A model for the prediction of bond strength is proposed with reference no mechanical model has been developed to interpret the
to corroded bars. Different confinement situations are considered, and test results; a key parameter is the confinement provided
the role of the interface pressure caused by bar expansion is studied. by transverse steel and concrete cover. Corresponding to
With reference to the ultimate bond stress (bond ultimate capacity), different confinement situations, the three following behaviors
the model is shown to reproduce some of the most documented test
results available in the literature very well. Initial bond increase
are observed:
and the final bond deterioration are also modeled, and the residual 1. Strong bond deterioration in the postcracking stage occurs
bond at corrosion levels beyond cover cracking is shown to remain for bars without any confinement stirrups (Clark and Saifullah
substantial, on the condition that sufficient confinement is provided by 1993; Auyeung, Balaguru, and Chung 2000; Mangat and
the stirrups. Finally, the model makes it possible to study bond Elgarf 1999) (Fig. 1(a));
mechanics under both mechanical and environmental actions, as
2. A sizeable residual bond strength though has been
required by the assessment of safety in existing deteriorated structures.
measured in specimens with transverse steel reinforcement
(Rodriguez, Ortega, and Garcia 1994; Cabrera and Ghoddoussi
Keywords: anchorage; bond; concrete; corrosion; cracking.
1992) (Fig.1(b)); and
INTRODUCTION 3. Bond strength may still be adequate after cover cracking
The effects of a severe environment on reinforced concrete in tests with transverse steel and efficient concrete cover
(RC) structures have been investigated thoroughly in the past (Al Sulaimani et al. 1990), and may even moderately increase
two decades, and now modeling the material and structural when skin reinforcement confining the bar is also provided
behavior with the primary objective of predicting the structural (Berra, Castellani, and Coronelli 1997) (Fig. 1(b)).
effects of different deterioration levels is becoming feasible. With regard to these issues, there is a still a question as to
In the case of bar corrosion, there is a strong interaction the extent to which the cracked cover can exert a sizeable con-
between the mechanical behavior of the reinforcement and finement on an anchored bar. To answer this, the development
the environment, since bar cross-section reductions, volume of the interface pressure associated with the confinement
changes in the reinforcement, concrete cover cracking, and around and along a bar needs further investigation.
modifications to the steel-concrete interface occur. Other effects that corrosion has on bond strength are a
When a pullout force is applied to a bar, equilibrium re- reduction of the friction coefficient and of the initial chemical
quires the development of bond stresses on the surface of the adhesion between the steel and the concrete, the reduction of
bar until bond breakdown takes place. In ribbed bars, bond the bearing area of the ribs, and the change of the actual rib
failure may be caused by the splitting of the cover or by the orientation (FIB 2000). These effects should all be imple-
shearing-off of the concrete keys between bar ribs (splitting mented into suitable models that predict bond strength at
and pullout failures, respectively). The former type of fail- different corrosion levels.
ure may be activated by corrosion product expansion, which
may partly or completely exhaust the hoop strength of the RESEARCH SIGNIFICANCE
cover, which then may split. The aim of this study is to investigate and model the
Low levels of corrosion that precede the cracking of the anchorage capacity of a corroded bar with reference to bar
cover (precracking stage) cause an increase in bond capacity position, cover thickness and splitting, and transverse reinforce-
of the bars because of the increase in friction between the ment. These problems have been widely investigated for
corroded bar and the concrete, and the pressure build-up nondeteriorated structures (FIB 2000), and design equations
around the bar. The appearance of cracks at the member are included in most codes, for example, ACI 318M-95
surface is considered by most authors to be a critical event (ACI Committee 318 1995). These equations, however,
for bond, because the confinement around the bar is impaired
must be modified to include corrosion effects to assess the
and bond efficiency decreases. Different trends of bond
anchorage capacity and judge whether bond failure may
deterioration for increasing corrosion levels beyond cover
become critical in a deteriorating structure. This represents
cracking (postcracking stage) have been measured in the
one of the mandatory verifications for the professional per-
numerous experimental studies devoted to the subject (Fig. 1).
forming the analysis of residual structural capacity.
These differences have been attributed to the different test
setup and specimens used (Table 1). The need has been declared
for a common standard test procedure that would be able to ACI Structural Journal, V. 99, No. 3, May-June 2002.
MS No. 01-089 received March 8, 2001, and reviewed under Institute publication
compare different experimental campaigns (FIB 2000). The policies. Copyright © 2002, American Concrete Institute. All rights reserved, includ-
variety of experimental specimens, however, allows for a ing the making of copies unless permission is obtained from the copyright proprietors.
Pertinent discussion will be published in the March-April 2003 ACI Structural Jour-
study of the influence of different parameters. Until now, nal if received by November 1, 2002.
