Machine learning-assisted shear horizontal guided wave propagation in functionally graded piezomagnetic sandwich structures with imperfect interfaces

This study develops an analytical and machine learning framework to investigate shear horizontal guided (SHG) wave dispersion in functionally graded piezomagnetic sandwich structures. The configuration comprises two dissimilar graded piezomagnetic layers joined through a spring-membrane imperfect interface and coated with surface sensing layers incorporating mass loading. Dispersion relations are derived from governing magnetoelastic equations and boundary and continuity conditions under magnetically open and short circuit conditions. Parametric analyses examine the influences of material gradation, magnetomechanical coupling, interface stiffness, membrane inertial density, and normalised density parameters. The results reveal pronounced sensitivity of dispersion characteristics to interfacial properties and bonding conditions, supporting potential interface characterisation through subsequent inverse analysis and experimental validation. Artificial Neural Network (ANN), Random Forest (RF), and Gaussian Process Regression (GPR) models are trained on analytically generated datasets, and a hybrid ANN-RF-GPR framework is developed for rapid dispersion prediction. Agreement between analytical and surrogate predictions supports predictive accuracy within the investigated dataset. The methodology provides an efficient computational basis for forward modelling and parametric investigation, with potential applications in nondestructive evaluation and structural health monitoring, subject to experimental verification and validated inverse identification procedures.