Scholarlink

<p>High-resolution, low sidelobe level microstrip antenna arrays are critical for modern radar systems. However, designing such arrays under manufacturing cost constraints remains challenging. This study presents a hybrid method to overcome the design challenge of X-band microstrip antenna arrays using standard FR4 substrates. Therefore a hybrid optimization methodology combining the Adaptive Neuro-Fuzzy Inference System (ANFIS), Physics-Informed Neural Networks (PINN), and Gradient-Penalised Wasserstein Generative Adversarial Networks (WGAN-GP) was employed. A total of 222 HFSS simulations were performed at 10.321&#xa0;GHz, with five design parameters (patch width, patch length, element spacing, and feed positions) mapped to five performance metrics (gain, return loss, bandwidth, efficiency, and sidelobe level). The ANFIS model learned complex electromagnetic relationships, while PINN ensured physics consistency through Maxwellssss equations and resonance constraints, and WGAN-GP generated optimized design candidates. The hybrid model achieved 12.7% average prediction error compared to HFSS (High Frequency Structure Simulator) simulations. The optimized 2 × 2 array design was fabricated on FR4 substrate (εr = 4.4, tanδ = 0.02, h = 1.6&#xa0;mm) and measured using a Keysight N5224A vector network analyzer. Measured results showed excellent impedance matching (S<sub>11</sub> = − 40.89&#xa0;dB at 10.2&#xa0;GHz) and 700&#xa0;MHz bandwidth. Side lobe levels between − 3.1 and − 4.6&#xa0;dB, while higher than ideal, were found to be consistent with the expected FR4 substrate limitations due to the high loss tangent. The results obtained throughout the entire process from design to measurement for practical and cost-constrained antenna design demonstrate the effectiveness of the proposed model.</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

Hybrid ANFIS-PINN-WGAN-GP Optimization for Low-Cost Fr4-Based X-Band Microstrip Antenna Array: Design, Fabrication, and Measurement

  • 期刊论文

  • 发表日期：

    2026年2月20日

查看全文
  • Çağlar Tozoğlu,
  • Ahmet Kayabaşı,
  • Rabia Toprak

摘要

High-resolution, low sidelobe level microstrip antenna arrays are critical for modern radar systems. However, designing such arrays under manufacturing cost constraints remains challenging. This study presents a hybrid method to overcome the design challenge of X-band microstrip antenna arrays using standard FR4 substrates. Therefore a hybrid optimization methodology combining the Adaptive Neuro-Fuzzy Inference System (ANFIS), Physics-Informed Neural Networks (PINN), and Gradient-Penalised Wasserstein Generative Adversarial Networks (WGAN-GP) was employed. A total of 222 HFSS simulations were performed at 10.321 GHz, with five design parameters (patch width, patch length, element spacing, and feed positions) mapped to five performance metrics (gain, return loss, bandwidth, efficiency, and sidelobe level). The ANFIS model learned complex electromagnetic relationships, while PINN ensured physics consistency through Maxwellssss equations and resonance constraints, and WGAN-GP generated optimized design candidates. The hybrid model achieved 12.7% average prediction error compared to HFSS (High Frequency Structure Simulator) simulations. The optimized 2 × 2 array design was fabricated on FR4 substrate (εr = 4.4, tanδ = 0.02, h = 1.6 mm) and measured using a Keysight N5224A vector network analyzer. Measured results showed excellent impedance matching (S11 = − 40.89 dB at 10.2 GHz) and 700 MHz bandwidth. Side lobe levels between − 3.1 and − 4.6 dB, while higher than ideal, were found to be consistent with the expected FR4 substrate limitations due to the high loss tangent. The results obtained throughout the entire process from design to measurement for practical and cost-constrained antenna design demonstrate the effectiveness of the proposed model.