Abstract: A conical origami structure generally has the characteristics of high strength, light weight, transformable configuration, and tunable stiffness. Furthermore, it significantly improves the weak out-of-plane stiffness of traditional and thin-walled origami structures, and thus has bright application prospects for extendable arms, deployable antennas, and energy-absorbing components. However, the current design of conical origami cannot fully consider engineering needs, or develop complex origami structures with specific geometries. Therefore, this work deduces the relationship between the four-fold creases through analytic geometry. Subsequently, based on the deep neural network, the origami creases are predicted and fitted with the geometric parameters. We establish an inverse design framework for four-fold conical origami structures, and realize the inverse design process from a given 3D structure to the 2D origami crease pattern. Computational analysis shows that all the designed conical origami structures are in good agreements with the analytical solutions. In addition, the physical origami model not only verifies the flat-foldability of the designed structure, but also enriches the possible structural configurations by certain operations, such as overlapping and mirroring. This design method is suitable for developing generalized conical origami, and can play a positive role in data-driven origami design.

Key words: four-fold, origami structure, deep neural network, parametric design, inverse design