Q-Factor Evaluation Accelerated by a Deep Neural Network

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21230%2F25%3A00384560" target="_blank" >RIV/68407700:21230/25:00384560 - isvavai.cz</a>

Výsledek na webu

<a href="https://doi.org/10.23919/EuCAP63536.2025.10999251" target="_blank" >https://doi.org/10.23919/EuCAP63536.2025.10999251</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.23919/EuCAP63536.2025.10999251" target="_blank" >10.23919/EuCAP63536.2025.10999251</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Q-Factor Evaluation Accelerated by a Deep Neural Network

Popis výsledku v původním jazyce

Antenna inverse design has gained significant popularity, yet it often faces challenges with slow convergence to optimal solutions. This paper presents a novel deep-learning algorithm for accelerated Q-factor evaluation based on the Method-of-Moments (MoM). The proposed deep neural network directly processes triangular antenna meshes to estimate the vector of current expansion coefficients, effectively replacing the traditional impedance matrix inversion required in MoM. The Q-factor is subsequently computed. For this purpose, we introduce QfResNET, a Residual Network (ResNET) architecture featuring an upsampling convolutional network and the GELU activation function. Antennas are represented as multi-channel 2D arrays, preserving the neighborhood relationships in the antenna mesh and labeled with Q-factors derived through MoM calculations. Our results demonstrate that QfResNET achieves a speed-up of up to 957% compared to MoM. However, for coarser antenna meshes, the deep neural network does not offer a significant advantage.

Název v anglickém jazyce

Q-Factor Evaluation Accelerated by a Deep Neural Network

Popis výsledku anglicky

Antenna inverse design has gained significant popularity, yet it often faces challenges with slow convergence to optimal solutions. This paper presents a novel deep-learning algorithm for accelerated Q-factor evaluation based on the Method-of-Moments (MoM). The proposed deep neural network directly processes triangular antenna meshes to estimate the vector of current expansion coefficients, effectively replacing the traditional impedance matrix inversion required in MoM. The Q-factor is subsequently computed. For this purpose, we introduce QfResNET, a Residual Network (ResNET) architecture featuring an upsampling convolutional network and the GELU activation function. Antennas are represented as multi-channel 2D arrays, preserving the neighborhood relationships in the antenna mesh and labeled with Q-factors derived through MoM calculations. Our results demonstrate that QfResNET achieves a speed-up of up to 957% compared to MoM. However, for coarser antenna meshes, the deep neural network does not offer a significant advantage.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

20201 - Electrical and electronic engineering

Projekt

<a href="/cs/project/GM21-19025M" target="_blank" >GM21-19025M: Optimální návrh v elektromagnetismu založený na lokální perturbaci přesných modelů</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2025

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

2025 19th European Conference on Antennas and Propagation (EuCAP)

ISBN

978-88-31299-10-7

ISSN

2164-3342

e-ISSN

2164-3342

Počet stran výsledku

5

Strana od-do

—

Název nakladatele

IEEE

Místo vydání

Anchorage, Alaska

Místo konání akce

Stockholm

Datum konání akce

30. 3. 2025

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

001507659900089