applied

sciences
Article
Pullout Response of Ultra-High-Performance
Concrete with Twisted Steel Fibers
Judong Ye and Guohua Liu *
College of Civil Engineering and Architecture, Zhejiang University, Hangzhou 310000, China;
yejudong@zju.edu.cn
* Correspondence: zjuliugh@zju.edu.cn; Tel.: +86-133-0651-5890




Received: 30 January 2019; Accepted: 12 February 2019; Published: 15 February 2019


Featured Application: The developed pullout force formula and the understanding of damage
mechanisms are expected to optimize the design of the ultra-high-performance fiber reinforced
concrete (UHPFRC) with twisted fibers.

Abstract: This paper aims to develop a pullout force formula and increase the understanding of the
damage mechanisms of ultra-high-performance fiber reinforced concrete (UHPFRC) with twisted steel
fibers (TSFs) through a pull-out test and finite element analysis (FEA). The formula was first obtained
through a theoretical force analysis with model assumptions that are based on the experimental data
in the literature. A microscale in-situ X-ray computed tomography (µXCT) was used to prepare 3D
images of the cross-section of concrete before and after TSFs with three embedment lengths were
pulled out. The tested pullout force values were used for comparison with the developed formula
values. The µXCT images show the concrete matrix was preserved after the TSF was pulled out,
indicating the stable pullout force values at the strain hardening stage was mainly caused by the fiber
untwisting. FEA results show this untwisting behavior occurs on the effective untwisting length of
TSF close to the exterior concrete surface. The theoretical formula values were found match well with
the testing data. The developed formula is potentially used to analyze the pullout behavior of TSF
with different geometries; thus, the design of the UHPFRC with TSFs can be optimized in the field.

Keywords: analytical model; finite element analysis; twisted steel fiber; ultra-high-performance
concrete

1. Introduction
Fibers with different types, shapes, aspect ratios have been used to enhance the mechanical
properties of concrete regarding its fracture toughness, flexural strength, tensile strength and energy
absorbing capacity under impact [1–8]. This improvement of mechanical properties is achieved by
producing the strain-hardening behavior and by increasing the strain capacity of the fiber-mixed
composite with supplementary cementitious materials such as silica fume and ground granulated
blast-furnace slag. It was reported that the maximum compressive strength and tensile strength of the
steel fiber reinforced concrete (FRC) could be as high as 292 MPa and 37 MPa, respectively and that
the strain at peak stress up to 1.1% was obtained [2]. These values are at least 5–10 times higher than
those of the conventional concrete without steel fibers. As a result, this ultra-high-performance fiber
reinforced concrete (UHPFRC) has been widely used in different structural members such as slabs,
beams, columns and beam-column joints for various construction applications [9].
Due to the superior performance and the increased applications of UHPFRC, extensive studies
have been performed to understand the mechanical properties of UHPFRC and the corresponding
reinforcement mechanism. The current studies focus on the optimization of the factors that influence

Appl. Sci. 2019, 9, 658; doi:10.3390/app9040658 www.mdpi.com/journal/applsci

Appl. Sci. 2019, 9, 658 2 of 17

the fiber-matrix bond strength, cementitious matrix packing density and the tensile response of
UHPFRC through either the direct tensile tests of dog bone specimens or the flexural bending
tests. These factors mainly include fiber type [10,11], fiber volume fraction [11], fiber embedment
length [12], fiber morphology [1,12–14], fiber orientation [15,16], loading rate [3,13], cementitious
material property [17], fiber-matrix interface property [12] and curing of cementitious material [12].
For example, Wille et al. [3] showed that energy absorption capacity significantly increased with the
fiber volume fraction of up to 3%. It should be noted that most of the above studies adopt material-level
experimental studies which however are often expensive and difficult to carry out, particularly at
meso/micro scales [18]. For instance, it is challenging to carry out strain measurements on short
discontinuous fibers, causing the difficulty in developing the bond-shear stress-slip relationship
between fibers and cementitious matrix. This could limit the detailed understanding of material’s
damage mechanisms and thus considerably limit the wider application of fibers.
Compared to the smooth or hooked steel fibers used as reinforcement in high-performance
cementitious composites, a relatively new type of steel fibers with twisted morphology has been
proposed to further enhance the mechanical properties of HPFRC [1,11,13]. This twisted steel fiber with
optimized section geometry (triangular and square section) was experimentally explored regarding its
superior performance in contributing the pull-out load and strain hardening capacity. Note that the
pull-out test is commonly used to evaluate the bond strength, pull-out response and strain hardening
behavior. The twisted fibers could theoretically provide a larger lateral surface area than smooth or
hooked steel fibers and thus increase frictional and adhesive bond forces at the fiber-matrix interface.
Consequently, the twisted fibers generated peak pullout forces on the order of 3-5 times those of
straight, smooth fibers [1]. Note that their high-performance-concrete strength values are generally
less than 100 MPa. Due to the complexity of fiber morphology, the fabrication of the fiber materials
becomes challenging. For example, only the rectangular-section twisted steel fibers are available
in Chinese material market, leading to the difficulty in investigating a variety of twisted steel fiber
geometries in experimental studies. Also, the pullout test with detailed measurements on fiber strain,
bond shear force, untwisting torque at meso/micro scales is still in an early stage. Several studies have
presented the typical pullout force response of HPFRC with twisted fibers [1,13], while a theoretical
or empirical formula of pullout force has not been established, particularly for UHPFRC with the
ultra-high compressive strength of more than 150 MPa.
The objective of this study is to develop a pullout force formula and thus to enhance the
understanding of the damage and fracture mechanisms of UHPFRC with twisted fibers through
a pull-out test and numerical modeling. The pull-out response of twisted fibers through the
cementitious matrix was first reviewed and then analytically modelled considering elastic-plastic
mechanics. The analytical model empirically assumes the tri-linear relationship of pullout force and
pullout deformation. A microscale in-situ X-ray computed tomography (µXCT) was used to provide
visual evidence for the failure pattern. The formula was verified by the pullout test results. The untwisting
mechanism of the single twisted fiber during the pullout process was also investigated by conducting
an image analysis and a finite element analysis (FEA). The developed pullout force formula and the
understanding of damage mechanisms are expected to optimize the design of the UHPFRC with
twisted fibers.

2. Analytical Model

2.1. Model Assumption

The twist geometry of twisted fiber and the physicochemical bond between fiber and concrete
matrix cause the pullout response complicated. To understand the general pullout behavior, the
typical pullout force data for the UHPFRC with twisted steel fibers were reviewed and shown in
Figure 1a. The data were collected from the existing experimental studies from Naaman and his
coworkers [1]; our experimental data will be used to verify the developed analytical model in Section
Appl. Sci. 2019, 9, 658 3 of 17

Experimental and verification. Based on the previous experimental data, we empirically assume that
the relationship of8,pullout
Appl. Sci. 2018, x FOR PEERforce
REVIEWand pullout displacement is a tri-linear response with three stages, as
3 of 17
shown in Figure 1b.
The thefirst
relationship
stage isofthepullout
linearforce and pullout
debonding displacement
stage, where the is apullout
tri-linearforce
response with three stages,
is increasing linearly with
as shown in Fig.1b.
the pullout displacement. During the pullout process, the shear force increases up to the fiber-matrix
The first stage is the linear debonding stage, where the pullout force is increasing linearly with
bond strength
the pullout and debonding
displacement. occurs
During theat the fiber-matrix
pullout interface
process, the shear after the
force increases up shear force exceeds the
to the fiber-matrix
bond strength. At the end of this elastic stage, the twisted fiber completes the
bond strength and debonding occurs at the fiber-matrix interface after the shear force exceeds debonding andthestarts to
untwist, entering into a plastic untwisting stage. The pullout force is assumed constant as Pf in this
bond strength. At the end of this elastic stage, the twisted fiber completes the debonding and starts
processtowhen
untwist, theentering into a embedment
remaining plastic untwisting stage.LeThe
length is pullout force isthe
larger than assumed
effectiveconstant length ∆l.
as Pf in this
untwisting
process when the remaining embedment length Le is larger than the effective untwisting length Δl.
Only the friction on the interface of the effective untwisting length is considered. In addition, the steel
Only the friction on the interface of the effective untwisting length is considered. In addition, the steel
fiber is fiber
assumed as an ideal plastic material and there is no damage in high-strength UHPC matrix
is assumed as an ideal plastic material and there is no damage in high-strength UHPC matrix
during during
the pull-out process.
the pull-out At the
process. At end of this
the end process,
of this process,the
thepullout
pullout force is dropped
force is droppedtotozero zero
withwith
a a high
dropping rate, entering a dropping stage.
high dropping rate, entering a dropping stage.

