Multi-frequency weak signal detection based on multi-segment cascaded stochastic resonance for rolling bearings

Abstract For rotating machinery, vibration signals excited by its faulty components provide rich condition information for its fault diagnosis and condition-based maintenance. However, strong noise severely influences the accurate detection of incipient faults. Thanks to the ability of enhancing weak input and suppressing the noise, the stochastic resonance (SR) has been applied to weak signal detection in some fields, and the improvement on its performance are still being concerned, especially in the mechanical engineering. For multi-frequency weak signals, this paper proposes an improved mechanism for the SR, called multi-segment cascaded stochastic resonance (MS-CSR). In this method, the input signal obtains segment enhancement by using some bistable SR models, and series connection of such a unit compose an improved cascaded SR (CSR) system, which can not only gradually enhance the weak signals of interest, but also pay more attention on the signal with relatively small amplitude at the initial. A modified measurement index, named alliance signal-to-noise ratio (ASNR) is defined to evaluate the detection performance of the proposed SR method, as well as the parameter selection for the MS-CSR system. In this index, a weight factor is introduced to influence the assignment of noise energy in the SR, so that the relatively weak signal in the multi-frequency input signal can obtain a high priority to make the resonance phenomenon happen and avoid the misdiagnosis. A simulated signal and an experimental vibration signal collected from a faulty bearing are used to verify the effectiveness of the proposed MS-CSR method. The results demonstrate that the MS-CSR is a useful tool for detecting weak signals with multiple characteristic frequencies.