Inverse design of personalized textiles with architected curved yarns

Textiles are widely used materials with high mechanical tunability arising from their hierarchical architecture. Recent advantages in additive manufacturing enable precise, yarn-level fabrication that allows control of yarn topology and fabrication of personalized textiles with architected curved yarns with a programmable mechanical response. Yet, both the forward and inverse design of these textiles remains challenging due to the large design space and the nonlinear, globally coupled mechanical behavior due to yarn connectivity. Here, we present on a new class of personalized textiles that are weaves with architected curved yarns and an inverse design framework for them that combines a compact Artificial Neural Network (ANN) representation of yarn orientation, a streamline-based weave generator, a calibrated simulation, and a gradient-based optimization method utilizing an ANN surrogate. Together, the inverse design framework enables the optimization of high-dimensional, personalized textile structures that are tailored to produce a target response using significantly fewer simulations than state-of-the-art methods. The proposed framework is experimentally validated and is shown to solve several inverse design problems, from stress-shielding textiles to shape morphing. These results demonstrate that data-efficient, surrogate-driven optimization can satisfactorily navigate within the design space of architected textiles, paving the way for automated, personalized design of functional textiles.