| 摘要: |
| 复合材料的有效电磁参数获取对研究其电磁性质具有重要作用。然而，基于S系数的传输反射方法不适用于含金属衬底的复合材料目标，导致其有效电磁参数难以检索。为解决该问题，提出了一种新型的信息融合有效电磁参数检索方法。首先，通过主成分分析（Principal Component Analysis，PCA）对有效电磁参数样本的S系数进行关联性分析和全局特征提取，简化模型复杂度。其次，构建了多输出回归信息融合网络（Multi-output Regression Information Fusion Network，MRIFN）模型对电磁参数进行检索。该模型结合了双向门控循环单元（Bidirectional Gated Recurrent Unit，BiGRU）处理主成分信息，并通过卷积层纵向融合特征信息，从而提高了检索性能。最后，通过非磁性吸波材料数据集和蜂窝有效电磁参数数据集验证了该检索方法有效性。结果表明，该方法对介电常数实部和虚部的平均绝对百分比误差（Mean Absolute Percentage Error，MAPE）分别为0.21%和0.57%。同时考虑介电常数和磁导率的综合影响，该方法对介电常数实部和虚部的 MAPE 分别为2.34%和9.28%，对磁导率实部和虚部的 MAPE 分别为2.40%和6.40%，验证了其有效性和泛化能力。 |
| 关键词: 复合材料 电磁参数 有效电磁参数检索 信息融合 深度学习 |
| DOI：10.20079/j.issn.1001-893x.240613006 |
|
| 基金项目:国家自然科学基金资助项目（62301215）；湖北省自然科学基金项目（2022CFA007）；湖北工业大学绿色工业科技引领计划项目（2022020801020267） |
|
| Method for Searching Effective Electromagnetic Parameters of Composite Materials with a Metal Substrate |
| ZHANG Fan,b,CHEN Xin,WU Minghu |
| (a.Hubei Key Laboratory for High-efficiency Utilization of Solar Energy and Operation Control of Energy Storage System;b.Hubei Engineering Research Center for Safety Monitoring of New Energy and Power Grid Equipment， Hubei University of Technology,Wuhan 430068,China) |
| Abstract: |
| Acquisition of effective electromagnetic parameters of composites plays an important role in studying their electromagnetic properties.However,the scattering (S) coefficient-based transmission reflection method is unsuitable for composite objects containing metal substrates,making their effective electromagnetic parameters difficult to retrieve.To solve this problem,a new method of information fusion and effective electromagnetic parameter retrieval is proposed.Firstly,Principal Component Analysis (PCA) is used to conduct correlation analysis and global feature extraction for S-coefficients of effective electromagnetic parameter samples,simplifying model complexity.Secondly,a Multi-output Regression Information Fusion Network (MRIFN) model is constructed for parameter retrieval.The model combines Bidirectional Gated Recurrent Unit (BiGRU) to process principal component information,vertically fusing feature information through convolutional layers,enhancing retrieval performance.Finally,datasets of non-magnetic absorbing materials and cellular effective electromagnetic parameters are used to validate the retrieval method.Results show Mean Absolute Percentage Error (MAPE) for real and imaginary parts of dielectric constant is 0.21% and 0.57%,respectively.Considering combined permittivity and permeability effects,MAPE of the proposed method for real and imaginary parts of permittivity is 2.34% and 9.28%,respectively,and for real and imaginary parts of permeability is 2.40% and 6.40%,respectively,verifying it搒 effectiveness and generalization ability. |
| Key words: composite material electromagnetic parameter effective electromagnetic parameter retrieval information fusion deep learning |
