International Journal of Adhesion & Adhesives 54 (2014) 131–142

Contents lists available at ScienceDirect

International Journal of Adhesion & Adhesives

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

Steel–concrete bonding connection: An experimental study

and non-linear finite element analysis
C. Meaud, B. Jurkiewiez n, E. Ferrier
Université de Lyon, Université Lyon 1, LGCIE, 82 Boulevard Niels Bohr, F-69622 Villeurbanne cedex, France

art ic l e i nf o a b s t r a c t

Article history: We present characterisation interface tests results obtained by push-out tests adapted to a Steel–
Accepted 19 May 2014 concrete bonded connection. The sizes of the concrete slabs and areas of the bonding zone are the
Available online 29 May 2014 primary studied parameters. It is shown that the behaviour observed during the push-out tests strongly
Keywords: depends on these parameters. Additionally, the measurements are compared to numerical results
Bonding obtained from 3D FE models accounting for the non-linear behaviour of the constitutive materials.
Steel–concrete composite element In particular, the numerical results confirm that the shear stress in the concrete slabs along the bonding
Push-out tests joint and near the interface is not uniform and varies with the level of the applied load. Moreover, the
Non-linear FE model shear stress may be accompanied by a strong compression stress perpendicular to the interface, which
is mainly due to the possible friction between the bottom face of the concrete slabs and the steel rigid
plate of the hydraulic press. This stress induces a favourable double state of stress in the bonding joint
and a high failure load. Therefore, these tests and numerical results show that the state of stress in a
push-out test is complex, varies with the load and seems to be most likely not representative of the state
of stress in a flexural member. Consequently, the push-out test can be used as a way of characterising a
Steel–concrete bonding connection that may be useful in comparing different adhesives or bonding
processes; however, this test should be used very carefully for beam designs.
& 2014 Elsevier Ltd. All rights reserved.

1. Introduction Steel–concrete composite beam with mechanical connectors, such

as studs. The behaviour of the flexural member exhibits sufficient
In Steel–concrete composite construction, the concrete slab and ductility and no shear failure if the bonding joint is properly
steel girders are typically connected by shear studs welded on the designed [3–7].
upper flange of the beams and hidden in fresh concrete. In the case To be used in Steel–concrete structures, the bonding connec-
of bridges for instance, this connection process requires specific tion should be first characterised, but unfortunately, no standard is
welding equipment and generally implies that the slab is cast available. In the case of repairing concrete structures with external
in situ with a heavy and expensive self-launching formwork. bonded carbon fibres reinforced polymers (CFRP), the character-
In addition, once the concrete is hardened, its shrinkage is isation may be performed by a double lap shear test [8–9].
restrained because the deck is rigidly fixed to the steel girder. Moreover, Berthet et al. [10] proposed a push-out test with a steel
This type of connection may induce uncontrolled and excessive plate measuring 5 mm in thickness. However, in both tests, the
cracking and reduce the building's quality and durability [1]. thicknesses of the bonded plates are too low and far from the size
For the past several years, a bonded connection between the of the girders used, for instance, in a bridge. As this dimension has
slab and steel girders has been examined as an alternative to shear a significant influence on the behaviour and on the state of stress,
studs [2]. This new connection device allows precasting the the push-out test principle has been chosen for this study and the
concrete deck in a factory and reducing the risk of cracking as sizes of the specimen are inspired from those of the shear studs
the shrinkage of the concrete slab is unrestricted during the initial given in the Eurocode guidelines [11–12] using a HEB profile.
weeks after casting. From a mechanical point of view, recent The only push-out tests on bonded connection between steel
studies indicated that bonding allows a similar behaviour as a and concrete that can be found in literature are those of Si Larbi
et al. [5]. These tests were performed with high performance
n
concrete (fc28 ¼67 MPa), and the dimensions of the concrete slabs
Corresponding author. Tel.: þ 33 472692130; fax: þ33 478946906.
E-mail addresses: [email protected] (C. Meaud),
and bonding area were 300  300  100 mm3 and 100  100 mm2,
[email protected] (B. Jurkiewiez), respectively. The observed failure was cohesive in the concrete
[email protected] (E. Ferrier). close to the interface, and the average ultimate shear stress was

http://dx.doi.org/10.1016/j.ijadhadh.2014.05.010
0143-7496/& 2014 Elsevier Ltd. All rights reserved.
132 C. Meaud et al. / International Journal of Adhesion & Adhesives 54 (2014) 131–142

