Construction and Building Materials 72 (2014) 434–443

Contents lists available at ScienceDirect

Construction and Building Materials

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

3D simulation of localized steel corrosion in chloride contaminated

reinforced concrete
Chong Cao ⇑,1
Department of Civil Engineering and Applied Mechanics, McGill University, 817 Sherbrooke Street West, Montréal, QC H3A 0C3, Canada

h i g h l i g h t s

 Pitting corrosion is examined through electrochemical corrosion analysis.

 Both circumferential and longitudinal macrocells contribute to localized corrosion.
 Macrocell corrosion controls pitting factor evolution.

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

Article history: Pitting corrosion of reinforcement often causes localized concrete cracking in chloride contaminated RC
Received 19 May 2014 structures. This paper presents a 3-D computational model to examine the localized corrosion of steel
Received in revised form 4 September 2014 rebar under chloride attack. Both circumferential and longitudinal macrocell corrosion have been
Accepted 21 September 2014
integrated into the numerical model. Modeling results show that macrocell corrosion controls the
Available online 11 October 2014
localized/pitting corrosion process. Highly localized corrosion pattern may occur due to extremely large
cathode/anode ratio. As corrosion front advances around reinforcement, macrocell corrosion may reduce
Keywords:
drastically and pitting factor also decreases. Parametric study indicates that chloride diffusion signifi-
Chloride
Localized corrosion
cantly influences pitting factor while longitudinal cathode/anode ratio controls localized macrocell
Pitting factor corrosion.
Macrocell Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction tribution shape governs the amount of corrosion products needed

Chloride-induced steel corrosion in Reinforced Concrete (RC)
structures normally takes localized/pitting pattern, as illustrated
in Fig. 1. Pitting corrosion causes localized iron dissolution by
forming intensively corroding anodes in chloride rich zones. The
self-propagating pitting cells are further aggravated by galvanic
coupling with nearby passive steel. Corrosion rusts accumulate
locally around the anodes and therefore concrete surrounding
active pitting zones suffers the most severe corrosion damage
[1–4]. The localized cracking pattern induced by non-uniform rust
expansion is schematically shown in Fig. 2a and b.
Service life prediction of RC bridges under pitting corrosion
attack requires accurate modeling of concrete deterioration pro-
cess. Corrosion rust distribution shape and electrochemical corro-
sion rate are the two key parameters for numerical simulation of
corrosion-induced concrete cover cracking: non-uniform rust dis-
⇑ Tel.: +1 514 398 2424.
E-mail address: chong.cao@mail.mcgill.ca
1
Formerly, MPhil Candidate and Visiting Scholar at with Hong Kong University of
Science and Technology, Clear Water Bay, Hong Kong, China.
to reach certain damage level, while localized corrosion current
density controls the rust expansion rate.
Empirical pitting factor a, representing the ratio between maxi-
mum pitting depth ppit (m) and uniform attack penetration x (m),
has been introduced to numerically describe both residual cross-
section of corroded rebar and non-uniform rust expansion shape
(Fig. 3). Experimental results indicate that pitting factor approxi-
mately falls in the range of 4–8 [5–8]. However, due to a lack of com-
prehensive understanding of localized corrosion mechanism in
concrete, previous quantitative depictions of pitting configuration
are mostly based on statistical characterization of maximum pit
depths by assuming a Gumbel distribution for pitting factor [9,10].
Moreover, empirical corrosion rate models are often adopted for
numerical modeling of concrete cracking and these models tend to
ignore the possible effects of localized macrocell corrosion on pit-
ting [11]. Nonetheless, numerical and experimental results have
shown that macrocell corrosion may significantly increase the iron
dissolution rate around localized anodes. The formation of pitting
corrosion are found to be mainly caused by macrocell corrosion cir-
culating around rebar circumference and along reinforcement
length [12–15].

