Deep convolutional neural network for segmenting μCT images of fiber reinforcements

The greatest challenge in creating digital material twins and FE mesh from μCT images of composite reinforcements is the lack of a robust and versatile tool for training μCT images. Here, we have used deep convolutional neural networks (DCNN) for segmenting μCT images of a multi-layer plain-weave fiber reinforcement. A set of raw 2D image slices extracted from the gray-scale volume of a single-layer reinforcement was used to train a DCNN using manually annotated images. The trained network was tested against the manually segmented ground truth images and it performed exceptionally well with a global accuracy of more than 96%. The trained DCNN was then used to segment unseen images from a multilayer stack of the fabric with good accuracy. The work presented here provides a robust and efficient framework of segmenting CT scan images of fiber reinforcements for generating digital material twins and FE mesh of fiber reinforcements.