1 Introduction

Thanks to their low density, excellent high-temperature sustainability, high specific stiffness and specific strength, carbon fiber-reinforced SiC-matrix composites (C/SiC composites) are being widely used for the structures to resist to high mechanical/thermal/thermal-mechanical loadings in the field of aerospace, astronautics, automotive, nuclear engineering, et al. [1,2,3,4,5,6,7]. Pores generally cannot be avoided in carbon fiber-reinforced composite material during some commonly used fabrication process, e.g. Chemical-Vapor Infiltration (CVI) process, and they were found to have an influence on the mechanical properties in the literature [8,9,10]. Thus it is important to reveal the relations between the damage mechanisms and the pores in order to improve the fabrication process and optimize the composite structures. Many experimental studies on damage mechanisms focus on carbon fiber-reinforced resin-matrix composites [10,11,12], while only a few focus on carbon fiber-reinforced SiC-matrix composites [1, 2]. The influence of pores on damage mechanisms in C/SiC composites fabricated by CVI process is still to be clarified.

X-ray Computed Tomography is an attractive tool to characterize the microstructures in the bulk of material non-destructively [13]. When sample mounted on a rotating stage between X-ray beam source and the detector is rotated through 180° (for synchrotron tomography) or 360° (for laboratory tomography), a set of radiographies are recorded and are then used to reconstruct the 3D image, in which the gray levels reflect the microstructure constituents in the bulk of sample. X-ray Computed Tomography has already been used in the field of experimental mechanics for the in-situ characterizations of damage evolutions in the bulk of specimen. Thus it provides a more reliable and direct way to study the damage mechanisms [14, 15].

An experimental protocol to study the influence of pores on damage mechanisms in a C/SiC composite fabricated by CVI process has been developed in the laboratory. The mechanical/thermal-mechanical tests realized by a special designed in-situ loading rig will be performed on C/SiC samples with in-situ observations by using X-ray Computed Tomography. The relations between the damage evolutions and pores will thus be established in the bulk. This paper focuses on the validation of the characterizations of pores and damages in the studied C/SiC composite by using X-ray Computed Tomography. It is necessary for the following in-situ tests. Two C/SiC specimens were characterized by using X-ray Computed Tomography in this paper: one is an undamaged specimen for the purpose of the characterizations of pores and another one is a damaged specimen after bending fatigue tests under high temperature in order to characterize the damages.

2 Experimental Methods

2.1 Material, Specimen and X-Ray Computed Tomography

Two-dimensional plain woven C/SiC composite (T300 carbon fiber) formed by CVI process has been studied [2, 4]. The specimens were in the shape of beam (3 mm × 6 mm × 60 mm). The two specimens shown in Fig. 1 were examined by using laboratory X-ray Computed Tomography:

Fig. 1
figure 1

Two C/SiC specimens examined by X-ray Computed Tomography: a undamaged specimen without suffering any loadings; (b) damaged specimen that fractured partially after a three point bending fatigue test under 500 °C (failure occurred after 4 × 105 cycles)

Full size image

One is an undamaged specimen (Fig. 1a) without suffering any mechanical/thermal tests. It was scanned by using a Phoenix x|Nanotom m X-ray nano CT system in Key Laboratory for Advanced Materials Processing Technology, Ministry of Education (Tsinghua University, Beijing, China) with a voxel size of 2.5 μm for the purpose of characterizations of pores as well as the microstructures in the bulk of material. A length of 5.8 mm in the beam specimen was scanned with the current resolution and a cube volume of 3 mm × 6 mm × 5.8 mm was obtained for material characterizations.

The other one is a damaged specimen (Fig. 1b) that fractured partially after a three point bending fatigue test (Sinusoidal loads, frequency: 100 Hz) under 500 °C (failure after 4 × 105 cycles) performed with a special designed testing machine in the laboratory. It was scanned by using an YXLON microfocus Computed Tomography in State Key Laboratory of Nonlinear Mechanics (Institute of Mechanics, Chinese Academy of Sciences, Beijing, China) with a voxel size of 5.0 μm in order to characterize the damages in the bulk of specimen. A length of 10.72 mm in the damaged beam specimen was scanned with the current resolution and it allows the final fracture area and its neighborhood areas being characterized.

2.2 Three-Dimensional Numerical Image Processing and Analysis

The three-dimensional numerical image processing and analysis were performed by using ImageJ [16] and Avizo 9.3 [17] software. The main steps are illustrated in Fig. 2.

Fig. 2
figure 2

The main steps of 3D digital image processing and analysis

Full size image

The reconstructed 16bit gray level 3D images were converted to 8bit gray level 3D images. In the conversion process, the gray level of the pixel having the maximum gray level in the original 16bit 3D image is defined as ‘255’ in the converted 8bit 3D image, while the gray level of the pixel having the minimum gray level in the original 16bit 3D image is defined as ‘0’ in the converted 8bit 3D image. Then the gray levels of the other pixels having the gray levels between the maximum level and the minimum level in the original 16bit 3D image are placed between ‘0’ and ‘255’ with a linear scaling in the converted 8bit 3D image.

