Use this URL to cite or link to this record in EThOS:
Title: The application of artificial intelligence to optimise electromagnetic structures
Author: Macpherson, Russell Iain
Awarding Body: University of Aberdeen
Current Institution: University of Aberdeen
Date of Award: 2004
Availability of Full Text:
Access through EThOS:
This thesis describes the use of artificial intelligence (AI) techniques applied to the problem of optimising requirements relating to specific electromagnetic structures. A micro-genetic algorithm (μGA) was coupled with the finite-difference time-domain (FDTD) method to optimise the placement of radiating and susceptive components within a shielded enclosure. This hybrid code was then used to optimise radio frequency identification (RFID) tags. Also presented here are the attempts to reduce the computational burden of the electromagnetics solver using artificial neural networks (ANNs) and fuzzy logic (FL) systems through the use of time-series prediction. The hybrid code was developed and verified in both two and three dimensions. It was then used to optimise the placement of a susceptor and multiple sources in order to minimise electromagnetic coupling between them. This optimisation was carried out for both single and multiple objectives. Testing of the optimised systems in the laboratory confirmed the validity of the pre­dicted pattern of results although the absolute magnitudes differed. Having confirmed the usefulness of the tool the optimisation of RFID tags was undertaken. Two performance metrics were used and these resulted in different tag geometries and material properties that maximally coupled electromagnetic energy from a source within a pipe full of water. During the course of this work the quality of the solutions from the μGA optimisation process was found to depend on the quality of the pseudorandom number generator (PRNG) employed by the μGA. The PRNG that produced the best results during many test simulations was carried forward and used in all subsequent optimisations. To reduce the computational burden of the FDTD solver time-series prediction using ANNs and FL was attempted. Both of these methods were unsuccessful in recursively predicting future values of a FDTD time-series. They were both, however, successful in non-recursively predicting into the future, this is unfortunately of limited, if any, use in this application. This thesis has presented an efficient, intelligent method for the placement of sources and susceptors within a shielded enclosure. The design tool produced has also been used to design RFID tags.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available