- Email: [email protected]
Micro-CT analysis of internal geometry of chopped carbon fiber tapes reinforced thermoplastics
Composites: Part A 91 (2016) 211–221
Contents lists available at ScienceDirect
Composites: Part A journal homepage: www.elsevier.com/locate/compositesa
Micro-CT analysis of internal geometry of chopped carbon fiber tapes reinforced thermoplastics Yi Wan a,⇑, Ilya Straumit b, Jun Takahashi a, Stepan V. Lomov b a b
The University of Tokyo, Department of Systems Innovation, Hongo 7-3-1, Bunkyo-ku, Tokyo, Japan KU Leuven, Department of Materials Engineering, Kasteelpark Arenberg 44, B-3001 Leuven, Belgium
a r t i c l e
i n f o
Article history: Received 25 April 2016 Received in revised form 15 September 2016 Accepted 8 October 2016 Available online 11 October 2016 Keywords: A. Polymer-matrix composites (PMCs) B. Microstructures D. CT analysis D. Microstructural analysis
a b s t r a c t Adequate characterization of the internal geometry of random fiber composites for prediction of their properties requires acquisition and analysis of vast amount of data. The present work proposes a method to measure and quantify the internal geometry of ultra-thin chopped carbon fiber tape reinforced thermoplastics (UT-CTT), a kind of randomly oriented strands, using the X-ray micro-CT. The two-scale fiber orientation distributions (orientation of the tapes and of the fibers inside the tapes), tape morphology, 3D structure and orientation misalignment were characterized using a specialized image processing methodology. The results indicate that the applied method can provide detailed information like primary orientation directions, tape waviness and splitting of the internal geometry of UT-CTT. UT-CTTs produced with different molding pressure were studied to confirm the capability of the methodology in distinguishing subtle structure changes caused by different molding conditions. Ó 2016 Elsevier Ltd. All rights reserved.
1. Introduction In recent years a significant progress has been made in the development of carbon fiber reinforced thermoplastics (CFRTP). Compared to thermoset polymers, thermoplastics exhibit superior cycle molding time and good in-plant recyclability. Randomly oriented short fiber systems generally show high capability of being manufactured in complex geometries without internal structural damage. Combined with their low manufacturing cost, random carbon fiber reinforced thermoplastics (rCFRTP) are regarded as potential substitutes for metallic materials in mass-production applications [1]. Advantageous thin ply effects of prepreg CF tapes on laminated composites mechanical performance have been verified by Sihn et al. [2] and Amacher et al. [3]. Composites molded by thin prepreg CF sheets show superior qualities in suppression of microcracking, delamination, and splitting damage for static, fatigue, and impact loadings. Randomly oriented strands (ROS) are one of the representative rCFRTP with high Vf, which makes possible achieving high mechanical properties [4–7]. X-ray micro computed tomography (micro-CT) has become a powerful non-destructive method for characterizing and visualizing the internal geometry of composite materials. With the ⇑ Corresponding author. E-mail address: [email protected] (Y. Wan). http://dx.doi.org/10.1016/j.compositesa.2016.10.013 1359-835X/Ó 2016 Elsevier Ltd. All rights reserved.
developments in digital image acquisition and image processing ability, the quantification of structural features, 3D model rebuilding together with the finite element (FE) mesh generation are considered to be the trends of current X-ray micro-CT research. rCFRTP generally possess complex internal geometry due to the interaction of the tape chunks during manufacturing. X-ray micro-CT provides a higher accuracy in the characterization of fiber orientation distribution (FOD), fiber length distribution (FLD), and multi-scale geometry quantification, compared with 2D image scanning method. The FOD and FLD of short glass fiber-reinforced phenolic foam with 2.7% fiber volume fraction (Vf) was characterized using Xray micro-CT method by Shen et al. [8]. FOD, FLD and 3D models of metal fiber networks and metal fiber foam were evaluated in [9,10]. X-ray micro-CT studies of short fiber composites are mainly concentrated on injection molded materials [11–16]. Nguyen et al. [14] used X-ray micro-CT and FE flow simulation to investigate the fiber orientation and flowability of injection molded glass fibers (GF) reinforced polyamide 6 (PA6). Sun et al. [15] studied the local fiber orientation and the effect of injection flow in short carbon fibers (CF) reinforced polypropylene (PP). Bernasconi et al. [12] calculated the local average orientation distribution and visualized 3D orientation models of GF/PA6 injection materials using the Mean Intercept Length (MIL) technique. The fiber morphology like the distributions of fiber length and fiber width were analyzed by Alemdar et al. [11]. The comparison between X-ray micro-CT and
212
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
2D image scanning method of short-GF/PA6 [16] and lignocellulosic fibers [17] composites were also studied. The FODs and 3D rebuilt models of short carbon-carbon felts [18] and wood fibers (WF) reinforced high-density polyethylene (PE) [19] were investigated. The combination of the reconstructed 3D model with the FE mesh generation and numerical analysis have been reported by several research groups [13,20,21]. The referenced studies did not consider ROS composites. The internal geometry of these materials differs significantly from the one of injection molded composites. In the ROS composites fiber orientation distribution has two-scales: orientation of the tapes and (mis)orientation of fibers within the tapes. This creates a situation similar to the characterization of fiber orientations in textile composites, where one has to distinguish between variability of the yarn directions and misalignment of fibers within the yarns [22]. In the present study, an image processing method based on structure tensor [23] is applied to determine the internal geometry of one kind of ROS made by water dispersed thin tapes, the ultrathin chopped carbon fiber tape reinforced thermoplastics (UT-CTT). The UT-CTT are developed in the framework of a Japan national project and considered a potential material for mass-production applications [1]. The proposed methods for reconstruction of multi-scale orientation distributions, tape morphology and 3D material models are verified for UT-CTT manufactured with two different molding conditions. The paper aims at detailed presentation of the image processing methodology, using a limited number of specimens. In future publications the methodology will be applied to material characterization with a necessary statistical robustness.
2. Materials and methods 2.1. Materials and specimens The UT-CTT used in this study are composed of randomly oriented UT-tape (oriented ultra-thin unidirectional prepreg tape). Spread carbon fiber tow (TR 50S, Mitsubishi Rayon Co., Ltd.) and Polyamid-6 (PA6, DIAMIRONTM C, Mitsubishi Plastics, Inc.) are used
to produce this tape. The fiber volume fraction (Vf) of UT-CTT is 55% in average. The UT-tape is considerably thin (44 lm in average) compared with the conventional one (about 150 lm or more), so it is called ‘ultra-thin’ (Fig. 1). The UT-tapes used in this study are provided by the Industrial Technology Center of Fukui Prefecture in Japan and are cut to 5 mm width and 6 mm length. To ensure a better tape distribution properties and to preserve the tape structure after molding, the intermediate UT-CTT sheets were manufactured by wet-type paper making process (Fig. 2). In the process tapes are put into a water container. The bottom of the container can be opened and has a filter. After the tapes are put into the container and randomly dispersed in the water, open the bottom of the container and filter out water to make the UT-CTT sheets. The UT-CTT sheets are stacked to manufacture UT-CTT plates with 2 mm thickness using compression molding. To identify the capability of the micro-CT method to recognize an internal geometry change, two different molding conditions are applied in this study, labeled as M10 (10 MPa molding pressure) and M3 (3 MPa molding pressure) (Fig. 3). After the UT-CTT plates are made, the specimens for X-ray measurement are prepared. To ensure the sufficient resolution of the X-ray images and provide reliable information on the internal geometry and fibers, the specimens must be cut in relatively small size. After each edge of the plates was cut off by 15 mm to eliminate the molding edge effect, the specimens with a size of 2 2 30 mm (thickness width length) were cut from both M10 and M3 plates arbitrarily by thin section cutting machine (EXAKT Advanced Technologies GmbH 310CP) for this study. 2.2. X-ray micro-CT scanning The internal structural information of UT-CTT was observed and collected by the 3D X-ray scan system TDM1000-II from Yamato Scientific Co., Ltd. During the observation, the specimens were fixed on a rotational stage (Fig. 4), and the distance between the rotation axis and the radiation source is set to 10 mm for reliable resolution of embedded fibers on the images. The scanned volume is a right circular cylinder because of the rotation, and the size of the cylinder is 0.88 mm in radius and 1.76 mm in height. To ensure sufficient image resolution during CT acquisition process, the
Fig. 1. Cross section of a conventional tape (a), UT-tape (oriented ultra-thin unidirectional prepreg tapes) (b) and surface of UT-CTT (ultra-thin chopped carbon fiber tape reinforced thermoplastics). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
213
Fig. 2. Wet-type paper making process for manufacturing UT-CTT sheets. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 3. Molding conditions M3 and M10. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
sample was positioned in the scanner’s field of view so that the imaged volume had a physical size of 1 1 1 mm and excluded the air surrounding the sample. The X-ray tube voltage was set to 40 kV and X-ray tube current was set to 40 lA for all the specimens. After the acquisition of the X-ray projection images (a rotation step of 0.24 degree and 25 min per full rotation of the sample), the 3D image was reconstructed by the image processing unit of the X-ray CT system. The pixel size of the reconstructed 3D micro-CT images was fixed to 4 lm, and 512 images stacked through out-of-plane direction for each specimen were used for image processing and calculations. The schematic of scanning process and scanned volume are illustrated in Fig. 5.
