Independent Component Analysis And Its Application To Vehicle Condition Monitoring

Vehicle condition monitoring and fault diagnosis play important role in Intelligent Vehicle System, and signal processing technique is the crucial link of vehicle condition monitoring and fault diagnosis, especially the research on the rotary parts in vehicles has been focused at home and abroad. Independent component analysis is a recently developed method, which focuses on extracting the higher-order statistical information from the data so that the components are not only statistically uncorrelated, but also as independent as possible. It can reveal the inherent characteristics of signals. This dissertation explores the independent component analysis and its applications to feature extraction, as well as pattern classification from incipient fault of vehicle rotary parts.In chapter 1, the significances of vehicle condition monitoring and the content of the research were pointed out.The basic theory of ICA is introduced in chapter 2. First, accounting the definition and fundamental properties of statistical independence, then establishing the ICA model. Second, focusing on the contrast function of measures of nongaussianity, such as kurtosis, negentropy, approximations of negentropy, mutual information.In chapter 3, on the foundation of research with FastICA algorithm, this dissertation mainly introduced and developed the extraction of the higher-order statistical information of one-dimensional vibration signal, which has excellent potential applications since it reveals the inherent characteristics of vibration signals.ICA convolutive model is studied in Chapater 4, reviews the general concept of ICA, especially the theory for convolutive mixtures, the model of convolutive mixture and two deconvolution structures: Recursive and Direct structures, then presents a ICA algorithm for convolutive mixtures based on RCTE threshold control criteria.In chapter5, the design process of on-line monitoring platform based on vibration measurement is introduced, such as the selection and fixation of vibration sensors, the design and realization of hardware and software.Chapter 6 gives the conclusions and the prospect about this study.