Académique Documents
Professionnel Documents
Culture Documents
ir)
www.elsevier.com/locate/ress
PII:
DOI:
Reference:
S0951-8320(14)00084-2
http://dx.doi.org/10.1016/j.ress.2014.04.016
RESS5036
To appear in:
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Declaration:
I, David BIGAUD, Senior Author of the paper, declare that the work therein described is original, is not
currently submitted to any other journal, and will not be so submitted so long as it is under consideration by
Reliability Engineering & System Safety.
This manuscript is:
A paper for a special issue on Accelerated Life Testing of Reliability Engineering & System Safety journal
Documents submitted:
1, Cover letter (RESS_Cover letter_David BIGAUD.pdf)
2, Highlights (RESS_Highlights_David BIGAUD.pdf)
3, Manuscript (RESS_Manuscript_David BIGAUD_Special_Issue_Accelerated_Testing.pdf)
4, Tables (RESS_Tables_David BIGAUD.pdf)
5, Figures (RESS_Figures_David BIGAUD.pdf)
Sincerely
David BIGAUD
Corresponding author
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
LARIS Laboratory, Angers University, 62 avenue Notre Dame du Lac, F-49000 Angers, France.
Abstract
Time-variant reliability analysis of RC highway bridges strengthened with carbon fiber reinforced polymer CFRP
laminates under four possible competing damage modes (concrete crushing, steel rupture after yielding, CFRP
rupture and FRP plate debonding) and three degradation factors is analyzed in terms of reliability index using
FORM. The first degradation factor is chloride-attack corrosion which induces reduction in steel area and concrete
cover cracking at characteristic key times (corrosion initiation, severe surface cover cracking). The second
degradation factor considered is fatigue which leads to damage in concrete and steel rebar. Interaction between
corrosion and fatigue crack growth in steel reinforcing bars is implemented. The third degradation phenomenon is
the CFRP properties deterioration due to aging. Considering these three degradation factors, the time-dependent
flexural reliability profile of a typical simple 15 m-span intermediate girder of a RC highway bridge is constructed
under various traffic volumes and under different corrosion environments. The bridge design options follow
AASHTO-LRFD specifications. Results of the study have shown that the reliability is very sensitive to factors
governing the corrosion. Concrete damage due to fatigue slightly affects reliability profile of non-strengthened
section, while service life after strengthening is strongly related to fatigue damage in concrete.
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
1. Introduction
RC structures behave inadequately over the time when placed in severe or aggressive environments. Such
environments could dramatically damage the concrete structure and decrease its durability. Thus, considering
damage in degrading reinforced concrete structures is of major interest in the context of reliability through the
service life analysis and maintenance. Due to numerous sources of uncertainty in the degradation models, a
probabilistic approach is suitable. Important degradation mechanisms are concrete carbonation [1] and corrosion of
the rebars due to chloride ingress [2]. This mechanism greatly influences the ageing of bridge structures submitted
to deicing salts, marine environment or under heavy traffic. The aim of the present article is to merge these
degradation influences, and especially corrosion, with fatigue effects.
Chlorides-induced corrosion that affects the reinforcing steel in concrete is a major problem facing civil
structures. In the last two decades degradation due to corrosion had been extensively studied. It has been achieved
from these studies that, corrosion causes reduction in steel area, reduction of steel properties - area, yield strength,
and ultimate strain (as reported by Cairns et al. [3] and Almusallam [4]), losses of bond between concrete and
steel bars, and, cracking and spalling of concrete cover (from Liu [5] and Li et al. [6]). Palsson and Mizra [7]
reported that up to 80% loss of cross-section areas for reinforcing a 40-years-old Canadian bridge demolished in
1999.
As far as fatigue damage is concerned, it mainly causes modifications of the concrete mechanical properties. As a
consequence, in the compression zone, strains in concrete fibers are permanently increased with the applied cycles
of live load while the stiffness is decreased as mentioned by El-Tawil et al. [8] and Zanuy et al. [9]. Such a
continuous deterioration leads to strengthening decision.
The strengthening of existing RC bridges is one of the current problems faced by the structural engineers when the
strength evaluation indicates a decrease in flexural or shear strength. In general, strengthening takes place as the
structural performance deteriorates throughout the service lifetime of the structure. In the recent years, the need for
strengthening or rehabilitation of reinforced concrete structures had been increased, due to the aging of these
structures, damage of these structures, modifications in design codes, or to support higher vehicle loads. In the past,
additional number of dowelling bars embedded in a new concrete volume adjacently casted to the old concrete
element was used as strengthening solution. More recently, steel plates attached to the concrete surface of tension
zone using adhesive materials or bolts have been proposed as alternative of the previous solution. Even more
recently, carbon fibre reinforced plastic CFRP materials have been used for the purpose of strengthening.
Laminates are externally bonded to extreme concrete tensile surface to produce additional bending strength. CFRP
strengthening provides many advantages comparing to the two former techniques such as ease of installation,
strength to weight ratio, high resistivity to exposure conditions, etc. ACI committee 440 [10] concludes in a 10160% increase of resistance when strengthened with CFRP laminates, even if a reduction in ductility requirements
ranges between (5-40%) may be observed.
The strengthening of an element eventually permits to enhance its reliability, often expressed in terms of
margin between load and resistance. This measure of the structure reliability has been used for many years and it
has been considered in design of structural systems that both loads and strengths are deterministic values. However,
deterministic approaches are not efficient for decisions taken under uncertainty induced from high loading level
(i.e., structure is close to the ultimate limit state) or when performances are dramatically reduced during time. A
more useful assessment can be provided by adopting probabilistic approaches which are used in safety analysis.
Here, reliability of structural systems can be defined as the probability that the structure under consideration has a
proper performance (strength, serviceability, or other limit state) throughout its lifetime.
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Reliability of non-strengthened concrete section damaged by corrosion or fatigue was discussed in many previous
publications (Val et al. [11], Petryna et al. [12], Betti et al. [13], Elachachi et al. [14], El-Hassan et al. [15], Marsh
and Frangopol [16]). The main objective of this article is to perform a time-variant reliability analysis for RC
bridge girders subjected to a coupled fatigue-corrosion deterioration process and strengthened with FRP of Carbon
fibers. Two types of corrosions were considered: general and pitting corrosions. The later were coupled with crack
growth that may take place due to fatigue effect on steel reinforcing bars.
The concrete girder has to be strengthened when the reliability index of the girder is deteriorated to a minimum
reliability index (min=3). The CFRP section is designed in order that the RC section, once repaired, reaches a target
reliability index T (3.8). Before CFRP strengthening installation, it is considered that corroded steel bars are fully
cleaned. Corrosion mechanism after strengthening continues from diffusion of new amounts of chlorides from
girder sides because the lower concrete surface is totally covered by CFRP laminates. Flexural ultimate limit state
with multiple competing failure modes is taken into account: concrete crushing, steel rebars rupture after yielding,
CFRP mid span debonding and CFRP rupture.
This approach has proven its accuracy to take into account the material non-linearity and has been validated in the
framework of static tests performed on several RC beams reinforced by composites. The same principle has been
applied by Bigaud et al. [20] for the case of pre-stressed RC beams and Plevris and Triantafillou [21] for modelling
the creep behaviour of RC beams reinforced with FRP. They implement the delayed properties of each constitutive
material to assess the beam section equilibrium. More recently Ferrier et al. [22] have improved the modelling by
taking into account the decreasing mechanical properties of all the constitutive materials subjected to fatigue (steel,
concrete, polymer adhesive and composite). The authors calculate the minimum and the maximum applied stresses
within each material under static initial conditions and consider that these stresses levels remain constant all along
the fatigue cycles. This simplified approach does not explicitly consider the cumulated damage and the actual
changing internal stresses during cycles. We correct this lack in this paper updating the repartition of stress within
materials every 10a cycles (a is an implementing natural number taken from 0 to 3). min and max correspond to
dead loads and combination of dead loads in addition to live loads respectively. Ultimate concrete strain which
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
decreases due to plastic strain exert from fatigue loading is calculated for the concrete layer that has maximum
compression strain.