2.3. Image processing and data analysis The internal geometry of the material was reconstructed from the CT image using a regular rectangular (voxel) mesh, where each element of the mesh contains a set of variables [23]. The CT image is converted from the native scanner format to a single threedimensional array I(x1, x2, x3) of 8-bit gray values, where the coordinates x1, x2, x3 are integer numbers. The analyzed variables are fiber orientation angles u and h in spherical coordinate system. u is in-plane orientation axis, with the Cartesian axes x1 and x2 lying in the plane of the sheet, and h is an angle of the vector with the axis x3, normal to x1 and x2 axes. The fiber orientation vector is
214
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
The integration is done over a local volume V of the CT image. The integration volume V has a cubical shape with a size of 17 17 17 pixels. The eigenvector e1 corresponding to the smallest eigenvalue k1 is the fiber orientation vector. The spherical angles were calculated from the eigenvector e1 (unit vector):
h ¼ arccosðe1;3 Þ
u ¼ arctan
e1;2 e1;1
ð4Þ ð5Þ
The voxel mesh therefore represents the two scalar fields h(x1, x2, x3) and u(x1, x2, x3) defined over the spatial domain of the material sample, represented in the CT image. For the purposes of present study, a special type of two-dimensional histograms was introduced, which includes the analyzed variable as one of the axes, and one of the spatial dimensions (x1, x2, x3) as the second axis of the histogram. A bin of this histogram shows frequency of occurrence of orientations in the range of the bin over a certain cross-section plane in the material. This type of histograms (further referred to as ‘‘unfolded” histogram) allows to represent the change of the distribution of a variable along a spatial coordinate. In this study it was applied to analyze through-thickness distribution of fiber orientations. The calculations were performed using a C# code VoxTex, developed in KU Leuven [23]. Inside the code, two-dimensional histograms were rendered using Root Data Analysis Framework v 5.34 (CERN). Fig. 4. Specimen on the rotational stage. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
calculated from the CT image I(x1, x2, x3) as a solution of the following eigenvalue problem:
Sei ¼ ki ei
ð1Þ
where S is the three-dimensional gray value structure tensor:
Z
Sij ¼ V
@I @I dv @xi @xj
ð2Þ
The derivatives in Eq. (2) were calculated using the 5-point central difference formula:
@I 1 ½Iðx 2; yÞ 8Iðx 1; yÞ þ 8Iðx þ 1; yÞ Iðx þ 2; yÞ @x 12 and similarly for the derivative by y coordinate.
ð3Þ
3. Results and discussions In the present study the internal geometry of UT-CTT were observed and analyzed by 3D X-ray micro-CT. During the step of voxel model construction in the image processing, the distance between voxels for averaging calculation, which defines a volume for averaging, called below ‘‘volume of interest” (VOI), needs to be determined related to the thickness direction of the image UT-CTT. The VOI’s throughthickness dimension is set to 11 pixels, i.e., 44 lm to ensure that the thickness of each VOI is close to the thickness of a single tape (44 lm in average). Other two dimension of VOIs are equal to the extents of the image. The VOIs are numbered according to their order in position along the thickness direction: VOI 1 to VOI 25. All volumes of interest have a hexahedral shape.
Fig. 5. Schematic of scanning process and scanned volume of the UT-CTT. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
215
Fig. 7. 3D micro model with visualized in-plane fiber orientation distribution of M10 UT-CTT, the Phi_XY (uXY) indicate in-plane orientation angle. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 6. Reconstructed 3D model of M10 UT-CTT with layered structure, concentric circles (ring artefacts) are observed at the side face. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
After 3D CT images were collected by the 3D X-ray scanner, the micro-structure of UT-CTT was reconstructed as illustrated in Fig. 6. The micro-structure model shows that the UT-CTT hold both features of laminate materials and short fiber composites. The model exhibit a layered structure, also the individual tapes in the UT-CTT show not-perfectly-flat orientations. Concentric circles (ring artefacts) can be observed at the side face of the model (see Fig. 6). Ring artefacts are usually suppressed at the stage of micro-CT acquisition by calibrating the detector; or during the reconstruction of the image, if the reconstruction software has the function of ring artefacts correction. However, they often cannot be eliminated completely and, if present after the reconstruction, cannot be corrected in the already reconstructed image. The presence of the ring artefacts is the case in the images acquired in the present work. Being unable to eliminate the artefacts we, as it is discussed in the subsequent sections, have ignored the spurious features caused by them in the final results.
ures hereinafter indicate the data density of the orientation. The orientation distribution shows special features. The angle hXY is concentrated around 90 degree with small dispersion, which means the UT-CTT is almost in-plane oriented. Several clusters of uXY are observed in the figure, and these clusters show irregularity in the in-plane angle. Because the fibers of UT-CTT are almost inplane oriented, so the distributions of uXY and hXY are ‘‘unfolded” as described in Section 2.3 – through the out-of-plane direction to analyze their change through the thickness. Fig. 9 shows the unfolded distribution of uXY (a) and hXY (b) through z axis. The unfolded uXY distribution indicates that the concentrated clusters of uXY in Fig. 8 do not mean local concentration of tapes in UTCTT. For example, the uXY distribution cluster from 20 degree to 80 degree in Fig. 8 is actually combined by three small clusters located at different z-positions (Fig. 9). On the other hand, although the thickness of VOI is set close to the thickness of a single tape, the orientations of layers are still not independent. Fig. 9(a) shows that small clusters still exist in the model. The size of the clusters are generally 3–4 VOI, also a larger cluster (VOI 15 to 22, 30–70 degree) is found. Because the tapes in UT-CTT are almost inplane oriented (Fig. 9(b)), this result means some tapes with same orientation may stick together during the wet-type paper-making process, or the tape waviness and tape splitting occurred locally.
