Akademik Veri Yönetim Sistemi
International Science and Engineering Symposium (KISES 2024) , Zonguldak, Türkiye, 10 - 11 Mayıs 2024, ss.79, (Özet Bildiri)
Recently, using accurate and efficient modelling techniques have become essential in engineering applications. Machine learning techniques have advantages in solving real-world problems and they are cost-effective for reliable predictions over expensive simulations. This paper focuses on building a machine learning model with the intent of Frequency Selective Surface (FSS) design. Artificial Neural Network based unit cell modelling for the transmission coefficient response of the unit cell in terms of its design parameters, was examined. Firstly, the geometry of the unit cell was designed using Electromagnetic (EM) simulation software and the dataset was obtained. Multilayer Perceptron (MLP), widely used in the solution of regression problems, proposes a suitable alternative for FSS design. The findings obtained from the MLP model and the results obtained from the EM simulation were compared. This comparison validated that the response of the EM simulation model was in good agreement with the proposed approach. As a result, research shows that the MLP is effective in predicting the transmission coefficient for the various FSSs.