The concrete constitutive model is the one described in Table 1. The steel bars present an elastic-plastic behavior
and the FRP shows a linear elastic one.
Table 1. Constitutive models of concrete, steel and FRP (and its interface with concrete).
2.2
Corrosion effects
As shown in figure 2, corrosion deterioration process can be divided into three stages; initiation, propagation
before surface cover cracking and propagation after cover cracking. This deterioration process results from
variation in corrosion current parameters through each stage and has been detailed by several authors (Stewart and
Rosowsky [23], Bastidas-Arteaga et al. [24], Yuan, Ji and Jiang [25]).
The service time duration of girders depends eventually on the two characteristic times tini and tsp corresponding
respectively to the corrosion initiation and the severe surface cover cracking times. Derivations of these two
characteristic times are provided in appendix. For time t tini+tsp, (ie after cracking), reduction of steel rebar radius
is observed.
For general corrosion (see legend "general" in figure 2), Vu and Stewart [26] and Bastidas-Arteaga et al. [27] have
proposed to evaluate the average losses of steel rebar radius with the following equation:
t
(1)
t ini
where, icorr(t) is the current density of steel corrosion cell (also derived in appendix).
Different variables, used for intermediate calculations leading to the expression of icorr(t), contribute to the
statistical nature of pav(t) (variables dst, Cs, Cth, Dcl,ref seen in appendix) Also, Vu and Stewart [26] have added that
it is necessary to multiply icorr(t) by a bias factor icorr (model error) of 1 with CoV taken at 0.2.
Equation 1 is valid for losses due to general corrosion since losses are uniform in both bar length and perimeter. As
far as pitting corrosion (see legend "pit shape" in figure 2) is concerned, losses impact particular small areas on bar
surface. Val and Trapper [28] distinguish two types of pitting corrosion along rebars length in a site survey of
repaired bridges; coarse (20-100mm) and fine (2-4mm) pits. In the present study, we considered the fine pits only.
Spherical pit configuration shown in figure 2 was used to determine the reduced cross-sectional area of pit area
(Apit). Maximum pit depth was taken equal to p(t)= pav(t), where is the pitting factor experimentally obtained.
This factor is time dependent and takes a value greater than unity (4 to 8 depending on the bar diameter). It will be
taken as probabilistic with a mean value of 7.1 in our further study (see 3.2).
It is worth mentioning that, corrosion did not affect steel area only, but its actions also extend to change steel
mechanical properties with time. It can be assumed that, yield strength is linearly proportional to the reduced crosssectional area As(t) such that (from [3]):
A (t )
f y ( t ) = 1 y s f yo
Aso
(2)
where, fyo is the steel yield stress (that will be considered as probabilistic parameter), Aso is initial bar area, and y is
an empirical coefficient.
Cairns et al [3] review twelve experimental studies which report an average value of y up to 0.017. Based on these
experimental data, average value of y equal to 0.012 will be considered.
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
f
c ,tot ( N ) = cu
E sec
0. 5
c ,tot ( N ) =
1.11 f cu
E sec
N
1 + 0.667
Nf
(3a)
(3b)
Where, Smax is the ratio of maximum stress to concrete strength (= max/fcu), N is the number of cycles, Esec is the
initial secant modulus (= max/c,0), c,0 is the concrete strain of the first cycle, tN is the load duration in hours.
In Eq.3, fatigue life Nf necessary to determine c,tot(N) is evaluated here according to Hsu [31] as:
max
= 1 0.0662(1 0.556 R )log N f + 0.0294 log T0
f cu (N )
(4)
Where R = c ,min c ,max and T0 is the cycle duration (sec per cycle).
Last term of Eq(3), Sc is the characteristic stress level expressed as:
S c = (S max + S min ) 2 + 1 2 2
(5)
Where, Smin is the ratio of minimum stress to concrete strength (= min/fcu). Maximum and minimum stresses are
related to total load (dead load + live load) and dead load only respectively.
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
c , p ( N ) = c ,tot ( N ) c ,e with c ,e =
max
Ec ,N
(6)
c ,ult ( N ) = ult ,0 c , p ( N )
(7)
N
Ec ,N = 1 0.33
Nf
.Ec ,1
(8)
Figure 3b. The uniform reduction of these characteristics is described by the model X = 1 D X , fat .X o . The
deterioration coefficient D X , fat may be expressed through a linear damage rule such as:
m
D X , fat =
k =1
nk ( k )
N f ( k )
(9)
where k represents the stress amplitude, nk the number of load cycles within the corresponding stress range
and N f the fatigue life. Dusicka, Itanib and Buckle [32] have observed that despite minor individual differences,
the overall fatigue strain life for all grades of steel exhibited similar behaviour.
As mentioned in 2.2, pitting corrosion initiated losses in steel area in terms of a small pit nucleation. This pit
propagates with time due to corrosion. However, another effect takes place based on the formed pit due to fatigue.
In this cases fatigue tends to increase the pit area based on fatigue crack growth concept leading to an excessive and
rapid cracking of steel bar. In order to know either the corrosion pit depth p(t) or fatigue crack growth size a(t) is
the governing degradation mode, two rates must be estimated. The first is the maximum pit corrosion rate which is
estimated as:
dp (t )
= 0.0116icorr (t )
dt
(10)
The second rate is the fatigue crack growth rate which is estimated by using Paris-Erdogan law:
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
da( t ) dt = f .C p .(K (t , a ))
(11)
Where, f is the load frequency (or the truck daily traffic), Cp and m are material coefficients related to crack growth
stage; at surface Cp=3.83x10-29 MPam.m-m/2+1 and m=20.863, and, through the volume Cp=3.16x10-12 MPam.m-m/2+1
and m=3.143. The stress intensity factor K has to be considered as dependent of time and crack size [27].
The crack growth (t), either by pitting corrosion or fatigue, depends on the competition between corrosion pit and
fatigue crack growth rates (see figure 4). At early stage, crack growth is governed by pitting corrosion (for t < tpt) as
dp(t)/dt>>>da(t)/dt, thus (t)=p(t). When da(t)/dt is greater than dp(t)/dt equals p(t) (for t = tpt), crack growth
becomes governed by fatigue and thus (t) = a(t). The time of transition corrosion pit to fatigue crack, tpt (see
figure 4), is defined as the time at which the crack growth rate reaches the maximum pit depth rate due to
corrosion. Such this definition can be estimated by solving numerically the following equation:
0.0116icorr (t pt ) = f .C p .(K (t pt , a ))
(12)
icorr(t) being statistical, it will contribute to the probabilistic nature of crack and pit rates.
2.3.3 Fatigue effects on FRP and its interface with concrete
As far as the FRP composite is concerned, effects of fatigue consist in modulus EFRP decrease (figure 3c) as
observed by Ferrier et al. [33]. According to the authors, deterioration coefficient DE FRP , fat can be expressed as
DEFRP , fat = 1 0.071.log10 ( N ) . The ultimate tensile strain FRP,ult being assumed constant during time by the
authors, strength fFRP will decrease with cycles according to the expression fFRP = FRP,ult.EFRP(N) =
FRP,ult.EFRPo. DE FRP , fat (where EFRPo is the initial FRP composite modulus).
