Durability Monitoring for Improved Service Life ...karaj/GPSC/Wk4EngineeringHL.pdfing structural...

Computer-Aided Civil and Infrastructure Engineering 26 (2011) 524–541

Durability Monitoring for Improved Service LifePredictions of Concrete Bridge Decks in Corrosive

Environments

D. Cusson,∗ Z. Lounis & L. Daigle

National Research Council Canada, Institute for Research in Construction, Ottawa, Canada

Abstract: The development of an effective strategy forthe inspection and monitoring of the nation’s criticalbridges has become necessary due to aging, increasedtraffic loads, changing environmental conditions, and ad-vanced deterioration. This article presents the develop-ment of a probabilistic mechanistic modeling approachsupported by durability monitoring to obtain improvedpredictions of service life of concrete bridge decks ex-posed to chlorides. The application and benefits of thisapproach are illustrated on a case study of a reinforcedconcrete barrier wall of a highway bridge monitoredover 10 years. It is demonstrated that service life predic-tions using probabilistic models calibrated with selectedmonitored field data can provide more reliable assess-ments of the probabilities of reinforcement corrosion andcorrosion-induced damage compared to using determin-istic models based on standard data from the literature.Such calibrated probabilistic models can help decisionmakers optimize intervention strategies as to how andwhen to repair or rehabilitate a given structure, thus im-proving its life cycle performance, extending its servicelife and reducing its life cycle cost.

1 INTRODUCTION

A large number of reinforced concrete (RC) bridges inCanada and the northern states of the United Statesare short and medium span bridges that exhibit seriouscorrosion-induced deterioration due to the use of de-icing salts on roads during winter. Many of these bridgeswere built during the postwar construction booms of the1950s, 1960s, and 1970s. Approximately 25% of themare considered deficient in terms of structural capac-

∗To whom correspondence should be addressed. Email: Daniel.

[email protected].

ity and functionality (U.S. DOT et al., 2007) as a re-sult of increased traffic loads, changing environmentalconditions, deterioration, and more stringent bridge de-sign codes. The widespread deterioration and some re-cent collapses of highway bridges (Inaudi et al., 2009)have highlighted the importance of developing effectivebridge inspection and maintenance strategies, includ-ing structural health and durability monitoring, whichcan help identify structural and durability problems be-fore they become critical, endanger public safety, andimpede traffic flow. The implementation of structuralhealth monitoring (SHM) programs can provide use-ful information on the physical health of bridges andtheir structural performance (Cruz and Salgado, 2009;Moaveni et al., 2009; Soyoz and Feng, 2009; Huanget al., 2010). Durability monitoring can supply valuabledata that can be used to calibrate service life predictionmodels.

Currently, the majority of highway bridges are in-spected at regular intervals (e.g., every 2 years) throughvisual inspections of the deck, superstructure, and sub-structure, which is followed by a mapping of the ob-served damage to a qualitative condition rating scale(e.g., 1–5; 0–9; etc.). More detailed and in-depth inspec-tions using nondestructive evaluation methods are con-ducted less frequently to supplement the data obtainedfrom visual inspection, especially for critical bridge el-ements to assess the level of corrosion, crack width, fa-tigue, spalling, delamination, stiffness, strength, etc.

1.1 Toward multiobjective management of highwaybridges

Cost optimization and life cycle cost optimization ofconcrete structures have been a subject of increasingresearch interest in recent years (Sarma and Adeli,

C© 2011 Crown in the right of Canada.DOI: 10.1111/j.1467-8667.2010.00710.x

KaraJ

Highlight

KaraJ

Highlight

KaraJ

Highlight

KaraJ

Highlight

KaraJ

Highlight

KaraJ

Highlight

KaraJ

Highlight

KaraJ

Highlight

KaraJ

Highlight

KaraJ

Highlight

KaraJ

Highlight

KaraJ

Highlight

KaraJ

Highlight

KaraJ

Highlight

Durability monitoring for improved service life prediction 525

1998; Sirca and Adeli, 2005). The approaches to mainte-nance optimization implemented in current bridge man-agement systems are based on single objective opti-mization, and more specifically on the minimization ofmaintenance costs or life cycle costs, which representsthe present value of all the costs incurred throughoutthe life cycle of a bridge structure, including the costsof design, construction, maintenance, repair, rehabilita-tion, replacement, demolition, and, in some instances,the users costs.

The management of bridges can be formulated asa multiobjective optimization problem as the bridgeowner or manager seeks to satisfy implicitly and simul-taneously several objectives, such as the minimizationof costs to owners and users, improvement of publicsafety and security, improvement of serviceability andfunctionality, minimization of maintenance time, mini-mization of traffic disruption, etc. The solution of such abridge management problem can be obtained by usingthe techniques of multicriteria or multiobjective opti-mization. Despite its relevance, the use of risk as a crite-rion for decision making in the management of highwaybridges is not easy given the complexities of assessingthe probability of failure, as well as the consequences offailure (Wenzel, 2009). In life cycle cost and cost-benefitanalyses, which have been implemented in many bridgemanagement systems currently used by different trans-portation agencies in North America, all losses, includ-ing fatalities, injuries, and social costs are quantified inmonetary terms, including the cost of human life.

It is clear from the above discussion that the exist-ing approaches to decision making have serious limita-tions as they express all losses in monetary terms andconsider only one criterion at a time, for example, min-imization of owner costs. Considering this issue, Lounisand Cohn (1995), Lounis and Vanier (2000), and Louniset al. (2009) proposed a multiobjective optimization ap-proach for decision making, which can incorporate allrelevant objectives, to help solve the bridge manage-ment problem. Such an approach enables a better evalu-ation of the effectiveness of preservation and protectionstrategies in terms of several objectives (safety, security,mobility, economy, cost, etc.) and determines the opti-mal solution that achieves the best trade-off between allof them (including conflicting ones, such as safety andcost). Figure 1 outlines a framework for the multiobjec-tive management of highway bridge structures. The de-velopment and implementation of this framework willprovide effective decision support to bridge owners andmanagers in optimizing the allocation of limited funds,as well as improving the life cycle performance of theirbridges.

SHM is one of the major components of this frame-work, which appears in two distinct phases: (i) risk-

based assessment of the current condition of the struc-ture, in combination with nondestructive evaluation(NDE) and load rating, which will help assess the phys-ical conditions of the bridge; and (ii) risk managementof identified safety-critical and security-critical bridges,along with other parallel activities such as visual in-spection, nondestructive evaluation, quantitative assess-ment, maintenance, strengthening, and protection. Thecomplete framework and multiobjective approach forthe effective management of bridges, which are outsidethe scope of this article, are presented in Lounis et al.(2009).

1.2 Toward structural health and durability monitoringof highway bridges

Structural health and durability monitoring, either withembedded sensors or by actual field testing, is an evolv-ing technology that can be used to monitor the con-dition of existing or new civil engineering structures.Depending on the importance and location of the struc-ture, available budgets and monitoring objectives, threemajor types of SHM can be adopted: (i) occasionalshort-term monitoring, to confirm design assumptions,monitor the structural behavior during a planned ex-ceptional event, or help in the planning of a more com-plex form of monitoring; (ii) periodic monitoring, toassess the parameters that are mainly influenced bycyclic events; and (iii) continuous monitoring, to assessthe performance over the life cycle of a critical struc-ture, including parameters that are influenced by staticand dynamic events (Braunstein et al., 2002). Continu-ous remote monitoring may be more cost-effective inthe long term than conducting frequent field testing,considering the cost of labor, the costs to the users, andtheir safety. In-depth information on the design of SHMsystems and applications on bridges can be found else-where (Mufti, 2001; Braunstein et al., 2002; Glisic andInaudi, 2007; Frangopol et al., 2008; Cardini and De-wold, 2009).

