Broadband EMI shielding performance (100 MHz—12 GHz) of diverse materials

Abstract

This study presents a data-driven predictive model for estimating Electromagnetic Interference (EMI) Shielding Effectiveness (SE) over a frequency range of 100 MHz to 12 GHz. A hybrid modelling method was developed that uses both material-specific parameterization and frequency–response correlations. The Ordinary Least Squares (OLS) regression technique was employed for model calibration, resulting in a linear model as a function of frequency. The framework was implemented in MATLAB and tested using six independent experimental datasets like conventional metals (copper, aluminium, steel, silver) and advanced materials (graphene, carbon composites). The dataset was randomly split into 70% training data and 30% testing data. Furthermore, tenfold cross-validation was applied during the training phase to enhance model reliability and minimize overfitting. The framework achieved a mean coefficient of determination (R²) of 0.762 ± 0.098 and Root Mean Square Error (RMSE) of 1.79 ± 0.68 dB on held-out test data. Graphene-based materials exhibited the highest predictability (R² = 0.914, RMSE = 1.12 dB), while aluminium and steel showed moderate accuracy (R² > 0.65). The 95% confidence interval for mean R² was [0.683, 0.840]. The complete code and datasets are provided as supplementary materials to ensure reproducibility and facilitate community-driven improvement. This validated computational tool enables rapid EMI shielding material selection for 5G/6G communications, Internet of Things (IoT) devices, and weight-constrained aerospace systems.