3.1. Internal geometry analysis Using the stacked micro-CT images of UT-CTT, the visualized 3D model of the material and histograms of fiber orientation distribution were calculated. In this section the results for M10 UT-CTT are used to demonstrate the analysis methodology. The visualized 3D micro model with in-plane fiber orientation of M10 UT-CTT is shown in Fig. 7. The angles are given in the global Cartesian coordinate system (X, Y, Z) identical to the Cartesian system (x1, x2, x3) introduced previously. The color in the model (as well in Fig. 11 (b)) indicated local in-plane fiber orientation angle uXY (Phi_XY), with degree (°) as the unit in this study. Orientation changes layer by layer, but the not-fully-flat in-plane orientation and waviness through out-of-plane direction (Z axis) indicated the UT-CTT also have orientation irregularity as it is the case with conventional rCFRTPs. The two-dimensional histogram where the in-plane (uXY) and out-of-plane (hXY, Theta_XY) fiber orientation distribution angle are combined is shown in Fig. 8, the color bar in this figure and fig-
Fig. 8. Two-dimensional histogram combined the in-plane (uXY) and out-of-plane (hXY) fiber orientation distribution in the total volume of the M10 UT-CTT specimen. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
216
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
Fig. 9. Unfolded histograms of uXY (a) and hXY (b) of M10 UT-CTT. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 10. Unfolded uXY histogram and the corresponding clusters in visualized 3D model of M10 UT-CTT. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
In addition, Fig. 9(a) indicates that the uXY is not randomly distributed, which is because the reconstructed 3D model of UT-CTT is smaller than the tapes due to the limited volume that can be imaged. The relatively small dimensions of the sample cannot ensure the statistical randomness of orientation distribution because of the size effect of the tapes. hXY shows lager scatter around the center layer of the model (Fig. 9(b)), this is considered to be caused not by the material feature but by the ring artefacts mentioned in the beginning of this section, which increase the isotropy of the calculated angles. Therefore the scatter in VOI 13 will be discarded from the analysis of the results. To study the clusters of uXY in detail, the 3D model with uXY distribution (Fig. 7) is combined with the unfolded uXY histogram through z axis (Fig. 9(a)). The subsets of the 3D model separated by the VOI were extracted following the clusters that appeared in the unfolded uXY histogram, which means that the 3D morphologies of the orientation concentration areas (the clusters in Fig. 9 (a)) can be specified and extracted from the general 3D model (Fig. 7). The VOI 4–7, 9–11 and 16–21 are shown in Fig. 10. After the subsets were extracted, the threshold of uXY is applied on the model to identify the fiber distributions in concentrated uXY and the threshold ranges are also illustrated in Fig. 10. The extracted 3D models with threshold of uXY and the corresponding areas in unfolded uXY histogram proved the assumption that the tapes can interact during the wet process and disturb the ideal uniformly random orientation distribution, which would exist if the placements of the tapes were independent. The existence of the uXY distribution clusters through z axis is due to the tapes with the same orientation sticking together during the wet-type paper making process, as well as tape waviness and tape splitting taking place locally during the compression molding process. The 3D model of VOI 4–7 demonstrates 4 different layers (tapes) with the same orientation distribution pattern. In contrast, the 3D model of VOI 9– 11 shows an integral part of tape with some scattered areas which
217
are considered to be the tape waviness and splitting. On the other hand, the 3D model of VOI 16–21 exhibits both structural features: tapes are stuck together at the top of the model, while scattered areas are observed on the bottom side. The combination of the 3D model with the unfolded uXY histogram shows a high capability for the quantitative internal geometry study of the UT-CTT with complex structural features. By extracting the subset models, the detail structures and tape positions are reconstructed visually. For the further research of the micro-structure and combination with simulations, the unfolded histograms were quantified. The average values and standard deviations (SD) of both uXY and hXY were calculated by each VOI (Fig. 11). The quantified orientation distribution of uXY reproduces the visualized histogram to a certain extent. The clusters in VOI 16–21 observed in the histograms are also shown as series of data with SD indicated by error bars. In contrast, the results of VOI 4, 6, 7 and 11 show significant mismatch between the peak values in Fig. 9(a) and the average value in Fig. 11(a). This is caused by the multimodality of the orientation distributions in these VOIs. The mismatch is caused by the multiple orientation concentrations in one VOI and the SD is significantly high even though the histogram show apparent orientation concentrations. On the other hand, when there is a single aligned integral part of tape in the VOI, the average value is perfectly matched with the peak in histogram, and the SD is small as the VOI 10. In addition, the average value and SD of hXY show concentrations same as the observed in the unfolded histogram. The increase of SD around the center layer is caused by the ring artefacts as discussed before. This section introduced a methodology for the internal geometry analysis of UT-CTT by micro-CT. It provides a convenient approach for quantified visualization modeling of the micro-3D structure with orientation information, which can give support for composite simulation modeling and numerical analysis. The assessable features were summarized and listed in Table 1.
Fig. 11. Average values and standard deviations (SD) of uXY (a) and hXY (b) of M10 UT-CTT.
218
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
Table 1 The assessable features of internal geometry of UT-CTT using micro-CT. Features
Required data
Example characterization
Planarity of layers
Unfolded histograms of hXY
The layer have good planarity if the hXY of corresponding VOI is close to 90°
In-plane fiber deviations in the tapes
Unfolded histograms of uXY
The fibers in tape have good in-plane alignment if the uXY of corresponding VOI is concentrated to certain angle
Out-of-plane fiber deviations in the tapes
Unfolded histograms of hXY
The fibers in tape have good out-of-plane alignment if the hXY of corresponding VOI is concentrated to certain angle close to 90°
‘‘splitting” of the tapes
Unfolded histograms of uXY
Tape splitting was occurred if the uXY of corresponding VOI is widespread and does not have significant concentration
‘‘sticking” of the tapes
Unfolded and quantified histograms of uXY
Tapes in several layers are considered was stuck if the corresponding adjacent VOI exhibit cluster of uXY distribution and the SD show similar value
Tape morphology after molding
Unfolded histograms of uXY and 3D model
The integral, fractional and split tapes’ morphology can be determined and built by combining the corresponding uXY concentrations in the VOI with thresholded subsets of 3D model
Fig. 12. 3D model (uXY visualized) (a) and two-dimensional histogram of uXY and hXY (b) of M3 UT-CTT. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
3.2. Internal geometry of UT-CTT produced with different molding pressure Now the methodology introduced in the previous section will be employed to characterize the micro-structural change caused by molding pressure: the differences between M3 and M10 speci-
mens, produced with the molding pressure of 3 and 10 MPa respectively. The M3 UT-CTT 3D model with uXY distribution and the uXY hXY two-dimensional histogram of M3 UT-CTT are illustrated in Fig. 12. The histogram of M3 UT-CTT shows better hXY concentration and the clusters of uXY are more scattered than M10 UT-CTT
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
219
Fig. 13. Unfolded histograms of uXY (a) and hXY (b) of M3 UT-CTT. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 14. Unfolded uXY histogram and the corresponding clusters in visualized 3D model of M3 UT-CTT. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
220
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
because of the much lower molding pressure. The twodimensional histogram of uXY hXY is also unfolded through outof-plane direction (Fig. 13). Compared with the uXY distribution of M10 UT-CTT (Fig. 9(a)), the uXY distribution of M3 UT-CTT (Fig. 13(a)) exhibits better concentration along the z axis. Although the orientation of each layer in M3 UT-CTT is still not independent, the clusters are relatively small (2–3 VOIs in general). The difference in hXY distributions of M3 (Fig. 13(b)) and M10 (Fig. 9(b)) UT-CTT is insignificant. The effect of the higher concentrated uXY distribution on the 3D structure of UT-CTT is also studied by subsets of the model with uXY thresholds (Fig. 14). Both the stuck tapes and tape splitting, which are observed in M10 model (Fig. 10), are also
observed in M3 model. The models with uXY thresholds of VOI 4–6 and VOI 6–8 in Fig. 14 show as the well aligned tapes stuck together, while the cluster of VOI 13–18 exhibit scattered orientation concentration areas with irregular model shape, which considered to be a structure disturbance caused by the tape splitting and tape waviness. But compared with the subset models of M10 UT-CTT, the M3 models show better orientation concentration based on the color of the models (subset model VOI 4–6 in Fig. 13). The representative X-ray images of both M10 and M3 UT-CTT were selected and illustrated in Fig. 15. Compared with the general one (Fig. 15(a)), the tape splitting mentioned before is clearly observed in M10 image as randomized single fibers (Fig. 15(b)). In
Fig. 15. X-ray images of a section of UT-CTT with well-preserved tapes (a, from M10 UT-CTT), M10 UT-CTT with tape splitting (b) and M3 UT-CTT with voids (c). Gray value in the images reflects the X-ray attenuation coefficient of the material, which is a function of its density and elemental composition and the gray value range of the images is from 696 to 5447. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 16. Histograms of average values and SD of uXY (a) and hXY (b).
