Machine Learning Approach for Predicting Electromagnetic Shielding Effectiveness of Metallic Grids in the Time and Frequency Domains

Accurate evaluation of electromagnetic shielding effectiveness (SE) is crucial for protecting modern electronic systems against electromagnetic interference (EMI) and transient disturbances such as electromagnetic pulses (EMP). This study investigates both time‐domain shielding effectiveness (TDSE) and frequency‐domain shielding effectiveness (FDSE) of metallic grid structures on dielectric substrates. TDSE metrics, including Peak SE and Derivative SE for electric and magnetic fields, quantify the attenuation of both field amplitude and its temporal rate of change under transient EM exposure. Full‐wave simulations using the Finite Integration Technique (FIT) in computer simulation technology microwave studio (CST‐MWS) were performed to generate datasets for training multilayer perceptron (MLP) neural networks. The MLP models map five structural and material parameters—aperture width, metal thickness, substrate thickness, relative permittivity, and loss tangent—to TDSE and FDSE responses. For TDSE prediction, the trained network achieves a root mean square error (RMSE) of 0.02215 and R 2 of 0.9846 on test data, demonstrating high predictive accuracy. For FDSE prediction across 1–4 GHz, the network provides close agreement with simulated spectra. Furthermore, a neural network‐based surrogate model is employed for rapid optimisation of metallic grid design under target shielding criteria. Comparisons with the Trust Region Framework in CST show that the surrogate‐based approach maintains high accuracy while significantly reducing computational time and optimisation cost. The proposed methodology enables efficient evaluation, prediction, and optimisation of metallic grid configurations for electromagnetic shielding applications under transient conditions.