Engineering Structures 118 (2016) 1–15

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Evaluation of the flexural behavior of composite beam with inverted-T

steel girder and steel fiber reinforced ultra high performance concrete
slab
Sung-Won Yoo a, Jinkyo F. Choo b,⇑
a
Department of Civil and Environmental Engineering, Woosuk University, Jinchon 355-803, Republic of Korea
b
Department of Civil Engineering, Konkuk University, 120 Neungdong-ro, Gwangjin-gu, Seoul 143-701, Republic of Korea

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

Article history: Huge efforts have been made to develop ultra high performance concrete (UHPC) and exploit its remark-
Received 17 November 2014 able properties. Among the achievements, various solutions were proposed to achieve optimal composite
Revised 17 March 2016 beams using steel fiber reinforced UHPC. In order to increase the economy in material, a composite beam
Accepted 21 March 2016
combining a slab made of UHPC and a steel girder without top flange is proposed. In this composite beam,
Available online 1 April 2016
the inverted-T steel girder requires the studs be arranged on the web for the composition with the UHPC
slab. Considering the absence of studies evaluating the flexural behavior of this new type of composite
Keywords:
beam, this study examines experimentally the flexural behavior with respect to the stud spacing and slab
Composite beam
UHPC deck
thickness. To that goal, eight composite beams with varying stud spacing and UHPC slab thickness were
Inverted-T steel girder fabricated together with two additional composite beams using slabs made of normal concrete without
Flexural behavior steel fiber reinforcement for comparison. In addition, an analysis method considering the tension-
Stud spacing softening behavior of steel fiber reinforced UHPC in the material model and the corresponding strain
Slab thickness compatibility conditions of the beam member is proposed. The comparison of the analytic and experi-
Tension softening mental results reveals the good accuracy of the predictions indicating that the empirical tension-
softening curve reflects reasonably the actual behavior of steel reinforced UHPC composite member.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction France and Germany developed a composite beam connecting an

Modern concrete structures require materials with higher
strength, higher performance and higher durability to deal with
the ever-increasing height and length of structures and resist to
harsher environmental conditions. However, conventional con-
crete is brittle and develops low tensile and compressive strengths.
Therefore, studies were undertaken worldwide on steel fiber rein-
forced ultra high performance concrete (UHPC) with compressive
strength higher than 180 MPa [1–3]. Compared to conventional
concrete, steel fiber reinforced UHPC is known to exhibit improved
deflection and flexural strength and post-cracking ductile behavior.
The exploitation of UHPC in bridge deck is expected to extend
significantly the lifetime and reduce significantly the weight of
the structure [1,4–6].
A review of previous studies dedicated to steel composite
beams using UHPC reveals that a European commission led by
⇑ Corresponding author. Tel.: +82 2 2049 6246; fax: +82 2 2201 0783.
E-mail addresses: imysw@woosuk.ac.kr (S.-W. Yoo), jfchoo@konkuk.ac.kr
(J.F. Choo).
inverted T-shape steel girder and a UHPC deck by giving a puzzle
shape to the top of the girder as shown in Fig. 1 [7]. The US Federal
Highway Administration conducted studies on the connection of
composite beams using an I-shape steel girder and UHPC waffle
deck panels by installing various shapes of shear connectors at
the top flange of the steel girder and by filling polymer concrete
in the shear connectors at the top of the steel girder and the shear
pockets arranged at the top of the precast waffle deck [2,8]
(see Fig. 2).
From this survey, it appears that the top flange of the steel
girder might be superfluous when the composite beam is formed
by composing a UHPC slab with the steel girder considering the
high stiffness developed by the UHPC deck and that the shear
connector usually applied for the construction of composite beams
can play the role fulfilled by the top flange of the steel girder. Based
upon this observation, this study applies an inverted T-shape steel
girder in which the upper flange of the girder is removed and
replaced by shear connector for the composition of the composite
beam shown in Fig. 3 [7,9–11].

http://dx.doi.org/10.1016/j.engstruct.2016.03.052
0141-0296/Ó 2016 Elsevier Ltd. All rights reserved.
2 S.-W. Yoo, J.F. Choo / Engineering Structures 118 (2016) 1–15

Table 1
Designation of test members and test variables.

Member designation Slab thickness (mm) Stud spacing (mm)

50-50 50 50
50-100 50 100
50-200 50 200
50-400 50 400
100-50 100 50
100-100 100 100
100-200 100 200
100-400 100 400
N50-100 50 100
N100-100 100 100
Fig. 1. Inverted-T steel girder with puzzle-shape [7].

