High-Precision Reconstruction and Rapid Adaptive Prediction Method of Nonuniform Spatial EM Coupling Distribution for Sensitive Passive Circuits

In this article, a high-precision reconstruction and rapid prediction for spatial electromagnetic coupling with adaptive optimization method among different sensitive passive circuits is proposed. To improve the low accuracy of source reconstruction caused by inappropriate initial values of dipole moment, the transfer matrix describing the relationship between the dipole and its radiation field is split. By combining it with the singular value decomposition (SVD) method, the ill-conditioning of matrix and the correctness of initial value are both improved. Subsequently, the dipole moments are optimized by genetic algorithm to find the optimal solution, enabling accurate characterization of the EMI source. However, the EMI-induced resonances, generated by the coupling between passive circuits, can lead to nonuniform radiation field distribution. Therefore, generalized neural network with K-means clustering method is used to construct an electromagnetic model library, which not only reduces the computational time but also accurately characterize the nonuniform near-field of sensitive passive circuits. Moreover, the noise coupling voltage is calculated, and the position of the victim circuit is adaptively optimized to achieve the minimum noise coupling. Finally, the effectiveness of the method is verified by numerical simulations and physical measurement results.