approximately 6 MPa. This value is similar to the tensile strength was approximately 2.0 70.6‰ (i.e 0.00270.0006). The actual
of the concrete. yield strength and failure strength were approximately 297 7
This paper aims to complete these experimental data in the 14 MPa and 4417 9 MPa, respectively.
case of ordinary C25/30 concrete type. In particular, we intend to
assess if the failure mode and the average ultimate shear stress 2.1.2. Adhesive
depend or not on the sizes of the concrete slabs or on the bonding The adhesive used for all push-out specimens was epoxy mixed
area as it had been shown in previous published paper for with silica (SIKADUR 30) because it suited the construction's
different materials and geometries (see [13] for instance). In the conditions along the span-life (hygrometry and temperature).
second part of the paper, a non-linear numerical FE model is Tensile tests were performed on 5 coupons according to the NFT
proposed in order to determine the state of stress and to localise 17-301 standard. The results showed that the resin exhibited high
the origin of the failure. The results are finally discussed. strength but was brittle. The failure stress, the secant elastic
modulus and the ultimate strain were approximately 25.4 7
2. Push-out tests 7.4 MPa, 11,76071374 MPa and 2.2 70.7‰ (i.e. 0.0022 70.0007),
respectively.
The experimental study consisted of 6 different tests series,
denoted PO-1-100, PO-1-300, PO2-100, PO2-300, PO3-100 and 2.1.3. Concrete
PO3-300 with 3 identical specimens for each series. The concrete used for all push-out specimens was a C25/30
concrete with a maximal size aggregate of approximately 8 mm.
2.1. Materials properties In one attempt in our laboratory, the slabs of the series PO1 were
cast horizontally as the series PO2 and PO3.
2.1.1. Steel For each of the three concreting phases, 12Φ16  32 cylindrical
The steel profile used for all the push-out specimens was a HEB specimens (diameter ¼160 mm  height ¼320 mm) were cast with
100 S275. Tensile tests were conducted on samples cut from the the slabs to conduct compression and tensile tests 28 days after
web and flange of an IPE 220 S275. The linear elastic limit strain concreting and the day of the push-out tests. Compression and
tension tests were performed on the cylindrical samples according
Table 1 to the NF P18-406 and the NF P18-408 standards respectively. The
Concrete properties. results are summarised in Table 1. It should be noted that in spite
of an identical test processes for the three series PO1 to PO3, the
PO1 PO2 PO3
compression and tensile strength at 28 days varied between 34
28 days fc (MPa) 36.7 70.8 37.17 0.9 33.77 1.4 and 37 MPa and 2.2 and 2.6 MPa, respectively.
ft (MPa) 2.6 70.1 2.6 7 0.4 2.2 7 0.1

Push-out tests day Age (days) 82 123 120 2.2. Fabrication of specimens and test set-up
fc (MPa) 39.3 70.7 43.4 7 1.4 38.8 7 1.0
ft (MPa) 3.2 70.6 3.2 7 0.2 3.0 7 0.1
The push-out specimens differed from each other either in the
fc: compression strength. size of the concrete slab or the bonding area as shown in Fig. 1. The
ft: tensile strength. specimens denoted POx-100 and POx-300 exhibited concrete slabs of

Longitudinal view

PO1-300 / PO2-300 / PO3-300 PO1-100 / PO2-100 PO3-100

50

HEB 100
70
50

HEB flanges
were cut
Bonding
a

area
50 25
50

100 100
300

Concrete 100 100

300 slab

Section

PO1-300 / PO2-300 / PO3-300 PO1-100 / PO2-100 / PO3-100

50

PO1-300 / PO1-100 / PO3-100 a = 100 mm

PO2-300 / PO2-100 a = 130 mm

a

150

PO3-300 a = 180 mm
300

100 100 100

100 100 100

Fig. 1. Push-out specimens geometry (all dimensions in mm).

 C. Meaud et al. / International Journal of Adhesion & Adhesives 54 (2014) 131–142 133

100  100  150 mm3 and 100  300  300 mm3 respectively. The
specimens denoted PO1-x, PO2-x and PO3-x exhibited a bonding
length of 100 mm, 130 mm and 180 mm respectively. Concrete slabs
were not reinforced. Before bonding, the concrete slabs were sanded,
and the steel surfaces were treated by sandblasting, acetone cleaning
and a primer was applied. The bond line thickness was not measured
but was estimated to less than 1 mm.
The measurements conducted during the tests were the fol-
lowing: the applied load, and for some specimens, the vertical
displacement, the vertical strains in the concrete slab and in the
steel profile and the differential displacement between the con-
crete slab and the steel profile near the interface and at mid-height
of the bonding zone (this will be called slip throughout the rest of
the paper). Load cell, gauges and Linear Variable Differential
Transformer sensors (LVDT sensors) were used with numerical
recording. A general view of the test device and locations of the
strain gauges are detailed in Figs. 2 and 3, respectively.
Fig. 3. General view of the test device.

2.3. Tests results

In all cases at failure, one slab was delaminated from the steel
HEB and a thin and pitted layer of concrete remained bonded on
the steel surface as shown in Fig. 4. Thus, the failure mechanism
was a shear failure, brittle and cohesive in the concrete near the
interface. This failure mechanism was the same to that depicted by
Si Larbi et al. [5] in the case of high performance concrete. During
the test, after a settling phase of the specimen, the load and the
slip increased up to the brittle failure, but not proportionally as
shown in Fig. 5 in the case of PO3-300 series. This may be
explained by the non linear behaviour of the materials as the
failure approached and the beginning of the shear failure mechan-
ism. The linear secant stiffness divided by the bonding area
ðK e ¼ F=u SÞ was 422,000 MN/m3. In the case of a headed stud
Φ ¼ 19 mm in diameter (widely used in bridge construction for
instance) with an associated surface of 12.5 Φ2 (minimal area
required for the welding), the linear secant stiffness divided by the

Fig. 4. Failure mode.

50 50

50 50

36000 PO3-300-3
32000 PO3-300-1

28000
15 13 PO3-300-2
24000
Force [daN]

PO1-300
20000

16000

12000

8000
50 50 50

50

4000
PO1-100
15 0
0 10 20 30 40 50 60 70 80 90 100
Slip [ m]
13
Top steel gauge Fig. 5. Slip measurements (PO3-300 specimens).
30

30

bonding area is approximately 33,237 MN/m3. This result confirms

90

90
100

100

that the bonding connection technique is between 6 to 9 times

90

more stiff than headed studs. However, it must be noted that in

50
50

the current study, the slip measurements exhibited significant

15 13 variation, and an image correlation technique should provide more
PO3-300 Bottom steel gauge
accurate information.
The strains are shown in Fig. 6. It appears that the maximum
Fig. 2. Strain gauges location (in mm). value of the shortening in the concrete slab and steel girder did
134 C. Meaud et al. / International Journal of Adhesion & Adhesives 54 (2014) 131–142

