Calculated Based On An Integrated Intelligent Analog Circuit Fault Diagnosis Method

With the rapid development of the electronic industry, the importance of the fault diagnosis of analog circuits is becoming more and more obvious and the research on theories and methods of the fault diagnosis are very popular. Generally speaking, fault diagnosis has pronounce signification for smooth working and trustable design of electronic devices or systems. In the past 40 years, the theories and approaches of the fault diagnosis in the field of circuits have attracted a lot of attention, but most of them are restrictive because of the variety and complexity of the analog circuits, especially the large-scale analog circuits with rigorous tolerance. The theory of the artificial neural networks has got a rapid development in recent years and the so-called computational intelligence which is typically represented by neural networks provides a powerful way for diagnose faults of analog circuits.In this thesis, we analyze the difficulties caused by tolerance and nonlinearity of the analog circuits, and then we propose a method for fault diagnosis based on integrated calculation of intelligence. The main work of the thesis is focused on two aspects: (1) studying the feature extraction method of a fault circuit, and (2) studying the methods of determining the structure of neural networks and the methods of determining the parameters of a neural network. For the first aspect, we investigate the methods of the principal component and the Rough-Set. In particular, we apply the theory of Rough-Set to simplify the constraints and apply the principal component analysis to reduce the dimension of the eigenvectors.By using numerical simulations and practical experiments, we get the conclusion as that the diagnostic accuracy obtained by the neural networks trained by integrated calculation of intelligence is better than the one obtained by the neural networks without integrated calculation of intelligence.The main contributions of this thesis are given as follows:1 By using the Multisim 7 circuit simulation software, we apply the Monte-Carlo method to obtain sample data for fault diagnosis of an analog circuit.2 Note that, in the fault diagnosis of an analog circuit, the method of the feature extraction plays an important role: it dominants the efficiency of fault diagnosis and it provides theoretical base for fault diagnosis in practical situation. In this thesis, we apply the theory of Rough-Set and the principal component analysis for this purpose.