http://dx.doi.org/10.1016/j.conbuildmat.2014.09.030
0950-0618/Ó 2014 Elsevier Ltd. All rights reserved.
 C. Cao / Construction and Building Materials 72 (2014) 434–443
In order to achieve a more precise simulation of pitting effects
on structural performance of RC bridges, both dynamic corrosion
rate and pitting shape should be accurately predicted. Although
there have been several attempts towards numerical modeling of
macrocell corrosion during the past two decades, existing
approaches are not able to simulate localized rust accumulation
along with pitting corrosion formation [16–19]. Currently, there
are few numerical methods for modeling the pitting corrosion evo-
lution in chloride contaminated RC structures.
The objective of this paper is to develop a 3-D computational
model for analysis of localized steel corrosion in chloride contam-
inated RC structures. The coupled micro- and macro-cell corrosion
process involved in typical chloride-induced pitting corrosion is
modeled through numerical polarization equations. Finite Element
Method (FEM) has been employed to numerically simulate corro-
sion potential distribution and corrosion current flow within con-
crete. Both circumferential and longitudinal macrocell corrosion
are simulated in order to capture the corrosion characteristic of
pitting corrosion in natural conditions. This numerical model has
been utilized to identify crucial factors influencing pitting corro-
sion pattern of reinforcing steel in concrete. Special attention has
been paid to the dynamic corrosion rate evolution and non-uni-
form rust expansion.
2. Corrosion model
In the case of chloride-induced steel corrosion in concrete, two
kinds of electrochemical corrosion cells may coexist during the
corrosion propagation stage: microcells leading to uniform/gener-
alized corrosion and macrocells causing pitting/localized corrosion
[13,20]. The coupled micro- and macro-cell corrosion behavior is
illustrated in Fig. 4a and b. For a better simulation of localized/
pitting corrosion in RC structures, the 3-D computational model
shall be developed by integrating both circumferential and longitu-
dinal macrocell corrosion.
2.1. Geometric configuration
The geometric configuration of a typical slab-on-girder bridge
under chloride attack is shown in Fig. 5a. In countries where deic-
ing salts are employed for winter maintenance of highway bridges,
bridge deck suffers the most severe corrosion damage and leakage
at expansion joints is a major cause of chloride attack for beam gir-
der. For bridges exposed to marine environment, beam girder may
also be exposed to airborne chlorides from seawater [1]. In all
cases, the corrosion problem can be numerically described by a
Fig. 1. Chloride-induced pitting corrosion in RC structures.
435
simplified model with concrete cover thickness of C (m), steel
diameter of D (m), slab/girder section thickness of W (m), bar spac-
ing of S (m) and slab/girder section length of L (m), as illustrated in
Fig. 5b. The length of localized corrosion zone is specified as La (m).
Specific points around the steel-concrete interface are refer-
enced by angle h. The upper surface of steel with shortest distance
to the concrete cover surface is referred to as ‘‘front’’ of the rebar
and assigned h = 0°, while the lower surface with longest distance
from the concrete cover surface is referred to as ‘‘back’’ of the rebar
and assigned h = ±180°. Thus, the chloride/oxygen transport dis-
tance X (m) around the rebar perimeter (0 6 h 6 180 ) shall be
expressed as
D
XðhÞ ¼ C þ ð1  cos hÞ ð1Þ
2
2.2. Chloride penetration
The transport process of chloride ions through porous concrete
cover can be numerically described by diffusion equation, written
as
@C Cl
¼ r  ðDCl rC Cl Þ ð2Þ
@t
where CCl (wt.% of cement) denotes chloride concentration in con-
crete, DCl (m2/s) represents diffusion coefficient of chloride, and t
(seconds) gives diffusion time. Different chloride exposure condi-
tions can be reflected by chloride concentration profiles on concrete
cover surface (C sCl ). Also, initial chloride concentration within con-
crete is denoted as C 0Cl .
The localized corrosion is confined to the area of steel where the
threshold chloride content (C crit Cl ) for depassivation is exceeded.
According to many experimental studies, the critical chloride con-
tent to trigger steel corrosion in concrete is dependent on several
factors including steel-concrete interface, binder type, steel poten-
tial, and PH of concrete pore solution, etc [21]. The threshold is
generally recognized among researchers as a distributed parame-
ter, which apparently arises because of the heterogeneous nature
of concrete [22]. Here, to simplify our analysis, the critical chloride
content in RC structures is chosen to be constant as 0.4% by weight
of cement.
The reinforcement serves as an impediment to inward chloride
migration and thus chloride ions build up at the top rebar surface.
This can lead to uneven chloride concentration around rebar
perimeter, as illustrated in Fig. 6. Previous experimental observa-
tions show that corrosion rusts tend to accumulate around rebar
circumference facing the concrete cover surface exposed to chlo-
ride attack [23,24]. Macrocell corrosion is responsible for promot-
ing localized corrosion, which will be demonstrated through
numerical analysis in this study.
2.3. Potential distribution
In order to simulate the potential distribution and electrical
current flow during the corrosion process, the concrete can be trea-
ted as a conductor with effective resistivity q (X m) [16–20]. The
potential distribution /(V) is governed by Laplace’s equation, writ-
ten as
 
1
r r/ ¼ 0 ð3Þ
q
According to Ohm’s law, the electrical current flux i (A/m2)
within the concrete electrolyte can be obtained as
1
i ¼  r/ ð4Þ
q
436 C. Cao / Construction and Building Materials 72 (2014) 434–443

H 2 O Cl − O 2

Anode

Fe(OH )2

OH − Fe + + OH −
− −
e e

Cathode Cathode
Concrete
Cover Protective layer
(PH > 13)

(a)

Cl −
H 2O O2 H 2O O2 H 2O O2

Concrete
OH − OH − Cover
(
(PH > 13))
Fe(OH )2
Protective
e− Fe + + Fe + + e− Layer

Cathode Anode Cathode

(b)
Fig. 2. Localized cracking pattern. (a) Around rebar circumference and (b) along rebar length, (Anode: Fe ? Fe++ + 2e; Cathode: O2 + 2H2O + 4e ? 4OH).

