Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/233813679

Service life of RC structures: Chloride induced corrosion: Prescriptive versus


performance-based methodologies

Article  in  Materials and Structures · January 2012


DOI: 10.1617/s11527-011-9765-2

CITATIONS READS

23 879

3 authors:

Pedro Faustino Antonio Costa


WSP Technical University of Lisbon
41 PUBLICATIONS   432 CITATIONS    19 PUBLICATIONS   601 CITATIONS   

SEE PROFILE SEE PROFILE

Francesca Lanata
Ecole Supérieure du Bois
53 PUBLICATIONS   350 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Monitoring View project

All content following this page was uploaded by Pedro Faustino on 03 February 2014.

The user has requested enhancement of the downloaded file.


Materials and Structures
DOI 10.1617/s11527-011-9765-2

ORIGINAL ARTICLE

Service life of RC structures: chloride induced corrosion:


prescriptive versus performance-based methodologies
Pedro F. Marques • António Costa •

Francesca Lanata

Received: 25 September 2010 / Accepted: 27 July 2011


Ó RILEM 2011

Abstract Reinforced concrete (RC) structures sub- the performance-based specifications. The classic
jected to aggressive environmental exposure condi- safety factor and recent probabilistic approaches have
tions are traditionally designed to satisfy safety, been used to estimate the service life of each
serviceability, durability and aesthetics requirements composition being compared to the target periods
throughout their operational design service life. This defined in the prescriptive specification. Numerical
is usually established using time-dependent mathe- calculations show that the results of the Partial Safety
matical models, developed through performance- Factor and a full probabilistic approach are distinctly
based methodologies in guidelines and European and different and consequently their convergence still
national standards. However, at present, in most cases, needs to be improved, due to the complexity of the
prescriptive methodologies are used. The objective of process of chloride penetration into the concrete, not
this paper is to compare, as regards chloride induced only the model but also the input values. When
corrosion, defined target periods of service life compared to performance-based approaches it would
according to a prescriptive methodology with service be expected that the prescriptive methodology would
life results of a performance-based methodology. In be more conservative due to its less quantified
the laboratory concrete specimens were manufactured information on concrete and environment properties,
having compositions according to a prescriptive though in this study this was not always true.
specification. These specimens were tested in order
to determine their performance properties (strength, Keywords Chloride corrosion  Durability 
chloride diffusion and capillary absorption). Test Performance-based methodology  Service life
results were included in the mathematical models of

1 Introduction
P. F. Marques (&)  A. Costa
IST—Technical University of Lisbon, Av. Rovisco Pais,
1049-001 Lisbon, Portugal Over the past years, relevant research has been carried
e-mail: faustino3@gmail.com out regarding durability of reinforced concrete (RC)
A. Costa structures. Apart from the European standards EN
e-mail: acosta@civil.ist.utl.pt 1990 [1] and EN 1992-1-1 [2] and the research
contribution of international associations [3, 4–6]
F. Lanata
that have been dealing with this topic for the last
GeM—University of Nantes, 2 rue de la Houssinière,
BP 92208, 44322 Nantes Cedex 3, France 30 years, there is a large number of publications
e-mail: Francesca.Lanata@univ-nantes.fr dedicated to the durability of concrete structures
Materials and Structures