The key benefits of SHM, including the monitoringof structural performance or long-term durability, canbe summarized as follows:

i. Improvement of public safety and extension of ser-vice life of bridges: SHM can enable bridge own-ers to remotely monitor the performance of theirstructures from a central site via Internet, therebyreducing the number of site visits for visual in-spections and destructive and/or nondestructivetesting. Therefore, SHM can provide continuousinformation on structural performance (e.g., ex-cessive deformations and stresses, yielding, foun-dation settlement) and durability (e.g., concrete

526 Cusson, Lounis & Daigle

Improve Life Cycle Performance of Critical Bridges

Safety

Performance objectives

Fatalities, injuries, illnesses, traffic disruption, pollution, asset destruction, socio-economic impacts

Haz

ard

sR

isk

Ass

essm

ent

Ris

k M

anag

emen

t

Extreme Shocks Extreme wind, flood, earthquake, vehicle

impact, ship impact, explosion, fire

Priority List of Critical Bridges Identify safety-critical and security-critical bridges

Bridge Preservation

Frequent or In-Depth Inspections

Continuous Monitoring

i

Maintenance/Strengthening

Protection of Critical Systems

Quantitative Assessment

Structural safety, vulnerability, robustness, serviceability, service life

Increasing Traffic

Environmental LoadsDeicing Salts, seawater, CO2,

temperature, wind

Health Mobility Security Environment

Quality Economy Social Equity

Consequences of Deterioration/Failure

Physical Condition

Multi-Objective Prioritization Safety, Health, Mobility, Security, Cost, etc…

Assess Current StateNon-destructive evaluation, load

rating, continuous health monitoring

Aging

Forecast Future StatesStochastic deterioration models

Stochastic models of loads

Health monitoring

Fig. 1. Framework for multiobjective management of aging highway bridges.

cracking, reinforcement corrosion, freeze-thawdamage). The access to such continuous informa-tion, rather than relying on periodic information(e.g., every 2nd year from conventional visualinspections) will allow timely decisions on cor-rective measures before problems become crit-ical and endanger public safety. Safety-criticalbridges should be monitored to ensure that thelikelihood of failure of critical load-bearing ele-ments is kept very low, especially for bridges inmajor urban centers, and nonredundant bridgesystems, for which failure can have catastrophicconsequences.

ii. Development and calibration of service life pre-diction models: The vital information acquiredby SHM can foster a better understanding ofdamage initiation and damage accumulation inbridge structures (Cusson et al., 2006; Butcherand Newhook, 2009) and calibration of service

life prediction models. In addition to the scarcityof field data available from full-size engineeringstructures, many key parameters used in servicelife prediction models are highly variable and un-certain, which may adversely affect the reliabil-ity of the predictions. The continuous updating ofservice life prediction models with selected fielddata can have a considerable impact on the plan-ning of preservation actions.

iii. Improved design and rehabilitation of concretestructures: Another benefit of SHM is to helpreassess the current live loads and environmen-tal loads on bridge structures. This will helpidentify any major variations from the histor-ical values that are assumed in bridge designcodes, thus reducing the uncertainty of critical de-sign parameters. This is becoming an importantissue given the growing concerns with climatechange and its potential impact on the safety and


serviceability of bridge structures due to in-creases in wind loads, flooding, thermal gradients,freeze-thaw cycles, deicing salt use, etc. Structuralhealth monitoring can also be used to ensure thestructural integrity of the bridge during the se-quential rehabilitation or replacement of dam-aged load-bearing elements (Carrion et al., 2009).

iv. Continuous security monitoring of critical bridges:A continuous video camera surveillance systemcan be installed to monitor critical bridges andtheir immediate surroundings to identify and re-spond to potential threats to the bridges andtheir components. Various types of activities canbe monitored depending on the type of bridge,such as car and truck traffic, movement of peo-ple, as well as boats, ships, and aircraft activities.Computer vision and pattern recognition technol-ogy can also be used to allow computers to pro-cess recorded images, watch for danger signs, andsend alarms to security officers. Such technolo-gies require high-speed communication systemsto deliver the information to remote security of-fices for analysis, response, reporting, and archiv-ing purposes. It is possible though to process andarchive images on site; however, this would notallow postcatastrophe analysis (FHWA, 2009).

v. Faster development and acceptance of new con-struction technologies: SHM can provide a struc-tured approach to assess the performance ofemerging innovative technologies applied indemonstration projects (thus promoting innova-tion), whether they are conducted on old struc-tures or new construction. Research on bridgesinvolves the use of sensors and data logging sys-tems for continuous monitoring of selected per-formance parameters (Tennyson et al., 2001; Cus-son et al., 2006). Sensors can provide detailedknowledge of (i) the conditions under which thestructure is being evaluated, including appliedloads, imposed displacements, and ambient en-vironment; and (ii) selected performance param-eters, which could provide insights on the per-formance of newly developed technologies andprotection systems, or changes in the behaviorof the structure as a result of strengthening orrepair.

1.3 Objectives

The objectives of this article are twofold: (i) to presenta probabilistic mechanistic modeling approach basedon durability monitoring to assess the life cycle perfor-mance of concrete bridge decks in corrosive environ-ments; and (ii) to demonstrate the effective use of se-lected data obtained from field monitoring to update

and improve the accuracy of mechanistic service lifeprediction models. A case study of the monitoring ofa concrete highway bridge barrier wall is presented andused to illustrate the approach and its benefits.

2 MECHANISTIC MODELING OF SERVICE LIFEOF RC BRIDGE DECKS SUBJECT TO

CHLORIDE ATTACK

2.1 Proposed durability monitoring and mechanisticmodeling approach

Corrosion of reinforcing and prestressing steel is a ma-jor problem for bridges in Canada and the northernstates of the United States. The deterioration inducedby corrosion accounts for a large portion of the mainte-nance expenditures of several bridge owners (U.S. DOTet al., 2007). For the structural performance and safetyassessment of bridges, several models have been devel-oped in the past, including Markovian models (whichare only qualitative), deterministic mechanistic mod-els, and probabilistic mechanistic models. However, forthe durability assessment and service life prediction ofbridge decks in corrosive environments, fewer modelshave been proposed and a large number of them are de-terministic models that do not consider the uncertaintyand variability associated with the key parameters gov-erning the initiation of corrosion and corrosion-induceddamage processes.

The proposed approach consists of the monitoringof critical durability parameters that govern the pro-posed mechanistic models for the service life of RCbridge decks exposed to chlorides. The field monitoringcould be periodic or continuous, using sensors and/orNDE methods ideally, or even limited destructive test-ing (e.g., cores), if absolutely necessary. The monitor-ing program could be conducted over a few years orduring the whole service life, depending on the impor-tance of the structure, the available monitoring budget,and other factors. Although long-term continuous mon-itoring is preferable, it will be shown in a case study(Section 3) that periodic monitoring of a new structureconducted over the first 10 years of its life can providereliable input data for the prediction of its remainingservice life (prediction models to be presented later inthis section).

For the case of RC bridge decks exposed to chloridesfrom deicing salts or seawater, four variables related toconcrete and steel properties have strong influences onservice life, namely

i. surface chloride content of concrete, which mayincrease over time and vary in space due to thecontinual use of deicing salts on roads, precipita-tion, and drainage;


ii. chloride diffusion coefficient of concrete, whichmay decrease over time due to the continuingcement hydration, decreasing concrete porosity,which may also vary in space;

iii. chloride threshold of the reinforcement, whichmay vary in time and space. It is also highly un-certain as it depends on environmental conditionsand properties of concrete and steel;

iv. corrosion rate of the reinforcement and rate ofdeterioration of the concrete structure, which areboth highly variable.

The considerable uncertainties associated with these pa-rameters governing the service life of RC bridges high-light the need for calibrating and updating service lifeprediction models with selected field monitoring data,and for the use of probabilistic analysis methods (to bepresented next).

In addition to the high variability and uncertainty ofthese parameters, this problem is compounded by thefact that just a few of these key parameters can actu-ally be monitored by the use of remote sensors. Al-though some analog sensors may be used to determinethe chloride threshold (indirectly) and the corrosionrate of the reinforcement, no commercial sensors arecurrently available to remotely determine the chloridecontent and chloride diffusion in concrete. Therefore,there is a strong need for the development and accep-tance of new sensors for durability monitoring to reducethe frequency of required field testing, including the useof destructive testing (such as taking concrete cores for

the determination of chloride content profiles in a struc-ture).

