Bearing fault diagnosis based on the active energy conversion of generalized stochastic resonance in fluctuating-frequency linear oscillator

Abstract

Noise-assisted stochastic resonance (SR) method has been extensively applied in bearing fault diagnosis. In order to cast off the basic constraint on nonlinearity in classical SR systems, and overturn the passive energy conversion from external noise to signal, a fluctuating-frequency linear oscillator (FFLO) is proposed and combined with the generalized scale transformation (GST) to overcome the small parameter limitation in this study. The results of output power amplification reveal that the proposed GST-FFLO system displays a rich variety of generalized SR (GSR) behaviors, which play an active role in the optimal energy conversion from internal regulatable noise to weak bearing fault signal. Moreover, it is also found that the undamped GST-FFLO system always produces more significant GSR peaks, thereby improving the precision and efficiency in the energy conversion. Finally, the experimental results demonstrate that the proposed method is always valid and exhibits the superiority in diagnosis performance and operating efficiency in several typical difficult cases, namely, impulse interference, low signal-to-noise ratio, and multiple faults.