2. Test setup

2.1. Test variables and test members

Fig. 2. Waffle deck UHPC panel system [2,8].
By adopting such configuration, the studs needed for the com-
position of the steel girder with the UHPC slab must be disposed
in the web of the girder because of the absence of the upper flange.
However, very few experimental and theoretical studies were ded-
icated to the behavior of the shear connectors welded on the web
of the girder and to the behavioral characteristics of the composite
beam with the inverted-T steel girder.
Accordingly, this study intends to evaluate the behavioral
characteristics of the shear connectors and the flexural behavior
of the composite beam using the inverted-T steel girder. To that
goal, eight composite beams with varying stud spacing and slab
thickness were fabricated and tested. Moreover, two additional
composite beams using the inverted-T steel girder and slabs made
of 50-MPa normal concrete without steel fiber reinforcement were
also fabricated for comparison with the UHPC composite beams.
The stud spacing and slab thickness are selected as test vari-
ables. Therefore, two thicknesses of 50 mm and 100 mm are
adopted for the UHPC slab and, four different lengths of 50, 100,
200 and 400 mm are chosen for the longitudinal spacing of the
studs leading to a total of 8 UHPC composite beam prototypes. In
addition, two supplementary composite beams using the
inverted-T steel girder and slabs made of 50-MPa normal concrete
without steel fiber reinforcement are also considered. All the test
members are fabricated with a span length of 2200 mm. The girder
presents a bottom flange with thickness of 13 mm and width of
225 mm, and its web is 8-mm thick and 177-mm high. In addition,
the UHPC slab is placed with thickness of 50 mm and 100 mm and
width of 248 mm at 137 mm along the height of the web.
The slab thickness is selected as test variable to examine the
effect of the embedment depth of the shear connector welded on
the girder web. The change in stud spacing intends to evaluate
the composite behavior of the composite beam and to derive the
appropriate spacing for the configuration shown in Fig. 3.
Table 1 gives the designation of the test members according to
the test variables. Fig. 4 indicates the dimensions of the test mem-
bers. Figs. 5–7 present the girder installed in the steel form, the
completed test members and a view of the loading test. In the sen-
sor layout shown in Fig. 7(a), a total of 18 strain gages with 9 con-
crete strain gages and 9 steel strain gages are attached to the front
and rear faces of the specimen and, two LVDTs (linear variable dif-
ferential transformer) are installed at 1/2 and 1/3 of the span to
measure the deflection. In addition, strain gages are also embedded
inside the specimen and attached to the top and side faces of all the
studs located inside the 400-mm wide area around the mid-span.

Fig. 3. Cross-sectional configuration of conventional composite beam (left) and composite beam with inverted-T shaped girder (right).
 S.-W. Yoo, J.F. Choo / Engineering Structures 118 (2016) 1–15
Fig. 4. Shape and dimensions of composite beam specimens.
Fig. 5. Inverted T-shape steel girder installed in form.
Fig. 6. Manufactured test members.
After placing, an error of maximum 2 mm was measured in the
dimensions of the UHPC slab with respect to the planned ones. This
very small error indicates the high precision of the fabrication of
the specimens.
2.2. Characteristics of concrete, UHPC, shear connectors and steel
girder
The mix composition of UHPC used in this study complies with
the Design Recommendations for Ultra-High Performance Concrete
3
Fig. 7. Loading test of composite beam specimen.
K-UHPC of the Korea Concrete Institute [12] as done in previous
works [11,13,14]. Tables 2 and 4 list the mix proportions of
K-UHPC and the corresponding properties obtained by cylinder
and flow tests, respectively. In Table 2, the proportions are
expressed in weight ratio by setting the weight of cement to 1. In
addition, steel fiber is admixed in a volume fraction of 2% with
two different diameters of 19.5 mm and 16.3 mm at ratio of 2:1,
respectively. Besides, the mix proportions of the 50-MPa concrete
used in the beam specimens N50-100 and N100-100 manufactured
for comparison with the specimens using UHPC are listed in Table 3.
The shear connector shown in Fig. 8 presents shank diameter
D = 13 mm, head diameter of 22 mm, and length H = 50 mm. From
direct tensile test, the yield strength and ultimate strength of the
shear connector are found to be 370 MPa and 470 MPa, respec-
tively. Moreover, the connection of the shear connector with the
steel girder is achieved by stud gun welding as usually done in con-
struction site so as to secure realistic conditions. The steel girder is
made of structural steel SM490. Referring to the data provided by
4 S.-W. Yoo, J.F. Choo / Engineering Structures 118 (2016) 1–15

Table 2
Mix proportions of K-UHPC in weight ratio [11,12].

Cement Zirconium Sand Filler Expansive agent Shrinkage reducing agent Plasticizer Water Antifoaming agent Steel fiber
(mm)
19.5 16.3
1 0.25 1.10 0.30 0.075 0.01 0.03 0.23 0.001 0.10 0.05

Table 3
Mix proportions of 50-MPa normal concrete.
S/a W/B Unit weight (kg/m3)
(%) (%)
Water Cement Fly Sand Gravel
ash
50 25 130 262 263 745 919
the manufacturer, the yield strength and ultimate strength of the
steel girder are 396 MPa and 554 MPa, respectively.
3. Test results and discussion
3.1. Crack and failure patterns of test members
Table 5 summarizes the crack and failure patterns observed
during the three-point loading test of the test members.
Figs. 9–14 show representative crack and failure patterns
observed in the composite beam specimens during the three-
point loading test. The observation of these patterns will help us
to identify eventual trend in the cracking and failure of the com-
posite beam according to the slab thickness and stud spacing.
In Fig. 9(a), member 50-400 with the thinnest 50-mm slab
thickness and the largest 400-mm stud spacing considered in this
study experienced axial cracking along the steel girder at the top
face of the slab. This crack pattern seems to be have been caused
by the tensile stress generated by the debonding of the studs at
mid-span, which distributed the load concentrated in a small area
around the stud to the neighboring concrete. In Fig. 9(b), the flex-
ural cracks in member 50-400 occurred on the lateral face of the
slab in a 400-mm wide zone around the shear connector at mid-
span. A total of 10 cracks with maximum width of 0.02 mm were
identified.
For member 100-400 with the thickest 100-mm slab thickness
and largest 400-mm stud spacing in Fig. 10, no axial crack was
observed at the top face of the slab but flexural cracks and
flexure-shear cracks propagated on the lateral face of the slab in
an 800-mm wide zone around the mid-span shear connector.
Approximately 20 cracks with maximum width of 0.03 mm were
identified.
For member 100-100 with 100-mm slab thickness and 100-mm
stud spacing in Fig. 11, no axial crack was observed at the top face
of the slab, and concrete failed by compression. Flexural cracks
occurred on the lateral face in a 1000-mm wide zone around the
shear connector at mid-span. Twenty-two cracks with maximum
width of 0.02 mm were identified.
Superplasticizer
In Fig. 12, member 50-50 with the thinnest slab thickness and
shortest stud spacing considered in this study showed no axial
1.5% of
crack on the top face of its slab and underwent compressive failure
cement weight of concrete at mid-span. Only 4 flexural cracks with maximum
width of 0.02 mm were detected at mid-span on the lateral face.
Figs. 13 and 14 show the failure pattern of the beam specimens
N50-100 and N100-100 using 50-MPa normal concrete and with-
out steel fiber reinforcement. The slabs made of normal concrete
and without steel fiber reinforcement experienced only one wide
crack near mid-span on their lateral face, and cracks propagating
along the axis of the girder on their top face. For the thinner slab,
a large number of wide cracks also developed around the connec-
tion with the steel girder. This situation can be attributed to the
absence of steel fiber and stresses the absolute necessity to arrange
reinforcing bars when steel fiber is not used.
The observation of the crack and failure patterns of the compos-
ite beams using UHPC slab indicates that larger stud spacing with a
thin slab induces axial cracks along the steel girder at the top face
of the slab. The occurrence of these axial cracks can be attributed to
the tensile stress generated by the redistribution of the load con-
centrated in a small area around the stud to the neighboring con-
crete. Moreover, the adoption of thick slab of 100 mm appears to
prevent the occurrence of axial cracking on the top face whereas,
except for the stud spacing of 50 mm, all the test members with
slab thickness of 50 mm underwent axial cracking on the top face.
For the test members with slab thickness of 50 mm, concrete
failure occurred by tension for stud spacing of 400 mm and, by
compression for stud spacing smaller than 200 mm. Besides, for
the test members with slab thickness of 100 mm, failure of con-
crete occurred by flexure for stud spacing of 400 mm and, by com-
pression for stud spacing smaller than 200 mm.
3.2. Load–displacement curves
Fig. 15 plots the load–displacement curves measured in the
loading test of the 10 considered test members. It appears that
the structural performance improves with larger slab thickness
as well as smaller stud spacing. In addition, the ultimate load
and behavior of the specimens using normal concrete are clearly
inferior to those exhibited by the members using UHPC, which
stress once again the critical role played by the steel fibers mixed
in UHPC.

