Construction and Building Materials 134 (2017) 220–227

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier.com/locate/conbuildmat

Technical note

Dispersion of carbon fibers in cement-based composites with different

mixing methods
Jie Gao a,b, Zhenjun Wang c,d,⇑, Ting Zhang c, Liang Zhou c
a
School of Highway, Chang’an University, Xi’an 710064, China
b
National and Local Joint Engineering Materials Laboratory of Traffic Engineering and Civil Engineering, Chongqing Jiaotong University, 400074, China
c
School of Materials Science and Engineering, Chang’an University, Xi’an 710061, China
d
Engineering Research Central of Pavement Materials, Ministry of Education of P.R. China, Chang’an University, Xi’an 710061, China

a r t i c l e i n f o a b s t r a c t

Article history: Mixing methods can significantly influence dispersion of carbon fibers in cement-based composites. In
Received 22 June 2016 this work, pre-mixing method (carbon fibers were added before cement) and after-mixing method (car-
Received in revised form 22 November 2016 bon fibers were added after cement) were adopted to prepare carbon-fiber-reinforced cement-based
Accepted 13 December 2016
composites (CFRC). X-ray computed tomography (CT) was used to obtain the cross-sectional images of
Available online 30 December 2016
successive layers in the composites. Scanning electron microscope (SEM) was adopted to capture
microstructure images of the composites. The CT images were continuously identified through their
Keywords:
grayscale peaks. The sub-mm pores were separated from carbon bundles to improve the accuracy of
Carbon-fiber-reinforced cement-based
composites
the identification results. Two-dimensional (2D) dispersion characteristics of two different mixing
Carbon fiber dispersion methods were quantitatively analyzed by extracting pixels area of each component. Meanwhile, three-
Carbon fiber bundles dimensional (3D) models were established by means of optimized CT images to obtain the volume frac-
Mixing methods tion of carbon fiber bundles. The results show that the average areas of uniformly dispersed area, which
obtains through pre-mixing method, are higher than those with after-mixing method. The dispersion
effect of pre-mixing method is superior than that with after-mixing method. In addition, the volume
fraction of carbon fiber bundles in specimens prepared through pre-mixing method was lower than that
prepared by after-mixing method. Thus, the contrast resolution of X-ray CT system can recognize
materials with over 0.3% density difference in CFRC; and the pre-mixing method is recommended in
the preparation of carbon-fiber-reinforced cement-based composites.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction difficult to achieve electric function roles of carbon fibers when

they are poorly dispersed [17–19]. Therefore, it is necessary to
Carbon-fiber-reinforced cement-based composites (CFRC) pos- quantitatively evaluate dispersion of carbon fibers in cement-
sess not only excellent mechanical properties [1–6], but also high based composites and recommend suitable mixing methods for
conductive and electromagnetic performances [7–9]. In recent preparing the carbon-fiber-reinforced cement-based composites.
years, CFRC have been extensively used as smart civil engineering To evaluate dispersion of carbon fibers in cement matrix, a tra-
materials for structural monitors, intelligent buildings and deicing ditional method was introduced in Technical Specification for Fiber
or snow-melting pavements [10–15]. However, carbon fibers tend Reinforced Concrete Structures in China (CECS38:2004) [20] and
to cluster during process of preparing the composites. Nowadays, Standard Test Method for Determination of Glass Fiber Content
pre-mixing method (carbon fibers were added before cement) in Glass Fiber Reinforced Concrete (ASTM C1229-94) [21]. Several
and after-mixing method (carbon fibers were added after cement) specimens with equal mass were separated from the fresh fiber
are usually adopted to prepare carbon-fiber-reinforced cement- reinforced cement paste and numbered. Each specimen was placed
based composites [16]. However, mixing methods of carbon fibers on a square hole sieve with a diameter of 75 lm. The cement
can evidently influence their dispersion in cement matrix. It is particles were washed out with water and carbon fibers were col-
lected, dried and weighed. The dispersion can be evaluated by vari-
ation coefficient of carbon fiber mass. In addition, Yang et al. [22]
⇑ Corresponding author at: School of Materials Science and Engineering, Chang’an introduced four methods to evaluate fiber dispersion, which were
University, Xi’an 710061, China.
fresh mixture method, scanning electron microscope method,
E-mail address: [email protected] (Z. Wang).

