Automated 3D digital reconstruction of fiber reinforced polymer composites

A 3D reconstruction system is described that generates digital reconstructions of an aerospace-grade graphite fiber reinforced composite in an accurate and highly automated way using high resolution X-ray computed tomography (CT) imagery. The 3D reconstruction system is based on algorithms from the multi-target tracking literature including a tracker using a global nearest neighbor data association architecture and Kalman filter, along with algorithms for track stitching and smoothing. Additionally, two new algorithms are developed to ensure that none of the virtual fibers in the 3D reconstruction overlap each other. First, is an algorithm for selecting among multiple possible tracks in the case of significant overlap. Second, is an algorithm that solves for an optimal, in the maximum a posteriori sense, adjustment for the virtual fiber locations and radii such that none of the virtual fibers overlap. The 3D reconstruction system is evaluated with X-ray CT data for an AS4/3501-6 graphite/epoxy composite. A representative sub-volume containing 629 fibers across 427 CT images, is used for evaluation. Additional algorithms are developed to manually correct any minor errors in the automatically generated 3D reconstruction leading to a fully accurate 3D reconstruction. In addition to being highly automated, the 3D reconstruction system is readily extensible to larger test specimens and other fiber arrangements.