X-ray computed tomography analysis of defects in 3D printed continuous carbon fibre-reinforced polymers aided by deep learning

3D-printed continuous carbon fibre-reinforced polymers (C-CFRPs) often suffer from higher porosity than conventionally manufactured composites. Here, the volume, distribution, and morphology of defects in 3D-printed C-CFRPs were investigated using X-ray computed tomography. The defects were automatically segmented based on the U-Net deep learning neural network and quantitatively analyzed. The defects are periodically distributed following the laminar structure, featuring ellipsoidal and net-like shapes. The long axes of the ellipsoidal-shaped pores are found to be generally aligned along the fibre direction in each layer, and these pores are more elongated in the top layer than in the bottom layer.