18000 36000 Test-1-Concrete

Test-1
Test-2-Steel Bottom steel
16000 Test-2-Concrete 32000 Test-2-Top steel
Test-3-Concrete
14000 28000
Test-2
Test-3
12000 24000 Concrete
Steel
Force [daN]

Force [daN]
Test-2
10000 20000
Bottom steel
Test-1-Steel
8000 16000
Test-3-Top steel
6000 Test-1-Concrete 12000
Test-3
4000 8000 Concrete

2000 PO1-300 4000 PO3-300

0 0
-250 -200 -150 -100 -50 0 50 100 150 200 250 300

0
50

45
0
50
00
50
00
0

0
0
0
0
0
0
0

0
0
0
-5

10
15
20
25
30
35
40

50
55
60
-2
-2

-1
-1
Strain [ m/m]
Strain [ m/m]

Fig. 6. Concrete strain and steel strain in PO1-300 and PO3-300.

Table 2 3. Finite element analysis

Push-out tests results.
3.1. Modelling details
Push out Concrete slab Bonding area Average shear failure
specimens dimensions strength
(mm3) (mm2) (MPa) In this study, the finite element programme Cast3M was used
to simulate push-out tests. Because of the complex geometry of a
PO1-100 100  100  150 2  100  100 2.0 70.5
push-out specimen, the finite element method was preferred to
PO1-300 100  300  300 6.8 70.2
PO2-100 100  100  150 2  100  130 1.8 70.4
analytical approaches [14–16]. Two dimensional (plane state of
PO2-300 100  300  300 7.6 70.1 stress) and 3D-simulations were initially conducted in the elastic
PO3-100 100  100  150 2  50  100 2.6 70.7 domain and the results are compared [17]. The shear state of stress
PO3-300 100  300  300 2  100  180 9.1 70.9 close to the interface in the 2D-model was very different from that
of the 3-D model. Thus the 3 D – model was selected to perform
non linear analysis. In order to limit the memory space and
analysis time, the thin bonding joint was modelled by interface
finite elements that had already been successfully used in other
not exceed 250 μm/m and 600 μm/m, respectively, and no long- applications (see [18,19]).
itudinal plastic strain had been measured during the test. In the As a result of the double symmetries of the specimen, only a
PO1-100 and PO1-300 series, the strains in the concrete slab and quarter of the push-out arrangement was modelled using finite
steel profile remained similar until the beginning of failure. At that elements available in the Cast3M library. The concrete slab and
point, the values significantly differed because of the stress steel element were meshed with a 4-node tetrahedral element
redistribution in the concrete from the micro cracked zones to (TET4). The bonding joint was meshed with a 4-node adhesive
the undamaged zone along the bonding joint. In the case of PO3- element (JOI4) assuming that the thickness bond line was 1 mm.
300, the measurements showed that the longitudinal strains were The size of the elements was adjusted based on the maximal
not constant along the bonding joint but varied linearly with capabilities of the computer to ensure convergence, maximal
respect to the vertical position. accuracy and independence of the results from the mesh (Fig. 7).
The average ultimate shear stresses (hereafter denoted “shear Consequently, the size of the elements was small near the inter-
strengths”) are given in Table 2. In the case of the 100  face (approximately 2.5 mm) and increased to about 20 mm in the
100  150 mm3 slabs (PO1-100, PO2-100, PO3-100), the shear remaining meshed volumes.
strengths were in the same order of magnitude as the tensile The loading was uniformly applied on the upper face of the
strength of the concrete and thus did not vary significantly with steel element. In addition to the symmetry planes, the boundary
the bonding length and slightly decreased as the width of the conditions of the contact area between the metallic press support
bonding surface increased. In the case of the 100  300  300 mm3 and the lower face of the concrete slab were chosen according to
slabs (PO1-300, PO2-300, PO3-300), the shear strengths were two different assumptions as follows:
greater than the tensile strength of the concrete and slightly
increased with the length of the bonding zone. As a conclusion, – Assumption denoted SF: only the vertical displacement of each
it appears that the average ultimate shear stresses depended node in the lower surface of the concrete part was assumed to
roughly on the size of the concrete slab and on the bonding be prevented in neglecting any force friction;
surface area and seemed to increase with the bonding length. – Assumption denoted F: all displacements of each node of the
However, as the standard variation may be significant, it is lower surface of the concrete part were assumed to be
suggested to increase the number of identical specimens in a prevented in accounting for high friction.
series to 5 or 6 for further tests. In addition, to increase the quality
of the specimens, it is recommended to saw the upper face of the The bonded joint was assumed to be elastic, isotropic and
concrete slabs perpendicularly to the bonding surface to limit the linear. The Von Mises criterion with the cinematic hardening rule
undesired effects caused by possible parallelism defaults. was used to represent the steel beam.
The various observations and measurements presented above In the case of the concrete, the choice of the constitutive law is
are beneficial. However, they can be complemented by numerical more sensitive because the slab was subjected to shear and
simulations allowing the determination of the state of stress, compression, and the law should consider the non-linear and
particularly close to the steel–concrete interface. The numerical irreversible behaviour with cracking in compression, as well as
simulation is detailed in the following paragraph. in tension. Thus, the behaviour of the concrete was approached by
 C. Meaud et al. / International Journal of Adhesion & Adhesives 54 (2014) 131–142 135

