This article was downloaded by: [University of Arizona]

On: 02 February 2015, At: 22:38

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Structure and Infrastructure Engineering:

Maintenance, Management, Life-Cycle Design and
Performance
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/nsie20

Capacity loss evaluation of reinforced concrete bridges

located in extreme chloride-laden environments
a a a
Azadeh Alipour , Behrouz Shafei & Masanobu S. Shinozuka
a
Department of Civil and Environmental Engineering , University of California , Irvine , USA
Published online: 04 Jan 2011.

To cite this article: Azadeh Alipour , Behrouz Shafei & Masanobu S. Shinozuka (2013) Capacity loss evaluation of reinforced
concrete bridges located in extreme chloride-laden environments, Structure and Infrastructure Engineering: Maintenance,
Management, Life-Cycle Design and Performance, 9:1, 8-27, DOI: 10.1080/15732479.2010.525243

To link to this article: http://dx.doi.org/10.1080/15732479.2010.525243

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
 Structure and Infrastructure Engineering
Vol. 9, No. 1, January 2013, 8–27

Capacity loss evaluation of reinforced concrete bridges located in extreme chloride-laden

environments
Azadeh Alipour*, Behrouz Shafei and Masanobu S. Shinozuka
Department of Civil and Environmental Engineering, University of California, Irvine, USA
(Received 23 April 2010; final version received 25 August 2010; accepted 13 September 2010; published online 4 January 2011)

This article provides a comprehensive procedure for the structural performance evaluation and life-cycle cost (LCC)
analysis of reinforced concrete highway bridges located in extreme chloride-laden environments. An integrated
computational methodology is developed to simulate the chloride intrusion and to estimate the corrosion initiation
time. The effects of various influential parameters on the chloride diffusion process are examined and the changes in
geometry and material properties of structural members are calculated over the entire life of the bridge. In order to
Downloaded by [University of Arizona] at 22:38 02 February 2015

evaluate the global structural degradation due to the corrosion mechanisms, an inventory of bridges with different
structural attributes are investigated. The extent of capacity loss is calculated using the moment-curvature and
nonlinear static (pushover) analysis. Results of this study are then utilised to find the LCC of bridges. Different
inspection and maintenance strategies are considered to minimise the total LCC, which includes the initial
construction cost, inspection and maintenance costs and service failure costs. The proposed approach indicates the
inspection and maintenance intervals in a way that the inspection and maintenance costs are optimised while the
safety of the bridge is ensured.
Keywords: reinforced concrete bridges; corrosion; structural performance; life-cycle cost; inspection and
maintenance strategies

highest chloride content near the surface with a

1. Introduction
From a long-term point of view, the durability of
reinforced concrete (RC) highway bridges is signifi-
cantly affected by the deterioration of their structural
members. When investigating the damaged bridges, the
deterioration caused by the corrosion of RC members
is usually found to be one of the main sources of
structural degradation which may eventually result in
the serviceability failure of bridges under service or
extreme loading conditions. An accurate estimation of
the extent of degradation during the structure’s life-
cycle provides both engineers and decision-makers
with valuable information which helps to ensure the
safety of bridges while reducing the associated costs.
Towards this goal, the current article focuses on the
corrosion process caused by the attack of chloride ions
and evaluates its effects on the life-cycle performance
and cost of RC bridges.
Chloride-induced corrosion is one of the deteriora-
tion mechanisms caused by the rapid intrusion of
chloride ions into the concrete. This mode of corrosion
is expected when the bridge is exposed to aggressive
environments (e.g. coastal environments or the appli-
cation of deicing salts). The penetration profile of
chloride ions in a RC member demonstrates the
decreasing trend towards the depth of the member.
The chloride transport mechanism in concrete is a
complex phenomenon that may occur in several forms,
such as ionic diffusion, capillary suction and permea-
tion. When the concentration of chloride ions in the
pore solution within the vicinity of reinforcing bars
becomes high enough to depassivate the protection
film of the reinforcement, the layers of rust start to
form on the reinforcing bar surface and the corrosion
of steel begins.
In this article, an integrated computational meth-
odology is developed to simulate the penetration of
chloride ions into the RC members. Through a
comprehensive study, the effects of various influential
parameters, such as water-to-cement ratio, ambient
temperature, relative humidity, concrete age, free
chloride content and binding capacity, are considered
to obtain a precise prediction of the chloride content at
different depths of RC members with the progression
of time. By comparing the chloride content values with
certain critical thresholds suggested in the literature,
the corrosion initiation time is estimated. After
corrosion initiation, the time-dependent characteristics
of corroded bridges are identified through the extent of
the cracking and spalling of the concrete cover,

*Corresponding author. Email: aalipour@uci.edu

ISSN 1573-2479 print/ISSN 1744-8980 online

Ó 2013 Taylor & Francis
http://dx.doi.org/10.1080/15732479.2010.525243
http://www.tandfonline.com
 Structure and Infrastructure Engineering
reduction of the steel bar cross-section area and
decrease in the yield strength of reinforcing bars.
Based on that, the remaining capacity of a group of
RC bridges with different structural attributes is
evaluated over the time using the moment-curvature
and nonlinear static (pushover) analysis. Furthermore,
the service life-cycle cost (LCC) of RC bridges is
carefully studied in order to find the maintenance
strategies which optimise the required inspection
intervals and repair costs
2. Chloride intrusion
The process of deterioration caused by the chloride-
induced corrosion is usually divided into two main
phases: initiation and propagation. As it can be seen in
Figure 1, no corrosion occurs during the initiation
Downloaded by [University of Arizona] at 22:38 02 February 2015
time. At this phase, only chloride ions gradually diffuse
through RC members towards the reinforcing bars.
After chloride concentration at the depth of reinfor-
cing bars reaches its critical level, the propagation
phase begins. This phase itself consists of two stages:
rust expansion and cracking. The structural degrada-
tion takes place after crack initiation.
Although no corrosion is seen during the initiation
period, it is essential to find when the chloride content
reaches the value at which the rust expansion and
cracking occurs. To determine the time to corrosion
initiation, different models have been proposed during
past three decades. The majority of mathematical
models in the literature are based on the Fick’s second
law (Tuutti 1982):
  
x
CCl ðx; tÞ ¼ Cs 1  erf pffiffiffiffiffiffiffiffiffi ð1Þ
2 DCl t
Figure 1. Schematic demonstration of the corrosion process
idealised to the initiation and propagation phases.
9
where CCl(x,t) is the chloride content at spatial
coordinate x and time t, Cs, the chloride content at
the surface of the structural member (at x ¼ 0), DCl,
the diffusion coefficient, and erf (.), the error function.
This model has been widely used in both deterministic
and probabilistic approaches to estimate the chloride
content level and corrosion initiation time. Chatterji
(1995) and Kong et al. (2002) questioned the accuracy
of this model mainly for two reasons: first, the chloride
penetration in the concrete is not governed by a linear
diffusion equation; second, the parameters of initial
and boundary conditions as well as the chloride
diffusion coefficient are assumed to be constant while
they vary spatially and temporally.
In order to improve Equation (1), Saetta et al.
(1993) developed a numerical approach using the
conventional diffusion theory (Fick’s first law) and
the mass conservation principal (Fick’s second law) to
model the chloride diffusion into the concrete. In their
proposed diffusion model, the coupled effects of
moisture and heat flows are also considered based on
the approach given in Bažant and Najjar (1972). Xi
and Bažant (1999) suggested a model for chloride
penetration in the saturated concrete, which takes into
account both the binding capacity and chloride
diffusivity of the concrete. Later, Kong et al. (2002)
used the Xi and Bažant’s model to examine the rate of
chloride ingress into the concrete by studying the
influence of water-to-cement ratio and curing time.
Ababneh et al. (2003) then used the work of Xi and
Bažant to develop a mathematical model for chloride
penetration in the non-saturated concrete. In the
present article, the authors have extended the previous
numerical models to comprehensively consider the
simultaneous effects of the most influential parameters,
such as chloride binding capacity, free chloride
content, ambient temperature and relative humidity
in addition to concrete aging.
2.1. Mathematical modelling of diffusion
It is generally assumed that diffusion is the dominant
mode of chloride intrusion into the concrete. Diffusion is
the motion of chloride ions within the pore solution
based on the Fick’s first law. The chloride ions
concentrated on the exposed concrete surface diffuse
into the cement matrix due to the ion concentration
gradient. The diffusion of chloride ions into the partially
saturated concrete can be described using the following
partial differential equation (Saetta et al. 1993):
JCl ¼ DCl we gradðCf Þ ð2Þ
where JCl is the flux of the chloride ions in a unitary
area in a unit of time (kg m72 of concrete), DCl, the
 10 A. Alipour et al.