BOND-STRENGTH MODEL
When an anchorage is loaded up to failure, the first radial
cracks develop in the direction where the cover thickness is
a minimum (primary splitting). These first cracks do not
correspond to a splitting failure, which is characterized by
the propagation of a new set of cracks that penetrate the cover
in a different direction, and the reaching of the bursting
capacity of the anchorage (final splitting). Both the primary
and final splitting cracks must be considered to determine
bond splitting failure. These definitions are included in the
Swedish Handbook for Concrete Design (Betonghandboken
1990), following the studies by Tepfers (1973) on anchorage
ultimate bond strength, and are herein extended to consider
the effects of corrosion.
In this paper, the corrosion-product expansion inside the
cavity occupied by the bar, and the ensuing crack formation,
are modelled first, that is, prior to the loading of the bar and
to the development of bond (as is generally the case in lab-
oratory tests). Conditions of uniform attack are considered, as
caused by the artificial corrosion processes used in various
experimental campaigns, to reproduce the effects produced
by carbonation in field conditions.
Bar-concrete pressure
Crack width calculation—Because the corrosion-product
expansion controls the opening of the splitting cracks around
the bar, a relation between the depth X of the corrosion attack
and the total crack width Wcr can be established, as proposed,
for instance, by Molina, Alonso, and Andrade (1993)
W cr = Σ i u i corr = 2 π t (1a)
max
Fc = β f ct c min b (3)
where
fct = concrete tensile strength;
Fig. 2—Corrosion depth X and bar expansion t. cmin = cover thickness; and
b = distance between two nodes.
As for β, this factor takes into account the nonuniform
distribution of the hoop stresses in the cover at the onset of
cracking (Tepfers 1973). No crack opening takes place until
the cracking force is exceeded, and thus the spring acts
initially as a fixed support. Once the support reaction in a
point exceeds the cracking force, the crack starts to open and
the spring is activated. The cohesive forces transmitted
across the crack model the restraining action of the cracks
(the splitting cracks develop a sizeable, cohesive strength
across their interface as long as the crack width is limited). A
Fig. 3—Corrosion-crack patterns: (a) corrosion cracks bilinear relation is herein adopted for evaluating Fc as a func-
smaller cover; and (b) corrosion cracks both sides. tion of crack opening (Maeda, Otani, and Aoyama 1995)
where max o
τ bu = k ( X ) p ( X ) + τ b ( X ) + µ ( X ) p corr ( X ) (10)
n = number of transverse ribs at section;
sr = rib spacing;
As for the mechanical properties of the oxides, the findings
Ar = rib area in the plane at right angles to bar axis; of various authors are summarized in FIB (2000). At low
cr = coefficient, depending on the rib shape and area; corrosion levels, oxides are firm and adhere to the bar surface;
δ = orientation of the ribs; at high corrosion levels, the interphase between the steel and
φ = friction angle between steel and concrete; and the concrete is flaky and weak. The friction coefficient is
fcoh = adhesion strength. close to 0.3, and the adhesion strength is close to the tensile
The first term to the right of Eq. (8) is the component of strength of concrete. Youlin (1992) carried out a series of
bond strength related to the splitting of the cover (splitting friction tests that showed a 10% reduction of the friction
component), while the second term τbo , a function of the coefficient for corrosion levels close to 100 µm, and adhesion-
interface cohesion, is defined as the non-splitting component. strength values close to 0.8 MPa. The friction coefficient and
the adhesion strength are given linear formulations, which
As a result, bond strength for splitting failure is a function are based on the above-mentioned values, but the formu-
of rib orientation and area, as well as of the friction and lations make it possible to introduce the corrosion depth
adhesion properties at the interface. As anticipated, this X when its value exceeds the value Xcr associated with
model is herein modified to include the effects of corrosion. cover through-cracking (X, Xcr in mm)
Because of bar oxidation, both the friction and the adhesive
stresses acting on the inclined rib faces undergo a quanti-
tative evolution, which may be described by changing the µ = tan φ = B – C ( X – X cr ) (11)
adhesion fcoh and the friction angle φ in Eq. (9). Seemingly,
the corrosion-induced modification of rib area Ar is modeled f coh = D – E ( X – X cr ) (12)
through the reduction of rib height hr. Finally, the transfer of
bond stresses along the bar surface between the ribs is
considered: this surface is now rusted, and therefore a An increase of friction at low levels of corrosion is not
pressure p(X) acts at the interface. By considering the friction considered in this paper because it occurs when X is close to
coefficient of rusted steel µ(X), an additional contribution to 10 to 15 µm, that is, when usual covers have not yet cracked.
