Entropy based feature extraction for motorbike engine faults diagnosing using neural network and wavelet transform

Abstract

The sound of working vehicle provides an important clue for engine faults diagnosis. Endless efforts have been put into the research of fault diagnosis based on sound. It offers concrete economic benefits, which can lead to high system reliability and save maintenance cost. A number of diagnostic systems for vehicle repair have been developing in recent years. Artificial Neural Network is a very demanding application and popularly implemented in many industries including condition monitoring via fault diagnosis. This paper presents a feature extraction algorithm using total entropy of 5 level decomposition of wavelet transform. The engine noise signal is decomposed into 5 levels (A5, D5, A4, D4, A3, D3, A2, D2, A1, D1) using Daubechies "db4" wavelet family. From the decomposed signals, the entropy is applied for each levels and the feature are extracted and used to develop a backpropagation neural network.