chloride diffusion coefficient, called diffusivity (m2 reactions and it has no effects on the transfer of
s71), we, the evaporable water content (m3 of evapor- chloride ions. On the other hand, the evaporable water,
able water per m3 of concrete), and Cf, the free chloride which is identified as the water held in concrete pores,
concentration (kg m73 of pore solution). In Equation actively takes part in the diffusion process. According
(2), the negative sign indicates that the diffusion to Han (2007), the evaporable water content can be
process is in the opposite direction to that of increasing calculated as the summation of the capillary pore
the chloride concentration. Using Equation (2), the water, wc (m3 of capillary pore water per m3 of
chloride mass conservation relationship (i.e. Fick’s concrete), and gel pore water, wg (m3 of gel per m3 of
second law) can be expressed as below (Saetta et al. concrete), as below:
1993):
we ¼ wc þ wg ð7Þ
@Ct w 
¼ divðJCl Þ ¼ div½DCl we gradðCf Þ ð3Þ
@t we ¼  0:36a c=g ð8Þ
c w

where Ct is the total chloride concentration in the

unitary volume of porous body (kg m73 of concrete), wg ¼ ð0:18aÞc=g ð9Þ
w
Downloaded by [University of Arizona] at 22:38 02 February 2015

and t, time (s). The total chloride refers to the acid-

soluble chloride in the concrete, which is the summa- where c is the cement content, w/c, the water-to-cement
tion of free chlorides and bound chlorides. The ratio, gw, the water density, and a, the degree of
relationship between the total, Ct, free, Cf, and bound, hydration. From Equations (7)–(9), the amount of
Cb, chloride content in the non-saturated concrete is as evaporable pore water can be estimated for any
follows: ordinary Portland cement (OPC) concrete. Consider-
ing different water-to-cement ratios, Figure 2 shows
Ct ¼ Cb þ we Cf ð4Þ the change in the amount of evaporable water content
over the time. It is assumed here that the cement
By substituting Equation (4) in Equation (3), the content is equal to 350 kg m73 and the temperature is
relationship for the diffusion of free chloride ions is 238C. As expected, the evaporable water content
obtained as: increases in the concretes with the higher levels of the
water-to-cement ratio. It can be also seen that the
@Cf amount of evaporable water decreases during the aging
¼ div½Da gradðCf Þ ð5Þ
@t process of concrete until it reaches a constant level
within less than 100 days. As a case in point, the
where Da is the apparent diffusion coefficient. This evaporable water content for a w/c of 0.5 experiences
coefficient is influenced by the adsorption phenomena no change when it reaches 0.136 (13.6%) after only 60
and can be expressed as: days.

1
Da ¼ F1 ðCb ÞDCl ¼   DCl ð6Þ
1 þ ð1=we Þ @Cb=@Cf

From Equation (6), it can be understood that the

effects of three parameters should be studied in order
to determine the value of apparent diffusion coefficient.
These parameters include: evaporable water content,
we, chloride binding capacity, @Cb =@Cf , and chloride
diffusion coefficient, DCl. These parameters are dis-
cussed here to provide the reader with more details on
the required assumptions and equations for the
purpose of accurate modelling of the diffusion process.

2.1.1. Evaporable water content

The water in the concrete is composed of evaporable
water, we, and non-evaporable water, wne. The non- Figure 2. Change in the evaporable water content for a
evaporable water is produced because of hydration range of water-to-cement ratios over the time.
 Structure and Infrastructure Engineering
2.1.2. Chloride binding capacity
As described earlier, there are two types of chloride in the
concrete: free chloride, Cf, which is dissolved in the pore
solution, and bound chloride, Cb, which is chemically and
physically bound to the cement hydrates. Between these
two, only free chlorides are responsible for initiating the
corrosion process (Tuutti 1982) although the effect of
chloride binding capacity should also be taken into
account. The chloride binding capacity characterises the
relationship between free and bound chloride ions in the
concrete. It is normally given at a constant temperature
and referred to as the binding isotherm.
According to Nilsson et al. (1993), by increasing
the chloride binding capacity, the rate of chloride ionic
transport in the concrete slows down since the amount
of available mobile ions (free chlorides) is reduced by
Downloaded by [University of Arizona] at 22:38 02 February 2015
binding mechanisms. Hence, this reduction can even-
tually delay the corrosion initiation. In order to
calculate the chloride binding capacity in Equation
(6), @Cb =@Cf , three different binding isotherms are
defined (Martin-Pèrez et al. 2001):
(1) Linear isotherm:
@Cb
Cb ¼ alin Cf ! ¼ alin ð10Þ
@Cf
where alin is a constant coefficient.
(2) Langmuir isotherm:
aL Cf @Cb aL
Cb ¼ ! ¼ ð11Þ
1 þ bL C f @Cf ð1 þ bL Cf Þ2
where aL and bL are the Langmuir binding
constants, which are determined by the regres-
sion analysis of experimental data. Tang and
Nilsson (1993) stated that the Langmuir
isotherm can provide acceptable results when
the concentration of free chloride is less than
1.773 kg m73 of pore solution.
(3) Freundlich isotherm:
b @Cb b
Cb ¼ aF Cf F ! ¼ aF bF Cf F1 ð12Þ
@Cf
where aF and bF are the Freundlich binding
constants estimated through the same fitting
procedure as explained for the Langmuir
binding constants. Tang and Nilsson (1993)
found that this relationship is appropriate
for chloride concentrations of more than
0.355 kg m73 of pore solution.
From a series of experimental tests, Martin-Pèrez
et al. (2001) estimated the coefficients of idealised
11
binding isotherms for concrete with an 8% evaporable
water content. In the present article, the same isotherm
constants have been used. It is also assumed that the
surface chloride content is for aggressive conditions
related to marine structures or bridge decks exposed to
deicing salts. Figure 3 shows the relationship between
the free and bound chloride contents using the three
mentioned isotherms. It can be seen that Langmuir and
Freundlich isotherms result in similar estimations,
while the linear isotherm underestimates the bound
chloride content when the free chloride content is less
than 65 kg m73 of pore solution.
Since only the free chloride content affects the
corrosion process, the change of chloride binding
capacity, @Cb =@Cf , with the free chloride content is
shown in Figure 4. It is evident that a more realistic
binding capacity can be obtained using Langmuir and
Freundlich isotherms, especially for the lower and
upper ranges of free chloride content. Referring to
Equation (6), the term F1(Cb), which relates the
apparent diffusion coefficient to the chloride diffusion
coefficient, can now be evaluated. This term takes into
account the chloride binding capacity and considers its
effects on the reduction of diffusion coefficient. It can
be understood from Figure 5 that the reduction effect
for the linear isotherm is constant (equal to 0.53 here)
and independent of the free chloride content. On the
other hand, Langmuir and Freundlich isotherms result
in reduction factors which increase from 0.20 to 0.90
over the range of free chloride content.
2.1.3. Chloride diffusion coefficient
The chloride diffusion coefficient, DCl, is influenced by
many parameters, such as concrete mix properties,
Figure 3. Relationship between free and bound chloride
contents using three binding isotherms, given required
binding constants.
 12 A. Alipour et al.
curing conditions and the chemical composition of
steel and concrete. Among all these parameters, it has
been approved that the change in concrete mix
properties, especially in the water-to-cement ratio,
has a significant influence on the chloride diffusion
coefficient (Bamforth and Price 1996, Papadakis et al.
1996, Snyder 2001). The level of water-to-cement ratio
directly affects both the capillary porosity and perme-
ability of the concrete, which may result in significant
changes in the diffusion rate of chloride ions.
A number of models have been developed to study
the effects of water-to-cement ratio on the chloride
diffusion coefficient. As a case in point, Collepardi
et al. (1970) found that when the water-to-cement ratio
of OPC concrete changes from 0.5 to 0.6, the DCl
increases from 1.7 to 3.3 (610712 m2 s71). In the other
Downloaded by [University of Arizona] at 22:38 02 February 2015
study conducted by Page et al. (1981), the effects of
Figure 4. Estimation of chloride binding capacity of the
concrete for a range of expected free chloride contents using
linear, Langmuir and Freundlich isotherms.
Figure 5. Effects of the free chloride content and chloride
binding capacity on the apparent chloride diffusion
coefficient of the concrete.
water-to-cement ratio as well as temperature on DCl
were examined. For a constant temperature of 258C,
they measured DCl as 2.6, 4.47 and 12.5 (610712 m2
s71) for the water-to-cement ratios of 0.4, 0.5 and 0.6,
respectively. The chloride diffusion coefficients ob-
tained from the aforementioned studies have been
compared in Figure 6. Review of all the available data
shows a similar trend of increase in DCl and indicates
that for the water-to-cement ratios in the common
range of 0.3–0.5, the chloride diffusion coefficients are
almost similar.
The logarithm of the diffusion coefficient can be
related to the logarithm of the water-to-cement ratio
by the linear equation below:
log DCl;ref ¼ a þ b logðw=cÞ ð13Þ
where DCl,ref is the reference diffusion coefficient which
does not still consider the effects of other parameters,
such as temperature, humidity, aging and free chloride
content. By taking into account the effects of these
parameters, the DCl,ref will be improved to the DCl in
the next section. In Equation (13), a and b are
empirical coefficients assumed equal to 710.6 and
1.9 according to the empirical model proposed by
Bamforth and Price (1996). This model will be used to
calculate the amount of DCl,ref. It can be seen that for a
water-to-cement ratio of 0.5, the reference diffusion
coefficient will be 6.73 (610712 m2 s71).
2.2. Parameters affecting the chloride diffusion
coefficient
According to Ahmad (2003), the parameters affecting
the chloride diffusion coefficient can be generally
Figure 6. Influence of water to cement ratio on the chloride
diffusion coefficient (a comparison among values suggested
by Collepardi et al. 1970, Page et al. 1981, Bamforth and
Price 1996, Papadakis et al. 1996, Snyder 2000).
 Structure and Infrastructure Engineering 13

