Wavelet Time Scattering Features for the Bearing DefectsClassification

Abstract

Rollingelementbearingisoneofthemostcriticalcomponentsinrotatingmachinerywithawiderangeofapplications.Bearingsareusedfromsmallhouseholddevicestolargemachinery.Localizeddefectssuchaspits,spall,dents,etc.aregeneratedon contact surfaces of bearings due to cyclic loading during operation. Failure of bearings may cause production loss due todowntime and breakdown in machinery. It is therefore essential to diagnose bearing defects at the early development stage. In thepresent work, we propose a new Wavelet Scattering Transform (WST) based technique for the bearing fault classification.Leveraging concepts from signal processing and wavelet transform, WST is used to extract features that are stable to minordeformation and time-invariant. This study demonstrates that WST features can be used with various classification techniques togain improved classification discrimination. For the validation of the proposed algorithm, experimental data for normal and defective bearing are obtained from case western reserve university bearing data centre. It has been observed that proposed wave lets cattering transform methodology provide the better approximation off adult diagnosis.