Gear fault feature extraction based on cICA-Kurtogram under variable speed

Abstract：Constrained independent component analysis (cICA) is very sensitive to source noise in the fault feature extraction of gear fault under varying speed.For this problem, a method combining cICA, computed order tracking (COT) and fast-Kurtogram was proposed.According to the instantaneous rotating frequency of the motor, the instantaneous meshing frequency of the fault gear was obtained to establish rectangular wave as reference signal.The time domain signal with more obvious fault features was extracted by cICA, namely the independent component of interest (ICI) .Then ICI in the time domain was converted into angular stationary signal by equal-angle resampling based on the instantaneous phase of the shaft, where the failed gear was located.Feature orders of fault gear were extracted in order spectrum.According to the fast-Kurtogram, the parameters of appropriate filter were selected and the ICI in the angular domain was filtered.The position of shaft, where the fault gear located, was determined by square envelope spectrum.Experimental results showed that this method was effective for monitoring and fault diagnosis of gearbox vibration under variable speed.

Keyword：variable speed gearbox;fault feature extraction;constrained indepent component analysis;computed order tracking;fast-Kurtogram