Reinforcement-concrete bond behavior: experimentation

in drying conditions and meso-scale modeling

A. Michoua,b , A. Hilairea , F. Benboudjemaa , G. Nahasa,c , P. Wynieckid ,
Y. Berthaude
a
LMT-Cachan, ENS Cachan, CNRS, Université Paris Saclay, Cachan, France
b
Sorbonne Universités, UPMC Univ Paris 06, Paris, France
c
Institut de Radioprotection et de Sûreté Nucléaire, Fontenay-aux-Roses, France
d
Nuvia Structure, Aix-en-Provence, France
e
Sorbonne Universités, UPMC Univ Paris 06, CNRS, Institut Jean le Rond
d’Alembert,Paris, France

Abstract

Reinforced concrete structures must fulfill serviceability functions in terms

of durability and bearing capacities. Steel-concrete interface plays a crucial
role in the cracking behavior (crack openings, spacing). Hence, its accurate
description should be considered in a relevant way. In this paper, a meso-
scale reinforcement-concrete bond model is proposed, based on an explicit
mesh of the interface area. This area implicitly considers the presence of
steel ribs and the progressive steel-concrete slip through two coupled plastic
criteria. The model is calibrated on pull-out tests, validated on long rein-
forced concrete ties in tension and compared with experimental tests carried
out for this study. In this experiment, local strain in reinforcement and sur-
face cracking mapping have been measured. Special attention is also devoted
to the effect of drying shrinkage and creep on the structural and cracking be-
havior. Their effects are significant, but greatly depend on the reinforcement

Email address: farid.benboudjema@ens-cachan.fr (F. Benboudjema)

Preprint submitted to Engineering Structures July 15, 2015

ratio.
Keywords: Reinforcement-concrete bond, Cracking, Digital Image
Correlation, Optic sensors, Drying shrinkage

1. Introduction

Reinforced concrete structures are used for various applications in civil

engineering. In some particular cases, in addition to the knowledge of their
structural strength, the issue of the durability and the tightness is involved.
Tightness and durability are closely related to the cracking behavior of the
structure. A flow of liquid or gas across the structure significantly increases in
the presence of cracks [1] [2] [3], leading to internal concrete degradation and
corrosion of reinforcement for example. The cracking evolution in reinforced
concrete structures mainly depends on the transfer of internal forces between
concrete and reinforcement, and consequently on the evolutive steel-concrete
bond. Considering the aforementioned feature in numerical simulations en-
ables an accurate prediction of the cracking behavior [4].
The mechanisms of the steel-concrete bond and its progressive degrada-
tion have been intensively studied. The bond resistance can be decomposed
in different stages [5]. For ribbed rebars, a relatively low physical and chem-
ical cohesion links up steel with concrete. Inclined cracks then appear at
steel ribs, leading to a first decrease of the bond stiffness. Concrete crushing
and shear cracks progressively propagate from the ribs until their coales-
cence (maximum and loss of bond stress). For large relative displacements
between steel and concrete, the residual bond resistance is provided by fric-
tion. These local degradation mechanisms have been experimentally studied

2
on reinforced concrete ties [6], on pull-out tests [7], and also numerically
studied [8]. In an overall approach, a lot of works on pull-out tests have been
conducted in order to study the parameters influencing the bond resistance:
concrete strength, diameter and geometry of the rebar, active and passive
confinement [9] [10].
Numerically, different approaches have been developed to model the be-
havior of the steel-concrete bond. The principle is to consider the non linear
evolution between the bond stress τ and the steel-concrete relative displace-
ment. Based on damage theory [11], or on plasticity theory [12], joint el-
ements between steel and concrete were developed. They ensure the load
transfer and an explicit relative displacement between concrete and rein-
forcement. These methods generally imply an important calculation cost for
concrete structures. Moreover, convergence problems can be identified, be-
cause non-penetration conditions must be imposed between both materials.
An other approach models the interface through a 2D and 3D explicit area
[13] [14] [15], using the plasticity theory to model the relative slip between
steel and concrete (Mohr-Coulomb or Drucker-Prager criterion). This tech-
nique is numerically convenient, but has difficulties in capturing the progres-
sive loss of bond stiffness during loading. We can also cite other interesting
studies: (i) one is based on internal forces and kinematic relations between a
truss-element rebar and plain concrete [16], (ii) an other one extends an ex-
plicit cracking model for concrete to the bond behavior, taking into account
the progressive degradation in shear [17], (iii) embedded or extended finite
element methods are also used on joints elements [18].
In this communication, a meso-scale bond model is proposed. Unlike

3
other approaches, it implicitly considers the influence of the geometry of the
steel rebar on the local bond degradation and bond slip. For sake of simplicity
in the modeling and calibration process, the interface is considered as an
explicit massive area around a truss-element rebar [14] [15] [19]. Compared
to joint elements, this approach remains in a continuous problem, without
contact resolution. It favors the numerical robustness and the convergence
of the calculations. A procedure for the calibration and the validation of
the bond model is proposed, and is based on experimental tests: standard
pull-out tests and tension tests on reinforced concrete ties. The behavior of
the specimens is locally analyzed by Digital Image Correlation for concrete
cracking, and optic sensors for strain measurements of the embedded rebars.
After calibration and validation, the meso-scale bond model can be used for
structural analysis.

2. General approach

The presentation is decomposed into three main parts:

• The bond behavior is first studied by experimental tests. Standard

pull-out tests are first carried out to quantify the local bond behav-
ior. Tension tests on long reinforced concrete ties are then performed.
They exhibit the coupled behavior between the steel-concrete interface
and the concrete cracking. The bond resistance actually influences the
number of cracks, the crack spacing and consequently the crack open-
ings.

• A meso-scale bond model is proposed. The identification of the bond

properties is performed on the experimental results of the pull-out tests.

4
 Characterization tests on steel and concrete serve to calibrate the nu-
merical steel and concrete behaviors.

• Finally, the identified steel, concrete and bond properties are simultane-
ously used for numerical simulations of the tension test. The results of
this structural analysis provide a validation of the proposed approach.
They also guarantee its numerical relevance and efficiency to design
reinforced concrete structures.