http://dx.doi.org/10.1016/j.conbuildmat.2016.12.047
0950-0618/Ó 2016 Elsevier Ltd. All rights reserved.
 J. Gao et al. / Construction and Building Materials 134 (2017) 220–227 221

measurement of electrical resistance and simulation experiment. volume content of carbon fiber (Vcf) in specimen can be calculated
The results showed that different methods could be used under dif- by Eq. (1).
ferent conditions. Woo et al. [23] compared conductivity of differ- mcf qcf
ent specimens to determine the uniformity of steel fiber v cf ¼  100% ð1Þ
mc qc þ mw qw þ mcf qcf
dispersion. Kim et al. [24,25] used graphics software and statistical
analysis software to obtain the dispersion coefficient of polyvinyl where mc, mw, mcf – mass of cement, water and carbon fiber. qc, qw,
alcohol fiber. However, the methods mentioned above can only qcf – density of cement, water and carbon fiber, which are shown in
achieved indirect analysis of dispersion but can not directly obtain Table 1 or Table 2.
the dispersion morphology of carbon fibers and the accurate vol- In this work, two mixing methods were used to prepare the
ume fraction of carbon fiber bundles. composites. For the pre-mixing method, the mixing water was
In this work, the first purpose is to evaluate the dispersion of weighed firstly. Then, carbon fibers were added and stirred for
carbon fibers with different mixing methods, which influences 60 s by a JJ-1/1A forced electric mixer with the rotation rate
design, preparation, property and application of CFRC. The second 1000r/min. In the following, cement was added and stirred by a
purpose is to describe the distribution morphology of carbon fiber JJ-5 cement mortar mixer in accordance with the pattern of fast
bundles in the cement matrix, which plays vital roles in the func- (120 r/m) – slow (60r/m) – fast (120 r/m), and each pattern was
tional sensitivity and mechanical property enhancement of CFRC. stirred for 30 s, respectively. The process of pre-mixing method is
illustrated in Fig. 1.
2. Experimental For the after-mixing method, the mixing water and cement
were first stirred by a JJ-5 cement mortar mixer (60 r/m) for 30 s.
2.1. Raw materials Then, the carbon fibers were added and stirred for 120 s (fast-
slow-fast pattern, 40 s each). The detail process of after-mixing
In this work, ordinary Portland cement was used and its method is illustrated in Fig. 2.
properties are shown in Table 1, which can meet requests in The stirred mixtures were poured into 40 mm  40 mm 
Standards of Common Portland Cement in China (GB 175-2007) 40 mm steel models and vibrated for 30 s in a vibration platform.
[26]. Polyacrylonitrile (PAN)-based carbon fibers were adopted The specimens were placed in a curing room with the temperature
and their properties are shown in Table 2. Tap water was used. of 20 ± 2 °C and the relative humidity >90%. After curing for 7 days,
the specimens were used for tests. In CT image acquisition and
2.2. Sample preparation and tests identification, the large density difference between carbon fiber
and water or air makes them easily-distinguished. Therefore,
2.2.1. Preparation of the specimens capture of CT images and recognition of carbon fibers can not be
In specimens, the proportion of various materials set as: influenced by the present of free water in specimens. Moreover,
cement:water:carbon fiber = 1:0.55:0.025 (in mass). Then the considering the grayscale varies with curing age significantly, the

Table 1
Properties of ordinary Portland cement.

Fineness Density Initial setting Final setting Flexural strength Compressive

time (h) time (h) (MPa) strength (MPa)
(m2/kg) (g/cm3) 3d 28 d 3d 28 d
320 3.105 3.1 4.9 5.9 7.9 21.7 55.4

Table 2
Properties of carbon fiber.

Diameter Length Density Tensile strength Young’s modulus Carbon content

(lm) (mm) (g/cm3) (MPa) (GPa) (%)
7.1 2–5 1.742 3900 230 95.3

Fig. 1. The process of pre-mixing method.