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
contrast, the image of M3 UT-CTT shows no tape splitting but voids as the molding defects (Fig. 15(c)). The tape splitting generated based on the mechanism that high pressure during the molding process will split randomly distributed tapes that were stirred during the paper making process. Consequently, the tape splitting was just observed in M10 UT-CTT. The voids appear because the low molding pressure cannot force the resin into the fissures along tapes. X-ray CT technics show potential to quantify voids volume and location of composites under different molding conditions [24], but there still have a number of problems to tackle for precise quantification and will discuss in further researches. The average values and SD of uXY and hXY are also quantified for M3 UT-CTT (Fig. 16). The SD of uXY in M3 UT-CTT are lower on average than that in M10 because less tape splitting occurred in low molding condition. The SD of hXY is slightly lower in M3 than in M10, which indicates that the higher molding pressure causes a higher tape waviness because the tape splitting decreases the microstructural regularity. The comparison between the results of M3 and M10 UT-CTT indicates that the introduced methodology exhibits a significant capability to evaluate the differences in the internal geometry caused by the difference in molding pressure. 4. Conclusions The image processing method, which is based on structure tensor analysis and implemented in VoxTex software of KU Leuven, has been applied to the determination of the internal geometry of UT-CTT with complex multi-scale structure. Features of FOD of in-plane and out-of-plane fiber directions can be interpreted to reveal structural features of the tapes placement and fiber misalignment. The clusters of uXY in VOI indicate the tape interactions, such as stuck tapes and tape waviness, generated during molding processes. The change of uXY distribution in a single VOI caused by the change of the molding pressure provided evidence of the tape splitting caused by high molding pressure. The combination of unfolded histograms with visualized 3D models gives a highly intuitive method for the understanding of the fibrous structure of the material. Both the 3D models and the histograms of uXY and hXY values can provide important information for further numerical analysis of material properties. Change of the molding pressure affects the tape placement patterns. With the increase of the molding pressure from 3 to 10 MPa, tape splitting and tape waviness may occur and the orientation clustering of uXY decreased. In the next stage of the research the image analysis methodology presented here will be applied to obtain a larger statistical sampling of UT-CTT specimens, manufactured with different molding pressures and different tape sizes to characterize in detail the manufacturing parameters – structure relations. The orientation information obtained by this X-ray micro-CT method will be further exploited for the numerical simulation of mechanical properties of UT-CTT. Acknowledgements Part of this study was supported by Research Fellowships of Japan Society for the Promotion of Science Grant-in-Aid for JSPS Research Fellow Grant Number 15J09248. S.V. Lomov is a Toray Professor (Toray Chair for composite materials, KU Leuven). The development of micro-CT facilities at the Department of Materials Engineering is supported by the Hercules foundation (AKUL/09/001 ‘‘Micro- and nano-CT for the hierarchical analysis
221
of materials); the facilities are maintained under the supervision of Prof Martine Wevers. Part of this study was conducted as Japanese METI project ‘‘the Future Pioneering Projects/Innovative Structural Materials Project” since 2013 financial year. Authors would like to express sincerely appreciation to the project members who have provided valuable information and useful discussions. References [1] Takahashi J, Ishikawa T. Next challenge in CFRTP for mass production automotive application. In: SEICO 14 SAMPE EUROPE 35th international conference and forum. Paris, France, 2014. [2] Sihn S, Kim RY, Kawabe K, Tsai SW. Experimental studies of thin-ply laminated composites. Compos Sci Technol 2007;67(6):996–1008. [3] Amacher R, Cugnoni J, Botsis J, Sorensen L, Smith W, Dransfeld C. Thin ply composites: experimental characterization and modeling of size-effects. Compos Sci Technol 2014;101:121–32. [4] Feraboli P, Peitso E, Deleo F, Cleveland T, Stickler PB. Characterization of prepreg-based discontinuous carbon fiber/epoxy systems. J Reinf Plast Compos 2009;28(10):1191–214. [5] Feraboli P, Peitso E, Cleveland T, Stickler PB. Modulus measurement for prepreg-based discontinuous carbon fiber/epoxy systems. J Compos Mater 2009;43(19):1947–65. [6] Feraboli P, Cleveland T, Ciccu M, Stickler P, DeOto L. Defect and damage analysis of advanced discontinuous carbon/epoxy composite materials. Compos Part A – Appl Sci Manuf 2010;41(7):888–901. [7] Landry B, Hubert P. Experimental study of defect formation during processing of randomly-oriented strand carbon/PEEK composites. Compos Part A: Appl Sci Manuf 2015;77:301–9. [8] Shen HB, Nutt S, Hull D. Direct observation and measurement of fiber architecture in short fiber-polymer composite foam through micro-CT imaging. Compos Sci Technol 2004;64(13–14):2113–20. [9] Tan JC, Elliott JA, Clyne TW. Analysis of tomography images of bonded fibre networks to measure distributions of fibre segment length and fibre orientation. Adv Eng Mater 2006;8(6):495–500. [10] Tsarouchas D, Markaki AE. Extraction of fibre network architecture by X-ray tomography and prediction of elastic properties using an affine analytical model. Acta Mater 2011;59(18):6989–7002. [11] Alemdar A, Zhang H, Sain M, Cescutti G, Mussig J. Determination of fiber size distributions of injection moulded polypropylene/natural fibers using X-ray microtomography. Adv Eng Mater 2008;10(1–2):126–30. [12] Bernasconi A, Cosmi F, Dreossi D. Local anisotropy analysis of injection moulded fibre reinforced polymer composites. Compos Sci Technol 2008;68 (12):2574–81. [13] Abdin Y, Lomov SV, Jain A, van Lenthe GH, Verpoest I. Geometrical characterization and micro-structural modeling of short steel fiber composites. Compos Part A – Appl Sci Manuf 2014;67:171–80. [14] Nguyen Thi TB, Morioka M, Yokoyama A, Hamanaka S, Yamashita K, Nonomura C. Measurement of fiber orientation distribution in injection-molded shortglass-fiber composites using X-ray computed tomography. J Mater Process Technol 2015;219:1–9. [15] Sun X, Lasecki J, Zeng D, Gan Y, Su X, Tao J. Measurement and quantitative analysis of fiber orientation distribution in long fiber reinforced part by injection molding. Polym Testing 2015;42:168–74. [16] Bernasconi A, Cosmi F, Hine PJ. Analysis of fibre orientation distribution in short fibre reinforced polymers: a comparison between optical and tomographic methods. Compos Sci Technol 2012;72(16):2002–8. [17] Hamdi SE, Delisée C, Malvestio J, Da Silva N, Le Duc A, Beaugrand J. X-ray computed microtomography and 2D image analysis for morphological characterization of short lignocellulosic fibers raw materials: a benchmark survey. Compos Part A: Appl Sci Manuf 2015;76:1–9. [18] Dietrich S, Gebert JM, Stasiuk G, Wanner A, Weidenmann KA, Deutschmann O, et al. Microstructure characterization of CVI-densified carbon/carbon composites with various fiber distributions. Compos Sci Technol 2012;72 (15):1892–900. [19] Saeed U, Rizvi G. Three-dimensional orientation of compression-molded highdensity polyethylene/wood fibers using X-ray micro-tomography. J Cell Plast 2015;51(1):45–57. [20] Faessel M, Delisée C, Bos F, Castéra P. 3D Modelling of random cellulosic fibrous networks based on X-ray tomography and image analysis. Compos Sci Technol 2005;65(13):1931–40. [21] Naouar N, Vidal-Salle E, Schneider J, Maire E, Boisse P. 3D composite reinforcement meso F.E. analyses based on X-ray computed tomography. Compos Struct 2015;132:1094–104. [22] Barburski M, Straumit I, Zhang XW, Weyers M, Lomov SV. Micro-CT analysis of internal structure of sheared textile composite reinforcement. Compos Part A – Appl Sci Manuf 2015;73:45–54. [23] Straumit I, Lomov SV, Wevers M. Quantification of the internal structure and automatic generation of voxel models of textile composites from X-ray computed tomography data. Compos Part A: Appl Sci Manuf 2015;69:150–8. [24] Nikishkov Y, Airoldi L, Makeev A. Measurement of voids in composites by Xray computed tomography. Compos Sci Technol 2013;89:89–97.
