Scholarlink

<p>This study proposes a compact 1 × 2 circular Microstrip patch antenna operating at 0.3 THz, addressing the increasing demand for terahertz (THz) antennas in biomedical, wearable, and ultra-fast wireless applications. The antenna employs Graphene as the radiating element and Kapton II as the substrate enhanced with photonic band gap (PBG) structures including cuboid, cylindrical, and hybrid cavities arranged in a square lattice. Among the five studied configurations, the hybrid PBG design (Antenna V) with dimensions of 1400 × 814&#xa0;μm² demonstrated the best overall trade-off. At a substrate thickness of 55&#xa0;μm, Antenna V achieves a reflection coefficient below − 50 dB, with a bandwidth of approximately 23&#xa0;GHz, a gain of 10.52 dBi, and radiation efficiency around 93.7%. Other substrate thicknesses result in variations in reflection, bandwidth, and efficiency, highlighting the trade-offs in optimizing the performance of Antenna V.To accelerate the design cycle and minimize simulation requirements, four machine learning models namelyLinear Regression (LR), Support Vector Regression (SVR), Random Forest (RF), and Artificial Neural Network (ANN), were trained on CST simulation data to predict antenna performance over varying substrate thicknesses. SVR demonstrated the most consistent predictive accuracy, with a mean absolute error (MAE) of 1.79&#xa0;GHz for bandwidth, 1.03 dB for return loss, 0.11 dBi for gain, and 2.62% for efficiency. ANN followed closely in overall performance, particularly in modeling nonlinear behaviors. In addition, a comparative analysis between CST results and model predictions confirmed SVR’s robustness and ANN’s suitability for complex behavior modeling.The integration of advanced materials, PBG structures, and predictive modeling demonstrates a promising path toward intelligent THz antenna design. The proposed workflow enables faster prototyping while ensuring high fidelity, paving the way for scalable and efficient miniaturized THz components for next-generation communication systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

High Performance Terahertz Circular Microstrip Antenna Optimized Using Machine Learning and Hybrid Photonic Band Gap Structures

  • 期刊论文

  • 发表日期：

    2026年5月27日

查看全文
  • Mustapha Reguig,
  • Allel Mokaddem,
  • Mohammed Belkheir,
  • Mehdi Rouissat,
  • Djamila Ziani

摘要

This study proposes a compact 1 × 2 circular Microstrip patch antenna operating at 0.3 THz, addressing the increasing demand for terahertz (THz) antennas in biomedical, wearable, and ultra-fast wireless applications. The antenna employs Graphene as the radiating element and Kapton II as the substrate enhanced with photonic band gap (PBG) structures including cuboid, cylindrical, and hybrid cavities arranged in a square lattice. Among the five studied configurations, the hybrid PBG design (Antenna V) with dimensions of 1400 × 814 μm² demonstrated the best overall trade-off. At a substrate thickness of 55 μm, Antenna V achieves a reflection coefficient below − 50 dB, with a bandwidth of approximately 23 GHz, a gain of 10.52 dBi, and radiation efficiency around 93.7%. Other substrate thicknesses result in variations in reflection, bandwidth, and efficiency, highlighting the trade-offs in optimizing the performance of Antenna V.To accelerate the design cycle and minimize simulation requirements, four machine learning models namelyLinear Regression (LR), Support Vector Regression (SVR), Random Forest (RF), and Artificial Neural Network (ANN), were trained on CST simulation data to predict antenna performance over varying substrate thicknesses. SVR demonstrated the most consistent predictive accuracy, with a mean absolute error (MAE) of 1.79 GHz for bandwidth, 1.03 dB for return loss, 0.11 dBi for gain, and 2.62% for efficiency. ANN followed closely in overall performance, particularly in modeling nonlinear behaviors. In addition, a comparative analysis between CST results and model predictions confirmed SVR’s robustness and ANN’s suitability for complex behavior modeling.The integration of advanced materials, PBG structures, and predictive modeling demonstrates a promising path toward intelligent THz antenna design. The proposed workflow enables faster prototyping while ensuring high fidelity, paving the way for scalable and efficient miniaturized THz components for next-generation communication systems.