means of a coupled damage-plasticity model as developed where the dot denotes the time derivative, E is the elastic stiffness
by Jason et al. with isotropic damage and a non-local approach tensor, m is the flow vector, k is the set of internal variables, h is
[20–24]. Plasticity and damage are coupled using the effective the plastic modulus and λ_ is the plastic multiplier (these functions
stress s0 defined as follows: and parameters are detailed in [20,22]). The plasticity is finally
s ¼ ð1  DÞ s0 ; ð1Þ described by the set of parameters ðr c ; r t ; p; Ah ; Bh ; C h ; γ ; α;
a; k0 Þpresented in Tables 3 and 4.
where s is the applied stress tensor, and D is the damage variable. The evolution of damage is defined according to the isotropic
The plasticity is governed independently by the classical set of model initially developed by Mazars in 1984 and modified by
equations in which the effective stress has been substituted by the Jason in order to take into account the plastic strain. The damage is
applied stress: controlled by an equivalent strain εeq defined as follows:
ε_ ¼ ε_ e þ ε_ p sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3
s0 ¼ E ε e εeq ¼ ∑ ð〈εei 〉 þ Þ2 ð3Þ
ε_ p ¼ λ_ mðs0 ; kÞ i¼1

k_ ¼ λ_ hðs0 ; kÞ ð2Þ 〈εei 〉 þ represents the ith positive principal value of the elastic strain
tensor. The damage D is defined by an evolution law which
7 distinguishes the response in tension and in compression:

6 if εeq Z εD0 then

Shear Stress D ¼ αt Dt þ αc Dc
5 Peeling stress εD0 ð1  At Þ At
Dt ¼ 1  
εeq exp½Bt ðεeq  εD0 Þ
Stress [MPa]

4 ð4Þ
3 εt 〈εe 〉
αt ¼ ∑ i 2i þ
i ¼ 1 εeq
3

else D ¼ 0
2

1
The parameters Dc and αc (compression damage) are determined
by expressions similar to those of Dt and αt in which the subscript t
0 is substituted by c. εD0 ¼ f t =E represents the initial thresh-old from
0 10000 20000 30000 40000 50000 60000 70000 which damage grows (ft: tensile strength).ðAt ; Ac ; Bt ; Bc Þ are four
Number of nodes constants calibrated for each concrete from either the material
Fig. 7. Patch test in the case of push-out specimen PO1-300. (The stresses are given tests results (compression and tension) or chosen according to the
in the concrete close to the end of the bonded joint for FE50% Fu). recommendations of Jason et al. [21].

Table 3
Parameters of the constitutive laws (Berthet's specimens [10]).

Adhesive : elastic and isotropic

E¼ 13,000 MPa ν ¼ 0.31

Steel : elastic plastic with hardening

E¼ 210,000 MPa ν ¼ 0.30 Yield strength fe ¼ 360 MPa Second tangent modulus H ¼400 MPa

Concrete : elastic plastic with isotropic damage (see Jason et al. [21] for more details)

Elastic phase Damage Plastic strain

E (MPa) ν Ac At Bc Bt εD0  10-4 rc  106 rt  106 γ p a k0 Ah  10-5 Bh  10  3 Ch  10  6 α

36,600 0.20 3.30 1.40 1800 11,000 1.3 130 13 0.99 0.3 1 0.1 7 3 3 0.5

Table 4
Parameters of the constitutive laws for the studied push-out specimens.

Adhesive : elastic and isotropic

E ¼13,000 MPa ν ¼ 0.31

Steel : elastic plastic with hardening

E ¼210,000 MPa ν ¼ 0.30 Yield strength fe ¼275 MPa Second tangent modulus H¼ 400 MPa

Concrete : elastic plastic with isotropic damage (See Jason et al. [21] for more details)

Elastic phase Damage Plastic strain

Push out specimens E (MPa) ν Ac At Bc Bt εD0  10-4 rc  106 rt  106 γ p a k0 Ah  10  5 Bh  10  3 Ch  10  6 α

PO1 37,300 0.2 1.27 0.90 1707 18,027 1 130 13 1 0 1 0.1 7 3 3 0.5
PO2 1.29 0.90 1707 18,664 1 150 13 1 0 1 0.1 7 3 3 0.5
PO3 1.28 0.90 1707 18,650 1 130 11 1 0 1 0.1 7 3 3 0.5
136 C. Meaud et al. / International Journal of Adhesion & Adhesives 54 (2014) 131–142

Longitudinal view Section

Steel plate

15
Bonding
area
300

300
Concrete
block
15

150 5
200 200

Fig. 8. Push-out specimens geometry of Berthet's series [10] (dimensions in mm).

18000
Gauge G6 FEM-F-Concrete
16000 FEM-F-Steel

14000
0 Gauge G7

-50 12000
FEM-SF

Force [daN]
-100 Gauge G8 Concrete
10000 Test-2
FEM-SF-Steel
-150 Steel
8000
Force [kN]

-200
6000
-250
Test-3-Steel
-300 4000 Test-2-Concrete
-350 Test-1-Concrete
2000
-400 Test-3-Concrete Test-1-Steel
0
-450 -250 -200 -150 -100 -50 0 50 100 150 200 250 300
-500 Strain [ m/m]
-1200 -1000 -800 -600 -400 -200 0
Strain [ m/m] Fig. 11. Concrete strain and steel strain in PO1-100. FEM-SF: Finite Element Method
with no friction assumption. FEM-F: Finite Element Method with friction
G6-Test G7-Test G8-Test G6-FEM G7-FEM G8-FEM
assumption.
Fig. 9. Strains in the steel plate—numerical results versus measurements (S-E-C).

18000
Test-2-Steel
16000 Test-2-Concrete
FEM-F-Steel
Test-3-Concrete
14000
z Test-3
12000 Steel
FEM-F
Force [daN]

10000 Concrete
Test-1-Steel
8000
150, 200 or 300 mm

Steel HEB
6000 Test-1-Concrete FEM-SF-Steel

4000

2000 FEM-SF-Concrete PO1-300

0
Interface -250 -200 -150 -100 -50 0 50 100 150 200 250 300
element
Strain [ m/m]

Fig. 12. Concrete strain and steel strain in PO1-300. FEM-SF: Finite Element
Method with no friction assumption. FEM-F: Finite Element Method with friction
x y assumption.

