Professional Documents
Culture Documents
Karin Lundgren , Per Kettil , Kamyab Zandi Hanjari , Hendrik Schlune &
Armando Soto San Roman
To cite this article: Karin Lundgren , Per Kettil , Kamyab Zandi Hanjari , Hendrik Schlune
& Armando Soto San Roman (2012) Analytical model for the bond-slip behaviour of
corroded ribbed reinforcement, Structure and Infrastructure Engineering, 8:2, 157-169, DOI:
10.1080/15732470903446993
Download by: [Indian Institute of Technology Guwahati] Date: 24 June 2017, At: 06:07
Structure and Infrastructure Engineering
Vol. 8, No. 2, February 2012, 157169
Corrosion of reinforcement aects the bond mechanism between reinforcement and concrete, and thus the
anchorage. Reliable models describing this are needed especially for assessment of the load-carrying capacity of
existing structures. This paper presents an analytical one-dimensional model for bond-slip response of corroded
reinforcement. The proposed model is an extension of the bond-slip model given in the CEB-FIP Model Code 1990,
and is practically applicable for structural analyses to determine the load-carrying capacity of corroded structures.
Furthermore, the anchorage length needed to anchor the yield force is calculated from the bond slip, using the one-
dimensional bond-slip dierential equation. Results of the proposed model are compared with experimental results
as well as results from an advanced three-dimensional nite element model. The suggested model is shown to give
results that are consistent with the physical behaviour.
Keywords: corrosion; bond slip; concrete; reinforcement; analytical model
In previous work presented by Soto San Roman is assumed to be in the elastic range (for the purpose of
(2006), the detailed 3D model of Lundgren (2005) was analysis of the bond-slip behavior within the ancho-
used for a parameter study resulting in 1D bond-slip rage length of the bar) according to
relations for dierent cases and dierent amounts of
corrosion penetration. From the results, it was noted
that it was possible to obtain the bond-slip response of du
s Ee; e 2; 3
corroded reinforcement by shifting the bond-slip curve dx
(see later Figure 7) of uncorroded reinforcement along
the slip axis, see Schlune (2006) or Lundgren et al. where E is the Youngs modulus, e is the strain and u is
(2007). The shift along the slip axis depended on the the displacement of the bar. The bond stress is here
degree of corrosion. Thus, the level of corrosion assumed to follow an elasto-plastic law
corresponds to a certain amount of slip. This can be
explained by assuming that corrosion has an eect t Ds sp ; 4
similar to that of pulling of a bar, i.e. in both cases
splitting stresses are induced in the concrete. Based on jtj tb k 5
these ndings, the 1D bond-slip model given in the
CEB-FIP Model Code 1990, CEB (1993), was ex- where D is the bond stiness, s is the slip, sp is the
tended to include corroded reinforcement. plastic slip, and tb is the bond strength, which is a
The purpose of this paper is to present the function of the hardening parameter k. Here, the
extension of the 1D bond-slip model given in the deformation of the surrounding concrete is neglected,
CEB-FIP Model Code 1990 to include corroded ribbed and hence the displacement of the bar is equal to the
reinforcement. Furthermore, the eect of corrosion on slip:
the anchorage length is examined, and comparisons to
test results and more detailed analyses are made. The us 6
presentation is organised as follows:
The boundary conditions for the problem of pull-out
(1) The basic 1D bond-slip dierential equation is of a bar with length L with a prescribed displacement
stated. uL are
(2) The 1D bond-slip model given in the CEB-FIP
Model Code 1990 is reformulated into a s0 0; uL uL 7
plasticity model.
(3) The model is rened with respect to the degree The solution of the dierential equation gives the
of connement. deformation and stress along the bar as well as the
(4) The eect of corrosion is introduced into the pull-out force.
model.