bond capacity is calculated. To summarize, bond strength The changes in the geometric characteristics of the bar ribs
can be formulated as follows affect Ar and cr in Eq. (10). The coefficient cr is taken equal
to 0.8 for crescent-shaped ribs. The change in the rib area is Table 2—Bursting capacity of anchorage
considered reducing its height hr of a quantity equal to the max max
corrosion attack X. Specimen type p fin , MPa p res ,* MPa
The model parameters regarding the deteriorating inter- Rodriguez, Ortega, and 7.7 —
face friction and cohesion have been estimated from the Garcia, Type 1
experimental results by Rodriguez, Ortega, and Garcia Rodriguez, Ortega, and
7.0 —
Garcia, Type 2
(1994); a description of the tests, together with statistical
information on the results and the estimation of the parameters Rodriguez, Ortega, and
7.0 3.0
Garcia, Type 3
in Eq. (11) and (12), is provided in Appendix A.
Al Sulaimani et al. 8.5 —
A flowchart of the steps followed by the model to evaluate
Berra, Castellani, and
bond strength is given in Fig. 7. Coronelli
15.8 —
*Only for specimens where final cracking is reached during corrosion phase.
COMPARISON WITH TEST RESULTS
The bond-strength values calculated by means of Eq. (10) (Fig. 8(a)). The corrosion pressure is evaluated by the mod-
reproduce the results of four experimental studies very well, el, from primary cracking (Point A in Fig. 8(b)) up to
as shown in Fig. 8, for bond strength. Also, the role played cracking of the cover in bending (Point B in Fig. 8(b)).
by the corrosion-induced pressure (which cannot be directly The contribution given by the stirrups to the bond strength
measured) is highlighted. The test results consider three for all corrosion levels is considered in the bursting ca-
confinement situations: max
pacity of the anchorage p fin (Table 2).
1. Bars with no transverse steel confinement;
2. Bars with transverse steel confinement (either in a central The results from a second experimental study (Al Sulaimani
or corner position) (Fig. 3); and et al. 1990) regarding bars confined with stirrups (6 mm stir-
3. Bars with transverse steel confinement and skin re- rups with 50 mm spacing [Table 1]) have been analyzed. The
inforcement. bars were in a central position in a beam specimen, as shown
1. Bars with no transverse steel confinement—The results in Fig. 3(a), with the lowest cover-to-bar diameter ratio of 2;
by Rodriguez, Ortega, and Garcia (1994) for specimens this determines the cracking of this cover first (primary split-
with no stirrups (Type 3 specimens; refer to Appendix A) ting). The variation of corrosion pressure for increasing crack
show strong bond deterioration for all levels of corrosion opening is shown in Fig. 8(c). The bond strength beyond pri-
considered in the tests (Fig. 8(a)). The model shows an ini- mary cracking remains larger than the noncorroded value. The
tial bond-strength increase, followed by a sharp decrease, maximum bursting capacity of the anchorage is not reached
max
with the whole process being governed by the evolution of (pmax in Eq. (10) is the same as for the noncorroded bar p fin ),
the pressure p(X); the initial increase was not measured by and the corrosion pressure gives a supplementary contribution
Rodriguez, Ortega, and Garcia, who considered only cor- to bond strength via the friction mechanism (the third term in
rosion levels in the post-cracking stage. The increase, Eq. (10)). The cover cracking in bending does not take place
however, was shown in similar tests that also covered low- for the corrosion levels considered; the corrosion pressure ini-
er corrosions levels (Clark and Saifullah 1993). The resid- tially decreases after primary cracking (Point A in Fig. 8(c))
ual bond strength depends on the first and second terms in and then rises steadily. The model predicts the end of this in-
Eq.(10) because the corrosion-induced pressure becomes crease to occur beyond X = 300 microns due to cracking in
negligible. In particular, for corrosion levels beyond the bending. Bond deterioration occurs for high corrosion levels,
propagation of the cracks, that is, when the cover is split due to the degradation of the interface and of the ribs’ height
max and inclination.
along the whole length, p res is due to the confining effect of
the concrete still remaining around the bar at bar pullout (Ta- 3. Bars with transverse steel confinement and skin reinforce-
ble 2). ment—The test results by Berra, Castellani, and Coronelli
2. Bars with transverse steel confinement (either in a cen- (1997) (Fig. 8(d)) with heavy confinement (Table 1) (6 mm
tral or corner position)—The results by Rodriguez, Ortega, stirrups with 100 mm spacing and skin reinforcement; bar
and Garcia (1994) (Fig. 8(b)) for bars with transverse steel position as in Fig. 3(a)) have been modeled considering only
confinement (Type 1 specimens; refer to Appendix A) ex- a partial stiffness loss after cover cracking (α = 0.5 in
hibit a remarkable residual bond strength, as is also evi- Eq.(7)). Note that the pressure and associated bond strength
denced by comparison with the results without stirrups increase at all corrosion levels, after cover cracking. Bond