Table 4
Properties of the concretes considered in this study.

Concrete type Cylinder Compressive strength (MPa) Elastic modulus (MPa) Ultimate strain, eu Slump flow (mm) Air content (%)
K-UHPC 1 187 45,385 0.00400 765 2.9
2 186 45,124 0.00402
3 176 44,491 0.00416
Average 183 45,000 0.00406
Normal concrete 1 59 27,621 0.00311 796 3.4
2 56 26,924 0.00314
3 62 28,255 0.00305
Average 59 27,600 0.00310
 S.-W. Yoo, J.F. Choo / Engineering Structures 118 (2016) 1–15 5

Fig. 8. Shear connector for direct tensile test. (a) Top face of UHPC slab

Besides, member 50-400 develops clearly lower flexural

strength than the other specimens using UHPC. Even if this situa-
tion can be attributed to a certain extent to experimental or fabri-
cation error, we should recall that there was only a difference
smaller than 2 mm in the fabricated specimens as compared to
the planned dimensions. Therefore, the main cause should be
found in the thin thickness of the slab and the excessively large
stud spacing rather than in the experimental error.
Despite of its excessively wide stud spacing, member 100-400
presents slightly smaller flexural strength owing to the contribu-
tion of its thicker slab. Nevertheless, the ultimate loads of all the
members with UHPC slab thickness of 50 mm are still slightly
larger than those of members N50-100 and N100-100 using
50-MPa normal concrete and stud spacing of 100 mm but there
remains room for a certain level of ductile behavior in view of
the large deflections observed. This ductile behavior can be attrib- (b) Side face of UHPC slab
uted to the contribution of the steel fiber in UHPC. Moreover, these
results also indicate that composite behavior cannot be realized in Fig. 9. Crack pattern of member 50–400.
case of excessively large stud spacing.
Eurocode-4 [15,16] defines the characteristic relative slip, duk, as
behavior limit of 6 mm recommended in Eurocode-4. Besides, the
0.9 du, where du is the relative slip measured at load PRK corre-
characteristic relative slip of member N50-100 is 4.98 mm and
sponding to the ultimate load reduced by 10%. Eurocode-4 assesses
does not satisfy the criterion of Eurocode-4 and that of member
ductile behavior when this characteristic relative slip is larger than
N100-100 satisfies merely the criterion with value of 6.47. These
6 mm and states that stable behavior is achieved by the composite
results verify once again the effect of the steel fiber and slab
member in such case. Table 6 arranges the crack load, the yield
thickness.
load at the bottom flange of the steel girder, the ultimate load,
the failure load, the relative slip at 90% of the failure load, and
the characteristic relative slip measured in the tests to conduct 3.3. Load–strain curves of UHPC slab
comparison with the stipulations of Eurocode-4 [16].
In view of Table 6, the test members develop characteristic Figs. 16 and 17 plot the load–concrete strain curves at the top
relative slip ranging between 9.41 and 15.16 mm, which indicate and bottom of the UHPC slab, respectively. The ultimate compres-
sufficient ductile behavior since these values far exceed the ductile sive strain of the UHPC slab is approximately 0.004, which is in

Table 5
Crack and failure patterns of test members.

