| Considering the fact that non-stationary variables exist in modern systems, whichcause the trouble with fault diagnosis using traditional signal processing method, andthose strategies can’t provide the status information of the entire system since theyusually deal with single variable. To overcome these difficulties, cointegration from theeconometrics is introduced to establish the modal of the long term equilibrium amongsystem variables, and applied to the task of fault diagnosis.This paper explores the application of cointegration used in fault diagnosis of thehydraulic servo system based on the previous work of condition monitoring. If systemvariables are integrated of first order, the system variables could establish a cointegrationmodal. The cointegration vector matrix is proposed as the characteristic parameter ofnon-stationary system fault diagnosis. The classification algorithm of supporting vectormachine (SVM) is used to train the model and test the accuracy of classification. Thisnew method is applied to the fault diagnosis of a simulated hydraulic servo system,5system variables are used as the cointegration variables. The test results indicate thatcointegration vector matrix has great fault diagnosis ability in a typical non-stationarysystem.Based on the cointegration modal, the stationary cointegration modal residuals(innovations) are used to detect the fault of the system because the residuals containvarious system information. In this paper, analysis both in time and frequency domain areused to analyze the innovations of the cointegration modal, specifically observing thetime-domain waveform and calculate the common statistics and also the power spectrumdensity of innovations. The results show that innovations are sensitive to fault status likestiffness failure and damping faults, but not effective on leakage fault. Wavelet variancemethod and wavelet power spectrum are successfully applied in the leakage faultdiagnosis. |
PDF Full Text Request