divided into internal and external parameters. The

internal parameters include: the concrete mixture
proportion, quality of ingredients, casting and curing
conditions and chemical composition of materials.
These parameters influence the reference diffusion
coefficient, DCl,ref, and as discussed in the previous
section, the water-to-cement ratio is the most critical
parameter among them. On the other hand, the
external parameters, such as ambient temperature,
relative humidity, aging and free chloride content also
have significant effects on the diffusion coefficient. The
dependency of chloride diffusion coefficient, DCl, on
the external parameters can be considered using the
multifactor law:

DCl ¼ DCl;ref F2 ðTÞF3 ðhÞF4 ðte ÞF5 ðCf Þ ð14Þ

Figure 7. Effects of temperature variation on the chloride
Downloaded by [University of Arizona] at 22:38 02 February 2015

diffusion coefficient for a range of water-to- cement ratios.

where F2(T) considers the temperature parameter,
F3(h), the relative humidity, F4(te), the age of concrete,
and F5(Cf), the free chloride content. The significance
and contribution of these parameters are described in
the following.
2.2.1. Ambient temperature
Based on Arrhenius law, Saetta et al. (1993) and Xi
and Bažant (1999) suggested an influence factor to take
into account the ambient temperature in the calcula-
tion of the chloride diffusion coefficient. This factor
compares the current temperature, T, with a reference
temperature, Tref, by using Equation (15):
  
E 1 1 culated and used to estimate the chloride diffusion
F2 ðTÞ ¼ exp  ð15Þ
R Tref T
obtained for the last 15 years (1995–2009) from the
National Oceanic and Atmospheric Administration
(NOAA). The range of temperatures in the Los
Angeles area can be seen in Figure 8. From this
database, it is evident that the temperature has a
periodic trend over the year. Hence, a sinusoidal
function is fit to the data:
Tð KÞ ¼ 291  15 sin ð2pt=365Þ ð16Þ
where t is in days (0  t  365). From the sinusoidal
function fitted to the local temperature data, the F2(T)
at the specific location of the structure can be cal-
coefficient.

where R is the gas constant (KJ/(mol 8K)), and E, the

activation energy of diffusion process (KJ/mol).
According to Page et al. (1981), the values of E for a
cement paste made of OPC depend on the water-to-
cement ratio and can be considered as 41.8 + 4.0,
44.6 + 4.3 and 32.0 + 2.4 (KJ/mol) for w/c of 0.4, 0.5
and 0.6, respectively. In Equation (15), T and Tref are
expressed in 8K and the reference temperature is
usually assumed to equal 296 8K. As it can be seen in
Figure 7, when the temperature is below 238C, F2(T) is
less than one and has a reduction effect on the chloride
diffusion coefficient. On the other hand, when the
temperature passes the reference temperature, it
accelerates the chloride diffusion process by increasing
the frequency of thermal vibrations of the diffusant (Xi
and Bažant 1999).
In order to use Equation (15), temperature data
should be gathered within a specific region where the
structure is located. In the current study, the daily
temperature data of the Los Angeles area have been
2.2.2. Relative humidity
Based on the semi-empirical expressions proposed by
Bažant and Najjar (1972), the following function
(Equation (17)) was used by Saetta et al. (1993) in
order to consider the effects of relative humidity, h, on
the chloride diffusion coefficient.
1
F3 ðhÞ ¼   4  ð17Þ
1h
1 þ 1hc
where hc is the critical humidity level at which F3(h)
becomes equal to the average of its maximum and
minimum values (here assumed to equal 0.75). It is
seen from Equation (17) that consideration of relative
humidity always applies a reduction factor (less than
1.0) to the chloride diffusion coefficient.
Similar to the ambient temperature, the local
humidity information should be obtained for use in
 14 A. Alipour et al.

Figure 8. Measured daily temperature for the Los Angeles area from 1995 to 2009 obtained from NOAA (left) and a sinusoidal
Downloaded by [University of Arizona] at 22:38 02 February 2015

function fitted to the available temperature data (right).

Equation (17). The average monthly relative humidity

data for the Los Angeles area have been collected to
find out its annual trend. The relative humidity is
periodic in nature and is repeated throughout the
years. Hence, it can be simulated by a half-sinusoidal
function as below:

hð%Þ ¼ 0:65 þ 0:13 sin ðpt=365Þ ð18Þ

where t is in days (0  t  365). The measured

monthly relative humidity and corresponding sinusoi-
dal function are shown in Figure 9 for one year. The
same curve can be used for the following years as well.

2.2.3. Age of concrete
Concrete aging may cause a reduction in the chloride
diffusion coefficient. Due to the progress of hydration
reactions with time, the porosity of cement decreases.
This slows down the diffusion process, especially
during the initial life of the concrete. According to
Martin-Pèrez et al. (2001), the aging effect of the
concrete on the chloride diffusion coefficient, F4(te), is
introduced as:
t m
ref
F4 ðte Þ ¼ ð19Þ
t dependent free chloride content is directly related to
where tref is the reference time (equal to 28 day), and m,
the empirical age factor assumed to equal 0.04.
Figure 9. Measured monthly relative humidity for Los
Angeles area obtained from NOAA and a half sinusoidal
function fitted to available humidity data.
and Bažant 1999, Kong et al. 2002). The modification
factor can be expressed as:
F5 ðCf Þ ¼ 1  kðCf Þn ð20Þ
pffiffiffiffiffi
where k and n are empirical parameters, equal to 70
and 0.5, respectively. Based on Equation (20), the time-
the diffusion coefficient and should be updated during
the diffusion process at desired time steps.

