Classification of High Frequency Impact Signal in Vibrational Analysis of Spur Gears by using Convolutional Neural Networks

Abstract

 Spur gears are one of the widely used  gears in a gearbox assembly. They often require  lubrication and replacement of pinion and gears as prone to  damage in high speed shafts with heavy loads and adverse  working conditions. These creates spalling and breakage of  gear tooth due to material fatigue from excessive loads and  also forms pitting corrosion due to reduced lubrication and  higher input shaft speeds. Vibrational analysis of these  rolling elements is necessary for monitoring the health of  gears periodically. These graphs provide a pattern  waveform over time to study the characteristic high  frequency impact noise signal peaks due to increased  vibrations from faulty sections. This paper depicts about  the implementation of convolutional neural networks to  analyze the vibrational graphs obtained at different rotating  speed of shafts for various gear ratios to plot the high  frequency impact noise and train the neural networks to  identify the peaks and classify among the faulty and  healthy spur gears and pinions for a better way to reduce  time in estimating the remaining average working life of  gears and perform adequate maintenance of components.