Composite periodic structure electromagnetic scattering characteristic calculation method based on graph neural network

By constructing graph structures and graph convolutional networks based on graph neural network methods, the problem of low efficiency in calculating the electromagnetic properties of large-scale finite-period structures is solved, and fast and accurate prediction of electromagnetic scattering characteristics is achieved.

CN120597715APending Publication Date: 2025-09-05SOUTHEAST UNIV
0 Cites 0 Cited by

Patent Information

Application Number
CN202510756630.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-06-09
Publication Date
2025-09-05

AI Technical Summary

Technical Problem

Existing commercial simulation software has low computational efficiency when dealing with large-scale finite-period structures, and each time the structural coefficients are adjusted, the grid needs to be re-divided and the matrix equations need to be solved, resulting in excessive computing time and memory consumption.

Method used

A graph neural network-based method is adopted to construct a graph structure and a graph convolutional network, use the feature matrix and adjacency matrix to represent the compound periodic structure, and train the graph neural network model to achieve rapid prediction of the electromagnetic scattering characteristics of the compound periodic structure.

Benefits of technology

The calculation time is greatly reduced, the efficiency of electromagnetic characteristics analysis of composite large-scale finite period structures is improved, and the accuracy and efficiency of online prediction are guaranteed.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a composite periodic structure electromagnetic scattering characteristic calculation method based on a graph neural network, and the method comprises the following steps: selecting a composite finite periodic structure with different periods, structure parameters, relative dielectric constants, incident wave angles and other parameters in a certain range, and constructing a data set; designing a drawing neural network in combination with prior physical information; the preprocessed data set is used for training and testing a network model and selecting an optimal model parameter; and finally, far-field scattering is calculated by using the fully trained graph neural network model. According to the method, prior physical information is fused into the graph neural network, so that the model can quickly predict the far-field scattering characteristics of the composite large-scale finite periodic structure with different structure parameters under multiple incident angles. The construction and reasoning process of the graph replaces establishment and solution of a complex and time-consuming impedance matrix equation, and the modeling analysis efficiency of the composite periodic structure is greatly improved.
Need to check novelty before this filing date? Find Prior Art