For that purpose, each specimen is fabricated with the same concrete
mixture and the same steel rebar, in order to limit the variability of both
the material and the bond properties. The samples are protected against
dessiccation up to 17 days and are tested after 110 days. They are therefore
exposed to drying conditions during 90 days (average relative humidity of 40
%). Concrete mixture presents a water-cement ratio equal to 0.48. Concrete
properties are measured on cylindrical samples: Young’s modulus, tensile
strength (splitting test) and compressive strength Ec , ft and fc , respectively)
and on notched concrete beams (fracture energy Gf ) [20]. The results are
presented in Table 1, as well as the mechanical properties of the steel rebar
: Young’s modulus Es , yield strength fy . The steel rebar and its idealized
geometry are presented in Figure 1. The distance between two consecutive
ribs is equal to 8 mm. The length and height of the ribs are 4 mm and 0.5
mm, respectively.
Specific attention is also devoted to the ”life” of the specimen before me-
chanical loading. Phenomena such as autogeneous shrinkage, drying shrink-
age, creep, induce a self-equilibrated stress field in the structure. Since the
delayed strains are restrained by the steel rebars, tensile stresses are induced

5
 Concrete Mean Value (min. - max.) Steel rebar
Ec 35 GPa (34.45 - 35.96) Es 200 GPa
ft 2.9 MPa (2.78 - 3.04) fy 500 MPa
fc 49.4 MPa (48.15 - 50.60)
Gf 94.6 J/m2 (81.5 - 104.5)

Table 1: Main material properties.

L = 8 mm
hr = 0.5 mm
lr = 4 mm

(a) (b)

Figure 1: Steel rebar of diameter d = 12 mm (a) and its idealized geometry (b)

in concrete, which may lead to a debonding at the steel-concrete interface,

the same way as the cracking and debonding at the aggregates-cement paste
interface [21]. The influence of the delayed strains is numerically quantified
on the studied specimens.
The numerical simulations are carried out with the help of the finite
element code Cast3m [22].

3. Experimental bond tests

With the aim of calibrating the developed bond model, standard pull-
out tests have been carried out. A single rebar of diameter d = 12 mm is
embedded into a 15d concrete cube. The anchorage length is equal to 5d
(Figure 2-b). This configuration provides a mechanism of pulling, instead

6
of splitting. Teflon sheets are added on the loaded surface, as detailed in
Figure 2-a, in order to avoid friction and spurious stresses in the specimen.
The connection between the specimen and the fixed support is managed by a
steel bearing (Figure 2-c) . It corrects a possible lack of co-axiality between

Spherical bearing sheets

Displacement sensor LVDT
Teflon sheets

(a)

15 d
LVDT
F 5d
d

15 d

(b) (c)

Figure 2: Configuration of the pull-out tests: (a) tested specimens, (b) geometrical di-
mensions, (c) steel bearing for axiality

the fixed loading axis (hydraulic jack of 15 kN capacity) and the axis of the
rebar. A displacement sensor is located at the unloading end of the rebar,
measuring its relative slip with concrete. If the bond stress τ is supposed to
be constant along the anchorage length l, it can be deduced from the applied
load F :
F
τ= (1)
πdl
Three samples are tested until large displacements, meaning until the

7
residual friction between steel and concrete (Figure 3). The results highlight
a four-step bond behavior. A first increase of bond stress without slip is
observed up to τ = 3 MPa (point A). The bond stiffness progressively de-
creases until a maximum bond stress τmax (about 13 - 14 MPa) (point B).
This stage is followed by a loss of resistance with a larger slip (point C). It
seems interesting to note the next increase of resistance for each test (point
D). It can be explained by the ”ribs by ribs” displacement of the rebar into
concrete, involving friction and residual ribs resistance. Indeed, the distance
between the point C and the point E is equal to the distance between two
consecutive ribs on the rebar.

15 Test 1
Test 2
Test 3
Bond stress τ [MPa]

B
10

D
5
A
C
E
0
0 2 4 6 8 10 12 14
Slip [mm]

Figure 3: Experimental results: pull-out tests (zoom within the range of small slips in the
figures 14 and 16)

8
4. Experimental tension tests on reinforced concrete ties

Tests on long reinforced concrete ties are then carried out (Figure 4).
Three 1.15 meter-long structures are tested in tension (Figure 5). The con-
crete cross section is equal to 10×10 cm2 . The length of the specimens favors
the localization of multiple transverse cracks during loading. A steel rebar
(identical to the rebar in pull-out bond tests, diameter d = 12 mm) is em-
bedded on 1 meter. The test is performed up to yielding of reinforcement.

Figure 4: RC tie in tension : experimental set up

1m 10 cm
F d F
10 cm

Figure 5: Geometric dimensions of reinforced concrete ties

4.1. Concrete cracking

Digital Image Correlation [23], noted DIC, is performed during loading

on each specimen. Two cameras are located in front of the sample and point
to a random speckle, obtained by paint spray across a perforated steel grill.

9
Pictures are taken during loading and their analysis by DIC then leads to
the evolution of the displacement field of the studied surface in two dimen-
sions. The measurement error can be evaluated a priori at about 4 µm. An
example is given in Figure 6-a, presenting the displacement field in the load-
ing direction of a half structure after cracking. Displacement gaps appear

(a)

−10

−11
Axis displacement [pixel]

−12
crack opening

−13

−14

−15 Measured displacement

Linear interpolations

0 500 1,000 1,500 2,000 2,500

AB axis [pixel]

(b)

Figure 6: Measurement of crack openings: (a) displacement field in loading direction after
cracking, (b) post-processing by multi linear regressions

during loading and represent the progressive localizations of each crack. The

10
axial displacement can be evaluated along a chosen axis (AB axis for exem-
ple, Figure 6-a). An automatic computation of multiple linear regressions
is performed between two consecutive cracks (Figure 6-b). For a crack i at
the position xi , let us name f and g the linear regressions at the left and
right edges of the discontinuity, respectively. The local opening δui can be
calculated as δui = g(xi ) − f (xi ) (Figure 6-b). The DIC is performed on one
surface only. The opening wi of the crack i is considered as the mean value
of local crack openings δui , which are calculated on the studied area (Figure
6-a). The pixel / mm conversion is independently calculated on the three
RC ties. On average, one pixel is equal to 250 µm.
The final cracking patterns are presented in Figure 7, after yielding of
reinforcement. Five or six cracks localize during loading, perpendicular to the
loading axis. These qualitative results can then serve as a comparison with
the numerical simulations. A quantitative characterization of the cracking