Concerning the interface FRP to concrete, effects of fatigue results in the reduction of the maximum normal load
supported by the FRP N FRP ,t max ( N ) as a function of applied cycles (figure 3d). Yao et al. [34] have shown that
this reduction is a consequence of the deterioration of the concrete compressive strength f cu ( N ) due to fatigue.
Nevertheless, considering in addition that FRP modulus will decrease with cycles, we finally notice that fatigue
will either induce a decrease or an increase of maximum strain to debonding debonding ( N ) under the comparative
variations of N FRP ,t max ( N ) and EFRP ( N ) within Equation (13):
debonding ( N ) =
N FRP ,t max ( N )
(13)
In our case, FRP Young modulus decreasing more rapidly than the maximum normal load - for the specific
calculated levels of stress within each constituent -, we eventually observe an increase of the maximum strain to
debonding.
debonding(N) will be probabilistic by nature since parameters such as fFRPo or EFRPo are statistical.
2.4 Effect of CFRP durability
2.4.1 Environmental durability only
The long-term performances of CFRP laminates, either modulus or ultimate strength, are affected by field
conditions (humidity, temperature, method of installation, curing, etc). Karbhari and Abanilla [35] have formulated
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
a simplified durability model based on Arrhenius acceleration law for ageing effects. The long-term performance
P(t) can thus be expressed in the general following form :
P( t )
B = DP ,ageing B = A.ln(t )
P0
(14)
With the aim to express the effect of moisture on the degradation of any grade of CFRP, the same authors propose a
three-step approach to formulate the long-term performance model. The first step consists in carrying out
accelerated tests; the performances of CFRP samples subjected to constant immersion in de-ionized water are
measured. The evolution of performance as a function of % gain of moist weight absorbed by CFRP specimen (P =
f(%moist)) is obtained. The second step consists in following the time-evolution of the moisture profile within the
same grade of CFRP under natural conditions (real conditions of moisture) and assuming a Fickian response. Thus,
this second step leads to express a % gain of moist weight under real moisture exposure as a function of time
(%moist = f(t)). From these two steps, it is finally possible to express the evolution of performance as a function of
time under real moist conditions (P = f(t)). In our subsequent reliability analysis study, we will consider the same
CFRP configuration as Karbhari and Abanilla [35], i.e. wet layup carbon/epoxy for external strengthening. For this
one, the authors give the following expression for the long-term ultimate strength fFRP(t) and modulus EFRP(t):
f FRP ( t )
(15a)
with A f FRP = - 0.03366 and B f FRP = 1.0607 and where, fFRPo is the initial CFRP ultimate strength, t is the time in
days.
EFRP ( t )
(15b)
with AEFRP = -0.0418 and BEFRP = 1.0607 and where EFRP0 is the initial CFRP modulus.
Here the values of constants B f FRP and BEFRP are not equal to the unity because the parameters of the formula
were obtained by fitting experimental data. This is considered acceptable as an assessment of predictive accuracy
of the model shows reliable predictions of the model especially for long periods of exposure conditions.
Since parameters fFRPo and EFRPo are probabilistic, fFRP(t) and EFRP(t) will show the sale nature.
Atadero et al. [36], in their study about the variability of CFRP properties (thickness tp, modulus EFRP and strength
fFRP), conclude that FRP modulus and strength are statistically correlated to CFRP thickness. CFRP properties used
in this study and reported in Table 3 were statistically de-correlated from its thickness by the authors. CFRP
strength and modulus statistical distribution were fitted respectively with Lognormal and Weibull distributions.
According to five data sets which were reported in [36, 37] it can be observed that the CFRP number of layers has
an insignificant effect on EFRPo, while fFRPo decreases with respect to number of layers by 0, 2, 11, and 23% (as an
average between the data sets reported) for 1, 2, 3, and 4 layers respectively. Since we assume in our study that the
strengthening is ensured by two FRP layers, we further consider a decrease of 2% for fFRP. The same authors also
show that the number of layers slightly affects the mean and coefficient of variation of CFRP modulus and
strength.
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
(16a)
(16b)
This simple approach does not consider any coupling between ageing and fatigue effects.
These empirical expressions can be questionable in the sense that degradation parameters D E FRP , fat , D f FRP , fat ,
D E FRP , ageing and D f FRP , ageing may be negative. But this problem happens at N=1.2x1015 cycles for D E FRP , fat and
D f FRP , fat , and, at t = 7.9x105 and 3.7x107 years respectively for D E FRP , ageing and D f FRP , ageing . Thus, we will
consider this empirical models as relevant.
The flow diagram of figure 5 illustrates how the design capacity of CFRP-strengthened RC-beams is estimated
considering all the deterioration models.
Figure 5. Scheme of corrosion-fatigue deterioration process in RC structures
c (t ) m ,cc . c ,ult (t )
(17a)
m,cc is the model error for concrete crushing. Ali [46] has shown through a comparison with an important database
of experimental results that this error follows a normal distribution, with a bias of 1.066 and a CoV of 0.121.
The second is steel rupture after yielding that may occurred after fatigue and combined effect of corrosion
which reduces the steel area -:
s (t ) s ,ult (t )
(17b)
(17c)
m,fr is the model error for FRP rupture. Ali [46] has shown this error follows a normal distribution, with a bias of
1.045 and a CoV of 0.102.
And the fourth is debonding (end or intermediate crack debonding) of FRP from the concrete surface when force
applied to the FRP reaches the maximum allowable force NFRP,debonding(t). This fourth ULS is condidered to be more
frequent than FRP rupture:
(17c)
m,fd is the model error for FRP rupture. Ali [46] has shown this error follows a normal distribution, with a bias of
1.023 and a CoV of 0.106.
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Chen and Teng [47] propose an analytical model for the maximum FRP laminates applied force N FRP ,debonding (t )
that causes debonding:
f cu (t )
(18a)
Where,
Le (t ) =
E FRP (t )t FRP
f cu (t )
; w =
1
2 bFRP b c
L
; l =
sin
1 + bFRP bc
2 Le
L Le
L < Le
(18b-c-d)
With bc the girder width, tFRP, the FRP laminate width. L is the bond length. We have compared this debonding
model to experimental results and have obtained a high correlation close to the unity (=0.958). This model gives a
correlation higher than the debonding model reported in ACI Committee 44 [10] which gives a correlation of about
0.938.
G( X i ( t )) = m R( t ) S ( t ) = m R( X 1 ( t ), X 2 ( t ),.....X n ( t )) ( DL DL + LL LL( t ))
(19)
Where, R(t) is the random section resistance function of system of random variables (geometric and material
properties), S(t) is the applied random loads. m, DL, and LL are random variables that correspond to uncertainties
existing in resistance model (ie model error), dead loads, live loads respectively. Plevris et al [21] propose to
consider a bias value taken at 1.01 and a coefficient of variation at 0.046.
The above limit state formulation shows its evolution as a function of time. In our paper, the time-dependent
reliability analysis results in distinct studies at consecutive time t or cycles N increments considering the evolution
of each variable Xi (e.g. icorr(t), fcu(N), EFRP(N), fy(t), etc) detailed in 2.