Fig. 3. Geometry of corroded steel bar.

For the boundary condition at the steel-concrete interface, the where imac (A/m2) denotes macrocell corrosion current density at
macrocell current flux must equal the total current flux passing active/passive zone, and n is the unit vector normal to the boundary
through the interface, expressed as surface.
  For other boundary conditions, the zero current flux through
1 
imac ¼ n  r/  ð5Þ concrete surface should be specified as
q interface
 C. Cao / Construction and Building Materials 72 (2014) 434–443 437

 
1
n r/ ¼ 0 ð6Þ
q
Macrocell

2.4. Numerical polarization

A

Microcell For chloride-induced corrosion of steel embedded in concrete,

three controlling mechanisms have been classified: chloride con-
A C trol, oxygen control and resistivity control [20]. Chloride concen-
tration controls the anodic polarization of reinforcing steel and
therefore dominates the corrosion initiation process. Oxygen trans-
port controls the cathodic polarization of reinforcing steel and is
often regarded as the most critical controlling mechanism in corro-
sion propagation stage. The oxygen diffusion process through con-
(a) crete can be numerically characterized by a limiting current
density iL (A/m2). Concrete resistivity mainly affects macrocell cur-
rent circulating between anodes and cathodes.
The anodes are under activation polarization (chloride control)
only while the cathodes are under combined activation polariza-
tion (chloride control) and concentration polarization (oxygen con-
Macrocell trol) [25], as illustrated in Fig. 7. Thus, we have the numerical
polarization equations as
iFe
/ ¼ /0Fe þ bFe log 0
ð7Þ
C C A C C i
|fflfflfflfflfflffl{zfflfflfflfflfflFe
ffl}
A C activation

for anodic reaction at the active zone (Fe ? Fe++ + 2e); and
iO2
Microcell / ¼ /0O2 þ bO2 log 0
2:303RT log iL ð8Þ
(b) iO2 zO2 F iL  iO2
|fflfflfflfflfflffl{zfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
activation concentration
Fig. 4. Coupled micro- and macro-cell corrosion modeling. (a) Around rebar
circumference and (b) along rebar length. for cathodic reaction at both the active and passive zones (O2 + 2H2
O + 4e ? 4OH). Here, iFe and iO2 gives corrosion current density
for anodic and cathodic reaction, respectively. F = 96,485 C/mol rep-

Deicing salts

Airborne chlorides

(a)

X Front Passive Active Passive C

0
– +

D
W 180
Back
La

S L
(b)
Fig. 5. Geometric configuration. (a) Slab-on-girder bridge and (b) simplified model.
438 C. Cao / Construction and Building Materials 72 (2014) 434–443

Concrete cover 20 mm
Steel diameter 16 mm
Bar spacing 100 mm
Section thickness 100 mm

Surface content 2 wt.% cement

Initial content 0 wt.% cement
Diffusion coefficient ×
Diffusion time 7 years
Corrosion front

Fig. 6. Rebar obstruction effect on chloride penetration (Chloride concentration: wt.% cement).

Table 1
Potential (V)

Polarization parameters.

Parameters Values
Anodic Tafel slope (bFe) 0.090 V/dec
Cathodic Tafel slope (bO2 ) 0.180 V/dec
Anodic equilibrium potential vs. SCE (/0Fe ) 0.780 V
ϕ i =i −i Cathodic equilibrium potential vs. SCE (/0O2 ) 0.160 V
Anodic exchange current density (i0Fe ) 3  104 A/m2
Cathodic exchange current density (i0O2 ) 1  105 A/m2
β