222 J. Gao et al. / Construction and Building Materials 134 (2017) 220–227

Fig. 2. The process of after-mixing method.

specimens for X-ray CT scanning should have the same curing age 3. Results and discussion
to conduct the separation strategy more precisely [27].
3.1. Optimization of X-ray CT images

2.2.2. X-ray CT image acquisition Captured images by X-ray CT were converted gray images of
The technical parameters of X-ray CT system used in this work BMP format files to obtain the original CT scanned images (Fig. 4
are shown in Table 3. It shows that the CT system possesses 24 lp/ (a)) with commercial image processing software (IPS). X-ray
mm spatial resolution, which means that it is able to detect the absorption coefficient is related with material density and X-ray
object whose diameter beyond 0.2 mm. Thus the main investigated energy grades. In theory, the same material possesses the same
object is the lager carbon fiber bundles in mm order. However, the X-ray absorption capacity under X-ray radiation with stable energy
contrast resolution of X-ray CT is 0.3%, which means that materials [28]. However, energy of X-ray often attenuates rapidly during the
with over 0.3% density difference can be easily distinguished. The penetration process, which leads to the uneven distribution of
carbon fiber bundles are composed of carbon fiber and air. The energy. Therefore, the original X-ray CT images were processed
density of carbon fiber is 1.742 g/cm3 and that of cement particle with IPS to eliminate negative impact factors.
is 3.105 g/cm3. Therefore, the contrast resolution is enough to dis- The first step is to filter, in which median filter (3  3 mode) in
tinguish them. spatial filter group was adopted to remove original image Gaussian
The scanning thickness was 2 mm by layer, 19 consecutive noise and thin edge pixels, to keep the foreground edge smoothly.
slices of CT images were obtained for each 40 mm cubic sample The second is to sharpen, which means that sharpen filter
in this work. The process of image acquisition is illustrated in (7  7 mode) was used to enhance the detail representation of
Fig. 3. In addition, S-4800 scanning electron microscope (SEM) fuzzy parts and improve the discrimination between the different
was adopted to capture images of the composites for the gray value gray scale of pixels. The last step is to optimize the image quality.
determination of different components. In order to improve accuracy of carbon fiber bundles identification,
brightness and contrast of the images were adjusted to highlight
Table 3 the characteristics of the component morphology, especially differ-
Technical parameters of X-ray CT system. ence between carbon fibers and cement paste. The optimized
images are shown in Fig. 4(b).
Parameters Results
Maximum tube voltage/kV 225
Effective scanning thickness/ mm 0.5–2 3.2. Identification of X-ray CT images
Spatial resolution/ mm 0.2
Contrast resolution/% 0.3 The grayscale of image essentially reflects density of materials
Field of image 1024  1024
[29]. The higher brightness parts in the image are tending to be
Image format BMP/DOCM
white, which indicates the higher density of materials. Meanwhile,

Fig. 3. The process of image acquisition.

 J. Gao et al. / Construction and Building Materials 134 (2017) 220–227 223

Fig. 4. CT image of the composites: a) original CT image; b) Optimized CT image.

19000
18000 Cement paste without CF
After-mixing method Peak B
17000
Pre-mixing method
16000
Peak C
Peak A

4000
Frequency

Peak E
0 2 4 6 8 10 12 14 16 18 20 22
2000
Gray value Peak H
Peak D
Peak F Peak G

0
0 50 100 150 200 250
Grayscale

Fig. 5. Grayscale frequency histogram of the CT image.

