基于GA-DBP算法的UWB天线优化建模
Optimization Modeling of UWB Antennas Based on GA-DBP Algorithm
李昊 1王昕2
作者信息
- 1. 辽宁工程技术大学电子与信息工程学院，葫芦岛，125000
- 2. 辽宁工程技术大学电子与信息工程学院，葫芦岛，125000
- 折叠
摘要
随着无线通信技术的快速发展，超宽带（UWB）天线在多频带适应性、体积小、低成本等方面的优势使其成为研究热点。然而，传统BP神经网络在天线参数建模中容易陷入局部最优，训练稳定性差，且对复杂非线性关系的拟合能力有限。本文提出了一种基于遗传算法优化的双隐藏层BP神经网络（GA-DBP），利用遗传算法的全局搜索能力自适应优化网络权重与偏置，并通过双隐藏层结构提升函数映射能力。以HFSS生成的UWB天线样本数据为基础，构建GA-DBP网络模型并与BP和GA-BP网络进行对比实验。结果表明，GA-DBP神经网络在迭代收敛速度、MAE值及S11参数拟合精度上均优于传统方法，实现了更高的预测精度和更强的泛化能力。该方法为基于神经网络的UWB天线优化建模提供了有效的技术途径。
Abstract
With the rapid development of wireless communication technology, ultra-wideband (UWB) antennas have attracted significant research attention due to their advantages in multi-band adaptability, compact size, and low cost. However, conventional BP neural networks tend to fall into local optima, exhibit poor training stability, and have limited capability in fitting complex nonlinear relationships when used for antenna parameter modeling. This paper proposes a double-hidden-layer BP neural network optimized by a genetic algorithm (GA-DBP), which leverages the global search capability of the genetic algorithm to adaptively optimize network weights and biases, and enhances the function mapping ability through a double-hidden-layer structure. Based on UWB antenna sample data generated by HFSS, the GA-DBP network model is constructed and compared with traditional BP and GA-BP networks through experimental validation. The results indicate that the GA-DBP neural network outperforms conventional methods in terms of convergence speed, mean absolute error (MAE), and S11 parameter fitting accuracy, achieving higher prediction accuracy and stronger generalization ability. This method provides an effective technical approach for neural network-based UWB antenna optimization modeling.关键词
超宽带天线（UWB）/BP神经网络/遗传算法（GA）/双隐藏层/天线优化建模Key words
Ultra-Wideband Antenna/Backpropagation Neural Network/Genetic Algorithm/Double Hidden Layer/Antenna Optimization Modeling引用本文复制引用
李昊,王昕.基于GA-DBP算法的UWB天线优化建模[EB/OL].(2026-06-08)[2026-06-13].http://www.paper.edu.cn/releasepaper/content/202606-38.学科分类
无线通信/电子电路/计算技术、计算机技术