where DO2 (m2/s) represents the effective diffusivity of oxygen in

concrete, zO2 ¼ 4 is the valency of cathodic reaction,
β
C sO2 ¼ 0:268 mol=m3 denotes constant oxygen concentration at con-
ϕ crete cover surface, and X (m) gives the oxygen diffusion distance
from concrete cover surface (Fig. 5b).
In order to solve the governing Laplace’s equation, the non-
i
linear polarization behavior of reinforcing steel should be specified
at the steel-concrete interface. For computational efficiency, an
explicit expression for macrocell corrosion current density in terms
i i i i log i (A⁄m ) of corrosion potential shall be obtained. Thus, inverse relations of
Eqs. (7) and (8) are needed for corrosion modeling.
The inverse expression of Eq. (7) can be directly obtained via
Bulter-Volmer equation and the anodic corrosion current density
Fig. 7. Corrosion polarization. is given in terms of corrosion potential as
!
0 /  /0Fe
iFe ¼ iFe exp 2:303 ð10Þ
bFe
resents Faraday’s constant, R = 8.314 J/(mol K) indicates universal
gas constant and T (K) is absolute temperature of concrete. Table 1 Due to the existence of concentration polarization term, Eq. (8)
lists the values for both anodic and cathodic polarization parame- is highly non-linear and does not admit explicit expression of
ters at standard condition (T = 293 K). cathodic corrosion current density. According to Gulikers [26],
By assuming a linear distribution of oxygen concentration the cathodic polarization can be approximately described as
through the concrete cover, the limiting current density iL (A/m2)  
//0O
can be approximately calculated as 0
iO2 exp 2:303 bO 2
2
iO2 ¼   ð11Þ
DO2 zO2 FC sO2 i0O //0O
iL ¼ ð9Þ 1 þ iL2 exp 2:303 bO 2
X 2
 C. Cao / Construction and Building Materials 72 (2014) 434–443 439

The exact expression of cathodic polarization behavior and sure on surrounding concrete [7,27,28]. Therefore, the average
a a a
approximate description (iL = 0.001, 0.01, 0.1 A/m2) are compared corrosion current densities (imic , imac and itotal ) have been obtained
in Fig. 8. It is clearly demonstrated that Gulikers’ approximation from the electrochemical corrosion analysis. Assuming that icorr
can capture the asymptotic behavior of non-linear cathodic polar- (A/m2) is the total corrosion current density that flows through
ization curve. the corroding steel, the corrosion attack penetration (reinforcing
Therefore, the macrocell corrosion current density can be steel radius reduction) xp (m) around the rebar circumference
obtained as can be approximately estimated as [13]
a a a Rt
imac ¼ iFe  iO2 i ðh; tÞdt
t i corr
A
  xp ðh; tÞ ¼ ð16Þ
! 0 //0O zFe  F  qs
0 iO2 exp 2:303 bO 2
0 /  /Fe 2
¼ iFe exp 2:303    ð12Þ where h (0 6 h 6 180 ) represents the circumferential angle,
bFe i0O //0O
1 þ iL2 exp 2:303 bO 2 A = 55.85 g/mol gives the atomic weight of the iron corroded, ti
2
(seconds) denotes the corrosion initiation time, zFe = 2 indicates
for the active steel bar; and the valency of anodic reaction, and qs = 7800 kg/m3 is the density
  of steel. The corrosion rust may expand in volume and gradually fill
0 //0O the porous interfacial transition zone (ITZ) between steel and con-
iO2 exp 2:303 bO 2
p p p p 2 crete. By ignoring the compressive deformation of rust and steel,
imac ¼ iFe  iO2 ffi iO2 ¼   ð13Þ
i0O //0O we may obtain
1 þ iL2 exp 2:303 bO 2
2
tr ðh; tÞ ¼ rv  xp ðh; tÞ ð17Þ
for the passive steel bar. It should be mentioned that the macrocell for the rust layer thickness (m); and
current density at the active zone is defined as positive because the
ur ðh; tÞ ¼ ðr v  1Þ  xp ðh; tÞ ð18Þ
current flows out of the active bar. Accordingly, the macrocell cur-
rent density at the passive zone is supposed to be negative as the for the radial rust expansion (m). Here, rv represents the volumetric
current flows in. expansion ratio of rust and is assumed to be 2.96 throughout the
Also, for the active rebar surface, we have whole corrosion process [29].
  The cumulative corrosion level g (%) can be calculated as
0 //0O
iO2 exp 2:303 bO 2 Rp
a 2 2 xp ðh; tÞdðh  rÞ
imic ¼   ð14Þ gðtÞ ¼ 0
ð19Þ
i0O //0O
1 þ iL exp 2:303 bO 2
2
1
4
pD 2
2
where D (m) represents the steel diameter, and r (m) denotes the
for microcell corrosion current density; and radius of corroded steel cross-section.
!
a 0 /  /0Fe
itotal ¼ iFe exp 2:303 ð15Þ 3. Numerical analysis
bFe
The electrochemical corrosion processes at steel-concrete inter-
for total corrosion current density.
face are simulated by numerical polarization equations along with
experimentally obtained corrosion parameters. The durability
2.5. Rust expansion
properties of concrete cover are characterized by its chloride diffu-
sion coefficient, electrical resistivity, oxygen diffusion coefficient
In the present study, pitting factor is assumed to be constant
and thickness. In order to simulate corrosion potential distribution
along rebar anodes and longitudinally variable accumulation of
and corrosion current flow within concrete, we have adopted the
corrosion rust will not be accounted for. This simplification is
Finite Element Method (FEM) together with adaptive meshing in
based on the common approach that circumferentially non-uni-
COMSOL MULTIPHYSICS. The computational procedures for local-
form rust distribution be utilized for modeling of expansive pres-
ized corrosion analysis are briefly illustrated in Fig. 9.

