Fault Classification of IC Engine Using Wavelet Energy Features and Geometric Mean Neuron Model

Abstract

The paper proposes a novel approach for fault classification in an Internal Combustion (IC) engine using wavelet energy features and geometric mean neuron model based neural networks. Live signals from the engine were collected with and without faults by using four industrial microphones. The acoustic signals measured for faulty engines were decomposed using wavelet transform. The energy of each decomposed signal was computed and used as a feature vector for further classification using GMN based neural networks.