Machine learning-assisted design of a microwave metamaterial absorber using PMMA epoxy/MWCNT/Fe3O4 nanocomposite for X-band applications

A PMMA-epoxy nanocomposite reinforced with multi-walled carbon nanotubes (MWCNTs) and Fe 3 O 4 nanoparticles is developed for multiband microwave metamaterial absorbers (MMAs). The hybrid nanocomposite exhibits stable dielectric behavior over a wide frequency range, with low losses at lower frequencies and enhanced dielectric and magnetic losses in the X-band, arising from interfacial polarization and magnetic dipolar effects. With this properties, a compact MMA is designed and demonstrates multiband absorption at 7.71, 7.94, 8.79, 8.89, and 13.30 GHz, achieving an average absorptivity of 96.16% with polarization-insensitive performance. Unlike conventional FR4-based absorbers, the proposed design combines magnetic-dielectric nanocomposite functionality with geometric compactness and angular stability. Furthermore, absorptivity is predicted using machine learning models such as CatBoost, Extra Trees, XGBoost, Random Forest, K-Nearest Neighbors, and an ensemble regressor, which reduce computational cost. The findings show that combining experimentally validated nanocomposites with data-driven optimization is an effective way to design high-performance, multiband absorbers for use in radar stealth, EMI shielding, and sensing.