Fault Feature Extraction For Bearing Clearance Of Reciprocating Compressor Based On CIELMD And RCMFE

Reciprocating compressors are used widely in industry,agriculture,transportation,national defense,especially petroleum and chemical production industries for its wide pressure rang,high compression efficiency and strong applicability.Sliding bearings are important components of the reciprocating compressor transmission mechanism,and their operating conditions have characteristics of high speed and heavy load.After long-term service,owing to manufacturing and assembly errors,collisions and friction during operation,the sliding bearing is bound to occur oversized clearance faults,which in turn causes the whole machine to produce severe vibration and result in huge economic losses and adverse social impacts.Therefore,in order to provide theoretical support for efficient,reliable,safe and stable operation of reciprocating compressors,this paper takes the clearance state of the reciprocating compressor sliding bearing as the object,and carries out the research on its vibration mechanism,fault feature extraction,and pattern recognition method according to the strong nonstationary,nonlinear,and characteristic coupling characteristics of its vibration signals.The main research contents are as follows.A multi-body dynamics model of reciprocating compressor with bearing clearance is established to study the mechanism of vibration transmission in bearing clearance.Based on the description of the bearing clearance mechanism of the reciprocating compressor,a rigidflexible multi-body dynamic model of a reciprocating compressor is established.After the model is verified by using the measured vibration data,the internal relationship between the collision force of the moving pair and the vibration response of the body is studied..Then the effect of bearing clearance variation on the dynamic performance of a reciprocating compressor was investigated,and the vibration transmission mechanism of the bearing clearance excitation is gained by the simulation of this model.A compound interpolation local mean decomposition method(CIELMD)is proposed based on the local strong non-stationary characteristics of the reciprocating compressor vibration signal.The interpolation envelope is the key factor affecting the accuracy of local mean decomposition method(LMD).For the local strong nonstationary characteristics of the signal,it is proposed to establish the local envelope by using cubic spline interpolation for stationary part of a signal,and establish a local envelope by using monotonic cubic Hermite interpolation for strong nonstationary part.By defining the signal nonstationary coefficients and the method of connecting the endpoints of different interpolation curves,a compound envelope construction algorithm is given.Furthermore,a local mean decomposition method based on the compound interpolation envelope is constructed.The study of strong nonstationary simulation signals shows that this method can significantly improve the decomposition accuracy of its PF component.The sevrer oversized bearing clearance fault signal of reciprocating compressor bearing is analyzed,and the results show that the characteristic frequency of PF component envelope spectrum is more significant.A nonlinear quantitative analysis method for bearing clearance fault state of reciprocating compressor is proposed based on refined composite multi-scale fuzzy entropy(RCFME).A refined composite multi-scale fuzzy entropy with better stability,precision and anti-interference is proposed by combining the concept of refine composite multi-scale entropy and fuzzy entropy,and is applied to quantify the nonlinear characteristics of different state signal to form fault feature.Application analysis of white noise and 1/f noise signals shows that RCFME entropy values are consistent,insensitive to data length,and the probability of undefined entropy is small.The research on the oversized bearing clearance fault signal of reciprocating compressor shows that the characteristic curves of different states are distinguishable.A fault feature optimization method for the reciprocating compressor bearing clearance state is presented based on Memetic algorithm.A series of PF components was decomposed from fault signals of different bearing clearance states by CIELMD method,and then some of which contain the main fault information is screened by the correlation coefficient,the RCFME method is used to quantitatively describe the PF component to form the state feature matrix,and the Memetic algorithm is used to optimize the element composition of the matrix with the largest average sample distance.In traditional binary tree support vector machine hierarchical structure,each subclassifier uses uniform parameter to train samples,however this strategy inhibite the performance of subclassifiers,then an improved algorithm is proposed for each subclassifier using independent parameter to train samples.The reciprocating compressor fault recognition results show that both the feature extraction method and the pattern recognition method improve the accuracy of fault state recognition.