2.2 Conceptual two-stage service life model

Figure 2 presents the different corrosion-induced dam-age mechanisms developing in a typical RC bridgedeck exposed to chlorides, which identifies two dis-tinct stages: a corrosion-initiation stage and a propaga-tion stage. This is a modified version of Tuutti’s sim-plified model (1982), which did not consider the effectof early-age cracking on chloride penetration duringthe corrosion-initiation stage. With time, each stage de-velops into higher levels of damage, which include: (i)early-age cracking of concrete due to restrained shrink-age (if any); (ii) initiation of reinforcement corrosion af-ter a relatively long period of chloride diffusion throughconcrete; (iii) internal cracking around the reinforcingbars due to the build-up of corrosion products; (iv) sur-face cracking due to further progression of corrosion-induced cracks; (v) spalling or delamination of theconcrete cover; and finally (vi) rehabilitation or replace-ment of the concrete deck, depending on the amountof concrete damage to the deck that can be toleratedby the bridge owner. Note that other conceptual servicelife models do exist, such as that proposed by Li et al.(2007), based on crack width and sectional strength, asopposed to Tuutti’s model which is based on the degreeof corrosion and damage accumulation.

Based on the conceptual model illustrated inFigure 2, the service life (ts) can be defined as the total

Fig. 2. Schematic description of the service life model.


Table 1AASHTO description of condition states for concrete bridge

decks with no surface protection and built with uncoatedreinforcement

Conditionstate Description

1 The surface of the deck has no patched area andno spall in the deck surface

2 The combined distress area (i.e., existing patches,spalling, and delamination) of the deck is lessthan 10%

3 The combined distress area of the deck isbetween 10% and 25%

4 The combined distress area of the deck isbetween 25% and 50%

5 The combined distress area of the deck is morethan 50%

time to reach a given corrosion-induced damage level,which is the sum of the corrosion-initiation time (ti)(also considering the effect of early age cracking), andthe propagation time (tp) corresponding to a given levelof damage, as follows:

ts = ti + tp (1)

Several different estimates of service life can be ob-tained by using this conceptual model. In North Amer-ica for instance, different departments of transporta-tion have adopted different criteria for defining the endof service life as a measure to limit excessive damageto the structure and as an indicator for the need ofa major rehabilitation. For example, the end of ser-vice life of a concrete highway bridge deck is generallydefined as the time to reach critical damage levels interms of spalling or delamination (e.g., AASHTO, 2002,Table 1). It is up to the bridge authority to decidewhich criteria to use based on many factors, includingthe desired levels of serviceability, functionality, andsafety, the costs of maintenance and repair, and avail-able funds. In this article, the end of service life of aconcrete bridge deck is defined as the time to onset ofspalling. It is not the intent of this article to relate theend of service life of a bridge to the time of possible col-lapse, because the focus of the proposed approach is ondurability of a bridge deck under service conditions, andnot the structural integrity of the bridge.

2.3 Prediction of chloride ingress into concrete

In uncracked concrete, the chloride ingress can be de-termined by using Crank’s solution of Fick’s second lawof diffusion (Crank, 1975):

C(x, t) = Cs(1 − erf (0.5x/√

Dct)) (2)

where C(x,t) is the chloride content at depth x after timet; Cs is the apparent surface chloride content (assumedconstant); Dc is the apparent chloride diffusion coeffi-cient (assumed constant); and erf is the error function.The above model also assumes that concrete is homoge-nous and diffusion is the main mechanism of chlorideingress into concrete. In porous solids such as concrete,chlorides can penetrate into concrete via different phys-ical mechanisms, such as diffusion, capillary absorption,electrical migration, and permeation due to hydraulicpressure heads, depending on the exposure conditionand moisture content (Kropp and Hilsdorf, 1995). Thelack of accuracy of Fick’s model has been recognizedfor a long time, as it has been used with the so-called“apparent” values of the surface chloride content andchloride diffusion coefficient, which are obtained by fit-ting the Fickian model to available field data. How-ever, this model has gained wide acceptance due to itssimplicity and practicality (Bamforth, 1998; Tang andGulikers, 2007).

In cracked concrete, chloride ingress through cracksmay lead to premature reinforcement corrosion andconcrete deterioration (De Schutter, 1999; Rodriguezand Hooton, 2003). To model the effect of crack-ing on chloride ingress, it is generally accepted thatthe chloride diffusion coefficient corresponding to un-cracked concrete needs to be modified into an apparentvalue representing the cracked medium (Francois andArliguie, 1999; Gerard and Marchand, 2000). Boulfizaet al. (2003) used a simplified smeared approach to esti-mate the effect of cracks on chloride ingress. It assumesthat chloride ingress into cracked concrete can be ap-proximated by using Fick’s second law of diffusion, inwhich an apparent diffusion coefficient (Dapp) is foundwith this equation:

Dapp = Dc + wcr

scrDcr (3)

where Dc is the chloride diffusion coefficient in un-cracked concrete; wcr is the crack width; scr is the crackspacing; and Dcr is the diffusion coefficient inside thecrack, which may be assumed to be 5 × 10−10 m2/s, ac-cording to Boulfiza et al. (2003).

2.4 Prediction of onset of corrosion and concretespalling

To predict the time of corrosion initiation (ti), Equation(2) can be rearranged by setting C(x,t) equal to the chlo-ride threshold (Cth), at which steel corrosion is expectedto initiate, and x equal to the effective cover depth of thereinforcement (dc), as follows:


Fig. 3. Uncracked and cracked thick-wall cylinder models ofcorroding RC bridges: (a) tensile stresses developed at crack

initiation; (b) propagation of internal cracks in thick-wallcylinder (adapted from Lounis and Daigle, 2008).

ti = d2c

4Dapp

[erf −1

(1 − Cth

Cs

)]2 (4)

The accumulation of corrosion products over timegenerates contact pressure between the rebar and thesurrounding concrete, and may initiate cracks if the ten-sile stress in the concrete reaches the tensile strength(f ′

t). The concrete cover and the corroding rebar canbe modeled as a thick-wall cylinder subjected to auniformly distributed internal pressure (pi), assumingthat concrete is a homogeneous elastic material, andthat corrosion products are equally distributed aroundthe perimeter of the reinforcing bars (Bazant, 1979;Tepfers, 1979). The radial stress (σr) and tangentialstress (σt) generated in the concrete cover by the expan-sion of the corrosion products can be estimated usingthe thick-wall cylinder model (Figure 3, Timoshenko,1956). This model allows the calculation of the increasesin the rebar diameter (�d) related to different stages ofcorrosion-induced damage (i.e., internal cracking, sur-face cracking, spalling, or delamination) by calculatingthe radial displacements at the inner surface of the cylin-der. Assuming that the external pressure induced by ex-ternal loads on the structure is small or negligible, theincrease in rebar diameter at the onset of spalling isgiven by

�d = pi

Ec

[1 + ν + d2

2dc(dc + d)

]d (5)

where d is the rebar diameter, dc is the concrete coverthickness, ν is Poisson ratio of concrete, Ec is the effec-tive elastic modulus of concrete taking into account theeffect of creep (Bazant, 1979). In the case of spalling,assuming a 45◦ plane of failure with respect to the con-crete surface, the internal pressure in the concrete be-fore reaching the concrete tensile strength (f ′

t) is givenby Bazant (1979)

pi = 2 f ′t

(dc

d+ 0.5

)(6)

Note that similar expressions of �d and pi have beendeveloped to model the other limit states illustrated inFigure 2 (i.e., onset of internal cracking, onset of surfacecracking, and onset of delamination), and can be foundelsewhere (Lounis and Daigle, 2008).

Assuming that, for a row of parallel reinforcing bars,the corrosion products are smeared over the entire area,thus the production rate of corrosion products per unitarea is given by

jr = mr

stp(7)

where s is the rebar spacing, tp is the propagation time,and mr is the mass of corrosion products defined byBazant (1979) as

mr = πd(�d)

2[

1ρr

− α

ρs

] (8)

where ρr is the density of corrosion products (assumedat 3,600 kg/m3 for Fe(OH)3); ρs is the density of steel(7,860 kg/m3); and α is the molecular weight ratio ofmetal iron to the corrosion product (assumed at 0.52).

