Wavelet Resonance Demodulation Method and Its Application in Fault Diagnosis of Railway Bearings

Abstract

Resonance demodulation method other than other diagnosis methods, such as AR model method and wavelet transform method, is adopted by engineers to diagnose the faults of rotating machinery, but its effectiveness depends to a great extent on how to choose the center frequency and the bandwidth of the filter. Here a new resonance demodulation method based on wavelet is presented, which applies time-frequency transform on the acceleration signals collected and extracts the time-energy signal from the time-frequency spectrum instead of filtering the original acceleration signals and then extracting the envelope. The key part of the method is that a criterion in terms of the minimum of kurtosis factor is proposed and the time-energy signal, which is equivalent to envelope, can be automatically derived from the time-frequency spectrum according to the criteria. The method is applied to the analysis of vibration signals of rolling bearings with outer-race inner ring faults and rolling element faults, and the results show that it performs effectively in fault diagnosis of freight car rolling bearings and performs more effectively than conventional fault diagnosis methods, say, demodulated resonance technology, to extract the fault characteristic.