140
Pitch=3.2 mm Pitch=4.2 mm
120
Pitch=6.3 mm Pitch=12.7 mm

100
Pullout Load /N

80

60

40

20

0
0 2 4 6 8 10 12
Fiber End Slip/mm

(a) (b)

Figure Figure 1. Typical

1. Typical pullout
pullout curve
curve andand
itsitssimplified
simplified trilinear
trilinearmodel
modelfor for
twisted steel fiber-UHPC
twisted matrix: matrix:
steel fiber-UHPC
(a) typical pullout curve; and (b) simplified trilinear model for twisted steel fiber-UHPC.
(a) typical pullout curve; and (b) simplified trilinear model for twisted steel fiber-UHPC.
2.2. Derivation
2.2. Model Model Derivation
The cross-section of a twisted fiber can be rectangular, triangular or polygonal. The commonly-
The cross-section of a twisted fiber can be rectangular, triangular or polygonal. The commonly-used
used rectangular and triangular shape are analyzed in this study. As the force analysis process is
rectangular
similarandfortriangular shape
the rectangular are analyzed
or triangular shape,inthe
this study.analysis
detailed As theprocess
force of
analysis process
rectangular is similar for
fiber shape
the rectangular or triangular
is first carried out, followedshape,
by the the detailed
summary analysis
of triangular process
fiber shape. of rectangular fiber shape is first
carried out, followed by the summary of triangular fiber shape.
2.2.1. Fiber with A Rectangular Section
2.2.1. Fiber As
with a Rectangular
shown in Fig. 2, for aSection
twisted steel fiber with the 2a × 2b rectangular section, the contact stress
of the matrix on fiber surface is assumed to be linearly distributed on a section. The contact stress at
As shown in Figure 2, for a twisted steel fiber with the 2a × 2b rectangular section, the contact
the corner point is N1 (y) and its distribution is antisymmetric in the other parallel face. Take dx to be
stress ofthe
theincrement
matrix on fiber
at the surface
long is assumed
side and to be
it is subjected to linearly distributed
compression onthe
stress from a section. Thematrix
cementitious contact stress
at the corner
as point is N1 (y) and its distribution is antisymmetric in the other parallel face. Take dx to be
the increment at the long side and it is subjected to 𝑥 compression stress from the cementitious matrix as
𝑝 = 𝑁 (𝑦) (1)
𝑏
x
p = ofNthe
where p is the compression stress, b is the length (y)long side. The line load can be obtained by (1)
b 1
integrating p along the long side as
where p is the compression stress, b is the length of the long side. The line load can be obtained by
𝑥
integrating p along the long side as 𝑝 = 𝑁 (𝑦)𝑑𝑥 (2)
Z b𝑏
x
pX = N1 (y)dx (2)
0 b
Considering the angle θ of the long side twisted surface (Figure 2), the component of the pressure
in the fiber axial y directions
Appl. Sci. 2019, 9, 658 4 of 17

Z b
x
py = N1 (y)sinθdx (3)
0 b
TheAppl.
normal force px on the twisted surface is
Sci. 2018, 8, x FOR PEER REVIEW 4 of 17
Z b
Considering the angle θ of the long side twistedxsurface (Fig.2), the component of the pressure in the
px = N1 (y)cosθdx (4)
fiber axial y directions 0 b
𝑥
As the fine sands were mixed in the𝑝 matrix
= 𝑁 (𝑦) 𝑠𝑖𝑛 𝜃coarse
without 𝑑𝑥 (3)
aggregates, the contact between TSF
𝑏
and matrix is relatively uniform. Thus, we assume that the sliding friction coefficient µ between the
The normal force px on the twisted surface is
matrix and the steel fiber is constant, the friction force in the fiber axial direction is
𝑥
𝑝 = Z b𝑁 (𝑦) 𝑐𝑜𝑠 𝜃 𝑑𝑥 (4)
𝑏 x
fy = µ · N1 (y)cosθdx (5)
As the fine sands were mixed in the matrix b
0 without coarse aggregates, the contact between TSF
and matrix is relatively uniform. Thus, we assume that the sliding friction coefficient μ between the
The friction
matrix coefficient
and the steel fiberµisisconstant,
determined by referring
the friction force in thetofiber
theaxial
experimental
direction is results by Baltay and
Gjelsvik [19]. The friction coefficient at concrete-steel 𝑥 interface is mainly determined by the processing
methods of steel and it was ranged from 𝑓 =0.35
𝜇 ⋅ to 𝑁 (𝑦) 𝑐𝑜𝑠 𝜃 𝑑𝑥 (5)
𝑏 0.58. In this model, the friction coefficient of 0.5 at
concrete-steelTheinterface for UHPFRC
friction coefficient was set inbythe
μ is determined modeltointhe
referring this study.
experimental results by Baltay and
The twisted
Gjelsvik surface
[19]. angle atcoefficient
The friction the shortatside 0
is θ . Thus,
concrete-steel the axial
interface resultant
is mainly force of the
determined four sides
by the
processing
(line load) Pf 0 is methods of steel and it was ranged from 0.35 to 0.58. In this model, the friction coefficient
of 0.5 at concrete-steel interface for UHPFRC was set in the model in this study.
The twisted surface
Z b
x angle at the short side is θ'. Thus,athe
Z
x axial resultant force of the

four sides
(line load)Pf𝑃0 =is2 N1 (y)(sinθ + µ cosθ )dx + 2 N1 (y) sinθ 0 + µ cosθ 0 dx (6)
0 b 0 b
𝑥 𝑥
𝑃 =2 𝑁 (𝑦)(𝑠𝑖𝑛 𝜃 𝜇 𝑐𝑜𝑠 𝜃)𝑑𝑥 2 𝑁 (𝑦)(𝑠𝑖𝑛 𝜃 𝜇 𝑐𝑜𝑠 𝜃 )𝑑𝑥 (6)
𝑏 𝑏

Figure
Figure 2. Force
2. Force analysis
analysis sketchof
sketch of twisted
twisted steel
steelfiber of rectangular
fiber section.
of rectangular section.

The untwisting
The untwisting pullout
pullout forceforce
Pf Pisf isthus
thus obtained
obtained by byintegrating 𝑃 P
integrating on0 the effective untwisting
f on the effective untwisting
length Δl:
length ∆l:
𝑥 𝑥
𝑃Z =  Z2 𝑁 (𝑦)(𝑠𝑖𝑛 𝜃 𝜇 𝑐𝑜𝑠 𝜃)𝑑𝑥 2 Z 𝑁 (𝑦)(𝑠𝑖𝑛 𝜃 𝜇 𝑐𝑜𝑠 𝜃 )𝑑𝑥 𝑑𝑦  (7)
∆l b x𝑏 𝑏a x
N1 (y) sinθ 0 + µ cosθ 0 dx dy


f =
PSimilarly, 2
the twistedN 1 ( y
torque)( sinθ
T can + µ
be cosθ ) dx
expressed + 2
as (7)
0 0 b 0 b
𝑥 𝑥
𝑇= 2 𝑁 (𝑦)(cos 𝜃 − 𝜇 𝑠𝑖𝑛 𝜃) ∙ 𝑥𝑑𝑥 2 𝑁 (𝑦)(cos 𝜃 − 𝜇 𝑠𝑖𝑛 𝜃 ) ∙ 𝑥𝑑𝑥 𝑑𝑦 (8)
Similarly, the twisted𝑏 torque T can be expressed as 𝑏
It Z
assumed
∆l
 Z that
b
the compressive stress evenly distributed
Z a  (△l
along the fiber axial direction
direction), that is, Nx1(y) = N1. Equations (7) and (8) can be simplified
x to 0 0


T= 2 N1 (y)(cos θ − µ sinθ )· xdx + 2 N1 (y) cos θ − µ sinθ · xdx dy (8)
0 0 b 𝑥 𝑥 0 b
𝑃 =2 𝑁 (𝑠𝑖𝑛 𝜃 𝜇 𝑐𝑜𝑠 𝜃)𝑑𝑥 𝑁 (𝑠𝑖𝑛 𝜃 𝜇 𝑐𝑜𝑠 𝜃 )𝑑𝑥 ∙ Δ𝑙 (9)
𝑏 𝑏
It assumed that the compressive stress evenly distributed along the fiber axial direction
𝑥 𝑥
𝑇 =N21 (y) =𝑁N(cos
(∆l direction), that is, 𝜃 − 𝜇 𝑠𝑖𝑛 𝜃)(7)
1 . Equations ∙ 𝑥𝑑𝑥 𝑁 (cos
and (8) can − 𝜇 𝑠𝑖𝑛 𝜃 ) to
be 𝜃simplified ∙ 𝑥𝑑𝑥 ∙ Δ𝑙 (10)
𝑏 𝑏
b Z a
Z 
x x 0 0
N1 sinθ + µ cosθ dx ·∆l


