Journal Pre-proofs

Compressive and flexural properties of ultra-high performance fiber-reinforced

cementitious composite: the effect of coarse aggregate

Fanghong Wu, Lihua Xu, Yin Chi, Yanqin Zeng, Fangqian Deng, Qian Chen

PII: S0263-8223(19)34534-9
DOI: https://doi.org/10.1016/j.compstruct.2019.111810
Reference: COST 111810

To appear in: Composite Structures

Received Date: 30 November 2019

Accepted Date: 13 December 2019

Please cite this article as: Wu, F., Xu, L., Chi, Y., Zeng, Y., Deng, F., Chen, Q., Compressive and flexural properties
of ultra-high performance fiber-reinforced cementitious composite: the effect of coarse aggregate, Composite
Structures (2019), doi: https://doi.org/10.1016/j.compstruct.2019.111810

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover
page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will
undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing
this version to give early visibility of the article. Please note that, during the production process, errors may be
discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

© 2019 Published by Elsevier Ltd.

 1 Compressive and flexural properties of ultra-high performance
2 fiber-reinforced cementitious composite: the effect of coarse aggregate
3 Fanghong Wu, Lihua Xu , Yin Chi, Yanqin Zeng, Fangqian Deng, Qian Chen
4 School of Civil Engineering, Wuhan University, Wuhan 430072, China

5 Abstract: The inclusion of coarse aggregate in the cementitious composite can improve the
6 mechanical property and simultaneously curb the cement demand thus lowering the production cost
7 in concrete built infrastructure. This study investigates the effects of different coarse aggregate types
8 and contents on the compressive and flexural properties of ultra-high performance fiber-reinforced
9 cementitious composite (UHPFRC). Four types of coarse aggregates (granite, basalt, limestone and
10 steel slag) and four contents denoted by the volume of mortar replacement (0%, 15%, 30% and 40%)
11 are the main variables. The test results have shown a significant increase in compressive strength and
12 elastic modulus of UHPFRC after moderate addition of coarse aggregate with only an expense of a
13 slight reduction in the flexural strength and post-peak ductility. The aggregate types having different
14 stiffness can be effective in either increasing the compressive strength or enhancing the post-peak
15 ductility, which is also true for the increase in aggregate content. With respect to the flexural
16 behavior, it is found that coarse aggregate type significantly affects the flexural strength and
17 equivalent flexural strength ratio, whilst no considerable variations in the first crack strength is
18 observed. Furthermore, an analytical model is proposed to generate the complete compressive
19 stress-strain curves of UHPFRC.

20 Keywords:
21 Ultra-high performance fiber-reinforced cementitious composite (UHPFRC); coarse aggregate;
22 compressive properties; flexural behavior; stress-strain relationship.

23 1. Introduction
24 Ultra-high performance fiber-reinforced cementitious composite (UHPFRC) is a novel cement-

Corresponding authors.
E-mail addresses: xulihua@whu.edu.cn (L.H. Xu), yin.chi@whu.edu.cn (Y. Chi).

1
 1 based materials, which is characterized by ultra-high strength, superior ductility and durability [1].
2 Currently, major applications of UHPFRC can be found from large-span bridges, critical civil
3 infrastructures [2] to military applications and nuclear plant, where high strength, good ductility,
4 durability and the resistance to blast and impact loading are urgently demanded [3,4]. It is widely
5 acknowledged that the ultra-high performance concrete (UHPC) or reactive powder concrete (RPC)
6 is produced with a high cement content, supplementary mineral admixtures (SMAs) including silica
7 fume, fly ash, granulated blast furnace slag (GBFS), metakaolin etc., quartz sand, superplasticizer
8 and steel fibers [5]. Moreover, it is also reported that some special curing techniques such as the hot
9 curing, steam curing and autoclaved curing can further improve the strength and homogeneity of
10 UHPC/RPC, as well as reduce the volume and size of void in the matrix [6,7]. However, the
11 production process would require exquisite selection of raw ingredients and decent curing condition
12 which is impractical in widespread application and particularly is of high cost and energy intensity in
13 manufacturing process [8] .
14 Coarse aggregate (CA) inclusion provides a novel route to reduce the demand of UHPC for
15 cement and other expensive SMAs in the construction industry without significant sacrifice in
16 compressive strength of UHPC. On the one hand, addition of CA can decrease the production cost of
17 UHPC, as it reduces the consumption of cementitious materials in unit volume [9]. On the other
18 hand, it is found that CA improves the autogenous shrinkage resistance [10] as well as increases the
19 compressive strength and elastic modulus of UHPC [11].
20 A large number of previous studies have claimed that the CA intrinsic features, e.g., content,
21 type, size, shape, grading, strength and quality, have considerable effect on the mechanical properties
22 of normal strength concrete (NC) [12,13], high strength concrete (HSC) [14,15] and self-compacting
23 concrete (SCC) [16,17]. The consistent conclusions regardless of HSC and SCC highlight that the
24 weaker aggregate is likely to cause a lower concrete strength and elastic modulus. The compressive
25 and splitting strength is increased as the aggregate strength increased, and the weaker aggregate may
26 initiate a more ductile behavior of concrete than the counterparts of the stronger aggregate.
27 With the widespread application of UHPC, the compressive strength, tensile strength, flexural
28 behavior and the stress-strain relationship of UHPC with different steel fiber characteristics, e.g.,
29 shape, length, diameter and content, and SMAs have been extensively investigated [18-20]. Besides,

2
 1 other findings have verified that the distribution and orientation of steel fiber also have a remarkable
2 effect on mechanical properties of UHPC [21,22]. Specifically, the influence of fiber characteristics
3 on compressive stress-strain relation of UHPC are also explored [23], where some empirical
4 expressions, considering the fiber characteristic parameters, are proposed to generate the complete
5 compressive stress-strain curve of UHPC [24]. Until recently, some available researches have
6 demonstrated that moderate dosage of CA can enhance the compressive strength, elastic modulus,
7 and high-temperature resistance of UHPC [9,25-27]. It is also interesting to note that CA inclusion
8 easy to cause the clustering and decrease the utilization efficiency of steel fiber, thus decreasing the
9 tensile and flexural behavior. The existing literatures have made preliminary investigation on a single
10 CA type and content on the mechanical properties of UHPC. However, to the best of the author’s
11 knowledge, there is limited information available on the compressive and flexural behavior of
12 UHPFRC with various CA types and contents, which the effects of CA type and content on the
13 evolution of compressive and flexural behavior of UHPFRC need further exploration.
14 The objective of this study is to investigate the influence of CA content and type on
15 compressive and flexural properties of UHPFRC. Compressive and flexural samples were fabricated
16 in conventional methods without special preparation techniques such as heat water or steam curing.
17 Four different CA contents by the volume of mortar and four CA types were considered to evaluate
18 the effects of CA content and type on compressive properties, such as compressive strength, elastic
19 modulus, strain at peak stress, compressive toughness ratio, as well as the uniaxial compressive
20 stress-strain relation, and flexural behavior of UHPFRC. Then, an analytical stress-strain model was
21 proposed to generate the complete compressive stress-strain curve of UHPFRC. Besides, the
22 scanning electron microscope (SEM) observation was performed for microstructural evaluations.

23 2. Experimental details
24 2.1. Raw materials and mix design
25 The raw materials for the fabrication of UHPFRC included Grade 52.5 of ordinary Portland
26 cement (OPC) with a fineness index of 380 m2/kg. Silica fume with an average particle size of
27 0.120.2 µm and a specific surface of 25 m2/g, and fly ash with an average particle size of 43 µm
28 and a bulk density of 2.34 g/cm3 were used. The chemical components of cement, silica fume and fly
3
 1 ash determined by X-ray fluorescence spectrometer (XRF, S4 Pioneer) are listed in Table 1. A
2 polycarboxylate-based superplasticizer (SP) with water-reducing efficiency of 30% and solid content
3 of 40% was used at a ratio of 2.53% by the mass of cementitious materials to improve the
4 flowability of fresh concrete. Copper plating straight steel fiber with a length of 13 mm and an aspect
5 ratio of 65 was used, which has a tensile strength of over 2200 MPa. Local river sand with maximum
6 size of 1.5 mm, a specific density of 2.65 g/cm3, and fineness modulus of 2.37 was used, and the
7 particle size distribution is shown in Fig. 1.
8 To evaluate the influence of CA type and content on compressive and flexural performance of
9 UHPFRC, four types of CA with particle size ranging from 5 mm to 16 mm were considered, i.e,
10 crushed granite, basalt, limestone aggregates and steel slag. The specific densities of these CAs were
11 2.72, 2.93, 2.65 and 3.22 g/cm3, respectively. The crushing indexes of the selected CAs, determined
12 in accordance with Chinese Standard GB/T 14685-2011 [28], were 7.70, 8.69, 11.73 and 14.65,
13 respectively. The pictures of CAs and their grading curves are shown in Fig. 2 and Fig. 1 (b),
14 respectively.
15 Table 1
16 Chemical components of cementitious materials (%).
Composition Na2O MgO Al2O3 SiO2 SO3 Fe2O3 P2O5 CaO K2O MnO ZnO SrO
Cement 0.079 2.14 4.5 19.58 3.06 3.119 0.128 64.94 0.75 0.127 0.024 0.148
Silica fume 0.068 0.224 0.354 92.87 1.26 0.113 0.11 0.213 0.332 0.008 0.019 0.005
Fly ash 0.552 0.575 30.63 48.74 0.706 2.611 0.247 2.44 1.25 0.016 0.013 0.060

