Fault Diagnosis Of Reciprocating Compressorindicator Diagram Based On Image Recognition

Nowadays,one of the commonly used equipment for the production sector is reciprocating compressors,equipment failure has a serious securiyt risk,in addition to affect production efficiency and reduce economic efficiency.Therefore,when a compressor failure occurs,a timely and effective fault problem solving is particularly important.Due to the structure of reciprocating compressor is relatively complex,the vibration generating the excitation source,the measured vibration signal is often mixed with other redundant information of the random signal and the impulse signal,through extracting vibration signal characteristics to identify the type of fault diagnosis is not very satisfactory.To solve this problem,this paper proposes a method of using pressure signal indicator diagram to study the fault diagnosis of reciprocating compressor.Generated by the in-cylinder pressure signal,indicator diagram truly records the operating conditions and reflects the pressure changes in the cylinder,when fault occurs,indicator diagram shapes varying.Therefore,this paper will research the fault diagnosis type of reciprocating compressor according to indicator diagram shape differences.First of all,on the basis of referring to a number of domestic and foreign literature,this paper comprehensive analysises the basic principles and application method of fault diagnosis technology,also presents the research status in fault diagnosis technology of reciprocating compressor,proposes the method of fault diagnosis based on image recognition of indicator diagram.Secondly,it presents the structure and rationale,as well as indicator diagram generating and normalizing of reciprocating compressor,comparas the characteristics difference between fault indicator diagram and normal condition.Thirdly,it proposes two fault feature extraction algorithm,they are invariant moments based on the whole area of indicator diagram and wavelet desvriptor based on the edge contour of indicator diagram,it introduces kernel principal component analysis to optimize the useful information of invariant moments,in order to shorten the required time of final fault diagnosis.Finally,it detailed presents the basic priciples and the choice of kernel function of support vector machine theory,the created multi-class classification can smartly identiy the fault type of reciprocating compressor indicator diagram,and a satisfied classification result is concluded.