Time-frequency demodulation analysis for gearbox fault diagnosis under nonstationary conditions

Abstract

Gearbox fault diagnosis is important for ensuring the safety and running quality of many sorts of machinery. When fault exists, gearbox vibration signals feature amplitude modulation as well as frequency modulation, resulting in complex frequency structure. Besides, gearboxes often work under nonstationary conditions due to time-varying speed/load, resulting in nonstationary vibration signals which makes conventional frequency domain analysis ineffective. In this paper, we propose a time-frequency demodulation analysis method for presenting time-frequency spectra of both amplitude envelope and instantaneous frequency of analyzed signal, which directly reflect the fault frequency components on time-frequency plane without complex sidebands. To fulfill the mono-component requirement of instantaneous frequency estimation by Hilbert transform, iterative generalized demodulation is effectively exploited to decompose multi-component nonstationary signals into mono-components. Both numerical simulated and lab experimental gearbox vibration signals under nonstationary conditions are analyzed by the proposed method, and gear faults are effectively detected.