17

100 100 Granite

Limestone
Percentage Passing (%)

Percentage Passing (%)

Basalt
80 80
Steel slag

60 60

40 40

20 20

0 0
0.01 0.1 1 10 1 2 4 8 16

Sieve size (mm) Sieve size (mm)

(a) Grading curve of river sand (b) Grading curves of coarse aggregates

4
 Fig. 1. Grading curves of fine and coarse aggregates

(a) Granite (b) Basalt (c) Limestone (d) Steel slag

2 Fig. 2. Four types of coarse aggregate
3 A total of thirteen branches of mixtures were designed, as presented in Table 2. Four series of
4 UHPFRC incorporating four types of CA were named as UG, UB, UL and US, respectively. In
5 addition, with respect to each series of UHPFRC mixtures considered three CA replacement levels
6 by the volume of mortar (15%, 30% and 40%). For example, regarding to the specimens containing
7 granite aggregate were designated as UG15, UG30 and UG40, respectively, corresponding to the CA
8 replacement levels of 15%, 30% and 40%. The referenced mixture without CA, named as UR00, was
9 designed. Additionally, the water-binder ratio was set at 0.18 and the steel fiber volume fraction was
10 fixed at 2%.
11 Table 2

12 Mixture proportions of UHPFRC.

Unit weight (kg/m3)

Mix No. Replacement level
Cement Silica fume Fly ash Water SP Steel fiber River sand CA
UR00 660.7 152.4 203.3 182.9 25.4 157 1118.1 0 0
UG15 554.9 128.1 170.8 153.7 21.3 157 939.2 408 15%
UG30 477.4 110.2 146.9 132.2 22.0 157 807.9 816 30%
UG40 406.9 93.9 125.2 112.5 18.8 157 688.6 1088 40%
UB15 554.9 128.1 170.8 153.7 21.3 157 939.2 425 15%
UB30 477.4 110.2 146.9 132.2 22.0 157 807.9 850 30%
UB40 406.9 93.9 125.2 112.5 18.8 157 688.6 1132 40%
UL15 554.9 128.1 170.8 153.7 21.3 157 939.2 398 15%
UL30 477.4 110.2 146.9 132.2 22.0 157 807.9 795 30%
UL40 406.9 93.9 125.2 112.5 18.8 157 688.6 1060 40%
US15 554.9 128.1 170.8 153.7 21.3 157 939.2 483 15%
US30 477.4 110.2 146.9 132.2 22.0 157 807.9 988 30%
US40 406.9 93.9 125.2 112.5 18.8 157 688.6 1288 40%
13 Remarks: the capital letters G, B, L, S in the specimen numbering denote the initial letter of granite, basalt,

5
 1 limestone and steel slag, respectively, R represents the mixture without CA. The number is the CA replacement

2 level.

3 2.2. Specimen fabrication

4 The UHPFRC specimens were prepared using a vertical planetary mixer with the maximum
5 speed of 120 rpm. The raw materials were weighed and placed into the mixer in the following order:
6 firstly, the river sand and CA were dry-mixed 3 min. Secondly, cement, silica fume and fly ash were
7 added to mix for 2-3 min to obtain a good dispersion of dry components. Thirdly, the water with
8 75% superplasticizer was added to mixer for mixing another 5 min, until a visible flowability was
9 obtained. Finally, the steel fibers were evenly distributed into the paste, and the remained
10 superplasticizer was added to continuously mixing another 3 min, until the steel fibers were
11 uniformly dispersed in fresh concrete.
12 The fresh paste was cast into the plastic molds, and compacted on a vibrating table about 1 min
13 to consolidate the paste. Meanwhile, the flowability of fresh concrete was measured. Then, all molds
14 were covered by plastic sheets to prevent moisture loss before demolding, and kept in laboratory at
15 room temperature. After 24 hours, all samples were demolded and wet cured in a curing tank at room
16 temperature until testing. Prior to the testing, the specimens were air-dried for 1 day. Specifically, for
17 each mixture, four batches of cubic specimens with the edge length of 100 mm were fabricated for
18 compressive strength test. A group of prisms with the dimension of 100×100×300 mm were prepared
19 for axial compressive stress-strain behavior test. One group of beam samples with the side length of
20 100 mm and length of 400 mm were cast for four-point bending test.

21 2.3. Testing procedures

22 2.3.1. Flowability testing
23 The flowability of all mixtures was measured by a standard slump flow test immediately after
24 mixing according to CECS 13:2009 [29]. A cone with an upper diameter of 100 mm, a bottom
25 diameter of 200 mm and a height of 300 mm was placed horizontally on a smooth plate. The fresh
26 UHPFRC was cast into the cone in three layers. Then, the cone was left vertically upward to allow
27 the fresh concrete to flow freely. The maximum diameter and the corresponding perpendicular
28 diameter of fresh concrete were measured, where the spread refers to the average value of the
6
 1 measured diameters.
2 2.3.2. Compressive strength testing
3 The cubic compressive strength tests were conducted using a universal testing machine (UTM)
4 with the maximum load capacity of 3000 kN. The compressive strength was measured at 3, 7, 28 and
5 90 days after the initial casting to quantify the acquisition of compressive strength with curing time.
6 The tests were determined in accordance with CECS 13:2009 [29] at a constant loading rate of 1.2
7 MPa/s.
8 To investigate the effects of CA content and type on uniaxial compressive stress-strain behavior
9 of UHPFRC, three prisms for each mixture were tested using a 2000 kN INSTRON 1346 testing
10 machine, the test procedure was in accordance with CECS 13:2009 [29]. The prismatic strength,
11 strain at peak stress, elastic modulus, and the stress-strain curves were measured. Two linear variable
12 displacement transducers (LVDTs), with a gage length of 20 mm, were vertically mounted at the
13 symmetrical location of specimen to obtain the axial deformation during the test. Displacement
14 control with a constant loading rate of 0.02 mm/min was implemented to acquire the complete
15 stress-strain curve of UHPFRC. The data of load was automated recorded by the testing system with
16 an acquisition frequency of 2 Hz, and the displacements of LVDTs were collected by DH 3816N
17 data acquisition system at the same frequency. The details of the test setup are plotted in Fig. 3. Both
18 ends of prismatic specimens were polished at a cutting machine before testing to ensure that the two
19 surfaces are sufficiently smooth and parallel, and to eliminate the effect of stress concentration at the
20 corner of the specimen on the stress-strain behavior in the test. Once the axial shortening of the
21 specimen attained 5 times of the deformation at peak load or reached up to 4.5 mm, the test was set
22 to terminate.

Load cell
Rigid block
LVDT

Loading
frame

Loading plate
Hydraulic
cylinder
23
7
 1 Fig. 3. Stress-strain test setup.

2 2.3.3. Flexural behavior testing

3 The flexural behavior of UHPFRC samples were measured by four-point bending testing using
4 a UTM with the load capacity of 300 kN. Two LVDTs with a gauge length of 50 mm were installed
5 on the both sides of sample, using a customized aluminum fixture, to record the mid-span deflection.
6 The clear span between the two supports was 300 mm, as shown in Fig. 4. The flexural test was
7 conducted by displacement control mode at a constant rate of 0.5 mm/min to obtain the complete
8 post-peak behavior of UHPFRC. The sampling modes of load and displacement were same as these
9 of stress-strain test.