Some appropriate filters are necessary to be applied to the converted 8bit 3D image prior to the following image segmentations. Two kinds of filter were adopted in this paper, i.e. Median filter and Anisotropic diffusion filter [17, 18]. Median filter is a simple edge-preserving smoothing filter and reduces the noise by replacing the value of a voxel by the median value of its specific neighborhood voxels. The principle of Anisotropic diffusion filter is more complex than Median filter but it works effectively to preserve strong edges and enhance the contrast of edges. The value of the current voxel is compared with the values of its six neighbors in order to determine the new value for the current voxel. The diffusion works if the difference is below the specific diffusion stop criterion.

The segmentation of carbon fiber tow, SiC matrix, pores and damage in the bulk of specimen was realized by using grayscale thresholding method, i.e. by selecting the voxels having gray levels between the minimum and the maximum levels that correspond to the segmented constituent. The pores and damages including cracks and delaminations have lowest gray level; the SiC matrix have the highest gray level due to its high density while the carbon fibers have the medium gray levels (Fig. 3). Some other morphological image processing, such as dilation, erosion, opening and closing, are necessary to be applied on the segmented binary 3D images in order to correct the segmentation errors due to noises in the original image. The reader could refer to Ref. [18] to learn some similar trials that used in this paper.

Fig. 3
figure 3

Tomography slice in two-dimensional plain woven C/SiC composite fabiracated by CVI process (undamaged specimen). The slice is perpendicular to the laminate plain

Full size image

The thresholded damages and/or pores were labeled at first for the image analysis. After labeling, voxels that belong to a same individual object are given by a unique gray level and different objects are marked with different gray levels. The measurements could be performed on each object in the labeled images by using quantification tools in Avizo software or ‘Analysis_3D’ plugin [19] in ImageJ software. The 3D morphologies of carbon fiber tow, SiC matrix, pore and damage in the bulk of specimen can also be visualized by using ‘Generate Surface’ and ‘Surface View’ tools in Avizo software or ‘3D Viewer’ plugin in ImageJ software.

3 Results

3.1 Three-Dimensional Characterization of Pores

The undamaged specimen without suffering any loading was used to characterize the pores and microstructures in the studied two-dimensional plain woven C/SiC composite. Fig. 3 is a tomography slice view in the undamaged specimen. The imaging resolutions used in this study (either 2.5 μm for undamaged specimen or 5.0 μm for damaged specimen) allow the carbon fiber tows, SiC matrix, pores and damages being visualized clearly in the reconstructed image. Fig. 3 shows that large pores are located inside the C/SiC matrix between laminates. The present study is performed in the micro scale, while the morphologies inside the fiber tow, including the individual fibers and the dispersed quite small pores inside each fiber tow, are not revealed clearly by using the current resolution (2.5/5.0 μm) that is above their sizes.

The morphologies of fiber tow, SiC matrix, and pore in different orthogonal directions are shown in Fig. 4a. Fiber tows present totally different morphologies in the laminate plain and in the plains perpendicular to the laminate as a result of material design and fabrication process. In order to visualize the 3D morphology of pores and their relations with fiber tow and SiC matrix, the pores and SiC matrix in a cube with a side length of 0.75 mm inside the specimen are extracted and are shown in 3D rendering in Fig. 4b-d. Even in a small cube, the pores presenting individual morphology in the 2D plain (Fig. 3) present a connected morphology in Fig. 4b: the pores located in the different laminates can also connect with each other and some examples are marked with circles. The SiC matrix present plain-like morphology and they connects a lot between different laminates as shown in Fig. 4c. If the 3D renderings of pores and SiC matrix are shown in the same time in Fig. 4d, it is found that almost all pores are surrounded by SiC matrix except to some dispersed quite small pores which may be real small pores located inside the fiber tows or be due to image noises.

Fig. 4
figure 4

a Orthogonal slices view inside the undamaged C/SiC specimen shows the morphologies of fiber tow, SiC matrix and pores in different orthogonal directions. A cube with a side length of 0.75 mm inside the specimen is shown in 3D volume rendering. Pores and SiC matrix in this cube are segmented separately and are shown in 3D rendering: (b) pores are shown in red solid with orthogonal slices; (c) SiC matrix are shown in yellow solid with orthogonal slices; (d) pores are shown in red solid with SiC matrix shown in yellow translucence, while fiber tows are concealed

Full size image

The pores in the whole scanned volume of undamaged C/SiC specimen are then segmented and are shown in 3D rendering in Fig. 5a. The 3D morphologies of pores located in several laminates near surface are visualized in Fig. 5b and c shown with the 3D volume rendering of the rest part of specimen. It is observed that the pores present highly interconnected network morphology in the plain between laminates.