3.1. Case study

The proposed case study is utilized to demonstrate the localized

Fig. 8. Comparison between exact cathodic polarization and approximate descrip-
tion proposed by Gulikers (iL = 0.001, 0.01, 0.1 A/m2).
steel corrosion mechanism in chloride contaminated RC structures.
The values for geometric parameters are listed in Table 2 and the
input material properties for numerical analysis are summarized
in Table 3. Selection of these parameters are described in the
author’s published work and the references cited therein [13].
Due to symmetry, only one quarter of the slab/girder section is
considered in the numerical analysis. As an illustrative example,
the 3D finite element model with corrosion potential distribution
for corrosion front h = 90° is shown in Fig. 10. In our FE corrosion
analysis, the geometry of corroded rebar is assumed to be approx-
imately the same with virgin steel, since the corrosion loss to cause
concrete cracking is relatively small compared to the original rebar
cross-section.
The numerical results of average corrosion current density at
a a a
the active zone (imic , imac and itotal ) are shown in Fig. 11. The model-
ing results show that the macrocell corrosion current density may
440 C. Cao / Construction and Building Materials 72 (2014) 434–443

Fig. 9. Flowchart of proposed computational model for localized corrosion analysis.

Table 2 Table 3
Geometric properties. Material properties.

Properties Values Properties Values

Concrete cover (C) 15 mm Chloride diffusion coefficient (DCl) 6  1012 m2/s
Steel diameter (D) 12 mm Surface chloride concentration (C sCl ) 2 wt.% of cement
Bar spacing (S) 60 mm Initial chloride concentration (C 0Cl ) 0 wt.% of cement
Slab/girder section thickness (W) 70 mm 0.4 wt.% of cement
Critical chloride concentration (C crit
Cl )
Slab/girder section length (L) 1000 mm
Oxygen diffusion coefficient (DO2 ) 6  109 m2/s
Localized corrosion length (La) 100 mm
Concrete resistivity (q) 200 X m

