A New Instantaneous Wavelet Bicoherence for Local Fault Detection of Rotating Machinery

Abstract

Practical vibration signals resulting from fault-included rotating machinery are nonlinear and nonstationary in nature, which bring new challenges for the linear representations and the stationarity assumption methods. Wavelet bicoherence (WB) is considered an effective method for these nonlinear signals analysis. However, the current WB model ignores the phase information of the nonlinear signals and is often estimated by integrating over the finite-time interval. Thus, its direct application may cause spurious bicoherence peaks and the transient information loss of the nonstationary signals. To overcome these limitations, a new instantaneous WB model is established to extend the application of WB for nonstationary signal analysis. In this method, the instantaneous biphase information is used first for bispectrum calculation, and then, the algorithm based on ensemble average of instantaneous phase randomization is introduced to eliminate the spurious bicoherence. Finally, the bicoherence is estimated in the time-frequency domain. The effectiveness of the proposed method is validated by mathematical discussion, simulations, and experiments. Results illustrate that, compared with the commonly used method, the proposed method provides an alternative solution for local fault detection of rotating machinery.