Fault diagnosis of an automotive air-conditioner blower using noise emission signal

Abstract

This paper presents a neural network system for automotive air-conditioner blower fault diagnosis using noise emission signals. The proposed system consists of three parts: data acquisition, feature extraction, and fault classification. First, the noise emission signals are obtained from a condenser microphone and recorded by a data acquisition system. The signals are split into several wavelet nodes without losing their original properties by wavelet packet decomposition (WPD) by entropy criterion. Meanwhile, the energy values are calculated from these nodes for feature extraction. Finally, the energy features are used as inputs to neural network classifiers for identifying the various fault conditions. The WPD integrated with energy features is an efficient method for feature extraction. The energy features are efficient in reducing the dimensions of feature vectors and in the time consumed for training and classifying. In the experimental work, the probabilistic neural network (PNN) is used to verify the performance and compared with the conventional back-propagation neural network (BPNN) technique. The experimental results demonstrated the proposed technique can achieve powerful capacity for estimating faulty conditions quickly and accurately.