摘要
天线优化设计中,由于经典优化算法一般需要对同一结构的成百上千组参数进行电磁仿真后才能得出最优结果,因此多参数、大范围的天线优化设计存在计算效率问题。该文利用Kriging模型拟合参数空间内样本点的电磁仿真数据,代替电磁仿真实现从结构参数到电磁响应的瞬时近似计算,通过初始阶段的多维均匀拉丁超立方采样算法(LHS-MDU)和优化循环中的新增采样策略,减少电磁仿真次数,提高优化设计效率。利用此方法调整矩形贴片天线的馈点位置和双频单极子天线的振子长度来优化工作频点与阻抗带宽,相比遗传算法,完成相同目标所需的电磁仿真次数分别减少了75%和84%。
During the process of antenna design and optimization, classical optimization methods often require hundreds or even thousands trials of different parameter combinations, which leads to a low efficiency in solving multi-parameter and large scale optimization problems. In this paper, a quick and approximate computation of the EM response can be realized though a Kriging model, which is created by fitting the simulation results to their structural parameters. The number of EM simulation needed can be reduced by Latin Hypercube Sampling for MultiDimensional Uniformity （LHS-MDU） at the initial stage and a candidate-selecting method in the following optimization loops. In order to optimize the resonant frequency and impedance bandwidth, the feed position of a rectangular patch antenna and the dipole lengths of a dual-band monopole antenna are adjusted by the proposed method and compared with the genetic optimization, the numbers of EM simulation are reduced by 75% and 84% respectively.
出处
《电子与信息学报》
EI
CSCD
北大核心
2014年第12期3021-3026,共6页
Journal of Electronics & Information Technology
关键词
天线优化
代理模型
KRIGING模型
多维均匀采样
Antenna optimization
Surrogate model
Kriging model
Multidimensional uniform sampling
