A piecewise nonlinear stochastic resonance method and its application to incipient fault diagnosis of machinery

Since noise is beneficial to improve weak fault characteristic extraction in nonlinear systems where a stochastic resonance occurs, a new piecewise nonlinear stochastic resonance (PNSR) method is proposed to enhance and extract the incipient fault signatures of machinery and ensure their reliable operation. In this new PNSR method, a piecewise nonlinear potential is employed to overcome the saturated shortcoming of classical bistable potentials, thereby improving and highlighting the performance of stochastic resonance in weak fault characteristic extraction. Moreover, the proposed PNSR method can realize the adjustment of the potential structure by a single parameter, it avoids the shortcoming that the imperfect matching classical bistable stochastic resonance (CBSR) parameters results in low output signal-to-noise ratio (SNR), and further identify fault characteristics are difficult. In addition, the output SNR of the PNSR method is deduced by using adiabatic approximate theory to investigate the effect of system parameter on the PNSR method, and even has higher output SNR compared with the CBSR method and the proposed PNSR method. Finally, the proposed PNSR method is validated by utilizing simulation and bearing experiments. All diagnostic results show that the proposed PNSR method can effectively extract weak fault characteristics, thereby achieving the incipient fault diagnosis of machinery. Compared with the CBSR method, the proposed method has higher spectrum peak values in fault characteristic frequencies. Therefore, it has better recognition degree in whole frequency spectrum.