decrease drastically (100 lA/cm2 ? 3 lA/cm2) as a result of the
reduction in cathode/anode ratio (more than 1000 ? 10) along
with anode extension around rebar circumference. However, the
microcell current density may not change so much, since microcell
corrosion is supported by the local cathodic reaction in active zone.
Pitting factor (a = ppit/x) decreases along with the drop in macrocell
corrosion, as illustrated in Fig. 12. This indicates that pitting forma-
tion is mainly caused by localized macrocell corrosion.
The geometric configuration of corroded rebar with non-uni-
form rust accumulation around its perimeter is shown in Fig. 13
(pitting attack xp and rust expansion ur have been scaled up by
15 times to allow for a better view). The modeling results clearly
demonstrate the localized corrosion pattern. The pitting factor is
obtained as 6.8, 3.3 and 1.9 for corrosion front at 45°, 90° and
180°, respectively. These results are in reasonable agreement with
previous numerical study by Jang and Oh [7], who have assumed
the pitting factor to be 8, 4 and 2 for each case. About 3 years after
corrosion initiation, the pitting factor drops to 1.4 as the steel rebar
corrodes uniformly around its whole circumference.
3.2. Parametric study
As the main objective of this study is to find out the crucial fac-
tors influencing localized/pitting corrosion of reinforcing steel in
 C. Cao / Construction and Building Materials 72 (2014) 434–443
Fig. 10. 3D finite element model (Corrosion potential vs. SCE [V]).
Fig. 11. Corrosion rate evolution.
Fig. 12. Pitting factor evolution.
concrete, a parametric study is performed and the effects of chlo-
ride diffusion and longitudinal cathode/anode ratio on localized
corrosion are carefully examined. The chloride diffusion coefficient
(DCl) is chosen to be 0.5  1012, 1  1012 and 6  1012 m2/s. The
total corrosion length (L) is chosen to be constant as 1000 mm
441
while the localized corrosion length (La) is specified as 100, 500
and 1000 mm, representing the long-term anode extension process
along rebar length. This may result in the longitudinal cathode/
anode ratio (L/La) as 10:1, 2:1 and 1:1. For L = La case, the problem
is reduced into a 2-D model as the reinforcing steel is corroding
along its whole length and only circumferential macrocell corro-
sion is considered. All the other parameters for geometry and
material properties are assumed to be the same with the case
study (Tables 2 and 3).
Chloride diffusion coefficient significantly affects pitting factor.
The corrosion initiation time (for each h) increases drastically along
with reduction in chloride diffusion coefficient. Much more time is
required for corrosion front advancement around rebar circumfer-
ence in case of relatively lower chloride permeability. This may
allow for prolonged propagation time of localized corrosion. There-
fore, the smaller the chloride diffusion coefficient is, the larger is
the pitting factor induced by macrocell corrosion. The time-variant
evolutions of pitting factor for variable chloride diffusion coeffi-
cients are compared in Fig. 14.
Longitudinal cathode/anode ratio mainly influences localized
macrocell corrosion. The elongation of corroding steel reduces
stretching cathodes along reinforcement. Macrocell corrosion
decreases drastically as local anodes receive much less cathodic
support from longitudinal cathodes. This may lead to decelerated
localized corrosion process, as illustrated in Fig. 15. For the
extreme 2-D case (L = La), macrocell corrosion vanishes as the
whole rebar circumference starts to corrode and microcell corro-
sion may finally dominate the corrosion process [13].
The parametric analysis suggests that concrete properties sig-
nificantly influence pitting shape and pitting factor is highly
depended on corrosion level. Previous probabilistic descriptions
of localized corrosion [9,10] have not accurately accounted for dif-
ferent concrete properties and corrosion levels. In contrast, deter-
ministic electrochemical corrosion analysis allows for clear
demonstration of localized corrosion mechanism by considering
variable durability properties of reinforced concrete.
4. Discussion
For chloride-induced steel corrosion in concrete, the measured
general corrosion current density normally falls in the range of
0.1–10 lA/cm2 [1], which is equivalent to corrosion attack pene-
tration rate of 1.16–116 lm/year (ppit (lm/year) = 11.6  icorr
(lA/cm2)). In the case of localized macrocell corrosion, the steel
442 C. Cao / Construction and Building Materials 72 (2014) 434–443

1 month 5 months 18 months 36 months

θ = 45 θ = 90 θ = 180 θ = 180
ppit = 22.0 m ppit = 61.4 m ppit = 153.0 m ppit = 248.1 m
α = 6.8 α = 3 .3 α = 1 .9 α = 1 .4
η = 0.11 % η = 0.63 % η = 2.75 % η = 5.86 %

Fig. 13. Corroded rebar with non-uniform rust expansion (scale factor for xp and ur: 15).

Fig. 16. Comparison between experimental results and numerical prediction of

Fig. 14. Effects of chloride permeability on pitting factor. corrosion rust distribution profile for 2-D case (L = La).

pattern in concrete crack initiation and early propagation stages.

