The Research On Method Of Fault Diagnosis For Nonlinear Networks Based On Volterra Series, Wavelet Analysis And Neural Networks

With the development of research on fault diagnosis of analog circuits in the past more than three decades, a series of diagnosis theory and methods has been formed. Because of the complication and difficulty of fault diagnosis for analog circuits, there are a lot of limitations and poor practicable in these theories and their applications. Especially for nonlinear analog circuit fault diagnosis as nonlinear elements exist in circuits, there are no generally proper math. model and common simulation program for nonlinear networks. So the unified method is lacking for computing fault character of nonlinear circuits. As appearance and rapid development of wavelet theory, while increasingly maturation of neural networks theory and its application, it has been becoming a hot studied project to apply wavelet analysis with neural networks to locate faults. The wavelet analysis is used to analyze and process fault signals at first, then neural networks is used to classify and locate faults. Many results of research show that it provides a new way for fault diagnosis of analog circuits.In the paper, it was principally studied for applying Volterra series theory to analyze and calculate fault response of nonlinear analog circuits, and for combining wavelet analyzing with neural networks classifying to locate faults. The main contents are as follows. Discrete recursive computation formulas of Volterra series solution of math. model for a large class of nonlinear dynamic networks were derived. The formulas provide a simple and fast numeral calculation method for computing fault response curves, and make simulation programming easy. A method of hybrid model identification for a large class of nonlinear dynamic networks was obtained by combining Volterra series with block-pulse function transform(BPFT). Based on the hybrid model, a fault diagnosis method of multi-preset models was proposed. We studied and proposed the method of fault diagnosis dictionary in frequency domain in which the core of frequency domain for Volterra series solution of nonlinear network model was taken as a fault feature. We derived a group of fault diagnosis equations for nonlinear resistance circuits, in which the quantity of fault mark was treated as unknown, and applied a kind of neural network to solve these equations to locate faults. HAAR wavelet features, such as time-frequency localization, compact support and zero smooth, and its decomposition and reconstruction were described. We used HAAR wavelet to abstract the feature of fault response of nonlinear analog circuits,...