Machine learning-based reflection coefficient and impedance prediction for a meandered slot patch antenna
This article presents a real-time tuning approach for the impedance matching circuit of a slotted patch antenna across a broad frequency spectrum. A regression-based machine learning (ML) model is developed to replace costly and time-consuming VNA-based measurements. The model predicts matching ci...
Saved in:
| Summary: | This article presents a real-time tuning approach for the impedance matching circuit of a slotted patch antenna
across a broad frequency spectrum. A regression-based machine learning (ML) model is developed to replace
costly and time-consuming VNA-based measurements. The model predicts matching circuit parameters, including inductance, capacitance, and reflection coefficient (S11), using antenna slot dimensions and operating frequencies. A meandered slot patch antenna is designed and tuned at 2.45 GHz using an EM simulator, with variations in slot length and gap enabling operation between 2 GHz and 3 GHz. A dataset of 500 samples,
including resonance frequency and S11 values, is generated, and an equivalent RLC circuit is modeled to
calculate inductance and capacitance. The random forest algorithm is applied, achieving a maximum prediction
error of 12.46 % for S11 and stable R2 accuracy of 0.98 validated through 10-fold cross-validation. The results
confirm that the ML-based approach provides fast and accurate predictions, requiring minimal computation time.
This makes it a practical solution for real-time impedance matching in advanced antenna systems, such as those
used in wireless communication applications. |
|---|