[7, 8]. Besides, there are continuously workshops and However, this approach is empirical and needs to be
conferences on the topic [9, 10]. The literature shows calibrated by physical models considering the vari-
that deterioration occurring long before the estimated ability of the environmental and materials parameters.
design limits of RC structures is of great concern For this reason, several papers have developed
nowadays. In Portugal, a significant number of mathematical models for service life prediction [15].
structures that were designed and built in the 1980s This is particularly important when concrete struc-
and 1990s for a working life of 50 years are now close tures are exposed to aggressive environments [16].
to their serviceability limits or have even passed Thus, marine environments are the most aggressive
beyond these and mainly because of steel reinforce- ones considering the high levels of chlorides and
ment corrosion [11, 12]. relative humidity. Coastal reinforced concrete struc-
In Portugal, construction of reinforced concrete tures are significantly affected and have to be
buildings started to become more widespread in the conceived taking into account these aspects [17].
1950s. Only for socioeconomic important structures, This work analyses and compares two alternative
such as bridges and hospitals, special design concerns specifications of the Portuguese National Annex to the
as regards durability were established, since Portu- NP EN 206-1 [18]—LNEC E464 [19] and LNEC
guese codes at the time considered chemical attack on E465 [20]—for the definition of concrete composition
concrete as the governing action when compared to and the corresponding design lifetime estimations:
corrosion of steel bars. In fact, Portuguese codes used
to define cover thickness of 10 mm for protected RC (a) Prescriptive method. The concrete composition
elements and 20 mm for non-protected elements— is defined on pre-established standard limits in
[13]—and few years later 30–40 mm depending on view of its anticipated environmental expo-
the chemical aggressiveness of the environment and sure—LNEC E464;
the concrete compressive strength—[14]. (b) Performance-based method. The compositions
Many structures, initially supposed to be rehabil- are defined according to specific tests and
itated or retrofitted after more than 50 years of usage, mathematical models for the determination of
are now in need of significant intervention, even the design lifetime in view of the anticipated
though they are still far from their original design environmental exposure. Two approaches are
service life. This will require an important econom- herein implemented:
ical effort that in future could be reduced or even
I. Partial safety factor approach—strictly fol-
avoided by using durable concrete materials and life-
lows LNEC E465.
cycle evaluation of concrete structures in design and
II. Probabilistic approach—the mathematical
retrofit phases.
models, parameters and assumptions for
Depending on the environmental aggressiveness,
the computation of lifetime are based on
the constituents of the concrete may vary when its mix
LNEC E465. The probabilistic estimation
composition is designed. As far as durability con-
of the design lifetime is carried out by the
cerns, the most relevant properties of the concrete are
Monte Carlo method.
associated with its composition in terms of cement
content, cement type—regarding different addi- The study will be carried out considering marine
tions—w/c ratio and also concrete cover to steel environment and its corresponding classes, according
reinforcement in the structural members. Anyway, to definitions currently given by the European
research is being done not only to study materials that standards—EN 1990 and EN 1992-1-1—and the
are sufficiently durable but also to estimate, in a more Portuguese National Annex Specifications of the NP
accurate way, the lifetime cycle of structures facing EN 206-1. In the following, the environmental
the aggressiveness of different environments. As a exposure class XS (corrosion induced by chlorides
matter of fact, the current codes provide only qual- from sea water) will be considered. Three subclasses
itative definitions of exposure and they fail to define are distinguished: XS1, areas exposed to airborne salt
the design life in relation to durability, i.e. achieving but not in direct contact with sea water; XS2,
an acceptable level of reliability of the structure permanently submerged areas; and XS3, tidal, splash
performance in its exposure environment as a whole. and spray zones.
Materials and Structures