Fig. 10. FE mesh of the quarter push-out specimen.

3.2. Comparison between numerical results and measurements

Numerical simulations were conducted for each push-out 3.2.1. Interface characterisation tests of Berthet et al. [10]
series using a typical iterative incremental procedure. If the The proposed model was first applied to simulate some of
calculation stopped because of non convergence, simulations are the push-out tests performed by Berthet et al. [10]. The specimens
started again at the last point reached in the previous stage with a are formed by a steel plate between two concrete blocks bonded
refined division of the loading step. The calculation stopped as the one to each other by means of a thin layer of epoxy resin (Fig. 8).
convergence cannot be ensured anymore with an accuracy of The parameters of the constitutive law of each material were
1  10  4. The corresponding load was designated as the numerical determined from the data given in the paper and are presented in
failure load. the Table 3.
 C. Meaud et al. / International Journal of Adhesion & Adhesives 54 (2014) 131–142 137

The numerical results obtained with the SF-assumption were In the case of the 100  100  150 mm3 slabs, the measure-
very similar to those obtained with the F-assumption. The average ments of the failure load, the slips and the strains are under-
ultimate shear stress was about 4.7 MPa and corresponded to a estimated, regardless of the SF or F assumption. Therefore, this
cohesive shear failure in the concrete close to the interface: this is model may not correctly simulate push-out tests with undersized
very close to tests results as the ultimate shear failure varied concrete slabs, which can most likely be explained by possible
between 4.6 and 5.2 with respect to the surface treatment. minor geometric defaults that are not considered, such as the
The strains measured in the steel plate were also compared to parallelism default. These negative effects are sufficiently substan-
the numerical results as shown in the Fig. 9. Numerical results are tial to induce premature failure in the case of small slabs.
close to measurements and show that strains decrease along the From a numerical point of view only, it can be observed that the
interface. That indicates the progressive transfer of the forces from average ultimate shear failure was approximately twice as high in
the steel plate to the concrete blocks. the F-case as in the SF-case, regardless of the size of the concrete
slabs or bonding area (Table 5). Furthermore, in the case of the
constant bonding area, the shear stress slightly depends on the
3.2.2. Push-out tests size of the concrete slabs. Finally, the ultimate shear stress tends to
The proposed model was also applied to the push-out tests increase with the bonding length in the F-case and tends to
described in paragraph 2. The mesh is shown in the Fig. 10. The decrease in the SF-case. However, these observations are trends
material data were determined from characterisation tests pre- that should be carefully considered, as the mesh and the concrete
sented in paragraph 2.1 and are summarised in Table 4. parameters were not strictly identical for each case.
The comparison between the numerical results and the mea- Fig. 15 shows the variation of the slip along the interface with
surements is presented in Figs. 11–14 and in Table 5. It can be the average ultimate shear stress for the different studied cases.
observed that for a large part of the behaviour, the model The curves indicated a similar shape for the F- and SF-cases but
correctly predicts the slip and longitudinal strains regardless of with a notably different level of stress. Additionally, it can be
the boundary assumptions of SF or F, which has a slight influence observed that small slab thickness induces a slight increase of the
on the results at low force. To the contrary, the failure loads and linear stiffness.
final values of the slips and strains strongly depend on this The proposed model provides also the state of stress in any
assumption. point of the push-out specimen and at any level of loading. The
In the case of the 100  300  300 mm3 slabs, the numerical evolution of the distribution of the damage in the concrete slab
failure load (and the ultimate average shear stress), calculated can thus be obtained. These variables that cannot be measured
with the assumption that F is similar to the test failure load, allow for analysis of the behaviour or to corroborate the observa-
seemed to indicate that friction was important for these dimen- tions, such as the failure modes.
sions. Furthermore, the slips and strains were accurately predicted. For all numerical simulations, the results show that the strains
in the adhesive joint remain lower than the measured ultimate
strain (see paragraph 2.1). Likewise, the stress in the steel HEB
36000
Test-1-Concrete Test-1-Bottom steel Test-2-Top steel never reaches the yielding stress. Consequently, we focus the
32000 Test-2-Bottom steel following results only on the concrete slab and, particularly, in
FEM-F
28000 Bottom steel the zone near the interface (x ¼100 mm: see Fig. 10).
Test-2
24000 Concrete
Force [daN]

FEM-SF
20000 Top steel Table 5
FEM-F Average shear failure strength.
16000 Concrete
FEM-F-Top steel
12000 Push out specimens Tests FEM
Test-3-Top steel
8000 FEM-SF F SF
Concrete
MPa MPa MPa
4000 Test-3 PO3-300
Concrete FEM-SF-Bottom steel
0 PO1-100 2.0 7 0.5 7.6 3.7
-250 -200 -150 -100 -50 0 50 100 150 200 250 300 350 400 450 500 550 600 PO1-300 6.8 7 0.2 6.7 3.9
Strain [ m/m] PO2-100 1.8 7 0.4 8.2 3.5
PO2-300 7.6 7 0.1 7.3 3.7
Fig. 13. Concrete strain and steel strain in PO3-300. FEM-SF: Finite Element PO3-100 2.6 7 0.7 10.1 5.3
Method with no friction assumption. FEM-F: Finite Element Method with friction PO3-300 9.17 0.9 8.2 3.0
assumption.

36000 36000 PO3-300

32000 PO2-300 32000
28000 28000
24000 24000 FEM-F
Force [daN]

Force [daN]

20000 FEM-F 20000

16000 16000
12000 12000 FEM-SF
FEM-SF
8000 8000
4000 4000
0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Slip [ m]
Slip [ m]

Fig. 14. Slip measurements versus numerical results at mid-height.