The calculation of reliability index requires the transformation of the random variable vector {X} from the
physical variable space to standard, centred and independent variables as explained by Lemaire et al. [48]. Thus,
the limit state G(Xi(t)) must be redefined, using the transformation u i = T ( X i ( t )) , as
H (u i ( t )) = G T 1 ( X i ( t )) . Each corresponding variable ui in the standardized space are non-dimensional and are
represented with a standard Gaussian distribution (with u=0 and u=1).
The structural safety can be estimated in terms of reliability index (or of probability of failure Pf(t)) which can be
determined from the optimization problem:
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
( t ) = 1 (Pf ( t )) = min
u
i =1
2
i
(20)
There are several algorithms to search the reliability index such as first order algorithms (ie Hasofer-LindRackwitz-Fiessler, HLRF, first order method), second order algorithms (ie Newton method and sequential
quadratic programming method) and hybrid algorithms. All these algorithms are based on iterative processes that
require the calculation of partial derivatives of the performance function H(u).
In the present study FORM algorithm is used to solve the optimization problem. It is the most frequently used in
the literature and permits to identify directly the contribution of each factor to the safety index, as it is the simplest
method and the most frequently used in the literature.
Noting u* the optimal solution, importance or sensitivity coefficients can be obtained:
i =
u i
(21a)
u =u *
{ } = {H (u * )}
*
( )
H u
(21b)
Sensitivity factors have two major purposes. First, the sign of the sensitivity factor shows the relationship between
the performance function and the physical variables. For example, a positive i means that the performance
function H(u1,,ui, un) decreases as the random variable ui increases. Second, these sensitivity factors show the
contributions of the random variables to the safety-index or failure probability. The physical meaning of sensitivity
factor implies the relative contribution of each random variable to the reliability index .
For each reliability simulation, it is necessary to reduce the number of parameters taken among a set of 32 variables
available (the 12th of Table 2 and the 20st of Table 3), otherwise computing time is too long and results are
questionable. It is question to select which are the variables to be considered as deterministic or probabilistic. For
this, in the cases where we have reliable information about the distribution laws of variables (otherwise we must
therefore consider by defaults the deterministic nature of the variable), we simulate their effects on the variability
of reliability index through an a priori sensitivity analysis. This analysis consists in calculating the value of
importance coefficient i = -/ui of variable ui considering that all other parameters u k i are deterministic. This
a priori analysis certainly does not show the potential interactions between variables but is a good indicator so as to
select which are the parameters to be considered as probabilistic or deterministic. We will finally consider
separately the deterministic variables with low values of sensitivity (12 variables of Table 2) and the probabilistic
variables that may claim a high sensitivity (20 variables of Table 3), whatever the case considered. Indeed,
according to the level of chloride concentration, to if the calculation considers the RC beam before strengthening,
or after strengthening, to the level of traffic frequency, the value of sensitivity may vary. But a set of 20
probabilistic variables is too large to carry out a relevant reliability study. Thus a second-step analysis is carried out
in order to classify the sensitivity coefficients of variables in order of importance. An excerpt of a sensitivity
analysis is shown in 3.4.
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
After strengthening, lower steel rebars are considered to be fully cleaned (chloride concentration through the cover
returns to zero), as concrete cover are considered to be repaired or changed. Chloride attack is considered to be
reinitiated from the sides of the girder and stopped at girder bottom. It will be considered that, strengthening time
corresponds to the time required for reliability index to reach a minimum allowed value (min = 3). From other
side, strengthening has to increase reliability index to a value greater than a target value (T = 3.8). CFRP laminates
width will be taken equal to beam width. CFRP laminates thickness will be taken equal to 7 mm. It will be
considered that, live load growths annually in magnitude by 0.005 and in frequency by 0.023 [9]. f (truck/day) can
be considered as a fraction of the total traffic flow (from 10 to 20%), this fraction depends on the class (rural,
urban etc) of highway AASHTO [49].
3.2.2 Reliability data
Considering both the above mentioned sensitivity analysis and a complete literature review, we have selected what
are the variables to be considered as deterministic or statistical. Tables 2 and 3 report this selection. Distribution
laws are specified for each statistical variable.
In order to consider dependency between concrete compressive strength fc and modulus Ec, Ec is expressed via fc as:
Ec = Ec ( 0.1 f c )2 / 3
(22)
It can be considered that fc and coefficient Ec are independent random variables [11, 36].
Atadero et al [36] study the variability of CFRP properties (thickness tp, modulus EFRP, strength fFRP). The study
concludes that, CFRP modulus and strength are statistically correlated to CFRP thickness. CFRP strength and
modulus statistical distribution were fitted by the authors. It was proved that, lognormal and Weibull distributions
are the best descriptor for CFRP modulus and strength respectively. CFRP properties used in this study and
reported in Table 3 were statistically decorrelated from their thickness.
As mentioned in 2.2 and in appendix, corrosion process is influenced by temperature and relative humidity
conditions. We consider for the simulations average monthly profiles for temperature and relative humidity in the
ranges (20 to 40C) and (0.6 to 0.75) respectively (see figure 7). We assume these profiles are constant all along
lifetime in the reliability analysis (i.e., for example, that the temperature for month mm the year y will be the same
for year y+1, and so on). The chosen profiles is relevant to cities located in Arabic Golf (e.g. Jeddah which is
located on the west coast of Saudi Arabia), which is characterized by its moist and hot conditions.
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Different values of the surface chloride concentration Cs ranging from 2 to 4 kg/m3 were considered. The
coefficient between average steel area loss for uniform corrosion and for pitting corrosion is considered to follow
a Gumbel law with a mean at 7.1 and a CoV of 0.11 [50].
Different values of truck numbers per day (f) were considered. Truck duration T0 is taken to 0.05 (seconds per
cycle).
Figure 7. Temperature (C) and Relative Humidity year profiles.
Figure 9. Reliability profile for strengthened section and Cs = 3 kg/m3 (no fatigue).
The strengthening by FRP increases highly the section reliability (see figure 9). This strengthening is done when a
target value of =3 is reached. The time required to reach this target value is defined as the time-to-repaired (TTR).
The increasing of reliability mainly relies on the optimization of the FRP thickness. It was found that strengthening
with 7 mm thickness of FRP laminates may increase reliability of the concrete section by about 1.3 () for both
general and pitting corrosion (with Cs =3 kg/m3).
After repair and up to the 90th year, the decrease of reliability is mainly related to the live load increase along time
and is almost linear. The loss of reliability index is around
-1.31x10-2 year-1. During this period, FRP debonding is the governing failure mode. A change of slope is observed
after 90th year (slope becomes -4.02x10-2 year-1). This is due to the fact that concrete crushing progressively
becomes the governing failure mode.
Other results are shown in Table 4. It illustrates the increase in service life (years) and reliability index for
strengthening section with 7 mm of FRP laminates. It can be observed that strengthening may double service life of
section.
Table 4. Increase in reliability index and service life.
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Figure 12. Reliability profile considering complementary fatigue crack growth effects combined with pitting corrosion
(pitting corrosion with Cs = 3 kg/m3 + fatigue of concrete + fatigue crack growth effect on steel)
Figure 13. Same reliability profile than figure 12 but considering now additional effect of fatigue of FRP (pitting corrosion
with Cs = 3 kg/m3 + fatigue of concrete + fatigue crack growth effect on steel + fatigue of FRP)
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Figure 14. Same reliability profile than figure 13 but considering now additional effect of FRP environmental degradation
(pitting corrosion with Cs = 3 kg/m3 + fatigue of concrete + fatigue crack growth effect on steel + fatigue of FRP + FRP
durability)
Table 5. Variables importance coefficients from TTR up to failure (or end of simulation at 100 years) at different traffic
frequencies f [truck/day] and with Cs=3 kg/m3
3.5 Discussion
The proposed reliability models could show the long-term beaviour and the deterioration scenarios included in FRP
strengthening of RC beam subjected to fatigue-corrosion damage effects. The proposed model considers two types
of corrosion: pitting and uniform. Moreover, fatigue effects were considered in the concrete, FRP and steel. A
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
coupled effect of both fatigue and pitting corrosion was also performed. The proposed model was performed using
FORM method which enables to show the importance of the considered random variables.
