Automated Design and Optimization of Wire Antennas Using Genetic Algorithms.
AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OH
Pagination or Media Count:
A Genetic Algorithm GA has been used in conjunction with the Numerical Electromagnetics Code, Version 2 NEC2 to create and optimize atypical wire antenna designs with impressive characteristics. Antenna design parameters are encoded into an ordered series of numbers andor symbols analogous to a biological chromosome. A cost function that quantifies how well a design meets the engineers specifications is created. The GA uses these to generate and evaluate a population of designs. The most successful designs are then promoted and mixed through mating and mutation, while poor designs are removed. This process, difficult to trap in local minima, continues until convergence criteria are met, generally yielding excellent designs with no user intervention or initial guesses. Three antennas have been optimized a monopole loaded with a modified folded dipole, the Yagi antenna, and the crooked-wire genetic antenna. Prior study of the loaded monopole had shown hemispherical coverage was possible. The GA found an asymmetric loaded monopole with an average variation in gain over the hemisphere of only 0.4dB, confirmed by measurement. GA-optimized Yagi antennas surpassed the gain of conventional Yagis by about 1dB, improvement also confirmed by measurement. The GA designed a Yagi with a beamwidth of 50 deg -60 deg, sidelobes nearly 25dB down, and a 14 bandwidth-specifications difficult to achieve using conventional techniques. The crooked-wire genetic antenna is several wires joined in series locations and lengths are determined by the GA. Optimization for hemispherical coverage with right-hand circular polarization RHCP produced highly unusual shapes unrealizable using a conventional approach. RHCP hemispherical coverage was achieved with less than 4dB variation. Measurements verify the results.
- Electrical and Electronic Equipment
- Computer Programming and Software