Substituting Equation (8) into Equation (7) and solv-ing for tp, the propagation time corresponding to the on-set of spalling can be determined as follows:

tp = πd(�d)

2s jr

[1ρr

− α

ρs

] (9)

The production rate of corrosion products (jr) is alsorelated to the corrosion-current density (icor), whichcan be monitored in the field, and the loss of rebardiameter due to corrosion, which was determined as0.023icortp by Andrade and Alonso (2001). Conse-quently, jr can also be expressed as

jr = 0.023πdIcor

2s[

1ρr

− α

ρs

] (10)

where the parameter 0.023 is an empirical conversionfactor from current density units of μA/cm2 to corrosionrate units of mm/year.

2.5 Modeling of uncertainty in service life prediction

Service life predictions may be obtained through asimple deterministic analysis, that is, an analysis thatdoes not consider the variability and uncertainty asso-ciated with the main parameters governing the differ-ent mechanisms of deterioration in RC bridge decks.


Deterministic analyses are based on mean or charac-teristic values of the variables and can only predict thetimes to reach the different stages of corrosion initia-tion and corrosion-induced damage caused by an aver-age condition. To account for the variability and uncer-tainty of the input parameters, more reliable service lifepredictions can be obtained by using reliability-basedmethods, with the key input parameters expressed interms of their mean values and coefficients of variation(COV).

If the investigated governing parameters exhibit con-siderable spatial variability within the given bridgedeck, the analysis may require the use of random fieldsinstead of random variables for their modeling (Van-marcke, 1983; Stewart et al., 2003; Lounis and Daigle,2008). The use of random fields, however, requires con-siderably more input data such as the correlation func-tion and scale of fluctuation (which are most often notavailable in practice), and adds considerable analyticaland computational complexity to the analysis.

If the variability of the parameters can be assumedto be completely random, that is, no systematic vari-ability and no spatial correlation, then the use of ran-dom variables for modeling the above parameters is ad-equate (Lounis and Daigle, 2008). Considering that theRC bridge deck can be divided into a very large numberof small surface areas with the same probability (P%) ofspalling after a number of years, it can be approximatedthat a certain percentage (P%) of them are spalled aftera given time. These probabilities can be used to makeinitial and approximate comparisons to the five con-dition states defined in the 2002 AASHTO guidelinesfor the assessment of concrete bridge decks (Table 1),which is used in the bridge management systems of sev-eral states in the United States and some provinces inCanada.

In general, the time-dependent probability of failure(i.e., reaching limit states such as critical chloride con-tent, onset of corrosion, onset of cracking, spalling, ordelamination) can be formulated as follows:

Pf (t) = P(ts < t) (11)

The determination of the time-dependent probabil-ity of failure is a very complex problem due to the highlevel of nonlinearity of the performance functions thatdescribe the above limit states, which can be modeledwith first-order reliability methods, second-order reli-ability methods, Monte Carlo simulation, or randomfields. In this approach, the advanced first-order relia-bility method was selected and used to determine theprobability of failure for the above described service lifeperformance functions. The determination of the prob-ability of failure can be formulated as a nonlinear op-timization problem, in which the solution minimizing

the distance from the origin to the failure surface (de-scribed by any of the above performance functions) issought. This minimum distance was defined as a mea-sure of the reliability index by Hasofer and Lind (1974).Once the reliability index (β) is determined, the proba-bility of failure can be estimated as follows:

Pf (t) = 1 − (β) (12)

where is the well-known standard normal cumulativedistribution function. An iterative numerical algorithmand a software package called SLAB-D were developedto solve the above nonlinear optimization problem andto determine the reliability index (Lounis and Daigle,2008). If the random variables are not normal, then theabove approach provides estimates of the probability offailure (Madsen et al., 1986). The validity of this ap-proach will be demonstrated later in the case study bycomparing its prediction to that obtained with a MonteCarlo simulation (Melchers, 1999).

2.6 Implementation of proposed approach in practice

The use of detailed reliability-based mechanistic modelsfor every asset of a large network of bridges can becomevery costly and unmanageable. To address the needof asset managers for decision support tools that areboth practical and reliable, Lounis and Madanat (2002)proposed a two-level decision process based on twotypes of deterioration models. At level-1, the simplifiedMarkovian-type cumulative damage models are used toforecast the overall deterioration and required main-tenance funds for both short- and long-term planningfor a network of bridges, and to identify the criticallydamaged systems. At level-2, reliability-based mecha-nistic deterioration models (such as the ones proposedin this article) are used for the analysis of critical assets,which were identified from the first level of manage-ment analysis or from specific detailed assessment. Inthe case of an RC bridge deck exposed to chlorides, theanalysis at level-2 can provide quantitative estimates ofthe damage level and relevant information to select themost cost-effective maintenance strategies for the dete-riorated system.

In the next section, the proposed approach will bedemonstrated in a case study of a reconstructed con-crete bridge barrier wall of a highway bridge locatednear Montreal, Canada. It should be emphasized how-ever that the approach can be applied on other mem-bers of a bridge, regardless of surface orientation or thetype and severity of the exposure conditions. As long asthe chloride content profile of the concrete and corro-sion rate of the reinforcement can be measured or mon-itored either periodically or continuously, then accuratepredictions can be achieved with the proposed predic-tion approach.


Ideally embedded sensors, or other NDE methods(such as proposed by Sengul and Gjorv, 2008), shouldbe used to obtain the desired information to updatethe prediction models presented above. In cases wheresensors or NDE could not be used for any reason, thebridge authority will have to weigh out the impact ofusing minimal destructive testing at carefully selectedlocations on the bridge deck (e.g., coring through mem-brane in deck slab) against the risk of relying on lessaccurate service life predictions. Alternatively, anotheroption may be considered by the bridge authority whendealing with limited or partial data. For example, Choiand Seo (2009) proposed an assessment procedure forcorrosion-damaged structures based on the imprecisereliability theory, which was developed for cases whereonly insufficient data are available.

3 CASE STUDY: LIFE CYCLE PERFORMANCEPREDICTION OF A CONCRETE BRIDGE

BARRIER WALL USING DURABILITYMONITORING

The purpose of this section is to illustrate the benefits ofdurability monitoring for improving service life predic-tions, and to point out that some of the input data thatare commonly assumed in typical prediction models canbe somewhat different from reality and lead to inaccu-rate predictions, as some parameters are uncertain andcan also vary widely in space and time.

3.1 Description of bridge deck and experimentalprogram

In 1996, the Ministry of Transportation of Quebec un-dertook the rehabilitation of the Vachon bridge, which

is a major highway bridge in Laval (near Montreal)Canada. Part of the rehabilitation consisted of rebuild-ing the severely corroded concrete barrier walls, ofwhich ten 34 m spans were selected for the applicationand evaluation of different corrosion-inhibiting systems.The barrier wall reinforcement consisted of 15 mm di-ameter bars, with eight longitudinal bars distributed inthe cross-section, and transverse reinforcement spacedat 230 mm, as illustrated in Figure 4. The concrete hada water–cement ratio (w/c) of 0.36 to ensure low per-meability, a cement content of 450 kg/m3, and a mean28-day strength of 45 MPa. On-site surveys of the bar-rier wall were performed annually from 1997 to 2006(in May/June to ensure similar environmental condi-tions), including measurements of corrosion potentialand corrosion rate of the reinforcement, and concreteelectrical resistivity in the barrier wall, of which the con-crete cover was 75 mm. For early detection of corrosion,sets of rebar ladders were embedded during construc-tion. The ladder bars had concrete cover thicknesses of13 mm, 25 mm, 38 mm, and 50 mm (Figure 4), allowingadditional corrosion measurements to be taken. Moredetails can be found in Cusson and Qian (2009).

3.2 Measurement and prediction of chloride ingressinto concrete

Concrete cores were taken from the bridge barrier wallsafter 1, 2, 4, 5, 8, and 10 years of exposure to de-icingsalts to test several parameters, including chloride con-tent. The total acid-soluble chloride contents at differ-ent depths in the concrete were measured accordingto the ASTM C114 (1999) standard procedure. Of the10 spans of barrier wall under study, 3 of them hadidentical concrete mix designs and similar concrete sur-face conditions (referred to as Spans 12, 19, and 21

Fig. 4. Cross-section of the reconstructed RC bridge barrier wall (Vachon Bridge, Laval, Canada).


