Study On Novel Transmission Line Equation And Knowledge-Based Neural Network Model

On the basis of classical transmission line theory, this dissertation presents a generalized transmission line equation (GTLE) to overcome the difficulties from practical application of classical transmission line equation. From theoretical derivation to practical application of the generalized transmission line equation, from 1D-GTLE to 3D-GTLE, from broadband characteristic of GTLE to equivalent characteristic between partial GTLE and whole GTLE, from NNKBN model to BKBNN model, the dissertation systematically investigates application of the generalized transmission line equation for electromagnetic problem, and extends potential of neural network model for application in EM CAD. The theoretical system is built for broadband knowledge-based neural network model (BKBNN) in electromagnetic problem.At first, the basic content of new transmission line equation is introduced. The theoretical derivation for new equation is not only introduced from Maxwell basic theory but also from basic microwave circuit theory. Through comparison of the radiation parameter extraction between right-angle bend and 45 ° mitered bend, it necessitates to reconstruct a new transmission line equation.Secondly, for the analysis of several electromagnetic problems, such as a five section microstrip filter, microstrip cross, microstrip via, the theoretical derivation from 1D-GTLE to 3D-GTLE is presented in this dissertation. The extracted parameters have two important characteristics. One is that all the parameters are independent of the line loads while the conventional equations are not like this. Another is that parameters L and C are very slowly changed with frequency. Thus, the GTLE will have good broadband characteristics from above two characteristics.Thirdly, for the analysis of a double step discontinuity and double folded stub microstrip structure, the partial GTLE can model the whole structure as well as the whole GTLE. Since the MoM solutions for the whole structure was used to establish partial GTLE, the partial GTLE has contained all information for the whole structure.At last, on the basis of basic neural network model, the quasi-static analytic formulae for simple electromagnetic structures (such as stripline crossover discontinuity and microstrip T-junction) are regarded as prior knowledge, and a novel three-layer neural work with knowledge-based neurons in hidden layer (NNKBN) is presented in this dissertation. Furthermore, the extracted parameters by using GTLE are regarded as prior knowledge in order to construct broadband knowledge-based neural network model (BKBNN) for microstrip T-junction and coplanar waveguide to microstrip (CPW-M) transition structure.