A Phaseless Source Reconstruction Method Based on Two-Stage Physics-Informed Neural Network

View at publisher Summarise with VIRAT
Authors: Zhao Feng, Shijian Zheng, Xiaoting Yuan, Shaoqiu Xiao, Hao Chen, Yin Li
Cited by: 0
References: 21

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In this work, a phaseless equivalent source reconstruction method based on a two-stage physics-informed neural network (PINN) is proposed. Unlike conventional dipole-based equivalent source artificial neural networks that rely on Green's functions, the proposed approach is capable of recovering the phase information of the equivalent sources while simultaneously predicting both near-field magnitude and phase. By leveraging the physical mapping relationship between equivalent sources and near-field magnitudes, the proposed framework enables the joint reconstruction of the magnitude and phase of the equivalent sources. Moreover, through a phase-information constraint, it effectively alleviates the multi-solution problem inherent in phase retrieval. Compared with traditional iterative phase-retrieval algorithm, the proposed method exhibits superior performance in both magnitude prediction accuracy and phase recovery capability. Two numerical examples are provided to validate the effectiveness of the proposed approach. Simulation results show that for a 2×2 patch antenna array and a microstrip line on glass-base substrate, the method achieves magnitude prediction errors of 0.7% and 2%, respectively, while recovering relatively accurate phase distributions. These results indicate that the proposed method is well suited for near-field measurement, as well as antenna analysis and design applications.

Fields of Study & Subjects

Electromagnetic Compatibility and Measurements Model Reduction and Neural Networks Antenna Design and Optimization

Publication details

Journal / source
IEEE Antennas and Wireless Propagation Letters
Volume
25
Issue
6
Pages
2376–2380
Publication date
25 March 2026
Language
EN