Finally, long-term predictions of bridge performance are inherently uncertain since it is very difficult to predict
with any degree of confidence long-term deterioration processes and increasing loads over the time. Hence,
prediction of the effect of deterioration processes on bridge performance can only be relatively accurate for
relatively short time periods: e.g. Regular intervals not exceeding 2 years as recommended by National Bridge
Inspection Standard (NBIS) [51]. However, over these short periods serviceability performance may change
dramatically
Conclusions.
The present study constitutes an elementary step towards a perspective reliability research for CFRP strengthening
of RC elements affected by corrosion and fatigue damages of RC elements.
In order to illustrate the model, an interior girder of a bridge located in different corrosion environments was
studied. Time variant reliability analysis of flexural limit of the girder was performed. Two deterioration factors
were considered: corrosion of steel reinforcement and fatigue. Fatigue damages were evaluated deterministically
using the fatigue serviceability limit state described in AASHTO specifications. Corrosion model includes
corrosion initiation, and propagation. Two types of corrosion were considered: pitting and uniform. Fatigue was
considered to affect the compression zone of the RC section, steel bars and FRP material after strengthening.
Fatigue crack growth based on the pit zone, which has been formed due to pitting corrosion was considered in the
model. Time to FRP strengthening was defined as the time that the reliability index reaches a minimum value
equals to 3.Four failure modes were considered: concrete crushing, steel failure after yielding, FRP rupture and
FRP intermediate crack debonding. In what concerns end failure of FRP plate, they were neglected in the model as
they can be totally omitted using anchoraged system.
For non-strengthened RC section, corrosion has the greatest effect on ultimate limit state especially pitting
corrosion. General corrosion has greater effect on section reliability at initial stages of the studied age (100 years).
An excessive damage is found at latest stages for pitting corrosion. It is possible to affirm that time to strengthening
(time for = min) is approximately more than 40 and 60 years for pitting and uniform corrosion when considering
and neglecting live load growth. Regarding the effect of fatigue of RC section, it was remarked that fatigue effect
have a slight effect on the corroded RC beam; either for pitting or uniform corrosion. However, catastrophic failure
for RC section was observed in the reliability profile when coupling fatigue crack growth with pitting corrosion.
This can be explained due to the excessive losses in steel are due to the coupling effect.
Applying FRP strengthening to the corroded beams, the observed failure mode after strengthening is the debonding
failure mode. In contrast, the other considered failure modes (concrete crushing of FRP rupture) were not observed.
It was found that strengthening efficiency period after strengthening with reliability index greater than 3 is
significantly affected by the amount of live load frequency. In addition, the excessive losses in steel area due to
coupling effect was totally omitted, as the stress intensity of steel bars decreases dramatically due to FRP
strengthening. FRP deterioration due to fatigue has a slight effect on reliability profile of strengthened section,
while durability almost returns a negligible effect.
The most perspective advantage of the proposed model is that it can show the effect of fatigue damage in
compression zone on the reliability profile. It was found that the reliability profile after strengthening changes
significantly with respect to load frequency, as the service life after strengthening decreases as the load frequency
increases. This decrease takes place due to the losses in the ultimate failure strain of concrete leading to a
transformation of the failure mode after strengthening for FRP debonding to concrete crushing.
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Appendix.
As shown in figure 2, corrosion deterioration process can be divided into three stages; initiation, propagation
before surface cover cracking and propagation after cover cracking.
The service time duration of girders depends eventually on the two characteristic times tini and tsp corresponding
respectively to the corrosion initiation and the severe surface cover cracking times. These times depends on many
parameters that are cited below. These parameters can present a great variability and have a high sensitivity on the
characteristic times. Some of them are then considered as statistical in this article.
Derivation of corrosion initiation time tini.
Time to initiate corrosion (tini) is the time required for the chloride ions concentration at the concrete cover level
C(c,tini) to reach a threshold value Cth at the contact surface between concrete and steel. Bastidas-Arteaga et al. [24]
propose a value of 0.90 kg/m3 for Cth. From resolution of the 1D Fick second law of diffusion process, it is
possible to formulate the chloride concentration inside concrete C(x,t) as a function of time (t) and depth (x) from
concrete surface and thus to obtain tini (a detailed derivation can be found in [15])
4 DCl
(A.1)
where, c is the clear concrete cover, erf is the error function, Cth is the threshold value of chloride concentration at
initiation time, Cs is the chloride concentration on concrete surface and Dcl is the chloride diffusion coefficient.
This later parameter depends on many others. To enlighten the paper, we direct the reader toward Val and Trapper
[28] so as to catch the influence of temperature, time and relative humidity and toward Vu and Stewart [26] to see
the effect of concrete mix proportions (aggregate to cement ratio, water to cement ratio), mass densities of cement
and aggregates, water diffusion coefficient in an infinite solution. These parameters are summed up in Tables 2 and
3.
Cs depends on the distance from construction site to the coast (d in km) as expressed in the following empirical
model [27]:
2.95 kg/m 3
d < 0.1km
d > 2.84
0.35
(A.2)
wc ( t ) =
4d s ( t )
( 1 c )( a / b ) + ( 1 + c )( b / a )
2bf ct
Eef
(A.3)
where, c is the Poisson ratio of concrete and Eef (=1/c+1) is the effective elastic modulus of concrete. (1) is the
tangential stiffness reduction factor, which is used to account for residual tangential stiffness of cracked concrete
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
and is related to the average tangential strain at the cracked surface and concrete properties. fct is the concrete
tensile strength.
To understand what represent a, b and ds(t), we must rely once again on the work of Li et al. [6] who have
experimentally verified that embedded bars and the surrounding concrete behave as a thick ring as shown in figure
2, where D is the bar diameter and do is the thickness of porous band around steel bar.
The inner and outer radii of thick ring are a= (D+2do)/2 and b=c+a respectively. Once corrosion process
starts, the corrosion rust layer formed around the steel bar fills the porous band do, the growth of rust layer
generates radial compression stresses which initiate radial cracks (see figure 2.c). The thickness of rust layer
required to start compression radial stresses can be determined as [5]
ds ( t ) =
Wr ( t ) 1
r
( d + d o ) rust st
(A.4)
where r is a coefficient related to corrosion type (assumed to be equal to 0.57 in [5]), r is specific weight of
corrosion products (in kg/m3), st is specific weight of steel, and Wr(t) is the amount of rust products (in kg/m).
As the rust layer grows thicker, the ionic diffusion distance increases, and the rate of rust production decreases
because the diffusion is inversely proportional to the oxide thickness. The rate of rust production can be written as
follows according to Liu [5]:
kp
dWr (t )
=
dt
Wr (t )
(A.5)
t
Wr ( t ) = 2 0.105( 1 / r )Dicorr ( t )dt
0
(A.6)
Finally, the current density of steel corrosion cell icorr can be expressed based on the Faradays law of
electrochemical equivalence [15]:
icorr ( t ) =
(A.7)
where, nO2 is the obtained electric number of O2 molecule participating in chemical reaction ( nO2 =4), F is the
Faradays constant, CO2 is the oxygen concentration on the surface ( CO2 =8.93.10-10 mol/mm3), x(t) is the depth in
concrete media that has chloride concentration equal to Cth at time t tini, and DO2 (t ) is the oxygen diffusion
coefficient in concrete. For t tini+tsp , DO2 (t ) is considered as constant before cover cracking by Papadakis et al.