Pf = 2 N (sinθ + µ cosθ )dx + (9)
0 b 1 0 b
Appl. Sci. 2019, 9, 658 5 of 17

b Z a
Z 
x x
N1 cos θ 0 − µ sinθ 0 · xdx ·∆l


T=2 N (cos θ − µ sinθ )· xdx + (10)
0 b 1 0 b
Appl. Sci.
Figure 2018, 8, x FOR
3 shows thePEER REVIEW relation of a pitch. For any point K on the steel fiber
geometric 5 ofsurface,
17 it
√
travels through a pitch l0 in the embedment length direction (y direction) by a distance of 2π x + a2 . 2
Fig. 3 shows the geometric relation of a pitch. For any point K on the steel fiber surface, it travels
We can through
obtain a pitch l0 in the embedment length direction √ (y direction) by a distance of 2𝜋√𝑥 𝑎 . We
can obtain
2π x2 + a2
sinθ = q (11)
2 ( x2 +𝑎a2 ) + l 2
2𝜋√𝑥
4π
𝑠𝑖𝑛 𝜃 = 0
4𝜋 (𝑥 𝑎 ) 𝑙 (11)
l0
cos θ = q (12)
4π 2 (𝑙x2 + a2 ) + l02
cos 𝜃 = (12)
4𝜋 (𝑥 √ 𝑎 ) 𝑙
2π x + b 2 2
sinθ 0 = q 2𝜋√𝑥 𝑏 (13)
𝑠𝑖𝑛 𝜃 = 2 ( x 2 + b2 ) + l 2 (13)
4π(𝑥
4𝜋 𝑏 ) 𝑙 0
𝑙
𝑐𝑜𝑠 𝜃 = l0 (14)
cosθ 0 = q4𝜋 (𝑥 𝑏 ) 𝑙 (14)
4π ( x + b2 ) + l02
2 2

Figure 3. Sketch
Figure of of
3. Sketch twisted
twistedsurface anglefor
surface angle for rectangular
rectangular section
section fiber. fiber.

Substituting Equations
Substituting (11)–(14)
Equations (11−14)into
into Equations (9)and
Equations (9) and (10),
(10), the the untwisting
untwisting pullout
pullout force
force and the and the
torque
torque can be can be obtained
obtained as as
2𝑁Δ𝑙 2𝜋√𝑥 𝑎 𝜇∙𝑙 2𝜋√𝑥 𝑏 𝜇∙𝑙  (15)
𝑃 = 𝑥∙ √ 𝑑𝑥 𝑥∙ √ 𝑑𝑥
𝑏 Z b
2N1 ∆l  2π4𝜋 x(𝑥2 + a𝑎2 )+ µ𝑙 ·l0 a 4𝜋 2π(𝑥 x𝑏2 )+ b𝑙2 + µ·l
Z
0
Pf = 2𝑁 Δ𝑙 x𝑙 · −
q𝜇 ∙ 2𝜋√𝑥 𝑎 dx + 𝑙 − x𝜇 ·∙ q
2𝜋√𝑥 𝑏 dx  (15)
𝑇= b 0 4π 2 ( x 2 + a 2 ∙ 𝑥+𝑑𝑥
) l 2 0 4π 2 ( x 2 +∙ 𝑥b 𝑑𝑥
2 ) + l 2 (16)
𝑏 4𝜋 (𝑥 𝑎 ) 𝑙 0 4𝜋 (𝑥 𝑏 ) 𝑙 0

By consideringthe Mises yield criterion, we can obtain

√ √ 
2N1 ∆l  l0 − µ·2π x𝜎2 + a2 x 2 + b2
Z b Z a
𝜏2 l − µ·2π
T= +1 q0
3 · x dx = · x2 dx  (17) (16)
𝜎 𝜎
q
b 0 2
4π 2 ( x2 + a2 ) + l0 0 2 2 4π ( x + b2 ) + l02
where: σp is the uniaxial tensile yield strength of steel fiber (2500 MPa in this paper); σ is the axial
normal stress ofthe
By considering steel fiber,yield
Mises σ = Pfcriterion,
/A, A = 4ab;we
τ iscan
theobtain
untwisting-torque caused tangential stress of
steel fiber.
The plastic limit torque M of the σ 2
rectangular τ 2
section
  beam is well known as
+3 =1 (17)
σ𝑀p = 4𝑏 𝜏 (3𝑎σ−
p 𝑏) (18)
3
where: σp isAtthe
theuniaxial
yield limittensile
state (i.e., T = M)
yield in the plastic
strength untwisting
of steel stage, MPa
fiber (2500 Equation (16) can
in this be rewritten
paper); σ is the axial
normal asstress of steel fiber, σ = Pf /A, A = 4ab; τ is the untwisting-torque caused tangential stress of
steel fiber. 𝑃 9𝑇
=𝜎 (19)
4𝑎𝑏 4𝑏 (3𝑎 − 𝑏)
The plastic limit torque M of the rectangular section beam is well known as

4b2 τ
M= (3a − b) (18)
3
Appl. Sci. 2019, 9, 658 6 of 17

At the yield limit state (i.e., T = M) in the plastic untwisting stage, Equation (16) can be rewritten as
2 2
Pf
 
9T
+ 2
= σp 2 (19)
4ab 4b (3a − b)

If the dimension of twisted fiber (i.e., a, b, l0 ), uniaxial tensile yield strength of twisted fiber σp ,
the effective untwisting length ∆l are known (note that ∆l can be estimated from the typical
Appl. Sci. 2018, 8, x FOR PEER REVIEW 6 of 18
pullout
curve shownIf in Figure 1b), the untwisting pullout force P and the torque T can be obtained
the dimension of twisted fiber (i.e., a, b, l0), uniaxialf tensile yield strength of twisted fiber σp,from the
Equations (15), (16) and (19).
the effective untwisting length Δl are known (note that Δl can be estimated from the typical pullout
curve shown in Fig.1b), the untwisting pullout force Pf and the torque T can be obtained from the
2.2.2. Fiber with a(15),
Equations Triangular Section
(16) and (19).

Figure
2.2.2.4Fiber
shows theA force
With analysis
Triangular sketch for the TSF with triangular section. The relevant equation
Section
forms of triangular section are consistent with those of rectangular section. By replacing the section
Fig. 4 shows the force analysis sketch for the TSF with triangular section. The relevant equation
variables, the untwisting pullout force and the torque for the UHPFRC with triangular-section twisted
forms of triangular section are consistent with those of rectangular section. By replacing the section
fibers can be obtained
variables, as:
the untwisting pullout force andZthe torque for the UHPFRC with triangular-section
0 3∆l · N1 a
twisted fibers can be obtainedPas: f = ( sinθ + µ cosθ ) xdx (20)
a
3Δ𝑙 ∙ 𝑁 0
𝑃′= Z (𝑠𝑖𝑛 𝜃 𝜇 𝑐𝑜𝑠 𝜃)𝑥𝑑𝑥 (20)
0 3∆l𝑎· N1 b 2
T = (cosθ − f sinθ ) x dx (21)
3Δ𝑙 ∙a𝑁 0
𝑇′ = (𝑐𝑜𝑠 𝜃 − 𝑓 𝑠𝑖𝑛 𝜃)𝑥 𝑑𝑥 (21)
where, 𝑎
q
where, 2π x2 + 31 a2
sinθ = r
1
 
4π2𝜋2 x𝑥 2 +313𝑎a2 + l02
𝑠𝑖𝑛 𝜃 =
1
4𝜋 (𝑥 l30 𝑎 ) 𝑙
cosθ = r  
4π 2 x2𝑙 + 13 a2 + l02
𝑐𝑜𝑠 𝜃 =
1 √
4𝜋 (𝑥 𝑎 ) 𝑙
3 (with section area
The beam plastic torque limit of triangular section 3a2 ) is:
The beam plastic torque limit of triangular section (with section area √3𝑎 ) is:
2τ
M0 =2𝜏 a3 (22) (22)
𝑀′ = 𝑎
33
Considering
Considering the yield
the yield limitlimit state
state (i.e.,T 0𝑇 == M
(i.e., 𝑀′)0 ) in
inthe
theplastic
plasticuntwisting stage,stage,
untwisting the untwisting
the untwisting
pullout force and the torque for the twisted fiber with triangular section can be obtained from the
pullout force and the torque for the twisted fiber with triangular section can be obtained from the
Equations (20), (21) and (22).
Equations (20), (21) and (22).

Figure
Figure 4. Force
4. Force analysis
analysis sketchof
sketch oftwisted
twisted steel
steelfiber of of
fiber triangular section.
triangular section.

