Fault Diagnosis Of Gas Pressure Regulators Based On CEEMDAN And Feature Clustering

With the development of the west to east gas transmission project and other major national projects,China’s annual natural gas consumption continues to grow.Gas supply system has become one of the essential infrastructures.As the core part of gas pipeline network,the safe and stable operation of gas regulator has become an important guarantee for the daily life of urban residents and the sustainable development of industry.Due to the large scale of the gas pipeline network,the daily maintenance and repair of the gas regulator is huge and tedious.Any fault of the gas regulator in the gas pipeline network may bring the risk of gas leakage or even explosion.Therefore,an effective fault detection and identification approach for gas pressure regulators has become an urgent problem to be solved.In the gas network,the stability of gas regulator outlet pressure signal is an important parameter to reflect its performance and working state.It is also the key of fault diagnosis.However,due to the influence of various interference factors in the gas pipeline,the signals collected by the gas regulator often show non-linear and nonstationary,which can not be simply identified.Therefore,this thesis paper proposes a fault diagnosis method which combines Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMAN)and Fuzzy c-Means Clustering(FCM)to diagnose the common fault types of gas regulator.In this thesis,the CEEMDAN method is used to decompose the pressure signal at the outlet of the regulator into a series of intrinsic mode functions;And then the characteristic vectors of typical faults are established according to the Hilbert marginal spectrum(HMS)constructed by IMF;By collecting the pressure data of typical fault types,the FCM clustering algorithm is used to construct the clustering center of each fault type;Finally,the gas pressure data of residential buildings is collected,and the fault types are analyzed by calculating the membership degree between the sample data and the clustering center Identify.In addition,EMD and STFT are used to extract the fault features of the gas regulator outlet pressure signal to verify the superiority of the proposed fault diagnosis method.The experimental results show that the fault recognition rate of STFT method is79.6%.The recognition rate of EMD method is 87.0%.Using the fault diagnosis method based on CEEMDAN which is proposed in this thesis,the fault types of all test samples are correctly identified,and the membership degree between the test samples and the corresponding fault types is basically stable in the range of 0.9 ～ 1.Compared with support vector machine(SVM),artificial neural network(ANN)and other methods derived from machine learning,this method is data-driven,does not need the basic model,requires less data,training time is short,and training cost is low.