[52] and can be expressed as:
'
(A.8)
Where, f ck' (= f cu 8) is the characteristic concrete cube compressive strength. It is proposed to reduce value
given by equation 7 by 50% for concrete exposed to relative humidity around 70% [49]. It is worth mentioning that
as fcu follows a normal distribution, the oxygen diffusion coefficient will follow a lognormal one.
It can be observed from equation A.7 in the second stage that corrosion current decreases with increasing the
depth of critical chloride concentration x(t) (see figure A.1). The described corrosion model simulates the corrosion
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
process in the nature since corrosion rate continues to decrease due to formation of rust layer around the corroded
surface. This layer expands with time, thus transportation of oxygen and moist is gradually delayed leading to
reduction in corrosion rate [25].
d0
D
ds(t)
r0
Ps(t)
Ps(t)
radial crack
c
(a)
(b)
(c)
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
References.
[1]
Sudret B. Probabilistic models for the extent of damage in degrading reinforced concrete structures. Reliability
Engineering and System Safety, 2008; 93: 41022.
[2]
Choe D-E., Gardonia P., Rosowskya D., Haukaas T. Probabilistic capacity models and seismic fragility estimates for
RC columns subject to corrosion. Reliability Engineering and System Safety, 2008; 93: 38393.
[3]
Cairns J, Plizzari. G A, Du Y, Law D W, Franzoni C. Mechanical Properties of Corrosion-Damaged Reinforcement.
ACI material journal, 2005; 102(4): 256-64.
[4]
Almusallam, A. A. Effect of degree of corrosion on the properties of reinforcing steel bars. Construction and Building
Material, 2001; 15: 361-68.
[5]
Liu Y. Modeling the Time-to-Corrosion Cracking of the Cover Concrete in Chloride Contaminated Reinforced
Concrete Structures. Ph. D thesis, Faculty of the Virginia Polytechnic Institute and State University, Virginia, USA; 117 pages,
1996.
[6]
Li C, Melchers R E, Zheng, J. Analytical model for corrosion-Induced Crack Width in Reinforced Concrete
Structures. ACI Structural Journal, 2006; 103(4): 479-87.
[7]
Palsson, R., Mizra, M. S. Mechanical response of corroded steel reinforcement of abandoned concrete bridges. ACI
structural journal, 2002; 99(2): 157-62.
[8]
El-Tawil A, Ogunc C, Okeil A, Shahawy M. Static and Fatigue Analysis of RC Beams Strengthened with CFRP
Laminates. Journal of composites for construction, 2001; 5(4): 258-67.
[9]
Zanuy C, Albajar L, Pablo de la Fuents. Sectional analysis of concrete under fatigue loading. ACI Structural Journal,
2009; 106(5): 667-677.
[10]
ACI Committee 440. ACI 440.2R-02: Guide for the design and construction of externally bonded FRP system for
strengthening concrete structures. Farmington Hills, MI : ACI 2002.
[11]
Val D V, Stewart M G, Melchers R E. Life-Cycle Performance of RC Bridges Probabilistic Approach. ComputerAided Civil Infrastructure Engineering, 2000; 15: 14-25.
[12]
Petryna Y S, Pfanner D, Stangenberg F, Krtzig W B. Reliability of reinforced concrete structures under fatigue.
Reliability Engineering and System Safety, 2002; 77: 253-261.
[13]
Betti, R., West, A. C., Vermaas, G., and Cao, Y. 2005. Corrosion and embrittlement in high-strength wires of
suspension bridge cables. Engineering Bridges, 2005; 10(151): 12p: doi: 10.1061/(ASCE)1084-0702(2005)10:2(151)
[14]
Elachachi, M., Breysse, D., Yotte S., Cremona, C. A probabilistic multi-scale time dependent model for corroded
structural suspension cables. Probabilistic Engineering Mechanics, 2006; 20(3): 235-45.
[15]
El-Hassan, J., Chateauneuf, A., and Bressolette, P. H. 2008. Effet du couplage des conditions climatiques sur la
corrosion des poutres en bton arm. Compte-rendu des 26mes rencontres Universitaires de Gnie Civil, Nancy, June 2008.
[16]
Marsh P., Frangopol D. Reinforced concrete bridge deck reliability model incorporating temporal and spatial
variations of probabilistic corrosion rate sensor data. Reliability Engineering and System Safety, 2008; 93: 394-409.
[17]
Soares RC, Mohamed A, Venturini WS. Reliability analysis of nonlinear reinforced concrete using the response
surface method. Reliability Engineering System Safety, 2002; 75:13756.
[18]
Neves R.A., Chateauneuf A., Venturinia W.S., Lemaire M. Reliability analysis of reinforced concrete grids with
nonlinear material behavior. Reliability Engineering and System Safety, 2006; 91: 73544.
[19]
Varastehpour, H. and Hamelin, P., "Strengthening of concrete beams using fiber-reinforced plastics" Materials and
Structures, RILEM, 1997; 30; 160-166.
[20]
Bigaud D., Ferrier E., Hamelin P., Beam Compo Design: computer-aided software for the design of reinforced and
pre-stressed concrete beams strengthened by composite materials, Proceedings of Advanced Composite Material in Bridges
and Structures 3, Ed. Jagmohan Humar, A. Ghani Razaqpur, 2000; 405-412.
[21]
Plevris N., Triantafillou T.C, Time dependent behavior of RC members strengthened with FRP laminates, Journal of
Structural Engineering, 1994; 120(3): 1016-1042.
[22]
Ferrier E., Bigaud D., Hamelin P., Bizindavyi L., Neale K. W. Fatigue Reliability of External Bonded CFRP Used for
Concrete Structure, Materials and Structures, 2005; 38(5) No. 275: 39 46.
[23]
Stewart, M. G. and Rosowsky, D. V. Time-dependent reliability of deteriorating reinforced concrete bridge decks.
Structural Safety, 1998; 20: 91-109.
[24]
Bastidas-Arteaga E, Sanches-Silva M, Chateauneuf A, Silva M R. Coupled reliability model of biodeterioration,
chloride ingress and cracking for reinforced concrete structures. Structural Safety, 2008; 30: 110-29.
[25]
Yuan Y, Ji Y, Jiang J. Effect of corrosion layer of steel bar in concrete on time variant rate. Materials and Structures,
2009; 42: 1443-50.
[26]
Vu K A T, Stewart M G. Structural reliability of concrete bridges including improved chloride-induced corrosion
models. Structural safety, 2000; 23: 313-33.
[27]
Bastidas-Arteaga E, Bressolette P, Chateauneuf A, Sanchez-Silva M. Probabilistic lifetime assessment of RC
structures under coupled corrosion deterioration processes. Structural safety, 2009; 31: 84-96.
[28]
Val D V, Trapper P A. Probabilistic evaluation of initiation time of chloride-induced corrosion. Reliability
Engineering and System Safety, 2008; 93: 364-372.
[29]
Barnes, R.A., and Mays, G.C. Fatigue performance of concrete beams strengthened with CFRP plates, Journal of
Composites for Construction, ASCE, 1999; 3(2): 63-72.