3. Experimental and Verification

Appl. Sci. 2019, 9, 658 7 of 17

3. Experimental and Verification

To verify the theoretical understanding of pullout force and enhance the understanding of
untwisting behavior as well as the mechanical performance of matrix, this section shows the
experimental work and the comparison between theory and experiment.

3.1. Concrete Samples

The UHPRC samples had water-binder ratio of 0.2 with ordinary Portland cement (P.O 52.5,
produced by Hangzhou Qianchao Cement Factory, China) and silica fume that was used as
cementitious materials. Quartz sand with particle size of 0.1-0.5 mm and quartz powder with an
average particle size of 26 µm were used as filling agents. Solid polycarboxylate-based superplasticizer
(BASF Melflux 1641F with high efficiency of plasticization) was mixed to improve the flowability
of mixture. The mix was designed according to the general mix design for UHPRC such as [20].
The detailed mix proportions of UHPRC are 822 kg/m3 cement, 904 kg/m3 quartz sand, 205 kg/m3
silica fume, 246 kg/m3 quartz powder, 205 kg/m3 water, 10 kg/m3 superplasticizer. The quartz sand,
cement, silica fume and quartz powder were first dry mixed for 30 s in a pan mixer. Then, water
and super-plasticizer were added and mixed for a further 3 min. The mixture was compacted on a
vibrating table with adjustable intensity. The samples were cast in cubic steel molds, the surface of
which was covered with wet hessian at room temperature for 24 h. Subsequently, the samples were
demolded and sealed cured at room temperature for 28 days. The compression test was carried out
and the results show that the average compressive strength reached up to 155 MPa, indicating the
ultra-high performance of the designed sample.

3.2. Concrete Samples with a Single Twisted Steel Fiber

The twisted steel fibers with rectangular section were used in the experiment, as illustrated in
Figure 3. The fiber length l, equivalent diameter De , fiber pitch l0 , tensile strength σp and elastic
modulus E are 45 mm, 0.5 mm, 0.53 × 0.37 mm2 , 0.1961 mm2 , 10 mm, 2500 MPa and 206 GPa,
respectively. Only rectangular-section twisted steel fiber in high-performance fiber concrete was
studied in this paper as the twisted steel fibers with triangular section cannot be purchased in the
Chinese market.
There are three fiber embedment lengths of 15, 20 and 25 mm and six replicate specimens
(T15/20/25-1, -2, -3, -4, -5, -6) were prepared for each fiber embedment length to obtain the
representative testing values. The minimum fiber embedment length of 15 mm is 50% higher than the
fiber pitch value of 10 mm; while the maximum fiber embedment length of 25 mm is 2.5 times fiber
pitch and approximately half of fiber length. This arrangement ensures the whole pullout force curve
can be completely obtained.
Eighteen fiber reinforced specimens in total were prepared in the following process: First, a
twisted fiber with a hook anchor was fixed in the middle of a steel mold and the concrete mixture
prepared in above Section Concrete samples was poured into the mold for compaction. After 24 h,
the first half specimen was demolded. The second half specimen with a certain embedment length
of fiber (i.e., 15, 20 and 25 mm) was then cast and carefully demolded after 24 h. The whole sample
was then conditioned for 28 days. As the concrete strength during the sample preparation is low,
cares should be taken to protect the fiber-matrix interface from being damaged during the operation.
To ensure no rotation of sample during the pullout process, both sides of sample were restrained
using a fixture. The single fiber pullout experiment was then carried out according to the testing
standard [21]. The loading actuator controls the displacement values, which are used as the fiber end
slip. Note that the maximum pullout force value of approximately 250 N is so small that the fiber
deformation can be negligible. Displacement control at a low pull velocity of 0.01 mm/s was used in
our quasi-static loading experiment.
Appl. Sci. 2019, 9, 658 8 of 17

The concrete matrix of samples before and after the pullout test was scanned and high-resolution
images were obtained by using the Nikon XTH320 µXCT facility at Zhejiang University. The scanning
of the specimen before pullout test was to observe the fiber-matrix interface; while the scanning of the
matrix after pullout was to observe the embedded channel of the twisted steel fiber.
A series of two-dimensional images were first obtained and then three-dimensional images
were reconstructed by using VGStudio MAX 3.0 software [22]. The images were then analyzed and
compared (before pullout vs. after pullout) to observe the damage situation of the fiber-matrix interface
and understand the pullout mechanism of twisted steel fiber. The following scanning parameters
were used: exposure time (500 ms), voltage (200 kV), current (100 µA) and scanning resolution
(approximately
Appl. 30 µm).
Sci. 2018, 8, x FOR PEER REVIEW 8 of 17

3.3.Experimental
3.3. ExperimentalResults
Results

3.3.1. Image Analysis

3.3.1. Image Analysis
Figure 5 shows the XCT images of the typical cross-sections of sample with fiber (before pullout,
Fig.5 shows the XCT images of the typical cross-sections of sample with fiber (before pullout,
white color) and without fiber (after pullout, black color). As expected, due to the twisting geometry,
white color) and without fiber (after pullout, black color). As expected, due to the twisting geometry,
the 2D images of twisted fibers show the uneven width of fiber channel. It is interesting to observe
the 2D images of twisted fibers show the uneven width of fiber channel. It is interesting to observe
that the fiber-matrix interface is relatively smooth without considerable damage. The fiber channel
that the fiber-matrix interface is relatively smooth without considerable damage. The fiber channel
area in the image with steel fiber before the pullout test is 12.154 mm22 , which is rather close to that
area in the image with steel fiber before the pullout test is 12.154 mm , which is rather close to that
without fiber after the pullout test (12.239 mm2 ). Note that the scanning resolution is approximately
without fiber after the pullout test (12.239 mm2). Note that the scanning resolution is approximately
30 µm, so that any micro-damage with features less than 30 µm is out of sight in the obtained images.
30 µm, so that any micro-damage with features less than 30 µm is out of sight in the obtained images.
As short microcracks less than 30 µm in concrete are related to concrete durability performance instead
As short microcracks less than 30 µm in concrete are related to concrete durability performance
of mechanical property performance [23,24], the matrix-fiber interface could be regarded to be not
instead of mechanical property performance [23,24], the matrix-fiber interface could be regarded to
damaged within the mechanical analysis scope of this study. While future study can be performed to
be not damaged within the mechanical analysis scope of this study. While future study can be
understand the influence of microcracks on the property of twisted steel fiber such as the crack-induced
performed to understand the influence of microcracks on the property of twisted steel fiber such as
steel corrosion after the mass transport channel is formed.
the crack-induced steel corrosion after the mass transport channel is formed.
During the axial pulling process of TSF from the matrix, the fiber was gradually moving out of the
During the axial pulling process of TSF from the matrix, the fiber was gradually moving out of
solid matrix. No metal piece was peeled off during the pullout process. The untwisting phenomenon
the solid matrix. No metal piece was peeled off during the pullout process. The untwisting
was found after around 1–3 min of pulling and the whole twisting shape disappeared completely after
phenomenon was found after around 1-3 minutes of pulling and the whole twisting shape
pullout, as shown in Figure 6. This untwisting phenomenon is consistent with the existing numerical
disappeared completely after pullout, as shown in Fig.6. This untwisting phenomenon is consistent
studies from researchers such as Ellis et al. [14] who numerically investigated the pullout response of
with the existing numerical studies from researchers such as Ellis et al. [14] who numerically
the single twisted fiber with a triangular cross-section. As the untwisting behavior inside the concrete
investigated the pullout response of the single twisted fiber with a triangular cross-section. As the
is invisible, we then conduct a numerical study in Section Numerical analysis to understand the
untwisting behavior inside the concrete is invisible, we then conduct a numerical study in Section
untwisting characteristics inside the concrete.
Numerical analysis to understand the untwisting characteristics inside the concrete.

Embedment length le
Before pullout experiment
(with steel fiber)

1 Pitch≈10mm Steel fiber

After pullout experiment

(steel fiber channel)

1000μm
Pullout direction

Figure 5. Typical XCT section scanning images of concrete matrix before and after twisted steel fiber
Figure
pullout5.experiment.
Typical XCT section scanning images of concrete matrix before and after twisted steel fiber
pullout experiment.
 1000μm
Pullout direction

Figure 5. Typical XCT section scanning images of concrete matrix before and after twisted steel fiber
Appl. Sci. 2019, 9, 658 9 of 17
pullout experiment.

Appl. Sci. 2018, 8, x FOR PEER REVIEW 9 of 17

Figure6.
Figure Comparisonof
6.Comparison oftwisted
twistedsteel
steelfiber
fibershapes
shapes before
before and
and after
after pullout.
pullout.