(5) The properties of the proposed model are
discussed. Plasticity formulation of the bond-slip constitutive model
(6) The numerical implementation of the model is in CEB-FIP model code 1990
described. The actual shape of the 1D bond strength function tb
(7) The proposed 1D model is compared with a 3D depends on the properties, the geometry and stress
model and experimental results. distribution of both the surrounding concrete and the
(8) The work is concluded. reinforcement bar. One proposal for such a bond-slip
function that tries to take the most important of these
eects into account is given in the CEB-FIP Model
Code 1990, CEB (1993). Here it is slightly reformu-
1D bond-slip model for uncorroded and corroded ribbed
lated into the format of theory of plasticity, which
bars
later allows a convenient treatment of corroded
1D bond-slip dierential equation reinforcement. In addition, this makes the model
The equilibrium equation along a reinforcement applicable for reversed and cyclic loading conditions,
bar is as for example in the vicinity of developing exural
crack. However, the result of the model for reversed
p d2 ds
pdt0 1 and cyclic loading has not yet been veried against
4 dx test data.
where d is the rebar diameter, s is the stress in the rebar The rst part of the bond strength curve is slightly
and t is the bond stress. The stress in the reinforcement reformulated to give an initial nite stiness, while the
Structure and Infrastructure Engineering 159
rest of the equations completely follow the CEB-FIP unconned cases as well as an intermediate interpo-
Model Code 1990: lated curve as dened in the next paragraph.
!
k 2 s1
tb k 1 12 tmax ; 0k< 8 Interpolation between the cases conned and
s1 2 unconned
s1 s The CEB-FIP Model Code 1990 gives parameters for
tb k tmax ; k < s2 9 the extreme cases conned (i.e. ductile pull-out failure)
2 2
0 1 and unconned (i.e. brittle failure due to cover
k s2 s21 cracking induced by the radial tensile stress); see
tb k tmax tmax tf @ t
A;
Table 1. Required cover and amount of transverse
s3 Df s2 s21
reinforcement needed to justify an assumption about
s1 tf conned concrete is also given in the model code. In the
s2 k < s3 10
2 D following, a method used to interpolate for intermedi-
ate cases, which often occur in practice, is described.
tf
tb k tf ; s3 k 11
D
2tmax
D 12
s1
*fcc in MPa.
160 K. Lundgren et al.
The bond strength versus hardening parameter response of corroded reinforcement by shifting the
curve for the intermediate case is, although the bond-slip curve of uncorroded reinforcement along
physical failure mechanisms are dierent, assumed to the slip axis. The explanation is the exhaustion of the
be the weighted sum of the bond-slip curves for the connement resistance by splitting stresses and crack-
conned and unconned cases, according to ing induced by the corrosion pressure, which is an
eect similar to that produced by bar loading. This
tb kuncor tb;conf 1 kuncor tb;unconf 14 observation can conveniently be introduced in the
plasticity formulation of the CEB-FIP bond-slip model
An example of an interpolated curve is shown in by making the hardening parameter dependent on
Figure 2. The interpolation factor is determined by both the plastic slip sp and the corrosion penetration x
according to (with sp and x having the same units,
kc=d e.g. mm)
kuncor max 15
kAsw
k sp ax 16
where kc/d is a factor that depends on the cover to bar
diameter ratio, and kAsw is a factor that depends on the where the parameter a is assumed to be a constant.
amount of transverse reinforcement Asw/s (see deriva- This was suggested by Schlune (2006); see also
tion in Figure 4) according to Figure 3. Lundgren et al. (2007). This is a straightforward way
The amount of transverse reinforcement Asw/s that to introduce the negative eect corrosion has on bond.
is assumed to correspond to pull-out failure, see Schlune (2006) found that the constant a was around
derivation in Figure 4, is estimated from the assump- 8.1 for the analysed cases. This value was also chosen
tions that the transverse reinforcement yields when the in the analyses described in this paper. Thus, this
maximum bond stress for pull-out failure is obtained, means that a corrosion penetration of around 120 mm
and that the splitting stresses equal the bond stresses. is assumed to cause splitting stresses equivalent to
Transverse reinforcement should only be accounted for those caused by a plastic slip of 1 mm.
