Skip to main navigation Skip to search Skip to main content

Genetic algorithm applications in EM problems

  • Yuan YUAN

Student thesis: Doctoral Thesis

Abstract

An artificial Intelligent (AI) design platform has been introduced in this thesis to achieve complex electromagnetic (EM) designs, meeting the great demands of emerging wireless communication technologies. The design platform runs on the Beowulf pc cluster as parallel design environment following “get what you ask” procedure based on the genetic algorithm optimization. For a given design, only the design specifications and initial conditions are required and the reverse design procedure of getting optimal parameters will be preceded automatically. Compared to the conventional “trial and error” design procedure, it is more efficient and reliable since no tedious repeatable work is needed and defects of un-convergence failure or trap of local optimum can be avoided. To provide more flexibility and versatility of its applications, accurate circuit and full-wave electromagnetic simulator cores are incorporated and the AI design platform can attain the optimal designs of integrated RF/microwave applications fulfilling strict design specifications as much as possible. In this evolution optimization design procedure, details are recorded in the constructed database for tracing optimal design and further understanding of the design. Those experience accumulated by the design platform can inspire more creative ideas for the future applications. What’s more, the AI design platform can provide optimal solutions for the system modeling problems by hierarchical GA of simultaneously systemic and parametric optimization. There is no prior information needed in this approach, so it can exploit and explore more system modeling possibility and achieve better solutions. The successes of the implementation of the AI design platform are verified by several excellent EM applications in this thesis. Broadband unequal power divider has optimal profile with novel and simple solution of high ratio unequal divider. Frequency selective surface is designed by optimally selecting unit-cell pattern and layers stacking manner. It is used for spatial filters and metal-material surface as perfect magnetic conductor. Ultra wide band (UWB) antenna has been used in strict performance in UWB operating frequency band. Besides fulfilling those specifications, the proposed novel strip UWB antenna almost has fixed phase center and is a traveling wave antenna. Antenna array design is a comprehensive procedure and specifically difficult. The profile and architecture of the proposed antenna array is original and optimally designed to have excellent performance satisfying with the strict design specifications of the WLAN communication technologies. Achievements of those challenging applications are the contributions to their application fields.
Date of Award4 Oct 2004
Original languageEnglish
Awarding Institution
  • City University of Hong Kong
SupervisorChi Hou CHAN (Supervisor)

Keywords

  • Genetic programming (Computer science)
  • Design and construction
  • Genetic algorithms
  • Artificial intelligence
  • Electromagnetic devices

Cite this

'