Effective Gear Teeth Defect Identification Using Multi-Domain Feature Extraction

Abstract

Condition monitoring using vibration measurement is a most commonly used non-destructive technique. This paper proposes application of statistical features of vibration signal for gear defect diagnosis. The advanced signal processing of acquired signals to find the most significant features of the defects in gear system is the aim of the present work. An experimental investigation that examines the diagnostic potential of multidomain features for gear teeth defect identification is presented. It is concluded that db44 is a better mother wavelet function in the time - frequency domain as compared to db4 or db10, while standard deviation, variance and an absolute maximum of continuous wavelet transform and discrete wavelet transform are better features for gear defect identification in addition to conventional time and frequency domain features for training purposes of intelligence systems.