Load sensor

Loading plate
100
Aluminum
fixture
100 LVDT

Suppor t
300 MTS testing
machine
10

11 Fig. 4. Four-point bending test setup

12 2.3.4. SEM observation

13 To reveal the interaction mechanism of CA type on compressive and flexural responses of
14 UHPFRC, SEM observation was conducted to investigate the microstructures of the hardened matrix
15 as well as the interfacial transition zone (ITZ) between CA and matrix using QUANTA 200. The
16 samples for SEM observation were taken from the small pieces after compressive strength tests.
17 Before imaging, the samples were dried for 6 hours at 80 ℃ in an oven, and then vacuumed and
18 plated with gold powder in an ion spatter. During the observation, the acceleration voltage was set to
19 15 kV.

8
 1 3. Results and discussion
2 3.1. Flowability of fresh UHPFRC
3 Fig. 5 shows the effects of CA content and type on flowability of fresh UHPFRC. It is seen that
4 the spread decreases as an increase in CA content. The spread value of the referenced mixture
5 (UR00) without CA was 345 mm. For the mixture containing crushed granite, in comparison to the
6 UR00, the spread was decreased by 21.7%, 30.4% and 36.2% corresponding to 15%, 30% and 40%
7 inclusion, respectively. This phenomenon is also confirmed by literature [5,26] that increasing the
8 CA content would decrease the flowability of UHPC. The reduction of flowability associated with
9 CA content was certainly due to the reduced content of cementitious materials as a lubricant for the
10 fine and coarse aggregates in fresh concrete [30]. In the meantime, the higher content of CA
11 produces a higher internal fraction between CA and paste because of the interlocking within the
12 angular CAs [31]. In addition, CA is more likely to cause the agglomeration of steel fibers around
13 CAs owing to the interlocking between fibers and CA, hence resulting in decreased flowability [32].
14 In contrast, the CA type had a mild effect on the flowability of UHPFRC. This is ascribed to the
15 similar particle size distributions and surface features of aggregates, especially for the crushed
16 granite, basalt and limestone aggregates. It is worth mentioning that the flowability of UHPFRC
17 containing steel slag was slightly greater than the counterparts of the other mixtures at a high CA
18 content. This may be attributed to the spherical surface and the preferable smoothness of steel slag,
19 which decreases the friction and interlocking effect.

36
Granite
Basalt
33
Limestone
Flowability (mm)

Steel slag
30

27

24

21
0 15% 30% 40%

20 CA replacement level

21 Fig. 5. Effects of CA content and type on flowability of fresh UHPFRC

9
 1 3.2. Effects of CA content and type on compressive properties of UHPFRC
2 3.2.1. Cubic compressive strength
3 The averaged cubic compressive strengths development curves for mixtures with various CA
4 types at a fixed 30% content are typically presented in Fig. 7. As expected, the cubic compressive
5 strengths of UHPFRC mixtures increase with curing time regardless of the type and content of CA.
6 The strength development of all specimens in the first week was considerably greater than that of
7 later period with 7-d compressive strength nearly reached up to 7080% of the 28-d strength.
8 Thereafter, a slight increment in compressive strengths at 90-d was observed in comparison with the
9 strengths at 28-d.
10 It is noted that there were no significant differences in the early compressive strengths among
11 UHPFRC with various CA types, especially for 3-d strengths. This is reasonable, since the hydration
12 of cement is incomplete during the early age that causes the strengths of hardened paste and
13 aggregate-matrix (A-M) ITZ significantly lower than that of CA. Thus, the early strength of
14 UHPFRC mainly depends on the strengths of hardened paste and A-M ITZ, rather than the intrinsic
15 strength of CA.
16 The effect of CA type on the compressive strength becomes distinct starting from 7-d curing
17 age. The highest compressive strength was obtained by the UHPFRC specimens containing granite
18 aggregate, while the lowest strength was measured by specimens prepared with steel slag at 28-d and
19 90-d. With respect to the mixtures at a fixed 30% content, the 28-d compressive strength of the
20 specimen prepared with crushed granite, basalt and limestone aggregates were 134.4, 126.3 and
21 122.0 MPa, respectively, which were 17.5%, 10.5% and 6.5% higher than that of the mixture with
22 same content of steel slag. This amplitude of increase conforms to the ranking of aggregate crushing
23 index values. Generally, the cracking patterns of the concrete under compression load primarily
24 emerge following: a) in the matrix, b) within the coarse aggregate, c) at ITZ between matrix and
25 aggregate [15,16]. With regard to UHPFRC, owing to the substantial strength improvement in the
26 hardened paste and ITZ, which is considered as higher as the intrinsic strength of CA [13], the failure
27 pattern under compression demonstrated that the fracture plane not only propagates via the hardened
28 paste or ITZ, but also passes through the aggregates, as shown in Fig. 6 [15]. Thus, the later
29 compressive strength of UHPFRC depends strongly on the quality of CA. According to the results of
10
 1 crushing index test, the order of crushing index values are ranked as: crushed granite (7.70)< basalt
2 (8.69)< limestone (11.73)< steel slag (14.65). The lower index value, the stronger aggregate.
3 Therefore, the compressive strength of UHPFRC specimens prepared with steel slag were
4 significantly lower than the others.

CA cr acking CA cr acking CA cr acking CA cr acking

(a) Granite (b) Basalt (c) Limestone (d) Steel slag

5 Fig. 6. Failure mode of cubic samples

6 Fig. 8 shows the compressive strengths at 28-d of UHPFRC samples with various types and
7 contents of CA. Regardless of the aggregate type, increasing the CA content from 0% to 30% leads
8 to a distinct increase in compressive strength. Since the aggregates create a rigid skeleton in the
9 hardened paste and restrict the material deterioration [8,11]. However, a lower strength of UHPFRC
10 mixture was obtained when the CA content increased from 30% to 40%. The reduction of strength
11 may be ascribed to the substantial increase of CA content significantly reduces the flowability of
12 UHPFRC, thus resulting in the increase of defects within concrete.

150 150
Granite
Compressive strength (MPa)
Compressive strength (MPa)

Basalt
135 Limestone 120
Steel slag
120 90

105 60 Granite
Basalt
Limestone
90 30
Steel slag
Without CA
75 0
3d 7d 28d 90d 0 15% 30% 40%
Curing time Coarse aggregate content

Fig. 7. Compressive strength development of UHPFRC Fig. 8. Effects of CA type and content on 28-d

with 30% CA compressive strength

13 3.2.2. Uniaxial compressive stress-strain relations

14 The representative stress-strain relation of UHPFRC under uniaxial compression is
11
 1 schematically shown in Fig. 9. The complete stress-strain curve can be divided into five stages:
2 (1) Linear elastic stage (OA). The stress increases linearly to 7085% of the peak stress, which
3 is strongly different from NC corresponding to 4050%. At this stage, the concrete can be regarded
4 as an elastic state, the responses of all specimens are nearly identical and the bridging effect of steel
5 fibers is not motivated. The type and content of CA have limited influence on material behavior at
6 this stage.
7 (2) Micro-cracks propagation stage (AB). During this stage (about 8095% of the peak stress),
8 the curve gradually deviates from the initial straight line, and the slope of curve decreases slowly.
9 The internal micro-cracks or pores in matrix began to propagate into fine cracks. However, due to the
10 existence of CA at the tips of fine cracks, it can moderately restrain the fine cracks from propagating.
11 The aggregate type has a mild impact on stress-strain curves under fixed CA contents, while the CA
12 content has a huge effect on the curves. Since at this stage the stress is relatively small, and is lower
13 than the strength of aggregates, the effect of CA with a higher content is more significant in resisting
14 micro-crack development. Therefore, a higher elastic modulus can be obtained for the specimens
15 incorporating higher CA content.
16 (3) Macrocrack propagation stage (BC). The micro-cracks evolve into multiple fine cracks and
17 then gather into a visible shear crack at the corner of the specimen when the stress reached the peak
18 stress. Some CAs in the cracking zone are fractured, where the steel fibers can bridge the cracks and
19 confine the specimen, hence, the sudden failure of specimen is effectively suppressed. During this
20 stage, the specimen behaves strain hardening behavior until the peak stress. Notably, a higher peak
21 stress and smaller peak strain can be obtained for the UHPFRC specimen incorporating stiffer CA.
22 (4) Unstable cracking stage (CD). After the stress exceeds the peak stress, a progressive strain
23 softening behavior can be observed at the post-peak stage. During this stage, the stress decreases
24 rapidly as the multiple fine cracks quickly coalesce into macrocracks, along with a severer concrete
25 crushing. A shear plane with an inclination angle of 4060 to the horizontal direction is produced,
26 where some CAs in this zone are fractured. At the end of this stage, some steel fibers are pulled out
27 from the matrix, which is reflected by the discontinuity point on the stress-strain curve.
28 (5) Stable cracking stage (DE). After the stress decreases to the inflection point (D point) on the
29 curve, the cracking development of the failure plane tends to stable, and a flat descending branch of

12
 1 stress-strain curve is obtained. At this stage, the macro-crack is gradually widened with an
2 occurrence of continuous fiber pullout. The stress decreases slowly, while the strain increases
3 rapidly.