the lower brightness parts in the image tends to be black, which where i, j is the pixel of rows and columns at the entire image,
indicates the lower density of the material [30–32]. The expression respectively; N, M is number of rows and columns at the pixel
of image information relies on the brightness of numerous rectan- matrix
gular pixels with different grayscale, which are divided into 255 Eq. (2) shows that the discrete function contains all information
levels (0 to 255) in this work. about the location and grayscale of each pixel; and the grayscale of
After CT image is stored as a gray image, the whole image is each pixel reflects the density of the corresponding point in the CT
formed by pixel matrix with different grayscale, which constitutes image. Accordingly, extractions of all the pixel matrixes from a CT
a discrete function f (i, j), as shown in Eq. (2). image are able to obtain frequency distribution of grayscale, which
8 9 are shown in Fig. 5.
> f ð1; 1Þ f ð1; 2Þ  f ð1; MÞ >
>
> >
> As shown in Fig. 5, two peaks are observed in grayscale
< f ð2; 1Þ f ð2; 2Þ  f ð2; MÞ =
f ði; jÞ ¼ ð2Þ frequency distribution of the cement pastes without carbon fibers,
>
>     > > which stand for two different components in CFRC. The pores
>
: >
;
f ðN; 1Þ f ðN; 2Þ    f ðN; MÞ (larger than sub-mm) in cement paste present the lowest grayscale

Table 4
Morphology and gray thresholds of different components in CFRC.

Different components Carbon fiber bundles Uniformly dispersed area Carbon fiber-free area Sub-mm pores
Morphology

Gray threshold
Pre-mixing 9–46 46–220 220–255 0–9
After-mixing 9–31 31–207 207–255 0–9
224 J. Gao et al. / Construction and Building Materials 134 (2017) 220–227

Fig. 6. The SEM images of CFRC specimen.

for that the air in pores has the lowest density. Thus, the peak A can components in the CFRC specimens prepared by pre-mixing and
be considered as grayscale peak of sub-mm pores and the gray after-mixing methods, respectively. Each grayscale peak represents
threshold of sub-mm pores is 0–9. The peak B is grayscale peak one component in the specimen. Therefore, three peaks indicate
of cement paste because it possesses the highest density, thus three different components of the specimen, which are a) fiber
the gray threshold of cement paste is 200–255. bundle, b) uniformly dispersed fiber area, and c) fiber-free area.
Furthermore, three of evident grayscale peaks are observed They are sequenced by their density (q), which are qa < qb < qc.
in Fig. 5, which can indicate that there exist three different Then the grayscale (g) order can be sorted as ga < gb < gc in order.

Fig. 7. The typical distribution characterization of each component in specimens with different mixing methods.
 J. Gao et al. / Construction and Building Materials 134 (2017) 220–227 225

Therefore, the peak C, peak D and peak E of the CFRC specimen images. On the contrary, the carbon fiber bundles formed by
with after-mixing method is carbon fiber bundles, uniformly after-mixing method present flaky-shaped distribution. Both the
dispersed area and carbon fiber-free area, respectively. Similarly, uniform dispersed area and the fiber-free area present clustered
the peak F, peak G and peak H of the CFRC specimen with pre- distribution, exhibiting no regularity as shown in Fig. 7.
mixing method is carbon fiber bundles, uniformly dispersed area After typical distribution characterization acquired, in order to
and carbon fiber-free area, respectively. quantitatively evaluating the carbon fiber dispersion, pixel area
It should be noticed that only pores with larger than sub-mm of each component in 19 successive slice images for both pre-
size can be observed in the captured CT images in this work. The mixing and after-mixing were counted to contrastively evaluate
sub-mm pores in cement paste show lower grayscale than carbon their dispersion by the software. Carbon fiber bundle area, uni-
fiber bundles due to the air in the pores with the lowest density. formly dispersed area and fiber-free area are represented as AXb,
The gray threshold of pores (peak A) in cement paste without car- AXu, AXf (pre-mixing) and AHb, AHu, AHf (after-mixing), respec-
bon fiber is within the thresholds of carbon fiber bundles in CFRC tively. The results are shown in Fig. 8.
with pre-mixing (peak F) and after-mixing (peak C) method,
respectively. In order to avoid the influence of gray threshold of 44 AHb AXb Mean AHb Mean AXb
pores on the identification results of carbon fiber bundles, it is nec- 40
essary to separate the pores from carbon bundles by deducting the 36
gray threshold of sub-mm pores from those of carbon fiber bun-
32
dles. Moreover, either the peak H of the CFRC with pre-mixing
28

