Neural network models of slotted waveguide directional couplers

Muller, Martinette (2001-12)

Thesis (MScEng)--University of Stellenbosch, 2001.

Thesis

ENGLISH ABSTRACT: The application of artificial neural networks to microwave circuits is investigated. A neural network model is developed for two parallel waveguides coupled by a longitudinal slot in the common broad wall. Training data is generated through a moment method solution of the integral equations that describe the structure. A systematic investigation of training options is carried out and the development of the model is described in detail. The model is evaluated and compared with an Adaptive Sampling Interpolation (ASI) Technique. The neural network is found to be less accurate than the ASI Technique at a much greater expense of development time and required user supervision.

AFRIKAANSE OPSOMMING: Die toepassing van neurale netwerke op mikrogolfbane is ondersoek. In Neurale netwerk-model is ontwikkel vir twee parallelle golfleiers met longitudinale gleufkoppeling in die gemeenskaplike bree wand. Data vir die opleiding van die netwerke is verkry deur In momentmetode-oplossing van die integraalvergelykings wat die struktuur beskryf. Verskillende ontwerpsopsies vir die netwerke is stelselmatig ondersoek en die ontwikkelingsproses van die netwerk is volledig beskryf. Die model is geevalueer en vergelyk met In Aanpasbare Monstering Interpolasietegniek (AMI). Daar is gevind dat die neurale netwerk minder akkuraat is as die AMI terwyl die koste aan ontwikkelingstyd en gebruikerstoesig hoer is.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/52354
This item appears in the following collections: