Machine learning assisted wrinkling design of hierarchical thin sheets

Wrinkling in thin sheets is a long-standing subject in mechanics. Although the mechanism and morphologies of wrinkles in homogeneous sheets have attracted intense interest and been well-studied, they have been elusive in inhomogeneous sheets. Here, we propose a data-driven framework for wrinkling design by trained back-propagation algorithm with a database of one hundred hierarchical thin sheets structures from finite element analysis. Results show that the wrinkling patterns can be classified into three categories, e.g., scattered wrinkling, decreased wrinkling and increased wrinkling. We further show that the “entropy” of wrinkle has an important role in wrinkling design in the hierarchical thin sheets. This study provides a novel mechanism and strategies for wrinkling design, broadening the promising wrinkling application of hierarchical or programmable thin sheets.