A cross-layer radio frequency system data modeling method and device

By employing a cross-level RF system data modeling method and utilizing neural networks to construct device-level and unit-level proxy models, the problems of hierarchical fragmentation and parameter synchronization difficulties in RF system modeling are solved, enabling efficient and accurate RF system design.

CN122334148APending Publication Date: 2026-07-03SOUTHWEST CHINA RES INST OF ELECTRONICS EQUIP
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SOUTHWEST CHINA RES INST OF ELECTRONICS EQUIP
Filing Date
2026-03-13
Publication Date
2026-07-03

AI Technical Summary

Technical Problem

Existing RF system modeling techniques suffer from problems such as hierarchical fragmentation, complex multi-physics co-simulation, low automation of data-driven modeling, and difficulty in synchronizing parameters across levels, resulting in low design efficiency and difficulty in guaranteeing accuracy.

Method used

A cross-level RF system data modeling method is adopted. By obtaining the scattering parameters of RF devices, performing defolding and normalization, training neural networks to construct device-level and unit-level surrogate models, and combining them with a signal processing framework to construct a time-domain behavior model of the RF channel, and performing simulation verification in a closed link.

Benefits of technology

It achieves seamless synchronization and integrated modeling of parameters across different levels, improving modeling efficiency by more than 50%, reducing human intervention errors, supporting rapid design iteration and early evaluation, and reducing the risk of design rework.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122334148A_ABST
    Figure CN122334148A_ABST
Patent Text Reader

Abstract

This invention discloses a method and apparatus for cross-level radio frequency (RF) system data modeling. The method includes: acquiring scattering parameters of RF devices and performing defolding and normalization preprocessing; training a first neural network based on the processed scattering parameters to construct a device-level surrogate model; training a second neural network based on the device-level surrogate model and electrical behavior data to construct a unit-level surrogate model; encapsulating the unit-level surrogate model to generate a time-domain behavior model of the RF channel; and verifying the accuracy of the RF channel time-domain behavior model in a test link. This invention achieves automated, high-precision, integrated modeling from device to system level.
Need to check novelty before this filing date? Find Prior Art