Test member Occurrence of longitudinal Crack on lateral face of slab Failure pattern
crack on slab
Extension from No. of cracks Crack width (mm)
mid-span (mm)
50-50 No Center 4 0.02 Flexural compression
50-100 Yes 400 8 0.02 Flexural compression
50-200 Yes 800 18 0.02 Flexural compression + Failure of concrete around stud
50-400 Yes 400 10 0.02 Flexural compression + Failure of concrete around stud
100-50 No 1000 13 0.02 Flexural compression
100-100 No 1000 22 0.02 Flexural compression
100-200 No 1000 13 0.02 Flexural tension
100-400 No 800 20 0.03 Flexural tension
N50-100 Yes Center 1 0.50 Failure of concrete around stud
N100-100 Yes Center 1 0.10 Failure of concrete around stud
6 S.-W. Yoo, J.F. Choo / Engineering Structures 118 (2016) 1–15
(a) Top face of UHPC slab
(b) Side face of UHPC slab
Fig. 10. Crack pattern of member 100–400.
good agreement with the results of previous studies on the mate-
rial behavior of UHPC [12–14]. It appears also that the structural
performance degrades with larger stud spacing because of the
bond failure of the concrete around the stud before UHPC could
develop its material capacity.
For the test members with slab thickness of 50 mm, compres-
sive strain developed at the bottom of the slab at early loading
due to the position of the neutral axis in the web of the steel girder.
With the increase of the load, debonding of the shear connection at
mid-span occurred and the members started to show non-
composite behavior resulting in the development of tensile strain
at the bottom of the slab. Besides, the test members with slab
thickness of 100 mm developed tensile strain at the bottom of
the concrete slab since early loading and until failure due to the
position of the neutral axis at the level of the concrete slab.
Finally, the members using normal concrete failed before con-
crete could develop its material performance in non-composite
behavior due to the absence of steel fiber. At failure, the compres-
sive stress acting in these members changed into tensile stress.
3.4. Load–steel girder strain curves
Figs. 18 and 19 plot the load–strain curves of the steel girder at
the top of the web and at the bottom flange, respectively. The yield
strain of the steel girder remains below 0.002, which is in good
agreement with the tensile test results of the material [12–14].
Here also, the structural performance degrades with larger stud
spacing because of the bond failure of the concrete around the stud
before UHPC could develop its material capacity.
(a) Top face of UHPC slab
(b) Side face of UHPC slab
Fig. 11. Crack pattern of member 100–100.
In Fig. 18, unlike the other test members, member 100-50
underwent failure without yielding. This peculiar behavior seems
to be due to the large thickness and excessively dense arrangement
of the studs, which made the compressive failure of the concrete
slab occur prior to the debonding of the concrete around the stud
or the yielding of the steel girder. Similar behavior is observed in
Fig. 19 for test member 100-400. In this case, the reason can be
attributed to the large thickness and the excessively scattered
arrangement of the studs, which made the concrete slab fail
through flexure-shear prior to the bond failure of the concrete
around the stud.
For the members using normal concrete, yielding of steel could
not be observed in the girder because the concrete slab failed first.
3.5. Evaluation of shear connector by horizontal shear force
Eurocode-4 recommends the smallest value between Eqs. (1)
and (2) be the design shear resistance of the headed shear
connector.
p d2 =4
0:8f u
PRD ¼ ð1Þ
cv
2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0:29ad f ck Ecm
PRD ¼ ð2Þ
cv
where cv = partial factor (1.25); d = stud shank diameter; fu = speci-
fied ultimate tensile strength of the material of the stud; fck =
 S.-W. Yoo, J.F. Choo / Engineering Structures 118 (2016) 1–15 7

(a) Top face of UHPC slab (a) Top face of normal concrete slab

(b) Side face of normal concrete slab

(b) Side face of UHPC slab
Fig. 13. Crack pattern of member N50-100.
Fig. 12. Crack pattern of member 50–50.

Fig. 20 plots the load–horizontal shear force curves of the

compressive strength of concrete; Ecm = elastic modulus of
concrete; and,
 
hsc
a ¼ 0:2 þ1 for 3 6 hsc 6 4
d
h of the graphs also exhibit clear difference and indicate the signifi-
a ¼ 1 for sc > 4
d
where hsc = overall nominal height of the stud.
Eq. (5) expresses the nominal shear resistance of one stud shear
connector embedded in the concrete deck recommended in
AASHTO-LRFD [17].
qffiffiffiffiffiffiffiffiffi
0
Q r ¼ /sc Q n ¼ /sc 0:5Asc f c Ec 6 /sc F u Asc
where /sc = resistance factor; Asc = cross-sectional area of stud;
Ec = modulus of elasticity of concrete; and, Fu = minimum specified
tensile strength of one stud.
Table 7 compares the horizontal shear force acting on the con-
crete slab computed directly using the strains measured at the
upper and lower chords of the concrete slab and the static strength
of the stud calculated using the formulae of Eurocode-4 and
AASHTO-LRFD. The horizontal shear strength obtained experimen-
tally is plotted in Fig. 20. Here, the values of the static strength of
the stud calculated using Eqs. (1)–(5) are compared directly to the
experimental results by discarding the safety factor. In the present
test, this static strength represents the physical resistance of the
stud itself owing to the outstanding performance of UHPC.
specimens. As a matter of fact, the horizontal shear force becomes
larger with thicker slab and denser arrangement of the studs. The
members using normal concrete without steel fiber reinforcement
show smaller horizontal shear force than the members using UHPC
ð3Þ except the members adopting stud spacing of 400 mm. The shapes
ð4Þ cant influence of the fiber reinforcement on the horizontal shear
force.
The comparison of the ultimate horizontal shear force and the
static strength of the studs predicted by Eqs. (1) and (2) reveals
that the average ratio is 3.08 for Eurocode-4 and becomes 2.47
for AASHTO-LRFD. The comparison of the experimental and pre-
dicted results according to the stud spacing shows that the ratio
ð5Þ of the experimental results to the design predictions increases with
thicker slab and larger stud spacing. This can be explained by the
reduction of the horizontal shear force caused by the non-
composite behavior resulting from larger stud spacing. The reason
for the larger experimental results obtained for thicker slabs can be
found in the fact that the current design formulae evaluate the
resistance of the stud itself as the static strength of the stud in
the case where UHPC develops outstanding performance. In other
words, for high performance concretes like UHPC, the current
design formulae cannot reflect the thickness of the slab. In the
future, new design formulae should be derived based upon exper-
imental data for high performance concretes like UHPC.
However, the regulations of Eurocode-4 and AASHTO-LRFD
used for the comparison in this study cannot consider the scale
effect produced by cyclic loading and scaled specimens. Therefore,
8 S.-W. Yoo, J.F. Choo / Engineering Structures 118 (2016) 1–15

(a) Top face of normal concrete slab

(a) Slab thickness = 50 mm

(b) Side face of normal concrete slab

Fig. 14. Crack pattern of member N100-100.

future study should be additionally conducted on full scale

specimens.
(b) Slab thickness = 100 mm
4. Material modeling for analysis Fig. 15. Load–displacement curves of test members.