Test 1

Test 2

Test 3

Figure 7: Experimental final cracking pattern - RC Ties

behavior can also be proposed. First, the sum of crack openings Sop is defined

11
as:
X
Sop = (wi ) (2)
i

where wi is the crack opening of crack i. This criterion is related to the

elongation of the specimens. It can be also presented as a durability criterion.
A flow by diffusion across a structure is linear with respect to the crack
opening. An other criterion is defined as the evolution of the maximum
crack opening wmax :
wmax = max(wi ) (3)

This criterion is more local, but must be theoretically equal to the mean crack
opening in the structure. Both evolutions are presented in Figure 8. Up to F
= 20 kN, the concrete presents an elastic behavior, without transverse cracks
on the studied surface of the specimens. The first crack appears at about F
= 20 kN. The other ones progressively localize between F = 20 kN and F =
40 kN. The crack openings then increase up to yielding of the steel rebar (F
= 60 kN).

4.2. Strain field of the steel rebar

The obtained cracking behavior is highly dependent on the steel-concrete

bond [16]. But the local behavior of the embedded rebar is often difficult to
evaluate experimentally. With this aim, an instrumentation of the rebar by
optical fibers is used on two samples (Test 2 & Test 3). This technique, devel-
oped at IFSTTAR, has recently shown to be efficient for strain measurements
on concrete structures, allowing crack detection [24] and monitoring of em-
bedded rebars [25], with low uncertainty (10−6 ) and high spatial resolution (1
mm). An optical fiber of diameter 250 µm is glued in an engraving along the

12
 4
Test 1
Test 2

Sum of crack openings [mm]

Test 3
3

0
0 10 20 30 40 50 60 70
Applied load F [kN]

(a)
900
Test 1
Maximum crack opening [mm]

Test 2
720 Test 3

540

360

180

0
10 20 30 40 50 60
Applied load F [kN]

(b)

Figure 8: Experimental cracking behavior of RC ties : sum of crack openings (a) and
maximum crack opening (b)

13
rebar (Figure 9). The instrumentation has the advantage to be few-intrusive.

Figure 9: Engraving on the rebar with an embedded optical fiber

It slightly impacts the steel-concrete interface and the mechanical behavior of

the rebar (the engraving represents 0,9 % of the steel cross section). The op-
tical fiber (the sensor) is coupled with an Optical Backscatter Reflectometer
OFDR, working in frequency domain. The analysis of the backscattered fre-
quency spectrum, compared to a reference measurement, leads to the strain
field along the fiber, and consequently along the rebar (assuming no variation
of temperature). The frequency domain provides high resolution measure-
ments (1 mm).
The strain field (x) is obtained at each loading step on the embedded
rebar. The Figure 10-a and Figure 10-b show the evolution of (x) before
the first transverse crack, after the first transverse crack and after the second
transverse crack, for the Test 2 and Test 3, respectively. Before cracking, the
strain field remains quasi-constant in the middle area, up to a threshold  =

14
 1,200
∆u(F = 18,2kN)

Strain field (x) (.10−6 )

1,000

800

600

400

200

0
0 0.2 0.4 0.6 0.8 1
Embedded length [m]
F = 20kN before cracking
F = 18,2kN after crack 1
F = 18,2kN after crack 2

(a)

∆u(F = 18,5kN)
Strain field (x) (.10−6 )

1,000

500

0
0 0.2 0.4 0.6 0.8 1
Embedded length [m]
F = 21kN before cracking
F = 18,5kN after crack 1
F = 19,4kN - after crack 2

(b)

Figure 10: Evolution of strain field on the embedded rebar - optical fibers measurements
: (a) Test 2, (b) Test 3
15
7×10−5 in the rebar. The progressive transfer of load between concrete and
reinforcement can be seen at both edges (15-cm length). In this area, the local
variations can be due to local degradation at the interface (drying shrinkage
restraint for example). The optic sensor then captures the successive cracks
along the specimen and the tension stiffening on both sides of the cracks.
The transfer length, on both sides of a crack, is also equal to about 15 cm.
By integration on the length of the embedded rebar (1 m), the overall
elongation ∆u is calculated at each loading step as:
Z 1
∆u(F ) = (x, F ) dx (4)
x=0

The mechanical behavior F - ∆u of the structures is presented in Figure

11. Up to F = 20 kN, the initial stiffness represents the elastic behavior of
the structure. Then, the stiffness progressively decreases, in accordance with
the cracking behavior (Figure 8). A similar behavior is observed for the two
tests.

5. Modeling of the steel - concrete bond

The approach is based on an explicit 3D meshing and modeling of the

interface. The interface area links up the steel rebar, modeled by 1D truss
elements, with concrete around it. Its diameter is taken as equal to the real
diameter of the rebar (Figure 12). The rebar-interface and interface-concrete
bonds are considered as perfect with coincident nodes. Damage theory is
used to model concrete cracking. A scalar damage variable D, ranging from
0 to 1, is considered in the stress-strain relation:

σij = (1 − D)Cijkl elas

kl (5)

16
 60

Applied load F [kN]

40

20

Exp : Test 2
Exp : Test 3
0
0 1 2 3 4
∆u on the embedded length [mm]

Figure 11: Mechanical behavior F - ∆u of the RC ties

where σij , Cijkl and elas

kl are the stress, elastic stiffness and elastic strain com-

ponents, respectively. The evolution of damage D is based on an equivalent

strain criterion, involving the equivalent strain eq introduced by [26] for con-
crete. The evolution of damage in tension can be considered as exponentially
decreasing [27] :
d0
D =1− · exp(−Bt (eq − d0 )) (6)
eq
where Bt is a parameter controlling the softening behavior and d0 = ft /Ec ,
with ft the tensile strength and Ec the Young modulus. The concrete behav-
ior, stress-strain σ11 - 11 in tension, is presented in Figure 13-a. According
to the softening behavior, an energetic regularization [28] prevents of mesh
dependency. The regularization is based on the parameter Bt which depends
on the size of the finite element h, fracture energy Gf and parameters for

