Extension of Harmonic Product Spectrum and Its Application in Bearing Condition Monitoring

Accurate condition monitoring and timely fault diagnosis are very important for rotating machinery equipment. The harmonic product spectrum (HPS) based on envelope spectrum has been proven to be an effective fault diagnosis tool by revealing the cyclostationarity of fault signals of rotating machinery. However, its performance under different interfering noises has not been well explored and revealed. In this paper, based on the recently proposed generalized envelope spectra, three new versions of HPS are first proposed using log-envelope spectrum, fractional envelope spectrum and squared envelope spectrum respectively. Subsequently, the effects of discrete harmonic noise, random impulse noise and Gaussian background noise on the performance of envelope spectrum-based HPS and three new versions of HPS are comparatively investigated by numerical simulation signals. The results show that different versions of HPS show different robustness to different interference noises. The robustness of HPS based on log-envelope spectrum to random impulse noise is stronger than that based on envelope spectrum and squared envelope spectrum. When Gaussian background noise is dominant, the four HPSs have similar capabilities to reveal cyclostationarity. Finally, four HPSs are applied to the condition monitoring of rolling bearings by virtue of their capability to characterize cyclostationarity. The experimental results show that the four HPSs can not only effectively monitor the degradation status of rolling bearings but also accurately diagnose early faults. This work provides new and effective tools for condition monitoring of rolling bearings.