A Convolution Neural Network-based Method for Designing Honeycomb Absorbing Material

• Published in: 2019 Photonics & Electromagnetics Research Symposium - Fall (PIERS - Fall) pp. 1401 - 1404
• Main Authors: Han, Lingyan, Kuang, Lei, Liu, Huan, Lu, Jianxia, Liu, Qing Huo
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2019
• Subjects: Coatings; Convolutional neural networks; Electromagnetics; Honeycomb structures; Optimization; Photonics; Training
• DOI: 10.1109/PIERS-Fall48861.2019.9021524

In this paper, we propose a fast method for optimizing multiple size parameters of honeycomb absorbing material including honeycomb height and dip coating thickness, based on deep convolutional neural networks (CNN). The reflection coefficients (S 11 ) of honeycomb absorbing material are simulated to generate CNN training data as inputs to the CNN in this paper. The network consists of six concatenated convolution-maximum pooling layers and three fully-connected layers. The convolutional layers and the fully-connected layers perform feature extraction and data regression, respectively. When the correlation coefficient between the predicted value and the true value of the cellular absorbing material using CNN inversion is close to 1.0, it means CNN performs well in inversion. The trained CNN model is used to optimize the honeycomb absorbing material in a given size range. Numerical results show that the optimized S 11 of honeycomb absorbing material is even about 10 dB lower than that of the honeycomb without optimizing at some frequencies. This CNN based method provides a faster and more efficient approach for electromagnetic analysis.