Fig. 1 Tuutti’s model for deterioration of structures

2 The corrosion process of steel in concrete


in marine environment
Fig. 2 Typical chloride profile for a concrete structure in
At present Tuutti’s model [21] is widely accepted as marine environment (available data on Atlantic Ocean with
the conceptual model for the deterioration of struc- high tides)
tures (Fig. 1). As a function of time, this model
clearly distinguishes an initiation time followed by a In contrast, when concrete members are com-
propagation time. The initiation time refers to the pletely submerged in water, or saturated, for example
penetration of the aggressive agents into the concrete under the sea level (XS2), even if the chloride content
cover, while the propagation time is related to the on the concrete surface is high, the presence of
evolution of different forms of deterioration after oxygen near the reinforcing steel will be quite
corrosion has been initiated. reduced, which leads to a very low corrosion intensity
The sum of initiation and propagation results in the (velocity) according to some authors [24].
service life period of the structure. The extent of both Figure 2 gives an impression of the chloride
these periods depends on the environmental exposure content at different depths in function of the exposure
condition of the structure or its members. For classes XS1, XS2 and XS3 [25].
example, the initiation period is related to chloride
content; the critical chloride content causing the
dissolution of the passive layer is highly related to the 3 Design service life—reliability criteria
electrical potential difference of reinforcing steel
along with the presence of moisture and oxygen. The estimation of the design service life of reinforced
In marine environment in particular, when reinforced concrete structures is based on reliability criteria set
concrete members are exposed to aerial (XS1), tidal, or by the European standard [1]. In this standard, three
even splash zones (XS3), the effects of continuous different reliability classes are specified—RC1, RC2
wetting and drying cycles involve severe conditions, and RC3—concerning the importance of a certain
such as a high concrete surface chloride content, structure evaluated in terms of consequences after
diffusion but also convection penetration during immer- failure (risk analysis). Each of these classes is
sion cycle and oxygen availability during drying represented by a maximum probability of failure Pf
periods. In these zones, the penetration of chloride ions and the corresponding minimum reliability index b,
(initiation period) is relatively slow, but as soon as they which takes into account the statistical variability in
reach the reinforcing steel, corrosion can develop with action effects and the uncertainties in resistances and
considerable intensity because there is a sufficient model assumption.
amount of moisture and oxygen (propagation period) Even though corrosion effects on reinforced con-
[22, 23]. As a consequence, the propagation period is crete structures may vary widely, from the point of
usually discarded for the estimation of the service life. view of the consequence levels, the performance-
Materials and Structures

based specification given in LNEC E465 considers working life of 100 years—structural class S6 (EN
only the serviceability limit states. However, Euro- 1992-1-1).
code 0 [1] accounts for these limit states solely for the The specification LNEC E465, following the
reliability class RC2. guidelines of Eurocode 2 (EN 1992-1-1), classifies
Table 1 gives the minimum reliability index b the minimum concrete cover—cmin,dur (Sect. 4.2)—
associated with reliability classes recommended by for reinforced concrete structures according to struc-
the Eurocode 0 and the LNEC E465 for the ultimate tural classes with which structures’ design working
and serviceability limit states. The corresponding life and type are associated.
maximum probability of failure Pf is also given. The defined limits according to the Portuguese
specification (LNEC E464) are presented in Table 2.
These can only be reduced in view of additional
4 Prescriptive definition of concrete measures (increase of strength class, use of stainless
characteristics steel, protective coatings on concrete and/or protec-
tive epoxy coating on reinforcing steel).
4.1 Concrete composition
4.2 Concrete cover
The prescriptive methodology LNEC E464 sets the
limits of the concrete constituents (maximum w/c The cover of reinforced concrete structural members
ratio, minimum cement dosage and cement type), the is set by the EN 1992-1-1—Eurocode 2. The
minimum compressive strength and the concrete uncertainty of this variable in relation to its execu-
cover thickness for a design working life of 50 years tion is also taken into account in accordance with the
(target period) under the environmental exposures in NP ENV 13670-1 [26]. Thus, besides the minimum
issue, considering the structural class S4 (EN cover cmin,dur (mm), Eurocode 2 defines the nominal
1992-1-1). cover cnom (mm). In order to take into account any
The same prescribed limits of the concrete com- expected deviation, cnom is the value that should be
position and 10 mm added to the 50 years concrete considered for the construction project drawings and
cover permit, according to this specification, a design specifications:

Table 1 Recommended
Reliability class ULSa SLSb
minimum values of b/
maximum values of Pf for Eurocode 0 LNEC E465 Eurocode 0 LNEC E465
each reliability class
RC3 4.3/0.001% – – 2.0/2.3%
a
RC2 3.8/0.007% – 1.5/6.7% 1.5/6.7%
Ultimate limit state
b
RC1 3.3/0.048% – – 1.2/11.5%
Serviceability limit state

Table 2 Limits for a working life of 50 years defined by LNEC E464


Cement type CEM IV/A (reference); CEM IV/B; CEM III/A; CEM III/B; CEM CEM I; CEM II/A
V; CEM II/B; CEM II/A-D
XS1 XS2 XS3 XS1 XS2 XS3