3.3.2. Untwisting Mechanism

3.3.2. Untwisting Mechanism
The above untwisting phenomenon is found to be related to the pullout force response and is
The above untwisting phenomenon is found to be related to the pullout force response and is
discussed in this section. The eighteen pullout force versus fiber displacement Pf -s curves of specimens
discussed in this section. The eighteen pullout force versus fiber displacement Pf -s curves of
with three different embedment lengths were obtained. The results show that these curve patterns are
specimens with three different embedment lengths were obtained. The results show that these curve
similar and thus the average data lines for each embedment length are shown in Figure 8. The patterns
patterns are similar and thus the average data lines for each embedment length are shown in Fig.8.
of the pullout curves are generally consistent with those in Figure 1, while the maximum pullout force
The patterns of the pullout curves are generally consistent with those in Fig.1, while the maximum
is around 200-250N that is more than one order of magnitude compared to the typical pullout force
pullout force is around 200-250N that is more than one order of magnitude compared to the typical
values collected from Naaman and his coworkers [1]. The compressive strength of UHPRC is 155 MPa,
pullout force values collected from Naaman and his coworkers [1]. The compressive strength of
more than approximately 55% higher the value of high-performance concrete tested by Naaman and
UHPRC is 155 MPa, more than approximately 55% higher the value of high-performance concrete
his coworkers. Thus, the ultra-high mechanical property of UHPRC could increase the high pullout
tested by Naaman and his coworkers. Thus, the ultra-high mechanical property of UHPRC could
force as shown in Figure 7. In addition, the matrix-fiber interface in UHPRC has a high bond strength,
increase the high pullout force as shown in Fig.7. In addition, the matrix-fiber interface in UHPRC
which could also contribute the high pullout force.
has a high bond strength, which could also contribute the high pullout force.
Although there are variations of pullout force for the six replicate samples, the general pattern
Although there are variations of pullout force for the six replicate samples, the general pattern
of pullout force-fiber displacement curve shows there are three distinguishable stages that can be
of pullout force-fiber displacement curve shows there are three distinguishable stages that can be
identified. Both our pullout experimental data and the testing data from Naaman and his coworkers
identified. Both our pullout experimental data and the testing data from Naaman and his coworkers
show a similar tri-linear-stage pattern, as shown in Figure 1. The model assumption of our study
show a similar tri-linear-stage pattern, as shown in Fig.1. The model assumption of our study (Section
(Section Modal assumption) is thus reasonable.
Modal assumption) is thus reasonable.

Figure 7. Averaged experimental pullout curves.

Figure 7. Averaged experimental pullout curves.
To understand the untwisting mechanism and the pullout mechanism, the three trilinear stages of
pullout force versusthe
To understand fiber displacement
untwisting curves are
mechanism andpresented in mechanism,
the pullout Figure 7 andthe
analyzed as:
three trilinear stages
of pullout force versus fiber displacement curves are presented in Fig.7 and analyzed as:
• Linear debonding stage: The pullout force increased linearly with the fiber end displacement,
similar to the elastic stage for the normal concrete [25]. This process takes around 1-3 minutes.
There is no relative displacement between matrix and TSF, indicating the matrix-fiber interface
remains intact with no debonding occurs. Thus, the strong bonding between fiber and its
Appl. Sci. 2019, 9, 658 10 of 17

• Linear debonding stage: The pullout force increased linearly with the fiber end displacement,
similar to the elastic stage for the normal concrete [25]. This process takes around 1–3 min. There is
no relative displacement between matrix and TSF, indicating the matrix-fiber interface remains
intact with no debonding occurs. Thus, the strong bonding between fiber and its surrounding
matrix produces a united resistance to the external pulling force from both TSF and matrix. By
the end of this linear process, debonding initiates at the matrix-fiber interface where the pullout
force-loading displacement relationship becomes weakly nonlinear.
• Plastic untwisting stage: After the debonding occurs, debonding slip rotation of TSF is believed
to start. This movement is a kind of screw-like movements at the embedded fiber part after
debonding. As there is no relative rotation at the sample ends due to the constraint from the
fixture, a torque is generated. This torque leads to the plastic ‘untwisting’ deformation in the fiber
part near the outlet and thus the stable untwisting pull-out force. This pull-out force hardening
stage was also reported by Wille et al. [17]. As shown in Figure 7, there are slight fluctuations
in the untwisting pullout force. The fluctuations may result from the inhomogeneity of the
matrix-fiber interface and the stress concentration on the rectangular corner of TSF. This stress
concentration may be difficult to penetration by the matrix [1]. The super-high strength of matrix
prevents the TSF from damaging the matrix due to the undesirable sites of stress concentration.
• Dropping stage: When the remaining embedment length of the fiber is less than the effective
untwisting length ∆l, the pull-out force rapidly decreases to zero when the torque is released.

3.3.3. Pullout Force Calculation

Due to the variations of pullout forces at different fiber displacements (as shown in Figure 7,
a representative pullout force is calculated for quantitative analysis and comparison study with
theoretical values. The average pull-out force P f of the whole pullout process (as shown in Figure 8) is
defined as, R
Ep P(s)ds
Pf = = (23)
le le
where E p is the total pull-out energy during the whole process, that is, the area enclosed by the pullout
curve; le is the fiber embedment length; s is the fiber pullout displacement; P(s) is the pull-out force
when the pullout displacement is equal to s. s1 and s2 are obtained from the two intersection points
between P f and the actual pullout curve. s1 represents the starting point of the untwisting stage
and s2 represents the end point of untwisting stage. Thus, the untwisting pullout force in the plastic
untwisting stage can be estimated as R s2
s P ( s )ds
Pf = 1 (24)
s2 − s1
The effective untwisting length ∆l is equal to the value of le minus s2 . Note the fiber axial extension
 
during pullout is not considered in the calculation of ∆l as the extension equal to the P f le / EA f is
negligible. For example, when the P f is 200 N and le is 25 mm, the extension equals 0.1 mm. According
to the above calculation method, the experiment results are analyzed and shown in Table 1. These
values are then compared to the those calculated by the developed formula of pullout force in Section
Analytical model.
 The effective untwisting length Δl is equal to the value of 𝑙 minus 𝑠 . Note the fiber axial
extension during pullout is not considered in the calculation of Δl as the extension equal to the
𝑃 𝑙 /(E𝐴 ) is negligible. For example, when the 𝑃 is 200 N and 𝑙 is 25 mm, the extension equals
0.1 mm. According to the above calculation method, the experiment results are analyzed and shown
in Table
Appl. 1. These
Sci. 2019, 9, 658 values are then compared to the those calculated by the developed formula
11 of of
17
pullout force in Section Analytical model.

Figure 8. Sketch of calculation method for pullout force.

Figure 8. Sketch of calculation method for pullout force.
Table 1. Results of single fiber pullout tests his is a table.

Ep Average Pullout Untwisting

le/mm ID s1 (mm) s2 (mm) ∆l (mm)
(N.mm) Force (N) Pullout Force (N)
Average line 2873 191.5 1.26 13.24 1.76 199.1
T15-1 3295 219.7 1.11 13.94 1.06 229.5
T15-2 2980 198.7 1.75 13.04 1.96 210.2
15 T15-3 2461 164.1 1.41 13.63 1.37 179.7
T15-4 2895 193.0 1.21 13.31 1.69 201.3
T15-5 2810 187.3 1.21 13.75 1.25 195.8
T15-6 2539 169.3 1.37 13.51 1.49 182.1
Average line 3905 193.3 1.27 18.28 1.72 203.3
T20-1 3476 173.8 1.54 18.65 1.35 180.1
T20-2 4355 217.8 1.51 18.68 1.32 223.9
20 T20-3 3993 199.7 2.17 18.32 1.68 207.5
T20-4 3763 188.2 1.21 18.25 1.75 197.7
T20-5 3873 193.7 1.01 18.13 1.87 201
T20-6 3965 198.3 1.12 18.2 1.8 207.8
Average line 4582 184.4 1.22 23.35 1.65 193.0
T25-1 4720 188.8 1.28 23.44 1.56 198.3
T25-2 4678 187.1 1.31 23.12 1.88 194.8
25 T25-3 4494 179.8 1.33 22.61 2.39 192.1
T25-4 4427 177.1 1.29 23.61 1.39 184.0
T25-5 4445 177.8 1.23 23.70 1.30 184.3
T25-6 4738 189.1 1.25 22.47 2.53 196.6

3.4. Comparison between Test and Analytical Solution

As the experimental data has a certain degree of discreteness, the mean value of each embedment
length is compared with the analytical results. To obtain the analytical results of untwisting pullout
force Pf , the dimension of twisted fiber (i.e., a=0.265 mm, b=0.185 mm, l0 =10 mm), the uniaxial tensile
yield strength of twisted fiber σp =2500 MPa, the effective untwisting length ∆l from Table 1 are inputted
into the Equations (15), (16) and (19).
Figure 9 shows the comparison between the testing values of the pullout force and the analytical
solution. The results are in good agreement. The mean values of the three groups of different
embedment lengths are 4.2%, 6.1% and 1.3% higher than the analytical solution. This might because
the analytical model neglected the tangential frictional resistance after the debonding between the
UHPC matrix and the twisted steel fiber; only the friction on the fiber-fiber interface at the effective
untwisting length is considered. The pullout force can reach 200 N, which is generally larger than
Appl. Sci. 2019, 9, 658 12 of 17

the existing
Appl. studies
Sci. 2018, 8, [13,14,17].
x FOR PEER REVIEWThis
increase may be due to the higher mechanical properties of
12 pure
of 17
concrete without fibers and the better matrix-fiber interface produced in this study.

