• 中国期刊全文数据库
  • 中国学术期刊综合评价数据库
  • 中国科技论文与引文数据库
  • 中国核心期刊(遴选)数据库
陈辉, 蔡晗, 王帅杰. 特征融合卷积神经网络的超表面建模方法[J]. 桂林电子科技大学学报, xxxx, x(x): 1-8. DOI: 10.3969/1673-808X.2022320
引用本文: 陈辉, 蔡晗, 王帅杰. 特征融合卷积神经网络的超表面建模方法[J]. 桂林电子科技大学学报, xxxx, x(x): 1-8. DOI: 10.3969/1673-808X.2022320
CHEN Hui, CAI Han, WANG Shuaijie. Metasurface modeling method based on feature fusion convolutional neural network[J]. Journal of Guilin University of Electronic Technology, xxxx, x(x): 1-8. DOI: 10.3969/1673-808X.2022320
Citation: CHEN Hui, CAI Han, WANG Shuaijie. Metasurface modeling method based on feature fusion convolutional neural network[J]. Journal of Guilin University of Electronic Technology, xxxx, x(x): 1-8. DOI: 10.3969/1673-808X.2022320

特征融合卷积神经网络的超表面建模方法

Metasurface modeling method based on feature fusion convolutional neural network

  • 摘要: 针对基于深度学习的超表面设计中输入输出维度不匹配导致的预测精度不高的问题,提出了一种基于数据融合的卷积神经网络超表面建模方法。利用软件联合仿真对超表面进行结构建模仿真,获得了超表面一维结构参数、二维图片及其对应的电磁响应组成的数据集;通过构建卷积神经网络来提取超表面二维图片特征,并将其与一维结构参数相结合,构成一个全新的特征矩阵,对此特征矩阵通过全连接层训练,前向网络预测超表面的S21参数,逆向网络预测超表面的结构参数;训练后的模型具有良好的预测能力,前向网络模型、逆向网络模型预测的均方误差分别为3.148×10−4、2.548×10−3;用遗传算法对网络模型进行优化,使用迁移学习对新的数据集加速训练。研究结果表明,经算法优化的网络避免了大量的建模操作和数值计算,且运算准确率高。

     

    Abstract: To address the issue of low prediction accuracy resulting from input and output dimension mismatch in deep learning-based metasurface design, a metasurface modeling method based on data fusion convolutional neural networks is proposed.. Firstly, the software co-simulation is utilized to model and simulate the structure of the superstrate, obtaining a dataset consisting of one-dimensional structural parameters, two-dimensional images of the superstrate, and corresponding electromagnetic responses. Then, a CNN is constructed to extract the features of the two-dimensional images of the superstrate, which are combined with the one-dimensional structural parameters to form a new feature matrix. This feature matrix is then trained with a fully connected layer to predict the S21 parameters of the superstrate with the forward network and to predict the structural parameters of the superstrate with the inverse network. The trained model has good prediction capabilities, with mean squared errors of 3.148×10−4 and 2.548×10−3 for the forward and inverse network models, respectively. Further optimization of the network model is achieved through genetic algorithms, and transfer learning is utilized to accelerate the training process with new datasets. The results show that the algorithm-optimized network effectively avoids a large number of modeling operations and numerical computations, while maintaining high computational accuracy.

     

/

返回文章
返回