17
 Rebar

Concrete Interface

Figure 12: 3D modeling of the interface area

damage initiation d0 and ft , tensile strength of concrete:

h ft
Bt = h d 0 f t
(7)
Gf − 2

A correction is also introduced to limit damage in shear loadings (parameter

β = 1.06, detailed in [29]). The concrete behavior, stress-strain σ12 - 12 in
pure shear, is presented in Figure 13-b. It exhibits a residual shear stress
after the elastic phase. In the case of a perfect bond, the local bond behavior
cannot be simulated, in comparison with the experimental results on pull-out
tests (Figure 14). Taking into account a steel-concrete interface is therefore
necessary.
In the classical 3D approaches [14] [15], a plastic behavior is considered
for the interface, following a homogeneous Von Mises criterion with softening
along the rebar axis : 
 σeq = K
(8)
 dK = H dp

where K is the plastic threshold, σeq is the Von Mises equivalent stress, H < 0
is the softening parameter and p is the cumulative plastic strain. The Young’s

18
 3

Stress σ11 [MPa]

0
0 0.5 1 1.5 2
Strain 11 ·10−3

(a) tension
2.5

2
Stress σ12 [MPa]

1.5

0.5

0
0 1 2 3 4 5 6
Strain 12 ·10−4

(b) pure shear

Figure 13: Concrete behavior in tension and pure shear

19
modulus and the Poisson’s ratio are also taken as equal to the concrete one,
in order to guarantee the accurate initial stiffness of the structure. Compared
to a perfect bond, the main advantage of this bond model is to reproduce the
maximum bond stress, by the calibration of the parameter K (Figure 14) .
Nevertheless, for small slip values, the loss of stiffness, before the maximum

20

15
Bond stress τ [MPa]

10

0
0 0.2 0.4 0.6 0.8 1
Slip [mm]
Exp : 3 tests
Num : perfect bond with truss elements
Num : classical 3D interface (Equation 8)

Figure 14: Pull-out tests: classical modelings

bond stress, is under-estimated, that could lead to an over-estimated transfer

length between steel and concrete in structural analysis.
In this communication, the philosophy of the 3D bond model is modified.
The modeling steps are based on the evolutive damage mechanisms at the
steel-concrete interface, relating to the steel geometry. At a low level of axial
tension in the steel rebar, the steel-concrete behavior can be characterized

20
by a chemical bond at the interface. The interface transfers the load from
the steel rebar to concrete by shear stresses, without slip between the two
materials. Once the cohesion is locally reached, a friction mechanism appears.
This step can be easily modeled by an perfect Von Mises criterion :

σeq = K1 (9)

The use of equation 9, over the length of the interface, can represent the bond
behavior of a smooth steel bar for example. However, a ribbed bar exhibits
an increase of bond resistance after the initial cohesion. That is mainly due
to the local anchorage of the steel ribs into concrete. The stiffness of the
interface decreases in this step. Indeed, radial cracks localize at steel ribs
and a relative displacement between the two materials can be measured.
The cracks later coalesce. The maximum bond resistance is then reached.
For larger slip values, the bond strength decreases until a final friction stage.
The approaches proposed in the literature do not take into account the local
influence of the steel ribs. The bond models rather show an homogenized
behavior, such as the 3D classical approach (equation 8). The proposed bond
model aims to consider the influence of the steel ribs, in order to capture the
local concrete degradation at the steel-concrete interface. Obviously, the
point is not to explicitly mesh the steel ribs in the numerical simulations.
They are implicitly considered by a second local criterion in the interface
area. Unlike the first overall one (equation 9), a von Mises criterion with
softening is locally added at the ribs locations. It is defined as :

 σeq = K2
(10)
 dK = H dp
2 2

21
where K2 is the initial local plastic threshold (K2 >> K1 ), H2 < 0 is the
softening parameter and p is the cumulative plastic strain. The model can
be seen as a periodic field of properties (Figure 15) and aims to quantify the
local mechanisms at the interface.
For the sake of simplicity, each finite element of the interface have the
same size along the axis of the interface (Figure 15-a-c). In the present case,
the axial thickness is fixed at 2 mm. Two layers of finite elements compose
the smooth zone between two consecutive ribs. The length of this zone is
consequently equal to 4 mm, in accordance with the measured geometry
of the rebar (Figure 1). The plastic threshold K1 is constant in this area
(Figure 15-b). Two layers of finite elements also compose the steel rib. Its
total length is equal to 4 mm, in accordance with the measured length of the
rib. An evolution of the plastic threshold and of the softening parameter is
imposed on these layers (Figure 15-b-d). For example, the plastic threshold
ranges from K1 to K2 in the first element and from K2 to K1 in the next one.
Consequently, the distance between two consecutive ribs is equal to 8 mm, as
experimentally measured. An assumption is made that lateral confinement
and its effect on bond properties are negligible in our study.
Finally, it seems important to note that the bond model is not regular-
ized. The mesh and the related bond properties are chosen according to the
geometry of the rebar (distance between the ribs mainly).

6. Calibration of the bond model

The developed bond model is calibrated on the experimental bond tests.

Concrete and steel properties are chosen in accordance with the measured

22
 K
K2 finite elements
K2
L
L
K1 K1
lr
rebar axis

(a) (b)

0 H lr rebar axis

L L
H2
H2
finite elements

(c) (d)

Figure 15: Periodic field of bond parameters, L: distance between two consecutive ribs

values (Table 1). The steel rebar is considered as elastic in this test. The
interface area presents a diameter of 12 mm. As previously described, its
mesh is related to the geometry of the steel rebar. Only three parameters
remain to be identified: the plastic thresholds K1 and K2 and the local
softening value H2 .
Threshold K1 : this parameter cannot be directly identified from the lo-
cal bond results. However, shear stresses τ are dominant at the interface.
√
The Von Mises equivalent stress can be therefore calculated as σeq = τ 3.
√
Consequently, the plastic threshold K1 is calculated as K1 = c 3, where c is
initial steel-concrete cohesion without lateral confinement. c can depend on

23
the steel roughness (presence of corrosion for example). Its value is chosen
from the literature as c = 1.5 MPa [30], obtained on a bond test developed
by Ouglova et al. [7]. Finally, according to the previous comments, K1 is
taken equal to 2.5 MPa.
Threshold K2 and softening H2 : Both parameters are identified by a
reverse analysis from the local bond results (Figure 3). K2 controls the
maximum bond stress. H2 controls the softening branch. The identification
leads to the following values : K2 = 200 MPa, H2 = -300 MPa.
These parameters provide the local bond behavior presented in Figure
16 for small slip values. The progressive loss of stiffness until the maximum
bond stress is simulated. In this phase, the bond behavior is controlled by
the local plastic behavior of the interface area and the concrete damage. The
damage field around the interface is shown in Figure 17-a and compared with
the classic approach (equation 8) in Figure 17-b.

