MOEA/D-based Optimization Techniques And Their Applications In Antenna Design

Antenna optimization is an important research direction in electromagnetic field. In practical engineering, most antenna structure optimization designs could be attributed to multiobjective optimization problems (MOPs). With the development of computational intelligence, MOEAs (multiobjective evolutionary algorithms) have their successful applications in many fields owing to their parallelism and the ability of global optimization. Whereas, there’s few work on antenna structure optimization by using MOEAs, especially on topological structure optimization. MOEA/D (MOEA based on decomposition) has the most popularity in computational intelligence society. In this dissertation, MOEA/D-based optimization techniques will be investigated and some of them will be utilized to design antennas. Furthermore, powerful scientific research value and practical application potential of MOEA/D in antenna structure optimization design will be shown.The research contents consist of the following five points:1. study basic algorithm theory of MOEA/D, and then connect it with electro-magnetic simulator or numerical code to construct MOEA/D-based antenna multi-objective optimization framework;2. propose MOEA/D-GO (MOEA/D combined with enhanced genetic operators) whose high-efficiency is verified by numerical instances, and then employ MOEA/D-GO to optimize Coplanar Waveguide feed (CPW-fed) band-notched microstrip antenna, fragment-type UHF RFID reader wire antenna with perfect impedance matching, strong and uniform magnetic distribution, and given far-field main-beam, and high isolation MIMO (multiple-input multiple-output) PIFAs (planar inverted-F antennas) with non-uniform fragment cells, the high-efficiency of MOEA/D-GO for designing fragment-type structures is verified through simulated and measured results;3. MOEA/D-GO-II (MOEA/D-GO combined with2-D median filtering) is proposed in order to reduce overload computational cost when MOEA/D-GO is used to optimize fragment-type structure, then it’s utilized to design high isolation MIMO PIFAs with uniform fragment cells, and finally comparison among several designs shows its high-efficiency when fragment-type microstrip structure with low loss is appreciated; 4. modified MOEA/D-DE (MOEA/D combined with differential evolution) is proposed to relieve the bad population diversity when small neighborhood size is appreciated to speed up optimization, and then its high-efficiency is verified by numerical instances, finally, it’s used to design multiband bow-tie microstrip antenna with over-lapping arms, and its high-efficiency is further verified by this antenna design example, meanwhile, the efficiency of modified MOEA/D-DE for designing multi-band antennas is verified through simulated and measured results;5. SL (statistic location information) operator is proposed to form MOEA/D-SL which gives novel idea for solving hybrid optimization problems; IOO (inverse onion operator) is proposed to form MOEA/D-IOO for solving those antenna design problems where loop structure is appreciated; then numerical tests of both MOEA/D-SL and MOEA/D-IOO are preliminarily implemented; and finally the population diversity of MOEA/D-SL is improved through modifying SL.In this thesis, MOEA/D-GO, MOEA/D-GO-II, and MOEA/D-IOO focus on antenna topology optimization, modified MOEA/D-DE focus on antenna size optimization, while MOEA/SL is the extension of all of them. It should be noticed that both MOEA/D-SL and MOEA/D-IOO have powerful algorithm theory value, and their algorithm principle remains to be further studied. Therefore, their applications on antenna design are not implemented directly in this dissertation. Through these researches on MOEA/D-based multiobjective optimization techniques, not only algorithm theory of MOEA/D is developed, but high efficiency design methods in antenna society are proposed.