反向
反问题
物理
计算机科学
电子工程
微带天线
反射(计算机编程)
拓扑优化
带宽(计算)
声学
光学
工程类
逆散射问题
拓扑(电路)
数学
材料科学
优化设计
控制理论(社会学)
数学分析
作者
Yulei Yang,Yihong Su,Mingtang Li,Yiming Fan,Shi-Wei Dong,Xian Qi Lin,Bingjie Zhang,Yong Fan,Yu Jian Cheng
标识
DOI:10.1109/lawp.2026.3705680
摘要
Traditional pixelated element design relies on full-wave simulation–driven optimization, which incurs prohibitive computational cost. This letter presents an efficient inverse design method based on a Deep neural network surrogate model to accelerate the optimization process. By incorporating physical constraints, a high-quality dataset is constructed to improve prediction accuracy. To further enhance performance and ensure smooth topological transitions, a genetic algorithm is combined with a graph-theory-based shortest-path search. Based on this framework, a dual-circularly polarized (dual-CP) pixelated reflectarray is designed, fabricated, and measured. Experimental results show peak gains of 26.46 dBi (LHCP) and 26.54 dBi (RHCP). Over a 10.1% relative bandwidth, the antenna maintains an axial ratio below 2 dB and a gain variation within 2 dB.
科研通智能强力驱动
Strongly Powered by AbleSci AI