Key points are not available for this paper at this time.
This study presents a mechanics-guided artificial neural network (ANN) framework for predicting recovery time in 4D-printed metamaterials. The model leverages physically meaningful descriptors, material class, stimulus type, activation temperature, and infill fraction, enabling generalization across multiple material systems while minimizing data dependence. A curated dataset spanning shape memory polymers, hydrogels, shape memory alloys, and photo-responsive composites was compiled for model development and validation. Unlike conventional approaches such as finite element modeling or analytical viscoelastic formulations, which require detailed constitutive laws and geometry-specific inputs, the proposed ANN captures recovery trends directly from mechanics-informed features, delivering accurate estimates at low computational cost. To demonstrate practical applicability, a lightweight interface was developed for rapid evaluation of recovery behavior with minimal calibration. The findings underscore how embedding domain knowledge into data-driven models enhances interpretability and efficiency. This proof-of-concept framework offers significant potential for accelerating material screening, guiding process design, and enabling adaptive control strategies in biomedical, aerospace, and soft robotic applications.
Aswin Karkadakattil (Fri,) studied this question.