Deep learning-driven ultra-stretchable kirigami metamaterials: towards surface texture modulation via buckling

Abstract

Kirigami demonstrates distinctive buckling instability behavior when subjected to tensile stress, bestowing the structure with exceptional stretchability and design versatility. Nonetheless, conventional design methodologies predominantly focus on unidirectional cutting kirigami structures, investigating their buckling instability and out-of-plane configurations derived from geometric symmetry. To enhance the functionality of kirigami and thoroughly explore the mechanisms of buckling behavior upon the disruption of geometric symmetry, as well as to comprehend the impact of geometry on programmability during reconfiguration, we have analyzed the buckling instability mechanisms of tessellated cutting kirigami structures. An innovative design strategy for kirigami is proposed, leveraging deep learning techniques to enable accurate predictions of complex nonlinear constitutive relationships. Our approach offers a programmable design framework, facilitating the targeted identification of optimal kirigami structural patterns based on tensile strain requirements, thereby enhancing the adaptability of the desired mechanical response and minimizing trial-and-error costs. Our results indicate a 94.29% accuracy in mechanical performance predictions using the proposed method. The geometric symmetry-breaking considerably broadens the design space for kirigami. Additionally, through cross-selection and functional design of predicted kirigami cells, information encoding and transmission via kirigami metasurfaces can be achieved. This paper presents a forward-looking kirigami design strategy that predicts required mechanical performance based on functional demands and enables functional configuration.