when it is ecient; therefore, the following two
conditions should be fullled:
Change of failure mode due to corrosion cracking
(1) A splitting crack will cross the transverse For corroding reinforcement, the failure mode can
reinforcement (may not be the case for e.g. change from pull-out to splitting failure, unless
horizontal splitting cracks between bars with sucient connement is provided by ecient trans-
small spacing). verse reinforcement. This change of failure mode
(2) A slip of the main bar will introduce substantial would not be included in the model if only Equation
strain in the transverse reinforcement (may not (16) was used. Instead, the change of failure mode is
be the case if the transverse reinforcement is not here accounted for by an interpolation factor kcor that
located outside the main bars in, for example, a depends on the ratio x/xcr, where x is the corrosion
slab). penetration and xcr is the corrosion penetration that
causes cover cracking. Thus, when the cover is cracked,
the remaining bond capacity depends only on the
transverse reinforcement. Before the cover is cracked,
Eect of corrosion the cover also contributes to the capacity. To get a
From the results presented by Soto San Roman (2006), numerically stable modelling, a linear decrease from
it was noted that it was possible to obtain the bond-slip the capacity including the eect of the cover to the
0:8 0:5
fee c 1:5 d
xcr 11 18
40 d 16
Comparison of results
Results of the proposed 1D model were compared with
results computed by the advanced 3D model as well as
experimental results reported in literature. Experimen-
tal results from Almusallam et al. (1996) and Fang
et al. (2004) were used. Note that the computed results
are based on material data from the experiments and
assumed model parameters, i.e. they are not calibrated
to t the data. Good bond conditions were assumed. It
can be noted that the parameters used in the bond-slip Figure 8. Dimensions and boundaries in the nite element
model from CEB-FIP model code which this work is analyses. Dimensions in mm.
Structure and Infrastructure Engineering 163
the corrosion process to varying corrosion levels. reinforcement both in the FE analyses and in the 1D
Thereafter, a deformation-controlled pull-out force model. It was almost constant for varying corrosion
was applied. The results used in this comparison are levels in the FE analyses, while it slowly decreased for
the consequent bond stress versus slip, where the bond increasing corrosion levels in the 1D model. The bond
stress has been calculated from the force by assuming capacity drop in the results of the 1D model for large
constant bond stress along the bar. For more details, corrosion penetrations (500 mm) is due to the fact that
see Lundgren (2007). corrosion alone causes a large eect on the hardening
To make sure that the proposed model can treat the parameter k; compare Equation (16). In conclusion,
most common cases regarding the connement around the 1D model corresponds well for low corrosion
the bar in a suciently good manner, the comparisons levels, but is conservative for larger corrosion penetra-
have been sub-divided into four dierent cases tions in this case. The same result is also seen when the
(following the structure given in Lundgren 2007): anchorage lengths from the dierent analyses are
compared.
(1) With transverse reinforcement; cover cracks at Fang et al. (2004) made concentric pull-out tests
uncorroded pull-out. with and without transverse reinforcement. The con-
(2) With transverse reinforcement; cover does not crete cover was 60 mm and the reinforcement bar
crack at uncorroded pull-out. diameter was 20 mm; thus the cover-to-diameter ratio
(3) Without transverse reinforcement; cover cracks was 3. The yield strength of the main bars was
at uncorroded pull-out. 521 MPa. In the tests including transverse reinforce-
(4) Without transverse reinforcement; cover does ment, 26 over an embedment length of 80 mm were
not crack at uncorroded pull-out. used. Results from these experiments are compared
with results of the 1D model in Figure 10. The drop in
An overview of all analyses is given in Table 2. bond capacity in the 1D model at a corrosion
penetration of 80 mm is caused by cover cracking.
