Paper Title
The Envelop-Based Heterodyne Bispectrum Analysis for Motor Bearing Defect Detection

Abstract
Vibration signals produced by defects in motor rolling element bearings present a highly non-linear and non- Gaussian behavior. These defects produce characteristic fault frequencies that cause amplitude modulation of vibration signatures and also tend to modulate the structural resonance. The key task to diagnose bearing faults is to detect and identify sidebands in the frequency spectrum. The conventional bispectrum has difficulty in identifying the sideband pairs of sum and difference frequencies simultaneously because the expression for the conventional bispectrum is insufficient to reveal the nonlinear coupling effect in the modulated signal. In order to overcome the shortcomings in the conventional bispectral analysis method, a new tool called “heterodyne bispectrum” is proposed, which could interpret the bispectral analysis of vibrations in the presence of amplitude modulation more precisely. Due to the great influence of the high-frequency structural resonance, the low-frequency bearing characteristic frequencies and their harmonics are easily smeared and unable to be identified. Therefore, a new approach based on the fusion of envelope analysis and heterodyne bispectrum, called the “envelope-based heterodyne bispectrum” is employed for fault detection and diagnosis of motor bearings. Numerical simulations as well as vibration signals from bearings containing outer-race faults are used to demonstrate the effectiveness of this proposed technique. Index Terms—Heterodyne bispectrum; Envelope analysis; bearings; defect detection.