4
5 Fig. 9. Typical stress-strain curve of UHPFRC under uniaxial compression
6 The complete compressive stress-strain curves for UHPFRC samples prepared with various
7 types and contents of CAs are illustrated in Fig. 10. It can be observed that the CA type had an
8 indiscernible effect on the stress-strain curves in the ascending segment at an identical CA content.
9 However, it is noted that the post-peak stage for the samples incorporating steel slag were flatter than
10 others.
11 For the effect of CA content, it is evidently altering both the ascending and descending stages of
12 the stress-strain curves. The variation of stress-strain curves with different CA contents for
13 specimens in UG series are shown in Fig. 10(d). The incorporation of CA could strengthen the initial
14 stiffness of UHPFRC, which the specimens incorporating CA had higher peak strengths and smaller
15 peak strains. Specifically, the sample with 30% CA had the highest peak stress and slope at the
16 ascending stage. Moreover, the moderate incorporation of CA could strengthen the post-peak ductile
17 behavior, the increasing of CA content would result in a smoother descending stage where the best
18 ductility was achieved by the mixture with 40% CA. In the contrary, a steep descending branch was
19 observed for the control specimen without CA.

13
 120
100 Granite Granite
Basalt 100 Basalt
Limestone Limestone
80
Steel slag 80 Steel slag

Stress (MPa)
Stress (MPa)

60
60

40 40

20 20

0 0
0 3 6 9 12 15 0 3 6 9 12 15
 
Strain ( ) Strain( )
(a) CA content of 15% (b) CA content of 30%

120
Granite 100 UR00
100 Basalt UG15
Limestone UG30
80
80 Steel slag UG40
Stress (MPa)
Stress (MPa)

60
60

40 40

20 20

0 0
0 3 6 9 12 15 0 3 6 9 12 15
 
Strain( ) Strain ( )

(c) CA content of 40% (d) UHPFRC with different CA contents

1 Fig. 10. Stress-strain curves of UHPFRC samples under uniaxial compression

2 3.2.3. Failure modes

3 The typical failure modes of prismatic UHPFRC samples under uniaxial compression are shown
4 in Fig. 11. It can be seen that all the prismatic specimens presented shear failure, the type and
5 content of CA had little effect on failure mode. Before the load reached the peak load, no visible
6 cracks appeared on the surface of the specimen. Once the peak load was reached, a visible crack was
7 produced firstly at the corner of specimen with a slight explosion, and then diagonally expanded
8 through the section to form a shear plane rather than a vertical splitting plane [24]. The inclination
9 angle of the shear plane along the horizontal direction was in the range of 4060. As the deflection
10 increased, the wedge blocks would cause a relative slip along the shear plane, but the steel fibers
11 could prevent this shear sliding due to the bridging and anchoring effect. Some steel fibers on the
12 shear surface were pulled out from the matrix with increasing straining, which led to a gradual loss

14
 1 of its load bearing capacity rather than the abrupt deterioration of specimen. After termination of test,
2 all samples were still maintained as integrity. It should be mentioned that all CAs on the shear
3 surface were fractured during failure process, as shown in Fig. 11(e).

(a) UG30 (b) UL30 (c) US30 (d) UB30 (e) CA cracking

4060

CAs cracking
4

5 Fig. 11. Failure modes of UHPFRC samples under uniaxial compression

6 3.2.4. Elastic modulus

7 The elastic modulus is an essential parameter of concrete material for the design and analysis of
8 structure, which can be calculated based on the stress-strain curve, given as [33]:

0.4  f ck  f1
9 Ec  (2)
 2  0.00005

10 where fck is the peak stress, f1 is the stress corresponding to the strain of 50 με , ε2 represents the strain

11 when the stress reaches 40% of the peak stress.

12 The elastic modulus of UHPFRC samples containing various types and contents of CA is shown
13 in Fig. 12. It is indicated that the CA type has obvious effect on the elastic modulus of UHPFRC.
14 The difference of the elastic modulus for the specimens prepared with granite and limestone
15 aggregates at an identical CA content is small, which is slightly higher than those of steel slag and
16 basalt concrete, with the lowest elastic modulus obtained steel slag, which were 35.51 GPa, 36.02
17 GPa and 40.87 GPa, respectively, when 15%, 30% and 40% steel slag were incorporated. The higher
18 elastic modulus of specimen may be attributed to the higher stiffness of these aggregates and the
19 better continuous granular skeleton of limestone aggregate. Whilst, the lowest elastic modulus of that
20 with steel slag owing to the soft nature of steel slag, which the steel slag has the highest crushing
21 index. Moreover, it is observed that the elastic modulus was increased by 27.49%, 55.74% and
15
 1 53.99%, respectively, as 15%, 30% and 40% granite aggregate were added, compared with that of
2 the control mixture without CA. This can be ascribed to the fact that CA creates a rigid skeleton in
3 the hardened matrix [11], which can restrain the material deterioration of UHPFRC specimens during
4 the loading, thus producing a higher elastic modulus.

50 50

40 40

Elastic modulus (GPa)

Elastic modulus (GPa)

30 30

46.09 45.57
20 Granite 20 37.73
Basalt
Limestone
29.59
10 Steel slag 10

0 0
15% 30% 40% 0 15% 30% 40%
Coarse aggregate content Coarse aggregate content

(a) UHPFRC with different types of CA (b) UHPFRC with different granite aggregate contents

5 Fig. 12 Elastic modulus of UHPFRC with different CA types and contents

6 To date, many models have been proposed to predict the elastic modulus of normal and high
7 strength concrete, as well as UHPC, as listed in
8 Table 3. We attempted here to evaluate the applicability of those models for predicting the
9 mechanical properties of UHPFRC in terms of the 28-d prismatic compressive strength and the
10 elastic modulus , as listed in Table 4. It is clear that the predictions of elastic modulus given by ACI
11 2014 [34] and CEB-FIB 95 [35] overestimate the test values, due to these codes are usually used to
12 predict the Ec of NC or HSC. The predicted values of Ec from GB 50010-2010 [36] and Graybeal [23]
13 are slightly lower than the experimental values, especially for the specimens containing a high CA
14 content. The predicted formula of Ec proposed by An [18] overestimates the measured values, as the
15 equation is not appropriate for a wide strength range of UHPC, it is only reliable to predict the Ec of
16 fine-grained UHPC and UHPFRC with the compressive strength ranging between 199 and 218 MPa.
17 The predictions Ec proposed by Ma [9] provide the best fitting with the test results. Thus, based on
18 the formulas given by Graybeal [23] and Ma [9], the following equations are proposed to predict the
19 Ec of UHPFRC in this study. The ratios of the predicted values to the experimental values are
20 presented in Table 4. The predictions can correlate well with the experimental data.

16
1 Ec  4202 f ck (3)

Ec  8948   f ck 
1/3
2 (4)

3 where fck represents prismatic compressive strength (MPa).

4

5 Table 3

6 The proposed equations of elastic modulus

Literature Equation of Ec (MPa) Description

Normal concrete, f c  41.4 MPa

'
ACI 318-14 [34] Ec  4730 f c'

Ec  21500   ( f c'  8) /10 

1/3
CEB-FIB [35] f c'  80 MPa

GB 50010-2010 [36] Ec  105 /  2.2  34.7 / f cu,k  Normal concrete

UHPC without coarse aggregate,

Graybeal [23] Ec  3840 f '
c
126  f c'  193 MPa

Ec  8800  f c' 
1/3
Fine-grained UHPC, 150  f c  180 MPa
'
Ma et al. [9]

Ec  9075  f c' 
1/3
Fine- grained UHPC, 199  f c  218 MPa
'
An et al. [18]

'
7 Remarks: Ec represents the elastic modulus, f c denotes the cylinder strength.