Fig. 5
figure 5

a 3D rendering of pores in the scanned volume of undamaged C/SiC specimen; (b) and (c) 3D rendering of pores located in several laminates near surface with the rest part of specimen shown in 3D volume rendering

Full size image

The quantification analysis of pores is then performed on the undamaged specimen. The volume fraction of pores in the scanned sample is about 7%. Due to the dispersion of the studied material, such a quantitative analysis in the limited scan volume (about 200 mm3) could not represent the property of the studied material. Even for the 3D characterizations performed on an Al alloy (Fig. 3–19 in [18]) that has a less dispersion than the studied composite material, the volume fraction of pores can have a difference of more than ten times when CT scans are performed on different parts of the same sample. Thus a large enough scan volume on different representative samples is necessary to acquire the representative volume fraction of pores in the studied material. The quantitative results given here and in the following in this paper are only to help understand the qualitative analysis better. More than 20,000 individual pores are found, however, large pores in small number occupy the majority volume:

  • The largest pore that form a highly interconnected networks between several laminates (Fig. 4b and c show part of this pore) occupies about 40% of the total volume of pores in the characterized volume;

  • The largest 13 pores of which each has a volume more than 0.01 mm3 occupy 62% of the total volume of pores;

  • The largest 86 pores of which each has a volume more than 0.001 mm3 occupy 84% of the total volume of pores;

  • The rest large amounts (more than 20,000) of dispersed small size pores of which each has a volume below 0.001 mm3 occupy only 16% of the total volume.

The dispersed small size pores may be attributed to one of the following four situations:

  • Dispersed individual pores inside the SiC matrix;

  • Individual pores inside the fiber tow;

  • Pores (either inside SiC matrix or inside fiber tow) that connect with other large pores through quite small “passage”, that below the voxel size by using the current resolution (example shown in Fig. 6). These small ‘passage’ parts cannot be imaged by using the current resolution and the connections are thus be cut off.

  • Noises in the reconstructed image.

Fig. 6
figure 6

Schematic drawing of the connection, of which the size is below the voxel size, between two separated pores in the tomography images

Full size image

3.2 Three-Dimensional Characterization of Damages

The damaged specimen fractured after high temperature three point bending fatigue test was used for the characterization of pores in the material. The 3D volume rendering of the damaged specimen and the 3D rendering of segmented cracks and pores are shown in Fig. 7. The main crack is perpendicular to the top surface and passes through many laminates. However, it could be observed in Fig. 7c that the main crack presents a triangle shape in the fracture plain: the main crack passes through almost all the laminates from top surface under fatigue loading in one side, i.e. the left side in the right bottom view in Fig. 7c, while it passes only a few laminates in another side, i.e. in the right side in the right bottom view in Fig. 7c. It indicates that the main crack may initiate at the left top corner in the right bottom view in Fig. 7c, where there is a stress concentration, and then propagate towards right and down under cyclic loadings.

Fig. 7
figure 7

a 3D volume rendering of the damaged specimen in the final fracture area (left) and in the whole scanned volume (right); (b) 3D rendering of main crack and pores in blue solid with 3D volume rendering of the damaged specimen; (c) half of the 3D volume rendering of the damaged specimen in (b) is concealed in order to show the 3D morphology of main crack, while the main crack and its neighbor pores in the fracture plain are shown in 3D rendering in the right bottom

Full size image

The damages in the bulk of the specimen are also analyzed slice by slice. Figs. 8 and 9 give some examples of tomography slices in different view directions. The main crack passes through both fiber tow and SiC matrix in Fig. 8. However, matrix cracks are also observed outside the main crack area. In additional, more voids are observed outside the main crack area than in the undamaged specimen characterized in Figs. 3 and 4a even only by depending on a subjectively direct visualization. These voids may be pores or damages however they cannot be distinguished in the current images. A quantification analysis is then performed on these voids outside the main crack area. The volume fraction of these voids outside the main crack area is more than 10% that is much higher than the reference volume fraction (7%) of pores characterized in the undamaged specimen (section 3.1). It could be ascribed to the following three cases:

  1. 1)

    The volume of original pores is enlarged under the high temperature fatigue loading;

  2. 2)

    The matrix breaks outside the main crack area increase the volume of voids;

  3. 3)

    There are delaminations outside the main crack area which also increase the volume of voids and the delaminations are indeed observed in the examples shown in Fig. 9.

