Defect detection in carbon fiber-reinforced composites using directional dark-field imaging and tomography

This paper proposes the use of circular X-ray grating interferometry as an effective technique for defect detection with potential applications for in-line inspection of carbon fiber-reinforced pultruded profiles used inside the load-carrying spar caps of wind turbine blades. A fuzzball defect in the pultruded profile is characterized as a demonstration. The method allows for large field-of-view quantification of local fiber alignment and relative fiber volume fraction. A two-dimensional through the thickness averaged distribution of the fiber orientation, the mean scattering, and fractional anisotropy are determined. Based on this, it is possible to determine the size of the defect as well as quantify the severity of the defect.


Introduction
Renewable energy technology, such as wind turbines, plays a critical role in reducing greenhouse gas emissions and combating climate change [1].The efficiency of wind turbines depends largely on the design of their components, particularly the blades, which convert the kinetic energy of wind into mechanical energy [2].One important component of wind turbine blades is the spar cap, which provides structural support and stiffness to the blade.Carbon fiber reinforced polymer (CFRP) is a widely-used material for manufacturing spar caps due to its high strengthto-weight ratio and fatigue resistance.
A new generation of wind turbine blades use spar caps manufactured from carbon fiberreinforced pultruded profiles.These profiles are manufactured with fixed geometry and used as building blocks in the spar cap design.The pultrusion process involves a continuous resin impregnation of carbon fibers which is subsequently pulled through a die, resulting in a composite material with a tunable cross-sectional shape [3].
The orientation of the carbon fibers within the pultruded profile significantly affects the mechanical performance [4].In particular, the alignment of the fibers with respect to the loading direction can greatly influence the strength and stiffness, as well as its fatigue resistance.Therefore, it is crucial to carefully control the fiber orientation during the pultrusion process to achieve the desired mechanical properties.This requires precise control of the fiber placement and orientation within the die, as well as optimization of the resin system to ensure complete impregnation of the fibers.
Common manufacturing defects in CFRPs include voids, porosities, and inconsistent fiber placement, which results in non-uniform mechanical properties.The pultrusion process is ideal for mitigating the occurrence and severity of these types of defects and thus yields more consistent properties.A more random and localized manufacturing defect in pultruded profiles includes fuzzballs, which are an accumulation of broken fibers in the carbon fiber yarn.The broken fibers form a small ball that is trapped in between the yarns and subsequently fed into the pultrusion die.Fuzzball defects can act as a foreign inclusion and thus affect the alignment of the surrounding fibers, which can have a detrimental effect on mechanical strength and, ultimately, the structural integrity of the wind turbine blade.While quality control measures are employed during the pultrusion process to minimize the occurrence of defects, their complete elimination remains a challenge.The use of advanced monitoring and inspection techniques is necessary to identify defects.
To characterize fiber misalignment in CFRP pultruded profiles, various techniques have been developed, including X-ray radiography, computed tomography (CT), ultrasonic testing, and optical microscopy [5].Ultrasonic testing can be used to determine the orientation of fibers and the presence of delamination, but its accuracy is limited by the size of the fibers and the presence of other defects.Optical microscopy can provide high-resolution images of the composite material, but it is limited to surface analysis and cannot detect defects embedded within the material without preparation of a cross-section of the specimen.X-ray radiography and CT provide a non-destructive way to visualize the internal structure of composite materials and identify defects such as voids, resin-rich areas, and fiber misalignment.Conventional characterization of Fiber-Reinforced Composites (FRC)s using X-rays is performed in the socalled absorption mode.This approach requires sufficient contrast to differentiate between the composite matrix and reinforcement materials as well as direct resolution of individual microfibers, which drastically limits the imaged field-of-view (FOV).Radiography images and tomographic reconstructions of composite materials are usually followed by either segmentation of individual fibers or statistical analysis of the fiber orientations [6,7].In the meantime, X-ray dark-field or scattering mode enables accessing anisotropic microstructure due to the local smallangle X-ray scattering.Scattering from FRCs is not limited by the resolution of the imaging system as it occurs on multiple length-scales ranging from nanostructures up to individual fibers and fiber bundles.
Grating interferometry is one of the effective techniques to measure the dark-field signal from FRCs.Design of the interferometer can be chosen to optimize the sensitivity to a certain length scale without restricting the overall FOV.With a conventional interferometer system for compact X-ray systems consisting of a set of linear gratings, the same sample area is measured multiple times in order to retrieve the dark-field signal.Moreover, due to the dependence of the signal on the relative orientation of the grating lines to the alignment of the microstructure, the optical elements are usually rotated in plane to provide orientation sensitivity [8].A novel design of the grating comprised of an array of circular unit cells allows single-shot measurement capabilities for omni-directional orientation sensitivity, significantly relaxing the instrumental complexity and total measurement time per image.This is particularly crucial for in-line inspection applications as well as for 3D tomographic fiber characterization that is obtained from a collection of 2D images [9,10,11,12].
Here we present an application of directional dark-field imaging using circular gratings on industrial samples of 5 mm thick pultruded carbon fiber profile based on 7 µm thick carbon fibers used in the load carrying pultruded profiles in large wind turbine blades.We show the potential of this method for rapid and robust in-line inspection for sub-millimeter scale characterization of fiber misalignment and realtive fiber density in CFRP pultruded profiles over a wide area.