2.2.4. Free chloride content 3. Estimation of corrosion initiation time

In spite of the consideration of free chloride content in The governing partial differential equation for free
the calculation of the chloride binding capacity, some chloride diffusion, given by Equation (5), cannot be
studies consider the effect of free chloride content solved without using numerical methods because of
individually on the chloride diffusion coefficient (Xi the dependence of Da on nonlinear, time-dependant
 Structure and Infrastructure Engineering
parameters. In this article, a finite difference algorithm
has been developed to solve Equation (5) as a
boundary value problem in different time steps
(Alipour 2010). This algorithm helps to study the
effects of various influential parameters on the chloride
intrusion process. As described in the earlier section,
these parameters include: evaporable water content
and chloride binding capacity, F1(Cb), ambient tem-
perature, F2(T), relative humidity, F3(h), age of
concrete, F4(te) and existing free chloride content,
F5(Cf). Considering all these parameters together in
each time step results in a more realistic analysis of the
chloride diffusion process and can provide an accurate
estimation of the corrosion initiation time.
The calculation time step for the developed finite
difference algorithm has been assumed equal to 1 day
and the algorithm continues until a specified chloride
Downloaded by [University of Arizona] at 22:38 02 February 2015
threshold value is reached at the depth of the
reinforcing bars. The time corresponding to this
threshold chloride content indicates the corrosion
initiation time. At the beginning of analysis (t ¼ 0),
the free chloride content within the depth of RC
member is considered to be zero from the assumption
that no chlorides have been added to the concrete
mixture.
The boundary values and initial conditions of
Equation (5) can be summarised as below:
For t ¼ 0 : Cf ¼ 0 at x 4 0
For t  0 : Cf ¼ Cs at x ¼ 0
where Cs is the surface chloride content. The surface
chloride content may depend on various parameters,
such as the composition of the concrete, location of the
structure, orientation of its surface, chloride concen-
tration in the environment and the general conditions
of exposure with regard to rain and wind (Bertolini
2008).
For the structures located in coastal zones, there
are different types of exposure to chloride ions. Some
structures are completely or partially submerged in the
sea water and some are within the tidal or splash zones.
There are also many structures located in a relatively
small distance from the coastline, which makes them
vulnerable to the sea salt spray (water-borne chloride
ions carried by the wind). Depending on which of the
above is the case, the amount of surface chloride
content may vary.
According to Glass and Buenfeld (2000) and
Martin-Pèrez et al. (2001), a surface chloride content
of 17.7 kg m73 (of pore solution) simulates the
complete submersion in seawater while this value
increases to 90 kg m73 (of pore solution) for tidal or
splash zones. The surface chloride content due to the
sea salt spray has smaller values and can be assumed
15
as a function of the distance from the coastline. Based
on a field study on 1158 bridges, McGee (1999)
suggested a range of 0.03–2.95 kg m73 (of concrete)
for surface chloride content at bridges located in the
coastal zones. Val (2004) also collected a set of data
from the Mediterranean coasts and indicated a surface
chloride content of 7 kg m73 (of concrete) for
structures directly placed on the coast. For the present
article, it is assumed that the highway bridges under
study are located at a close distance from the coast and
the sea salt spray is the only mode of exposure to
chloride ions. As a result, the surface chloride content
can be taken equal to 5 kg m73 (of concrete) on
average.
Using all aforementioned assumptions, the devel-
oped finite difference algorithm calculates the free
chloride content at different depths and time steps.
This algorithm is repeated for three different iso-
therms, including: linear, Langmuir and Freundlich.
Assuming the water-to-cement ratio to equal 0.5, the
free chloride content at the depth of 50 mm is shown in
Figure 10. After obtaining the free chloride content at
each time step, the bound and total chloride contents
can be calculated using Equations (10)–(12). The
change in free, bound and total chloride contents
during a 30-year period can be also seen in Figure 10
for three different isotherms.
As mentioned earlier, the chloride diffusion coeffi-
cient in Equation (6) is updated at each time step of the
developed algorithm. This update is because of the
effects of time-varying parameters, such as chloride
binding capacity, concrete age, temperature, humidity
and free chloride content. Figure 11 shows the chloride
diffusion coefficient for Langmuir and Freundlich
isotherms. The study of influential parameters shows
Figure 10. Change in free, bound, and total chloride
contents during a 30-year period using three different
binding isotherms.
 16 A. Alipour et al.

Figure 11. Chloride diffusion coefficient calculated using Langmuir and Freundlich isotherms.
Downloaded by [University of Arizona] at 22:38 02 February 2015

that the evaporable water content and chloride binding

capacity, F1(Cb), age of concrete, F4(te) and free
chloride content, F5(Cf), vary monotonically with
time. As a result, the strong fluctuation of the chloride
diffusion coefficient is mainly due to the seasonal
variation in the ambient temperature, F2(T), and
relative humidity, F3(h). In addition, the general
increasing trend of the diffusion coefficient and
F1(Cb) are found to be similar which implies the
significance of the chloride binding capacity of the
concrete.
The calculated free chloride content at different
time steps can be used to estimate the corrosion
initiation time, tinit. The corrosion initiation time is
determined as the time when the chloride concentra-
tion near the reinforcing bars reaches the threshold
chloride concentration. This means: Figure 12. Summary of data available in literature for the
critical chloride concentration required to initiate the
Cf ðtinit ; dc Þ ¼ Ccritical ð21Þ corrosion process.

where dc is the depth at which the reinforcing bars are

placed (usually equal to the concrete cover depth). In
Equation (21), the Ccritical is the threshold chloride
concentration causing depassivation of the concrete
protection film and initiation of the corrosion process.
There have been many research efforts during the past
three decades to determine an appropriate threshold
for the critical chloride content. It had been first
suggested that the critical value should be determined
by investigating the free chloride concentration, but
the study of Glass and Buenfeld (2000) on the chemical
aspects of chloride binding capacity showed that the
bound chloride should also be taken into account. As a
result, the threshold value is expected to represent the
total chloride content. Figure 12 demonstrates a
summary of data available in the literature regarding
the measured or suggested values for the critical
chloride concentration. For this study, the critical
chloride concentration is considered to be 1% of the
cement weight which is assumed to equal 350 kg m73.
The total chloride content profiles given in Figure
10 can be used to evaluate the corrosion initiation
time. Assuming the threshold value as 3.5 kg m73 (1%
of cement weight), the initiation time can be estimated
as 9.33, 12.66 and 10.40 years for the linear, Langmuir
and Freundlich isotherms, respectively. It is evident
that different chloride binding isotherms result in some
change in the estimation of the corrosion initiation
time, which indicates the importance of a reasonable
choice for the chloride binding model. Considering
only the Freundlich isotherm, as the middle range
 Structure and Infrastructure Engineering
isotherm, the initiation time for different cover depths
of 40, 50 and 60 mm would respectively be 7.23, 10.40
and 14.20 years (Figure 13). It can be seen that the
values of corrosion initiation time obtained from the
developed algorithm lie well within the range of 7–20
years, observed by Kong et al. (2002).
It is worth mentioning that the computational
model discussed in this section carefully studies the
diffusion process and estimates the corrosion initiation
time through a deterministic approach. Although the
developed model significantly reduces the uncertainties
associated with the simplifying assumptions, it does
not consider the uncertainties corresponding to the
model parameters as independent random variables.
Based on the existing literature, parameters affecting
the corrosion initiation time have high variability.
Hence, the effects of parameter variability are con-
Downloaded by [University of Arizona] at 22:38 02 February 2015
sidered in this section to improve the deterministic
model to a more reliable probabilistic approach.
Towards this goal, those parameters which have
the most influence on the corrosion initiation time are
studied. Assuming no variation in concrete cover
depth, selected random variables in the current study
include: reference chloride diffusion coefficient, DCl,ref,
surface chloride content, Cs and chloride threshold
value, Ccritical. As stated in Equation (13), the chloride
diffusion coefficient is a function of water-to-cement
ratio. A normal distribution function can be assigned
to the water-to-cement ratio with the mean value of 0.5
and coefficient of variation (COV) of 3–5%, according
to ACI (2009). Using Equation (13), a lognormal
distribution with the mean of 6.73 (610712 m2 s71)
and COV of 0.076 is obtained for the chloride diffusion
coefficient. The surface chloride content is also
Figure 13. Estimation of corrosion initiation time for
different cover depths of 40, 50 and 60 mm using
Freundlich isotherm.
17
assumed to have a lognormal distribution with the
mean of 5 kg m73 and COV of 0.50.
As mentioned earlier, the corrosion initiation time
is defined as the time required to reach the chloride
threshold value at a specific depth of the RC member.
With this definition, the chloride content at the level of
reinforcing bar can be considered as a demand
parameter, D, and the chloride threshold value for
corrosion initiation can be assumed as a resistance
parameter, R. Employing a probabilistic approach, the
demand parameter can be expressed as a probability
distribution function at each time step by taking into
account the uncertainties associated with influential
parameters, such as chloride diffusion coefficient and
surface chloride content. On the other hand, the
resistance parameter is an indicator of capacity of the
RC member before corrosion initiates and can be
identified by the chloride threshold value, Ccritical. In
the current study, a normal distribution function with
the mean of 3.5 kg m73 (equal to 1% of cement
weight) and COV of 0.20 has been assumed for the
chloride threshold value.
To predict the probability distribution function of
demand at different time steps, a set of 10,000 Monte
Carlo simulations is performed to calculate the total
chloride content, Ct, at the level of reinforcing bars. In
these simulations, the chloride diffusion coefficient and
surface chloride content are random variables with
lognormal distributions. Considering the probability
distribution of chloride content (demand) obtained
from Monte Carlo simulations at each time step, the
measure of risk in terms of probability of exceeding the
capacity level can be expressed as:
pf ¼ P½Ccritical < Ct 
Z 1  Z ct 
¼ fc critical ðccritical Þdccritical fct ðct Þdct
Z0 1 0
¼ Fccritical ðct Þfct ðct Þdct ð22Þ
0
where Fccritical(ct) is the cumulative distribution function
of Ccritical evaluated at ct. By repeating the procedure
of Equation (22) for different time intervals during the
life-cycle of bridge, the cumulative distribution func-
tion of corrosion initiation time is calculated. This
function is shown in Figure 14 and it can be seen that
the expected value of corrosion initiation time is 11.50
for the concrete cover depth of 50 mm. It is under-
stood that there is a slight difference between the
corrosion initiation time obtained from probabilistic
analysis (11.50 years) and the time calculated from the
deterministic approach (10.40 years). By increasing the
number of simulations, the corrosion initiation time
from the probabilistic analysis is expected to get closer
 18 A. Alipour et al.

