Gearbox fault diagnosis: A higher order moments approach

Abstract

The paper presents a higher order moment (HOM) based gearbox fault diagnosis in which the logarithmic amplitude of HOM (LHOM) is proposed as a new feature for gearbox fault diagnosis rather than traditional time domain features. Three benchmark gearboxes datasets at different localised fault conditions are taken to generalise the results obtained from the LHOM approach. The results demonstrate that the LHOM exhibits a linear trend with order of moments and is directly proportional to the severity of faults beyond the fifth order moment. At the end, different case studies are performed by using five popular ML-classifiers and three feature selection techniques on both types of features (conventional and proposed). The results from three distinct gearbox datasets under different localised fault conditions demonstrate that the LHOM features can also be used as extracted features for diagnosing gearbox faults.