8 Table 4

9 Experimental results of UHPFRC

Mixture fck Ec 0 ACI CEB- GB Graybeal Ma An Equation Equation
code (MPa) (GPa) (µε) 318-4 FIP 50010 (3) (4)
UR00 81.43 29.59 3198 1.44 1.46 1.29 1.06 1.29 1.33 0.78 0.76
UG15 94.51 37.73 2547 1.22 1.20 1.03 0.90 1.06 1.10 0.92 0.90
UG30 98.03 46.09 2834 1.02 1.00 0.85 0.75 0.88 0.91 1.11 0.96
UG40 96.62 45.57 2694 1.02 1.00 0.86 0.75 0.89 0.91 1.10 0.90
UB15 96.83 37.15 2837 1.25 1.23 1.05 0.92 1.09 1.12 0.90 1.16
UB30 104.08 42.44 2611 1.14 1.11 0.93 0.84 0.98 1.01 0.99 1.12
UB40 106.32 46.02 2642 1.06 1.03 0.86 0.78 0.91 0.93 1.06 1.01
UL15 87.08 37.93 3091 1.16 1.17 1.01 0.86 1.03 1.06 0.97 0.93
UL30 94.52 46.49 2827 0.97 0.96 0.83 0.71 0.85 0.88 1.16 1.11
UL40 92.24 45.28 2700 1.00 1.00 0.86 0.74 0.88 0.91 1.12 0.90
US15 84.41 35.51 2924 1.22 1.23 1.08 0.90 1.09 1.12 0.92 1.12
US30 89.66 36.02 2917 1.24 1.24 1.07 0.91 1.09 1.13 0.91 1.01

17
 US40 91.51 40.87 2751 1.11 1.10 0.95 0.82 0.97 1.00 1.01 1.09
Mean value 1.14 1.13 0.98 0.84 1.00 1.03 1.00 0.997
Standard deviation 0.132 0.141 0.132 0.098 0.124 0.129 0.110 0.117
1

2 3.2.5. Peak strain

3 Fig. 13 summarizes the tested strains at peak stress of UHPFRC mixtures for different contents
4 and types, which range from 2500 to 3200ε. It is generally observed that the incorporation of CA
5 reduces the peak strain, however, no clear trend can be identified with respect to the effect of CA
6 content.
7 The strain εc corresponding to peak stress of UHPFRC is also assessed by the abovementioned
8 literatures, as listed in Table 5. The comparisons of the test results with the predicted values of the
9 given models are shown in Fig. 13. It is seen that all the models underestimate the experimental

10 result of the control specimen without CA ( f ck  81.34 MPa,  0  3189με ). The models proposed

11 by Graybeal [23] and Guo et al. [37] overestimate the experimental results of UHPFRC, whilst the
12 predictions of An et al. [18] are prominently lower than the test results. The predicted values from
13 EC 4 [38] are slightly lower than these of experimental values. The predicted values given by
14 CEB-FIB [35] and Wang et al. [39] coincide well with the current test results. Therefore, the
15 equation in CEB-FIB [35] is recommended to estimate the strain at peak stress of UHPFRC.

16 Table 5

17 The predictions of strain at peak stress

Literature Predicted model of strain ε0 Description

 0  0.7  f 0 
0.31
CEB-FIB [35] f 0  120 MPa

 0  2  0.085  f 0  50 
0.53
EC4 [34] 50 ≤ f 0 ≤ 100 MPa

UHPC without coarse aggregate,

Graybeal [23]  0  f 0 /  5.97  0.26  f 0 
126  f 0  193 MPa

Fine-grained UHPC,
An et al. [18]  0  0.0257  f 0.96
0
199  f 0  218 MPa

Wang et al. [39] 

 0  0.5  1.95  0.01491 f 0  0.763  4 f 0
 f 0  200 MPa

18
 80 ≤ f 0 ≤ 150 MPa,
Guo et al. [37]  0   6.7264  f 0  2460.9  103
UHPC without coarse aggregate

3500 3500

3000
3000
2500

Peak strain c ()

Peak strain ()

2500 2000
Test results
2000 Granite 1500 CEB-FIB (1995)
Basalt EC2 (2004)
Limestone 1000 Graybeal (2007)
1500 Steel slag An et al. (2017)
500 Wang et al. (2016)
Without CA
Guo et al. (2017)
1000 0
0 15% 30% 40% 80 85 90 95 100 105 110
Coarse aggregate content Compressive strength fc' (MPa)

Fig. 13. Peak strain of UHPFRC mixtures with different Fig. 14. Comparison of peak strain with the predictions

contents and types of CA of the existing models

1 3.2.6. Compressive toughness ratio

2 Toughness is an index used to assess the energy dissipation capacity of the concrete structure
3 during the deformation, which can be calculated by integrating the area under the stress-strain curve
4 when the strain reaches a specific strain. The toughness ratio (TR), first proposed by Ezeldin and
5 Balaguru [40] and Nataraja et al. [41], is used to evaluate the effect of steel fiber characteristics on
6 toughness of fiber-reinforced concrete. It can be calculated by the ratio of toughness until the strain
7 reached 5 times of the peak strain to that of a rigid plastic material, as described in equation (5). The
8 ultimate concrete strain of 0.003 is adopted, recommended by ACI 318-14 code [34] for concrete
9 structure.

TF5
10 TR  (5)
0.015  f 0

11 where TR represents the toughness ratio, TF5 denotes the area under stress-strain curve up to five
12 times of the peak strain, f0 is the peak stress of the control mixture UR00.
13 Fig. 15 shows the compressive toughness ratios of UHPFRC mixtures calculated using Eq. (5).
14 As can be seen, the incorporation of CA strengthens the compressive toughness ratio, which
15 increases with an increase in CA content, except for the UHPFRC mixtures with steel slag. This can
16 be attributed to the combined effect of CAs and steel fibers [25,26]. Since the volume fraction of

19
 1 steel fiber remained constant, the addition of CA would result in a denser distribution of steel fibers
2 in unit area, especially around the aggregates. This might effectively inhibit the propagation of
3 cracks and deterioration of UHPFRC materials, and triggering the initiation of multi-cracks in the
4 matrix. From Fig. 10, it is observed that the sample containing steel slag has a steeper descending
5 stage for a higher CA content.
6 As shown, the aggregate type has a significant effect on compressive toughness ratio. The
7 UHPFRC specimens prepared with steel slag have a higher compressive toughness ratio compared to
8 these of the mixtures with other three types of aggregates at the CA content of 15% and 30%.
9 Likewise, the TR of UHPFRC specimens incorporating limestone is also greater than the mixtures
10 with granite and basalt aggregates. During the failure process of the samples under uniaxial
11 compression, the features of aggregate significantly influence the cracking behavior of concrete
12 [13,14]. Since the stiffness and strength of granite and basalt are significantly greater than those of
13 limestone and steel slag, the prior two aggregates are likely to generate stress concentration at the
14 interface of A-M, thus resulting in the earlier fracture of aggregate at post-peak stage. Whilst, the
15 limestone and steel slag are known to weaker than the other two aggregates, which the ductile
16 behavior at the post-crack stage is more accessible for UHPFRC samples prepared with weaker
17 aggregates.

0.6 0.6

0.5 0.5

0.4
Toughness ratio

0.4
Toughness ratio

0.3 0.3
0.564
0.2 0.442 0.45
0.2 0.395
Grante
Basalt
0.1 0.1
Limestone
Steel slag
0.0 0.0
15% 30% 40% 0% 15% 30% 40%
Coarse aggregate content Coarse aggregate content

(a) CA type (b) CA content

18 Fig. 15. Effects of CA type and content on compressive toughness ratio of UHPFRC

19 3.3. Prediction of stress-strain relationship under uniaxial compression

20 A precise analytical model used to predict the complete stress-strain curve of concrete is

20
 1 necessary for structural design and analysis. The existing empirical compressive stress-strain
2 relationships of concrete can be expressed as a polynomial, exponential, trigonometric function and
3 rational fraction [24,39-42]. Most of these models can precisely predict the compressive stress-strain
4 curves of plain and fiber-reinforced concrete or UHPC without CA. Carreira and Chu [43] adopted a
5 rational fraction to describe the stress-strain behavior of plain concrete under uniaxial compression,
6 as given by Eq.(6). Ezeldin and Balaguru [40] and Nataraja et al. [41] modified this model to
7 describe the stress-strain relationship of normal-strength steel fiber-reinforced concrete.

fc   c /  0 
8  (6)
fc   1   c /  0 
'

9 where fc and c represent the stress and strain, f c' is the peak stress, 0 is the strain corresponding to

10 peak stress,  denotes material parameter to describe the shape of stress-strain curve.
11 The stress-strain relationships of UHPFRC under uniaxial compression can be described by a
12 rational fraction based on the modified model given by Carreira and Chu [43]:

   c /  0 
 , 0  x 1
fc    1   c /  0 


13 
f ck  A   c /  0  (7)
, x 1
 A  1   /  B
 c 0

1
14 = , 0  x 1 (8)
f ck
1-
0E

15 where fck is the prismatic compressive strength, A and B are the material parameters to describe the
16 shape of stress-strain curves, used to consider the effect of aggregate content on descending branch,
17 E and ε0 are given by equation (3) and the equation in CEB-FIB [35], respectively.
18 The value of A can be used to describe the flatness of the descending stage, the larger of this
19 value means a steeper descending stage. As indicated in Fig.9, the shape of the descending branch
20 for UHPFRC mixtures relates to the CA content. The higher aggregate content produces a flatter
21 descending stage, however, the aggregate type, except for steel slag, has a small effect on the
22 descending stage. Hence, based on the experimental results, the values of A and B can be obtained
23 using a best fitting statistical analysis. The fitting results indicated that the value of A is related to CA

21
 1 content, but the value of B approximates a constant of 2.17 as follows:

2 A  0.17  2.56  Va , B  2.17 (9)

3 where Va represents the CA content (%).