250
Appl. Sci. 2018, 8, x FOR PEER REVIEW 12 of 17
Theoretical values Testing values
200
Pullout force: N 250
150 Theoretical values Testing values
200
Pullout force: N

100
150
50
100
0
50
15 20 25
0 Embedment length: mm
15 20 25
Figure 9. Comparison between the testing values and theoretical values of pullout forces.
Embedment length: mm
Figure 9. Comparison between the testing values and theoretical values of pullout forces.
4. Numerical Analysis
Figure 9. Comparison between the testing values and theoretical values of pullout forces.
4. Numerical Analysis
In this section, a numerical modeling of UHPC with single TSF is presented. The aim of this
analysis wassection,
to understand
4.InNumerical
this Analysis the invisible
a numerical untwisting
modeling of UHPC behavior of TSFTSF
with single inside the concrete,
is presented. which
The aim cannot
of this
be observed through the above experiment.
analysis was to understand the invisible untwisting behavior of TSF inside the concrete, which cannot
In this section, a numerical modeling of UHPC with single TSF is presented. The aim of this
be observed through
analysis was the above
to understand the experiment.
invisible untwisting behavior of TSF inside the concrete, which cannot
4.1. Structure Model
be observed through the above experiment.
4.1. Structure
The FEAModel model for UHPC with single TSF is established by using ABAQUS and illustrated in
4.1. 10a
Figure Structure
and Model
its meshing is shown in Figure 10b. The model is composed of three parts, that is,
The FEA model for UHPC with single TSF is established by using ABAQUS and illustrated in
concreteThe matrix,
FEA TSF
model andfortheir interface. TheTSF
TSFishas an embedment
by usinglength of 20 mm, a free endin with
Fig.10a and its meshing is UHPC
shownwith single
in Fig.10b. established
The model is composed ABAQUS and
of three parts, illustrated
that is, concrete
Fig.10a
a short andof
length its1meshing
mm andis ashown in Fig.10b.
rectangular The model
section 0.37 mm2of. To
of 0.53is×composed three parts, that
perform is, concrete
a dense meshing on
matrix, TSF and their interface. The TSF has an embedment length of 20 mm, a free end with a short
matrix, TSF
the concrete areaandclose
their interface. The TSF has anwith
embedment length of 20asmm,
big aasfree end with a short
length of 1 mm and a to TSF, a solid
rectangular column
section of 0.53×a diameter
0.37 mm 2twice
. To perform athe equivalent
dense meshing diameter
on the
length
of TSF was of 1 mm
partitioned.and a rectangular section of 0.53× 0.37 mm 2. To perform a dense meshing on the
concrete area close to TSF, a solid column with a diameter twice as big as the
concrete area close to TSF, a solid column with a diameter twice as big as the equivalent diameter of equivalent diameter of
The partitioned.
TSF TSF
was interface between concrete matrix and TSF was modelled as interfacial transition zone
was partitioned.
(ITZ),The
theinterface
thickness of which
between is twice the equivalent diameter of TSF. The steel fiber element is (ITZ),
C3D8
The interface betweenconcrete
concretematrix
matrix andandTSF
TSFwaswas modelled
modelled as interfacial
as interfacial transition
transition zone
zone (ITZ),
hexahedral
the the
thickness element
thickness of ofwhichwith average
whichisistwice
twice the
side length
the equivalent
of 90 µm. The
diameterof of
equivalent diameter TSF.
concrete
TSF.
TheThe
and
steelsteel
ITZ
fiberfiber
elements
element
element
are
is C3D8
C3D4
is C3D8
tetrahedral
hexahedral elements
element and
with their side
average length
side increases
length of 90 from
µm. 50
The µm to
concrete
hexahedral element with average side length of 90 µm. The concrete and ITZ elements are C3D4 1.5 mm
and ITZwhen they
elements are
are away
C3D4 from
the TSF.
tetrahedralTheelements
tetrahedral meshing
elements sensitivity
andtheir
and waslength
theirside
side analyzed
length and the
increases
increases from results
from 5050
µm µmshow
to 1.5 thatmm
mm
to 1.5 thiswhen
when fine
theymeshing are used
are away
they from infrom
away this
simulation
the TSF. is
Thesufficient
meshing to produce
sensitivity stable
was pullout
analyzed force
and the results.
results show that
the TSF. The meshing sensitivity was analyzed and the results show that this fine meshing used in this fine meshing used in
this simulation is sufficient to produce stable pullout force results.
this simulation is sufficient to produce stable pullout force results.

(a) (b)
Figure
Figure 10.10.The
TheFEA
FEAmodel
model for
for UHPC
UHPCwith
withsingle TSF,
single (a) (a)
TSF, illustration, (b) meshing.
illustration, (b) meshing.
(a) (b)
4.2. Material Model10. The FEA model for UHPC with single TSF, (a) illustration, (b) meshing.
Figure

4.2. Material Model

Appl. Sci. 2019, 9, 658 13 of 17

4.2.
Appl.Material
Sci. 2018, Model
8, x FOR PEER REVIEW 13 of 17

The Elastic Perfectly Plastic model of steel fiber was adopted, as shown in Figure 11. The density,
The Elastic Perfectly Plastic model of steel fiber was adopted, as shown in Fig.11. The density,
Young’s modulus, Poisson’s ratio and yield stress of steel fiber are 7800 kg/m3, 206 GPa, 0.3 and
Young’s modulus, Poisson’s ratio and yield stress of steel fiber are 7800 kg/m3, 206 GPa, 0.3 and 2500
2500 MPa, respectively.
MPa, respectively.

Figure 11.
Figure 11. The Elastic Perfectly Plastic
Plastic model
model of
of steel
steel fiber.
fiber.

Testing
Testingresults
resultsininexperimental
experimentalsection
sectionshow
show that
that the
the matrix
matrix channel
channel waswas not
not damaged
damaged after after the
the
steel fiber was pulled out, indicating that the steel fiber debonded from the concrete-fiber
steel fiber was pulled out, indicating that the steel fiber debonded from the concrete-fiber interface interface
without
without damaging
damaging the the concrete.
concrete. Therefore,
Therefore, the
the concrete
concrete material
material was
was simulated
simulated byby using
using its its Elastic
Elastic
Perfectly
Perfectly Plastic model with density
density of
of 2400
2400 kg/m
kg/m 3,3Young’s
, Young’smodulus
modulusofof 4646 GPa,
GPa, yield
yield stress
stress (close
(close to
to the compressive strength) of 155 MPa and Poisson’s
the compressive strength) of 155 MPa and Poisson’s ratio of 0.2. Theratio of 0.2. The stress-strain behavior of this
this
Elastic
Elastic Perfectly
Perfectly Plastic
Plastic model
model for
for concrete
concreteisissimilar
similarto tosteel,
steel,as
asshown
shownin inFigure
Fig.11.11.
The
The friction
friction model
model of of matrix-fiber
matrix-fiber interface
interface was
was simulated
simulated by by using
using load-independent,
load-independent, isotropicisotropic
Coulomb friction law. If the equivalent shear stress reach the critical shear value
Coulomb friction law. If the equivalent shear stress τeqeq reach the critical shear value τcrit
τ τ , the interface
crit the interface

starts
starts debonding.
debonding. q
τeq = τ 2 + τ22 (25)
𝜏 = 𝜏1 𝜏 (25)
τcrit = µpcontact (26)