Minimum nominal 45 50 55 45 50 55
cover (mm)
Maximum (w/c) 0.55 0.55 0.45 0.45 0.45 0.40
Minimum cement 320 320 340 360 360 380
dosage (kg/m3)
Minimum strength class C30/37 C30/37 C35/45 C40/50 C40/50 C40/50
LC30/37 LC30/37 LC35/38 LC40/44 LC40/44 LC40/44
Materials and Structures

cnom ¼ cmin;dur þ Dcdev ð1Þ clear understanding of in-service durability perfor-


mance of a structure. As a consequence, specification
where Dcdev (mm) is the expected deviation which LNEC E465 has introduced a performance-based
depends on the quality control, and in the Portuguese approach requiring a careful and realistic assessment
case (NP ENV 13670-1) 10 mm is the required value, of the interrelation between design, durability along
regardless of the level of quality control. with future maintenance and repair. The conceptual
The minimum concrete cover cmin,dur may be basis of such an approach is to ensure that the
interpreted as a 5% fractile characteristic value [5] required performance is maintained throughout the
and NP ENV 13670-1), where Dcdev = 1.645s being intended life of the structure along with the optimi-
s the standard deviation of a concrete cover having a zation of the incurred lifetime costs [27]. According
normal distribution. In this case the standard devia- to these specifications, the design of concrete com-
tion of the cover would correspond to 6 mm. The positions can be carried out through performance-
nominal cover cnom is assumed to represent the mean based indicators as an alternative to the definition of
value. the quantities of its constituents. Indeed, there are no
The Portuguese prescriptive specification LNEC limits whatsoever for the constituents’ type and
E464 is included in the National Annex of the quantity because each composition has to be tested
Portuguese standard NP EN 206-1. Hence, this and its results analysed. Furthermore, some of the
document determines the requisites of the composi- results have to be included in mathematical models in
tion of reinforced concrete members, and specifies order to perform a lifetime estimation regarding the
the nominal cover which includes the minimum cover type of action—chlorides [28] or carbonation [29]. In
cmin,dur and its deviation, Dcdev. the following, the existing service-life predictive
Table 3 gives the minimum concrete cover cmin,dur models based on durability indicators for chloride-
recommended by the Eurocode 2 for the prescribed induced corrosion will be presented.
structural classes. The structural classes considered
herein and adopted by the Portuguese specification 5.1 Modelling of the initiation period
LNEC E464 are: class S4—as the basis for a target
period of 50 years—and class S6—as the basis for a The defined serviceability limit state is reached when
target period of 100 years. a certain amount of chlorides—commonly referred to
as the critical chloride content—penetrates into the
concrete cover at the level of the reinforcing steel
5 Performance-based definition of design service level; at this depth, there is then a high probability of
life depassivation. This chloride content, called CR in the
performance-based method (LNEC E465), is consid-
The prescriptive nature of the LNEC E464 results in a ered equivalent to the threshold value of a determined
lack of insight into their consequences and this means composition in an exposure time equal to the
that, at the structural design stage, there is lack of initiation period and a depth penetration equal to
cover depth.
The specification LNEC E 465 defines the CR
Table 3 Minimum cover (mm)—cmin,dur according to EN
1992-1-1 values in a general way according to each exposure
class and the w/c ratio. The recommended values are
Structural class Reinforced concrete Pre-stressed concrete
given in Table 4.
XS1 XS2 XS3 XS1 XS2 XS3

S1 20 25 30 30 35 40 Table 4 Chloride content by weight of cement—CR


S2 25 30 35 35 40 45
Water/cement XS1; XS2 XS3
S3 30 35 40 40 45 50
S4 35 40 45 45 50 55 w/c B 0.30 0.6% 0.5%
S5 40 45 50 50 55 60 0.30 \ w/c B 0.40 0.5% 0.4%
S6 45 50 55 55 60 65 w/c [ 0.40 0.4% 0.3%
Materials and Structures

