Segmentation Tools for Delineating the Profiles of Pulled-out Fibre Bundles in In-Situ Tomograms of Translaminar Fracture of Thin-Ply Composites

  • AhmadvashAghbash, Sina (KU Leuven)
  • Rojas Gomez, Camilo (KU Leuven)
  • Broggi, Guillaume (Ecole Polytechnique Fédérale de Lausanne)
  • Aydemir, Abdullah (Ecole Polytechnique Fédérale de Lausanne)
  • Argyropoulos, Alexios (EPFL, North Thin Ply Technology SARL)
  • Cugnoni, Joël (EPFL, HEIG-VD)
  • Michaud, Véronique (Ecole Polytechnique Fédérale de Lausanne)
  • Mehdikhani, Mahoor (KU Leuven)
  • Swolfs, Yentl (KU Leuven)

Please login to view abstract download link

The translaminar fracture toughness of a fibre-reinforced composite reflects its damage tolerance under longitudinal tension. Fibre pull-out process is one of the main energy-dissipation contributors to translaminar toughness. Our previous work [1] proposed and successfully tested a downscaled compact tension specimen for translaminar fracture analysis of thin-ply composites. The in-situ experiments, conducted at TOMCAT beamline of the Swiss Light Source (Paul Scherrer Institut), are followed by reconstruction of tomographic volumes. The U-net convolutional neural network in Dragonfly software is employed for semantic segmentation. For the first time, the in-situ pulled-out bundle profiles (see Fig. 1), their maximum height and the interrelationship with the associated translaminar fracture toughness are analysed per lay-up configuration. Finally, the effect of fibre-by-fibre hybridisation and/or ply-blocking in altering the fracture toughness, is assessed.