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
[30]
Holmen J.O. Fatigue of concrete by constant and variable amplitude loading. Fatigue of concrete structures 1982;
American Concrete Institute, Detroit: 71-110.
[31]
Hsu, T. T. C. Fatigue of Plain Concrete. ACI journal, 1981; 78(4): 292-304.
[32]
Dusicka P., Itanib A.M., Buckle I.G. Cyclic response of plate steels under large inelastic strains. Journal of
Constructional Steel Research, 2007; 63: 156164.
[33]
Ferrier E., Bigaud D., Clment J.C., Hamelin P. Fatigue-loading effect on RC beams strengthened with externally
bonded FRP. Construction and Building Materials, 2011; 25(2): 539-46.
[34]
Yao J., Teng J.G., Chen J.F. Experimental study on FRP-to-concrete bonded joints. Composites: Part B, 2005; 36:
99113.
[35]
Karbhari V M, Abanilla M A. Design factors, reliability, and durability predictions of wet layup carbon/epoxy used in
external strengthening. Composites Part B, 2007; 38: 10-23.
[36]
Atadero A A, Lee L, Karbhari V M. Consideration of material variability in reliability analysis of FRP strengthened
bridge decks. Composites Structures, 2005; 70: 430-43.
[37]
Atadero, A. A. and Karbhari, V. M. Sources of uncertainty and design values for field-manufactured FRP. Composites
structures, 2009; 89: 83-93.
[38]
Steward M.G. Probability of serviceability failure of reinforced concrete office floor beams. In: Lemaire, Favre,
Mebarki, editors. Applications of statistics and probability. Rotterdam: Balkema, 1995. p.277-283.
[39]
Quan Q., Gengwei Z. Calibration of reliability index of RC beams for serviceability limit state of maximum crack
width. Reliability Engineering and System Safety, 2002; 75: 359-66.
[40]
Picard A., Massicote B. Serviceability design of prestressed concrete bridges. Journal of Bridges Engineering, 1999;
4(1): 48-55.
[41]
Ashour, A.F., El-Refaie, S.A., and Garrity, S.W. Flexural strengthening of RC continuous beams using CFRP
laminates. Cement & Concrete Composites, 2004; 26: 765-775.
[42]
Bogas, J.A. and Gomes, A. Analysis of the CFRP flexural strengthening reinforcement approaches proposed in fib
bulletin 14. Construction and Building Materials, 2008; 22(10): 2130-2140.
[43]
Aram, M.R. and Czaderski, C. Debonding failure modes of flexural FRP-strengthened RC beams. Composites: Part B,
2008; 39: 826-841.
[44]
Pham, H. P. and Al-Mahidi, R. Reliability analysis of bridge beams retrofitted with fibre reinforced polymers.
Composite Structures, 2008; 82: 177-84.
[45]
Teng, J.G., Smith, S.T., Yao, J., and Chen, J.F. Intermediate crack-induced debonding in RC beams and slabs.
Construction and Building, 2003; 17: 447-462.
[46]
Ali O. Time-dependent reliability of FRP strengthened reinforced concrete beams under coupled corrosion and
changing loading effects. A thesis submitted at Angers University, October 2012, 325p.
[47]
Chen, J. F. and Teng, J.G. Anchorage strength model for FRP and steel plates bonded to concrete. Journal of
Structural Engineering, 2001; 127 (7): 784-91.
[48]
Lemaire, M., Chateauneuf, A., and Mitteau, J. C. Fiabilit des structures, couplage mcano-fiabiliste statique. Paris
France by Hermes Science. 2005; 506p.
[49]
American Association of state of Highway Transportations Officials. AASHTO LRFD Bridge Design Specifications.
2002, Washington DC: AASHTO.
[50]
Stewart, M. G., and Al-Harthy, A. Pitting corrosion and structural reliability of corroding RC structures: Experimental
data and probabilistic analysis. Reliability Engineering and System Safety, 2008; 93: 373-82.
[51]
Chen, W., and Duan, L. Bridges Engineering: Construction and Maintenance. Taylor & Francis Group, LLC, United
States. 2003; 236p.
[52]
Papadakis, V. G., Vayenas, C. G., Fardis, N. M. Physical and chemical characteristics affecting the durability of
concrete. ACI Material Journal, 1991; 8(2):186-196.
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
List of Tables
Table 1. Constitutive models of concrete, steel and FRP (and its interface with concrete).
Table 2. Material constant (deterministic).
Table 3. Probabilistic parameters of the variables.
Table 4. Increase in reliability index and service life.
Table 5. Variables importance coefficients from TTR up to failure (or end of simulation at 100 years) at
different traffic frequencies f [truck/day] and with Cs=3 kg/m3
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Table 1. Constitutive models of concrete, steel and FRP (and its interface with concrete).
Variable.
Es
c
p
c
do
steel
r
r
c
a
ac
wc
Value
201300
22
18
0.2
12.5x10-3
78.5
36
0.57
29
26
5.14
0.45
Units
MPa
kN/m3
kN/m3
--mm
kN/m3
kN/m3
--kN/m3
kN/m3
-----
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Variable
fcu
Biasa
1
1.066
1
Covb
0.15
0.121
0.08
Ec
distribution
Normal
Normal
Normal
units
MPa
-----
nominal
35
1
2.15.104
fyo
dst
Cs
Cth
Normal
Normal
Lognormal
Uniform
MPa
mm
kg/m3
kg/m3
0.1
0.05
0.5
Dcl,ref
Lognormal
cm2/s
500
1
1212
1
2-4
1
Range (0.6 to 1.2)
with Cov of 0.19
2.10-8
1
icorr
Normal
Lognormal
Normal
Gumbel
Lognormal
Weibull
Normal
Normal
Normal
Normal
Normal
Normal
--mm
mm
--GPa
MPa
-----
1
40
Eq. 3
7.1
71
802.46
1
1
0.2
0.15
0.2
0.17
0.11
0.11
0.102
0.106
0.1
0.25
0.18
0.1
m,cc
c
wc
EFRP
fFRP
m,fr
m,fd
DL
SL
LL
IL
Depends on analysis
1
1
1
1
1
1
1.045
1.023
1.05
1
1
1
0.75
Unifo
rm
Increase in
Time To Repair TTR (year)
Total service life (year)
Increase in
Time To Repair TTR (year)
Total service life (year)
Pittin
g
Corrosion type
4
1.389
35.43
63.77
1.384
40.51
74.12
Table 5. Variables importance coefficients from TTR up to failure (or end of simulation at 100 years) at
different traffic frequencies f [truck/day] and with Cs=3 kg/m3
ui
f
50
fcu
fy
dst
Ec
EFRP
fFRP
m,cc
m,fr
0.09:0.06
0.58:0.49
0.03:0.03
0.08:0.05
0:-0.04
0.01:0
0:0.23
0:0
500
0.07:-0.05
0.59:0.06
0.07:0.03
0.15:0.06
0.06:-0.03
0:-0.01
0.01:0.25
0:0
1000
ui
f
50
0.04:-0.07
0.59:0
0.08:0.03
0.18:0.08
0.06-0.02
0.01:-0.01
0.01:0.27
0:0
m,fd
icorr
LL
IL
DL
SL
0.17:0.01
0:0
0.05:0.06
-0.73:-0.81
-0.05:-0.05
-0.12:-0.14
-0.08:-0.09
500
0.18:0.01
0:0
0.05:0.07
-0.72:-0.92
-0.05:-0.06
-0.13:-0.14
-0.08:-0.09
1000
0.18:0
0:0
0.06:0.08
-0.72:-0.93
-0.05:-0.06
-0.13:-0.15
-0.09:-0.10
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Figures
Nc
neutral
axis
s
FRP
FRP
strain
NS
NFRP
stress
force
Figure 1. Fiber section of RC section strengthened with FRP plates is discretized in several layers.
