Gearbox fault diagnosis using acoustic signals, continuous wavelet transform and adaptive neuro-fuzzy inference system

Abstract

This paper proposes a method of diagnosing faults in a single stage spur gearbox based on an analysis of acoustic signals acquired under various fault conditions. The time domain acoustic signals acquired from the gearbox are converted into a number of angular domain signals, each representing one revolution of the gearbox drive shaft. The resultant angular domain signals are averaged in order to improve the signal to noise ratio. The angular domain averaged signals thus obtained are decomposed using continuous wavelet transform. A range of optimal scales is then identified based on the energy-Shannon's entropy ratio of continuous wavelet coefficients. The wavelet amplitude maps pertaining to the various gear health conditions are segmented into 6 parts and continuous wavelet coefficients from optimal scales fed directly to the ANFIS in the form of data samples. The results demonstrate that acoustic signals and ANFIS can effectively be utilized to diagnose the condition of the gearbox. © 2018 Elsevier Ltd