Position/mm
method or the peak E of the CFRC with after-mixing method is
included in the peak B of cement paste without carbon fibers, 24
which indicates that the peak H stands for the carbon fiber-free 20
area of specimen with pre-mixing. In addition, the peak E can be 16
considered as grayscale peak of carbon fiber-free area in specimen 12
with after-mixing method. Finally, based on the indication of gray 8
threshold, different component of each slice can be continuously
4
identified through gray threshold of each component. The results
are shown in Table 4 1x10
4
2x10
4
3x10
4
4x10
4
5x10
4
6x10
4

Pixels
3.3. Analysis of microstructure of CFRC
(a) Areas of carbon fiber bundles
In Fig. 6, it can be observed that the carbon fiber bundles are 44
present concentrated distribution. The carbon fiber bundles con- AHu AXu Mean of AHu Mean of AXu
40
tain carbon fibers and air and show the lowest brightness because 36
of the lowest density in contrast to that of cement paste or carbon
32
fibers. The uniformly dispersed fiber area is composed of cement
Positon/mm

28
particles with a small amount of carbon fiber. The density of
cement particles is 3.105 g/cm3. Therefore, the uniformly dispersed 24
fiber area presents a significantly higher brightness than carbon 20
fiber bundles. On the other hand, Fig. 6 also illustrated that the 16
existence of fiber-free area. The carbon fiber-free area possesses 12
the highest brightness because of the highest cement density 8
among raw materials.
4
Furthermore, the component is represented by a grayscale
threshold instead of a fixed value in this work for the reason that 4.0x10
4
6.0x10
4
8.0x10
4
1.0x10
5
1.2x10
5
1.4x10
5

the same component can possess slight grayscale difference due

Pixels
to slight density difference. Therefore, the identification results of
carbon fiber bundles and uniformly dispersed fibers can be well (b) Areas of uniform dispersed carbon fiber
determined by combination of grayscale and SEM images.
In addition, the identification of X-ray CT images is the founda-
AHf AXf Mean of AHf Mean of AXf
tion for dispersion evaluation by 2D images and 3D models. In this 40
study, the carbon fiber dispersion with different mixing methods 35
was determined in 2D images first, then distribution morphology
30
Position/mm

of carbon fiber bundles was described in 3D models.

25
3.4. 2D dispersion characteristics of carbon fibers
20

In order to study the two-dimensional dispersion characteris- 15

tics of carbon fibers in the specimens, the optimized CT images
10
were automatically segmented in the IPS by the gray thresholds
shown in Fig. 5. Accordingly, three components were labeled with 5
different colors and pixel area of each component was obtained by
4 4 4 4
the IPS. The typical distribution characterization of each compo- 0 1x10 2x10 3x10 4x10
nent in specimens is shown in Fig. 7. Pixels
Fig. 7 shows that carbon fiber bundles formed by pre-mixing (c) Areas of fiber-free
method present an island distribution in the uniform area.
Meanwhile, the fiber-free areas are distributed on the edge of the Fig. 8. Area statistic results of each component.
226 J. Gao et al. / Construction and Building Materials 134 (2017) 220–227

