Data-driven and interpretable stiffness modeling of deployable origami structures

Deployable origami structures exhibit highly tunable stiffness characteristics, yet efficient and interpretable modeling across different deployment states remains challenging. This study proposes a data-driven sym bolic learning framework to model the axial and bending stiffness of deployable origami structures based on key geometric parameters, including fold thickness, panel thickness, and fold width. Kolmogorov Arnold Networks are employed to identify explicit analytical expressions directly from data, avoiding the limitations of black-box machine learning approaches. After systematic simplification guided by variable ranges and physical relevance, compact and interpretable stiffness models are obtained. The results re veal that axial stiffness is strongly influenced by fold thickness and panel thickness, reflecting the role of folds as load-transferring and rotational-resisting components. In contrast, bending stiffness is dominated by panel thickness and fold width, while the contribution of fold thickness is negligible due to the panel controlled bending inertia. By introducing deployment ratio as a governing parameter, unified stiffness models valid across multiple deployment states are established. The proposed approach provides an efficient and physically transparent alternative to finite element–based stiffness evaluation.