of 10–25 mA cm72 has been suggested by Rodriguez

et al. (1994). This range corresponds to the high
reinforcement corrosion risk because it is larger than 1
mA cm72 (Andrade et al. 1993). In this study, the lower
bound of this range, 10 mA cm72, is assumed for the
icorr.
Substituting all mentioned assumptions in Equa-
tion (23), the consumed mass of steel per unit length of
reinforcing bar, DMloss (g cm71), subjected to the
corrosion for a time step, Dt (s), can be obtained from
Equation (25), as below:

2:893  109 pD0 icorr Dt ðt ¼ tinit Þ

DMloss ðtÞ ¼
2:893  109 pDred ðtÞicorr Dt ðt > tinit Þ
ð25Þ

Figure 14. Cumulative distribution function of corrosion

Downloaded by [University of Arizona] at 22:38 02 February 2015

initiation time for the cover depth of 50 mm. 4.1. Crack initiation time
To predict the time to crack initiation, the ratio of
to the deterministically estimated time. A more Mloss to the initial mass of steel, M0, should be
detailed discussion on the comparison between deter- calculated at each time step. The percentage of steel
ministic and probabilistic approaches can be found in mass loss, m(t), is given by:
Alipour et al. (2010).
Mloss ðtÞ
mðtÞ ¼ 100 ð26Þ
M0
4. Crack initiation and propagation
The RC members lose their strength and ductility The time corresponding to the point that the m(t)
because of crack initiation and propagation. Hence, equals to the mcritical is marked as the time to crack
the study of corrosion-induced cracks is necessary to initiation. The mcritical represents the threshold value of
estimate the extent of structural degradation over the steel loss for crack initiation and can be calculated
life-cycle of structures. In order to calculate the using the formula developed by El Maadawy and
corrosion crack initiation time, the Faraday’s Law is Soudki (2007), as below:
used. As it can be seen in Equation (23), the mass of
400
steel consumed over the time, Mloss, is related to the mcritical ¼ ½dc fct ðc þ 1 þ nÞðD0 þ 2d0 Þ
amount of current (A) that flows through the electro- D20 ðg  1ÞEeff
chemical corrosion cell. þ D0 d0 Eeff  ð27Þ

DMloss Z where D0 is the initial rebar diameter, d0, the thickness

¼ Icorr ð23Þ
Dt nF of porous zone around the rebar, and dc, the concrete
cover depth. In Equation 27 the rebar properties are
where Z is the atomic weight of the ion being dissolved represented by the Poisson’s ratio, n, and a coefficient
(i.e. Fe ¼ 55.85 gr), F, the Faraday’s constant for consideration of the mass density of steel and rust,
(96,500 A s), and n, the valence of reaction (usually g. The effects of concrete properties have also been
taken as 2). For a unit length of anodic bar, the current taken into account by the concrete’s effective elastic
due to corrosion, Icorr (A cm71), can be related to modulus, Eeff, and the tensile strength of concrete, fct.
the current per unit area of the reinforcing bar, icorr In Equation (27), the c is calculated as below:
(A cm72), using Equation (24):

