【作者】 吴铮；

【导师】 顾长青；

【作者基本信息】 南京航空航天大学 ， 电磁场与微波技术， 2008， 硕士

【摘要】 体积小、重量轻、外观优美、结构紧凑的微带天线在移动通信等领域得到了广泛应用,如何实现其小型化、宽带化、多频和多功能是当今的研究热点。另外, MEMS开关加载可重构天线因其可以用一个物理口径实现多个天线的功能而越来越受到人们的关注。如何对这些天线进行优化设计是本文所想探索的。近年来,利用遗传算法对天线进行优化设计一直是研究的热点。做为一种通用的搜索方法,遗传算法在进行天线优化时,不受初始值限制,程序通用性强且能够取得较好的区域搜索和空间扩展的平衡。在使用遗传算法进行天线优化时,需要通过解析方法或者全波分析构造所需要的适应度函数,但是前者对于复杂结构天线难以实现,后者耗时较长。针对上述问题,本文引入神经网络,通过对有限个样本的全波分析结果进行训练从而建立天线结构参数和性能参数之间的非线性映射关系。一旦神经网络模型建成,天线的性能参数值就可以通过此映射关系直接得出,相比于全波分析所耗时间几乎可以忽略。本文对于神经网络和遗传算法的结合进行了研究,并将其用于天线结构的优化设计。对于给定天线原型结构和设计要求,首先利用部分正交实验法选出部分天线结构参数样本,用高频仿真软件HFSS进行仿真并将其结果输出,然后运用BP神经网络对这些样本进行训练,建立结构参数和天线性能之间的映射。最后运用遗传算法,根据天线设计要求定出相应的适应函数表达式,通过一代代的遗传进化最终得到所需要的天线的结构参数。文中对有寄生单元的倒F天线,带探针且上部开槽的三角形贴片天线等进行了优化设计,效果较好。对于开关加载分形可重构天线进行了尝试,为以后课题的深入研究进行积累和铺垫。更多还原

【Abstract】 Microstrip antennas have several advantages for mobile communications in size,weight and compact structure. How to make them compact and broadband is a valuable focus. Besides, fractal antennas and antennas loading MEMS (Micro-Electro-MechanicalSystems ) switches have attracted more and more attentions.Today, optimizing the design of antennas using Genetic Algorithms(GA) is a focus.As a common search method,GA can be used easily in the antenna design and get good effect without the boundary in initialization values.When using GA,we get the fitness values by some mathematics methods or simulations.But the former method is too complicate to realize when the antenna is intricate in structure ,and the latter would consuming a lot of time.To solving this problem,we bring forward the using of neural network (NN) to establish a good mapping relation between the structure and the performance,in this way ,we can save much time.In this paper,carried on a research on the integrating of GA and NN and using them in the optimization of antennas .First we picked some samples and calculate the performance parameter with High Frequency Structure Simulator (HFSS) and output the results.Then establish the NN modal using the former results to map the relation between the structure and the performance .After that,we use the GA to search the one satisfying our requirement ,adopting different fitness functions.By utilizing this set of optimization program, several antennas including PIFA with parasite elements, triangle shape microstip antenna and so on.Also try to explore in a fractal MEMS Reconfiguable antenna.更多还原