Fig. 15. Effects of longitudinal cathode/anode ratio on localized macrocell
corrosion.
corrosion rate may even reach as much as 1.16 mm/year for
100 lA/cm2. This indicates that the presence of pitting corrosion
can facilitate intensified cross-section loss at local anodes and thus
cause significant serviceability concern for RC bridges. Localized
pitting corrosion has been identified as the dominant corrosion
According to accelerated corrosion tests by Torres-Acosta and
Sagüés [30], the amount of localized corrosion attack needed to
crack the concrete cover is 30–272 lm, depending on specimen
dimensions (concrete cover C, reinforcing bar diameter D and ano-
dic length La). On the other hand, long-term corrosion experiments
by Vidal et al. [31] show that attack penetrations of 40–140 lm
may generate 0.2–2.0 mm cracks and the pitting factor is esti-
mated at a = 8 according to Rodriguez et al. pitting model.
Here, it is necessary to mention that the effects of corrosion-
induced concrete cover longitudinal cracks on subsequent
corrosion process have not been considered in the present study.
Extensive cracking of concrete cover may alter oxygen/chloride/
moisture transport parameters and the electrochemical corrosion
process will be significantly affected [20]. Therefore, the proposed
localized corrosion model should be limited to the case before
severe cracking appears.
Moreover, experimental results show that concrete cracking is
usually accompanied by half or quarter corroded rebar
[23,24,27,28]. Corrosion may not spread around the whole circum-
ference of rebar until severe cracking occurs during the long-term
natural corrosion process, in which case longitudinal cracks may
transform localized/pitting corrosion into generalized/uniform
corrosion by substantially changing the steel-concrete interface
properties [32,33]. Consequently, pitting factor of 4–8 or even
larger are expected for use in concrete cracking simulation.
 C. Cao / Construction and Building Materials 72 (2014) 434–443
Another drawback of the proposed method is that the effects of
corrosion current on chloride ions movement have not been
considered. Since macrocell current is flowing out of anodes into
cathodes, Cl are electro-migrating to the localized anodes. Mean-
while, migration of H+ from active zone onto adjacent passive rebar
leads to localized acidification and the reduced pH will cause the
active zone to spread laterally. Multi-ions approach shall be
employed towards modeling transport of Cl and H+ under macro-
cell corrosion in future study.
The circumferentially non-uniform corrosion rust accumulation
for the 2-D case (L = La) can be verified against test data (Fig. 16),
for which experimental details have been explained in the author’s
previous work [13]. Regarding with the 3-D localized steel corro-
sion, there have been very few test results available in literature.
Further experimental study needs to characterize the non-homo-
geneous geometric properties of corroded reinforcement when
partial length and partial perimeter of steel is corroding.
5. Conclusions
This paper presents a numerical method for deterministic anal-
ysis of localized steel corrosion in concrete. Simulation of coupled
microcell and macrocell corrosion process around/along steel-
concrete interface has been achieved by employing Finite Element
modeling in COMSOL MULTIPHYSICS. Both circumferential and
longitudinal macrocell corrosion are integrated into the 3D compu-
tational model. Geometric characterization of corroded rebar as
well as rust distribution profile has been numerically obtained
through electrochemical corrosion analysis.
Modeling results show that macrocell corrosion is responsible
for pitting pattern formation. During the dynamic corrosion pro-
cess, macrocell corrosion may reduce drastically as corrosion front
advances around reinforcement. Pitting factor also drops together
with the decreasing macrocell corrosion current. After steel rebar
starts to corrode around its whole perimeter, circumferential
macrocell vanishes and longitudinal macrocell may dominate the
localized corrosion process.
Cathode/anode ratio has been found to be a crucial factor gov-
erning the localized macrocell corrosion. Highly localized corrosion
pattern may occur due to extremely large cathode/anode ratio.
Also, chloride diffusion coefficient has significant impact on pitting
factor evolution. Lower chloride permeability can facilitate local-
ized corrosion by slowing down corrosion anode extension process
around rebar circumference.
Potential applications of the proposed localized corrosion
model may include modeling of concrete cover cracking induced
by pitting corrosion. The non-uniform build-up of corrosion rust
may cause much higher concentrated expansive pressure around
local anodes and lead to accelerated concrete cracking. Compared
with existing empirical non-uniform rust expansion models for
localized corrosion, our proposed computational framework
provides valuable information for the evolution of both localized
corrosion rate and non-uniform rust expansion around rebar
perimeter.
Acknowledgements
This paper is written in memory of Prof Moe M S Cheung (1945-
2014), the late Chair Professor in the Department of Civil Engineer-
ing at Hong Kong University of Science and Technology. The author