Thereafter, the remaining capacity is mainly due to the
With transverse reinforcement; cover cracks at stirrups. No similar drop can be seen in the experi-
uncorroded pull-out mental results; thus the connement provided by the
In the nite element analyses with transverse reinforce- transverse reinforcement appears to be underestimated
ment and cover cracking at uncorroded pull-out, the with the 1D model. A corresponding increase in
concrete cover was 40 mm; thus the cover-to-diameter anchorage length can be seen in the results from the
ratio was 2.0. The results from these analyses are 1D model.
compared with results of the 1D model in Figure 9. In Lundgren (2005), it was concluded that for the
The drop in bond stress in the bond-slip curves from case with transverse reinforcement and cover cracking
the FE analyses is due to a splitting crack reaching the at uncorroded pull-out, corrosion causes a small bond
outer surface; however, the transverse reinforcement decrease, or does not inuence the bond capacity. The
kept the structure together, thus causing a ductile 1D model described here shows reductions in bond
behaviour for low corrosion levels. Regarding the capacity; how large the reduction is depends on the
bond capacity, it was about 15 MPa for uncorroded relation between the cover and the amount of
*Note:
1. With transverse reinforcement; cover cracks at uncorroded pull-out.
2. With transverse reinforcement; cover does not crack at uncorroded pull-out.
3. Without transverse reinforcement; cover cracks at uncorroded pull-out.
4. Without transverse reinforcement; cover does not crack at uncorroded pull-out.
164 K. Lundgren et al.
Figure 9. Comparison between results from FE analyses and 1D model for the case with c/d 2, transverse reinforcement and
cover cracking at uncorroded pull-out. (a) Bond-slip curves for dierent corrosion penetrations (in mm); (b) maximum bond
stress versus corrosion penetration; and (c) anchorage length needed to anchor the yield force versus corrosion penetration.
Figure 10. Comparison with experimental results by Fang et al. (2004), with c/d 3, ribbed bars and stirrups. (a) Bond-slip
curves for dierent corrosion penetrations (in mm); (b) maximum bond stress versus corrosion penetration; and (c) anchorage
length needed to anchor the yield force versus corrosion penetration.
transverse reinforcement. It can be concluded that the with a concrete cover of 80 mm, giving a cover-to-
1D model gave conservative results. diameter ratio of 4.0. The results from these analyses
are compared with results of the 1D model in
Figure 11. Both the FE analyses and the 1D model
With transverse reinforcement; cover does not crack at resulted in ductile bond-slip behaviour. Furthermore,
uncorroded pull-out the bond capacities decreased slightly and the ancho-
The FE analyses with transverse reinforcement and no rage length increased slightly when the cover was
cover cracking at uncorroded pull-out were carried out cracked; in the 1D model this occurred at a corrosion
Structure and Infrastructure Engineering 165
penetration of 100 mm, while in the FE analyses at to-diameter ratio of 3. Results from these experiments
around 270 mm. Again, it can be concluded that the 1D are compared with results of the 1D model in
model gave results that are consistent with the physical Figure 13. The bond-slip behaviour for uncorroded
behaviour, and are on the safe side. Since no tests with reinforcement is too ductile compared with the
transverse reinforcement and cover large enough to experimental results; however, already for low corro-
avoid cracking at uncorroded pull-out were found in sion levels the failure mode became brittle and
the literature, no comparisons for this case to corresponded better. The drop in bond capacity in
experimental results were made. the 1D model at a corrosion penetration of 80 mm is
due to cover cracking. In the experiments, the decrease
of bond capacity was more gradual.
Without transverse reinforcement; cover cracks at In Lundgren (2005), it was concluded that for the
uncorroded pull-out case without transverse reinforcement and where the
In axi-symmetric FE models for this situation, the cover cracks at uncorroded pull-out, bond capacity
concrete cover was 40 mm, i.e. similar to the corre- decreases already for low corrosion levels. Here it was
sponding case with transverse reinforcement. The shown that the 1D model also gave this type of results.
results from these analyses are shown in Figure 12.