138 C. Meaud et al. / International Journal of Adhesion & Adhesives 54 (2014) 131–142

10 PO3-100 10
9 9 MEF-SF
PO2-100
8 8
PO1-100
Shear stress [MPa]

Shear stress [MPa]

7 7
PO2-300
6 PO1-300 6 PO3-100
5 5
PO1-100
4 PO3-300 4 PO2-300
3 3
2 2 PO2-100 PO3-300
1 MEF-F 1
PO1-300
0 0
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
Slip [ m] Slip [ m]

Fig. 15. variation of the slip at mid-height with the average ultimate shear stress. (numerical results only).

10 10
5 5
0 0
-5 -5
-10 -10
XX

XX
-15 -15
-20 -20
-25 y=0 -25
y=0
-30
-35
F -30 SF
-35
-40 -40
200 210 220 230 240 250 260 270 280 290 300 200 210 220 230 240 250 260 270 280 290 300
z (mm) z (mm)
Fu Fu 97 Fu 94 Fu 86 Fu 67 Fu 56 Fu Fu 95 Fu 93 Fu 92 Fu 90 Fu 85 Fu 80

6 6
4 4
2 2
0 0
-2 -2
XX

XX

-4 -4
-6 -6
-8
y = 25 mm -8
y = 25 mm
-10
-12
F -10 SF
-12
200 210 220 230 240 250 260 270 280 290 300 200 210 220 230 240 250 260 270 280 290 300
z (mm) z (mm)
Fu Fu 97 Fu 94 Fu 86 Fu 67 Fu 56 Fu Fu 95 Fu 93 Fu 92 Fu 90 Fu 85 Fu 80

10 10

5 5

0 0

-5 -5
XX

XX

-10 -10

-15 y = 50 mm -15 y = 50 mm
-20 F -20 SF
-25 -25
200 210 220 230 240 250 260 270 280 290 300 200 210 220 230 240 250 260 270 280 290 300
z (mm) z (mm)
Fu Fu 97 Fu 94 Fu 86 Fu 67 Fu 56 Fu Fu 95 Fu 93 Fu 92 Fu 90 Fu 85 Fu 80

Fig. 16. Numerical results for PO1-300, Normal stress sxx (MPa) in x E100 mm.

The state of stress and damage are presented in Figs. 16–19, accompanied by perpendicular stress sxx, which can result in
only in the case of PO1-300 (for the sake of concision), for different unexpectedly high values.
loads. It can be observed that, regardless of the friction assumption The damage variable D begins to increase at a load of approxi-
(F or SF), the stresses and damage are not uniform along the mately 60% Fu, and the damage progressively spreads on the entire
bonding area. They exhibit a small number of peaks and they vary bonding surface as the load increases. Next, the damage induces
during the loading up to failure. Furthermore, it can be observed stress redistribution from the more damaged zones to the less
that near the interface, the axial stress szz and shear stress τxz are damaged zones or to undamaged zones. As the bonding area
 C. Meaud et al. / International Journal of Adhesion & Adhesives 54 (2014) 131–142 139

5 5
0 0
-5 -5
-10
-10
-15
ZZ

ZZ
-15
-20 y = 25 mm -20
y = 25 mm
-25
-30 -25
-35 F F -30 SF
-40 -35
200 210 220 230 240 250 260 270 280 290 300 200 210 220 230 240 250 260 270 280 290 300
z (mm) z (mm)
Fu Fu 97 Fu 94 Fu 86 Fu 67 Fu 56 Fu Fu 95 Fu 93 Fu 92 Fu 90 Fu 85 Fu 80

Fig. 17. Numerical results for PO1-300. Axial stress szz (MPa) in xE 100 mm.

0 0

-2 -2

-4 -4

-6 -6

-8 -8

-10
y=0 -10 y=0
-12 -12
-14 F -14
SF
-16 -16
200 210 220 230 240 250 260 270 280 290 300 200 210 220 230 240 250 260 270 280 290 300
z (mm) z (mm)
Fu 60 Fu 70 Fu 85 Fu 90 Fu 95 Fu Fu Fu 95 Fu 93 Fu 92 Fu 90 Fu 85 Fu 80

0 0

-2 -2

-4 -4

-6 -6

-8 -8
-10 -10
y = 50 mm
-12
-14
F y = 50 mm -12
-14 SF
-16 -16
200 210 220 230 240 250 260 270 280 290 300 200 210 220 230 240 250 260 270 280 290 300
Longueur du joint de colle (mm) z (mm)

Fu 60 Fu 70 Fu 85 Fu 90 Fu 95 Fu Fu Fu 95 Fu 93 Fu 92 Fu 90 Fu 85 Fu 80

Fig. 18. Numerical results for PO1-300. Shear stress τxz [MPa] in xE100 mm.

1.0 1.0
0.9 0.9
F
Damage d in concrete
Damage d in concrete

0.8 0.8
0.7 0.7 SF
0.6 0.6
0.5 0.5
0.4 y=0 0.4
0.3 0.3 y=0
0.2 0.2
0.1 0.1
0.0 0.0
200 210 220 230 240 250 260 270 280 290 300 200 210 220 230 240 250 260 270 280 290 300
z(mm) z (mm)
Fu Fu 97 Fu 95 Fu 85 Fu 70 Fu 58 Fu Fu 95 Fu 93 Fu 92 Fu 90 Fu 85 Fu 80

1.0 1
0.9 0.9
Damage d in concrete

Damage d in concrete

0.8 0.8
0.7 y = 50 mm 0.7 SF y = 50 mm
0.6 0.6
0.5 0.5
0.4
0.3
F 0.4
0.3
0.2 0.2
0.1 0.1
0.0 0
200 210 220 230 240 250 260 270 280 290 300 200 210 220 230 240 250 260 270 280 290 300
z (mm) z (mm)
Fu 95 % Fu 85 % Fu 70 % Fu 58 % Fu Fu Fu 95 Fu 93 Fu 92 Fu 90 Fu 85 Fu 80

Fig. 19. Numerical results for PO1-300. Damage d in x E100 mm.

