Improved empirical wavelet transform method based on spectrum trend for gearbox fault signal processing

Abstract

The empirical wavelet transform decomposes the signal by segmenting the Fourier spectrum and constructing an adaptive filter bank and then segments the spectrum by detecting the maximum value in the spectrum. However, the number of mono-components is unknown or difficult to confirm when processing actual gearbox fault signals, the result of decomposition will be neither reasonable nor conducive to demodulation analysis if the maximum value is concentrated in one component. To solve this problem, a concept of an amplitude modulation–frequency modulation signal in the form of a wave crest is proposed. The spectrum trend of the Fourier spectrum of signal is extracted by the empirical mode decomposition method, and then the number of mode components is presupposed and the spectrum is segmented according to the extracted spectrum trend. The extraction method of the spectrum trend based on improved empirical wavelet transform based on spectrum trend method is proposed for gearbox fault diagnosis. The experiment, including the vibration test platform of horizontal driving machine and the processing of experimental signals based on the improved empirical wavelet transform based on spectrum trend method are also given. According to the simulation and experiment results, it can be clearly found that the spectrum trend can describe the distribution of signal spectrum well, and the components obtained by the improved empirical wavelet transform based on spectrum trend method are more suitable than original empirical wavelet transform method for demodulation analysis.