4 The comparisons between the predictions of the proposed model with experimental full
5 stress-strain curves are exhibited in Fig. 16. By contrast, the predictions of the proposed model in
6 this paper are consistent well with the experimental curves, particularly at the descending branch.

90 105
Proposed model Proposed model
75 90 Test results
Test results
75
60

Stress (MPa)
Stress (MPa)

60
45
45
30
30
15 15

0 0
0 5 10 15 0 3 6 9 12 15
 
Strain (  Strain ( 

(a) UR00 (b) UG15

105
105 Proposed model
Proposed model
90 Test results
90 Test results
75
75
Stress (MPa)
Stress (MPa)

60
60

45 45

30 30

15 15

0 0
0 3 6 9 12 15 0 3 6 9 12 15
 
Strain (  Strain ( 

(c) UG30 (d) UG40

Fig. 16. Comparison of proposed model stress-strain curves with the experimental results

7 3.4. Effects of CA content and type on flexural properties of UHPFRC

8 3.4.1. Flexural behavior
9 Fig. 17 illustrates the effects of CA type and content on flexural responses of UHPFRC. As can
10 be seen, the curves can be divided into three segments:

22
 1 (1) Elastic stage: the load is linearly increasing with continuous deformation, the matrix and
2 steel fiber sustain the flexural load jointly at this stage, while the bond stress between them is
3 relatively small. The initial behaviors of all specimens with different types and contents of CAs are
4 similar.
5 (2) Cracking stage: after the load is increased up to the first cracking load, the load usually
6 exhibits a mildly drop with a slight cracking sound, followed by a nonlinear increase with the
7 increment of deformation. At this stage, the flexural load is mainly sustained by steel fibers through
8 the friction and mechanical interlock.
9 (3) Post-cracking stage: the load is gradually decreased as the deformation increases, and a
10 zig-zag shaped curve is shown, indicating that some steel fibers are gradually pulled out from matrix
11 along with a crisp click sound. At the descending stage, the flexural load is only undertaken by the
12 steel fibers, and the crack width is broadened with an increase in deformation. Many steel fibers in
13 the crack zone are gradually pulled out from matrix, thus, resulting in a lower available fiber amount
14 to sustain the flexural load.

Ⅰ- Elastic stage Ⅰ- Elastic stage

40 40 Ⅱ- Cracking stage
Ⅱ- Cracking stage
Ⅲ- Post-cracking stage Ⅲ- Post-cracking stage
30 Granite 30 Granite
Load (kN)

Load (kN)

Basalt Basalt
Limestone Limestone
20 Steel salg 20 Steel salg

10 10

Ⅰ Ⅱ Ⅲ Ⅰ Ⅱ Ⅲ
0 0
0 1 2 3 4 5 0 1 2 3 4 5
Displacement (mm) Displacement (mm)

(a) 15% CA (b) 30%CA

23
 40
Ⅰ- Elastic stage 40 UR00
Ⅱ- Cracking stage UG15
30 Ⅲ- Post-cracking stage UG30
30 UG40
Granite

Load (kN)
Load (kN)

Basalt
20 Limestone
20
Steel salg

10 10

0 Ⅰ Ⅱ Ⅲ
0
0 1 2 3 4 0 1 2 3 4 5
Displacement (mm) Displacement (mm)

(c) 40% CA (d) Effect of CA content on flexural behavior

1 Fig. 17. Effects of CA content and type on flexural load-displacement curves of UHPFRC

2 As illustrated in Fig. 17, all specimens exhibited good flexural ductility and deflection-
3 hardening after the initial cracking due to the incorporation of 2% steel fibers, which were
4 significantly different from that without steel fiber with a sudden drop after reaching the peak load
5 [6]. Overall speaking, the CA type and content have a small effect on the load-deflection (P-Δ)
6 curves at the pre-cracking stage, whilst significant effect on the cracking and post-cracking stages is
7 noted. Specifically, the UHPFRC specimens with steel slag possess the lowest peak load. While,
8 there is a distinct difference for the effect of other three CA types on P-Δ curves at different CA
9 contents. For instance, with respect to mixtures with 15% CA, the highest peak load is obtained by
10 the specimens containing basalt aggregate. Nevertheless, for the mixtures with 30% CA, the
11 specimen with granite aggregate has the highest peak load and the most voluminous descending
12 stage. Furthermore, Fig. 17(d) shows the influence of CA content on the P-Δ curves for specimens in
13 UG series. The CA content is found to have insignificant effect at elastic stage, however, the curves
14 start to deviate at the cracking and post-cracking stages. It is noticed that UG specimens with higher
15 CA content have a lower flexural strength and inferior ductility. The weakened post-cracking
16 flexural behavior is largely attributed to the reduction of steel fiber utilization efficiency after the
17 incorporation of CA [11].
18 3.4.2. Flexural strength of UHPFRC
19 The effects of CA content and type on the first crack and flexural strength of UHPFRC are
20 presented in Fig. 18. The first crack and flexural strengths are determined according to ASTM

24
 1 C1609 [44], where the first crack strength denotes the first-peak stress and the flexural strength refers
2 to the greatest value on the P-Δ curve, the strength under four-point flexural load can be calculated

3 by the following formula: f  PL / bh 2 , where P stands for flexural load, L refers the clear span

4 length, b and h represent the width and height of specimen section.

5 It can be seen from Fig. 18 that the influence of CA type on the first crack and flexural strengths
6 for UHPFRC mixtures is fluctuating. The CA type had a greater effect on flexural strength than that
7 of the first crack strength. Generally, when CA content was lower than 30%, UHPFRC mixtures with
8 basalt and granite aggregates have higher first crack and flexural strengths compared with that
9 containing limestone and steel slag. The mixtures with basalt aggregate inclusions show the highest
10 first crack strengths at both CA contents of 15% and 30%. The smallest first crack and flexural
11 strengths are observed in steel slag UHPFRC mixtures. However, the opposite results are observed
12 with respect to the UHPFRC mixtures with 40% CA content, where the first crack and flexural
13 strengths of the specimens containing steel slag and limestone are slightly higher than that containing
14 granite and basalt aggregates. In the meantime, the strength differences for specimens prepared with
15 various types of CAs are relatively small.
16 The addition of CA decreases the flexural strengths, in comparison to the referenced mixture
17 UR00, the flexural strengths of specimens were decreased by 19.9%, 9.9% and 30.7%, with the
18 incorporation of 15%, 30% and 40% granite aggregate, respectively. This may be ascribed to the
19 incorporation of CA decreases the utilization efficiency and bonding stress of steel fibers [26], where
20 the load after increased up to cracking stage, the flexural load is mainly sustained by the steel fibers.
21 Nevertheless, the effect of CA content on the first crack strength is not clear.

12
First crack strength (MPa)

10
Flexural strength (MPa)

10
8
8
6
6
4
Granite 4 Granite
Basalt Basalt
2 Limestone 2 Limestone
Steel slag Steel slag
0 0
15% 30% 40% 0 15% 30% 40%
Coarse aggreagte content Coarse aggreagte content

25
 (a) First crack strength (b) Flexural strength

1 Fig. 18. Effects of CA type and content on first cracking and flexural strength of UHPFRC

2 3.4.3. Equivalent flexural strength ratio

3 The equivalent flexural strength ratio, according to ASTM C1609 [44], can be used to reflect
4 the ability of energy consumption of fiber-reinforced concrete. The following equations are used to
5 calculate equivalent flexural strength ratios of the deflections up to L/300 and L/150:

300  T300
6 RT,300  (10)
f1bh 2

150  T150
7 RT ,150  (11)
f1bh 2

8 where RT,300 and RT,150 are the equivalent flexural strength ratios corresponding to the deflections of
9 L/300 and L/150, T300 and T150 represent the areas under the load-deflection curve when the
10 deflections reach up to L/300 and L/150, fl refers to the first crack load, b and h stand for the width
11 and height of specimen section, respectively.
12 Fig. 19 shows the effects of CA content and type on equivalent flexural strength ratio of
13 UHPFRC. It can be seen in Fig.19(a) that the incorporation of CA slightly strengthens the equivalent
14 flexural strength ratio corresponding to the deflection of L/300. However, the trend is not clear for
15 the effect of CA content on RT,300 with respect to the UHPFRC specimens containing different CA
16 types. This is attributed to the fact that the flexural load at pre-cracking stage is mainly resisted by
17 the matrix rather than the steel fiber [8]. However, the incorporation of CA could improve the initial
18 stiffness of concrete [26], hence, improving the ductility of concrete at initial elastic stage. As
19 evidently shown in Fig. 17, the areas under flexural load-deflection curves up to the deflection of
20 L/300 are primarily composed by the pre-cracking segment, and that of L/150 mainly includes the
21 pre-cracking, cracking and part of post-cracking stage.

