Limitations of vibration demodulation analysis application in machine fault diagnosis

Demodulation analysis is widely used to diagnose the faults of machines fixed gear and bearing. From theoretical analysis, we know the existing demodulation methods have three limitations as following: when we demodulate two time-domain adding signals without modulating information (fault information), the subtraction of the two signals' frequencies will display as the result of demodulation; in generalized detection-filtering demodulation analysis, the effect of frequency aliasing may arise as a result of getting absolute value or detection; in several new Zoom demodulation methods, because digital low-pass filtering can not be carried out in the process of Zoom sub-sampling, it is possible to occur the phenomenon of frequency aliasing at low frequency range, which is the overlap of the higher harmonics of modulation frequency. This paper presents an optimized Hilbert Transform demodulation method based upon complex analytic band-pass filter, which can overcome the three limitations if the analytical frequency of modulation and filter's bandwidth satisfy a strict mathematics relation between themselves. Meanwhile, we can integrate band-pass filtering, Hilbert transform and sub-sampling as a whole to obtain a fast operation. It is an optimized demodulation method.