where τ 1 and τ 2 are two orthogonal tangential 𝜏 =shearing

𝜇𝑝 stress on the interface, pcontact is the normal
(26)
force at interface, µ is the friction coefficient, which is determined by referring to the experimental
where τby
results 1 and τ2 are
Baltay andtwo orthogonal
Gjelsvik [19]. tangential shearing
The friction stress at
coefficient onconcrete-steel
the interface, interface
pcontact is the normal
is mainly
force at interface,
determined by the μprocessing
is the friction coefficient,
methods of steelwhich
and it is
wasdetermined
ranged from by 0.35
referring to Thus,
to 0.58. the experimental
the friction
results by Baltay and Gjelsvik [19]. The friction coefficient at concrete-steel
coefficient of 0.5 at concrete-steel interface was set in the model in this study. interface is mainly
determined by the processing methods of steel and it was ranged from 0.35 to 0.58. Thus, the friction
4.3. Loadingof
coefficient and
0.5Boundary Conditioninterface was set in the model in this study.
at concrete-steel
A reference point was coupled with TSF and then the displacement loading was applied to the
4.3. Loading and Boundary Condition
reference point. The four concrete side faces were simply supported, the left side of concrete was fixed
and theA reference
right sidepoint was coupled
of concrete was notwith TSF and then the displacement loading was applied to the
restrained.
reference point. The
The pullout fourwas
process concrete side faces
simulated were simply
as a quasi-static supported,
process thedynamic
by using left side explicit
of concrete was
method
fixed
in and the right side of
ABAQUS/Explicit. concrete
The was nottechnology
mass scaling restrained. was applied to decrease computation cost.
The pullout
To ensure processpullout
the quasi-static was simulated
process isasmaintained,
a quasi-static
theprocess by using
kinetic energy dynamic
of TSF explicit
should method
be sufficiently
in ABAQUS/Explicit. The mass scaling technology was applied to decrease computation
small. This is achieved by maintaining the ratio of kinetic energy internal energy of less than 5%. cost. To
ensure the quasi-static
Consequently, the fiber pullout processatisthe
displacement maintained, the kinetic
reference point energy of as
was controlled TSF should
0.01 mm/s.be sufficiently
small. This is achieved by maintaining the ratio of kinetic energy to internal energy of less than 5%.
4.4. Numerical Results
Consequently, the fiber displacement at the reference point was controlled as 0.01 mm/s.
The FEA results of fiber pullout force were then compared with the average experimental results
4.4.the
for Numerical Results
fiber embedment length of 20 mm, as shown in Figure 12. The numerical results match well
with the testing results in thepullout
The FEA results of fiber Linear force
debonding stage
were then and the early
compared stage
with the of fiber
average untwisting; results
experimental while,
the
for pullout
the fiberforce slightly length
embedment decreased
of 20and
mm,stabilized
as shown with furtherThe
in Fig.12. pullout displacements
numerical fromwell
results match 6 mm to
with
20 mm, causing a lower simulated pullout force compared to the testing value. The higher
the testing results in the Linear debonding stage and the early stage of fiber untwisting; while, thetesting
pullout force slightly decreased and stabilized with further pullout displacements from 6 mm to 20
mm, causing a lower simulated pullout force compared to the testing value. The higher testing
pullout force may be caused by the creep-induced resistance from the 28-day curing concrete, while
this creep was not considered in our FEA model. In addition, the tangential frictional resistance after
Appl. Sci. 2019, 9, 658 14 of 17

pullout force may be caused by the creep-induced resistance from the 28-day curing concrete, while
Appl.Sci.
Appl. Sci.2018,
2018,8,8,xxFOR
FORPEER
PEERREVIEW
REVIEW 1414ofof1717
this creep was not considered in our FEA model. In addition, the tangential frictional resistance after
the debonding
thedebonding
the between
debondingbetween
betweenthethe UHPC
theUHPC matrix
UHPCmatrix and
matrixand the
andthe twisted
thetwisted steel
twistedsteel fiber
fiberisis
steelfiber not
isnot considered
notconsidered
consideredinin the
inthe FEA
theFEAFEA
model; while
model;while
model; minor
whileminor tangential
minortangential frictional
tangentialfrictional resistance
frictionalresistance may
resistancemay still
maystill exist,
stillexist,producing
exist,producing the
producingthe extra pullout
theextra
extrapulloutforce
pulloutforce
forcein
the testing
ininthe results.
thetesting
testingresults.
results.
250
250

200
200
force/N
Pullout force/N

150
150
Experimental
Experimental
Pullout

Modeling
Modeling
100
100

50
50

00
00 55 10
10 15
15 20
20
FiberEnd
Fiber EndSlip/mm
Slip/mm
Figure 12. Comparison between numerical and experimental results for the case with fiber embedment
Figure 12.
Figure 12. Comparison
Comparison between
between numerical
numerical and
and experimental
experimental results
results for
for the
the case
case with
with fiber
fiber
length of 20 mm.
embedmentlength
embedment lengthofof20
20mm.
mm.
Figure 13 shows the evolution of Mises stress response over the pullout course at pullout
Fig.13 shows
Fig.13
displacements shows
of 0,thethe mm,
4.0 evolution
evolution
8.0 mm, ofof12.0
Mises
Misesmm. stress
stress response
16.0 response
mm over
over
and 20.0 theInterestingly,
the
mm. pullout course
pullout course
the atat pullout
pullout
untwisting
displacements
displacements of
of 0,
0, 4.0
4.0 mm,
mm, 8.0
8.0 mm,
mm, 12.0
12.0 mm.
mm. 16.0
16.0 mm
mm and
and 20.0
20.0 mm.
mm. Interestingly,
Interestingly,
phenomenon was found within the effective untwisting fiber length of around 1.7 mm away from the the
the untwisting
untwisting
phenomenon
phenomenon
exterior wassurface.
was
concrete foundwithin
found within
The theeffective
the effective
untwisting untwistingfiber
wasuntwisting
completed fiberlength
when length ofofaround
the whole around 1.7pulled
1.7
fiber was mmaway
mm away from
from
out at the
the
the exterior
exterior concrete
concrete surface.
surface. The
The untwisting
untwisting was
was completed
completed when
when the
the whole
whole fiber
fiber
end of the simulation. The above observation is consistent with our experimental study and others was
was pulled
pulled out
out atat
the
the end
end
such asofof the
theet
Ellis simulation.
simulation. The above observation is consistent with our experimental study
al. [14]. The above observation is consistent with our experimental study and others and others
suchas
such asEllis
Ellisetetal.
al.[14].
[14].

Figure 13. The Mises stress evolution over the pullout process.
Figure13.
Figure 13.The
TheMises
Misesstress
stressevolution
evolutionover
overthe
thepullout
pulloutprocess.
process.
The above numerical and testing results reveal the strain hardening behavior of UHPFRC with
TSFs The
The abovenumerical
above
is caused numerical
by the fiber and
and testingresults
testing
untwisting results reveal
at thereveal theuntwisting
the
effective strainhardening
strain hardening behavior
lengthbehavior
near theofofUHPFRC
UHPFRC
exterior with
with
concrete
TSFsisiscaused
TSFs causedbybythe
thefiber
fiberuntwisting
untwistingatatthe
theeffective
effectiveuntwisting
untwistinglength
lengthnear
nearthe
theexterior
exteriorconcrete
concrete
surface.This
surface. Thissupports
supportsthetheassumption
assumptionininour
ourtheoretical
theoreticalmodel
modelthat
thatonly
onlyconsiders
considersthe
thefriction
frictionononthe
the
fiber-fiberinterface
fiber-fiber interfaceatatthe
theeffective
effectiveuntwisting
untwistinglength,
length,neglecting
neglectingthe
thetangential
tangentialfrictional
frictionalresistance
resistance
afterthe
after thedebonding
debondingbetween
betweenthe theUHPC
UHPCmatrix
matrixand andthe
thetwisted
twistedsteel
steelfiber.
fiber.Table
Table11displays
displaysthat
thatthe
the
Appl. Sci. 2019, 9, 658 15 of 17

surface. This supports the assumption in our theoretical model that only considers the friction on the
fiber-fiber interface at the effective untwisting length, neglecting the tangential frictional resistance
after the debonding between the UHPC matrix and the twisted steel fiber. Table 1 displays that the
tested effective untwisting length is insensitive to the fiber embedment length. To further optimize
the pullout behavior, further study on the strategies of extending effective untwisting length can be
thus performed.
The friction at matrix-fiber interface and its adhesive bond are critical to increase the pullout
force. Consequently, the matrix-fiber interface with optimum property is needed. For the same mix
proportion, an optimum mixing and casting is thus required to produce an intact interface with a
minimum number of pores and other defects. Figure 5 shows a number of pores that were caused
during the casting process, though we conducted a careful casting procedure, similar to studies such
as [23]. The dense mixture with a low water-binder ratio might increase the difficulty of escaping the
bubble during the casting. Future study can be carried out to improve the casting quality of UHPFRC
and investigate the influence of air bubbles on the matrix-fiber interface and thus the pullout force.
The significant strain hardening in the plastic untwisting stage causes an optimized property of
UHPFRC with TSFs. This strain hardening causes a significant increase in energy dissipation area
enclosed by the pullout curves. Thus, the twisting is an ideal way to improve the strain hardening
property and the UHPFRC with TSFs may be applied well in the concrete structures subjected to
earthquakes or shocking to absorb dynamic energy.

