基于神经网络的非线性电路故障诊断的研究-控制工程专业论文.docxVIP

硕士学位论文Abstract 硕士学位论文 Abstract Fault diagnosis of non-linear circuit iS a comprehensive and crisscross discipline of forword developing．It is also all application subject based on new theories，new technologies and new methods．Due to more and more studies of scientists on this field，the fault chagnosis for nonlinear resistive circuits and nonlinear dynamic circuits is also deeply discussed and studied in the paper． First of all，according to the basic principle of circuit fault diagnosis and neural network theory,the fault location methods of nonlinear circuits via BPNN are proposed and examples for diagnosis are also presented． Secondly,to acquire multi-value solutions of nonlinear resistive networks，the method of canonical piece wise·linearization is introduced with emphasis．These solutions are taken as sample set of BPNN to identify and classify the faults．In order to simplify the structure of BPNN structure，decrease the training time and improve the accuracy of diagnosis，the wavelet analysis iS applied as preprocessor ofBPNN． Finally,for a large class of nonlinear dynamic networks，which are expressed by uniform math model，the Volterra series solution of the model in time domain was transformed to numerical recursive formnlas in block—pulse domain by applying BPF(Block—Pulse Function)．The response of nonlinear dynamic circuit can be calculated precisely by computer simulation as SBT(Simulation before Test)．It provides foundation not only for software analysis of computer,but for a new way ofnumcrically computing for a class ofnonlinear dynamic networks． Keywords：Non-linear circuits，fault diagnosis，Back Propagation Neural Networks 03PNN)，canonical piece-wise linemSzation，wavelet transform． V 硕士学位论文 硕士学位论文 原创性声明 原创性声明 本人声明，所呈交的学位论文是本人在导师指导下进行的研究 工作及取得的研究成果。尽我所知，除了论文中特别加以标注和致谢 的地方外，论文中不包含其他人已经发表或撰写过的研究成果，也不 包含为获得中南大学或其他单位的学位或证书而使用过的材料。与我 共同工作的同志对本研究所作的贡献均已在在论文中作了明确的说 明。 作者签名：驻 日期： 型年卫月丛日 关于学位论文使用授权说明 本人了解中南大学有关保留、使用学位论文的规定，即：学校 有权保留学位论文，允许学位论文被查阅和借阅；学校可以公布学位 论文