Fast Electromagnetic Scattering Calculation for UAV Targets Based on Graph Neural Networks

To solve the problem of insufficient efficiency in electromagnetic simulation of UAV targets, this paper introduces the Barycentric subdivision method and graph neural network and proposes a fast electromagnetic scattering computational method. This method uses the impedance matrix and right-hand side terms from the method of moments as the node inputs for the graph network, which preserves the electromagnetic principles during computation and ensures the generalization ability. Additionally, the graph network is used to predict updates of the basis function coefficients, accelerating the solution of the matrix equation. Finally, the calculation results of the proposed method, MoM based on direct method and MoM based on BiCGSTAB are compared. The results demonstrate that the proposed method reduces the time consumption by an order of magnitude while maintaining accuracy, proving its effectiveness.