Adaptive Generalized Demodulation Transform Based Rolling Bearing Time-varying Nonstationary Fault Feature Extraction

Abstract

Fault impulse amplitude and fault impulse intervals of rolling bearing are time-varying under nonstationary conditions. The existing research mainly focuses on solving the problem of spectral smearing caused by time-varying impulse intervals, and few research on the problem of time-varying impulse amplitude. Under low rotational speeds, the magnitude of the fault impulses is small, and it is easily overwhelmed by noise, which makes it difficult for rolling bearing fault diagnosis. As such, an adaptive generalized demodulation transform (GDT) based rolling bearing fault diagnosis method under nonstationary conditions is proposed. The reset criterion is developed to improve the GDT, whose conversion factor is adjustable and the reset factor is optimal. As a result, the time-varying fault-related frequency curve is adaptively concentrated on the highest point instead of the starting point. The generalized characteristic index (GCI) model is defined via the hypothesis and the time-frequency ridge, extracted from the envelope time-frequency representation of filtered signal. The fault-related frequency ridge is transformed by the adaptive GDT, combined with spectrum based quantitative characterization and the GCI, the rolling bearing is diagnosed without the rotational speed measurement. The analysis results of simulated and measured signal demonstrate the effectiveness of the proposed method.