Theoretically, carbon fibers can be considered completely dis- cannot be determined well. Therefore, 3D reconstruction model
persed if areas of carbon fiber bundles and areas of the fiber-free of the specimen was developed with the optimized CT images by
reach zero; and uniform dispersion areas of carbon fibers are commercial 3D processing software (3DPS), whose settings were
approaching the total image pixel area. The higher value between 2 mm layer distance, optimized model quality and 1.2 smoothing
AXu and AHu indicates the better dispersion effect of carbon fibers. edge factor. Reconstructed 3D model of the specimen with dimen-
As to 2D images shown in Fig. 8, Mean AXb, Mean AXu and Mean AXf sions of 38 mm  40 mm  40 mm. Then, 3D models of carbon
of the composites prepared with pre-mixing method are 0.43, 1.72 fiber bundles were conducted on the basis of the gray threshold
and 0.30 times of the Mean AHb, Mean AHu, and Mean AHf of the of carbon fiber bundles. The results are shown in Fig. 9.
composites prepared with after-mixing method, respectively. In Fig. 9 shows that feature of carbon fiber bundles in 3D models is
conclusion, the dispersion effect of carbon fibers through pre- more intuitive than that of 2D images. It presents island distribu-
mixing method is better than that through after-mixing method tion after pre-mixing method and concentrative sheet distribution
in the same mixing duration. The excellent distribution of carbon through after-mixing method. The volume of original model, orig-
fibers during the pre-mixing process can be attributed to the fact inal 3D model with pre-mixing and after-mixing methods were
that surface of carbon fiber is well wetted previously. Therefore, 6.080  104 mm3, 2.031  103 mm3 and 5.042  103 mm3, respec-
the friction force among carbon fibers is reduced due to the pres- tively. In addition, on contrast to the reconstructed 3D model,
ence of water films on surface of carbon fibers. Consequently, the volume fraction of carbon fiber bundle model with the meth-
despite that several carbon fiber bundles are exhibiting, most car- ods of pre-mixing and after-mixing were Vpm = 3.34% and
bon fibers are dispersed uniformly in the cement matrix. Vam = 8.29%, respectively. The result indicates that the dispersion
After incorporation of carbon fiber into the cement matrix with effect of pre-mixing method is much better than that of after-
after-mixing method, larger bundles cannot be dispersed into mixing method. It is worth noting that the values of Vpm and Vam
smaller bundles completely because that the carbon fibers in large are much higher than that carbon fiber volume content (1.2%).
bundles are difficult to be wetted. Along with mixing, a small por- Furthermore, the carbon fiber bundle exhibits a loose structure,
tion of carbon fiber in the edge of the bundle is dispersed into the which is composed of undispersed carbon fibers and air. This
matrix under the friction action. However, most of them are still phenomenon results in that cement paste is difficult to exist the
clustered. Therefore, after-mixing method is not well able to pre- bundle interior and fill the space among carbon fibers because of
pare the composites with uniform dispersed carbon fibers. the barrier of carbon fiber bundles.

4. Summary and conclusions

3.5. 3D dispersion characteristics of carbon fiber bundles
To quantitatively evaluate the fiber dispersion in carbon fiber
Although the dispersion effect of two different mixing methods cement-based composites prepared by pre-mixing and after-
can be determined by carbon fiber distribution in two-dimensional mixing method, X-ray CT was used to obtain digital images of
images, distribution of carbon fibers in three-dimensional space specimens. The following conclusions can be drawn.

Fig. 9. 3D model of carbon fiber bundles in CFRC.

