基于多人工神经网络的超材料微带天线智能设计  

Intelligent Design of Metamaterial Microstrip Antenna Based on Multiple Artificial Neural Networks

作　　者：王永刚 董焱章 Wang Yonggang;Dong Yanzhang(School of Automobile Engineering,Hubei University of Automotive Technology,Shiyan 442002,China;Hubei Key Laboratory of Automotive Power Train and Electronic Control,Shiyan 442002,China)

机构地区：[1]湖北汽车工业学院汽车工程学院,湖北十堰442002 [2]汽车动力传动与电子控制湖北省重点实验室,湖北十堰442002

出　　处：《湖北汽车工业学院学报》2022年第2期45-48,54,共5页Journal of Hubei University Of Automotive Technology

基　　金：国家自然科学基金青年科学基金(11502075);汽车零部件技术湖北省协同创新项目(2015XTZX0401);湖北汽车工业学院博士科研启动基金(BK201501)。

摘　　要：依据超材料微带天线的增益、E面半功率波瓣宽度和H面半功率波瓣宽度建立3个人工神经网络,得到卦线长度、断口间距、卦线宽度、卦线间距和覆层高度,建立统一的多人工神经网络。对超材料微带天线进行智能设计,并对核心尺寸设计结果进行结构重分析。结果表明:基于多网络的训练集、验证集和测试集均方误差均小于1‰,相比于常规的单人工神经网络,其预测精度有了明显提升;设计的3个天线性能指标的绝对误差达标。Three artificial neural networks were established based on the gain, the half-power beamwidth of the E-plane and the half-power beamwidth of the H-plane from the metamaterial microstrip antenna. The Bagua line length, fracture length of the broken line, Bagua line width, Bagua line distance and cover height were obtained, and an unified artificial neural network was established. The metamaterial microstrip antenna was intelligently designed, and the core size design results were reanalyzed. The test results show that the mean square errors on the training set, validation set and test set based on multiple networks are all below 1‰. Compared with the conventional single artificial neural network, its prediction accuracy is significantly improved. The absolute errors of the three designed antenna performance indicators all meet the standards.

关 键 词：超材料 微带天线 人工神经网络 增益 半功率波瓣宽度 

分 类 号：TN822[电子电信—信息与通信工程]