26
 Equivalent flexural strength ratio, R150
Equivalent flexural strength ratio, R300

0.54
1.7
0.52
1.6

0.50
1.5

0.48 Granite
Granite 1.4
Basalt Basalt
Lime stone Lime stone
0.46 1.3 Steel slag
Steel slag
0 15% 30% 40% 0 15% 30% 40%
CA content CA content

(a) Up to the deflection of L/300 (b) Up to the deflection of L/150

1 Fig. 19. Effects of CA type and content on equivalent flexural strength ratios of UHPFRC

2 As illustrated in Fig. 19(b), for the case of deflection up to L/150, the equivalent flexural
3 strength ratio is approximately linearly decreasing with the increment of CA content. For instance,
4 the equivalent flexural strength ratio corresponding to L/150 of the reference mixture without CA is
5 1.72, which is decreased by 7.0%, 11.4% and 15.7%, respectively, when 15%, 30% and 40% granite
6 aggregates are added. This may be ascribed to the lower utilization efficiency of steel fibers due to
7 the inhomogeneous distribution of fibers caused by the increment of CA content. Furthermore, it is
8 also observed that the CA type considerably affected RT,150 of UHPFRC mixtures. With respect to
9 identical CA content, the lowest RT,15 values had been obtained for the mixtures prepared with steel
10 slag, while the highest values were observed for the mixtures containing basalt aggregate.

11 3.5. Microstructure investigation

12 Fig. 20 shows the SEM images of hardened paste, A-M ITZ and fiber-matrix ITZ. It indicates
13 that the hardened paste has a very dense microstructure for all UHPFRC mixtures prepared with
14 diverse types of aggregate, as shown in Fig. 20(e). Besides, there are some voids can be observed in
15 the hardened paste of granite and limestone UHPFRC samples. Many spherical fly ash particles,
16 bonded well with the hardened paste, can be found in the matrix. This can be ascribed to the low w/b,
17 high cement content, as well as the filling effect of silica fume and fly ash, generating a fairly
18 compact matrix. Moreover, the pozzolanic effect of fly ash can produce the second hydration, and
19 slightly increase the long-term strength of UHPFRC [45].
20 The SEM images of A-M ITZ indicate that the UHPFRC samples have a very compact

27
 1 interfacial zone regardless of the aggregate type, and there are no evident pores can be found in the
2 ITZ. Additionally, it can be seen that all the specimens also have a compact fiber-matrix ITZ. This
3 may be attributed to the hydrophilic surfaces of CA and steel fiber, thus the cement around these
4 surfaces can be comprehensively hydrated, so that the hydration products and the mineral admixtures
5 properly fill the space between the matrix and aggregates or fibers.
6 The distinctions of morphological features for different types of aggregates are illustrated in
7 Fig. 20. The texture of granite aggregate is more homogeneous, a distinct boundary between
8 aggregate and matrix can be observed compared with the other aggregates. Conversely, the steel slag
9 has the worst homogeneity. Whereas the bond of the matrix and steel slag is the closest, and some
10 hydration products enter the aggregate surface due to the hydrophilicity and coarser surface of steel
11 slag. This could be verified by the ITZ of steel slag and matrix, which has the least distinct interface
12 among the four types of aggregates. Additionally, steel slag has a plumper and spherical
13 morphology, nevertheless the other three types of aggregates equipped some sharp corners, easy to
14 generate stress concentration and better clamping effect with the matrix. Thus, the low stiffness and
15 plump shape of steel slag caused a lower compressive strength and better post-peak behavior of
16 UHPFRC specimens. This may be attributed to the formation of multiple cracks, leads to the
17 concrete material becoming more energy-consuming for the sample prepared with steel slag.

Matrix
Matrix ITZ Limestone
aggregate

ITZ
ITZ
Matrix
Granite aggregate Basalt aggregate

(a) Granite (b) Basalt (c) Limestone

28
 Matrix Matrix

Steel fiber

ITZ
Fly ash Fly ash ITZ
Steel slag
Matrix

(d) Steel slag (e) Hardened paste (f) ITZ of steel fiber and matrix

Fig. 20. SEM images of UHPFRC

1 4. Conclusions
2 This study investigated the effects of coarse aggregate content and type on compressive and
3 flexural properties of UHPFRC. Based on the experimental results presented in this study, the
4 following conclusions can be drawn:
5 (1) The aggregate type and content have a remarkable effect on the compressive strength and
6 elastic modulus of UHPFRC. The UHPFRC specimens prepared with stronger aggregate have higher
7 compressive strength and elastic modulus. Besides, the compressive strength increases with an
8 increase in CA content up to 30%, but decreases as the CA content further increases up to 40%. With
9 respect to UHPFRC, there is an optimal CA content to obtain a maximal compressive strength.
10 (2) The elastic modulus and the strain corresponding to peak stress are evaluated by compared
11 with the available models. New formulas are proposed to predict the elastic modulus of UHPFRC,
12 limiting to the cubic compressive strength ranges between 100 MPa and 150 MPa. The equation in
13 CEB-FIB can be used to estimate the strain corresponding to peak stress of UHPFRC.
14 (3) The CA type and content have an inconsiderable influence on the linear elastic stage
15 behavior of UHPFRC, but significant on the post-peak stage of the uniaxial compressive stress-strain
16 curve. The descending stages of steel slag UHPFRC specimens are remarkably flatter than these of
17 the other aggregate UHPFRC. A smoother descending stage of the stress-strain curve is expected for
18 the mixture with a higher CA content. An analytical model based on the formulas given by Carreira
19 and Chu is developed to predict the complete stress-strain relationship of UHPFRC under uniaxial

29
 1 compression.
2 (4) The effects of CA type and content on the compressive properties are greater than that on
3 flexural behavior. The CA type and content have a mild effect on pre-cracking flexural behavior,
4 while considerable effect on cracking and post-cracking responses. Overall, UHPFRC mixtures
5 prepared with basalt and granite aggregates have higher first-crack and flexural strengths when CA
6 content less than 30%. Besides, the incorporation of CA could mildly strengthen the equivalent
7 flexural strength ratio for the case of deflection to L/300, but significantly decreases the equivalent
8 flexural strength ratio corresponding to the deflection of L/150.

9 Data availability
10 The raw/processed data required to reproduce these findings cannot be shared at this time due to
11 technical or time limitations.

12 Declaration of Competing Interest

13 The authors declare that they have no known competing financial interests or personal
14 relationships that could have appeared to influence the work reported in this paper.

15 Acknowledgements
16 This work was supported by the Key Project of National Natural Science Foundation of China
17 (Grant NO. 51738011) and the National Natural Science Foundation of China (Grant No. 51978538).
18

19 References
20 [1] Yoo DY, Banthia N. Mechanical properties of ultra-high-performance fiber-reinforced concrete: A review.

21 Cem Concr Compos 2016;73:267-280.

22 [2] Yoo DM, Hwang HH, Kim ST. An Experimental Study on the shear connection for UHPC Deck Bridge.

23 Compos Res 2011;24:29-33.

24 [3] Li PP, Yu QL. Responses and post-impact properties of ultra-high performance fibre reinforced concrete under

25 pendulum impact. Compos Struct 2019;208:806-815.

30
 1 [4] Xu JC, Wu CQ, Xiang HB, Su Y. Behaviour of ultra high performance fibre reinforced concrete columns

2 subjected to blast loading. Eng Struct 2016;118:97-107.

3 [5] Wang C, Yang CH, Liu F, Wang CJ, Pu XC. Preparation of Ultra-High Performance Concrete with common

4 technology and materials. Cem Concr Compos 2012;34:538-544.

5 [6] Wu ZM, Shi CJ, He W. Comparative study on flexural properties of ultra-high performance concrete with

6 supplementary cementitious materials under different curing regimes. Constr Build Mater 2017;136: 307-313.

