Automated generation of labeled synthetic training data for machine learning based segmentation of 3D-woven composites

Written by: Johan Friemann, Lars P. Mikkelsen, Carolyn Oddy and Martin Fagerström

Published on: July, 2024

Summary

Composite parts with 3D-textile reinforcement show promise in high-performance applications. For widespread use, accurate material characterisations are required. Characterisation of the textile architecture in the as-manufactured state may be performed with X-ray CT. Due to the similarity between the chemical composition of carbon fibres and epoxy based matrices, the contrast of X-ray CT scans is poor. Therefore, segmentation with classical methods is difficult or even impossible. Alternatively, machine learning based segmentation approaches may be used. One drawback of machine learning-based algorithms is the need for large datasets whose ground truth labellings require extensive manual labour. This can be circumvented by utilising automatically labelled synthetic X-ray CT data. In this work, a novel pipeline that generates synthetic CT image datasets, with automatically labelled ground truths, is developed. The pipeline is entirely based on free and/or open source software. It is demonstrated that segmentation model, trained on only such data, is able to accurately segment a real X-ray CT scan of a 3D-reinforced carbon fibre composite sample. A pixel-wise agreement of 88% is reached when compared to a manual segmentation. This implies potentially large time savings in segmentation tasks, which could accelerate characterisation of textile composites in their as-manufactured state.

Award

Best paper award for tomography outreach tools: Vidal, F. P., Afshari, S., Ahmed, S., Atkins, C., Béchet, É., Corbí Bellot, A., ... Tugwell-Allsup, J. (2024). Müller, B., & Wang, G. (Eds.). "X-ray simulations with gVXR as a useful tool for education, data analysis, set-up of CT scans, and scanner development". Developments in X-Ray Tomography XV, SPIE Optics & Photonics 2024. DOI: 10.1117/12.3025315

Publications

Friemann, J., Mikkelsen, L. P., Oddy, C., & Fagerström, M. (2025). Synthetic, automatically labelled training data for machine learning based X-ray CT image segmentation: Application to 3D-textile carbon fibre reinforced composites. Composites Part B: Engineering, p. 112656. doi:10.1016/j.compositesb.2025.112656
Friemann, J., Mikkelsen, L. P., Oddy, C., & Fagerström, M. (2024). Automated generation of labeled synthetic training data for machine learning based segmentation of 3D-woven composites. Proceedings of the 21st European Conference on Composite Materials: Special sessions (pp. 333–338). doi:10.60691/yj56-np80
Vidal, F. P. et al (2024, Nov). X-ray simulations with gVXR as a useful tool for education, data analysis, set-up of CT scans, and scanner development. Developments in X-Ray Tomography XV, SPIE Optics & Photonics 2024. doi:10.1117/12.3025315 .

Citations

@article{FRIEMANN2025112656,
  author = {Friemann, Johan and Mikkelsen, Lars P. and Oddy, Carolyn and 
    Fagerstr{\"o}m, Martin},
  doi = {10.1016/j.compositesb.2025.112656},
  issn = {1359-8368},
  journal = {Composites Part B: Engineering},
  keywords = {Segmentation; X-ray CT; 3D-textile reinforced composites; 
    Machine learning; Open source software},
  pages = {112656},
  title = {{Synthetic, automatically labelled training data for machine 
    learning based X-ray CT image segmentation: Application to 3D-textile 
    carbon fibre reinforced composites}},
  year = {2025},
  note = {**XCT**, **ML**, **Segmentation**, **Material science**}
}

@inproceedings{Joahn2024ECCM,
  author = {Friemann, Johan and Mikkelsen, Lars P. and Oddy, Carolyn and 
    Fagerstr\"{o}m, Martin},
  title = "Automated generation of labeled synthetic training data for machine 
    learning based segmentation of {3D}-woven composites",
  booktitle = "Proceedings of the 21st European Conference on Composite 
    Materials: Special sessions",
  year = 2024,
  volume = 8,
  pages = {333-338},
  doi = {10.60691/yj56-np80}
}

@inproceedings{Vidal2024SPIE,
  author = {Franck P. Vidal and Shaghayegh Afshari and Sharif Ahmed and
    Carolyn Atkins and {\'E}ric B{\'e}chet and Alberto Corb{\'i} Bellot and
    Stefan Bosse and Younes Chahid and Cheng-Ying Chou and Robert Culver and
    Lewis Dixon and Johan Friemann and Amin Garbout and Cl{\'e}mentine Hatton and
    Audrey Henry and Christophe Leblanc and Alberto Leonardi and
    Jean Michel L{\'e}tang and Harry Lipscom and Tristan Manchester and
    Bas Meere and Simon Middleburgh and Iwan Mitchell and Liam Perera and
    Mart{\'i} Puig and Jenna Tugwell-Allsup},
  title = {{X-ray simulations with gVXR as a useful tool for education, 
    data analysis, set-up of CT scans, and scanner development}},
  volume = {13152},
  booktitle = {Developments in X-Ray Tomography XV},
  editor = {Bert M{\"u}ller and Ge Wang},
  organization = {International Society for Optics and Photonics},
  publisher = {SPIE},
  pages = {131520W},
  keywords = {X-ray imaging, Computed tomography (CT), Simulation, 
    Digital twinning, Graphics processor unit (GPU) programming, 
    Image registration, Digitally reconstructed radiograph (DRR), 
    Machine learning},
  year = {2024},
  doi = {10.1117/12.3025315},
}