ðD0 þ 2d0 Þ2
c¼ : ð28Þ
pD0 icorr ðt ¼ tinit Þ 2dc ðD0 þ dc Þ
Icorr ðtÞ ¼ ð24Þ
pDred ðtÞicorr ðt > tinit Þ
The Equation (27) has been obtained based on the
where D0 is the original diameter of steel bar, Dred(t), basic principles of elasticity which relate the radial
the reduced diameter during the corrosion process, and pressure to the concrete displacement. El Maadawy
tinit, the corrosion initiation time calculated in the and Soudki (2007) have made three assumptions in
previous section. For the icorr in Equation (24), a range developing this formulation: (1) the concrete around
 Structure and Infrastructure Engineering
the reinforcing bar is a thick-walled concrete cylinder;
(2) the concrete is a homogeneous material; (3) there is
a porous zone at the steel-to-concrete interface, that
the corrosion products must fill before they can expand
and create pressure on the concrete surrounding the
reinforcing bar.
4.2. Degradation of corroded RC members
According to AASHTO Guide Manual for the
condition evaluation of highway bridges (2005), it is
necessary to consider the modified geometry and
material properties of the structural members for the
condition assessment of deteriorated bridges. For this
purpose, this article specifically studies the reduction
rate of the diameter and yield strength of the steel
reinforcing bars in corroded RC members. Further-
Downloaded by [University of Arizona] at 22:38 02 February 2015
more, the effects of corrosion crack propagation are
considered for the calculation of confined concrete
properties.
Taking the steel mass density, rs, equivalent to 7.8
(g cm73), the change in the volume of corroded steel
per unit length, DVloss (cm3 cm71), can be calculated
as:
step-by-step analysis and it is estimated to be 51 days
DMloss 3:7091010 pD0 Dt ðt¼tinit Þ
DVloss ðtÞ¼ ¼
rs 3:7091010 pDred ðtÞDt ðt>tinit Þ
ð29Þ
The initial condition of the system is assumed such that
there is no corrosion in the reinforcing bars prior to
the corrosion initiation time, tinit. As a result, the
consumed mass and volume of the corroded steel can
be calculated from Equations 25 and 29, using the
original diameter of the reinforcing bar. However, for
the next steps of the corrosion process, the reduced
rebar diameter should be used to calculate the
consumed values. The reduced rebar diameter after
each time step of corrosion is calculated as:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4DVloss ðtÞ
Dred ðtÞ ¼ D20  : ð30Þ
p
The residual strength of corroded reinforcing bars was
investigated experimentally by Du et al. (2005a,b).
They conducted both accelerated and simulated
Table 1. Reduced mass diameter and yield strength of reinforcing bars at 5-year time intervals.
Timea (year) 0 5 10 15
(Mloss/M0)6100 0.00 6.38 12.24 18.10
D (mm) 35.80 34.75 33.61 32.46
fy/fy0 1.00 0.97 0.94 0.91
Note: aafter corrosion initiation.
19
corrosion tests on the steel bars embedded in concrete
and concluded that the strength of bars decreases
significantly with chloride penetration. Their test
results are in reasonable agreement with other studies,
such as Andrade et al. (1991), Lee et al. (1996) and
Morinaga (1996). Therefore, the empirical equation
below proposed by Du et al. (2005a,b) is used to
estimate the reduction in the yield strength of corroded
reinforcing bars.
fy ðtÞ ¼ ð1  0:005mðtÞÞfy0 ð31Þ
where fy(t) is the yield strength of corroded reinforce-
ment at each time step, fy0, the yield strength of non-
corroded reinforcement, and t, the time elapsed since
corrosion initiation. The reduced rebar cross-section
(from Equation (30)) and yield strength (from Equa-
tion 31) are given in Table 1 for different time intervals.
These values are used for the capacity evaluation of the
RC bridges in this study.
The time in which the concrete starts cracking was
defined earlier as the time when the steel loss reaches
the critical level (mcritical) obtained from Equation (27).
In the present article, this time is calculated through a
(0.14 year) for the structures under consideration. To
calculate the crack width after crack initiation, wcrack
(mm), the analytical equation proposed by Vidal et al.
(2004) can be used:
wcrack ðtÞ ¼ KðDAs ðtÞ  DA0 Þ ð32Þ
where DAs(t) is the steel loss of the rebar cross-section
during the corrosion process (mm2), DA0, the steel loss
of the cross section immediately after crack initiation
(mm2), and K, an empirical coefficient assumed to
equal 0.0577. The crack width of 0.3 mm has been
considered here as one of the first serviceability limits.
The time in which the crack width exceeds 0.3 mm
occurs after 117 days (0.32 year). Comparing the time
to crack initiation (0.14 year) as well as the time to
exceed the crack width of 0.3 mm (0.32 years) to the
time to corrosion initiation (10.40 years, using
Freundlich isotherm), it can be recognised that the
two former values are negligible when compared to the
latter one. Hence, since the crack initiation occurs
20 25 30 35 40 45 50
23.96 29.82 35.68 41.54 47.40 53.26 59.12
31.32 30.17 29.03 27.88 26.74 25.59 24.45
0.88 0.85 0.82 0.79 0.76 0.73 0.70
 20 A. Alipour et al.
shortly after the corrosion initiation time, this article
assumes the corrosion initiation time as the service-
ability threshold time. Furthermore, it is widely
accepted that the crack width of more than 1 mm
indicates the performance failure of the concrete cover.
The time corresponding to this crack width has also
been calculated equal to 542 days (1.48 year) after
corrosion initiation time. Since the capacity of
structures under study will be evaluated every 5 years
after the corrosion initiation time, it is assumed that
the concrete cover is destroyed at the time of the first
analysis interval.
5. Inventory of RC bridges under study
In order to evaluate the effects of the corrosion process
on the performance of RC bridges, an inventory of
Downloaded by [University of Arizona] at 22:38 02 February 2015
bridges has been developed for the current study. These
bridges represent the RC bridges located in the Los
Angeles area and can satisfy the current design require-
ments. To better quantify the structural degradation and
capacity loss due to the corrosion process, the bridges
with two equal spans are considered in the present paper.
The bridges under study have three different span
lengths of 15, 30 and 45 m, representing the short-,
medium- and long-span bridges, respectively. The deck
width is constant for all the bridges and it is designed for
four traffic lanes. The deck section consists of concrete
Figure 15. A schematic view of the two-span bridges under study.
box girders with slopped external girders. The total
concrete cross-section of deck is approximately 12 m2.
The bridges have only one interior bent with two circular
columns. Both columns of each bent have the same
diameter and height. The diameter of columns varies
from 1.3 to 1.9 m based on the bridge span length. Each
of short-, medium- and long-span categories also
includes bridges with three column height levels of 7.5,
10.0 and 12.5 m. Figure 15 illustrates a schematic
view of the bridges under study.
5.1. Modelling assumptions
The structural analysis of RC bridges has been carried
out with the OpenSees software (2009). As an object-
oriented finite element software, OpenSees is used to
assess the structural performance of the bridges under
corrosion scenarios defined in the earlier sections of
this article. The developed bridge models consist of
superstructure elements, columns, abutments and fixed
foundations.
5.1.1. Superstructure
The superstructure is modelled using linear-elastic
elements. In the ordinary bridges, the columns and
abutments are usually expected to experience the
inelastic behaviour. As a result, the superstructure
 Structure and Infrastructure Engineering
remains in the elastic range and no nonlinearity is
needed to be assigned to the superstructure elements.
Furthermore, since the concrete superstructure always
experiences some cracks due to loading conditions, the
cracked section properties should be calculated before
analysis. According to the Caltrans seismic design
criteria (SDC 2006), the flexural stiffness of concrete
box girder superstructures can be modified by a factor
of 0.50 to 0.75 in order to reflect the effects of this type
of cracking.
5.1.2. Column
The columns are modelled using inelastic beam-
column elements. In these elements, the plasticity has
been concentrated over a specified length of element
ends and it is used to model the plastic hinge zone of
Downloaded by [University of Arizona] at 22:38 02 February 2015
the columns (OpenSees 2009). The column cross-
section is modelled using the fibre section module
available in OpenSees, while the reinforcing bars of
concrete section are modelled in two layers considering
the expected steel properties. The stress–strain rela-
tionship of confined concrete is developed using
Mander’s model (Mander et al. 1988, Priestley et al.
1996) in order to capture the capacity and behaviour of
the bridge columns accurately.
5.1.3. Abutment
The model for abutment consists of a rigid element
with the length equal to the superstructure width,
connected to the superstructure centerline through a
rigid joint. At both ends of the rigid element, there are
three spring components that carry the nonlinear
response of the abutment in longitudinal, transverse
and vertical directions. In each direction, a zero-length
element is assigned with a nonlinear stiffness which is
dependent upon the material properties of the abut-
ment backfill. The stiffness and strength of these
springs are determined from Section 7.8.1 of Caltrans
SDC (2006). It should be mentioned that a gap element
has also been added to the abutment model in the
longitudinal direction to consider the 5 cm expansion
joint of the bridges.
5.2. Capacity evaluation of deteriorated bridges
The developed bridge models are used to evaluate the
effects of the corrosion process on the performance of
typical RC bridges. As described in the previous
sections, the properties of structural members change
after the corrosion initiation time and a gradual
capacity loss is expected for the entire bridge. The
deteriorated properties can be mainly found at the
concrete cover and the reinforcing bars. Because of
21
crack propagation, the concrete cover is destroyed
shortly after the crack initiation time, as calculated in
Section 4.2. Furthermore, considering Equations (30)
and (31), the diameter and yield strength of steel bars
decrease over the time according to Table 1. As it can
be seen from the moment-curvature analysis of the
column cross-section, the mentioned degradations
directly reduce the load bearing capacity of columns.
Since the columns have a key role in resisting lateral
loads, any reduction in column capacity causes a
significant capacity loss for the entire bridge system. In
the present article, the pushover analysis is used as a
measure to illustrate how the lateral load resistance of
a bridge drops after the corrosion initiation time.
The moment-curvature (M7j) analysis is used to
calculate the plastic moment capacity of ductile
concrete members based on expected material proper-
ties. This analysis derives the curvatures associated
with a range of moments for a concrete cross section
based on the principles of strain compatibility and the
equilibrium of forces. The M7j curve can be idealised
with a bilinear elastic perfectly-plastic curve to
estimate the yield and plastic curvature as well as the
moment capacity of column cross section. In the
current study, the M7j curves have been developed
for all bridge models considering the mentioned
changes in member properties. The moment-curvature
analysis has been repeated 11 times for each bridge in
order to take into account the structural degradation
during the life-cycle of bridge, which is over a period of
50 years. The analysis starts with the intact bridge
(before corrosion initiation time) and continues with
the deteriorated bridge in consecutive 5-year time
intervals.
Figure 16 shows the M7j curves for the short-
span bridge with the column diameter of 1.3 m and
column height of 7.5 m. It is evident that as the bridge
deteriorates, the M7j curve has a drop in the yield
and plastic moment capacities. The trend of change in
the plastic moment capacity of different bridge cases
can be seen in Figure 17 for a 50-year life-cycle after
the corrosion initiation time. Since the moment-
curvature analysis is independent of column height,
the plastic moment capacity is the same for all bridges
that have the same column cross-section. From Figure
17, it can be concluded that the plastic moment
capacity of short-, medium- and long-span bridges
under study decreases by 77, 85 and 92% after 50
years. It is clear that in the case of not taking
appropriate maintenance actions, the risk of structural
failure increases drastically.
The second measure to evaluate the lateral capacity
of the bridge is the nonlinear static (pushover) analysis.
In the pushover analysis, monotonically increasing
lateral forces are applied to the nonlinear model of the
 22 A. Alipour et al.
entire bridge until the displacement of the control node
at the deck level exceeds the target displacement
(assumed to equal 1.00 m). For the pushover analysis,
both material and geometric nonlinearities are con-
sidered and the gravity loads are applied to the
structure before the analysis starts. This procedure
evaluates the overall strength of bridge, typically
measured through the base shear capacity.
All the developed bridge models have been
analysed using the pushover analysis procedure. The
pushover curves for the short-span bridge with the
column diameter of 1.3 m and column height of 7.5 m
is shown in Figure 18. It is evident from this figure that
as the bridge deteriorates, the yield point happens at a
smaller displacement and the base shear capacity
decreases significantly. The base shear capacity is
Downloaded by [University of Arizona] at 22:38 02 February 2015
Figure 16. Moment-curvature plots for the short-span
bridge with the column diameter of 1.3 m and column
height of 7.5 m, deteriorated over 50 years.
Figure 17. Reduction in the plastic moment capacity of
short-, medium-, and long-span bridge cases for a 50-year
life-cycle after the corrosion initiation time.
defined here as 99% of the maximum base shear of
the bridge obtained from pushover analysis. The
changes in base shear capacity of all the bridge cases
have been shown in Figure 19. This figure summarises
the capacity loss of bridges with different structural
attributes. The drop in base shear capacity can be seen
in 10-year time intervals, which has an average value of
49, 51 and 54% for the short-, medium- and long-span
bridges over 50 years.
6. Life-cycle cost analysis
The LCC of a structure is defined as the total cost
of the structure from the beginning of planning
for construction to the end of its service life time.
Figure 18. Pushover curves for the short-span bridge with
the column diameter of 1.3 m and column height of 7.5 m,
deteriorated over 50 years.
Figure 19. Change in the base shear capacity of bridges
under study over 50 years.
 Structure and Infrastructure Engineering 23