Contents lists available at ScienceDirect
Composites: Part A journal homepage: www.elsevier.com/locate/compositesa
Micro-CT analysis of internal geometry of chopped carbon fiber tapes reinforced thermoplastics Yi Wan a,⇑, Ilya Straumit b, Jun Takahashi a, Stepan V. Lomov b a b
The University of Tokyo, Department of Systems Innovation, Hongo 7-3-1, Bunkyo-ku, Tokyo, Japan KU Leuven, Department of Materials Engineering, Kasteelpark Arenberg 44, B-3001 Leuven, Belgium
a r t i c l e
i n f o
Article history: Received 25 April 2016 Received in revised form 15 September 2016 Accepted 8 October 2016 Available online 11 October 2016 Keywords: A. Polymer-matrix composites (PMCs) B. Microstructures D. CT analysis D. Microstructural analysis
a b s t r a c t Adequate characterization of the internal geometry of random fiber composites for prediction of their properties requires acquisition and analysis of vast amount of data. The present work proposes a method to measure and quantify the internal geometry of ultra-thin chopped carbon fiber tape reinforced thermoplastics (UT-CTT), a kind of randomly oriented strands, using the X-ray micro-CT. The two-scale fiber orientation distributions (orientation of the tapes and of the fibers inside the tapes), tape morphology, 3D structure and orientation misalignment were characterized using a specialized image processing methodology. The results indicate that the applied method can provide detailed information like primary orientation directions, tape waviness and splitting of the internal geometry of UT-CTT. UT-CTTs produced with different molding pressure were studied to confirm the capability of the methodology in distinguishing subtle structure changes caused by different molding conditions. Ó 2016 Elsevier Ltd. All rights reserved.
1. Introduction In recent years a significant progress has been made in the development of carbon fiber reinforced thermoplastics (CFRTP). Compared to thermoset polymers, thermoplastics exhibit superior cycle molding time and good in-plant recyclability. Randomly oriented short fiber systems generally show high capability of being manufactured in complex geometries without internal structural damage. Combined with their low manufacturing cost, random carbon fiber reinforced thermoplastics (rCFRTP) are regarded as potential substitutes for metallic materials in mass-production applications [1]. Advantageous thin ply effects of prepreg CF tapes on laminated composites mechanical performance have been verified by Sihn et al. [2] and Amacher et al. [3]. Composites molded by thin prepreg CF sheets show superior qualities in suppression of microcracking, delamination, and splitting damage for static, fatigue, and impact loadings. Randomly oriented strands (ROS) are one of the representative rCFRTP with high Vf, which makes possible achieving high mechanical properties [4–7]. X-ray micro computed tomography (micro-CT) has become a powerful non-destructive method for characterizing and visualizing the internal geometry of composite materials. With the ⇑ Corresponding author. E-mail address: [email protected] (Y. Wan). http://dx.doi.org/10.1016/j.compositesa.2016.10.013 1359-835X/Ó 2016 Elsevier Ltd. All rights reserved.
developments in digital image acquisition and image processing ability, the quantification of structural features, 3D model rebuilding together with the finite element (FE) mesh generation are considered to be the trends of current X-ray micro-CT research. rCFRTP generally possess complex internal geometry due to the interaction of the tape chunks during manufacturing. X-ray micro-CT provides a higher accuracy in the characterization of fiber orientation distribution (FOD), fiber length distribution (FLD), and multi-scale geometry quantification, compared with 2D image scanning method. The FOD and FLD of short glass fiber-reinforced phenolic foam with 2.7% fiber volume fraction (Vf) was characterized using Xray micro-CT method by Shen et al. [8]. FOD, FLD and 3D models of metal fiber networks and metal fiber foam were evaluated in [9,10]. X-ray micro-CT studies of short fiber composites are mainly concentrated on injection molded materials [11–16]. Nguyen et al. [14] used X-ray micro-CT and FE flow simulation to investigate the fiber orientation and flowability of injection molded glass fibers (GF) reinforced polyamide 6 (PA6). Sun et al. [15] studied the local fiber orientation and the effect of injection flow in short carbon fibers (CF) reinforced polypropylene (PP). Bernasconi et al. [12] calculated the local average orientation distribution and visualized 3D orientation models of GF/PA6 injection materials using the Mean Intercept Length (MIL) technique. The fiber morphology like the distributions of fiber length and fiber width were analyzed by Alemdar et al. [11]. The comparison between X-ray micro-CT and
212
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
2D image scanning method of short-GF/PA6 [16] and lignocellulosic fibers [17] composites were also studied. The FODs and 3D rebuilt models of short carbon-carbon felts [18] and wood fibers (WF) reinforced high-density polyethylene (PE) [19] were investigated. The combination of the reconstructed 3D model with the FE mesh generation and numerical analysis have been reported by several research groups [13,20,21]. The referenced studies did not consider ROS composites. The internal geometry of these materials differs significantly from the one of injection molded composites. In the ROS composites fiber orientation distribution has two-scales: orientation of the tapes and (mis)orientation of fibers within the tapes. This creates a situation similar to the characterization of fiber orientations in textile composites, where one has to distinguish between variability of the yarn directions and misalignment of fibers within the yarns [22]. In the present study, an image processing method based on structure tensor [23] is applied to determine the internal geometry of one kind of ROS made by water dispersed thin tapes, the ultrathin chopped carbon fiber tape reinforced thermoplastics (UT-CTT). The UT-CTT are developed in the framework of a Japan national project and considered a potential material for mass-production applications [1]. The proposed methods for reconstruction of multi-scale orientation distributions, tape morphology and 3D material models are verified for UT-CTT manufactured with two different molding conditions. The paper aims at detailed presentation of the image processing methodology, using a limited number of specimens. In future publications the methodology will be applied to material characterization with a necessary statistical robustness.
2. Materials and methods 2.1. Materials and specimens The UT-CTT used in this study are composed of randomly oriented UT-tape (oriented ultra-thin unidirectional prepreg tape). Spread carbon fiber tow (TR 50S, Mitsubishi Rayon Co., Ltd.) and Polyamid-6 (PA6, DIAMIRONTM C, Mitsubishi Plastics, Inc.) are used
to produce this tape. The fiber volume fraction (Vf) of UT-CTT is 55% in average. The UT-tape is considerably thin (44 lm in average) compared with the conventional one (about 150 lm or more), so it is called ‘ultra-thin’ (Fig. 1). The UT-tapes used in this study are provided by the Industrial Technology Center of Fukui Prefecture in Japan and are cut to 5 mm width and 6 mm length. To ensure a better tape distribution properties and to preserve the tape structure after molding, the intermediate UT-CTT sheets were manufactured by wet-type paper making process (Fig. 2). In the process tapes are put into a water container. The bottom of the container can be opened and has a filter. After the tapes are put into the container and randomly dispersed in the water, open the bottom of the container and filter out water to make the UT-CTT sheets. The UT-CTT sheets are stacked to manufacture UT-CTT plates with 2 mm thickness using compression molding. To identify the capability of the micro-CT method to recognize an internal geometry change, two different molding conditions are applied in this study, labeled as M10 (10 MPa molding pressure) and M3 (3 MPa molding pressure) (Fig. 3). After the UT-CTT plates are made, the specimens for X-ray measurement are prepared. To ensure the sufficient resolution of the X-ray images and provide reliable information on the internal geometry and fibers, the specimens must be cut in relatively small size. After each edge of the plates was cut off by 15 mm to eliminate the molding edge effect, the specimens with a size of 2 2 30 mm (thickness width length) were cut from both M10 and M3 plates arbitrarily by thin section cutting machine (EXAKT Advanced Technologies GmbH 310CP) for this study. 2.2. X-ray micro-CT scanning The internal structural information of UT-CTT was observed and collected by the 3D X-ray scan system TDM1000-II from Yamato Scientific Co., Ltd. During the observation, the specimens were fixed on a rotational stage (Fig. 4), and the distance between the rotation axis and the radiation source is set to 10 mm for reliable resolution of embedded fibers on the images. The scanned volume is a right circular cylinder because of the rotation, and the size of the cylinder is 0.88 mm in radius and 1.76 mm in height. To ensure sufficient image resolution during CT acquisition process, the
Fig. 1. Cross section of a conventional tape (a), UT-tape (oriented ultra-thin unidirectional prepreg tapes) (b) and surface of UT-CTT (ultra-thin chopped carbon fiber tape reinforced thermoplastics). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
213
Fig. 2. Wet-type paper making process for manufacturing UT-CTT sheets. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 3. Molding conditions M3 and M10. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
sample was positioned in the scanner’s field of view so that the imaged volume had a physical size of 1 1 1 mm and excluded the air surrounding the sample. The X-ray tube voltage was set to 40 kV and X-ray tube current was set to 40 lA for all the specimens. After the acquisition of the X-ray projection images (a rotation step of 0.24 degree and 25 min per full rotation of the sample), the 3D image was reconstructed by the image processing unit of the X-ray CT system. The pixel size of the reconstructed 3D micro-CT images was fixed to 4 lm, and 512 images stacked through out-of-plane direction for each specimen were used for image processing and calculations. The schematic of scanning process and scanned volume are illustrated in Fig. 5.
