7th Conference on Industrial Computed Tomography, Leuven, Belgium (iCT 2017)

www.iCT2017.org

Improved local fiber orientation from µCT scans of fiber reinforced composites

Dascha Dobrovolskij1, Michael Godehardt1, Frieder Heieck2, Alexander Rack3, Katja Schladitz1, Simon Zabler4 1Fraunhofer-Institut für Techno- und Wirtschaftsmathematik, Kaiserslautern, Germany, e-mail:

dascha.dobrovolskij@itwm.fraunhofer.de, michael.godehardt@itwm.fraunhofer.de, katja.schladitz@itwm.fraunhofer.de 2IFB - Institut für Flugzeugbau, Universität Stuttgart, Germany, e-mail: frieder.heieck@ifb.uni-stuttgart.de

3European Synchrotron Radiation Facility, ID19, Grenoble, France, e-mail: alexander.rack@esrf.fr 4Simon Zabler, Fraunhofer-Entwicklungszentrum Röntgentechnik, Würzburg, Germany, e-mail:

simon.zabler@iis.fraunhofer.de

Abstract 2nd order orientation tensors as needed as input for CAD simulation programs can be obtained from micro computed tomography image data via local orientation analysis. There are several well established methods for calculating the fiber orientation in each voxel, based on the structure tensor, on the Hessian matrix of 2nd order gray value derivatives, on moments of inertia, and on filtering with anisotropic Gaussians. For 3D image data, the former two have been proven to be preferable both in terms of accuracy and computing time. Here, we show the applicability and usefulness of the method based on the Hessian matrix in two highly relevant application cases: Region of interest scans of a glass fiber reinforced composite components and synchrotron radiation µCT scans of braided carbon fiber rovings.

Keywords: glass fiber reinforced composites, carbon fiber rovings, image analysis

1 Orientation tensors from local fiber orientations 2nd order orientation tensors can be easily deduced from 3D image data once local directional information is available by averaging the pairwise products of the orientation vector components over a small subvolume. For estimating these local fiber orientation vectors, several methods have been applied to CT image data of glass fiber reinforced composites. In [1,2] the local fiber direction is found as the one yielding the strongest filter response when filtering with anisotropic Gaussian filters with a prolate filter mask. In [3] the ellipsoid of inertia is calculated from discrete distance transforms. Its major axis yields the local fiber direction. The structure tensor [4] exploits first order gray value derivatives (gradients) and finds the local fiber direction as the one corresponding to the smallest eigenvalue. That is, the local fiber direction is the one where the gradient is minimal. In other words, the fiber direction is the spatial direction in which the gray values change the least. Finally, the Hessian matrix of 2nd order gray value derivatives, describing the local curvature of the gray values can be used, too. The desired local fiber direction again corresponds to the smallest eigenvalue as this is the direction of least change of gray values. All four methods have been applied successfully. The quantitative evaluation based on synthetic image data in [5] revealed clear advantages of the latter two methods based on gray value derivatives in terms of precision and run-time over the anisotropic Gaussians and w.r.t. noise sensitivity over the moments of inertia method. All four compared methods rely on the fibers being bright compared to the background (matrix material), an assumption holding true for CT images of glass fibers in a polymer matrix (GFRP). However, in practical applications with densly packed, nearly parallel fibers it turns out to be favorable to take into account the direction information carried by the elongated gaps in between, too. In the case of the Hessian matrix based method applied here, this just requires to order the eigenvalues w.r.t. their absolute values thus covering the case of dark fibers on bright background, too [6].

2 Glass fiber reinforced composite components An important and recently available data acquisition method is the region of interest (ROI) CT scan of GFRP components. Due to security standards and the quality management of produced components it is an important task to analyze the fiber distribution and orientation in the component. ROI-CT allows the high resolution CT measurements needed to fulfill this task without the typical restriction on the sample size caused by the need to place the sample completely within the X-ray beam. Here we present the local fiber orientation analysis for a subregion of a GFRP brake pipe clip. A vizualisation of the results is shown in Figure 1, where the right side shows the difference of the local orientation w.r.t. to the measured global mean fiber direction.

7th Conference on Industrial Computed Tomography, Leuven, Belgium (iCT 2017)

www.iCT2017.org

Figure 1: Analysis results for a region of interest CT data set. Left: 3D visualization of segmented fiber system. Right: Deviation of local fiber direction from global mean fiber direction (blue: deviations up to 20°, yellow: deviations larger than 20°). Analysis and visualization

with Fraunhofer ITWM’s MAVI.

4 Carbon fiber rovings The second application case we report on are braided carbon fiber rovings. The analysis of carbon fiber roving structures is particularly important due to security relevant applications e.g. in aircraft industry. The carbon fibers within the rovings are densely packed. However, the use of synchrotron radiation yielding increased gray value contrast, extremely low noise level, and the opportunity to exploit additionally the in-line phase contrast, enables local orientation analysis. The results are precise enough to separate the rovings based on the gained diections. This separation can be used to analyze each roving type separately, e.g. w.r.t. undulation, see also [8].

Figure 2: Cross-sectional sllice views of the reconstructed SRµCT image data of braided carbon fiber rovings. Left: original. Center and

right: rovings separated by orientation.

Acknowledgements This work was partially funded by the Fraunhofer Society via MEF 3D Volant. We thank Andreas Senn, global Core & Application Engineering Manager, ITW Engineered Fasteners & Components for providing the brake pipe clip.

References [1] K. Robb, O. Wirjadi, K. Schladitz K, Fiber orientation estimation from 3D image data: Practical algorithms,

visualization, and interpretation. In: Proc. Int. Conf. Hybrid Intelligent Systems, 2007. [2] O. Wirjadi, K. Schladitz, A. Rack, T. Breuel T, Applications of anisotropic image filters for computing 2d and 3d-fiber

orientations. In: Proc. 10th Europ. Congr. Stereology and Image Analysis, 2009. [3] H. Altendorf , D. Jeulin D, 3d directional mathematical morphology for analysis of fiber orientations. Image Analysis

Stereology 28, 143–53, 2009. [4] M. Krause, J. Hausherr, B. Burgeth, C. Herrmann, W. Krenkel, Determination of the fibre orientation in composites

using the structure tensor and local X-ray transform. J Material Science, 45, 888–96, 2010. [5] O. Wirjadi, K. Schladitz, P. Easwaran, J. Ohser, Estimating fibre direction distributions of reinforced composites from

tomographic images, Image Analysis and Stereology, 2016, submitted [6] A.F. Frangi, W.J. Niessen, K.L. Vincken, M.A. Viergever. Multiscale vessel enhancement filtering. In Medical Image

Computing and Computer-Assisted Intervention - MICCAI'98, W.M. Wells, A. Colchester and S.L. Delp (Eds.), Lecture Notes in Computer Science 1496, 130-137, 1998.

[7] F. Hermann, Investigation of the 3D-fiber architecture of dry and impregnated carbon fiber textiles by means of computed tomography, BSC thesis, Institut für Flugzeugbau, Universität Stuttgart, 2016

[8] F. Heieck, P. Middendorf, Effect of the cover factor of 2D biaxial and triaxial braided carbon composites on their in-plane mechanical properties. In: Proc. 17th European Conference on Composite Materials, Munich, 2016.