A weighting function for improvement of spectral coherence based envelope spectrum

Abstract Envelope analysis plays an essential role in fault diagnostics of rotating machinery. One popular method to attain the envelope spectrum is to integrate the spectral coherence over a chosen spectral frequency band. Considering the recently developed algorithm for fast estimation of spectral coherence, the present study proposes a weighting function used for the integration. Noting that the spectral coherence for a cyclostationary signal appears with a periodic pattern along the cyclic frequency axis, the weighting function is designed by calculating the autocorrelation over the cyclic frequency domain of the spectral coherence. This improves the envelope spectrum estimation, as spectral bands with high autocorrelations are amplified in the course of integration. Without consideration of fault characteristics, the proposed idea alleviates the burden of selecting the frequency band for integration. Experimental measurements from the cavitating propeller and faulty bearing demonstrate its usage in challenging scenarios such as noisy situations or multiple types of faults.