Instance segmentation and statistical analysis of large-field-of-view X-ray computed tomography data of fiber weave structures

We present methods for the geometric characterization of hierarchical 3D microstructures of continuously carbon fiber reinforced composites made from stacked woven textile preforms, using X-ray CT and statistical image analysis. As the representative volume of these composites is rather large (in-plane side length > 20 mm) and, simultaneously, contains very small details (< 10 µm), we propose to divide the characterization procedure into three steps, considering (1) single rovings, (2) single weave layers, and (3) full laminates. In this paper, we focus on the geometry of single rovings and a single weave layer. Special emphasis is put on deviations from idealized shapes, including deformation of the commonly assumed elliptical cross section of single rovings, which results from the interaction between the undulating orthogonal rovings. We show that roving cross sections exhibit periodic changes in shape corresponding to their position in the weave structure. Using a mathematical modeling approach, we quantify the variability between rovings.