Application of Sound Signature Emitted by Rolling Element Bearing For Identification of Faulty State

Abstract

Machinery fault identification is an important part of condition based maintenance activity within the industries. Among several techniques, Sound based signal analysis is less explored technique for identification of machinery faults. In the present work, the problem of fault identification of rolling component bearing is taken. The acoustic signatures of healthy and defective bearings have been obtained from experimental test rig for variable radial load and shaft rotation speed. The aquisition has been done using commercially available smartphones and studio grade microphones. The aquired signals with suitable signal processing (Hilbert Transform) are converted in the frequency domain for identification of specific location of fautls. The statistical features like Kurtosis, RMS, Skewness and Crest Factor are extracted from time domain signals from all aquired sound signatures. A Support Vector Machine (SVM) is trained using these paramters for classification of healthy and faulty condition of bearing. The result of classification shows clear bifurcation between healthy and faulty bearings based on sound signature emitted.