A machine learning approach to predicting mechanical behaviour of non-rigid foldable square-twist origami

Origami-inspired metamaterials offer unlimited possibilities and broad application prospects in science and engineering. In particular, non-rigid foldable origami structures potentially provide more favorable character-istics than rigid foldable ones owing to the flexibility of thin sheets. However, the previous rigid analytical so-lutions are not sufficient to reflect the actual motions of non-rigid origami because of facet deformation, which prevents the widespread promotion of non-rigid origami applications. In this study, a machine learning approach has been one of first proposed for predicting the mechanical behaviours of a non-rigid foldable square-twist origami pattern (Type 1). Decision tree and multilayer perceptron were applied to classify multi-stability and predict the transition point, maximum strain energy, and energy barrier of the square-twist origami structure. Both methods efficiently provided accurate results for various geometric and material parameters. It is worth mentioning that the dimensional analysis added to this approach can reduce the computational time and increase the accuracy of the predicted results. Due to the same folding principle and similar parameters of origami pat-terns, the machine learning approach proposed herein is a promising alternative for other complex origami engineering problems when analytical and empirical solutions are unavailable.

Automated summary

Origami-inspired metamaterials have unlimited possibilities and broad applications in science and engineering. Machine learning approaches can accurately predict the mechanical behaviors of non-rigid origami structures, providing strong support for their applications.