A Novel Fault Diagnosis of Bearing Approach using a Hybrid of Empirical Mode Decomposition and Discrete Wavelet Transform

摘要

To ensure a flawless working of a machine, fault diagnosis and condition monitoring is very critical. Talking about automobile industry, aviation industry, or robotics, mechanical inspection is essential. Bearing being one of the most important components in almost every mechanical system from spindle of a lathe to the wheel of an aircraft. So, health management of a bearing needs more attention than any other mechanical component. In this paper, a novel approach for signal processing using a hybrid of discrete wavelet transform (DWT) and empirical mode decomposition (EMD), for fault diagnosis of bearing has been proposed. In the first step, DWT is used to split the raw vibration signal into certain frequency sub-bands, and in the next step EMD is used to decompose the selected frequency band into a number of intrinsic mode functions (IMFs) and a residue. Afterwards, feature ranking and selection was performed using One-way ANOVA and Kruskal Wallis in MATLAB. Then for classification unit, support vector machine (SVM) and artificial neural network (ANN), techniques have been adopted. The results using the proposed strategy were quite satisfying. This concept can be used in future to inspect and predict a range of different faults including gear and rotor.