Influence of yarn deformation on the mechanical behavior of 2.5D 1/2-twill warp-lining woven composites via an OCRNet+HRNet based parametric modeling method

The 2.5D 1/2-twill warp-lining woven composites (2.5D-1/2T-WLWC) are a highly promising material, but the precise extraction of its meso-structural parameters and the rapid construction of high-fidelity models remain key challenges limiting its engineering applications. In this work, a semantic segmentation technique in the field of computer vision, i.e., OCRNet + HRNet network, is introduced, and a parametric and automated modeling method considering yarn extrusion is proposed on this basis, which introduces a parameter λ for yarn extrusion control and a model interference correction method for yarn node control. Subsequently, the rationality of the proposed modeling method is verified in terms of geometrical parameters, mechanical properties and damage modes, and the influence of the parameter λ on the mechanical behavior of the material is analyzed. The results show that the average comparison error of geometric parameters is only 5.16 %, the maximum prediction error of mechanical properties is only 3.51 %, the predicted stress-strain curve is basically consistent with the experimental curve, and the key features of the ultimate damage morphology coincide with the actual fracture. It is also revealed that as parameter λ increases, the stiffness and strength of the material under warp loading increase, while they exhibit an initial increase followed by a decrease under weft loading. Therefore, by regulating parameter λ, the internal stress distribution of the material can be optimized, and its mechanical properties and damage resistance can be improved, which has a significant engineering reference value for material design and molding process.