Application of multi-scale noise tuning parameter-induced stochastic resonance for planetary gearbox diagnosis

Abstract

ABSTRACT Multi-scale noise tuning parameter-induced stochastic resonance method (MNTPSR-1/f β ) is proposed for planetary gearbox diagnosis. This method is divided into two steps. The first step is called multi-scale noise tuning (MNT) and the second step called parameter-induced stochastic resonance (PSR). MNT based on wavelet packet transformation tunes the noise to an approximate 1/f β noise which can improve PSR effect. PSR enhances the weak high-frequency fault characteristic signal. Some simulations are designed for analysing MNTPSR-1/f β performance such as the MNT effect, anti-noise capability, and optimal frequency response. A fault diagnosis strategy for the planetary gearbox is proposed. To validate the proposed method, two pre-planting fault tests of sun gear and ring gear are carried out. For comparison, PSR without MNT, MNTPSR-1/f and MNTPSR-1/f β are used to analyse the fault signals. The simulations and case validations prove that our proposed method is the most effective. Abbreviations: wavelet packet transformation (WPT), stochastic resonance (SR)