Bearing defect diagnosis by stochastic resonance based on woods-saxon potential

Abstract

The interference from background noise makes it difficult to identify the incipient bearing defect via vibration analysis. Stochastic resonance (SR) is a nonlinear phenomenon which is characterized by that the output signal can be enhanced with the assistance of the proper noise. The Large Parameter Bistable SR (LPBSR) method is commonly used in the bearing fault diagnosis. However, the LPBSR method requires signal tuning to satisfy the small parameter requirement (both amplitude and frequency are far less than 1), which implies the inherent structure of the input signal is modified and the effect of SR may be further affected. Additionally, a barrier exists in the bistable model, which indicates that the particle motion in the bistable potential is unstable and then the external noises induced by the unstable motion are introduced in the output signal. This chapter proposes a new strategy to realize bearing defect diagnosis, that is, a Woods-Saxon potential instead of a conventional bistable potential is utilized to achieve SR and to enhance the output signal-to-noise ratio (SNR). In the proposed method, the output SNR can be optimized just by tuning the parameters of the Woods-Saxon potential. This method overcomes the limitation of the small parameter requirement of the classical bistable SR, and can thus detect a high driving frequency. Furthermore, the smooth Woods- Saxon potential leads to a more stable particle motion compared to the bistable potential, which provides a more regular output waveform and reduces the unexpected noises at the same time. The proposed method has yielded more effective results than the traditional methods, which was verified by means of a practical bearing vibration signal carrying defect information.