4.1. Compressive behavior model of UHPC

Table 6
Recall that steel fibers were admixed with a unique length of Crack, yield, ultimate, failure load and characteristic slip measured in loading test.

13 mm but two different diameters of 19.5 mm and 16.3 mm in Test Crack Yield Ultimate Failure du duk
proportion of 2:1 for a volume fraction of 2%. Compression and member load (kN) load load (kN) load (kN) (mm) (mm)
tension tests were conducted on the steel fiber reinforcing the (kN)

specimens from which results are exploited to establish the mate- 50-50 – 428.1 573.2 554.2 11.72 10.54
rial model. The density of the steel fiber is 7500 kg/m3 and the 50-100 – 431.0 550.9 522.9 11.70 10.53
50-200 413.7 408.7 450.9 425.3 10.87 9.78
yield strength is 2500 MPa.
50-400 145.9 278.9 281.4 277.4 16.84 15.16
The compressive strength and elastic modulus of UHPC are 100-50 730.6 567.2 807.4 806.4 12.28 11.05
determined from the compressive stress–strain curves based on 100-100 642.4 558.7 755.8 752.4 11.70 10.53
the measured load–displacement curves. In view of the results of 100-200 476.2 535.9 583.2 563.2 10.77 9.69
the compressive strength test in Table 4 for the mix proportions 100-400 307.5 – 435.4 425.6 10.45 9.41
N50-100 – – 234.8 191.2 5.53 4.98
of Table 2, the average compressive strength is 183 MPa and the N100-100 – – 423.5 411.5 7.19 6.47
elastic modulus is 45,000 MPa. The specimens behaved linearly
until their sudden failure at compressive strength at which the
ultimate strain (eu) reached approximately 0.00406. This observa-
tion is in conformity with the well-known linear behavior before 4.2. Tensile behavior model of UHPC
the ultimate strain of the stress–strain curve of steel fiber rein-
forced UHPC reported in the literature. Accordingly, the compres- The tension softening curve of concrete can be obtained by
sive stress–strain relationship is modeled by means of the linear direct tensile test or by 3-point loading test on notched beams.
curve plotted in Fig. 21. The direct tensile test necessitates special equipment and presents
 S.-W. Yoo, J.F. Choo / Engineering Structures 118 (2016) 1–15
(a) Slab thickness = 50 mm
(b) Slab thickness = 100 mm
Fig. 16. Load–concrete strain at top of UHPC slab.
difficulties in maintaining stable loading conditions during the test.
In addition, the experiments conducted by our research team
showed that less than 10% of data could be acquired by the direct
tensile test. Besides, the 3-point bending test on notched beams, as
a standard testing method proposed by RILEM [18] and JCI [19], is
known to be particularly adapted for steel fiber reinforced concrete
and is applied in this study to describe the tension softening
behavior.
The tension softening model applied here is obtained through
analysis based upon the fictitious crack model of Hillerborg [20]
shown in Fig. 22. The tension softening curve is represented by
means of the multi-linear curve shown in Fig. 23. The cohesive
stress r(a, x) in the fictitious crack model of Fig. 22 is expressed
as a multi-linear function of the crack mouth opening displace-
ment (CMOD, d(a, x)). At the intersection (dk) of the kth and
(k + 1)th lines, the slopes of the CMOD and tension softening curve
are determined by optimizing the load obtained from the experi-
mental load and crack equation. At each analysis step, the consti-
tutive equations are formulated for all the CMODs so as to
compute the optimal tension softening curve.
The relationship between the sectional force and the deforma-
tion of a member subjected to bending is computed based upon
9
(a) Slab thickness = 50 mm
(b) Slab thickness = 100 mm
Fig. 17. Load–concrete strain at bottom of UHPC slab.
the relation between the curvature of the section and the CMOD.
This corresponds to the concept of converting the actual stress-
CMOD relationship into the stress–strain of the cracked section.
The Association Française du Génie Civil (AFGC) [21] suggests to
compute the tensile stress-CMOD relation based on the inverse
analysis of the load–CMOD curves measured during three-point
bend test conducted on notched specimens. Accordingly, three-
point bend test on 3 notched prismatic specimens fabricated with
dimensions of 100 mm  100 mm  400 mm was performed to
evaluate the flexural tensile behavior of UHPC (Fig. 24). The notch
was arranged at mid-span with a depth of 10 mm. The CMOD at
the notch was measured at each loading step using clip gages so
as to draw out the load–CMOD curves shown in Fig. 25.
Fig. 26 presents the numerical analysis model adopted for the
inverse analysis. The model of the flexure-tension specimen
involves 1076 triangular elements disposed symmetrically with
reference to mid-span. Fig. 27 plots the tension-softening curves
obtained by inverse analysis using the load–CMOD curves mea-
sured on the flexure-tension specimens.
Table 8 arranges the average values of the tensile strength and
strain resulting from the inverse analysis. In Table 8, ftj is the ten-
sile strength, ee is the elastic strain corresponding to the strain at
10 S.-W. Yoo, J.F. Choo / Engineering Structures 118 (2016) 1–15

