基于BiLSTM代理模型的RFID标签天线快速多目标设计

Fast Multi-objective RFID Tag Antenna Design Based on BiLSTM Surrogate Model

为快速完成超高频(UHF)标签天线多目标设计,构建了基于双向长短期记忆(BiLSTM)神经网络的新型天线性能预测代理模型,并使用带精英策略的非支配排序遗传算法(NSGA-II)配合静态惩罚对RFID标签天线进行多目标寻优。首先按照RFID标签天线延伸的趋势将天线结构序列化并依次输入BiLSTM网络结构,结合响应集训练出适用于RFID标签天线的代理模型。进一步地,为最大程度搜索可行域内,尤其是可行域边界的所有标签天线形式,在NSGA-II算法中引入静态惩罚,最终完成RFID标签天线低回波损耗、小型化、宽频带的多目标快速设计。设计实例表明,该方法在超高频RFID标签天线领域的预测精度、计算代价等方面综合表现优于现有天线设计方法,具有适用性和实用价值。

In order to quickly complete the multi-objective design of ultrahigh frequency(UHF)tag antenna,a new antenna performance prediction surrogate model based on BiLSTM neural network is constructed in this paper,and the non-dominated sorting genetic algorithm with elite strategy(NSGA-II)is used to optimize the RFID tag antenna under static punishment.Firstly,the antenna structure is serialized and input into BiLSTM network in accordance with the trend of RFID tag antenna extension,and the surrogate model suitable for RFID tag antenna is developed in combination with response training.Further,in order to search all tag antenna forms within the feasible domain,especially the boundary of the feasible domain to the maximum extent,static penalty is introduced in NSGA-II algorithm to finally complete the rapid multi-target design of RFID tag antenna with low return loss,minification and broadband.The design example shows that the method is superior to the existing antenna design methods in terms of prediction accuracy and calculation cost,it has applicability and practical value.