In both FE and 1D model analyses, the bond-slip
response was brittle due to the lack of connement, Without transverse reinforcement; cover does not crack
and became even more brittle for increasing corrosion at uncorroded pull-out
levels. Also the bond capacity decreased with increas- In the FE analyses without transverse reinforcement
ing corrosion penetrations. As can be seen in Figure 12, and no cover cracking at uncorroded pull-out, corro-
the agreement between the 1D model and the FE sion alone cracked the cover at a corrosion penetration
analyses is rather good. One dierence is that for large around 215 mm. This can be compared with the 1D
corrosion penetrations, the bond capacity became model, which resulted in corrosion cracking at a
almost zero in the FE analyses, resulting in very large corrosion penetration of 98 mm. Corrosion cracking
anchorage length. Since the 1D model has a small caused a sharp drop in bond capacity in both types of
residual bond capacity, the anchorage length became analyses, as can be seen in Figure 14. Furthermore, the
large for deep corrosion penetrations, although not as bond-slip behaviour changes from ductile to brittle at
large as in the FE analysis results. corrosion cracking, also in both types of analyses. The
The concentric pull-out tests without transverse dierence in corrosion penetration at cracking is
reinforcement made by Fang et al. (2004) were here explained by the geometry of the modelled specimen
used for comparison. The test specimens had a cover- in the FE analysis (see Figure 8); the concrete outside
Figure 11. Comparison between results from FE analyses and 1D model, with c/d 4, transverse reinforcement, no cover
cracking at uncorroded pull-out. (a) Bond-slip curves for dierent corrosion penetrations (in mm); (b) maximum bond stress
versus corrosion penetration; and (c) anchorage length needed to anchor the yield force versus corrosion penetration.
166 K. Lundgren et al.
Figure 12. Comparison between results from FE analyses and 1D model, with c/d 2, without transverse reinforcement, cover
cracks at uncorroded pull-out. (a) Bond-slip curves for dierent corrosion penetrations (in mm); (b) maximum bond stress versus
corrosion penetration; and (c) anchorage length needed to anchor the yield force versus corrosion penetration.
Figure 13. Comparison with experimental results by Fang et al. (2004), with c/d 3, ribbed bars without stirrups. (a) Bond-slip
curves for dierent corrosion penetrations (in mm); (b) maximum bond stress versus corrosion penetration; and (c) anchorage
length needed to anchor the yield force versus corrosion penetration.
the bonded zone kept the structure together, and thus bond capacity became almost zero in the FE analyses,
delayed corrosion cracking. This eect is not included resulting in very large anchorage length. Since the 1D
in the 1D model. For the anchorage length, the model had a small residual bond capacity, the
behaviour is similar to the case without transverse anchorage length became large for deep corrosion
reinforcement and no cover cracking at uncorroded penetrations, although not as large as in the FE
pull-out. Thus, for large corrosion penetrations, the analysis results.
Structure and Infrastructure Engineering 167
Almusallam et al. (1996) have carried out cantilever bond tests were conducted, and the corrosion penetra-
bond tests with large concrete covers and no active tion was measured by the weight loss method. The
transverse reinforcement. Their specimens were yield strength of the bar was here assumed to be
152*254*279 mm3, and had a 12 mm diameter bar 600 MPa. Again, corrosion cracking caused a sharp
with 102 mm embedment length. They were subjected drop in bond capacity, both in experiments and in the
to accelerated corrosion at varying levels; thereafter 1D model; see Figure 15. Furthermore, the bond-slip
Figure 14. Comparison between results from FE analyses and 1D model, with c/d 4, without transverse reinforcement, no
cover cracking at uncorroded pull-out. (a) Bond-slip curves for dierent corrosion penetrations (in mm); (b) maximum bond
stress versus corrosion penetration; and (c) anchorage length needed to anchor the yield force versus corrosion penetration.