 J. Gao et al. / Construction and Building Materials 134 (2017) 220–227 227

a) The X-ray CT image optimized processing has significant [8] F.J. Baeza, O. Galao, E. Zornoza, P. Garc E S, Effect of aspect ratio on strain
sensing capacity of carbon fiber reinforced cement composites, Mater. Des. 51
effect to improve image quality. Identified gray thresholds
(10) (2013) 1085–1094.
of carbon fiber bundle, uniform dispersed area and fiber- [9] Wu Yao, Bing Chen, Wu Keru, Smart behaviour of carbon fiber reinforced
free area, were put forward depending upon its gray scale cement-based composite, J. Mater. Sci. Technol. 19 (3) (2003) 239–242.
to categorize the components in images. [10] Egemen Teomete, Measurement of crack length sensitivity and strain gage
factor of carbon fiber reinforced cement matrix composites, Measurement 74
b) The spatial resolution of X-ray CT system can be used to (10) (2015) 21–30.
distinguish carbon fiber bundles which are in mm order. [11] Baoshan Huang, Xingwei Chen, Xiang Shu, Effects of electrically conductive
However, the contrast resolution of X-ray CT system can additives on laboratory-measured properties of asphalt mixtures, ASCE J.
Mater. Civil Eng. 21 (10) (2009) 612–617.
recognize materials with over 0.3% density difference in [12] Kezhi Li, Chuang Wang, Hejun Li, Effect of chemical vapor deposition
CFRC. treatment of carbon fibers on the reflectivity of carbon fiber-reinforced
c) The accuracy of the identification results of different com- cement-based composites, Compos. Sci. Technol. 68 (5) (2008) 1105–1114.
[13] K.T. Lau, Structural health monitoring for smart composites using embedded
posites is highly dependent on their grayscale thresholds, FBG sensor technology, Mater. Sci. Technol. 30 (13A) (2014) 1642–1654.
which can be well determined by their grayscale peaks [14] Xiaojun Wang, D.D.L. Chung, Self-monitoring of fatigue damage and dynamic
and SEM images. The identification result of carbon fiber strain in carbon fiber polymer-matrix composite, Compos. Part B Eng. 29B (1)
(1998) 63–73.
bundles should be optimized for decreasing influence of [15] Sherif Yehia, Christopher Tuan, Thin conductive concrete overlay for bridge
sub-mm pores on the identification accuracy of the carbon deck deicing and anti-icing, Transp. Res. Rec. 1698 (2000) 45–53.
fiber bundles. [16] Jian Zhou, Shunzhi Qian, Guang Ye, Oguzhan Copuroglu, Klaas van Breugel,
Victor C. Li, Improved fiber distribution and mechanical properties of
d) The effect of pre-mixing and after-mixing methods can be
engineered cementitious composites by adjusting the mixing sequence, Cem.
determined by comparing its pixel area of uniform dispersed Concr. Compos. 34 (3) (2012) 342–348.
carbon fibers (AXu and AHu). The higher value between AXu [17] Burak Felekoglu, Kamile Tosun-Felekoglu, Ravi Ranade, Qian Zhang, Victor C.
and AHu, the better dispersion method it is. As results, the Li, Influence of matrix flowability, fiber mixing procedure, and curing
conditions on the mechanical performance of HTPP-ECC, Compos. B Eng. 60
dispersion effect of pre-mixing method is superior to that (4) (2014) 359–370.
of after-mixing method with the same mixing duration. [18] Yilmaz Akkaya, Surendra P. Shah, Bruce Ankenman, Effect of fiber dispersion
e) The 3D model of carbon fiber bundles is an effective mea- on multiple cracking of cement composites, J. Eng. Mech. 127 (4) (2001) 311–
316.
surement to obtain its volume fraction and dispersion mor- [19] Chuang Wang, Kezh Li, Hejun Li, Gengsheng Jiao, Jinhua Lu, Dangshe Hou,
phology. The carbon fiber bundle’s volume of the composites Effect of carbon fiber dispersion on the mechanical properties of carbon fiber-
prepared by pre-mixing method is lower than that of the reinforced cement-based composites, Mater. Sci. Eng.: A 487 (1–2) (2008) 52–
57.
composites prepared by after-mixing method. Therefore, [20] CECS38:2004, Technical specification for fiber reinforced concrete structures,
the pre-mixing method is recommended for preparing the General Administration of Quality Supervision, Inspection and Quarantine of
carbon-fiber-reinforced cement-based composites. the People’s Republic of China, Beijing, 2004.