Even though chloride penetration into concrete Table 6 Values of khor


Distance to coast khor
occurs through different processes—permeation, as a function of the dis-
tance of the concrete 0 1.0
absorption and diffusion—depending on the exposure surface to the coast
conditions, the models currently used for the estima- 1 km 0.6
tion of the initiation period under chloride attack are
based on diffusion only.
Table 7 Values of ktemp as a function of the temperature of the
Thus, the modelling equations compute the pen- concrete surface
etration depth considering diffusion through the cover
Temperature 0°C 10°C 15°C 20°C 25°C 30°C 35°C
by means of Fick’s second law [30], though taking
into account that the diffusion coefficient diminishes ktemp 2.2 1.5 1.2 1.0 0.8 0.7 0.6
with the exposure time [11]:
 
Cðx; tÞ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi t n
0
x ¼ 2erf 1 1  Dapp ðtÞt ð2Þ Dapp ðtÞ ¼ kD;c kD;RH kD;T D0 ! Dapp ðtÞ
CS t n t
0
¼ kD0 ð4Þ
where x is the penetration depth (m) corresponding to t
a chloride content of C(x, t); Dapp(t) is the diffusion where D0 is the potential coefficient of diffusion
coefficient of chlorides in concrete, as a function of (migration) obtained from laboratory tests according
exposure time (m2/s); C(x, t) is the chloride content to NT Build 492 [31] at the concrete’s reference age
in percentage by weight of cement (%), at the depth of t0 = 28 days (m2/s), kD,c is the factor that accounts
x (m) after an exposure time t (s); CS is the chloride for the influences of curing (Table 8), kD,RH considers
content in percentage by weight of cement (%) at the the influence of the relative humidity RH (Table 9),
external concrete surface (x = 0), assumed to be kD,T accounts for the temperature influence
constant over time; erf-1 is the inverse of the error (Table 10) and n is the concrete’s ageing factor as
function. regards the chloride’s penetration in time (Table 11).
The following equations allow the determination Once the potential coefficient of apparent diffusion
of CS and Dapp(t) (LNEC E465): D0 and the concrete cover have been evaluated, by
CS ¼ Cb 2:5ðw=cÞkvert khor ktemp ð3Þ developing Eqs. 2 and 4 and knowing that, at the
limit, C(x, t) will be represented by the critical
where Cb is the surface chloride content that accounts chloride content CR, the initiation period ti, which
for the salinity of the seawater at the Portuguese coast substitutes t, may be obtained by the following:
(21 g/l) with a temperature of (16 ± 2)°C and is
defined as 3% for the classes XS2 and XS3 and 2%
for the class XS1, w/c is the water/cement ratio, kvert
and khor are coefficients related to environmental
Table 8 Values of kD,c as a function of curing days
exposure considering the concrete location to sea
level and distance to sea coast respectively (Tables 5, Number of curing days kD,c
6). ktemp is the coefficient that accounts for the
Normal 2.4
concrete temperature (Table 7).
Permanent contact with water 0.75
Formwork with controlled 1.0
permeability and 3 days
of wet curing

Table 5 Values of
Exposure kvert
kvert as a function of
class
exposure classes Table 9 Values of kD,RH
(height relative to Exposure class kD,RH
XS1 0.7 as a function of exposure
mean tide level) classes
XS2 1.0 for 1 m depth XS1 0.4
1.4 for 25 m depth XS2 1.0
XS3 1.0 XS3 1.0
Materials and Structures

Table 10 Values of kD,T


Temperature (°C) kD,T
The following equations and definitions are pro-
as a function of concrete posed by the specification LNEC E465 of the
temperature 30 1.5 National Annex of the NP EN 206-1:
25 1.2
20 1.0 5.2.1 Faraday’s law
15 0.8
10 0.75 x ¼ 0:0115Icorr tp ð6Þ
0 0.4
[33] where x (mm) is the reduction of the radius of the
ordinary steel reinforcement, Icorr (lA/cm2) is the
corrosion rate, and tp (years) is the propagation period.
Table 11 Values of n as a function of exposure classes and
cement type 5.2.2 Empirical equation for the estimate
Exposure class n of the radius reduction causing first surface
cracking
CEM I/II CEM III/IV
 