Figure 15. Comparison with experimental results from Almusallam et al. (1996) with c/d 5.3. (a) Bond-slip curves for
dierent corrosion penetrations (in mm); (b) maximum bond stress versus corrosion penetration; and (c) anchorage length needed
to anchor the yield force versus corrosion penetration.
168 K. Lundgren et al.
behaviour changed from ductile to brittle at corrosion has spalled o totally, it might not be safe to use
cracking, also both in experiments and in the 1D the model to estimate required anchorage
model. In the tests, a rst crack was noted at a lengths.
corrosion penetration of 140 mm, and at a corrosion . For large corrosion penetrations, e.g. x greater
penetration of 200 mm, a second crack reached the than 400500 microns, where the model may give
edge. In the 1D model, corrosion cracking occurred for non-conservative results, see point above, the
a lower corrosion penetration, around 90 mm, thus model may need renements to take into account
causing a decrease in bond capacity and an increase in interface deterioration such as rib reduction and
anchorage length. For large corrosion penetrations, subsequent change of the friction angle.
the anchorage length is slowly decreasing. This is
due to the area decrease of the reinforcement bar, Future work will concentrate on testing the model in
which decreases the yield force that should be applications to beams that fail due to bond. Further-
anchored. more, investigations concerning the anchorage length
In Lundgren (2005), it was concluded that for this for large corrosion penetrations are needed.
case, corrosion causes a small increase in bond
capacity until the cover cracks; thereafter the bond
capacity decreases abruptly. Also the ductility de-
creases after cover cracking. The suggested 1D model Notation
was shown to give results that are consistent with this Asw Area of transverse reinforcement
physical behaviour, and in most cases they are on the D Bond stiness
safe side. However, for large corrosion penetrations, E Youngs modulus
the calculated anchorage length might not be on the L Length
safe side. This is due to the residual bond capacity a Parameter relating the corrosion pene-
from the CEB-FIP Model Code 1990, which in the tration to the hardening parameter
present model will remain also for very large corrosion c Concrete cover
levels. It can be questioned whether that is really d Reinforcement bar diameter
reasonable. fcc Concrete compressive strength
kAsw Factor depending on the amount of
transverse reinforcement
Conclusions kcor Factor for corroded case
This paper has presented an analytical 1D model for kc/d Factor depending on the cover-to-bar-
bond-slip response of corroded reinforcement. The diameter ratio
proposed model is an extension of the 1D bond-slip kuncor Factor for uncorroded case
model given in the CEB-FIP Model Code 1990. The s Slip; distance between stirrups
model is practically applicable for structural analyses s 1, s 2, s 3 Slip values, model parameters
to determine the load-carrying capacity of corroded sp Plastic slip
structures. x Coordinate along reinforcement bar;
The capability of the model was examined by corrosion penetration
comparisons to experimental results as well as to xcr Corrosion penetration that causes
results from an advanced 3D nite element model. cover cracking
From this study, the following conclusions can be u Displacement of the bar
drawn: uL Prescribed displacement
e Strain
. The computed results, in terms of bond-slip k Hardening parameter
curves, maximum bond stress and anchorage s Stress in the rebar
length needed to anchor the yield force, give a t Bond stress
qualitatively reasonable response compared with tb Bond strength
the experiments; i.e. the results are consistent tb, conf. Bond strength for conned case
with the physical behaviour. tb,unconf. Bond strength for unconned case
. The suggested model gives results that are on the tf Remaining bond strength, model
safe side for most cases. It should be noted, parameter
however, that for large corrosion penetrations tmax Maximum bond strength, model
and no or small amount of transverse reinforce- parameter
ment, the calculated anchorage length might not tmax, conf. Maximum bond strength for conned
be on the safe side. Thus, if the concrete cover concrete
Structure and Infrastructure Engineering 169