关键词: 故障诊断, 减压阀, 功率谱, 双谱切片, 小波包, BP网络

Abstract:

Compared with AR power spectrum,bispectrum can eliminate Guassian noise effectively, and preserved singals' phase information. Both of the AR power spectrum and bispectrum obtained from  normal singals and fault singals which were acquired from a pressure relief valve's vibration show different peaks characteristics. This paper extracted features from  power spectrum and bispectrum' slices of the two kinds of singals, then the features were used as the input parameter of a BP neural network to diagnose faults of the valve. The experimental results indicate clearly the performance difference of bispectrum and AR power spectrum.

Key words: fault diagnosis, relief valve; , power spectrum, bispectral slice, wavelet packet, BP network