5. Conclusions
In this paper, a theoretical formula for pullout force was first derived according to the force analysis
of twisted steel fiber (TSF) of rectangular section and triangular section. The model assumptions
were explicitly made based on the experiment data obtained from the existing studies. Then the
experiment on the ultra-high-performance fiber reinforced concrete (UHPFRC) with TSF at three
embedment lengths was performed to gain insights into the strain hardening characteristics due
to untwisting mechanism. A simplified numerical study was also performed to understand the
untwisting behavior inside the concrete. The microscopic X-ray tomography was used to compare the
internal microstructure of the UHPC matrix before and after pullout. The following conclusion can
be made:
• The pullout process of a single TSF in UHPFRC can be divided into a linear debonding stage,
plastic untwisting stage and a dropping stage, so the pullout force versus displacement curve can
be simplified as a trilinear curve.
• A significant strain hardening characteristic occurred due to the plastic deformation of TSF and
its untwisting deformation during the pullout process. This untwisting deformation occurred on
the fiber (within the effective untwisting fiber length) close to the exterior concrete surface.
• The fiber channel inside the matrix was not damaged after pullout, indicating a strong pullout
resistance of matrix-fiber interface due to the high mechanical strength of matrix and strong
bonding between matrix and TSF.
• The pullout force is independent of the embedment length of TSF and dependent of the geometry
of fiber cross-section as well as the friction at matrix-fiber interface and its adhesive bond.
• The developed analytical formula was verified by the experimental data and thus can be used
to gain insights into the mechanism of pullout process and to understand the parameters that
influence the mechanical property of UHPFRC with TSFs.

Author Contributions: Conceptualization, J.Y. and G.L.; methodology, J.Y. and G.L.; software, J.Y.; validation, J.Y.
and G.L.; formal analysis, J.Y. and G.L.; investigation, J.Y.; resources, G.L.; data curation, J.Y.; writing—original
draft preparation, J.Y.; writing—review and editing, G.L.; visualization, J.Y.; supervision, G.L.; project
administration, G.L.; funding acquisition, J.Y. and G.L.
Funding: This project was supported by the funding from the National Natural Science Foundation of China
(Grant No. 51379185) and National Basic Research Program of China (973 Program, 2013CB036901).
Appl. Sci. 2019, 9, 658 16 of 17

References
1. Naaman, A.E. Engineered steel fibers with optimal properties for reinforcementof cement composites. J. Adv.
Concr. Technol. 2003, 1, 241–251. [CrossRef]
2. Wille, K.; Naaman, A.E.; El-Tawil, S.; Parra-Montesinos, G.J. Ultra-high performance concrete and fiber
reinforced concrete: Achieving strength and ductility without heat curing. Mater. Struct. 2012, 45, 309–324.
[CrossRef]
3. Wille, K.; El-Tawil, S.; Naaman, A.E. Properties of strain hardening ultra high performance fiber reinforced
concrete (UHP-FRC) under direct tensile loading. Cement Concr. Compos. 2014, 48, 53–66. [CrossRef]
4. Yu, R.; Spiesz, P.; Brouwers, H.J. Energy absorption capacity of a sustainable Ultra-High Performance Fibre
Reinforced Concrete (UHPFRC) in quasi-static mode and under high velocity projectile impact. Cement Concr.
Compos. 2016, 68, 109–122. [CrossRef]
5. Venkateshwaran, A.; Tan, K.H.; Li, Y. Residual flexural strengths of steel fiber reinforced concrete with
multiple hooked-end fibers. Struct. Concr. 2018, 19, 352–365. [CrossRef]
6. Zhao, M.; Zhao, M.; Chen, M.; Li, J.; Law, D. An experimental study on strength and toughness of steel fiber
reinforced expanded-shale lightweight concrete. Constr. Build. Mater. 2018, 183, 493–501. [CrossRef]
7. Qu, D.; Cai, X.; Chang, W. Evaluating the Effects of Steel Fibers on Mechanical Properties of Ultra-High
Performance Concrete Using Artificial Neural Networks. Appl. Sci. 2018, 8, 1120. [CrossRef]
8. Kang, S.T.; Kim, J.; Lee, B. Effects of Water Reducing Admixture on Rheological Properties, Fiber Distribution
and Mechanical Behavior of UHPFRC. Appl. Sci. 2019, 9, 29. [CrossRef]
9. Wang, D.; Ju, Y.; Zheng, W.; Shen, H. Seismic Behavior and Shear Bearing Capacity of Ultra-High Performance,
Fiber-Reinforced Concrete (UHPFRC) Beam-Column Joints. Appl. Sci. 2018, 8, 810. [CrossRef]
10. Ferreira, S.R.; Pepe, M.; Martinelli, E.; de Andrade Silva, F.; Toledo Filho, R.D. Influence of natural fibers
characteristics on the interface mechanics with cement based matrices. Compos. Part B Eng. 2018, 140,
183–196. [CrossRef]
11. Kim, M.J.; Kim, S.; Yoo, D.Y. Hybrid Effect of Twisted Steel and Polyethylene Fibers on the Tensile
Performance of Ultra-High-Performance Cementitious Composites. Appl. Sci. 2018, 10, 879. [CrossRef]
12. De Andrade Silva, F.; Mobasher, B.; Soranakom, C.; Toledo Filho, R.D. Effect of fiber shape and morphology
on interfacial bond and cracking behaviors of sisal fiber cement based composites. Cement Concr. Compos.
2011, 33, 814–823. [CrossRef]
13. Kim, D.J.; El-Tawil, S.; Naaman, A.E. Loading rate effect on pullout behavior of deformed steel fibers.
ACI Mater. J. 2008, 105, 576. [CrossRef]
14. Ellis, B.D.; McDowell, D.L.; Zhou, M. Simulation of single fiber pullout response with account of fiber
morphology. Cement Concr. Compos. 2014, 48, 42–52. [CrossRef]
15. Lee, Y.; Kang, S.T.; Kim, J.K. Pullout behavior of inclined steel fiber in an ultra-high strength cementitious
matrix. Constr. Build. Mater. 2010, 24, 2030–2041. [CrossRef]
16. Alkaysi, M. Strength and Durability of Ultra-High Performance Concrete Materials and Structures.
Ph.D. Thesis, University of Michigan, Ann Arbor, MI, USA, 2016.
17. Wille, K.; Naaman, A.E. Effect of ultra-high-performance concrete on pullout behavior of high-strength
brass-coated straight steel fibers. ACI Mater. J. 2013, 110, 451. [CrossRef]
18. Qsymah, A.; Sharma, R.; Yang, Z.; Margetts, L.; Mummery, P. Micro X-ray Computed Tomography
Image-based Two-scale Homogenisation of Ultra High Performance Fibre Reinforced Concrete. Constr. Build.
Mater. 2017, 130, 230–240. [CrossRef]
19. Baltay, P.; Gjelsvik, A. Coefficient of friction for steel on concrete at high normal stress. J. Mater. Civ. Eng.
1990, 2, 46–49. [CrossRef]
20. Hassan, A.M.; Jones, S.W.; Mahmud, G.H. Experimental test methods to determine the uniaxial tensile and
compressive behaviour of ultra high performance fibre reinforced concrete (UHPFRC). Constr. Build. Mater.
2012, 37, 874–882. [CrossRef]
21. China Association for Engineering Construction Standardization. Test Methods Used for Steel Fiber Reinforced
Concrete; CECS13:89; China Association for Engineering Construction Standardization: Beijing, China, 1989;
pp. 30–35. (In Chinese)
22. Volume Graphics Gmbh. VGStudio MAX 3.0; Volume Graphics Gmbh: Heidelberg, Germany, 2014.
Appl. Sci. 2019, 9, 658 17 of 17

23. Wu, Z.; Wong, H.S.; Buenfeld, N.R. Effect of confining pressure and microcracks on mass transport properties
of concrete. Adv. Appl. Ceram. 2014, 113, 485–495. [CrossRef]
24. Wu, Z.; Wong, H.S.; Buenfeld, N.R. Influence of drying-induced microcracking and related size effects on
mass transport properties of concrete. Cement Concr. Res. 2015, 68, 35–48. [CrossRef]
25. Popovics, S. A numerical approach to the complete stress-strain curve of concrete. Cement Concr. Res. 1973, 3,
583–599. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).