Sample
The current study is based on a carbon fiber-reinforced polymer matrix pultruded profile, where a fuzzball defect has purposely been introduced in the pultrusion process line with the intention of studying the effect on mechanical properties.The fuzzball defect is manufactured by adding a ball of accumulated broken carbon fibers in between the carbon yarns in the pultrusion line.The fuzzball forces the surrounding carbon fibers to compact and misalign relative to the global fiber direction, ultimately changing the local fiber-matrix structure (Fig. 1.a and b).Fuzzballs can, in reality, occur in any position in the pultruded profile but have been placed near the surface in order to visually inspect the defect characteristics (Fig. 1.c and d).The fuzzball defect presented in this work is approximately 16 mm in diameter prior to manufacturing the pultruded profile.In reality, fuzzball defects can have any size and are only limited by the dimensions of the pultrusion die.Smaller fuzzballs may be hidden, while larger fuzzballs are typically visible on the surface.

Imaging setup
A schematic overview of the measurement setup at Xnovo Technology is shown in Fig. 2. A fiber-reinforced polymer component is positioned between the X-ray source and detector, and a circular grating optics is placed at a specific location depending on the desired image resolution.The probed FOV, length scale, and resolution of the setup can be adjusted accordingly.The unit cell size of the optical element determines the smallest pixel size of the reconstructed image, which is typically few orders of magnitude larger than the fiber diameter.For the experiment, a microfocus X-ray source was operated at 40 kV and 20 W. A large FOV CMOS detector with 50 µm pixel size was used to capture X-ray images.A circular grating optics prototype with 222.75 µm unit cell size was placed equidistantly between the source and detector providing a 6 × 4 cm 2 effective FOV.The sample was raster scanned (via single scan line) in the proximity of the grating via step-scan mode with 10 seconds per single FOV.To extract the contribution of the scattering signal from fibers, a reference image was captured at the beginning of the scan, followed by a series of images of the sample at different positions of the scan line.For comparison, a typical absorption contrast tomography of a CFRP sample with 1 µm pixel (voxel) size over 2 × 2 × 2 mm 3 volume takes between 10-20 hours of scanning time [6,12].

Signal processing
A detailed description of the directional scattering signal retrieval algorithm is presented in [9,10].For each unit cell in the grating, the algorithm produces a 2D scattering distribution within angular bins uniformly spread over 180 degrees.The scattering distribution can be subsequently redefined by three eigenvectors v 1 , v 2 , v 3 and the corresponding eigenvalues λ 1 , λ 2 , λ 3 via principal component analysis.The eigenvector with the lowest eigenvalue is considered as the preferential local fiber orientation, because the scattering is weakest along the fiber orientation.The mean scattering (MS) can provide an estimate of the relative local density of fibers when defined as where max(λ 1 + λ 2 + λ 3 ) indicate the volumetric maximum of the sum of the eigenvalues.Fractional anisotropy (FA) is a metric commonly used in diffusion tensor imaging to quantify the degree of directionality, or anisotropy, of water diffusion within tissues, including those with fiber structures such as white matter in the brain.FA reflects the organization and coherence of the fibers, with higher FA values indicating greater directionality and alignment of the fibers.Thus, the more directional the fiber structures, the higher the FA value.The FA value is computed as [13]: 1293 (2023) 012016 IOP Publishing doi:10.1088/1757-899X/1293/1/012016 where λ = (λ 1 + λ 2 + λ 3 )/3.Areas with a high volume fraction of well-aligned fibers will exhibit both high FA and high MS, whereas regions with a high volume fraction of randomly distributed fibers will have low FA but high MS.In the case of unidirectional CFRP pultruded profiles, local changes in MS and FA values are useful for indicating defects that alter the density and orientation of the fibers, respectively.

