Bearing envelope analysis window selection Using spectral kurtosis techniques

Envelope Analysis is a well-known signal processing technique for bearing fault detection. However, improper window selection can result in poor fault detection performance. Using a known fault data set, we quantify the performance of spectral kurtosis (SK) and envelope kurtosis (EK) as a technique for setting an optimal frequency and bandwidth window for the envelope analysis. We establish a measure of effectiveness (MOE): the correlation of fault energy with total spall length. With this MOE, we evaluate the ability of SK/EK to predict the optimal envelope analysis window.