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

Johan Friemann; Lars P. Mikkelsen; Carolyn Oddy; Martin Fagerström

doi:10.60691/yj56-np80

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

Johan Friemann^*, Lars P. Mikkelsen, Carolyn Oddy, Martin Fagerström

^*Corresponding author for this work

Chalmers University of Technology

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

29 Downloads (Pure)

Abstract

A novel pipeline for the generation of synthetic tomograms of woven composite materials, to be used for training of machine learning based segmentation algorithms is presented. The pipeline is completely based on open source software and heavily utilizes the graphical processing unit for fast data generation. The proposed method generates a surface mesh of the woven geometry, scans it, reconstructs the scan, and generates a voxel labeling of the generated tomogram. It is demonstrated that the method can generate images that show good agreement with experimentally produced x-ray computed tomography images of a 3D-woven carbon fiber reinforced polymer composite.

Original language	English
Title of host publication	Proceedings of the 21st European Conference on Composite Materials : Special sessions
Editors	Christophe Binetruy, Frédéric Jacquemin
Volume	8
Publication date	2024
Pages	333-338
ISBN (Electronic)	978-2-912985-01-9
DOIs	https://doi.org/10.60691/yj56-np80
Publication status	Published - 2024
Event	21^st European Conference on Composite Materials - Nantes, France Duration: 2 Jul 2024 → 5 Jul 2024

Conference

Conference	21^st European Conference on Composite Materials
Country/Territory	France
City	Nantes
Period	02/07/2024 → 05/07/2024

Keywords

X-ray computed tomography
Woven composites
Segmentation
Synthetic training data

Access to Document

10.60691/yj56-np80Licence: CC BY-NC

FulltextFinal published version, 502 KBLicence: CC BY-NC

OpenUrl availability

Full text

Cite this

@inproceedings{5954cf12f4244b4c99f42150e1f3929d,

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

abstract = "A novel pipeline for the generation of synthetic tomograms of woven composite materials, to be used for training of machine learning based segmentation algorithms is presented. The pipeline is completely based on open source software and heavily utilizes the graphical processing unit for fast data generation. The proposed method generates a surface mesh of the woven geometry, scans it, reconstructs the scan, and generates a voxel labeling of the generated tomogram. It is demonstrated that the method can generate images that show good agreement with experimentally produced x-ray computed tomography images of a 3D-woven carbon fiber reinforced polymer composite.",

keywords = "X-ray computed tomography, Woven composites, Segmentation, Synthetic training data",

author = "Johan Friemann and Mikkelsen, {Lars P.} and Carolyn Oddy and Martin Fagerstr{\"o}m",

year = "2024",

doi = "10.60691/yj56-np80",

language = "English",

volume = "8",

pages = "333--338",

editor = "Christophe Binetruy and Fr{\'e}d{\'e}ric Jacquemin",

booktitle = "Proceedings of the 21st European Conference on Composite Materials",

note = "21<sup>st</sup> European Conference on Composite Materials, ECCM21 ; Conference date: 02-07-2024 Through 05-07-2024",

}

Friemann, J, Mikkelsen, LP, Oddy, C & Fagerström, M 2024, Automated generation of labeled synthetic training data for machine learning based segmentation of 3D-woven composites. in C Binetruy & F Jacquemin (eds), Proceedings of the 21st European Conference on Composite Materials: Special sessions. vol. 8, pp. 333-338, 21^st European Conference on Composite Materials, Nantes, France, 02/07/2024. https://doi.org/10.60691/yj56-np80

Automated generation of labeled synthetic training data for machine learning based segmentation of 3D-woven composites. / Friemann, Johan; Mikkelsen, Lars P.; Oddy, Carolyn et al.
Proceedings of the 21st European Conference on Composite Materials: Special sessions. ed. / Christophe Binetruy; Frédéric Jacquemin. Vol. 8 2024. p. 333-338.

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

TY - GEN

T1 - Automated generation of labeled synthetic training data for machine learning based segmentation of 3D-woven composites

AU - Friemann, Johan

AU - Mikkelsen, Lars P.

AU - Oddy, Carolyn

AU - Fagerström, Martin

PY - 2024

Y1 - 2024

N2 - A novel pipeline for the generation of synthetic tomograms of woven composite materials, to be used for training of machine learning based segmentation algorithms is presented. The pipeline is completely based on open source software and heavily utilizes the graphical processing unit for fast data generation. The proposed method generates a surface mesh of the woven geometry, scans it, reconstructs the scan, and generates a voxel labeling of the generated tomogram. It is demonstrated that the method can generate images that show good agreement with experimentally produced x-ray computed tomography images of a 3D-woven carbon fiber reinforced polymer composite.

AB - A novel pipeline for the generation of synthetic tomograms of woven composite materials, to be used for training of machine learning based segmentation algorithms is presented. The pipeline is completely based on open source software and heavily utilizes the graphical processing unit for fast data generation. The proposed method generates a surface mesh of the woven geometry, scans it, reconstructs the scan, and generates a voxel labeling of the generated tomogram. It is demonstrated that the method can generate images that show good agreement with experimentally produced x-ray computed tomography images of a 3D-woven carbon fiber reinforced polymer composite.

KW - X-ray computed tomography

KW - Woven composites

KW - Segmentation

KW - Synthetic training data

U2 - 10.60691/yj56-np80

DO - 10.60691/yj56-np80

M3 - Article in proceedings

VL - 8

SP - 333

EP - 338

BT - Proceedings of the 21st European Conference on Composite Materials

A2 - Binetruy, Christophe

A2 - Jacquemin, Frédéric

T2 - 21<sup>st</sup> European Conference on Composite Materials

Y2 - 2 July 2024 through 5 July 2024

ER -

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

Abstract

Conference

Keywords

Access to Document

OpenUrl availability

Fingerprint

Cite this