Fault Diagnosis of Pressure Relief Valve Based on Power Spectrum

In normal and fault status, the vibration signals of pressure relief valve shows different peaks characteristics. Wavelet packet can decompose signal's low and high frequency components simultaneously, then a signal's features can be extracted after its wavelet packet decomposition and reconstruction. The BP network can realize complex nonlinear mapping relations, and it can complete the duty of complex pattern recognition. This paper uses the features extracted by wavelet packet from AR power spectrum as the BP network input, then diagnose the valve's faults, and the result is satisfactory.