2.3. Image processing and data analysis The internal geometry of the material was reconstructed from the CT image using a regular rectangular (voxel) mesh, where each element of the mesh contains a set of variables [23]. The CT image is converted from the native scanner format to a single threedimensional array I(x1, x2, x3) of 8-bit gray values, where the coordinates x1, x2, x3 are integer numbers. The analyzed variables are fiber orientation angles u and h in spherical coordinate system. u is in-plane orientation axis, with the Cartesian axes x1 and x2 lying in the plane of the sheet, and h is an angle of the vector with the axis x3, normal to x1 and x2 axes. The fiber orientation vector is
214
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
The integration is done over a local volume V of the CT image. The integration volume V has a cubical shape with a size of 17 17 17 pixels. The eigenvector e1 corresponding to the smallest eigenvalue k1 is the fiber orientation vector. The spherical angles were calculated from the eigenvector e1 (unit vector):
h ¼ arccosðe1;3 Þ
u ¼ arctan
e1;2 e1;1
ð4Þ ð5Þ
The voxel mesh therefore represents the two scalar fields h(x1, x2, x3) and u(x1, x2, x3) defined over the spatial domain of the material sample, represented in the CT image. For the purposes of present study, a special type of two-dimensional histograms was introduced, which includes the analyzed variable as one of the axes, and one of the spatial dimensions (x1, x2, x3) as the second axis of the histogram. A bin of this histogram shows frequency of occurrence of orientations in the range of the bin over a certain cross-section plane in the material. This type of histograms (further referred to as ‘‘unfolded” histogram) allows to represent the change of the distribution of a variable along a spatial coordinate. In this study it was applied to analyze through-thickness distribution of fiber orientations. The calculations were performed using a C# code VoxTex, developed in KU Leuven [23]. Inside the code, two-dimensional histograms were rendered using Root Data Analysis Framework v 5.34 (CERN). Fig. 4. Specimen on the rotational stage. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
calculated from the CT image I(x1, x2, x3) as a solution of the following eigenvalue problem:
Sei ¼ ki ei
ð1Þ
where S is the three-dimensional gray value structure tensor:
Z
Sij ¼ V
@I @I dv @xi @xj
ð2Þ
The derivatives in Eq. (2) were calculated using the 5-point central difference formula:
@I 1 ½Iðx 2; yÞ 8Iðx 1; yÞ þ 8Iðx þ 1; yÞ Iðx þ 2; yÞ @x 12 and similarly for the derivative by y coordinate.
ð3Þ
3. Results and discussions In the present study the internal geometry of UT-CTT were observed and analyzed by 3D X-ray micro-CT. During the step of voxel model construction in the image processing, the distance between voxels for averaging calculation, which defines a volume for averaging, called below ‘‘volume of interest” (VOI), needs to be determined related to the thickness direction of the image UT-CTT. The VOI’s throughthickness dimension is set to 11 pixels, i.e., 44 lm to ensure that the thickness of each VOI is close to the thickness of a single tape (44 lm in average). Other two dimension of VOIs are equal to the extents of the image. The VOIs are numbered according to their order in position along the thickness direction: VOI 1 to VOI 25. All volumes of interest have a hexahedral shape.
Fig. 5. Schematic of scanning process and scanned volume of the UT-CTT. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
215
Fig. 7. 3D micro model with visualized in-plane fiber orientation distribution of M10 UT-CTT, the Phi_XY (uXY) indicate in-plane orientation angle. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 6. Reconstructed 3D model of M10 UT-CTT with layered structure, concentric circles (ring artefacts) are observed at the side face. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
After 3D CT images were collected by the 3D X-ray scanner, the micro-structure of UT-CTT was reconstructed as illustrated in Fig. 6. The micro-structure model shows that the UT-CTT hold both features of laminate materials and short fiber composites. The model exhibit a layered structure, also the individual tapes in the UT-CTT show not-perfectly-flat orientations. Concentric circles (ring artefacts) can be observed at the side face of the model (see Fig. 6). Ring artefacts are usually suppressed at the stage of micro-CT acquisition by calibrating the detector; or during the reconstruction of the image, if the reconstruction software has the function of ring artefacts correction. However, they often cannot be eliminated completely and, if present after the reconstruction, cannot be corrected in the already reconstructed image. The presence of the ring artefacts is the case in the images acquired in the present work. Being unable to eliminate the artefacts we, as it is discussed in the subsequent sections, have ignored the spurious features caused by them in the final results.
ures hereinafter indicate the data density of the orientation. The orientation distribution shows special features. The angle hXY is concentrated around 90 degree with small dispersion, which means the UT-CTT is almost in-plane oriented. Several clusters of uXY are observed in the figure, and these clusters show irregularity in the in-plane angle. Because the fibers of UT-CTT are almost inplane oriented, so the distributions of uXY and hXY are ‘‘unfolded” as described in Section 2.3 – through the out-of-plane direction to analyze their change through the thickness. Fig. 9 shows the unfolded distribution of uXY (a) and hXY (b) through z axis. The unfolded uXY distribution indicates that the concentrated clusters of uXY in Fig. 8 do not mean local concentration of tapes in UTCTT. For example, the uXY distribution cluster from 20 degree to 80 degree in Fig. 8 is actually combined by three small clusters located at different z-positions (Fig. 9). On the other hand, although the thickness of VOI is set close to the thickness of a single tape, the orientations of layers are still not independent. Fig. 9(a) shows that small clusters still exist in the model. The size of the clusters are generally 3–4 VOI, also a larger cluster (VOI 15 to 22, 30–70 degree) is found. Because the tapes in UT-CTT are almost inplane oriented (Fig. 9(b)), this result means some tapes with same orientation may stick together during the wet-type paper-making process, or the tape waviness and tape splitting occurred locally.
