Research On Fault Diagnosis For Electro-hydraulic System Based On Wavelet Analysis

There are two main difficulties in fault detection and diagnosis of electro-hydraulic servo (EHS) system. The most difficult is in lack of fault signature tool, which comes from the features of the EHS system such as time varying, non-linear and coupling of fluid and solid. In fault diagnosis, abrupt signal is often connected with a kind of faults. In addition, EHS system integrates electrical and mechanical parts into a tiny package. Even one knows there is fault in the system, sometimes he is not sure where the fault is. This problem has not been overcome for a long time.This paper tries to provide a solution to this problem. A wavelet analysis is introduced to detect and diagnose faults occurred in the EHS system. As well known, wavelet analysis is referred to as a mathematical microscope. It takes advantage to Fourier transform in its multi-resolution analysis both in time domain and in frequency domain. It is the reason why wavelet analysis finds so wide application in the area of signal processing and fault diagnosis. It is also the reason why it is applied in fault detection and diagnosis of EHS system, because the time response and the frequency performance are of the same importance in this type of system.After comparing the FFT(fast fourier transform) and WT(wavelet transform), this paper gives the selection firstly. WT is chosen to analyse the pressure response signal from the EHS. Then, this approach is applied to detect and diagnose several kinds of faults in the EHS system. One is null oscillation occurred in the EHS valve, the second is the failure in the torque motor circuit, and the third is the load disturbance in the actuator. All these fault symptoms are extracted by wavelet analysis. Besides, the faulty conditions are compared with faulty free conditions in the form of wavelet transform.But there are some problems existing on wavelet analysis when directly applied for control systems. As a solution, this paper develops a wavelet model based approach (WMBA). In this approach, a model in DWT form is built for the EHS system at its healthy condition. This model is used to generate residual signal by comparing the healthy signal and the signal in the running system. By processing the residual signal, faults from incipient fault to serious fault can be detected effectively. As the conclusion, wavelet analysis can be a powerful tool in fault diagnosis of control systems.