Sound-assisted vibration resonance for weak fault detection in machinery

In nonlinear models, stochastic resonance (SR) is one of the most famous nonlinear phenomena in the past three decades. It completely overturns the traditional concept of “noise is useless”. SR can effectively use the energy of noise to enhance weak signals of interest. Compared with SR, vibration resonance (VR) uses a high-frequency excitation to soften the stiffness of the nonlinear models, change their equilibrium points to bifurcate, thereby enhancing weak useful signals. However, in the field of mechanical weak fault diagnosis, the early naturally developed damage is relatively weak and overwhelmed by a large number of interference frequencies from noise and other signals excited by normal components. In the face of this situation, the traditional VR method that simply uses the high-frequency excitation and vibration signals to enhance weak damage features is not far good enough. Therefore, this paper proposes a method to boost the mechanical weak fault detection by exciting VR using sound-assisted vibration signals. The coupling coefficient is used to control the degree of assistance of the sound signal, and the response amplitude gain is used as an indicator to quantify the weak signal detection performance, so as to improve the weak fault detection capability under extremely low signal-to-noise ratio conditions. Then, the effectiveness of this proposed method is verified by bearing fault experiment. The experimental results show that the proposed method can effectively improve the capability of weak mechanical fault detection. Finally, its superiority is further demonstrated by comparing it with the underdamped time-delay feedback VR method, the classic denoising method and the fast spectral kurtosis method.