(a) Slab thickness = 50 mm

(a) Slab thickness = 50 mm

(b) Slab thickness = 100 mm

Fig. 18. Load–steel girder strain at top of girder web. (b) Slab thickness = 100 mm
Fig. 19. Load–steel girder strain at bottom of flange.

initiation of cracking, e0.3 is the strain for a CMOD of 0.3 mm, e1% is
the strain when the CMOD reaches 1% of the height of the speci-
Table 7
men, elim is the limit strain corresponding to zero tensile stress, Comparison of experimental ultimate horizontal shear force and predicted resistance
and rbt and r1% are respectively the stresses corresponding to e0.3 load without safety factor.
and e1%. The average tensile strength is found to be 11.7 MPa,
Test (a): Horizontal (b): Stud (a)/(b) (c): Stud (a)/(c)
which is significantly higher than that observed in normal member shear force force by Eq. force by Eq.
concrete. from test (kN) (1) (kN) (2) (kN)
On the other hand, the relation between the CMOD (wi) and the 50-50 1206 799 1.51 998 1.21
fictitious plane strain (ewi) in the flexural crack section shown in 50-100 997 399 2.50 499 2.00
Fig. 28 can be expressed as follows. 50-200 564 200 2.83 250 2.26
50-400 347 100 3.48 125 2.78
wi ¼ lw ewi ¼ ðbhÞewi ð6Þ 100-50 1686 799 2.11 998 1.69
100-100 1125 399 2.82 499 2.26
where lw = factor depending on the sectional dimensions and 100-200 824 200 4.13 250 3.30
expressed in term of the height (h) of the beam; and, b is a constant. 100-400 528 100 5.29 125 4.23
Average – – 3.08 – 2.47
The relation between the tensile strain and the CMOD is as
N50-100 702 399 1.76 499 1.41
follows. N100-100 745 399 1.88 499 1.49
wi
ecr ¼ eu þ ð7Þ
bh
f tj w
Eq. (7) reflects the tensile stress–strain relationship proposed by
e¼ þ ð8Þ
Eij lc
AFGC [21] for steel fiber reinforced UHPC and the characteristic
values must be computed. Finally, the tensile stress–strain can be where ftj = tensile strength; Eij = elastic modulus; w = CMOD; and,
expressed as follows. lc = characteristic length of concrete. Here, lc = ⅔h for rectangular
 S.-W. Yoo, J.F. Choo / Engineering Structures 118 (2016) 1–15 11

Fig. 21. Compressive stress–strain model of steel fiber reinforced UHPC.

(a) Slab thickness = 50 mm

Fig. 22. Fictitious crack model [20].

(b) Slab thickness = 100 mm

Fig. 20. Load–horizontal shear force curves.

of T-shaped section with h = height of the section. This characteris-

Fig. 23. Multi-linear tension-softening curve.
tic length is conceptually similar to bh in Eq. (6).
Fig. 29 plots the resulting tensile stress–strain model computed
from the tensile stress-CMOD curve of UHPC.

5. Analysis results and discussion

5.1. Sectional analysis method

The sectional analysis method is applied to analyze the load–

deflection or moment–curvature of the member. The section is
modeled as a multi-layered cross-section as shown in Fig. 30.
The concept of the sectional analysis method was applied in the
studies of Thomas and Ramaswamy [23] and Yuguang et al. [24] for
the evaluation of the structural performance of steel fiber
reinforced concrete. In addition, Kooiman et al. [25] proposed the
sectional analysis method applying multi-layered section to ease
practically and reliably the evaluation of the post-cracking behav-
ior of steel fiber reinforced concrete. For the analysis purpose,
the member cross-section is subdivided into multiple layers and,
the distribution of the strain along the cross-section is assumed Fig. 24. Three-point bend test on notched UHPC specimen.
12 S.-W. Yoo, J.F. Choo / Engineering Structures 118 (2016) 1–15

Fig. 25. Load–CMOD curves measured during 3-point bend test on notched UHPC
Fig. 28. Stress–strain distribution in flexural crack section [22].
specimens.

Fig. 26. Numerical analysis model of flexure-tension specimen for inverse analysis.

Fig. 29. Tensile stress–strain model of steel fiber reinforced UHPC.

to be linear as shown in Fig. 30. The deformation at the top and

bottom of the section can be determined using the curvature and
the depth of the neutral axis. Since the curvature expresses the
change in the slope per unit length of the member, it is assumed
to be equal to the slope of the strain in the cross-section. The
curvature in the cross-section is increased by a constant increment
at each analysis step. Moreover, the position of the neutral axis is
assumed for each curvature to compute the strain distribution in
the cross-section. The strains at the top and bottom of the beam
section are expressed as follows by means of the strain compatibil-
ity conditions.
etop ¼ c /
ð9Þ
Fig. 27. Tension-softening curves obtained by inverse analysis. ebottom ¼ ðh  cÞ/

Table 8
Tensile data obtained by inverse analysis.

Specimen ftj (MPa) ee e0.3 e1% elim rbt (MPa) r1% (MPa)
1 11.7 0.00026 0.00206 0.00626 0.02625 11.7 11.3
2 14.1 0.00031 0.00211 0.00631 0.02625 14.1 13.2
3 9.1 0.00020 0.00200 0.00620 0.02625 9.1 8.1
Average 11.7 0.00026 0.00206 0.00626 0.02625 11.7 10.9
 S.-W. Yoo, J.F. Choo / Engineering Structures 118 (2016) 1–15 13

Fig. 30. Multi-layered section and distributions of stress and strain.

where / = curvature; and, h = height of section. attributed to the error in the material models of concrete and steel.
After having assumed the strain distribution, the stress in each Nevertheless, the cases where the analysis gives larger values than
layer is computed based upon the stress–strain relations of the the experimental results indicate that these members did exhibit
steel fiber reinforced concrete and steel reinforcement established
in the material modeling. Then, the section forces can be calculated
at each layer considering that the resultants of the sectional forces
in all the layers are in equilibrium.

