AI-driven design and electrothermal actuation of 4D-Printed carbon fiber-reinforced composites

Abstract

This study presents a data-driven framework for the design, prediction, and optimization of electrothermally actuated composite structures based on continuous carbon-fiber reinforced shape memory polymers (CCF-SMPs). A Multiphysics model was developed and calibrated against experimental thermal-field, unfolding-trajectory, and recovery-force measurements. The resulting simulation–experiment dataset was then used to train and evaluate a surrogate model. Continuous fibers contributed to improved thermal uniformity, reduced activation gradients, and increased recovery force relative to pure SMP through a coupled electro-thermo-mechanical effect, rather than through thermal conductivity alone. A surrogate model trained on hybrid experimental-simulation datasets was developed to predict key outputs, including temperature evolution, unfolding angle, recovery force, and energy consumption, and was evaluated against independent experimental measurements. The framework was further validated on architected composite actuator units with varied fiber orientations and hinge geometries. AI-predicted force-displacement responses showed less than 7% stiffness error and less than 5% peak-force error relative to experiments. Finally, the optimizer was applied to variable-thickness morphing structures, achieving a 35% reduction in actuation time, a 28% improvement in geometric accuracy, and a 22% reduction in energy consumption. These results demonstrate the potential of AI-guided design for next-generation high-performance morphing composite systems.