140 C. Meaud et al. / International Journal of Adhesion & Adhesives 54 (2014) 131–142

exhibits total and /or excessive damage, the convergence status of 3.2.3. Parametric study
the calculation cannot be achieved, and the numerical failure load In the numerical results presented above, the influence of the
is considered as reached. bonding length on the behaviour was studied. However, it should
Additionally, it can be noted that the shear stress distribution is be noted that the distance between the bottom of the slab and the
almost constant in the central zone of the bonding joint up to
approximately 80% of the failure load with a small peak at both
extremities of the interface. As failure approaches, these peaks
increase and can achieve a high value greater than the tensile
strength of the concrete, which is possible only because of the
perpendicular compression stress sxx at the same location. Thus,
regardless of the assumption of S or SF, the bonding joint is not
subjected to a simple shear state of stress but is submitted to shear
and perpendicular (tension or compression) stresses.
The main difference between the F-case and SF-case lies in the
sign, the distribution and the intensity of the perpendicular stress Shear +
sxx. In the F-case, sxx is mainly a compression stress, whereas in Compression
the SF-case, it is a tension stress. Consequently, in the latter case,
the damage occurs at a lower load and increases faster than in the Friction
F-case; hence, the failure load is lower.
The numerical simulations were carried out for all push-out
specimens. The results show that the size of the concrete slabs does
not significantly influence the behaviour and, particularly, the state of Fig. 21. Strut effect in push out test in case of friction.
stress in the concrete slab near the interface, as illustrated in Fig. 20.
The bonding length (L) induces similar distributions in the z-direction.
Further calculations were conducted to investigate the influ-
ence of several parameters of the constitutive laws. For instance, in
the SF-case, the maximal load mainly depends on the parameter
εDO, which initiates the failure whereas this parameter has a week
influence in the F-case (see more details in [17]). Finally, it can be

200 mm
concluded that the response of the push-out specimen mainly Bonded
depends on the concrete constitutive law. Thus, the parameters 100 mm
area
should be carefully selected as explained by Jason et al. [21].
Therefore, preventing all displacement between the steel press
support and the bottom of the concrete slab caused by friction h
results in a shear state of stress and perpendicular compression in
the bonding joint (illustrated in Fig. 21),which can explain the high
h : freeheight(100, 50, 25 or 10 mm)
failure load corresponding to an average ultimate shear stress
greater than the tensile strength of the concrete. Fig. 22. Definition of the free height.

0 0
-2 -2
Shear stress [MPa]
Shear stress [MPa]

-4 -4
-6 -6
-8 -8
-10
-10
-12
-14 PO1 -12 PO2
-16 -14
-18 -16
-20 -18
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
z/L z/L
PO2_300_F, Fu PO2_100_F, Fu
PO1_100_F PO1_300_F PO1_100_SF PO1_300_SF
PO2_300_SF, Fu PO2_100_SF, Fu

0
-2
Shear stress [MPa]

-4
-6
-8
-10
-12
-14
PO3
-16
-18
-20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
z/L
PO3_300_F PO3_300_SF

Fig. 20. Influence of the bonded area and the size of the concrete slab on the ultimate average shear stress distribution.
 C. Meaud et al. / International Journal of Adhesion & Adhesives 54 (2014) 131–142 141

0 0

-2 -2

Shear stress [MPa]

Shear stress [MPa]

F
-4 -4

-6 -6

-8 -8

-10 -10
SF
-12 -12
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
z/L z /L
mm 10 mm 25 mm 50 mm 100 mm 10 mm 25 mm 50 mm 100

Fig. 23. Ultimate shear stress for different values of free height.

9 correlation technique may allow for more accurate and complete

With friction (F)
8 measurement of the slip and strain near the interface.
7
A 3D non-linear FE model is proposed to investigate the push-
out test. This model allows for accurately predicting the behaviour
6
in the case of the 300  300  100 mm3 concrete slabs, assuming
[MPa]

