| Authors | B. Sivashanmugavalli , R. Mahalakshmi, S. Anusiya |
| Affiliations |
B.S. Abdur Rahman Crescent Institute of Science and Technology, Vandalore, Chennai, India |
| Е-mail | sivashanmugavalli@crescent.education |
| Issue | Volume 18, Year 2026, Number 4 |
| Dates | Received 20 March 2026; revised manuscript received 19 August 2026; published online 21 August 2026 |
| Citation | B. Sivashanmugavalli, R. Mahalakshmi, S. Anusiya, J. Nano- Electron. Phys. 18 No 4, 04010 (2026) |
| DOI | https://doi.org/10.21272/jnep.18(4).04010 |
| PACS Number(s) | 84.40.Ba |
| Keywords | Triband monopole antenna, Machine learning, FNN, CNN (3) , HFSS (15) , Optimizer, Antenna design. |
| Annotation |
This paper presents a machine learning-driven optimization strategy for designing a compact triband monopole antenna operating at 2.4 GHz, 3.5 GHz, and 5.8 GHz, suitable for WLAN, WiMAX, and sub-6 GHz 5G communication. A wide range of antenna models with varied slot patterns were simulated in Ansys HFSS to create a dataset of return-loss (S11) characteristics. Three deep-learning structures a 3-layer Feedforward Neural Network (FNN), a 5-layer FNN, and a Convolutional Neural Network (CNN) were employed to predict optimal slot dimensions, trained with Adam, Nadam, and AdamW optimizers. The 5-layer FNN (Nadam) achieved a prediction accuracy of 66.66 %, while the 3-layer FNN (Adam) produced the best overall match for triband response. To further refine predictions, Differential Evolution (DE) optimization was applied to tune geometric parameters. HFSS re-validation confirmed improved antenna behavior with higher gain, lower S11, and reduced manual design effort. The proposed ML-assisted hybrid framework demonstrates a scalable, time-efficient, and accurate approach for multiband antenna optimization. |
|
List of References |