Stochastic Resonance Array for Designing Noise-Boosted Filter Banks to Enhance Weak Multi-Harmonic Fault Characteristics of Machinery

Abstract

Noise in stochastic resonance (SR) is beneficial to enhance weak useful information and has been widely stuided and applied to real-world signal detection. In tristable SR, however, the depth of middle potential well is not equal to those of both left and right potential wells and adjusting the system parameters would change the depth and width of three potential wells simultaneously. Therefore, the role of potential-well width and potential-barrier height or potential-well depth in a triple-well potential on SR has not been investigated separately yet and the effect of a both uniform depth and different depth between left/right potential well and middle well in a triple-well potential on SR has not been explored yet. Meanwhile, SR is used to weak signal detection and early fault diagnosis, but it ofter processes a complex signal into a sine shaped one, resulting in the loss of multi-harmonic signature embedded in the raw signal. To advoid this, this paper investigates underdamped SR induced by a symmetric triple-well potential with a uniform depth. Then, a SR array based noise-boosted filter bank method is proposed to enhance weak multi-harmonic fault characteristics for early fault diagnosis of machinery. Theoretical results using spectral amplification factor to quantify SR indicate that tristable SR induced by a symmetric triple-well potential with a uniform depth is supperior to that induced by a symmetric triple-well potential with a different depth, increasing the potential-well width would maximize the spectral amplification factor of the tristable SR but keep the optimal noise intensity unchange basically, and increasing the potential-well depth would decrease the spectral amplification factor of the tristable SR but enlarge the optimal noise intensity obviously. Experimental results demonstrate that the proposed method is able to enhance weak multi-harmonic fault characteristics for early fault diangosis of roller bearings. Comparing with accugram and minimum entropy deconvolution combined with spectral kurtosis methods, the proposed method could detect weak fault characteristics and their harmonics succesfully but other two methods fail to extract fault signature.

Keywords: Mechanical fault diagnosis, Weak signal detection, Stochastic resonance, Nonlinear filters, The benefits of noise