Enhancing Electromagnetic Parameter Modeling of Irregularly Shaped Composite Structures with Partitioned Equivalent Modeling and Near-Field Scanning

To characterize local electromagnetic parameters in aircraft composite skins containing geometric irregularities, such as curved surfaces and slots, this paper proposes a partitioned equivalent modeling (PEM) method. The proposed method provides a parametric basis for low-frequency electromagnetic shielding effectiveness analysis of complex structures. Within the PEM method, an irregular structure is decomposed into homogeneous equivalent partitions; local equivalent electromagnetic parameters are then extracted through near-field scanning measurements, S-parameter-based inversion, and a physics-guided deep neural network. Numerical simulations show that the proposed method maintains low S-parameter reconstruction errors in the 10 MHz-20 MHz frequency range, with magnitude MSE values at the 10⁻⁶ level. Experimental validation using a near-field measurement system further shows that more than 89% of the scanned regions have relative errors below 0.20%, with a maximum average error of 0.57%. These results indicate that the proposed method offers an effective approach for low-frequency near-field characterization and quantitative modeling of geometrically complex composite structures.