基于KNN-XGBOOST堆叠模型在PCB RFID天线阻抗预测的研究  

KNN-XGBOOST-based stacked model for PCB RFID antenna impedance prediction

作　　者：姜延坤 洪涛[2] 章吉丽 JIANG Yankun;HONG Tao;ZHANG Jili(College of Energy Environment&Safety Engineering,China Jiliang University,Hangzhou 310018,China;College of Quality and Standardization,China Jiliang University,Hangzhou 310018,China)

机构地区：[1]中国计量大学能源环境与安全工程学院,浙江杭州310018 [2]中国计量大学质量与标准化学院,浙江杭州310018

出　　处：《现代电子技术》2024年第19期14-20,共7页Modern Electronics Technique

基　　金：浙江省基础公益研究计划项目(LGG22E050011)。

摘　　要：针对传统的天线仿真建模过程中需要的天线阻抗耗时长问题,文中提出一种基于KNN-XGBOOST模型的天线阻抗预测方法。现有研究大多为单一预测算法,旨在通过对比寻求预测效果更优的算法。首先通过ANSYS仿真软件收集大量的PCB RFID天线阻抗设计数据,然后结合影响阻抗中天线长度和频率共8个有效特征,以KNN和XGBOOST两种算法作为基模型,线性回归作为元模型,构建了一个堆叠集成学习模型。在实验过程中,通过交叉验证和网格搜索技术,对模型的超参数进行了精细调优,以确保模型能够达到最优的预测性能。实验结果显示,与单一的KNN和XGBOOST模型相比,KNN-XGBOOST模型的均方根误差降低了30%~70%,R^(2)提高了10%。在预测PCB RFID天线的阻抗实部和虚部时,KNNXGBOOST模型具有较高的准确率和较低的预测误差,证明了其在电磁仿真设计优化中的应用价值。In view of the time-consuming antenna impedance required in the traditional processes of antenna simulation and modeling,an antenna impedance prediction method on the basis of KNN-XGBOOST-based model is proposed.Most of the existing studies focus on the single prediction algorithms.These studies aim to find out the algorithms with better prediction effect by comparative analysis.Initially,a substantial data of PCB(printed circuit board)RFID(radio frequency identification)antenna impedance design is collected with simulation software ANSYS.Subsequently,leveraging eight influential features including antenna length and frequency,a stacked ensemble learning model is constructed on the basis of constructing the base model with algorithms KNN(K-nearest neighbor)and XGBOOST(eXtreme Gradient Boosting),and the meta-model with linear regression.In the experiment,fine-tuning of the model's hyperparameters is implemented via cross-validation and grid search techniques,so as to ensure that the model can reach the optimal predictive performance.The experimental results demonstrate that the root mean square error(RMSE)of the KNN-XGBOOST-based model is reduced by 30%~70%,and its R-squared(R^(2))is increased by 10% in comparison with those of the KNN-based model and XGBOOST-based model.When predicting the real and imaginary parts of PCB RFID antenna impedance,the KNN-XGBOOST-based model exhibits higher accuracy rate and lower prediction error,which verifies its application value in the optimization of electromagnetic simulation design.

关 键 词：PCB RFID天线 阻抗预测 KNN算法 XGBOOST算法 融合堆叠 电磁仿真 

分 类 号：TN821-34[电子电信—信息与通信工程]