15
Bond stress τ [MPa]

10

5
Exp : 3 tests
Num : classical 3D interface
Num : ”ribbed” 3D interface
0
0 0.2 0.4 0.6 0.8 1 1.2
Slip [mm]

Figure 16: Pull-out tests: effect of the developed model with a ”ribbed interface”

24
 (a) (b)

Figure 17: 3D modeling of pull-out tests, damage field around the interface at τ = 8 MPa:
(a) developed ”ribbed” interface, (b) classical 3D approach

In the developed approach, inclined cracks can be observed at ribs loca-

tions. The model is therefore more representative of the local bond degra-
dation, compared to the classical 3D approach. The full bond behavior is
presented in Figure 18. The final residual friction for large displacements is
not reproduced in the model, but (i) this stage is rarely reached in structural
applications, and (ii) if a discontinuity appear in the structure, only the rebar
must transfer the local load across the crack. Modeling the final friction by
the 3D interface would involve the participation of the interface in addition
to the rebar. Nevertheless, the model would not be able to represent the final
increase of resistance after 6 mm slip.

25
 15 Test 1
Test 2
Test 3

Bond stress τ [MPa]

Num : ”ribbed” 3D interface
10

0
0 2 4 6 8 10 12 14
Slip [mm]

Figure 18: Pull-out tests: numerical bond behavior compared to experimental results

7. Validation of the bond model

7.1. Introduction

The bond model is validated on the tension tests. With a concern for
representativity, the specimens are fabricated with the same concrete and
the same steel rebar as those used in the bond tests. Consequently, the first
numerical simulation takes only into account: (i) the measured materials
properties (Table 1), and (ii) the bond parameter as equal to those previously
identified (K1 = 2.5 MPA, K2 = 200 MPa, H2 = -300 MPa). A perfect
elastic-plastic behavior is taken for the steel rebar.
Especially for the tension test, the localization process is highly depen-
dent on the material heterogeneities. Numerically, uniform concrete proper-
ties lead to a diffuse damage field in the sample. Concrete heterogeneities
has therefore to be considered, in order to localize cracks in the specimen.

26
A random field on concrete tensile strength is generated by the Turning
Band Method [31]. The 3D correlated field presents an isotropic exponential
covariance. The mean value is taken as equal to the measured one ft , a stan-
dard deviation equal to 5% and a correlation length equal to three times the
maximum aggregate size (6 cm) [32].
The result F - ∆u of the first numerical simulation is presented in Figure
19. The initial stiffness shows a good correlation with experiments. Multiple
cracks appear during loading. But the major drawback of this simulation
is the significant over-estimation of the first crack initiation (F = 33 kN).
Generally, the concrete tensile strength is artificially decreased in order to
capture the experimental behavior [33]. However, the delayed strains can
have a significant influence on the mechanical response of a structure [34]
[35]. The influence of the concrete delayed strains is here studied, in order
to explain the experimental results.

7.2. Influence of drying shrinkage

Experimental measurements of autogenous shrinkage, drying shrinkage,

mass loss, basic and drying creep are performed on 7x7x28 cm3 samples and
used in the present simulations. The samples were exposed to an average
relative humidity of 45 % ± 5 and an average temperature of 25 degrees
Celsius ± 1. Autogenous shrinkage is not taken into account because of
its negligible evolution compared to drying shrinkage. The specimens are
protected against dessiccation up to 17 days and are tested 110 days after
pouring. They are exposed to an average relative humidity of 40 % and an
average temperature of 22 degrees Celsius. Concrete is considered as mature
during this period in the calculations (mechanical properties of concrete do

27
 40

30
Applied load F [kN]

20

10 Num
Exp : Test 2
Exp : Test 3
0
0 0.2 0.4 0.6 0.8 1
∆u on the embedded length [mm]

Figure 19: First numerical tension test - comparison with experimental results

not evolve). The total concrete strain is calculated according to the equation
11 where elas , sh , bc , dc are the respective elastic, drying shrinkage, basic
creep and drying creep strain :

tot = elas + sh + bc + dc (11)

The shrinkage strain is a function of time and of the size of the specimen
(diffusive phenomenon). In order to focus on reinforcement restraint, the
drying shrinkage strain sh
spe , imposed to the reinforced concrete tie, is ob-

tained according to the equation 12 (diffusion controlled process) [36]. This

equation considers the drying shrinkage strain as a function of the parameter
√
t/ri : √ √
sh t sh t
spe ( ) = test ( ) (12)
rspe rtest
where ri is the notional size, i=spe for the RC tie and i=test for the 7×7×28
cm3 sample for which the delayed strains are measured on. The notional size

28
is defined as 2 × Sdr /pdr , where Sdr is the drying surface, and pdr is the
drying perimeter (rspe = 5 cm, rtest = 3.5 cm). Consequently, the shrinkage
is supposed uniform in the specimen. This assumption implies that the auto-
restraint caused by the drying shrinkage gradient is neglected. The imposed
drying shrinkage strain sh
spe is presented in Figure 20. The creep modeling (see
Drying shrinkage strain sh (.10−6 )

300

200

100
7*7*28 cm3 specimens - sh
test
RC Tie - sh
spe
0
0 20 40 60 80 100
Drying time [day]

Figure 20: Drying shrinkage strain imposed to the structure

[37] for the description of the model) considers basic and drying creep strains
in concrete. The drying creep is assumed to be linear and proportional to
the drying shrinkage [38]. The basic creep parameters are identified from the
experimental creep tests in compression. The specific creep in tension and
compression are considered as equal, because of the lack of consensus in the
literature [37]. Specific basic creep strain (bc /σ) after 22 days is measured
as 9.6× 10−6 MPa−1 . The stresses, induced by the drying shrinkage strain,
are partially released by the creep strains. Hence, creep phenomena has to be
automatically considered, to avoid an over-estimation of the drying influence.

