Surrogate assisted optimization of S-type shape memory polymer lattice metastructures for enhanced energy absorption and recovery

Shape memory polymers have attracted growing attention in recent years due to their unique capability of combining high energy absorption with reversible deformation. These characteristics make them promising candidates for lightweight, multifunctional structures in aerospace, automotive, and biomedical applications. In this study, a lattice structure with S-type unit cells (a geometry known for its curved, arc-shaped configuration) is fabricated from polylactic acid (PLA) shape-memory polymer. The thermo-mechanical cycle of deformation, cooling, unloading, and reheating was simulated and experimentally validated to characterize the time-dependent shape-memory behavior. Data extracted from finite element simulations were employed to train an artificial neural network (ANN) surrogate model, which provided accurate predictions of mechanical performance. Subsequently, a genetic algorithm (GA) was applied to optimize lattice geometry with the dual objectives of maximizing energy absorption and enhancing recovery efficiency. Results demonstrated approximately 80% shape recovery upon reheating above T g . The proposed integrated numerical-machine learning-optimization framework highlights the potential of combining high-fidelity physics-based modeling with data-driven surrogate models for the efficient design of advanced energy-absorbing metastructures with excellent recoverability. The novelty of this study does not lie in the geometric shape of the structure itself, but rather in the first-time combination of: (i) experimental and numerical characterization of its full thermo-visco- hyper-elastic shape memory behavior, (ii) development of an ANN surrogate trained on high-fidelity finite element data, and (iii) multi-objective GA optimization specifically targeting the concurrent maximization of energy absorption and shape recovery.