5 that the displacement of the bottom surface of the concrete slab is

completely prevented because of friction. Accordingly, the discre-
moy u

3 Without friction (SF) pancies between the measurements and numerical results are
minor. As a result of the strut effect, this boundary condition
2
induces a shear state of stress accompanied by a high perpendi-
1
cular compression stress. This complex and favourable state of
0 stress delays the failure and allows the achievement of an average
0 20 40 60 80 100 120
ultimate shear stress greater than the tensile strength of concrete.
Free height h [mm]
However, this assumption should be checked by further tests, such
Fig. 24. Influence of the free height on the average ultimate shear stress. as measuring the displacement at the bottom of the slab. Further
simulations show that this strut effect may be reduced if the free
height is lower than 50 mm.
bottom of the steel HEB, which is called the “free height” (see Thus, in its present state, the push-out test seems to exhibit a
Fig. 22), decreases as the bonding length increases; hence, the two small bias to characterise the steel–concrete bonding connection.
parameters vary. To identify the only influence of the free height, If this test can be used because of its simplicity or the possibility of
further simulations were conducted on the PO1-200 series exhi- comparing different bonding processes, it is questionable if the
biting 200  200  100 mm3 concrete slab (the other assumptions results or their use in terms of average shear stress may be
were kept unchanged). The bonding length was maintained at a beneficial for design purposes. In fact, the favourable strut effect
constant of 100 mm, whereas the position of the steel HEB with present in the push-out test can probably not be developed to the
respect to the concrete slab was modified, as shown in Fig. 22. same order of magnitude for a flexural member. Therefore, it is
Therefore, the free height has been given 4 different values, as important to improve the characterisation of the interface in
follows: 100, 50, 25 and 10 mm. reducing or eliminating the strut effect, by decreasing the variation
The numerical results, presented in Figs. 23 and 24, confirm the of the shear stress along the interface and ensuring the indepen-
influence of the free height on the behaviour. In particular, the dence of the results from the sizes of the slabs or bonding areas.
failure load and shear stress in the concrete near the interface tend
to decrease as the free height decreases in the F-case, whereas
the values increase if no friction is taken into account (SF-case). References
If the free height is lower than 50 mm, the failure loads are
approximately the same in both cases (S and SF); however, the [1] Lebet JP, Ducret JM. Experimental and theoretical study of the behaviour of
average ultimate shear stress remains greater that the tensile composite bridges during construction. J Construct Steel Res 1998;46:69–70.
strength of the concrete. [2] Steel Concrete Bridges – A guide for novel structures – National program
MIKTI (France): Presses des Ponts; 2010.
[3] Bouazaoui L, Perrenot G, Delmas Y, Li A. Experimental study of bonded steel
concrete composite structures. J Construct Steel Res 2007;63:1268–78.
[4] Bouazaoui L, Jurkiewiez B, Delmas Y, Li A. Static behaviour of a full-scale Steel–
concrete beam with epoxy bonding connection. Eng Struct 2008;30:1981–90.
4. Conclusions [5] Si Larbi A, Ferrier E, Jurkiewiez B, Hamelin P. Static behaviour of a steel–
concrete beam connected by bonding. Eng Struct 2007;29:1034–42.
The Steel–concrete bonding connection has been investigated [6] Si Larbi A, Ferrier E, Hamelin P. Concrete to steel lap joint failure criteria under
combined shear and peeling stress. J Construct Steel Res 2009;65:386–94.
by push-out tests. The results indicate that the behaviour and [7] Jurkiewiez B, Meaud C, Michel L. Non linear behaviour of steel concrete epoxy
average ultimate shear stress depend on the size of the concrete bonded composite beams. J Construct Steel Res 2011;67:389–97.
slabs and on the bonding length. In order to limit small geometric [8] Japanese Concrete Institute. Report II of Research Committee on continuous
fiber reinforced. vol. 5; 1998.
defaults, such as the parallelism of either interfaces, and to
[9] Ferrier E, Hamelin P. Long-time concrete composite interface characterization
increase the reliability of such push-out tests, the use of under- for reliability prediction of RC beams strengthened with FRP. Mater Struct
sized concrete slabs is not advised, and an increase in the number 2001;35(253):564–72.
of identical specimens to 5 or 6 is recommended. Moreover, [10] Berthet J, Yurdtdas I, Delmas Y, Li A. Evaluation of the adhesion resistance
between steel and concrete by push out test. Int J Adhes Adhes 2011;31:75–83.
fabrication of the specimens could be improved by sawing the [11] Eurocode 4. Calculation of the steel–concrete composite structures and National
bottom face of the concrete for instance. Finally, an image Application Document, part 1: general rules and buildings rules; 1997.
142 C. Meaud et al. / International Journal of Adhesion & Adhesives 54 (2014) 131–142

[12] Eurocode 4. Calculation of the steel–concrete composite structures and [19] Kim S, Nguyen H. Finite element modelling of push-out tests for large stud
National Application Document, part 2: composite bridges; 2000. shear connectors. J Construct Steel Res 2009;65:1909–20.
[13] Da Silva LFM, Das Neves PJC, Adams RD, Wang A, Spelt JK. Analytical models of [20] Jason L. Relation endommagement-perméabilité pour les bétons. [PhD thesis].
adhesively bonded joints—Part II: comparative study. Int J Adhes Adhes Nantes, France: Ecole Centrale de Nantes; 2004.
2009;29:331–41. [21] Jason L, Huerta A, Pijaudier-Cabot G, Ghavamian S. An elastic plastic damage
[14] Hart-Smith LJ. The design of adhesively bonded joints. Adhes Sci Eng formulation for concrete: application to elementary tests and comparison with
2002;2:725–77. an isotropic damage model. Comput Methods Appl Mech Eng 2006;195(52):
[15] Bigwood DA, Crocombe AD. Non-linear adhesive bonded joint design analysis. 7077–92.
Int J Adhes Adhes 1990;10:31–41. [22] Krayani A. Contributions to the nonlinear modelling of the mechanical beha-
[16] Adams RD, Comyn J, Wake WC. Structural adhesive joints in engineering. 2nd viour of concrete and of the reinforced and prestressed concrete structures.
ed.. London: Chapman & Hall; 1997.
[PhD thesis]. Nantes, France: Ecole Centrale de Nantes; 2007.
[17] Meaud C. Analyse multi échelles des connections par collage—application aux
[23] Pijaudier-Cabot G, Bazant Z. Non local damage theory. J Eng Mech 1987;113:
elements structuraux multi matériaux fleches. [PhD thesis], 1. Lyon, France:
1512–33.
Université Claude Bernard Lyon; 2013.
[24] Mazars J, Pijaudier-Cabot G. Continuum damage theory: application to con-
[18] Castagnetti D, Dragoni E. Standard finite element techniques for efficient stress
crete. J Engineering Mech 1989;115:345–65.
analysis of adhesive joints. Int J Adhes Adhes 2009;29:125–35.