Detection of Gear Faults in Variable Rotating Speed Using EEMD Decomposition and Instantaneous Frequency

Abstract

When a local gear fault is presented, both the amplitude and phase of the tooth meshing vibration are modulated. If the rotating speed of the shaft is invariable, the gear-fault-induced modulation phenomenon which manifest as frequency sidebands equally spaced around the meshing frequency and its harmonics in vibration spectra. The Hilbert transform has been widely used in demodulation of such signals and has given good results. However, under variable rotating speed of the shaft, the meshing frequency and its harmonic and the sidebands vary with time and hence the vibration signal becomes non-stationary. The use directly of the Hilbert transform doesn’t allow detecting the variation of the rotating machine and its harmonics which reflect the gear fault. In this study, we propose to use first the ensemble empirical decomposition (EEMD) which is particularly suitable for processing non stationary signals. By using EEMD the signal can be decomposed into a number of IMFs which are mono component, and then we use the Hilbert transform to calculate the instantaneous frequency and the envelope of each IMF. To identify the fault, we apply the ensemble empirical decomposition (EEMD) in second time to the instantaneous frequency to obtain mono component frequency and we calculate the spectrum of each IMF to evaluate the frequency. In this works, to validate this strategy, we analyze simulated signals for healthy and faulty gear boxes when the speed of machine is regular and variable; these models are based on the models of McFadden.