Design, Simulation, Fabrication, and Parametric Optimization of Microstrip Antennas Using Deep Learning Algorithms

This review examined four studies that together cover much of the ML/DL toolbox used in microstrip antenna design: a decision-tree-plus-fuzzy-inference             framework  for classification and parameter estimation, a multi-output regression and ResNet study of coupled dual-band geometry, an ANN-based study of operational optimization for 5G smart antennas, and a PRISMA-based survey of the wider field. Rather than grouping them by antenna type, the review followed the order in which a design project is actually carried out: design fundamentals, simulation-based dataset generation, DL modeling, parametric optimization, and fabrication and validation. The main finding, is that none of the reviewed case studies checks its DL predictions against fabricated, measured hardware. The broader survey shows this is not a quirk of the three case studies. It is a pattern across the field. A second finding is that many reported accuracy figures are too good to take at face value. MAPE values near 0% and R² values are more consistent with small, correlated datasets and possible overfitting than with a model that has captured antenna behavior almost perfectly. We flagged these figures throughout the review and did not treat them as settled results. The proposed framework addresses both problems. It has six stages: analytical initialization, constrained dataset generation, forward and inverse DL modeling, surrogate-assisted optimization, fabrication and measurement, and an optional feedback stage. Its defining contribution is fabrication and measurement, where a single geometry is fabricated and its DL-predicted, EM-simulated, and VNA-measured performance are compared directly. Because this project covers fabrication and parametric optimization as well as design and simulation, it can make the comparison that the existing literature leaves out. The framework is intended to serve as the starting structure for the methodology chapter.

[1]    R. Ramasamy and M. A. Bennet, “An Efficient Antenna Parameters Estimation Using Machine Learning Algorithms,” Progress In Electromagnetics Research C, vol. 130, pp. 169–181, 2023. [2]  R. Gadhafi, A. Copiaco, Y. Himeur, K. Afsari, H. Mukhtar, K. Ghanem, and W. Mansoor, “Exploring the Potential of Deep-learning and Machine-learning in Dual-band Antenna Design,” IEEE Open Journal of the Computer Society, 2024, doi: 10.1109/OJCS.2024.3463190. [3]  P. Kavitha and A. Thatikonda, “Optimizing 5G Smart Antenna Design Parameters using Deep Learning,” International Journal of Intelligent Systems and Applications in Engineering, vol. 12, no. 22s, pp. 2073–2080, 2024. [4]  N. Sarker, P. Podder, M. R. H. Mondal, S. S. Shafin, and J. Kamruzzaman, “Applications of Machine Learning and Deep Learning in Antenna Design, Optimization and Selection: A Review,” IEEE Access, vol. 11, pp. 103890–103915, 2023, doi: 10.1109/ACCESS.2023.3317371.