Inferring material properties from FRP processes via sim-to-real learning

Abstract Fiber reinforced polymers (FRP) provide favorable properties such as weight-specific strength and stiffness that are central for certain industries, aerospace or automotive manufacturing. Liquid composite molding (LCM) is a family of often employed, inexpensive, out-of-autoclave manufacturing techniques. Among them, resin transfer (RTM), offers high degree automation. Herein, textile preforms saturated by fluid polymer matrix in closed mold.Both impregnation quality level fiber volume content crucial importance the final part quality. We propose to simultaneously learn three major (fiber permeability X Y direction) presented three-dimensional map based on sequence camera images acquired flow experiments compare CNNs, ConvLSTMs, Transformers. Moreover, we show how simulation-to-real learning can improve digital twin FRP manufacturing, compared simulation-only models sparse real data. The overall best metrics are: IOU 0.5031 Accuracy 95.929 %, obtained pretrained transformer models.