Machine learning-assisted energy-aware adaptive vibration control of functionally graded smart composites with Terfenol-D layers

This work presents a machine learning-guided optimization framework for functionally graded smart composite structures integrated with magnetostrictive layers for adaptive vibration control. The functionally graded host is coupled with Terfenol-D/Galfenol layers to combine material gradation effects with magnetic-field-driven actuation. High-fidelity numerical simulations are conducted across a wide design domain, including gradation index, magnetostrictive layer thickness, magnetic field intensity, boundary conditions, geometric ratio, and excitation frequency. The resulting physics-based dataset is used to train surrogate machine learning models for predicting natural frequency, vibration amplitude, frequency response characteristics, and vibration suppression efficiency. These trained models are further embedded into an optimization scheme to determine material and actuation parameters for enhanced vibration attenuation. By replacing repeated expensive simulations with rapid predictive models, the proposed strategy establishes an effective connection between artificial intelligence and smart composite design. The optimized configurations exhibit higher dynamic stiffness, lower vibration response, and improved adaptive control capability compared with non-optimized designs. Feature-sensitivity analysis confirms that material gradation, magnetostrictive layer thickness, and magnetic field intensity are the most influential parameters controlling vibration suppression. The developed framework offers an efficient AI-assisted route for inverse design and performance optimization of functionally graded magnetostrictive composites for aerospace, robotic, precision mechanical, and vibration-isolation applications.