Cc þ Ct þ T s ¼ 0
Z Z
ð10Þ
rc dAc þ rs dAs ¼ 0
Ac As

where Cc and Cs are respectively the compressive force and tensile

force sustained by concrete; and, Ts is the tensile force sustained
by steel.
The sectional force, that is the moment, can be calculated based
on the stress distribution in the cross-section satisfying the
equilibrium.
Z Z
M¼ rc ydAc þ rs ydAs ð11Þ
Ac As

where rc is the stress of concrete; and, rs is the stress of steel.

The moment M obtained by Eq. (11) and the curvature / can be
converted into the load P acting on the member of length L and the (a) Slab thickness = 50 mm
deflection d at mid-span follows.

4M
P¼
L   ð12Þ
1  cos h 1 1
d¼ ; where h ¼ sin L/
/ 2

5.2. Load–deflection relation

Fig. 31 plots the load–deflection curves of the members

obtained by sectional analysis together with the experimental
curves for comparison purpose. As stated in Section 3.2 for the
experimental results, the structural performance is seen to
enhance with thicker slab thickness and with shorter stud spacing.
It can be predicted that the specimens with stud spacing shorter
than 100 mm will exhibit perfect composite behavior. Taking into
account that perfect composite behavior of steel and concrete was
assumed in the analysis, the experimental and analytic results
exhibit good agreement and indicate similar behavior of the com-
posite beam. However, in Fig. 31, it appears that the analytical
(b) Slab thickness = 100 mm
results, which should be conservative, are slightly smaller than
the experimental results corresponding to the members having Fig. 31. Comparison of analytic and experimental load–displacement curves of
stud spacing shorter than 100 mm. This slight discrepancy can be composite beams.
14 S.-W. Yoo, J.F. Choo / Engineering Structures 118 (2016) 1–15

non-composite behavior after the bond failure of the concrete

around the studs. These factors explain the position of the analytic
curve in the middle of the experimental curves in Fig. 31.

5.3. Load–concrete compressive strain relation

Fig. 32 compares the load–concrete compressive strain curves

of the members obtained by sectional analysis with the experi-
mental curves.
The load–concrete compressive strain curves show patterns
similar to the load–deflection relation observed in the previous
section. However, there is some difference between the analytical
and experimental results in the increase of the concrete strain
according to the thickness of the slab. This error can be explained
by the utilization of the average values of the concrete batches in
the material model.

5.4. Load–steel tensile strain relation

(a) Slab thickness = 50 mm
Fig. 33 compares the load–steel tensile strain curves of the
members obtained by sectional analysis with the experimental
curves.

(b) Slab thickness = 100 mm

Fig. 33. Comparison of analytic and experimental load–steel tensile strain curves.

(a) Slab thickness = 50 mm

Here also, results similar to the load–displacement relation can
be observed. Especially, since steel is less prone to variation than
concrete, steel is seen to yield at practically identical strain except
for the stud spacing of 400 mm. However, in view of the analytic
results, the actual yield strength of steel is slightly superior to
the 396 MPa notified by the manufacturer. This difference can be
attributed to the non-consideration of the post-yielding strain
hardening of steel in the material model of steel.
The comparison shows that the analytic results are in relatively
good agreement with the experimental moment–curvature results
for the composite beam with steel fiber reinforced UHPC. This indi-
cates that the tension-softening curve computed from the material
tests reflects reasonably the actual behavior of the composite beam
with steel fiber reinforced UHPC. Consequently, the material model
of the steel fiber reinforced UHPC considering the tension-softening
characteristics and the flexural behavior analysis proposed in this
study are sufficiently accurate to predict reasonably the flexural
resistance of the composite beam with steel fiber reinforced UHPC.