The LCC analysis provides a framework that helps to
allocate appropriate resources for design, construction
and operation of the structure. The focus of the current
study is on the LCC analysis of RC bridges located in
extreme chloride-laden environments. The extent of
structural capacity loss due to the corrosion process
was shown in the previous section and it can be clearly
understood that the regular inspection and mainte-
nance of bridges are necessary in order to avoid any
structural failure. From the resources point of view, it
is also important to optimise the inspection and
maintenance schedules in a way that the total cost of
structure is minimised while the structure satisfies the
performance requirements.
The LCC of a bridge consists of a one-time initial
cost associated with design and construction of the
bridge and regular inspection and maintenance costs
Downloaded by [University of Arizona] at 22:38 02 February 2015
long-span bridges under study can be seen in Table 2.
Their inspection and maintenance costs as well as their
failure costs will be discussed in the upcoming sections.
6.1. Inspection and maintenance costs
The cost of inspection and maintenance is expected to
be incurred at regular time intervals, Dt. The inspection
cost is calculated as:
X
n
CIN ¼ SzðiDtÞ ð35Þ
i¼1
where S is the cost of each inspection, n, the number of
maintenance intervals, and z, the discount factor from
Equation (34). It can be seen that by increasing the
number of inspections, n, during the life-cycle of

necessary at certain time intervals. The general formula the bridge, the inspection cost rises accordingly. For
for LCC analysis can be expressed as follows: the bridges under study, the inspection cost is assumed to
equal 0.5% of the construction cost. The maintenance
LCC ¼ Cc þ ½CIN þ CM þ CuM  þ ½Csf þ Cusf  ð33Þ cost over a structure’s life span can be expressed as:
X
n
where Cc is the initial construction cost, CIN, the CM ¼ MzðiDtÞ ð36Þ
inspection cost, CM, the maintenance cost, CuM , the i¼1
user cost associated with the maintenance procedure,
Csf, the bridge service failure cost and Cusf , the user cost where M is the cost of maintenance activity in base
associated with the probable service failure. The year prices. The maintenance cost is usually assumed
inflation is also taken into account by dominating between 0.5% and 1.0% of the construction cost, but
future maintenance expenditures in base year prices. since it is expected that the maintenance cost increases
These recurrent maintenance costs are combined by as the bridge ages, it is assumed here that the
weighing them according to a discount factor that maintenance cost has a linear increasing trend from
takes into account the time value of the money. The 0.5 to 1.0% during the structure’s service life time.
discount factor, z(t), is defined as: In addition to the direct maintenance costs, the user
cost associated with the temporary closure of facilities
zðtÞ ¼ ð1 þ rÞt ð34Þ should also be considered. The actual user cost during
the regular maintenance work depends on the extent
where r is the discount rate indicating the expected and duration of service disruption (Chang and
market rate of return on an investment. The choice of Shinozuka 1996). This can be expressed as:
discount rate is often disputable in LCC analysis. In X
n
practice, the discount rate ranges from 2 to 8%. CuM ¼ tm bm uzðiDtÞ ð37Þ
Generally, choosing a high discount rate favours short i¼1
service life time while a low discount rate encourages a
longer service life time.
The initial cost of construction is assumed to be
the summation of the costs of the bridge components. Table 2. LCC of short-, medium-, and long-span bridges
under study, including initial construction cost, inspection
The cost of the deck is computed as the total area
and maintenance costs and service failure costs.
of the deck multiplied by a cost per unit deck area.
Caltrans Contract Cost Data (2008) suggests $380 per Bridge case Short-span Medium-span Long-span
m2 for deck construction cost. For bridge piers, the
Cc $361,151 $500,117 $639,083
construction cost includes the cost of steel work and CIN $122,557 $169,716 $216,874
concrete work. After calculating the total volumes of CM $154,744 $214,288 $273,832
the steel and concrete, the pier cost can be estimated by CuM $95,171 $95,171 $95,171
assuming $4 per kg of steel and $840 per m3 of concrete Csf $90,085 $124,749 $159,412
Cusf $45,471 $45,471 $45,471
according to Caltrans Contract Cost Data (2008). The Ctotal $869,179 $1,149,512 $1,429,843
average total construction cost of short-, medium- and
 24 A. Alipour et al.

where tm is the duration of maintenance activities, and

bm, the index of usage disruption (0  bm  1). For
example, if maintenance entails closure of one of the
two lanes of a bridge, bm would be 0.5. In Equation
(37), u is the unit user cost which depends on the
volume and type of traffic crossing the bridge as well as
the availability of convenient alternative routes. The
user cost also typically includes the increased costs
associated with travel delays and accidents. In the
current study, it is assumed that for the maintenance
programme, only 1/4 of bridge is closed at each period
of time for one week. Hence, the average hourly user
cost is calculated to be $8.75 per vehicle. The annual
traffic is assumed to be 100,000 h vehicle year71, which
makes the user cost of the maintenance programme
Figure 20. Probability of service failure due to the
equal to $875,000. It is also assumed here that the
corrosion process for various inspection intervals.
Downloaded by [University of Arizona] at 22:38 02 February 2015

inspection procedure is so short that it leads to

negligible disruptions on bridge traffic, and as a result,
it causes no user cost.
6.2. Service failure costs
The expected value for the service failure cost of a
bridge can be calculated using Equation (38):
X
n Cusf ¼
Csf ¼ Cf zðiDtÞDpf ðiÞ
i¼1
where Cf is the repair cost due to the service failure,
assumed to be equal to 20% of the construction cost.
In order to obtain the expected service failure cost, the
repair cost should be multiplied by the relevant
probability of failure during each time interval of
the bridge life-cycle. Since the current article studies
the effects of the corrosion process on LCC of
bridges, the probability of failure due to the corrosion
process between (i71)-th and i-th time intervals,
Dpf(i), has been calculated using a recursive formula
suggested by Val and Stewart (2003):
X
i1
Dpf ðiÞ ¼ pðiDtÞ  p½ði  1ÞDt þ Dpf ðjÞfp½ði  jÞDt
j¼1
 p½ði  j  1ÞDtg
where p is the cumulative distribution function for the
time of service failure (Figure 14). Since the crack
initiation and propagation time (calculated in Section
4.2) is small comparing to the corrosion initiation time,
the service failure is assumed to occur after the
corrosion initiation time. The values of Dpf(i) have
been calculated for different inspection intervals and
are shown in Figure 20 for the entire life-cycle of
the bridge, which is assumed to be 50 years after the
corrosion initiation time. It is evident that as the
number of inspections decreases, the probability of
failure due to the corrosion process increases. Accord-
ing to Chang and Shinozuka (1996), the expected user
cost associated with the probable service failure can be
calculated as below:
X
n
tsf bsf uzðiDtÞDpf ðiÞ ð40Þ
ð38Þ i¼1
where tsf is the duration of repair activities after failure,
and bsf, the usage disruption parameter due to the
service failure. In this study, it is assumed that it takes
one month for the bridge to be repaired and during this
time half of the bridge will be closed to traffic.
6.3. Optimised inspection and maintenance intervals
Based on the total construction cost of the short-,
medium- and long-span bridge cases, the inspection
and maintenance costs as well as the service failure
costs of all the bridges are calculated and the average
costs are summarised in Table 2. The values of Table 2
are obtained with the assumption of 4% annual
inflation rate and scheduling annual inspection and
maintenance programme. This table shows that the
ð39Þ user associated costs, CuM and Cusf , do not change for
different bridge types, which indicates that they are
independent of the construction cost.
In order to evaluate the effects of inspection and
maintenance intervals on the LCC of bridges, different
inspection and maintenance strategies have been
examined. In this study, it is assumed that the
inspection and maintenance intervals are the same
and can be scheduled every 1, 2, 3, 4 or 5 years. Table 3
indicates the average inspection, maintenance and
service failure costs for the medium-span bridge
inspected and maintained according to the proposed
 Structure and Infrastructure Engineering 25

Table 3. Effects of inspection/maintenance intervals on the life cycle cost of the medium-span bridge.

Interval Annual 2-year 3-year 4-year 5-year

Cc $500,117 $500,117 $500,117 $500,117 $500,117
CIN $169,716 $86,887 $54,368 $39,966 $32,798
CM $214,288 $112,618 $69,391 $51,288 $43,249
CuM $95,171 $48,723 $30,488 $22,412 $18,392
Csf $124,749 $237,982 $340,877 $431,840 $500,949
Cusf $45,471 $86,744 $124,249 $157,405 $182,595
Ctotal $1,149,512 $1,073,071 $1,119,490 $1,203,028 $1,278,100

Figure 22, considering the decreasing trend of inspec-

tion and maintenance costs and increasing trend of the
service failure costs. From this figure, it can be
understood that the total LCC of the bridge can be
minimised if the inspection and maintenance intervals
Downloaded by [University of Arizona] at 22:38 02 February 2015

are scheduled for every 2 or 3 years. This schedule

optimises the inspection and maintenance costs while
ensuring the safety of the bridge.