> 0.3 mm
Pit shape
t ini
corrosion losses
t pt
fatigue
losses
p(t)
a(t)
Time
Figure 4. Crack rate and pit rate competition.
General
Composite
Concrete
c.
a.
fFRP
frp
min
max
fcu
decrease of Ec
with N
p(N)
FRP,ult
decrease of EFRP
with N
}curves N0
curve N=0
curve N0
curve N=0
Co(N=0) C,ult(N=0)
tot(N)
e
Interface
Steel
d.
b.
ratio
fu
fy
S,ult
Ncyc
NFRP,max(N 0) / NFRP,max(N=0)
debonding(N 0) / debonding(N=0)
So
decrease of fy
and fu with N
}curves N0
curve N=0
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Reliability index
5
4,5
corrosion
initiation
4
3,5
Cs=2 kg/m
3
2,5
2
Cs=3 kg/m
1,5
1
general
pitting
0,5
0
0
20
40
60
Years
Figure 8. Reliability profile for non strengthened section (no fatigue).
80
100
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
400 mm
dlane2
= 0.9m
250mm
300 mm
bFRP=300mm
825mm
14
25mm
112@150
bs
4320mm
20mm
dp=
1200mm
ds=800mm
Lane 3 width 3600 mm
bc=400mm
c
2200 mm
1800 mm
dlane1
= 0.9m
160 mm
300 mm
250 mm
Figure 6. Bridge cross-section and trucks position - Reinforcements and dimensions details of the interior beam.
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
4,5
concrete
crushing
Reliability index
4
corrosion
initiation
3,5
strengthened
3
repairing
2,5
2
1,5
1
general
pitting
0,5
non-strengthened
0
0
20
40
60
80
100
Years
Figure 9. Reliability profile for strengthened section and Cs = 3 kg/m3 (no fatigue).
5
strengthened
4,5
Reliability index
4
corrosion
initiation
3,5
3
repairing
2,5
2
1,5
non-strengthened
50 truck/day
500 truck/day
0,5
1000 truck/day
0
0
20
40
60
80
100
Years
Figure 10. Reliability profile for fatigue of concrete added to general corrosion effects (general corrosion with Cs = 3 kg/m3 +
fatigue of concrete)
5
strengthened
4,5
4
corrosion
initiation
Reliability index
3,5
3
repairing
2,5
2
1,5
50 truck/day
non-strengthened
500 truck/day
0,5
1000 truck/day
0
0
20
40
60
80
100
Years
Figure 11. Reliability profile for fatigue of concrete added to pitting corrosion effects (pitting corrosion with Cs = 3 kg/m3 +
fatigue of concrete)
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
5
corrosion
initiation
4,5
strengthened
Reliability index
3,5
3
repairing
2,5
2
1,5
non-strengthened
1
50 truck/day
0,5
500 truck/day
1000 truck/day
0
0
20
40
60
80
100
Years
Figure 12. Reliability profile considering complementary fatigue crack growth effects combined with pitting corrosion (pitting
corrosion with Cs = 3 kg/m3 + fatigue of concrete + fatigue crack growth effect on steel)
5
strengthened
4,5
4
corrosion
initiation
Reliability index
3,5
3
repairing
2,5
2
1,5
50 truck/day
non-strengthened
500 truck/day
0,5
1000 truck/day
0
0
20
40
60
80
100
Years
Figure 13. Same reliability profile than figure 12 but considering now additional effect of fatigue of FRP (pitting corrosion
with Cs = 3 kg/m3 + fatigue of concrete + fatigue crack growth effect on steel + fatigue of FRP)
5
4,5
Reliability index
4
3,5
3
2,5
2
1,5
0,5
0
0
20
40
60
80
100
Years
Figure 14. Same reliability profile than figure 13 but considering now additional effect of FRP environmental degradation
(pitting corrosion with Cs = 3 kg/m3 + fatigue of concrete + fatigue crack growth effect on steel + fatigue of FRP + FRP
durability)
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Tables
Table 1. Constitutive models of concrete, steel and FRP (and its interface with concrete).
Variable.
Es
c
p
c
do
steel
r
r
Value
201300
22
18
0.2
12.5x10-3
78.5
36
0.57
Units
MPa
kN/m3
kN/m3
--mm
kN/m3
kN/m3
---
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
c
a
ac
wc
kN/m3
kN/m3
-----
29
26
5.14
0.45
Variable
fcu
Biasa
1
1.066
1
Covb
0.15
0.121
0.08
Ec
distribution
Normal
Normal
Normal
units
MPa
-----
nominal
35
1
2.15.104
fyo
dst
Cs
Cth
Normal
Normal
Lognormal
Uniform
MPa
mm
kg/m3
kg/m3
0.1
0.05
0.5
Dcl,ref
Lognormal
cm2/s
500
1
1212
1
2-4
1
Range (0.6 to 1.2)
with Cov of 0.19
2.10-8
1
icorr
Normal
Lognormal
Normal
Gumbel
Lognormal
Weibull
Normal
Normal
Normal
Normal
Normal
Normal
--mm
mm
--GPa
MPa
-----
1
40
Eq. 3
7.1
71
802.46
1
1
0.2
0.15
0.2
0.17
0.11
0.11
0.102
0.106
0.1
0.25
0.18
0.1
m,cc
c
wc
EFRP
fFRP
m,fr
m,fd
DL
SL
LL
IL
Depends on analysis
1
1
1
1
1
1
1.045
1.023
1.05
1
1
1
0.75
Unifo
rm
Increase in
Time To Repair TTR (year)
Total service life (year)
Increase in
Time To Repair TTR (year)
Total service life (year)
Pittin
g
Corrosion type
4
1.389
35.43
63.77
1.384
40.51
74.12
Table 5. Variables importance coefficients from TTR up to failure (or end of simulation at 100 years) at different
traffic frequencies f [truck/day] and with Cs=3 kg/m3
ui
fcu
fy
dst
fFRP
EFRP
Ec
m,cc
m,fr
f
50
0.09:0.06
0.58:0.49
0.03:0.03
0.08:0.05
0:-0.04
0.01:0
0:0.23
0:0
500
0.07:-0.05
0.59:0.06
0.07:0.03
0.15:0.06
0.06:-0.03
0:-0.01
0.01:0.25
0:0
1000
ui
f
50
0.04:-0.07
0.59:0
0.08:0.03
0.18:0.08
0.06-0.02
0.01:-0.01
0.01:0.27
0:0
m,fd
icorr
LL
IL
DL
SL
0.17:0.01
0:0
0.05:0.06
-0.73:-0.81
-0.05:-0.05
-0.12:-0.14
-0.08:-0.09
500
0.18:0.01
0:0
0.05:0.07
-0.72:-0.92
-0.05:-0.06
-0.13:-0.14
-0.08:-0.09
1000
0.18:0
0:0
0.06:0.08
-0.72:-0.93
-0.05:-0.06
-0.13:-0.15
-0.09:-0.10
Downloaded from Iran library: (www.libdl.ir) | Sponsor: Tehran Business School (www.tbs.ir)
Highlights
We propose a method to follow the time-variant reliability of strengthened RC beams.
We consider multiple competing failure modes of CFRP strengthened RC beams.
We consider combined degradation mechanisms (corrosion, fatigue, ageing of CFRP).