0

5

10

15

20

25

0 10 20 30 40 50 60 70 80

Tota

l ch

lori

de

co

nte

nt (

kg

/m3)

Depth (mm)

10-year exposure to deicing salts

Field data (measured on 3 test spans)

Best fit of field data

Predictions based onCs & Dc from literature

Apparent Cs

Fig. 5. Measured and predicted profiles of total chloridecontent after 10 years.

Table 2Severity classification of corrosive environment

(Weyers, 1998)

Apparent surfaceSeverity chloride content

Low 0.0–2.4 kg/m3

Medium 2.4–4.7 kg/m3

High 4.7–5.9 kg/m3

Severe 5.9–8.9 kg/m3

according to the bridge deck construction drawings).Figure 5 presents the profile of the total chloride contentmeasured in the concrete after 10 years of exposure todeicing salts (each data point is a mean value obtainedfrom two cores). The best-fit curve was obtained by aregression analysis using Crank’s solution of Fick’s 2ndlaw of diffusion (Equation 2). For the regression, val-ues of Cs and Dc were adjusted until the sum of errorsbetween the solution and the field data reached a mini-mum. The chloride contents measured at a mean depthof 6 mm from the concrete surface were not consideredin the regression analysis because diffusion near the sur-face is not the only main mechanism governing chlorideingress in concrete (as discussed above).

From the field data obtained over a period of 10 years,a mean apparent surface chloride content of 22.7 kg/m3

and a mean apparent chloride diffusion coefficient of0.63 cm2/year were determined. In reality, the highestnear-surface chloride content was measured to be atleast 16.5 kg/m3 in the barrier wall (Figure 5), whichis already much higher than the maximum value of8.9 kg/m3 suggested by Weyers (1998) for geographicalregions where severe exposures to deicing salts are ex-pected (Table 2). It is noted that these guidelines weredeveloped for bridge decks located in the United States

and may not apply to regions like Canada or othernorthern countries, where more deicing salts may beused during longer and colder winter periods.

Similarly, the mean value of the apparent chloridediffusion coefficient of 0.63 cm2/year determined after10 years for the concrete barrier wall was found to belarger than those obtained from the literature for a con-crete with similar w/c. For example, Figure 5 shows thepredicted chloride profiles using Equation (2) with Cs =8.9 kg/m3, as suggested by Weyers (1998) for severe ex-posure conditions, and (a): Dc = 0.21 cm2/year basedon data provided by Dhir et al. (1990) on a concretevery similar to that used in this case study, or (b): Dc =0.38 cm2/year determined with the empirical model ofBoulfiza et al. (2003) for a concrete with a w/c = 0.36containing no supplementary cementing materials.

It can be seen that the chloride profiles are largelyunderestimated when compared to the measured fielddata, even for a short period of 10 years. These dis-crepancies can be explained by the large fluctuations ofmany factors influencing chloride ingress into concrete,including concrete mixture formulation, hydration andcuring characteristics, temperature and humidity con-ditions, and surface chloride content. It may be con-cluded that the determination of the chloride profile fora given concrete structure, using selected literature val-ues from similar concretes tested under similar environ-ments, may result in inaccurate estimates, leading to po-tential overestimations of the remaining service life ofthe structure.

Figure 6 presents the evolution of the apparent sur-face chloride content over time, where it is shown to in-crease significantly over time up to a maximum valueof 22.7 kg/m3 after 10 years. Figure 6 also presentsthe time variation of the apparent chloride diffusion

0.0

0.5

1.0

1.5

2.0

2.5

0

5

10

15

20

25

0 2 4 6 8 10 12

Ap

p. C

l d

iffu

sio

n c

oe

ffic

ien

t (cm

2/y

r)

Ap

p. s

urf

ace

Cl co

nte

nt (

kg

/m3)

Time (years)

Best fit ofapparent Cs

Best fit ofapparent Dc

Each point is an average of 6 test results

Fig. 6. Apparent surface chloride contents and apparentchloride diffusion coefficients determined over a

period of 10 years.


Table 3Mean values of apparent surface chloride contents and apparent chloride diffusion coefficients used in comparative analysis

Selected data from Selected data from field measurements (see Figure 6)Parameter the literature (case 1) (case 2) (case 3)

Cs 8.9 kg/m3 Increasing from 10.8 to 22.7 kg/m3 in 10 years, and 22.7 kg/m3

constant at 22.7 kg/m3 for the remaining time

Dc 0.21 cm2/year, or Decreasing from 1.17 to 0.63 cm2/year in 10 years, and 0.63 cm2/year0.38 cm2/year constant at 0.63 cm2/year for the remaining time

coefficient, where it is observed to decrease by a factorof 2 from year 2 to year 10. This reduction could be ex-plained in part by the continuing cement hydration andcorresponding reduction in concrete porosity.

Considering that most chloride diffusion predictionmodels use constant values of Cs and Dc from the lit-erature or initial field data, the above observations sug-gest that prediction models may also yield inaccuratepredictions if input values of Cs and Dc are not reg-ularly updated with field monitoring data. To investi-gate this point further, the time variation of the chlo-ride content was estimated for three cases: case 1 usingconstant values of Cs and Dc from the literature for asimilar concrete; case 2 using the varying values of Cs

and Dc determined from field data measured over 10years; and case 3 using the constant values of Cs andDc determined from field data measured at a time of 10years. Table 3 gives the values used in this comparativeanalysis.

As varying values of Cs and Dc were used in case 2,Equation (2) could not be used to determine the chlo-ride profile over time. Instead, it was determined bysolving the one-dimensional partial differential equa-tion of Fick’s 2nd law of diffusion:

∂C∂t

= ∂

∂x

(Dc

∂C∂x

)(13)

The numerical solution of Equation (13) was obtainedwith a one-dimensional model of the concrete barrierwall using the finite difference method with the follow-ing central operator:

Ci,�t = Ci + 0.25 [Ci−1 − 2Ci + Ci+1] and �t = �x2

4Dc

(14)

where Ci,�t is the chloride content in concrete at the ithlayer for a given time increment �t; and �x is the thick-ness of the concrete layer (which was set to 5 mm inthis case). Note that this particular diffusion problemcould also be solved in a similar manner using othertechniques, for example, cellular automata, which is aspecial class of evolutionary algorithms (Vorechovskaet al., 2009).

Figure 7 presents the estimated chloride contents atdifferent concrete depths over time for case 1, in whichthe values of Cs and Dc were selected from the litera-ture for a similar concrete in a similar environment. Forexample, depending on the selected value of Dc, a chlo-ride content of 1.8 kg/m3 (Cth according to CEB, 1992)would take 20–35 years to accumulate in this concreteat a 50 mm depth, or 45–80 years at a 75 mm depth. Fig-ure 8 presents the same calculations for case 2, in whichvarying and more severe values of Cs and Dc based onfield measurements were used. It can be seen that thepredicted chloride contents are much higher than thosefrom case 1. For example, in case 2, a chloride contentof 1.8 kg/m3 would take only 5 years to accumulate inthis concrete at a 50 mm depth, or only 12 years at a75 mm depth, which is quite different from the predic-tions for case 1.

To demonstrate the effect of using constant versusvarying input values on the predictions of chloride con-tent, the results of case 3 (using the constant field valuesof Cs and Dc determined at an age of 10 years) are com-pared in Figure 8 to those of case 2 (using the varyingfield values determined over a period of 10 years). Itcan be seen that chloride profiles are slightly differentin the very early years; however, not enough to invali-date the predictions of the models using constant valuesof Cs and Dc. In this case study, when using low valuesof chloride threshold (e.g., 1.8 kg/m3), case 2 predictionsare slightly more conservative than case 3 predictions byonly 2 or 3 years. Predictions of high chloride contentsobtained in the long term do not seem to be significantlyaffected by the choice of input values from case 2 or case3, as long as the values are based on field measurementsmade on the structure of interest.

The above analysis clearly illustrates the need forfield monitoring to update current prediction models forwhich the accuracy of the predictions rely on the qualityof the input data. For example, Figure 9 compares themeasured chloride profile obtained at an age of 10 years(data points) to three estimated chloride profiles for anage of 10 years (curves) using Equation (2) updated withCs and Dc values measured at ages of 2 years, 5 years,and 10 years. It can be observed that the use of more


Fig. 7. Predictions of chloride profiles for cases 1a and 1b (with Cs and Dc selected from the literature).