Figure 21. Ratio of different cost items to the initial
construction cost.
Figure 22. Total LCC of the medium-span bridge obtained
for different inspection and maintenance intervals.
time schedules. The ratio of different cost types to the
initial construction cost are depicted in Figure 21.
Referring to Table 3 and Figure 21, when the
inspection intervals increase, the costs associated with
the inspection and maintenance decrease but this
causes a significant increase in the service failure cost.
The total LCC of the bridge has been shown in
7. Conclusions
RC highway bridges are continuously exposed to
different environmental stressors during their service
life time. Chloride-induced corrosion is one of dete-
rioration mechanisms that causes serious structural
degradation and may result in the service failure of the
entire bridge. This article develops a comprehensive
framework to study the chloride intrusion mechanisms
and predict the progress of corrosion in RC members.
Through this study, the effects of various influential
parameters, such as water-to-cement ratio, ambient
temperature, relative humidity, concrete age, free
chloride content and binding capacity, are carefully
considered for an accurate estimation of the chloride
penetration profile in deteriorating structural members
over the time.
The corrosion initiation time is estimated by
comparing the chloride content values in the vicinity
of reinforcing bars the with critical thresholds. After
the determination of the corrosion initiation time, the
rate of reduction in geometry and material properties
due to the corrosion process is calculated at different
time steps. This article specifically considers the
reduction rate of diameter and yield strength of
reinforcing bars in corroded members. The crack
initiation and propagation in concrete cover is also
investigated in order to modify the confinement
assumptions required for capacity evaluation. Based
on the updated structural characteristics of corroded
members, the capacity loss of a group of RC bridges
with short-, medium- and long spans is evaluated. In
addition to the structural degradation, the effects of
the corrosion process on the LCC of bridges are also
studied. From the detailed calculation of construction,
 26 A. Alipour et al.
inspection, maintenance and service failure costs of the
bridges, this article examines various inspection and
maintenance strategies and suggests the optimised
inspection and maintenance intervals. The proposed
procedure minimises the total LCC of the bridge while
maintains its expected structural performance.
References
AASHTO Guide Manual for Condition Evaluation and
Load and Resistance Factor Rating (LRFR) of Highway
Bridges, 2005. Washington, DC: American association of
state highway and transportation officials.
Ababneh, A., Benboudjema, F., and Xi, Y., 2003. Chloride
Penetration in Non-saturated Concrete. Journal of
Materials in Civil Engineering, 15 (2), 183–191.
ACI Manual of Concrete Practice, 2009. Farmington Hills,
MI: American concrete institute.
Ahmad, S., 2003. Reinforcement corrosion in concrete struc-
Downloaded by [University of Arizona] at 22:38 02 February 2015
tures, its monitoring and service life prediction: a review.
Cement and Concrete Composites, 25 (4–5), 459–471.
Alipour, A., 2010. Life-cycle performance assessment of
highway bridges under multi-hazard conditions and envir-
onmental stressors. Ph.D. Dissertation. Department of
Civil and Environmental Engineering, University of
California, Irvine, CA.
Alipour, A., Shafei, B., and Shinozuka, M., 2010. Determi-
nistic and probabilistic evaluation of time to corrosion
initiation of bridges located in coastal areas. Proceedings
of the fifth International conference on bridge maintenance,
safety and management (IABMAS), Philadelphia, PA,
July 11–15.
Andrade, C., Alonso, C., and Molina, F.J., 1993. Cover
cracking as a function of rebar corrosion: Part I:
Experimental test. Materials and Structures, 26 (3),
345–353.
Andrade, C., et al., 1991. Remaining life time of reinforced
concrete structures: effect of corrosion in mechanical
properties of the steel, life prediction of corrodible
structures. Proceedings of the international symposium
of the national association of corrosion engineers, Section
12, 1–11, Cambridge, UK.
Bamforth, P.B. and Price, W.F., 1996. An international review
of chloride ingress into structural concrete. Middlesex,
UK: Taywood Engineering Ltd. Technology division.
Rep. No. 1303/96/9092.
Bažant, Z.P. and Najjar, L.J., 1972. Nonlinear water
diffusion in unsaturated concrete. Material and Struc-
tures, 5 (25), 3–20.
Bertolini, L., 2008. Steel corrosion and service life of
reinforced concrete structures. Structure and Infrastruc-
ture Engineering, 4 (2), 123–137.
Caltrans Contract Cost Data, 2008. Sacramento, CA:
California department of transportation.
Caltrans Seismic Design Criteria, 2006. Sacramento, CA:
California department of transportation.
Chang, S. and Shinozuka, M., 1996. Life cycle cost analysis
with natural hazard risk. ASCE Journal of Infrastructure
Systems, 2 (3), 118–126.
Chatterji, S., 1995. On the applicability of Fick’s second law
to chloride ion immigration through Portland cement
concrete. Cement and Concrete research, 25 (2), 299–303.
Collepardi, M., Marciallis, A., and Turriziani, R., 1970.
Kinetic of penetration of chloride ions into concrete.
Industria Italiana del Cemento, 4, 157–164.
Du, Y.G., Clark, L.A., and Chan, A.H.C., 2005a. Residual
capacity of corroded reinforcing bars. Magazine of
Concrete Research, 57 (3), 135–147.
Du, Y.G., Clark, L.A., and Chan, A.H.C., 2005b. Effect of
corrosion on ductility of reinforcing bars. Magazine of
Concrete Research, 57 (7), 407–419.
El Maaddawy, T. and Soudki, K., 2007. A model for
prediction of time from corrosion initiation to corrosion
cracking. Journal of Cement and Concrete Composites, 29
(3), 168–175.
Glass, G.K. and Buenfeld, N.R., 2000. The influence of
chloride binding on the chloride induced corrosion risk
in reinforced concrete. Corrosion Science, 42 (2), 329–
344.
Han, S.H., 2007. Influence of diffusion coefficient on chloride
ion penetration of concrete structures. Journal of Con-
struction and Building Materials, 21 (2), 370–378.
Kong, J.S., et al., 2002. Reliability analysis of chloride
penetration in saturated concrete. Journal of Probabilistic
Engineering Mechanics, 17 (3), 302–315.
Lee, H.S., Tomosawa, F., and Noguchi, T., 1996. Effect of
rebar corrosion on the structural performance of single
reinforced beams. Durability of Building material and
components. London, UK: E & FN Spon, 571–580.
Mander, J.B., Priestly, M.J.N., and Park, R., 1988.
Theoretical stress-strain model for confined concrete.
Journal of the Structural Division, 114 (8), 1804–1826.
Martin-Perez, B., Pantzopoulou, S.J., and Thomas, M.D.A.,
2001. Numerical solution of mass transport equations in
concrete structures. Journal of Computers and Structures,
79 (13), 1251–1264.
McGee, R., 1999. Modeling of durability performance of
Tasmanian bridges. Proceedings of the ICASP8 confer-
ence. Sydney, Australia, 297–306.
Morinaga, S., 1996. Remaining life of reinforced concrete
structures after corrosion cracking. Durability of building
material and components. London, UKE & FN Spon,
127–136.
Nilsson, L.O., Massat, M., and Tang, L., 1993. Chloride
binding capacity and bonding isotherms of OPC pastes and
mortars. Cement and Concrete Research, 23 (2), 247–253.
OpenSees Development Team, 2009. OpenSees: open system
for earthquake engineering simulation. Berkeley, CA:
OpenSees.
Page, C.L., Short, N.R., and El Tarras, A., 1981. Diffusion
of chloride ions in hardened cement pasta. Cement and
Concrete Research, 11 (3), 395–406.
Papadakis, V.G., et al., 1996. Mathematical modeling of
chloride effect of concrete durability and protection
measures. Concrete repair, rehabilitation and protection.
London, UK: E & FN Spon, 167–174.
Priestly, M.J.N., Seible, F., and Calvi, J.M, 1996. Seismic
design and retrofit of bridges. USA: John Wiley and Sons
INC.
Rodriguez, J., Ortega, L.M., and Garcia, A.M., 1994.
Assessment of structural elements with corroded reinfor-
cement. Proceedings of the corrosion and corrosion
protection of steel in concrete. Swamy Edition, Sheffield,
UK: Sheffield Academic Press, 171–185.
Saetta, A.V., Schrefler, B.A., and Vitaliani, R.V., 1993. The
carbonation of the concrete and the mechanism of moisture,
heat and carbon dioxide flow through porous material.
Journal of Cement and Concrete Research, 23 (4), 761–772.
Snyder, K.A., 2001. Validation and Modification of the
4SIGHT Computer Program. NISTIR 6747. USA:
National Institute of Standards and Technology.
 Structure and Infrastructure Engineering
Tang, L. and Nilsson, L., 1993. Chloride binding capacity
and binding isotherms of OPC pastes and mortar.
Cement and Concrete Research, 23 (2), 247–253.
Tuutti, K., 1982. Corrosion of steel in concrete. CBI Research
Report No. 4.82. Stockholm. Sweden: Swedish Cement
and Concrete Research Institute.
Val, D.V., 2004. Aspects of corrosion in reinforced concrete
structures and its influence on structural safety. Report
No. 2002950. National Building Research Institute.
Downloaded by [University of Arizona] at 22:38 02 February 2015
27
Val, D. and Stewart, M.G., 2003. Life-cycle cost analysis of
reinforced concrete structures in marine environments.
Journal of Structural Safety, 25 (4), 343–362.
Vidal, T., Catel, A., and Francois, R., 2004. Analyzing crack
width to predict corrosion in reinforced concrete. Cement
Concrete Research, 34 (1), 165–174.
Xi, Y. and Bažant, Z.P., 1999. Modeling chloride penetration
in saturated concrete. Journal of Materials in Civil
Engineering, 11 (1), 58–65.