XS1 0.55 0.65 1 c
x¼ 74:5 þ 7:3  17:4ftd ð7Þ
XS2 0.45 0.55 1000 /0
XS3 0.55 0.65
[34] where ftd (MPa) is the concrete tensile strength
(obtained from the Brazilian test), with values of 3
"  2 #1n
1 and 4 MPa which can be considered in view of the
2 1 CR  Ci 1 compressive strength class (EN 1992-1-1) of each
ti ¼ erf 1 ð5Þ
c CS  Ci kD0 t0n proposed concrete composition. Cover depth is
represented by c (mm) and /0 (mm) is the initial
With c (m) being the concrete cover depth and Ci the diameter of ordinary reinforcement bar.
initial chloride content, by weight of cement, in the
concrete composition. 5.2.3 Influence of the corrosion type on steel section
reduction
5.2 Modelling of the propagation period
/0  / ¼ ax ð8Þ
The propagation period corresponds to the onset of being a B 10 for localized (pitting) corrosion (asso-
corrosion of steel reinforcement within the concrete. ciated with chlorides).
This causes the loss of steel bars section as well as the Considering y (%) as the relative reduction of the
volume increase of the corrosion product. The first steel reinforcement radius calculated from:
results into loss of strength of the steel reinforcement
itself while the latter causes internal stresses against /0  / y
¼ / ð9Þ
the surrounding concrete and loss of bond between 2 100 0
concrete and steel [32]. In either case cracks may The final equations resulting from the previous
easily develop. Furthermore, cracking might allow considerations are as follows:
easier access of oxygen and moisture to the steel
y/0
reinforcement, which may well increase the corrosion tp ¼ ð10Þ
1:15aIcorr
rate.
The estimate of the propagation period depends on where y (%) is obtained as follows:
the definition of different levels of corrosion, estab-  
c 0:2
lished as limits, depending, in its turn, on crack y ¼ 74:5 þ 7:3  17:4ftd ð11Þ
/0 /0
width. Although the modelling of the propagation is
somehow difficult due to the complexity of the The Icorr values used to obtain the propagation period
factors involved, it can be simplified by the quanti- depend on the different corrosion levels (Table 12)
fication of the corrosion rate. and on the exposure classes (Table 13).
Materials and Structures

5.3 Partial safety factor approach Table 14 Safety factor c


Reliability LNEC E465
values for each reliability
class Serviceability
class
The partial safety factor approach employs Eq. 5 for Limit State
the modelling of the initiation period as presented in
RC3 2.8
the performance-based specification (LNEC E465).
RC2 2.3
The method is based on a deterministic computation
RC1 2.0
using calibrated reduction factors (safety factors) in
order to introduce the probabilistic nature of the
problem (semi-probabilistic approach). The definition ti
of the safety factors by the referred specification is tL ¼ þ tp ð12Þ
c
based on the assumption that the lifetime of a
structure is represented by a lognormal statistical where ti is the initiation period obtained from Eq. 5, c
distribution [6] with a coefficient of variation (COV) is the safety factor according to Table 14 and tp is the
of 0.5 [6]. propagation period given in Table 15.
For the majority of the variables in the modelling The estimation of the service life period using the
equation mean values are used, however, for the performance-based methodology (safety factor
concrete cover the deterministic calculus considers a approach) has been carried out for the target periods
characteristic value cmin,dur (see Sect. 28) [5]. The tg of 50 and 100 years, considering the exposure
initiation period obtained by a deterministic calcula- classes XS1, XS2 and XS3. As previously mentioned,
tion is then divided by a safety factor c associated the used concrete cover values, following the limits
with a required reliability level (see Sect. 3; of the prescriptive specification, are those associated
Table 14) and added to the predefined minimum with the following structural classes: S4 for a 50 year
propagation periods (LNEC E465). target life and S6 for a 100 year target life [1].
In view of the referred definitions, the specification
LNEC E465 specifies different minimum propagation 5.4 Probabilistic approach
periods—tp—for each exposure class and different
minimum required target periods—tg—(Table 15). The limit state function with respect to chloride
Considering that the requirement for service life ingress is defined by Eq. 13. The probabilistic
design is tL [ tg, it is finally possible to calculate the analysis of lifetime distribution is carried out using
design service life tL as: the statistical parameters of the involved variables
(mean and standard deviation). The mean values of
each variable are based on the experimental pro-
Table 12 Corrosion levels as a function of corrosion rate gramme and LNEC E465, while the values adopted
(LNEC E465) for the standard deviation are based on fib [5] and Val
Corrosion rate Section loss (lm/year) Corrosion level and Trapper [35].
(lA/cm2) The implementation of the probabilistic calculus
for the design lifetime has been carried out by means
\0.1 1.15 Negligible
of the Monte Carlo method. The random variables of
0.1–0.5 1.15–5.75 Low the limit state function have been considered with
0.5–1.0 5.75–11.5 Moderate their probability distribution according to various
[1.0 [11.5 High reference documents [36–38].

