Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Field Reconstruction for High-Frequency Electromagnetic Exposure Assessment Based on Deep Learning

Miao Cao, Zicheng Liu, Bazargul Matkerim, Tongning Wu, Changyou Li, Yali Zong, Bo Qi

TL;DR

The paper tackles near-field RF exposure assessment for millimeter-wave 5G by reconstructing field distributions to compute IPD and verify safety compliance. It marries physics-based initial reconstructions (PWEM or ISM) with a Residual U-Net refinernent trained on diverse, full-wave simulated data to yield accurate evaluation-plane fields. The approach achieves low average relative errors in both the electric field (4.57%) and psIPD (2.97%), demonstrates robustness across reconstruction distances up to $6\lambda$ and sampling densities down to $24\times 24$, and provides a thorough uncertainty analysis for practical measurement conditions. Experimental validation with IEC/IEEE 63195 horn antennas corroborates performance gains over classical methods and supports practical deployment for mmWave RF safety compliance.

Abstract

Fifth-generation (5G) communication systems, operating in higher frequency bands from 3 to 300 GHz, provide unprecedented bandwidth to enable ultra-high data rates and low-latency services. However, the use of millimeter-wave frequencies raises public health concerns regarding prolonged electromagnetic radiation (EMR) exposure. Above 6 GHz, the incident power density (IPD) is used instead of the specific absorption rate (SAR) for exposure assessment, owing to the shallow penetration depth of millimeter waves. This paper proposes a hybrid field reconstruction framework that integrates classical electromagnetic algorithms with deep learning to evaluate the IPD of wireless communication devices operating at 30 GHz, thereby determining compliance with established RF exposure limits. An initial estimate of the electric field on the evaluation plane is obtained using a classical reconstruction algorithm, followed by refinement through a neural network model that learns the mapping between the initial and accurate values. A multi-antenna dataset, generated via full-wave simulation, is used for training and testing. The impacts of training strategy, initial-value algorithm, reconstruction distance, and measurement sampling density on model performance are analyzed. Results show that the proposed method significantly improves reconstruction accuracy, achieving an average relative error of 4.57% for electric field reconstruction and 2.97% for IPD estimation on the test dataset. Additionally, the effects of practical uncertainty factors, including probe misalignment, inter-probe coupling, and measurement noise, are quantitatively assessed.

Field Reconstruction for High-Frequency Electromagnetic Exposure Assessment Based on Deep Learning

TL;DR

The paper tackles near-field RF exposure assessment for millimeter-wave 5G by reconstructing field distributions to compute IPD and verify safety compliance. It marries physics-based initial reconstructions (PWEM or ISM) with a Residual U-Net refinernent trained on diverse, full-wave simulated data to yield accurate evaluation-plane fields. The approach achieves low average relative errors in both the electric field (4.57%) and psIPD (2.97%), demonstrates robustness across reconstruction distances up to and sampling densities down to , and provides a thorough uncertainty analysis for practical measurement conditions. Experimental validation with IEC/IEEE 63195 horn antennas corroborates performance gains over classical methods and supports practical deployment for mmWave RF safety compliance.

Abstract

Fifth-generation (5G) communication systems, operating in higher frequency bands from 3 to 300 GHz, provide unprecedented bandwidth to enable ultra-high data rates and low-latency services. However, the use of millimeter-wave frequencies raises public health concerns regarding prolonged electromagnetic radiation (EMR) exposure. Above 6 GHz, the incident power density (IPD) is used instead of the specific absorption rate (SAR) for exposure assessment, owing to the shallow penetration depth of millimeter waves. This paper proposes a hybrid field reconstruction framework that integrates classical electromagnetic algorithms with deep learning to evaluate the IPD of wireless communication devices operating at 30 GHz, thereby determining compliance with established RF exposure limits. An initial estimate of the electric field on the evaluation plane is obtained using a classical reconstruction algorithm, followed by refinement through a neural network model that learns the mapping between the initial and accurate values. A multi-antenna dataset, generated via full-wave simulation, is used for training and testing. The impacts of training strategy, initial-value algorithm, reconstruction distance, and measurement sampling density on model performance are analyzed. Results show that the proposed method significantly improves reconstruction accuracy, achieving an average relative error of 4.57% for electric field reconstruction and 2.97% for IPD estimation on the test dataset. Additionally, the effects of practical uncertainty factors, including probe misalignment, inter-probe coupling, and measurement noise, are quantitatively assessed.

Paper Structure

This paper contains 21 sections, 10 equations, 18 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Incident Power Density
  4. Plane-Wave Expansion Method
  5. Inverse Source Method
  6. Deep Learning Enhancement
  7. Dataset generation
  8. Training strategy
  9. Results and discussion
  10. Electric field reconstruction accuracy
  11. Effect of training strategy
  12. Effect of classical algorithm selection
  13. Effect of reconstruction distance
  14. Effect of sampling density
  15. IPD distribution and psIPD evaluation
  16. ...and 6 more sections

Figures (18)

  • Figure 1: Schematic of the proposed field reconstruction method.
  • Figure 2: The applied U-Net neural network architecture with residual connections, R-U-Net.
  • Figure 3: Types of antennas used for data generation, (a) Square array, (b) rectangular array, and (c) horn antennas.
  • Figure 4: Comparison of predicted electric fields obtained by different training methods for three types of antennas.
  • Figure 5: (a) Front view (screen hidden) and (b) back view (backplate hidden) of the studied phone model.
  • ...and 13 more figures