A predictive deep-learning approach for homogenization of auxetic kirigami metamaterials with randomly oriented cuts

This paper describes a data-driven approach to predict mechanical properties of auxetic kirigami metamaterials with randomly oriented cuts. The finite element method (FEM) was used to generate datasets, the convolutional neural network (CNN) was introduced to train these data, and an implicit mapping between the input orientations of cuts and the output Young's modulus and Poisson's ratio of the kirigami sheets was established. With this input-output relationship in hand, a quick estimation of auxetic behavior of kirigami metamaterials is straightforward. Our examples indicate that if the distributions of training and test datasets are close to each other, a good prediction is achievable. Our efforts provide a fast and reliable way to evaluate the homogenized properties of mechanical metamaterials with various microstructures, and thus accelerate the design of mechanical metamaterials for diverse applications.