of reciprocating compressor fault signals.The comparison with the sample entropy shows that the dispersion entropy has a better feature extraction effect,its entropy curve is more stable,and the cross between different fault entropy curves is less,which further verifies the superiority of the distributed entropy method in the feature extraction of the reciprocating compressor.Finally,according to the above feature extraction results,the intelligent pattern recognition method of the support vector machine is used to classify and identify the faults of the reciprocating compressor.In the process of classification and identification,a parameter selection method using LIBSVM 3.25 toolbox is proposed to address the problem of empirical selection of support vector machine parameters.The results show that the fault diagnosis method based on a support vector machine and dispersion entropy can accurately identify the transmission mechanism faults,but the fault identification rate for the valve is low.Therefore,the data processing method of lifting the wavelet is used to analyze the valve data.The results show that this method improves the recognition accuracy of valve fault and realizes the diagnosis of different fault types.