Fig. 8. Predictions of chloride profiles for cases 2 and 3 (with Cs and Dc obtained from durability monitoring of bridge).

recent updates of Cs and Dc (e.g., year 5 values insteadof year 2 values) in the chloride diffusion model canconsiderably improve the prediction of the long-termchloride profile (in this case 10 years). It is also inter-esting to note that predictions using the year 2 data inthis case study provided a much better estimate of thelong-term chloride profile than the predictions based onthe literature values selected for a similar concrete un-der similar exposure conditions (see Figure 5).

3.3 Observation and prediction of corrosion initiationand concrete spalling

Figure 10 presents a sensitivity analysis of the times toinitiate corrosion and concrete spalling (using Equa-tions 1–10), depending on several factors: (i) coverthickness; (ii) chloride threshold; and (iii) corrosionrate. At the 75 mm depth (location of the main rein-forcing bars in the barrier wall), the prediction indicates


0

5

10

15

20

25

0 10 20 30 40 50 60 70 80

Tota

l ch

lori

de

co

nte

nt (

kg

/m3)

Depth (mm)

10-year exposure to deicing salts

Field data (measured on 3 test spans)

Prediction of Year-10 chloride profilebased on Cs & Dc measured at 10 yrs

Prediction of Year-10 profilebased on Cs & Dc measured at 5 yrs

Prediction of Year-10 profilebased on Cs & Dc measured at 2 yrs

Depth of main reinforcement

Fig. 9. Predictions of 10-year chloride profiles based on different measurement times.

0

5

10

15

20

25

30

35

40

45

50

55

60

0 1 2 3 4 5 6 7 8 9 10 11 12

Tim

e (y

ears

)

Chloride Threshold (kg/m3)

- Corrosion initiation(from measured Dc & Cs)

- Spalling (Icor=0.25μA/cm2)


CEB1992

Rebars at 75 mm:

Rebars at 25 mm:



Field observations after 10yrs:- at 75mm: no app. corrosion;- at 25mm: light corrosion,

no spalling yet.

- Corrosion initiation(from measured Dc & Cs)

Fig. 10. Sensitivity analysis of time to rebar corrosion and time to concrete spalling.

a time to corrosion initiation of 15 years based on thethreshold value suggested by CEB (1992). No significantcorrosion was in fact observed on sections of reinforcingbars cut from the barrier wall after 10 years (Cusson andQian, 2009).

The literature shows a strong disagreement amongresearchers on the range of values to use for the chlo-ride threshold of conventional reinforcing steel in con-crete. For example, the following values have been re-ported: 0.6–4.9 kg/m3 (Stratfull et al., 1975), 0.7–5.25

kg/m3 (Vassie, 1984), 6.3–7.7 kg/m3 (Lukas, 1985), 0.87kg/m3 (West and Hime, 1985), and 1.05–1.75 kg/m3

(Henriksen, 1993). There are many reasons explainingthese differences, including the variability of severalconcrete properties and exposure conditions, such aspH level, w/c, type and content of cementing materials,temperature, relative humidity (Tuutti, 1993), as well asthe ambiguity associated with the definition of chloridethreshold (Alonso et al., 2000). Another factor affect-ing the prediction accuracy is the high variability of the


Table 4Mean values and coefficients of variation used in the probabilistic analysis

Parameter Mean COV (%) Description of data used for calculation of mean and COV values

Cs (kg/m3) See Table 3 (C. 2), or Figure 6 14 16 measurements taken in barrier wall after 10 years (excludingtest sections with surface coating/sealer)

Dc (cm2/year) See Table 3 (C. 2), or Figure 6 28 16 measurements taken in barrier wall after 10 years (excludingtest sections with surface coating/sealer)

Cth (kg/m3) 1.8 30 No field data; mean value suggested by CEB (1992); assumingCOV for high variability

dc (mm) 75 20 4 measurements along height of barrier wallS (mm) 230 4 57 longitudinal measurements between stirrups of barrier wall

(see Figure 11a)Icor (μA/cm2) 0.25 40 52 measurements taken in barrier wall at cracks after 10 years

(see Figure 11b)

corrosion rate, which depends on many factors, includ-ing the location on the reinforcement relative to a crack,and the yearly variations of ambient temperature, oxy-gen content, and ambient relative humidity (Shiessl andRaupach, 1990). Over a year, the corrosion rate maybe expected to reach maximum values during the warmseason and minimum values during the cold season, andpossibly negligible values if the concrete is frozen.

At a depth of 75 mm in the barrier wall, the mod-els predicted concrete spalling to occur after 20 years ofexposure (Figure 10), based on the moderate corrosionrate (Icor) of 0.50 μA/cm2 (Cox et al., 1997) and on theCEB (1992) chloride threshold value. In fact, the meancorrosion rates measured after 10 years in the bridgebarrier walls (near cracks) were 0.25 μA/cm2 for the75 mm deep main reinforcement and 0.30 μA/cm2 forthe 25 mm deep test rebars. With this field data, andassuming a minimum chloride threshold of 1.8 kg/m3,the models predicted the onset of spalling after at least25 years for the 75 mm deep bars.

As mentioned above, the considerable uncertaintyand variability associated with the key parameters gov-erning the service life of RC bridges under corrosiveenvironments require the use of probabilistic analysismethods for a reliable prediction of their remainingservice life. The prediction models presented above inSection 2 require input data for different parametersrelated to the properties of concrete, steel, the environ-ment and geometry of the reinforcement. From the dataacquired during the field study at the Vachon bridge, itwas possible to estimate the average values and COVof many key parameters of interest. This information ispresented in Table 4. It can be seen that different pa-rameters have quite different COVs. For instance, barspacing had the smallest COV of 4% because it is re-lated to geometry, thus to the quality of workmanship.The corrosion rate had the largest COV of 40% as it

varies widely in space and time due to a number of dif-ferent factors as mentioned earlier. Figure 11 presentshistograms of the distribution of rebar spacing and cor-rosion rate measured in the barrier wall. From this in-formation, the assumption of normal distribution in aprobabilistic analysis is found reasonable.

Finally, Figure 12 presents the probabilities of steelcorrosion and concrete spalling for depths of 25 mmand 75 mm in the bridge barrier wall. For these cal-culations, the models (Equations 1–12) were used andfed with input values determined from the field data ob-tained after 10 years of exposure to deicing salts. Themean value of the chloride threshold to initiate rein-forcement corrosion was assumed to be 1.8 kg/m3 (con-verted from the CEB 1992 threshold value of 0.4% bymass of cement). Also, one of the curves determined bythe models presented in this article was recalculated us-ing the well-known Monte Carlo simulation techniqueusing 50,000 iterations. The purpose of this compari-son was to verify the validity of the advanced first-orderreliability method (Section 2.5) for solving nonlinearproblems.

As expected, the results show that as time increases,the probabilities of initiating corrosion and concretespalling increase, and more rapidly for the thinner con-crete cover and for the higher assumed corrosion rateof the steel reinforcement. As mentioned before, theseprobabilities may be used by bridge inspectors to assessthe condition of the deck by adapting to existing guide-lines (e.g., AASHTO, 2002) and to decide upon the besttime to take corrective actions (local repair, major reha-bilitation, or replacement). For example, if the bridgeowner’s policy is to never let a bridge deck (or any partof it) deteriorate more than condition state 3 (Table 1),the maximum probability of spalling could be set to 25%before undertaking any major corrective actions. In thiscase, a bridge deck, having concrete and environment


0

5

10

15

20

25

30

200 210 220 230 240 250 260

Fre

qu

ency

Rebar spacing (mm)

0

5

10

15

20

25

0.0

0

0.0

5

0.1

0

0.1

5

0.2

0

0.2

5

0.3

0

0.3

5

0.4

0

0.4

5

0.5

0

Fre

qu

ency

Corrosion rate (μA/cm2)

Fig. 11. Histograms of selected parameters measured in barrier wall: (a) rebar spacing between stirrups; (b) current density attransverse shrinkage cracks after 10 years.

