Vibration Analysis of Gearbox Fault Diagnosis using DWT and Statistical Features

Abstract

The gearbox is one of the critical subsystems in any rotating machinery, which plays a significant role in machine-driven power transmission in terms of change in speed and torque. It plays a vital role in patching different industrial functionalities. The advent of developing different gear technologies and the requirement to fulfil the desired mechanical benefits lead to add more importance to the gearbox health condition monitoring from various types of fault occurrences at an earlier stage. This study presents the vibration analysis of gearbox fault diagnosis using discrete wavelet transform (DWT) and statistical features. It is observed that, using wavelet reconstruction in the fault diagnosis better fault classification is achieved. The fault diagnosis has presented with an emphasis in time-domain followed by two different approaches. The approach-1 is illustrated as windowing of raw signal, feature extraction and feature classification using support vector machine (SVM). In the approach-2, after windowing the raw signal - each window of original vibration signal has converted into wavelet coefficients reconstructed signals (without leaving the time domain) using discrete wavelet transform (DWT) at different levels of decomposition followed by approach-1. The fault diagnostic accuracy of SVM is presented with 100 Monte-Carlo -runs to validate the consistency in the accomplished result. By observing the success rates in two approaches, it is clear that approach-2 with wavelet coefficient’s reconstructed signal is providing better classification accuracy, which can be practically deployed to diagnose the gearbox fault.