6. Conclusions
(b) Slab thickness = 100 mm

Fig. 32. Comparison of analytic and experimental load–concrete compressive strain This paper evaluated experimentally the behavioral characteris-
curves. tics of a new type of composite beam using inverted-T steel girder
 S.-W. Yoo, J.F. Choo / Engineering Structures 118 (2016) 1–15
and steel fiber reinforced UHPC slab through test conducted on 8
test members presenting varying slab thickness and stud spacing,
and two supplemental test members using 50-MPa normal
concrete slabs. In addition, an analysis method considering the
tension-softening behavior of steel fiber reinforced UHPC in the
material model was proposed to predict the flexural behavior of
the composite beam. The results can be summarized as follows.
(1) The test members without steel fiber reinforcement and
using 50-MPa normal concrete experienced numerous large
cracks at the top face of the slab and directed along the axis
of the steel girder and failed suddenly. Such cracking and
failure patterns emphasized the critical role of the steel fiber
reinforcement and stressed the absolute necessity to arrange
reinforcing bars in the slab when steel fiber is not used.
(2) For the test members using UHPC slab, large stud spacing
with thin slab favored the occurrence of axial cracks propa-
gating along the steel girder at the top face of the slab. The
occurrence of these axial cracks could be attributed to the
tensile stress generated by the redistribution of the load
concentrated in a small zone around the stud to the wider
neighboring concrete.
(3) The test members using UHPC developed characteristic
relative slip ranging between 9.41 and 15.16 mm, which
indicated sufficient ductile behavior since these values far
exceeded the ductile behavior limit of 6 mm stipulated in
Eurocode-4. Besides, the test members using normal con-
crete did not or did merely satisfy the limit of Eurocode-4
with characteristic relative slip ranging between 4.98 and
6.47 mm. This once again verified the reinforcing role of
the steel fiber.
(4) For high performance concretes like UHPC, the current
design formulae provided by Eurocode-4 and AASHTO-
LRFD cannot reflect the thickness of the slab. This stressed
the need to derive new design formulae based upon experi-
mental data for high performance concretes like UHPC.
(5) A moment–curvature analysis method was proposed by
applying material model considering the tension-softening
behavior of steel fiber reinforced UHPC. The tensile strength
of steel fiber reinforced UHPC obtained by inverse analysis
was 11.7 MPa, which is significantly larger than that of nor-
mal strength concrete.
(6) The comparison of the analytic and experimental results
showed good agreement for the composite beam with steel
fiber reinforced UHPC slab. This indicated that the tension-
softening curve computed from the material tests reflects
reasonably the actual behavior of the composite beam with
steel fiber reinforced UHPC.
Consequently, the material model of the steel fiber reinforced
UHPC considering the tension-softening characteristics and the
flexural behavior analysis proposed in this study are sufficiently
accurate to predict reasonably the flexural resistance of the com-
posite beam with steel fiber reinforced UHPC.
Acknowledgements
This research was supported by a grant (13SCIPA02) from Smart
Civil Infrastructure Research Program funded by Ministry of Land,
15
Infrastructure and Transport (MOLIT) of Korea government and
Korea Agency for Infrastructure Technology Advancement (KAIA).
References
[1] Russell HG, Graybeal BA. Ultra-high performance concrete: a state-of-the-art
report for the bridge community. US Department of Transportation. Federal
Highway Administration. Publication No. FHWA-HRT-13-060; June 2013.
[2] John H, George S. The implementation of full depth UHPC waffle bridge deck
panels. In: Federal highway administration highways for LIFE technology
partnerships program; 2010.
[3] Naaman AE, Chandrangsu K. Innovative bridge deck system using high-
performance fiber-reinforced cement composites. ACI Struct J 2004;101
(1):57–64.
[4] Yang IH, Joh C, Kim BS. Flexural strength of large-scale ultra high performance
concrete prestressed T-beams. Can J Civ Eng 2011;38:1185–95.
[5] Graybeal BA. Flexural behavior of an ultra high-performance concrete I-girder.
J Bridge Eng 2008;13(6):602–10.
[6] Yoo DY, Banthia N, Yoon YS. Flexural behavior of ultra-high-performance fiber-
reinforced concrete beams reinforced with GFRP and steel rebars. Eng Struct J
2016;111:246–62.
[7] European Commission. EUR 25321 – Prefabricated enduring composite beams
based on innovative shear transmission (preco-beam). RFSR-CT-2006-00030,
Final Report; 2009.
[8] Ghasemi S, Zohrevand, Mirmiran A, Xiao Y, Mackie K. A super lightweight
UHPC-HSS deck panel for movable bridges. Eng Struct J 2016;113:186–93.
[9] Lee KC, Joh C, Choi ES, Kim JS. Stud and puzzle-strip shear connector for
composite beam of UHPC deck and inverted-T steel girder. J Korea Concr Inst
2014;26(2):151–7.
[10] Feldman M, Hechler O, Hegger J, Rauscher S. Fatigue behavior of shear
connectors in high performance concrete. Int Conf Compos Constr Steel Concr
VI 2008:78–91.
[11] Yang IH, Joh C, Kim BS. Structural behavior of ultra high performance concrete
beams subjected to bending. Eng Struct J 2010;32:3478–87.
[12] Korea Concrete Institute. Design recommendations for ultra-high performance
concrete K-UHPC. KCI-M-12-003, Korea; 2012.
[13] Yang IH, Joh C, Lee JW, Kim BS. Torsional behavior of ultra-high performance
concrete squared beams. Eng Struct J 2013;56:372–83.
[14] Park JS, Kim YJ, Cho JR, Jeon SJ. Characteristics of strength development of
ultra-high performance concrete according to curing condition. J Korea Concr
Inst 2013;25(3):295–304.
[15] European Committee for Standardization, CEN 1994-4-4. Eurocode-4: Design
of composite steel and concrete structures, part 1-1: general rules and rules for
buildings. CEN; 2004.
[16] European Committee for Standardization, CEN 1994-2. Eurocode-4: Design of
composite steel and concrete structures, part 2: general rules and rules for
bridges. CEN; 2005.
[17] American Association of State Highway and Transportation Officials (AASHTO).
AASHTO LRFD bridge design specifications, 4th ed. Washington DC; 2007.
[18] RILEM TC 162-TDF. Recommendations of RILEM TC 162-TDF: test and design
methods for steel fibre reinforced concrete: bending test. Mater Struct
2002;35(253):579–82.
[19] JCI-S-002-2003. Method of test for load–displacement curve of fiber reinforced
concrete by use of notched beam. Jpn Concr Inst.
[20] Hillerborg A, Modeer M, Petersson PE. Analysis of crack formation and crack
growth in concrete by means of fracture mechanics and finite elements. Cem
Concr Res 1976;6(6):773–82.
[21] Association Française du Génie Civil. Bétons fibrés à ultra-hautes
performances. Association Française du Génie Civil, SETRA, France; 2002.
[22] Spasojević A. Structural implications of ultra-high performance fibre-
reinforced concrete in bridge design. Ph.D. thesis No. 4051, Ecole
Polytechnique Fédérale de Lausanne, Lausanne, Switzerland; 2008.
[23] Thomas J, Ramaswamy A. Crack width in partially prestressed T-beams having
steel fibers. ACI Struct J 2006;103(4):568–76.
[24] Yuguang Y, Walraven JC, Uiji JD. Study on bending behavior of an UHPC
overlay on a steel orthotropic deck. In: Proc. 2nd international symposium on
UHPC, Germany. p. 639–46.
[25] Kooiman AG, van der Veen C, Walraven JC. Modelling the post-cracking
behaviour of steel fibre reinforced concrete for structural design purposes.
HERON 2000;45(4):275–307.