3.1. Internal geometry analysis Using the stacked micro-CT images of UT-CTT, the visualized 3D model of the material and histograms of fiber orientation distribution were calculated. In this section the results for M10 UT-CTT are used to demonstrate the analysis methodology. The visualized 3D micro model with in-plane fiber orientation of M10 UT-CTT is shown in Fig. 7. The angles are given in the global Cartesian coordinate system (X, Y, Z) identical to the Cartesian system (x1, x2, x3) introduced previously. The color in the model (as well in Fig. 11 (b)) indicated local in-plane fiber orientation angle uXY (Phi_XY), with degree (°) as the unit in this study. Orientation changes layer by layer, but the not-fully-flat in-plane orientation and waviness through out-of-plane direction (Z axis) indicated the UT-CTT also have orientation irregularity as it is the case with conventional rCFRTPs. The two-dimensional histogram where the in-plane (uXY) and out-of-plane (hXY, Theta_XY) fiber orientation distribution angle are combined is shown in Fig. 8, the color bar in this figure and fig-
Fig. 8. Two-dimensional histogram combined the in-plane (uXY) and out-of-plane (hXY) fiber orientation distribution in the total volume of the M10 UT-CTT specimen. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
216
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
Fig. 9. Unfolded histograms of uXY (a) and hXY (b) of M10 UT-CTT. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 10. Unfolded uXY histogram and the corresponding clusters in visualized 3D model of M10 UT-CTT. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
In addition, Fig. 9(a) indicates that the uXY is not randomly distributed, which is because the reconstructed 3D model of UT-CTT is smaller than the tapes due to the limited volume that can be imaged. The relatively small dimensions of the sample cannot ensure the statistical randomness of orientation distribution because of the size effect of the tapes. hXY shows lager scatter around the center layer of the model (Fig. 9(b)), this is considered to be caused not by the material feature but by the ring artefacts mentioned in the beginning of this section, which increase the isotropy of the calculated angles. Therefore the scatter in VOI 13 will be discarded from the analysis of the results. To study the clusters of uXY in detail, the 3D model with uXY distribution (Fig. 7) is combined with the unfolded uXY histogram through z axis (Fig. 9(a)). The subsets of the 3D model separated by the VOI were extracted following the clusters that appeared in the unfolded uXY histogram, which means that the 3D morphologies of the orientation concentration areas (the clusters in Fig. 9 (a)) can be specified and extracted from the general 3D model (Fig. 7). The VOI 4–7, 9–11 and 16–21 are shown in Fig. 10. After the subsets were extracted, the threshold of uXY is applied on the model to identify the fiber distributions in concentrated uXY and the threshold ranges are also illustrated in Fig. 10. The extracted 3D models with threshold of uXY and the corresponding areas in unfolded uXY histogram proved the assumption that the tapes can interact during the wet process and disturb the ideal uniformly random orientation distribution, which would exist if the placements of the tapes were independent. The existence of the uXY distribution clusters through z axis is due to the tapes with the same orientation sticking together during the wet-type paper making process, as well as tape waviness and tape splitting taking place locally during the compression molding process. The 3D model of VOI 4–7 demonstrates 4 different layers (tapes) with the same orientation distribution pattern. In contrast, the 3D model of VOI 9– 11 shows an integral part of tape with some scattered areas which
217
are considered to be the tape waviness and splitting. On the other hand, the 3D model of VOI 16–21 exhibits both structural features: tapes are stuck together at the top of the model, while scattered areas are observed on the bottom side. The combination of the 3D model with the unfolded uXY histogram shows a high capability for the quantitative internal geometry study of the UT-CTT with complex structural features. By extracting the subset models, the detail structures and tape positions are reconstructed visually. For the further research of the micro-structure and combination with simulations, the unfolded histograms were quantified. The average values and standard deviations (SD) of both uXY and hXY were calculated by each VOI (Fig. 11). The quantified orientation distribution of uXY reproduces the visualized histogram to a certain extent. The clusters in VOI 16–21 observed in the histograms are also shown as series of data with SD indicated by error bars. In contrast, the results of VOI 4, 6, 7 and 11 show significant mismatch between the peak values in Fig. 9(a) and the average value in Fig. 11(a). This is caused by the multimodality of the orientation distributions in these VOIs. The mismatch is caused by the multiple orientation concentrations in one VOI and the SD is significantly high even though the histogram show apparent orientation concentrations. On the other hand, when there is a single aligned integral part of tape in the VOI, the average value is perfectly matched with the peak in histogram, and the SD is small as the VOI 10. In addition, the average value and SD of hXY show concentrations same as the observed in the unfolded histogram. The increase of SD around the center layer is caused by the ring artefacts as discussed before. This section introduced a methodology for the internal geometry analysis of UT-CTT by micro-CT. It provides a convenient approach for quantified visualization modeling of the micro-3D structure with orientation information, which can give support for composite simulation modeling and numerical analysis. The assessable features were summarized and listed in Table 1.
Fig. 11. Average values and standard deviations (SD) of uXY (a) and hXY (b) of M10 UT-CTT.
218
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
Table 1 The assessable features of internal geometry of UT-CTT using micro-CT. Features
Required data
Example characterization
Planarity of layers
Unfolded histograms of hXY
The layer have good planarity if the hXY of corresponding VOI is close to 90°
In-plane fiber deviations in the tapes
Unfolded histograms of uXY
The fibers in tape have good in-plane alignment if the uXY of corresponding VOI is concentrated to certain angle
Out-of-plane fiber deviations in the tapes
Unfolded histograms of hXY
The fibers in tape have good out-of-plane alignment if the hXY of corresponding VOI is concentrated to certain angle close to 90°
‘‘splitting” of the tapes
Unfolded histograms of uXY
Tape splitting was occurred if the uXY of corresponding VOI is widespread and does not have significant concentration
‘‘sticking” of the tapes
Unfolded and quantified histograms of uXY
Tapes in several layers are considered was stuck if the corresponding adjacent VOI exhibit cluster of uXY distribution and the SD show similar value
Tape morphology after molding
Unfolded histograms of uXY and 3D model
The integral, fractional and split tapes’ morphology can be determined and built by combining the corresponding uXY concentrations in the VOI with thresholded subsets of 3D model
Fig. 12. 3D model (uXY visualized) (a) and two-dimensional histogram of uXY and hXY (b) of M3 UT-CTT. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
3.2. Internal geometry of UT-CTT produced with different molding pressure Now the methodology introduced in the previous section will be employed to characterize the micro-structural change caused by molding pressure: the differences between M3 and M10 speci-
mens, produced with the molding pressure of 3 and 10 MPa respectively. The M3 UT-CTT 3D model with uXY distribution and the uXY hXY two-dimensional histogram of M3 UT-CTT are illustrated in Fig. 12. The histogram of M3 UT-CTT shows better hXY concentration and the clusters of uXY are more scattered than M10 UT-CTT
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
219
Fig. 13. Unfolded histograms of uXY (a) and hXY (b) of M3 UT-CTT. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 14. Unfolded uXY histogram and the corresponding clusters in visualized 3D model of M3 UT-CTT. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
220
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
because of the much lower molding pressure. The twodimensional histogram of uXY hXY is also unfolded through outof-plane direction (Fig. 13). Compared with the uXY distribution of M10 UT-CTT (Fig. 9(a)), the uXY distribution of M3 UT-CTT (Fig. 13(a)) exhibits better concentration along the z axis. Although the orientation of each layer in M3 UT-CTT is still not independent, the clusters are relatively small (2–3 VOIs in general). The difference in hXY distributions of M3 (Fig. 13(b)) and M10 (Fig. 9(b)) UT-CTT is insignificant. The effect of the higher concentrated uXY distribution on the 3D structure of UT-CTT is also studied by subsets of the model with uXY thresholds (Fig. 14). Both the stuck tapes and tape splitting, which are observed in M10 model (Fig. 10), are also
observed in M3 model. The models with uXY thresholds of VOI 4–6 and VOI 6–8 in Fig. 14 show as the well aligned tapes stuck together, while the cluster of VOI 13–18 exhibit scattered orientation concentration areas with irregular model shape, which considered to be a structure disturbance caused by the tape splitting and tape waviness. But compared with the subset models of M10 UT-CTT, the M3 models show better orientation concentration based on the color of the models (subset model VOI 4–6 in Fig. 13). The representative X-ray images of both M10 and M3 UT-CTT were selected and illustrated in Fig. 15. Compared with the general one (Fig. 15(a)), the tape splitting mentioned before is clearly observed in M10 image as randomized single fibers (Fig. 15(b)). In
Fig. 15. X-ray images of a section of UT-CTT with well-preserved tapes (a, from M10 UT-CTT), M10 UT-CTT with tape splitting (b) and M3 UT-CTT with voids (c). Gray value in the images reflects the X-ray attenuation coefficient of the material, which is a function of its density and elemental composition and the gray value range of the images is from 696 to 5447. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 16. Histograms of average values and SD of uXY (a) and hXY (b).
