Reciprocating Compressor Fault Early Warning And Diagnosis

Reciprocating compressor is the core operating equipment in petrochemical industry.The research on fault early warning and diagnosis can not only ensure the stable operation of the equipment,but also have important significance for the protection of equipment and personnel safety.The reciprocating compressor of an offshore natural gas production platform of CNOOC(China National Offshore Oil Corporation)is taken as the research object in this paper.Based on the research and summary of various feature extraction methods and time series data prediction model based on neural network,a time series prediction method based on E-CRBM(Error Continuous Restricted Boltzmann Machines)and a fault diagnosis method based on STF-DBN(Spatio-Temporal Features Fusion based on Deep Belief Network)are proposed.And they are applied to the fault early warning and diagnosis of reciprocating compressor.The specific research contents of this paper are as follows:In order to solve the problems of delayed prediction results in traditional time series prediction model,a time series prediction model based on CRBM(Continuous Restricted Boltzmann Machines)and NN(Neural Network)is proposed.CRBM is used to extract time series features and reduce sample dimension.The time feature is transformed by NN.Finally,CRBM is used to reconstruct the data to complete the prediction.The experimental results show that this method can effectively suppress the prediction lag.In order to further improve the accuracy of the prediction model,E-CRBM is proposed based on CRBM.On the basis of analyzing the feature error sequence obtained by NN,E-CRBM adds noise to the hidden layer of CRBM,in which the noise probability distribution is obtained by fitting the probability distribution of the error sequence.Compared with CRBM,E-CRBM can enhance the robustness of the prediction model and improve the overall prediction accuracy.Therefore,it is used as the trend prediction algorithm of compressor operating parameters.After parameter prediction,a fault diagnosis method based on STF-DBN is proposed to solve the problem of feature extraction and few fault samples of reciprocating compressor.This method fuses the information collected by various types of sensors of reciprocating compressor.In order to improve the sensitivity of the diagnosis model to faults,features are extracted from time and space,and the weights of features are allocated.The actual operation data of reciprocating compressor on offshore platform are used to test.The results show that the method can effectively detect the faults of gas valve and motor,and can give early warning 3 hours in advance.