0.0

0.2

0.4

0.6

0.8

1.0

0 10 20 30 40 50 60 70 80 90 100

Pro

ba

bili

ty o

f co

rro

sio

n / s

pa

llin

g

Time (years)

75-mm cover

25-mmcover

Initiation of steel corrosion(Cs = 22.7 kg/m3; Dc = 0.63 cm2/yr; Cth = 1.8 kg/m3)

Onset of concrete spalling (Icor = 0.50 μA/cm2)

Onset of concrete spalling (Icor = 0.25 μA/cm2)

First-order reliability method (all solid lines)

Monte Carlo simulation (with 50,000 iterations)

Fig. 12. Estimated probability of steel corrosion and concrete spalling for concrete cover depths of 25 mm and 75 mm.

that are similar to those of the investigated concrete bar-rier wall, would reach condition state 3 after 18 years ifthe actual corrosion rate was 0.25 μA/cm2, or 14 yearsif the corrosion rate was 0.50 μA/cm2. An optimal de-cision could then be made by selecting the preservationaction that yields the lowest life cycle cost (Lounis andDaigle, 2008).

4 SUMMARY AND CONCLUSIONS

The corrosion-induced deterioration of highway bridgescan have serious consequences in terms of reducedsafety, serviceability, and durability. This article pro-

posed a probabilistic modeling approach based on dura-bility monitoring for improving the life cycle perfor-mance predictions of aging concrete bridge decks builtin corrosive environments. Its application and benefitswere demonstrated on a case study of rebuilt RC barrierwalls on a highway bridge near Montreal, Canada. Thefollowing conclusions and recommendations are sug-gested:

• The mechanistic models of the proposed approach re-quire input data on four main variables (i.e., surfacechloride content, chloride diffusion coefficient, chlo-ride threshold, and corrosion rate) for the determi-nation of different limit states along the life cycle ofa bridge deck, including early-age cracking, onset of


corrosion, internal/surface cracking, spalling, and de-lamination.

• The above variables and other key parameters (e.g.,concrete cover thickness) do vary considerably inspace and/or time, and some of these cannot be mea-sured accurately. Relatively high coefficients of vari-ation between 20% and 40% have been measured inthe field.

• Given the considerable uncertainties associated withthese parameters governing the service life of RCbridges exposed to chlorides, the use of probabilisticanalysis methods are required.

• The advanced first-order reliability method is a sim-ple and practical approach that enables the modelingof uncertainties in service life prediction and requiresa minimum amount of data. The validity of this sim-plified approach was demonstrated by a comparisonwith the Monte Carlo simulation method.

• In the case study, it was shown that some of the keyinput data that are commonly used in service life pre-diction models could be very different from realitydue to their high variability and uncertainty. For ex-ample, values of apparent chloride content in con-crete were measured up to 23 kg/m3, which is about2.5 times higher than the maximum values suggestedin the literature for the most severe exposure to de-icing salts.

• It was also shown that values of apparent surfacechloride content in concrete increase over time al-though values of apparent chloride diffusion coef-ficient decrease over time, contrasting with somecommon assumptions used in service life modeling.However, it was shown that the use of constant valuesmeasured at the age of 10 years could provide com-parable estimates of service life compared to modelsusing varying values.

• It was demonstrated that service life predictionscould be improved significantly by updating the mod-els with selected field monitoring data from embed-ded sensors or on-site corrosion surveys, as opposedto using selected data from the literature. The pro-posed approach can be used on any RC elements ofbridge decks as long as the governing corrosion pa-rameters could be monitored on site and fed to theprobabilistic mechanistic prediction models.

• Recommendations were provided for the applica-tion of the proposed approach to a given network ofbridges. A two-level decision process based on twotypes of deterioration models was suggested, in whichcritically damaged bridges are first identified by us-ing simplified Markovian cumulative damage mod-els, and then individually analyzed using the proposeddurability monitoring and probabilistic mechanisticmodeling approach.

REFERENCES

AASHTO. (1998), Guide for Commonly Recognized (CoRe)Structural Elements, American Association of State High-way and Transportation Officials, Washington DC, 54 pp.Rev. 2002

Alonso, C., Andrade, C., Castellote, M. & Castro, P. (2000),Chloride threshold values to depassivate reinforcing barsembedded in a standardized OPC mortar, Cement and Con-crete Research, 30(7), 1047–55.

Andrade, C. & Alonso, C. (2001), On-site measurements ofcorrosion rate of reinforcements, Construction and Build-ing Materials, 15(2–3), 141–5.

ASTM C114. (1999), Standard Test Methods for ChemicalAnalysis of Hydraulic Cement, American Society for Test-ing and Materials, West Conshohocken, PA, 30 pp.

Bamforth, P. B. (1998), Spreadsheet model for reinforcementcorrosion in structures exposed to chlorides, in O. E. Gjørv,K. Sakai, and N. Banthia (eds.), Concrete under Severe Con-ditions 2: Environment and Loading, Vol. 1, E & FN Spon,London, pp. 64–75.

Bazant, Z. P. (1979), Physical model for steel corrosion in con-crete sea structures—theory, Journal of the Structural Divi-sion, ASCE, 105(ST6), 1137–53.

Boulfiza, M., Sakai, K., Banthia, N. & Yoshida, H. (2003), Pre-diction of chloride ions ingress in uncracked and crackedconcrete, ACI Materials Journal, 100(1), 38–48.

Braunstein, J., Ruchala, J. & Hodac, B. (2002), Smart struc-tures: fiber-optic deformation and displacement monitor-ing, in 1st International Conference on Bridge Maintenance,Safety and Management, IABMAS, Barcelona, Spain, 8 pp.

Butcher, A. M. & Newhook, J. P. (2009), Long-term monitor-ing of the confederation bridge, in 4th International Confer-ence on Structural Health Monitoring of Intelligent Infras-tructure (SHMII-4), Zurich, Switzerland, 8 pp.

Cardini, A. J. & DeWolf, J. T. (2009), Long-term structuralhealth monitoring of a multi-girder steel composite bridgeusing strain data, Structural Health Monitoring, 8(47), 47–58.

Carrion, F. J., Lopez, J. A., Quintana, J. A. & Orozco, P. R.(2009), Structural monitoring of cable stayed bridge duringrehabilitation, in 4th International Conference on StructuralHealth Monitoring of Intelligent Infrastructure (SHMII-4),Zurich, Switzerland, 7 pp.

CEB. (1992), Design Guide for Durable Concrete Structures,2nd edn., Thomas Telford Publishers, London, UK.

Choi, H.-H. & Seo, J. (2009), Safety assessment usingimprecise reliability for corrosion-damaged structures,Computer-Aided Civil and Infrastructure Engineering, 24,293–301.

Cox, R. N., Cigna, R., Vennesland, O. & Valente, T.; COST509. (1997), Corrosion and Protection of Metals in Contactwith Concrete, Final report, European Commission, Direc-torate General Science, Research and Development, Brus-sels, EUR 17608 EN, ISBN 92-828-0252-3, 148 pp.

Crank, J. (1975), The Mathematics of Diffusion, 2nd edn. Ox-ford University Press, London, 414 pp.

Cruz, P. J. S. & Salgado, R. (2009), Performance of vibration-based damage detection methods in bridges, Computer-Aided Civil and Infrastructure Engineering, 24(1), 62–79.

Cusson, D. & Qian, S. (2009), Ten-year field evaluation of cor-rosion inhibiting systems in concrete bridge barrier walls,ACI Materials Journal, 106(3), 291–300.


Cusson, D., Qian, S. & Hoogeveen, T. (2006), Field perfor-mance of concrete repair systems on highway bridge, ACIMaterials Journal, 103(5), 366–73.

De Schutter, G. (1999), Quantification of the influence ofcracks in concrete structures on carbonation and chloridepenetration, Magazine of Concrete Research, 51(6), 427–35.

Dhir, R. K., Jones, M. R., Ahmed, H. E. H. & Seneviratne, A.M. G. (1990), Rapid estimation of chloride diffusion coeffi-cient in concrete, Magazine of Concrete Research, 42, 177–85.

FHWA. (2009), iFlorida Model Deployment Final EvaluationReport, Report No. FHWA-HOP-08-050, 274 pp.

Francois, R. & Arliguie, G. (1999), Effect of microcrackingand cracking on the development of corrosion in reinforcedconcrete members, Magazine of Concrete Research, 51(2),143–50.