[21] ASTM C1229-94, Standard Test Method for Determination of Glass Fiber
Content in Glass Fiber Reinforced Concrete (GFRC) (Wash-Out Test), ASTM
International, West Conshohocken, PA, 2015.
Acknowledgements [22] Yuanxia Yang, Methods study on dispersion of fibers in CFRC, Cem. Concr. Res.
32 (5) (2002) 747–750.
[23] L.Y. Woo, S. Wansom, N. Ozyurt, B. Mu, S.P. Shah, T.O. Mason, Characterizing
The authors thank the supports from National and Local Joint fiber dispersion in cement composites using AC-Impedance Spectroscopy,
Engineering Materials Laboratory of Traffic Engineering and Civil Cem. Concr. Compos. 27 (6) (2005) 627–636.
Engineering, Chongqing Jiaotong University (No: LHSYS-2016- [24] Jin.-Keun Kim, Jeong.-Su Kim, Gee Joo Ha, Yun Yong Kim, Tensile and fiber
dispersion performance of ECC (engineered cementitious composites)
002), the Fundamental Research Funds for the Central Universities produced with ground granulated blast furnace slag, Cem. Concr. Res. 37 (7)
(No. 310831153504 and 310831163113), and Science and Technol- (2007) 1096–1105.
ogy Project of Fujian Province in P.R. China (No. 2014H0017). [25] Bang Yeon Lee, Jin-Keun Kim, Jeong-Su Kim, Yun Yong Kim, Quantitative
evaluation technique of Polyvinyl Alcohol (PVA) fiber dispersion in engineered
cementitious composites, Cem. Concr. Compos. 31 (6) (2009) 408–417.
References [26] GB 175–2007. Standards of Common Portland Cement in China. General
Administration of Quality Supervision, Inspection and Quarantine of the
[1] Chuang Wang, Kezhi Li, Hejun Li, Lingjun Guo, Gengsheng Jiao, Influence of CVI People’s Republic of China, Beijing; 2007.
treatment of carbon fibers on the electromagnetic interference of CFRC [27] E. Gallucci, K. Scrivener, A. Groso, M. Stampanoni, G. Margaritondo, 3D
composites, Cem. Concr. Compos. 30 (6) (2008) 478–485. experimental investigation of the microstructure of cement pastes using
[2] Xiang Shu, Ryan K. Graham, Baoshan Huang, Edwin G. Burdette, Hybrid effects synchrotron X-ray microtomography (mu CT), Cem. Concr. Res. 37 (3) (2007)
of carbon fibers on mechanical properties of Portland cement mortar, Mater. 360–368.
Des. 65 (1) (2015) 1222–1228. [28] Wenyuan Ren, Zhenjun Yang, Rajneesh Sharma, Ch Zhang, Philip J. Withers,
[3] D.D.L. Chung, Cement reinforced with short carbon fibers: a multifunctional Two-dimensional X-ray CT image based meso-scale fracture modelling of
material, Compos. B Eng. 31 (6–7) (2000) 511–526. concrete, Eng. Fract. Mech. 133 (2014) 24–39.
[4] Ryan K. Graham, Baoshan Huang, Xiang Shu, Edwin G. Burdette, Laboratory [29] M. Shibata, S. Okazaki, I. Ujike, Acquisition of capillary pore structure by X-ray
evaluation of tensile strength and energy absorbing properties of cement ct and visualization of flow by numerical analysis, Adv. Mater. Res. 845 (2014)
mortar reinforced with micro- and meso-sized carbon fibers, Constr. Build. 163–167.
Mater. 44 (7) (2013) 751–756. [30] A. Bhasin, A. Izadi, S. Bedgaker, Three dimensional distribution of the mastic in
[5] Michael F. Petrou, David Parler, Kent A. Harries, Dimitris C. Rizos, asphalt composites, Constr. Build. Mater. 25 (10) (2011) 4079–4087.
Strengthening of reinforced concrete bridge decks using carbon fiber- [31] Zhenjun Wang, Jingjing Xiao, Evaluation of air void distributions of cement
reinforced polymer composite materials, J. Bridge Eng. 13 (5) (2008) 455–467. asphalt emulsion mixes using X-ray computed tomography scanner, J. Test.
[6] Eva Mnahoncakova, Roman Vejmelka, Milena Jirickova, Robert Cerny, Pavla Eval. 40 (3) (2012) 485–490.
Rovnanikova, Patrik Bayer, Thermal and hygric parameters of carbon-fiber- [32] Kalehiwot Nega Manahiloh, Balasingam Muhunthan, Masoud Kayhanian,
reinforced cement composites after thermal and mechanical loading, J. Seyoum Yami Gebremariam, X-ray computed tomography and
Building Phys. 29 (2) (2005) 121–143. nondestructive evaluation of clogging in porous concrete field samples, J.
[7] H. Manuela Chiarello, Raffaele Zinno, Electrical conductivity of self-monitoring Mater. Civil Eng. 24 (8) (2012) 1103–1109.
CFRC, Cem. Concr. Compos. 27 (4) (2005) 463–469.