Demodulated synchrosqueezing S-transform and its application to machine-fault diagnosis

Abstract

Abstract The time–frequency analysis (TFA) technique has been viewed as a useful tool for processing non-stationary signals in the field of industrial machinery. Rub-impacts of a rotor system will cause vibration of the rotor and stator, thus any vibration signal with rub-impacts will be accompanied by high-frequency oscillation characteristics. In this paper, a novel TFA algorithm, termed a demodulated synchrosqueezing S-transform (DSSST), is proposed to extract the strong time-varying features in rub-impact vibration signals. The DSSST method is based on a modified S-transform, and introduces a pre-processing technique, a demodulation algorithm, to partially demodulate the oscillated modes for rub-impact identification. Meanwhile, a synchrosqueezing transform is utilized to further sharpen the time–frequency representation. Assisted by the proposed method, the rub-impact phenomenon and its impact frequency are clearly recognized through experimental and real validations.