had enjoyed happy research and study life with Prof Cheung in
HKUST for three years. Prof Cheung’s wisdom and kindness will
always be remembered by his students.
443
References
[1] Bertolini L, Elsener B, Pedeferri P, Polder R. Corrosion of steel in concrete:
prevention, diagnosis, repair. 2nd ed. Weinheim: WILEY-VCH Verlag GmbH &
Co. KGaA; 2013.
[2] Andrade C, Alonso C. Corrosion rate monitoring in the laboratory and on-site.
Constr Build Mater 1996;10(5):315–28.
[3] Angst U, Elsener B, Jamali A, Adey B. Concrete cover cracking owing to
reinforcement corrosion – theoretical considerations and practical experience.
Mater Corros 2012;63(12):1069–77.
[4] Apostolopoulos CA, Demis S, Papadakis VG. Chloride-induced corrosion of steel
reinforcement – mechanical performance and pit depth analysis. Constr Build
Mater 2013;38:139–46.
[5] Rodriguez J, Ortega LM, Casal J, Diez JM. Corrosion of reinforcement and service
life of concrete structures. Durab Build Mater Compon 1996;7(1):117–26.
[6] Val DV, Melchers RE. Reliability of deteriorating RC slab bridges. ASCE J Struct
Eng 1997;123(12):1638–44.
[7] Jang BS, Oh BH. Effects of non-uniform corrosion on the cracking and service
life of reinforced concrete structures. Cem Concr Res 2010;40(9):1441–50.
[8] González JA, Andrade C, Alonso C, Feliu S. Comparison of rates of general
corrosion and maximum pitting penetration on concrete embedded steel
reinforcement. Cem Concr Res 1995;25(2):257–64.
[9] Stewart MG. Spatial variability of pitting corrosion and its influence on
structural fragility and reliability of RC beams in flexure. Struct Safety
2004;26(4):453–70.
[10] Stewart MG. Mechanical behavior of pitting corrosion of flexural and shear
reinforcement and its effect on structural reliability of corroding RC beams.
Struct Safety 2009;31(1):19–30.
[11] Otieno M, Beushausen H, Alexander M. Prediction of corrosion rate in
reinforced concrete structures – a critical review and preliminary results.
Mater Corros 2012;63(9):777–90.
[12] Kim CY, Kim JK. Numerical analysis of localized steel corrosion in concrete.
Constr Build Mater 2008;22(6):1129–36.
[13] Cao C, Cheung MMS. Non-uniform rust expansion for chloride-induced pitting
corrosion in RC structures. Constr Build Mater 2014;51:75–81.
[14] Raupach M. Chloride-induced macrocell corrosion of steel in concrete –
theoretical background and practical consequences. Constr Build Mater
1996;10(5):329–38.
[15] Warkus J, Raupach M. Modeling of reinforcement corrosion – corrosion with
extensive cathodes. Mater Corros 2006;57(12):920–5.
[16] Balabanić G, Bićanić N, Ðureković A. Mathematical modeling of
electrochemical steel corrosion in concrete. ASCE J Eng Mech
1996;122(12):1113–22.
[17] Kranc SC, Sagüés AA. Detailed modeling of corrosion macrocells on steel
reinforcing in concrete. Corros Sci 2001;43(7):1355–72.
[18] Redaelli E, Bertolini L, Peelen W, Polder R. FEM-models for the propagation
period of chloride induced reinforcement corrosion. Mater Corros
2006;57(8):628–35.
[19] Warkus J, Raupach M. Modeling of reinforcement corrosion – geometrical
effects on macrocell corrosion. Mater Corros 2010;61(6):494–504.
[20] Cao C, Cheung MMS, Chan BYB. Modelling of interaction between corrosion-
induced concrete cover crack and steel corrosion rate. Corros Sci
2013;69:97–109.
[21] Angst U, Elsener B, Larsen CK, Vennesland Ø. Critical chloride content in
reinforced concrete – a review. Cem Concr Res 2009;39(12):1122–38.
[22] Hartt WH. Service life projection for chloride-exposed concrete reinforced
with black and corrosion-resistant bars. Corrosion 2012;68(8):754–61.
[23] Yuan YS, Ji YS. Modeling corroded section configuration of steel bar in concrete
structure. Constr Build Mater 2009;23(6):2461–6.
[24] Zhao YX, Hu BY, Yu J, Jin WL. Non-uniform distribution of rust layer around
steel bar in concrete. Corros Sci 2011;53(12):4300–8.
[25] Jones DA. Principles and Prevention of Corrosion. 2nd ed. New Jersey: Prentice-
Hall; 1996.
[26] Gulikers J. Numerical modelling of reinforcement corrosion in concrete. In:
Böhni H, editor. Corrosion in reinforced concrete structures. Florida: CRC
Press; 2005. p. 71–90.
[27] Zhao YX, Karimi AR, Wong HS, Hu BY, Buenfeld NR, Jin WL. Comparison of
uniform and non-uniform corrosion induced damage in reinforced concrete
based on a Gaussian description of the corrosion layer. Corros Sci
2011;53(9):2803–14.
[28] Tran KK, Nakamura H, Kawamura K, Kunieda M. Analysis of crack propagation
due to rebar corrosion using RBSM. Cem Concr Compos 2011;33(9):906–17.
[29] Suda K, Misra S, Motohashi K. Corrosion products of reinforcing bars
embedded in concrete. Corros Sci 1993;35(5–8):1543–9.
[30] Torres-Acosta AA, Sagüés AA. Concrete cracking by localized steel corrosion –
geometric effects. ACI Mater J 2004;101(6):501–7.
[31] Vidal T, Castel A, François R. Analyzing crack width to predict corrosion in
reinforced concrete. Cem Concr Res 2004;34(1):165–74.
[32] Zhang R, Castel A, François R. The corrosion pattern of reinforcement and its
influence on serviceability of reinforced concrete members in chloride
environment. Cem Concr Res 2009;39(11):1077–86.
[33] Zhang R, Castel A, François R. Concrete cover cracking with reinforcement
corrosion of RC beam during chloride-induced corrosion process. Cem Concr
Res 2010;40(3):415–25.