29
 Ninety days of drying are taken into account on a quarter of RC tie. The
Figure 21 shows the damage and axial stress fields after 90 days. Along the

Figure 21: RC tie : stress and damage fields in concrete, after 90 days in drying conditions
(quarter of structure) - the steel rebar and the interface are not represented

interface, it can be observed a slight concrete degradation (D ' 0.3) due to

the strain incompatibility between concrete and the couple rebar-interface.
However, considering a uniform drying shrinkage strain, the rebar restraint
induces significant tensile quasi-uniform axial stresses in the specimen. The
axial stresses are equal to about 0.75 MPa (=1/4 ft ). The mechanical re-
sponse of the tension test will be therefore influenced by the ”initial” state
of stress in the specimen. The stress field is not uniform at the interface, due
to the heterogeneities of the bond properties. The threshold K1 is reached
after 90 days of drying. That explains the stress localizations at the steel
ribs, which can favor the crack initiations in the specimen.
The tension test is then numerically carried out. The mechanical re-
sponse is plotted in Figure 22. The response is greatly impacted by the
drying shrinkage strain in the specimens. The first cracking force decreases

30
 40

30
Applied load F [kN]

20

Num : without drying

10 Num : with drying 90 days
Exp : Test 2
Exp : Test 3
0
0 0.2 0.4 0.6 0.8 1
∆u on the embedded length [mm]

Figure 22: Numerical tension test on RC ties, with or without considering drying (among
90 days)

by about 30 %, in accordance with the initial internal stresses before loading.

The first crack establishes the course of the further crack initiations. It is
also observed a slight loss of initial stiffness if drying. That is explained by
concrete degradation at the steel-concrete interface before loading (Figure
21).
Finally, the presented results show the importance of considering the de-
layed strains in structural analysis. It mainly controls the first crack initiation
along the specimen. However, the restraint phenomenon is reinforcement-
ratio dependent. A specific analysis must be therefore conducted for each
specific structure.

31
7.3. Mechanical behavior before the transverse localizations
Up to the first transverse crack (F ' 21 kN), the simulation shows an
quasi-constant stiffness (Figure 22). This phase can be considered as the
elastic structural behavior of the specimen. However, concrete degradation
can occur in this phase. Damage is initially initiated at the interface by
the drying phase (Figure 21). An additional local degradation can also be
observed at the edges during the tension test (Figure 23). It is localized at the
steel ribs and it decreases along the axis of the interface. The comparison of
the numerical damage field with the experimental results cannot be directly
carried out. Cracking is only analyzed on the external surface. But the optic
sensor provides the experimental axial strain field of the steel rebar. At F
= 20 kN, before transverse cracking, the experimental result is compared
with the simulated one (Figure 24). The strain field is quasi-constant in
the middle area. The strength of these results lies in the measurement of
the transfer length at the edges. The simulation is able to reproduce the
progressive anchorage of the rebar into the massive concrete. Hence, the
relevance of the bond model is validated in this phase.

7.4. Mechanical behavior after cracking

In this part, the full mechanical behavior is analyzed. With the aim of
providing an objective comparison with experimental results, several simula-
tions, respectively named simulation 1 - simulation 2 and simulation 3, are
proposed. Each simulation is characterized by a new generation of the ran-
dom field on concrete tensile strength (mean value ft , coefficient of variation
5 %, correlation length 6 cm, as previously mentioned). The results F - ∆u
are presented in Figure 25. The initial stiffness is similar for each simulation.

32
Figure 23: Damage field at the steel-concrete interface before the first transverse crack F
= 20 kN

1,000
Exp : Test 3
Num : ”ribbed” 3D interface
Strain field (x) (.10−6 )

500

0
0 0.2 0.4 0.6 0.8 1
Embedded length [m]

Figure 24: Strain field of the embedded rebar before the first transverse crack - F = 20
kN

33
That is logical, since the random tensile strength do not influence the elastic
behavior of concrete. Several load drops appear during loading due to cracks.
The first crack initiation and the localization process depend on the random
field, in accordance with the experimental results. The present simulations
provide a accurate modeling of the structural behavior.

60
Applied load F [kN]

40

20
Exp : 2 tests
Num : simulation 1
Num : simulation 2
Num : simulation 3
0
0 0.5 1 1.5 2 2.5
∆u on the embedded length [mm]

Figure 25: Tension test, mechanical behavior with drying conditions

Numerically, crack openings are obtained from the strain field. This
method is developed by (Matallah et al., 2010) [39] and implemented in finite
element code Cast3m. In each finite element, the strain field  is decomposed
in an elastic strain e and an inelastic strain i :

1 −1 −1
iij = ij − eij = Cijkl σij − Cijkl σij (13)
1−D

The local crack opening w is deduced from the inelastic strain as :

w = ni h iij nj (14)

34
where n is the normal vector from the crack and h is the size of the finite
element. The fields of crack openings are illustrated in Figure 26. Five or

Figure 26: Numerical tension tests - crack openings at steel yielding (F = 56 kN)

six cracks appear along the length of the specimens. These numerical results
show a good correlation with the experimental cracking patterns (Figure 7).
Their respective position depends on the random concrete properties. These
results also valide the robustness of the structural modeling, according to the
observed cracking patterns.
Quantitatively, the opening wi of the crack i is calculated as the mean
value of the local crack openings on the surface of the specimens. The evo-
lutions of Sop and wmax are shown in Figure 27 and Figure 28, respectively.
The experimental results are compared with the numerical ones.
The model reproduces the evolution of Sop , from the first crack initiation
(about 20 kN) until yielding of reinforcement. However, it slightly under-
estimates the experimental behavior before yielding (F = 40 - 60 kN), in

35
 4
Exp : 3 tests
Num : simulation 1

Sum of crack openings [mm]

Num : simulation 2
3 Num : simulation 3

0
0 10 20 30 40 50 60 70
Applied load F [kN]

Figure 27: Evolution of Sop during loading, experimental and numerical comparison

900
Exp : 3 tests
Maximum crack opening [mm]

Num : simulation 1
720 Num : simulation 2
Num : simulation 3
540

360

180

0
10 20 30 40 50 60
Applied load F [kN]

Figure 28: Evolution of wmax during loading, experimental and numerical comparison

accordance with the slightly over-estimated numerical stiffness in the same

phase (Figure 25).