Results
The measured omni-directional fiber orientation data is visualized using a color wheel, where each color corresponds to an orientation within a half-circle, and the brightness represents the degree of anisotropy.The overview of the fiber orientation (see Fig. 3.a) reveals predominant unidirectional alignment across the majority of the pultruded profile.However, a distinct region is observed where the local fiber orientation deviates from the main orientation, indicating the presence of a fuzzball defect that extends approximately 20 mm along the pulling direction.To illustrate the impact of this defect on fiber misalignment, an angular deviation map is presented (see Fig. 3.b), showcasing the increased variation in fiber orientation around the defect.The mean value of fiber misalignment within this region (see Fig. 3.c) is close to zero indicating a symmetrical shape of the defect, while the standard deviation of fiber misalignment is larger than in the non-defect region.
In a conventional absorption contrast (Fig. 4.a), the fuzzball defect doesn't exhibit sufficient contrast relative to the rest of the profile due to the small difference in material density of fibers and resin.However, the map of MS (Fig. 4.b) provides more insights into the effect of the fuzzball on local fiber density.Notably, within the affected region, the MS values are elevated compared to the rest of the pultruded profile, suggesting an increased local fiber density.Simultaneously, the FA in the same area (Fig. 4.c) is lower, indicating a more randomly distributed local fiber orientation.Orthogonal line profiles across the defect area (Fig. 4.d-e) reveal dimensions of the defect for different data modalities.Both the mean scattering and fractional anisotropy exhibit higher definition of the defected area via line profiles than conventional absorption contrast.This can ultimately lead to more reliable metrological studies of fuzzball defects.To demonstrate the potential of using circular-grating based optics for in-line inspection of CFRP pultruded profiles, a fuzzball defect segmentation from mean scattering is shown in Fig. 5. Here, the segmented defect is overlaid with the greyscale mean scattering map.

Conclusion
The fiber orientation in carbon fiber-reinforced pultruded profiles plays a critical role in determining their mechanical properties and overall performance in wind turbine blades.Various factors during the manufacturing process, such as resin cure rate, temperature, and pulling speed, can affect fiber orientation and ultimately influence the performance of the composite material.Optimization of the pultrusion process and post-processing techniques can result in improved mechanical properties and ultimately lead to more efficient wind turbines, contributing to a more sustainable future.Detecting and minimizing defects that cause fiber misalignment, is also crucial for ensuring optimal performance of the spar cap.This paper demonstrates that advanced techniques such as dark-field X-ray microscopy using circular gratings yield promising results in rapid and reliable detection of defects and characterization of fiber misalignment.
The findings of this study have implications for the design and optimization of wind turbine blades.The fast and accurate characterization of fiber orientation enables the development of more precise numerical models and simulations, which can aid in predicting the mechanical behavior of the spar caps under different loading conditions.This, in turn, facilitates the design of blades with improved structural integrity, enhanced energy capture, and increased overall efficiency.It is important to note that while this study is focused on carbon fiber-reinforced pultruded profiles, similar principles and considerations apply to other manufacturing methods and composite structures.The role of fiber orientation in determining the performance of composites is a fundamental aspect across various industries, including aerospace, automotive, and marine applications.

Figure 1 .
Figure 1.Carbon fiber-reinforced pultruded profile with a fuzzball defect at the surface.a) Microscopy image of fuzzball defect cross-section where carbon fibers appear white and the resin dark gray.Resin rich areas within the defect are black.b) Zoom of microscopy image with clear out-of-plane fiber misalignment w.r.t. the global fiber direction.c) Top view of 10 cm wide sample cut aimed for mechanical testing.d) Zoom of fuzzball defect with annotated defect characteristics.

Figure 2 .
Figure 2. Schematic representation of the experimental setup for omni-directional dark-field imaging.The X-rays from the source illuminate the grating consisted of an array of circular unit cells.The sample is scanned at different positions along the scanning axis shown as black arrow.Reference grating fringes are measured by the detector prior to the placement of the sample under X-rays.

Figure 3 .
Figure 3. a) omni-directional fiber orientation, b) fiber orientation misalignment wrt pulling direction, c) histogram of fiber misalignment within boxes indicated in (b): around the fuzzball defect and non-defected area as red and blue boxes, respectively .

Figure 4 .
Figure 4. Maps of absorption (a), mean scattering (b) and fractional anisotropy (c).Horizontal (d) and vertical (e) line profiles indicated in (a, b, c) as white and black dashed lines, respectively.

Figure 5 .
Figure 5. Example of defect segmentation based on dark-field image data: a) greyscale image of mean scattering, b) overlaid mask of the defect