Y. Wan et al. / Composites: Part A 91 (2016) 211–221
contrast, the image of M3 UT-CTT shows no tape splitting but voids as the molding defects (Fig. 15(c)). The tape splitting generated based on the mechanism that high pressure during the molding process will split randomly distributed tapes that were stirred during the paper making process. Consequently, the tape splitting was just observed in M10 UT-CTT. The voids appear because the low molding pressure cannot force the resin into the fissures along tapes. X-ray CT technics show potential to quantify voids volume and location of composites under different molding conditions [24], but there still have a number of problems to tackle for precise quantification and will discuss in further researches. The average values and SD of uXY and hXY are also quantified for M3 UT-CTT (Fig. 16). The SD of uXY in M3 UT-CTT are lower on average than that in M10 because less tape splitting occurred in low molding condition. The SD of hXY is slightly lower in M3 than in M10, which indicates that the higher molding pressure causes a higher tape waviness because the tape splitting decreases the microstructural regularity. The comparison between the results of M3 and M10 UT-CTT indicates that the introduced methodology exhibits a significant capability to evaluate the differences in the internal geometry caused by the difference in molding pressure. 4. Conclusions The image processing method, which is based on structure tensor analysis and implemented in VoxTex software of KU Leuven, has been applied to the determination of the internal geometry of UT-CTT with complex multi-scale structure. Features of FOD of in-plane and out-of-plane fiber directions can be interpreted to reveal structural features of the tapes placement and fiber misalignment. The clusters of uXY in VOI indicate the tape interactions, such as stuck tapes and tape waviness, generated during molding processes. The change of uXY distribution in a single VOI caused by the change of the molding pressure provided evidence of the tape splitting caused by high molding pressure. The combination of unfolded histograms with visualized 3D models gives a highly intuitive method for the understanding of the fibrous structure of the material. Both the 3D models and the histograms of uXY and hXY values can provide important information for further numerical analysis of material properties. Change of the molding pressure affects the tape placement patterns. With the increase of the molding pressure from 3 to 10 MPa, tape splitting and tape waviness may occur and the orientation clustering of uXY decreased. In the next stage of the research the image analysis methodology presented here will be applied to obtain a larger statistical sampling of UT-CTT specimens, manufactured with different molding pressures and different tape sizes to characterize in detail the manufacturing parameters – structure relations. The orientation information obtained by this X-ray micro-CT method will be further exploited for the numerical simulation of mechanical properties of UT-CTT. Acknowledgements Part of this study was supported by Research Fellowships of Japan Society for the Promotion of Science Grant-in-Aid for JSPS Research Fellow Grant Number 15J09248. S.V. Lomov is a Toray Professor (Toray Chair for composite materials, KU Leuven). The development of micro-CT facilities at the Department of Materials Engineering is supported by the Hercules foundation (AKUL/09/001 ‘‘Micro- and nano-CT for the hierarchical analysis
221
of materials); the facilities are maintained under the supervision of Prof Martine Wevers. Part of this study was conducted as Japanese METI project ‘‘the Future Pioneering Projects/Innovative Structural Materials Project” since 2013 financial year. Authors would like to express sincerely appreciation to the project members who have provided valuable information and useful discussions. References [1] Takahashi J, Ishikawa T. Next challenge in CFRTP for mass production automotive application. In: SEICO 14 SAMPE EUROPE 35th international conference and forum. Paris, France, 2014. [2] Sihn S, Kim RY, Kawabe K, Tsai SW. Experimental studies of thin-ply laminated composites. Compos Sci Technol 2007;67(6):996–1008. [3] Amacher R, Cugnoni J, Botsis J, Sorensen L, Smith W, Dransfeld C. Thin ply composites: experimental characterization and modeling of size-effects. Compos Sci Technol 2014;101:121–32. [4] Feraboli P, Peitso E, Deleo F, Cleveland T, Stickler PB. Characterization of prepreg-based discontinuous carbon fiber/epoxy systems. J Reinf Plast Compos 2009;28(10):1191–214. [5] Feraboli P, Peitso E, Cleveland T, Stickler PB. Modulus measurement for prepreg-based discontinuous carbon fiber/epoxy systems. J Compos Mater 2009;43(19):1947–65. [6] Feraboli P, Cleveland T, Ciccu M, Stickler P, DeOto L. Defect and damage analysis of advanced discontinuous carbon/epoxy composite materials. Compos Part A – Appl Sci Manuf 2010;41(7):888–901. [7] Landry B, Hubert P. Experimental study of defect formation during processing of randomly-oriented strand carbon/PEEK composites. Compos Part A: Appl Sci Manuf 2015;77:301–9. [8] Shen HB, Nutt S, Hull D. Direct observation and measurement of fiber architecture in short fiber-polymer composite foam through micro-CT imaging. Compos Sci Technol 2004;64(13–14):2113–20. [9] Tan JC, Elliott JA, Clyne TW. Analysis of tomography images of bonded fibre networks to measure distributions of fibre segment length and fibre orientation. Adv Eng Mater 2006;8(6):495–500. [10] Tsarouchas D, Markaki AE. Extraction of fibre network architecture by X-ray tomography and prediction of elastic properties using an affine analytical model. Acta Mater 2011;59(18):6989–7002. [11] Alemdar A, Zhang H, Sain M, Cescutti G, Mussig J. Determination of fiber size distributions of injection moulded polypropylene/natural fibers using X-ray microtomography. Adv Eng Mater 2008;10(1–2):126–30. [12] Bernasconi A, Cosmi F, Dreossi D. Local anisotropy analysis of injection moulded fibre reinforced polymer composites. Compos Sci Technol 2008;68 (12):2574–81. [13] Abdin Y, Lomov SV, Jain A, van Lenthe GH, Verpoest I. Geometrical characterization and micro-structural modeling of short steel fiber composites. Compos Part A – Appl Sci Manuf 2014;67:171–80. [14] Nguyen Thi TB, Morioka M, Yokoyama A, Hamanaka S, Yamashita K, Nonomura C. Measurement of fiber orientation distribution in injection-molded shortglass-fiber composites using X-ray computed tomography. J Mater Process Technol 2015;219:1–9. [15] Sun X, Lasecki J, Zeng D, Gan Y, Su X, Tao J. Measurement and quantitative analysis of fiber orientation distribution in long fiber reinforced part by injection molding. Polym Testing 2015;42:168–74. [16] Bernasconi A, Cosmi F, Hine PJ. Analysis of fibre orientation distribution in short fibre reinforced polymers: a comparison between optical and tomographic methods. Compos Sci Technol 2012;72(16):2002–8. [17] Hamdi SE, Delisée C, Malvestio J, Da Silva N, Le Duc A, Beaugrand J. X-ray computed microtomography and 2D image analysis for morphological characterization of short lignocellulosic fibers raw materials: a benchmark survey. Compos Part A: Appl Sci Manuf 2015;76:1–9. [18] Dietrich S, Gebert JM, Stasiuk G, Wanner A, Weidenmann KA, Deutschmann O, et al. Microstructure characterization of CVI-densified carbon/carbon composites with various fiber distributions. Compos Sci Technol 2012;72 (15):1892–900. [19] Saeed U, Rizvi G. Three-dimensional orientation of compression-molded highdensity polyethylene/wood fibers using X-ray micro-tomography. J Cell Plast 2015;51(1):45–57. [20] Faessel M, Delisée C, Bos F, Castéra P. 3D Modelling of random cellulosic fibrous networks based on X-ray tomography and image analysis. Compos Sci Technol 2005;65(13):1931–40. [21] Naouar N, Vidal-Salle E, Schneider J, Maire E, Boisse P. 3D composite reinforcement meso F.E. analyses based on X-ray computed tomography. Compos Struct 2015;132:1094–104. [22] Barburski M, Straumit I, Zhang XW, Weyers M, Lomov SV. Micro-CT analysis of internal structure of sheared textile composite reinforcement. Compos Part A – Appl Sci Manuf 2015;73:45–54. [23] Straumit I, Lomov SV, Wevers M. Quantification of the internal structure and automatic generation of voxel models of textile composites from X-ray computed tomography data. Compos Part A: Appl Sci Manuf 2015;69:150–8. [24] Nikishkov Y, Airoldi L, Makeev A. Measurement of voids in composites by Xray computed tomography. Compos Sci Technol 2013;89:89–97.