Font Size: a A A

Study On Learning Algorithms Of FNN And Application In Susceptibility Assessment Of Electronic Systems

Posted on:2005-07-10Degree:MasterType:Thesis
Country:ChinaCandidate:F HanFull Text:PDF
GTID:2168360122980323Subject:Circuits and Systems
Abstract/Summary:PDF Full Text Request
It is a difficult problem to assess high power microwave (HPM) susceptibility ofelectronic devices or electronic systems. HPM susceptibility is difficult to assess because itis stochastic in nature. There are many parameters which affect the result of HPMinteraction with targets. From the point of view of math, HPM susceptibility evaluating is anonlinear problem. FNN is a powerful tool which is used to deal with nonlinear problems. In this paper, wediscuss the drawback of learning algorithm of FNN based on T-S model. A learningalgorithm is proposed to deal with the defects. Finally, the method of using fuzzy sets theoryand FNN to evaluate HPM susceptibility of electronic devices are proposed. The thesis is organized as follows: In chapter 1, a brief review of the development of fuzzy neural networks (FNN) ispresented. The possibility and necessity of applying of FNN to evaluate HPM susceptibilityof electronic systems is discussed. In chapter 2, the basic theory, the structure and learning algorithms of FNN based onT-S model are introduced. The drawback of learning algorithms is discussed. In chapter 3, a new method to minimize a quadratic function is proposed. The methodis proved and compared with conjugate gradient (CG) method. It can be used in learningalgorithm of FNN based on T-S model and the robust of the learning algorithm is improved. In chapter 4, a hybrid training method of FNN based on T-S model is proposed. Usingpartial least square (PLS) regression in the train method, the constraint of original methodare break. In chapter 5, the fuzzy neural network is applied to evaluate the failure thresholds ofelectronic devices as a function of the parameters of HPM. A method is presented toevaluate the possibility distribution of electronic device failure by applying the possibilitytheory. Combining the possibility theory and prediction ability of the fuzzy neural network,the possibility distribution of electronic device failure is provided. In chapter 6, the conclusion of the thesis and the work which we prepare to do arediscussed.
Keywords/Search Tags:fuzzy neural network, learning algorithms, partial least square regression, susceptibility assessment
PDF Full Text Request
Related items