Direct modeling of the elastic properties of single 3D printed composite filaments using X-ray computed tomography images segmented by neural networks

This study introduces a new method for creating accurate microscale finite element (FE) models of 3D printed
composites. The approach involves utilizing conventional micro-computed tomography (micro-CT) and neural
network algorithms and is applied to single 3D printed composite filaments that are reinforced with Kevlar
fibers. Initially, images from micro-CT scans are processed using the YOLOv7 (you only look once) algorithm
to differentiate the fibers in the micro-CT images, resulting in an accurate representation of the fibers in the
microstructure. The fibers are then integrated into representative volume elements (RVEs) that are simulated
using the FE method to predict the effective elastic properties of the 3D printed composite. The results
are compared with experiments and indicate that this approach leads to accurate predictions of the elastic
properties. Additionally, it is demonstrated that the printed filaments display transversely isotropic behavior,
with the axis of isotropy aligned with the length of the printed filament. These findings highlight the potential
of this approach for ameliorating the design and production of 3D printed composites.