7 [7] Yazıcı YH, Yardımcı MY, Yiğiter H. Mechanical properties of reactive powder concrete containing high

8 volumes of ground granulated blast furnace slag. Cem Concr Compos 2010;32:639-648.

9 [8] Sobuz HR, Visintin P, Mohamed Ali MS, Sing M. Manufacturing ultra-high performance concrete utilising

10 conventional materials and production methods. Constr Build Mater 2016;111:251-261.

11 [9] Ma JX, Orgass M, Dehn F, Schmidt D, Tue N. Comparative investigation on ultra-high performance concrete

12 with and without coarse aggregates. In: Proceedings of the International Symposium on Ultra High

13 Performance Concrete, Kassel, Germany. September 2004. p. 221-228.

14 [10] Yoo DY, Kang ST, Lee JH, Yoon YS. Effect of shrinkage reducing admixture on tensile and flexural

15 behaviors of UHPFRC considering fiber distribution characteristics. Cem Concr Res 2013;54:180-190.

16 [11] Xu LH, Wu FH, Chi Y, Cheng P, Zeng YQ, Chen Q. Effects of coarse aggregate and steel fiber contents on

17 mechanical properties of high performance concrete. Constr Build Mater 2019;206:97-110.

18 [12] Meddah MS, Zitouni S, Belâabes S. Effect of content and particle size distribution of coarse aggregate on the

19 compressive strength of concrete. Constr Build Mater 2010;24:505-512.

20 [13] Wu KR, Chen B, Yao W, Zhang D. Effect of coarse aggregate type on mechanical properties of

21 high-performance concrete. Cem Concr Res 2001;31:1421-1425.

22 [14] Beshr H, Almusallam AA, Maslehuddin M. Effect of coarse aggregate quality on the mechanical properties of

23 high strength concrete. Constr Build Mater 2003;17:97-103.

24 [15] Beushausen H, Dittmer T. The influence of aggregate type on the strength and elastic modulus of high strength

25 concrete. Constr Build Mater 2015;74:132-139.

26 [16] Uysal M. The influence of coarse aggregate type on mechanical properties of fly ash additive self-compacting

27 concrete. Constr Build Mater 2012;37:533-540.

28 [17] Beygi MHA, Kazemi MT, Nikbin IM, Javad VA, Rabbanifar S. The influence of coarse aggregate size and

29 volume on the fracture behavior and brittleness of self-compacting concrete. Cem Concr Res 2014;66:75-90.

31
 1 [18] An LH, Fehling E. Influence of steel fiber content and aspect ratio on the uniaxial tensile and compressive

2 behavior of ultra-high performance concrete. Constr Build Mater 2017;153:790-806.

3 [19] Wu ZM, Shi CJ, He W, Wu LM. Effects of steel fiber content and shape on mechanical properties of

4 ultra-high performance concrete. Constr Build Mater 2016;103:8-14.

5 [20] Yoo DY, Kim SW, Park JJ. Comparative flexural behavior of ultra-high-performance concrete reinforced with

6 hybrid straight steel fibers. Constr Build Mater 2017;132:219-229.

7 [21] Yoo DY, Banthia N, Kang ST, Yoon YS. Effect of fiber orientation on the rate-dependent flexural behavior of

8 ultra-high-performance fiber-reinforced concrete. Compos Struct 2016;157:62-70.

9 [22] Huang HH, Gao XJ, Li LS, Wang H. Improvement effect of steel fiber orientation control on mechanical

10 performance of UHPC. Constr Build Mater 2018;188:09-721.

11 [23] Graybeal BA. Compressive behavior of ultra-high-performance fiber-reinforced concrete. ACI Mater J

12 2007;104(2):146-52.

13 [24] Wu ZM, Shi CJ, He W, Wang DH. Uniaxial Compression Behavior of Ultra-High Performance Concrete with

14 Hybrid Steel Fiber. J Mater Civ Eng 2016;28:6016017.

15 [25] Li PP, Yu QL, Brouwers HJH. Effect of coarse basalt aggregates on the properties of Ultra-high Performance

16 Concrete (UHPC). Constr Build Mater 2018;170:649-659.

17 [26] Liu JZ, Han FY, Cui G, Zhang QQ. Combined effect of coarse aggregate and fiber on tensile behavior of

18 ultra-high performance concrete. Constr Build Mater 2016;121:310-318.

19 [27] Missemer L, Ouedraogo E, Malecot Y, Clergue C. Fire spalling of ultra-high performance concrete: From a

20 global analysis to microstructure investigations. Cem Concr Res 2019;115:207-219.

21 [28] GB/T 14685-2011. Pebbie and crushed stone for building. Beijing, China;2011.

22 [29] CECS 13:2009. Standard test methods for fiber reinforced concrete. Beijing, China;2009.

23 [30] Hassan AMT, Jones SW, Mahmud GH. Experimental test methods to determine the uniaxial tensile and

24 compressive behaviour of ultra high performance fibre reinforced concrete (UHPFRC). Constr Build Mater

25 2012;37:874-882.

26 [31] Jang SJ, Yun HD. Combined effects of steel fiber and coarse aggregate size on the compressive and flexural

27 toughness of high-strength concrete. Compos Struct 2018;185:203-211.

28 [32] Yu R, Spiesz P, Brouwers HJH. Mix design and properties assessment of Ultra-High Performance Fibre

29 Reinforced Concrete (UHPFRC). Cem Concr Res 2014;56:29-39.

32
 1 [33] ASTM C 469M-14. Standard test method for static modulus of elasticity and poisson’s ratio of concrete in

2 compression. American Society of Testing and Materials, West Conshohocken, United States;2010.

3 [34] ACI 318-14. Building Code Requirements for Structural Concrete and Commentary. American Concrete

4 Institute, Farmington Hills, USA;2014.

5 [35] CEB-FIB. High Performance Concrete-Recommended Extensions to the Model Code 90-Research Needs.

6 International Federation for Structural Concrete, Lausanne, Switzerland;1995.

7 [36] GB 50010-2010. Code for design of concrete structures. Beijing, China; 2010.

8 [37] Guo XY, Kang JF, Zhu JS. Constitutive relationship of ultra high performance concrete under uniaxial

9 compression. J Southeast Univ 2017;02:369-376 (in Chinese)

10 [38] EN 1994-4-1. Eurocode 4: Design of composite steel and concrete structures-Part 1-1: General rules and rules

11 for buildings. European Committee for Standardization, Brussels, Belgium;2004.

12 [39] Wang Z, Wang JQ, Liu TX, Zhang F. Modeling seismic performance of high strength steel ultra-high

13 performance concrete piers with modified Kent-Park model using fiber elements. Adv Mech Eng 2016;8:1-13.

14 [40] Ezeldin AS, Balaguru PN. Normal- and high-strength fiber-reinforced concrete under compression. J Mater

15 Civ Eng 1992;4: 415-429.

16 [41] Nataraja MC, Dhang N, Gupta AP. Stress-strain curves for steel-fiber reinforced concrete under compression.

17 Cem Concr Comp 1999;21:383-390.

18 [42] Mansur MA, Chin MS, Wee TH. Stress-strain relationship of high-strength fiber concrete in compression. J

19 Mater Civ Eng 1999;11:21-29.

20 [43] Carreira DJ, Chu KH. Stress-strain relationship for plain concrete in compression. J Am Concr Inst

21 1985;82:797-804.

22 [44] ASTM C1609/C1609M-12. Standard Test Method for Flexural Performance of Fiber-Reniforced Concrete

23 (Using Beam With Third-Point Loading). West Conshohocken, United States;2012.

24 [45] Yu YR, Spiesz P, Brouwers HJH. Development of an eco-friendly Ultra-High Performance Concrete (UHPC)

25 with efficient cement and mineral admixtures uses. Cem Concr Compos 2015;55:383-394.

33
 1 Author Contribution Statement
2
3 Fanghong Wu: Conceptualization, Methodology, Visualization, Investigation, Formal analysis,
4 Writing-Original Draft
5 Lihua Xu: Supervision, Project administration
6 Yin Chi: Writing - Review & Editing
7 Yanqin Zeng: Investigation
8 Fangqian Deng: Visualization
9 Qian Chen: Investigation
10

34
1 Highlights

2  The stiffer coarse aggregate, the higher compressive strength of UHPFRC

3  Moderate dosage of coarse aggregate can improve the compressive strength and elastic modulus
4  The inclusion of coarse aggregate weakens the flexural behavior
5  An analytical model is proposed to predict the compressive stress-strain curves of UHPFRC
6

35