Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Fiber bundle tracking method to analyze sheet molding compound microstructure based on computed tomography images

: Schöttl, L.; Weidenmann, K.A.; Sabiston, T.; Inal, K.; Elsner, P.


NDT & E International 117 (2021), Art. 102370, 8 S.
ISSN: 0963-8695
Fraunhofer ICT ()

Discontinuous fiber reinforced polymers (DicoFRP) like Sheet Molding Compounds (SMC) are frequently applied in modern lightweight designs, due to their good formability and mechanical properties at low density. The DicoFRP microstructure affects the mechanical properties and has to be taken into account in the design process. X-ray Micro-Computed Tomography (μCT) systems acquire volumetric images of microstructures in a non-destructive way. The fiber orientation is determined by using state-of-the-art image processing tools. The resolution of volumetric images and the specimen size are directly coupled due to the cone-beam μCT geometry. In order to identify all individual fibers, high resolution images are needed and consequently, only small specimen volumes are scanned. The present contribution makes use of the property that fibers are arranged as bundles within SMC. Consequently, fiber bundles instead of individual fibers are used to analyze microstructures in order to overcome the μCT-cone-beam-related conflict between sample size and image resolution. The fiber bundles are determined by means of orientation data and an introduced tracking method, facing the challenge of crossing fiber bundles. Subsequently, the tracked fiber bundles, which are related to the same mesoscopic fiber bundle are merged by using a hierarchical agglomerative clustering method. The presented method is applied to a typical SMC microstructure. This contribution introduces an image processing method for analyzing SMC microstructures based on fiber bundles, opening up the possibility to investigate large specimen volumes. This enables to characterize representative SMC microstructures and utilize the data for art modeling approaches.