Automated image segmentation of 3D printed fibrous composite micro-structures using a neural network

A new, automated image segmentation method is presented that effectively identifies the micro-structural objects (fibre, air void, matrix) of 3D printed fibre-reinforced materials using a deep convolutional neural network. The creates training data from physical specimen composed single, straight fibre embedded in cementitious matrix with voids. specific micro-structure this strain-hardening composite (SHCC) obtained X-ray micro-computed tomography scanning, after which ground truth mask sample constructed by connecting each voxel scanned to corresponding object. network trained identify fibres oriented arbitrary directions through application augmentation procedure, eliminates time-consuming task human expert manually annotate these data. predictive capability methodology demonstrated via analysis practical SHCC developed for concrete printing, showing well capable adequately identifying complex micro-structures arbitrarily distributed and fibres. Although focus current study on materials, proposed can also be applied other such as plastics. identified may serve input dedicated finite element models allow computing their mechanical behaviour function composition.