36
 The evolution of wmax is also under-estimated. Theoretically, each crack
opening must have the same value. The average crack opening is consequently
equal to the maximum crack opening. This point is numerically checked,
thanks to a ”perfect” uniaxial loading and homogeneous bond properties
over the length of the structure. Nevertheless, the crack opening field is ex-
perimentally more heterogeneous, with a larger value of the maximum crack
opening. This observation can be related to the experimental heterogeneity
of the concrete, which leads to heterogeneous bond properties.
An other criterion can also be defined as the sum of cubic crack openings:
X
3
Sop = (wi3 ) (15)
i

It couples the number of cracks and their respective opening. This criterion
increases the participation of large cracks, compared to smaller one. It can
also be seen as durability criterion, since a transfer by permeation is related
to the cubic crack opening. The results are presented in Figure 29. The
numerical simulations are in accordance with the experimental evolution of
3
Sop . However, a low under-estimation can be noted. These results confirm
the previous conclusions on the distribution of the crack openings along the
specimens. The experimental distribution of crack openings is larger than
the distribution in numerical simulations.
The main difference between numerical and experimental behavior is the
tension stiffening between steel and concrete. For example, the bond stresses
at the interface can be compared after the second transverse crack (Figure
30). Experimentally and numerically, the bond stress τ is calculated as a

37
 Sum of cubic crack openings [mm3 ]
100

10−1

10−2

10−3
Exp : 3 tests
Num : simulation 1
10−4 Num : simulation 2
Num : simulation 3
10−5
10 20 30 40 50 60
Applied load F [kN]

3
Figure 29: Evolution of Sop during loading, experimental and numerical comparison

10
Shear stress τ (x) [MPa]

−5

Exp: Test 3
−10 Num: simulation 1

0 0.2 0.4 0.6 0.8 1

Embedded length [m]

Figure 30: Bond stresses along the steel-concrete interface after the second transverse
crack

38
function of the axial strain  of the steel rebar :
S Es d
τ (x) = (16)
2 π r dx
where S is the area of the steel cross section, Es is the steel Young’s mod-
ulus, r is the radius of the rebar (6 mm). Bond stresses numerically and
experimentally reverse on both sides of a given crack (positive on the left,
negative on the right). It can be noted that the tension stiffening is symmet-
ric on both sides of a crack and identical for both cracks in the numerical
simulation. The same conclusion cannot be made for the experimental re-
sults, due to the heterogeneous concrete properties at the interface (random
location of aggregates for example). That conclusion explains the measured
distribution of crack openings during loading. Hence, the main issue would
be to favor an heterogeneous crack distribution in the numerical simulation.
A proposition would consist to randomize the present periodic field of bond
properties, and mainly the second thereshold K2 and the softening parameter
H2 . That would locally change the transfer length between both materials
and consequently force the competition between cracks. However, a lot of
bond tests must be made in order to calibrate the random bond properties.

8. Conclusions

With the sake of durability, the modeling of reinforced concrete structures

is more and more focused on their cracking behavior. The cracking behavior
is mainly controlled by the embedded reinforcement and its associated bond
behavior.
New experimental results on the bond behavior are presented: pull-out
tests and tension tests on reinforced concrete ties. Displacement field mea-

39
surements (Digital Image Correlation) and strain field measurements of the
embedded rebar (optic sensors) provide local information on the progres-
sive degradation of the structure: concrete cracking, transfer length, tension
stiffening. The experimental results aim to valide the proposed numerical
approach.
A meso-scale reinforcement-concrete interface has been developed in this
communication. The associated bond model leads to a fine description of the
concrete degradation at the interface : local inclined cracks at the steel ribs,
slip between the two materials. Unlike existing approaches, the influence of
the steel ribs is implicitly taken into account. The local transfer of internal
forces from the steel rebar to concrete is representative of experimental mea-
surements. The bond model can be used for structural applications whose
cracking is highly dependent on the behavior of the steel-concrete interface:
4-point bending tests on reinforced concrete beams for example.
Specific attention is devoted to the influence of the concrete delayed
strains on the mechanical response of structures. The restraint of drying
shrinkage shows a significant effect on the cracking behavior. It has to be
taken into account in order to accurately design reinforced concrete struc-
tures.

9. Acknowledgement

The authors would like to thank Dr. Aghiad Khadour (IFSTTAR, France)
for his technical support in the use of the optical fiber sensors for strain
measurements.

40
10. References

[1] Z. P. Bažant, S. Sener, J.-K. Kim, Effect of cracking on drying perme-

ability and diffusivity of concrete, ACI Materials Journal 84 (5) (1987)
351–357.

[2] X. Jourdain, J.-B. Colliat, C. De Sa, F. Benboudjema, F. Gatuingt,

Upscaling permeability for fractured concrete: meso–macro numerical
approach coupled to strong discontinuities, International Journal for Nu-
merical and Analytical Methods in Geomechanics 38 (5) (2014) 536–550.

[3] A. Djerbi, S. Bonnet, A. Khelidj, V. Baroghel-Bouny, Influence of

traversing crack on chloride diffusion into concrete, Cement and Con-
crete Research 38 (6) (2008) 877–883.

[4] H. Wu, R. Gilbert, Modeling short-term tension stiffening in reinforced

concrete prisms using a continuum-based finite element model, Engi-
neering Structures 31 (10) (2009) 2380–2391.

[5] P. Soroushian, K.-B. Choi, Local bond of deformed bars with different
diameters in confined concrete, ACI Structural Journal 86 (2) (1989)
217–222.

[6] Y. Goto, Cracks formed in concrete around deformed tension bars, in:
ACI Journal Proceedings, Vol. 68, ACI, 1971, pp. 244–251.

[7] A. Ouglova, Y. Berthaud, F. Foct, M. François, F. Ragueneau, I. Petre-

Lazar, The influence of corrosion on bond properties between concrete
and reinforcement in concrete structures, Materials and Structures 41 (5)
(2008) 969–980.

41
 [8] A. Daoud, O. Maurel, C. Laborderie, 2d mesoscopic modelling of bar–
concrete bond, Engineering Structures 49 (2013) 696–706.

[9] R. Eligehausen, V. Bertero, E. Popov, Local bond stress-slip relation-

ships of deformed bars under generalized excitations, Report No. EERC
83/23, Earthquake Engineering Research Center, University of Califor-
nia, Berkeley, Calif., (1983).

[10] B. S. Hamad, Bond strength improvement of reinforcing bars with spe-

cially designed rib geometries, ACI Structural Journal 92 (1) (1995)
3–13.

