Research On Fault Diagnosis Method Of Crank Pin Of Reciprocating Compressor

The structure of reciprocating compressor is complex,and the normal and safe operation of crankshaft connecting rod mechanism is related to the production efficiency of enterprises.The fault rate of crank-connecting rod mechanism of reciprocating compressor is relatively high.It is difficult to obtain information of fault sudden because of reciprocating and rotating motion,and its vibration signal shows non-stationary characteristics.Traditional diagnosis methods are difficult to identify.In this paper,Bentley reciprocating machine model is used to carry out fault simulation experiments,select fault features and establish diagnosis model to identify crank pin status.The main research contents are as follows:(1)Fault simulation experiment of crank pin of reciprocating compressor.The basic structure and working principle of reciprocating compressor,the composition of crankshaft and connecting rod mechanism,stress condition,principle and fault mechanism are introduced.Using Bentley reciprocating machine model,the simulation experiment scheme,experiment process and steps of crank pin fault are designed,and the vibration acceleration signals of each crank pin obtained from the experiment are analyzed.(2)Establishing feature selection model.Firstly,based on time-domain analysis and frequency-domain analysis,fault diagnosis of clearance between connecting rod bushing and crank pin of reciprocating compressor is carried out.Then,feature extraction of time-domain and frequency-domain signals is carried out.Combining with feature selection method of compensation distance evaluation technology,time-domain waveform and vibration signal data of normal,slight and severe wear of crank pin are selected,and feature selection model is input.The more sensitive characteristic indicators were obtained.(3)The SVM parameter optimization model based on genetic algorithm is constructed.Emphasis is laid on the classification of three and six types of samples.Using the simulation experimental data of reciprocating compressor crank pin fault,the training set and test set samples after feature selection are normalized and input into GA-SVM model.Compared with PCA-SVM,BP-SVM,SVM and GA-SVM model without feature selection,the effectiveness and superiority of GA-SVM model based on compensation distance evaluation technology is proved.