Frangopol, D. M., Strauss, A. & Kim, S. (2008), Bridge reli-ability assessment based on monitoring, Journal of BridgeEngineering, ASCE, 258–70.

Gerard, B. & Marchand, J. (2000), Influence of cracking onthe diffusion properties of cement-based materials. Part I:influence of continuous cracks on the steady-state regime,Cement and Concrete Research, 30(1), 37–43.

Glisic, B. & Inaudi, D. (2007), Fibre Optic Methods for Struc-tural Health Monitoring, John Wiley & Sons, Ltd., WestSussex, UK, 262 pp.

Hasofer, A. M. & Lind, N. C. (1974), Exact and invariantsecond-moment code format, Journal of the EngineeringMechanics Division, ASCE, 100(EM1), 111–21.

Henriksen, C. F. (1993), Chloride corrosion in Danish bridgecolumns, in L.-O. Nilsson (ed.), Chloride Penetration intoConcrete Structures, Goterborg, Sweden.

Huang, R. Y., Mao, I. S. & Lee, H. K. (2010), Exploringthe deterioration factors of bridges: a rough set theory ap-proach, Computer-Aided Civil and Infrastructure Engineer-ing, 25(7) 517–29.

Inaudi, D., Bolster, M., Deblois, R., French, C., Phipps, A., Se-basky, J. & Western, K. (2009), Structural health monitor-ing system for the new I-35W St. Anthony Falls Bridge, in4th International Conference on Structural Health Monitor-ing of Intelligent Infrastructure (SHMII-4), Zurich, Switzer-land, 9 pp.

Kropp, J. & Hilsdorf, H. K. (1995), Performance Criteria forConcrete Durability, E&FN SPON, London.

Li, C.-Q., Mackie, R. I. & Lawanwisut, W. (2007), A risk-costoptimized maintenance strategy for corrosion-affected con-crete structures, Computer-Aided Civil and InfrastructureEngineering, 22, 335–46.

Lounis, Z. & Cohn, M. Z. (1995), An engineering approach tomulticriteria optimization of bridge structures, Computer-Aided Civil and Infrastructure Engineering, 10(4), 233–8.

Lounis, Z. & Daigle, L. (2008), Reliability-based decision sup-port tool for life cycle design and management of highwaybridge decks, in Annual Conference of the TransportationAssociation of Canada, Toronto, 19 pp.

Lounis, Z., Daigle, L., Cusson, D. & Almansour, H. (2009), Amulti-objective approach for the management of aging crit-ical highway bridges, in Aging Infrastructures Workshop,Columbia University, New York City, 20 pp.

Lounis, Z. & Madanat, S. M. (2002), Integrating mechanis-tic and statistical deterioration models for effective bridgemanagement, 7th ASCE International Conference on Ap-plications of Advanced Technologies in Transportation,Cambridge, MA, 513–20.

Lounis, Z. & Vanier, D. J. (2000), A multi-objectiveand stochastic system for building maintenance manage-ment, Computer-Aided Civil and Infrastructure Engineer-ing, 15(5), 320–9.

Lukas, W. (1985), Relationship between chloride content inconcrete and corrosion in untensioned reinforcement onAustrian bridges and concrete road surfacings, Betonwerkund Fertigteil-Technik, 51(11), 730–4.

Madsen, H. O., Krenk, S. & Lind, N. C. (1986), Methods ofStructural Safety, Prentice-Hall Inc., Englewood Cliffs, NJ.

Melchers, R. E. (1999), Structural Reliability and Prediction,John Wiley and Sons, Chichester.

Moaveni, B., Conte, J. P. & Hemez, F. M. (2009), Uncertaintyand sensitivity analysis of damage identification results ob-tained using finite element model updating, Computer-Aided Civil and Infrastructure Engineering, 24(5) 320–34.

Mufti, A. (2001), Guidelines for structural health monitoring,ISIS Manual No. 2, http://www.isiscanada.com/manuals/No2 contents.pdf.

Rodriguez, O. G. & Hooton, R. D. (2003), Influence of crackson chloride ingress into concrete, ACI Materials Journal,100(2), 120–6.

Sarma, K. & Adeli, H. (1998), Cost optimization of con-crete structures, Journal of Structural Engineering, ASCE,124(5), 570–8.

Sengul, O. & Gjorv, O. E. (2008), Electrical resistivity mea-surements for quality control during concrete construction,ACI Materials Journal, 105(6), 541–7.

Shiessl, P. & Raupach, M. (1990), Influence of concrete com-position and microclimate on the critical chloride content inconcrete, in C. L. Page, K. W. J. Treadaway, and P. B. Bam-forth (eds.), Corrosion of Reinforcement in Concrete, Soci-ety of Chemical Industry, Elsevier Applied Science, Lon-don, UK, pp. 49–58.

Sirca, G. & Adeli, H. (2005), Cost optimization of prestressedconcrete bridges, Journal of Structural Engineering, ASCE,131(3) 380–8.

Soyoz, S. & Feng, M. Q. (2009), Long-term monitoring andidentification of bridge structural parameters, Computer-Aided Civil and Infrastructure Engineering, 24(2), 82–92.

Stewart, M., Faber, M. H. & Gehlen, C. (2003), Temporal andspatial aspects of probabilistic corrosion models, in FIB-JCSS Workshop on Probabilistic Modeling of DeteriorationProcesses in Concrete, Lausanne, Switzerland, 269–78.

Stratfull, R. F., Jurkovich, W. J. & Spellman, D. L. (1975),Corrosion testing of bridge decks, Transportation ResearchRecord, No. 539, pp. 50–9.

Tang, L. & Gulikers, J. (2007), On the mathematics of time-dependent apparent chloride diffusion coefficient in con-crete, Cement and Concrete Research, 37(4), 589–95.

Tennyson, R. C., Mufti, A. A., Rizkalla, S., Tadros, G. & Ben-mokrane, B. (2001), Structural health monitoring of innova-tive bridges in Canada with fiber optic sensors, Smart Mate-rials and Structures, No. 10, pp. 560–73.

Tepfers, R. (1979), Cracking of concrete cover along anchoreddeformed reinforcing bars, Magazine of Concrete Research,31(106), 3–12.

Timoshenko, S. P. (1956), Strength of Materials. Part II: Ad-vanced Theory and Problems, 3rd edn., D. Van NostrandCompany, Princeton, NJ.

Tuutti, K. (1982), Corrosion of Steel in Concrete, Report No.4, Swedish Cement and Concrete Research Institute, Stock-holm, Sweden.


Tuutti, K. (1993), The effect of individual parameters on chlo-ride induced corrosion, in L.-O. Nilsson (ed.), ChloridePenetration into Concrete Structures, Goteborg, Sweden,pp. 18–25.

U.S. DOT, FHWA, FTA. (2007), 2006 Status of the Nation’sHighway Bridges, and Transit: Conditions & Performance,Report to Congress.

Vanmarke, E. (1983), Random Fields, MIT Press, Cambridge.Vassie, P. (1984), Reinforcement corrosion and the durability

of concrete bridges, in Proceedings of the Institution of CivilEngineer, Part I, No. 76, Issue 3, pp. 713–23.

Vorechovska, D., Podrouzek, J., Chroma, M., Rovnanıkova,P. & Teply, B. (2009), Modeling of chloride concen-

tration effect on reinforcement corrosion, Computer-Aided Civil and Infrastructure Engineering, 24, 446–58.

Wenzel, H. (2009), From structural health monitoring to riskbased infrastructure management, in 4th International Con-ference on Structural Health Monitoring of Intelligent In-frastructure (SHMII-4), Zurich, Switzerland, July 22–24,13 pp.

West, R. E. & Hime, W. G. (1985), Chloride profiles in saltyconcrete, Material Performance, 24(7), 29–36.

Weyers, R. E. (1998), Service life model for concrete struc-ture in chloride laden environments, ACI Materials Journal,95(4), 445–53.

Durability Monitoring for Improved Service Life ...karaj/GPSC/Wk4EngineeringHL.pdfing structural...

Documents

Transcript of Durability Monitoring for Improved Service Life ...karaj/GPSC/Wk4EngineeringHL.pdfing structural...