[11] B. Richard, F. Ragueneau, C. Cremona, L. Adelaide, J. L. Tailhan,

A three-dimensional steel/concrete interface model including corrosion
effects, Engineering Fracture Mechanics 77 (6) (2010) 951–973.

[12] J. V. Cox, L. R. Herrmann, Development of a plasticity bond model

for steel reinforcement, Mechanics of Cohesive-frictional Materials 3 (2)
(1998) 155–180.

[13] H. Reinhardt, J. Blaauwendraad, E. Vos, Prediction of bond between

steel and concrete by numerical analysis, Matériaux et Construction
17 (4) (1984) 311–320.

[14] A. Leroux, Modèle multiaxial d’endommagement anisotrope: gestion

numérique de la rupture et application à la ruine de structures en béton
armé sous impacts, Ph.D. thesis, École normale supérieure de Cachan-
ENS Cachan (2012).

42
[15] B. Kolani, Comportement au jeune âge des structures en béton armé
à base de liants composés aux laitiers, Ph.D. thesis, Université de
Toulouse, Université Toulouse III-Paul Sabatier (2012).

[16] A. Casanova, L. Jason, L. Davenne, Bond slip model for the simulation
of reinforced concrete structures, Engineering Structures 39 (2012) 66–
78.

[17] T. Phan, J.-L. Tailhan, P. Rossi, P. Bressolette, F. Mezghani, Numerical

modeling of the rebar/concrete interface: case of the flat steel rebars,
Materials and structures 46 (6) (2013) 1011–1025.

[18] A. Ibrahimbegovic, A. Boulkertous, L. Davenne, D. Brancherie, Mod-

elling of reinforced-concrete structures providing crack-spacing based on
x-fem, ed-fem and novel operator split solution procedure, International
Journal for Numerical Methods in Engineering 83 (4) (2010) 452–481.

[19] A. Sellier, G. Casaux-Ginestet, L. Buffo-Lacarrière, X. Bourbon, Or-

thotropic damage coupled with localized crack reclosure processing: Part
II: Applications, Engineering Fracture Mechanics 97 (2013) 168–185.

[20] RILEM, Determination of the fracture energy of mortar and concrete

by means of three-point bend tests on notched beams, Materials and
Structures 18 (106) (1985) 285–290.

[21] F. Lagier, X. Jourdain, C. De Sa, F. Benboudjema, J.-B. Colliat, Nu-

merical strategies for prediction of drying cracks in heterogeneous ma-
terials: Comparison upon experimental results, Engineering Structures
33 (3) (2011) 920–931.

43
[22] Cast3m (2011) http://www-cast3m.cea.fr/.

[23] G. Besnard, F. Hild, S. Roux, Finite-element displacement fields anal-

ysis from digital images: application to Portevin–Le Châtelier bands,
Experimental Mechanics 46 (6) (2006) 789–803.

[24] J.-M. Henault, M. Quiertant, S. Delepine-Lesoille, J. Salin, G. Moreau,

F. Taillade, K. Benzarti, Quantitative strain measurement and crack
detection in RC structures using a truly distributed fiber optic sensing
system, Construction and Building Materials 37 (2012) 916–923.

[25] A. Khadour, F. Baby, A. Herrera, F. Taillade, P. Marchand, P. Rivillon,

A. Simon, M. Quiertant, F. Toutlemonde, Distributed strain monitor-
ing of reinforcement bars using optical fibers for shm, in: CONSEC13-
Seventh International Conference on Concrete under Severe Conditions–
Environment and Loading, 2013, p. 1620.

[26] J. Mazars, A description of micro-and macroscale damage of concrete

structures, Engineering Fracture Mechanics 25 (5) (1986) 729–737.

[27] P. H. Feenstra, Computational aspects of biaxial stress in plain and

reinforced concrete, Ph.D. thesis, TU Delft (1993).

[28] A. Hillerborg, M. Modéer, P.-E. Petersson, Analysis of crack formation

and crack growth in concrete by means of fracture mechanics and finite
elements, Cement and concrete research 6 (6) (1976) 773–781.

[29] G. Pijaudier-Cabot, J. Mazars, J. Pulikowski, Steel-concrete bond anal-

ysis with nonlocal continuous damage, Journal of Structural Engineering
117 (3) (1991) 862–882.

44
[30] B. H. Tran, Y. Berthaud, F. Ragueneau, Essais PIAF: Pour identifier
l’adhérence et le frottement, 18e Congrès Français de Mécanique (Greno-
ble 2007).

[31] G. Matheron, The intrinsic random functions and their applications,

Advances in applied probability (1973) 439–468.

[32] K. Haidar, G. Pijaudier-Cabot, J.-F. Dubé, A. Loukili, Correlation be-

tween the internal length, the fracture process zone and size effect in
model materials, Materials and structures 38 (2) (2005) 201–210.

[33] P. Mivelaz, Etanchéité des structures en béton armé. fuites au travers

d’un élément fissuré, Ph.D. thesis, EPFL Thesis [in French] (1996).

[34] G. Kaklauskas, V. Gribniak, D. Bacinskas, P. Vainiunas, Shrinkage influ-

ence on tension stiffening in concrete members, Engineering Structures
31 (6) (2009) 1305–1312.

[35] V. Gribniak, G. Kaklauskas, R. Kliukas, R. Jakubovskis, Shrinkage ef-

fect on short-term deformation behavior of reinforced concrete–when it
should not be neglected, Materials & Design 51 (2013) 1060–1070.

[36] J. A. Almudaiheem, W. Hansen, Effect of specimen size and shape on

drying shrinkage of concrete, ACI Materials Journal 84 (2) (1987) 130–
135.

[37] A. Hilaire, F. Benboudjema, A. Darquennes, Y. Berthaud, G. Nahas,

Modeling basic creep in concrete at early-age under compressive and
tensile loading, Nuclear Engineering and Design 269 (2014) 222–230.

45
[38] B. Gamble, L. Parrott, Creep of concrete in compression during drying
and wetting, Magazine of concrete research 30 (104) (1978) 129–138.

[39] M. Matallah, C. La Borderie, O. Maurel, A practical method to estimate

crack openings in concrete structures, International Journal for Numeri-
cal and Analytical Methods in Geomechanics 34 (15) (2010) 1615–1633.

46