Fig. 8
figure 8

Tomography slices of the damaged specimen in the plain parallel with the laminates are shown with different distances to the top surface shown in Fig. 7(a)

Full size image
Fig. 9
figure 9

Examples of tomography slices in different view directions (in the plains perpendicular to the laminate plain) in the damaged specimen: (a) one slice about 1.5 mm below the side surface shown in Fig. 7(a); (b) one slice parallel with the front surface shown in Fig. 7(a) (about 3 mm to the main crack area)

Full size image

4 Discussions

The pores in the studied two-dimensional plain woven C/SiC composite fabricated by CVI process have been thoroughly characterized in 3D. Although the pores present dispersed void-like morphology by 2D characterizations, such as traditional optical microscope or Scanning Electronic Microscope (refer to literature [2, 6], similar with the tomography slice shown in Fig. 3), they are observed in 3D characterizations to form a highly interconnected plain networks inside SiC matrix between laminates and even connect through several laminates to form a 3D interconnected network architecture. However, they may have much more connections than the current observations but these connections may have a size below the voxel size by using the current resolution thus be cut off, such as an example illustrated in Fig. 6. A higher resolution in nanoscale that could reveal the quite small size pores will result in a smaller scanning volume and thus will obstruct to get a reorganization of the highly interconnected 3D architecture in the present study focusing in the microscale.

The large pores should be paid attention at first even their amounts are much smaller than the small pores but they occupy the majority volume of pores. Large pores are also easier to induce stress concentration under mechanical loadings that may accelerate the nucleation of damage. In additional, an attractive application of the studied C/SiC composite material is to resist to the thermal-mechanical loadings. The oxidation of C/SiC composite material has been stressed in high temperature environment [1,2,3, 20]. The interconnected 3D architecture of large pores may provide the ‘passages’ for air flow and thus accelerate the oxidations. While such pore ‘passages’ are generally surrounded by SiC matrix (Fig. 4c) that is an antioxidation material and thus obstructs the contact between the ‘passages’ of air and the fiber tows. However, even the material surface may be protected by antioxidant coat and the fiber tows are surrounded by SiC matrix, air may flow into these porosity ‘passages’ and touch with fiber tows once the SiC matrix and/or the antioxidant coat are broken under mechanical loadings. Nevertheless, the oxidation ablation of fiber tow is not observed obviously in the tomography image of the damaged specimen after high temperature bending fatigue test, because it is found in the literature [2] that the oxidation of T300 carbon fiber becomes obvious when temperature is above 600 °C that is higher than the testing temperature (500 °C) in this paper.

The 3D characterization of damages in the damaged specimen highlights the effectivity of X-ray Computed Tomography for the damage analysis. The 3D morphology of the main crack is visualized while it is difficult to be understood by using the traditional fracture analysis. If one separates the two half parts of the partial fractured specimen for the fracture analysis, it is hard to distinguish the fracture surface caused by fatigue loading and that by separation after test. Even if the fracture analysis works on the fracture area, i.e. the main crack area, it is unable to reveal the damages (such as delaminations and matrix breaks) outside the final fracture area (Figs. 8 and 9) examined by using X-ray Computed Tomography. However, there is a drawback for the characterizations of damages in the damaged specimen in this paper: the voids caused by the damages (such as delaminations and cracks) or the deformations of pores due to loadings and the voids belongs to the fabrication pores existing before tests cannot be distinguished in the current tomography image. This drawback will be made up in the following in-situ tests: the tomography image taken before the test on the testing specimen will help recognize the fabrication pores in the following tomography images taken during tests. Thus the validation of the characterizations of pores and damages in this paper is successful.

5 Conclusions

In order to study the influence of pores on damage mechanisms in C/SiC composites by using in-situ observations with X-ray Computed Tomography, the three dimensional characterizations of pores and damages by using X-ray Computed Tomography are validated in this paper. X-ray Computed Tomography is found to be very effective to characterize the pores and damages in the studied material:

  1. (1)

    The pores in the studied material form a highly interconnected plain networks inside SiC matrix between laminates and even connect through several laminates to form a 3D interconnected network architecture. The SiC matrix enclosing the networks of pores could obstruct the contact between the air in the networks of pores and the fiber tows, however once the SiC matrix and/or the antioxidant coat are broken under mechanical loadings, the 3D interconnected network architecture may help the contact of air with fiber tows and thus accelerate the oxidations of carbon fibers at high temperature.

  2. (2)

    The main damages types including fiber break, matrix break and delamination are identified in the damaged C/SiC specimen after a bending fatigue test under 500 °C. The main crack may initiate at a corner of the specimen and then propagate in the plain perpendicular to the laminates by both fiber breaks and matrix breaks. Delaminations and matrix breaks are also observed in the bulk outside the fracture area. Oxidations of carbon fiber tows are not observed obviously under the current 500 °C.

The mechanical/thermal mechanical tests with in-situ observations by using Synchrotron Radiation Computed Tomography will then be performed on the studied C/SiC composite in order to visualize the damage evolutions in the bulk of material.