Fault Diagnosis Method Of Reciprocating Compressor Based On MCA And CMFE

Reciprocating compressors are commonly used in the petroleum and chemical industries,and are the main tools for the compression and transportation of coal gas,natural gas,ethylene and other dangerous gases,in the event of a major failure,the leakage of dangerous gases will not only lead to huge economic losses,but may also result in casualties.Therefore,in order to avoid the occurrence of the above things,it is necessary to conduct research on fault diagnosis and monitoring of reciprocating compressors.But the reciprocating compressor is not only complicated in structure,but also has a harsh working environment,and the vibration signal has non-stationary and non-linear characteristics.Therefore,how to separate critical fault information from such a complex signal has become a research on fault diagnosis and monitoring of reciprocating compressors and is also a problem that scientific researchers have to solve in recent years.This paper proposes a fault diagnosis method for reciprocating compressors based on the combination of morphological component analysis(MCA)and compound multi-scale fuzzy entropy(CMFE)for the characteristics of reciprocating compressor vibration signals.In the traditional MCA algorithm,the sparse dictionary is fixed and does not have adaptability.Once the sparse dictionary does not match the signal components,it will greatly affect the decomposition result of the MCA,resulting in the inability to separate critical fault information,therefore,this paper proposes a MCA decomposition method based on kurtosis-optimized sparse dictionary.The wavelet sparse dictionary optimized for kurtosis can be used to match the shock components well.The optimized wavelet sparse dictionary and discrete sine and cosine dictionary are used to reconstruct the fault shock component and the resonance component,which can effectively separate the fault shock signal from the noise and other interference signals.Multi-scale fuzzy entropy(MFE)is proposed by combining the concept of multi-scale analysis and fuzzy entropy theory.Compared with fuzzy entropy,MFE can reflect the complexity and self-similarity of time series under different scale factors.But in the ordinary MFE algorithm,the MFE under different scale factors is defined as the fuzzy entropy of the first coarse-grained time series under the scale factor.In this way,for short time series analysis,MFE will often cause undefined fuzzy entropy values at larger time scales.To make up for this deficiency,this paper proposes composite multi-scale fuzzy entropy(CMFE)to improve algorithm accuracy.The method in this paper uses MCA to decompose the vibration signal of the reciprocating compressor bearing to obtain the shock component containing the fault feature,then,the impact component is quantitatively calculated by CMFE,the fault feature vector is extracted,and finally the fault feature vector is input into a support vector machine(SVM)for recognition and classification,and the fault type is judged.Experimental results show that this method can accurately extract fault feature information and diagnose the fault type.First,consult relevant literature to understand the structure and failure mechanism of reciprocating compressors,and grasp the status and development trends of fault diagnosis of reciprocating compressors at home and abroad.Secondly,the principle and implementation steps of MCA are studied.In view of the shortcomings of traditional MCA fixed sparse dictionary that does not have adaptability,the MCA algorithm that uses kurtosis to optimize wavelet sparse dictionary is proposed,which can effectively separate fault impact and interference components.The experimental results of analog signals and measured vibration signals of reciprocating compressors show that the improved MCA decomposition method can accurately and efficiently analyze non-stationary signals.Then,the principle and algorithm flow of sample entropy,fuzzy entropy and multi-scale fuzzy entropy are explained,aiming at the problem that multi-scale fuzzy entropy often causes undefined entropy values for short time series analysis at larger time scales,a composite multi-scale fuzzy entropy is proposed,to improve the algorithm accuracy of multi-scale fuzzy entropy.The experimental data of reciprocating compressor failure shows the effectiveness and superiority of compound multi-scale fuzzy entropy in feature extraction.Finally,taking the 2D12 reciprocating compressor as the research object,this paper introduces the structure and working principle of the reciprocating compressor,the fault mechanism of bearing and gas valve and the arrangement of measuring points,and gives the fault diagnosis method and specific process.This method first analyzes and decomposes the sparse dictionary based on morphological component analysis of the fault signal,then,the composite multi-scale fuzzy entropy quantitative calculation is performed on the fault impact component obtained by the decomposition to construct the feature vector,and finally,the support vector machine is used for classification and recognition.Experimental results show that the method in this paper has high recognition accuracy and can effectively diagnose different faults.