Automatic detection and classification of defects in woven composites manufactured by RTM based on X-ray microtomography

We present an automatized methodology for analysis of defects in Resin Transfer Molded manufactured composites. The experimental microstructures scanned with X-ray microtomography are first segmented with ak-means unsupervised clustering algorithm. The information on phase attribution is then used to extract envelopes of the woven reinforcement. The surface mesh obtained in both phase and structure segmentation steps is then used to retrieve specific geometric and spatial features. The spatio-morphological types of residual voids are finally identified in a multi-dimensional feature space.