Table 15 Minimum propagation periods of corrosion until


first surface cracking (marine environment)

Table 13 Exposure classes and corrosion levels of steel Service life tp estimated (years)
reinforcement (LNEC E465) (target) period—tg
XS1 XS2 XS3
XS1 XS2 XS3
50 years 0 40 0
Moderate Negligible High 100 years 0 80 0
Materials and Structures

As for the partial safety factor analysis, the marine environment—XS1, XS2 and XS3, the con-
concrete cover specified for a target period tg of crete compositions were made in order to respect
100 years is obtained from the one specified for demanding limits with respect to water/cement ratio,
50 years plus 10 mm. cement dosage, and cement type.
Equation 13 expresses the limit state function used The performance of these compositions was ana-
for the implementation of the Monte Carlo method lysed considering the testing results, as regards
where k represents the model uncertainty: mechanical strength [39], capillary absorption [40]
and non-steady chloride migration [31], even though
gð xÞ ¼ tL  tg
only the latter was included in the mathematical
("  2 #1n
1

2 1 CR  Ci 1 models of the performance-based specification LNEC


¼k erf 1 E465 to estimate the lifetime of concrete elements.
c CS  Ci kD0 t0n
)
y/0 6.2 Concrete composition
þ  tg ð13Þ
1:15aIcorr
The concrete compositions were mixed using
Thus, the probability of failure may be expressed as cements that follow the NP EN 197-1 [41]. This
the probability that the limit state function is standard establishes the cement type according to the
negative: proportion of clinker and other constituents (lime
Pf ¼ P½gð xÞ\0 ð14Þ filler, silica fume, fly ashes, pozzolans, blast furnace
slag, etc.). Table 16 shows the cement types used in
the four concrete compositions under study, as well
6 Experimental program as the cement dosage and the w/c ratio.

6.1 Introduction 6.3 Results

As mentioned before, the objective of this experi- The determination of the compressive strength at the
mental work is to evaluate the properties of different age of 28 days was carried out following the standard
concrete compositions, concerning their durability as NP EN 12390-3 (Table 17). Regarding chlorides, the
a corrosion protection to steel reinforcement. experimental procedure for the determination of the
From the requisites of the prescriptive specifica- coefficient of migration followed NT Build 492,
tion LNEC E464, based on the exposure classes for which included cylindrical specimens with 100 mm

Table 16 Concrete
Type of cement Constituents XS1; XS2 XS3
compositions respecting the
limits of LNEC E464 and Dosage w/c Dosage w/c
NP EN 197-1 (kg/m3) (kg/m3)

CEM II/A-D 8% silica fume 320 0.55 340 0.45


[87% clinker
CEM III/A 60% blast furnace slag 340 0.55 340 0.45
[35% clinker
CEM IV/A 20% fly ash 340 0.55 340 0.45
10% silica fume
[65% clinker
CEM IV/B–V 40% fly ash 340 0.55 340 0.45
[55% clinker
CEM V/A 25% blast furnace slag 340 0.55 340 0.45
25% fly ash
Exposure classes XS1, XS2 [45% clinker
and XS3
Materials and Structures

Table 17 Concrete experimental cubic compressive strength: Table 19 Concrete coefficient of capillary absorption: tests
test results results
Concrete XS1; XS2 XS3 Concrete XS1; XS2 XS3
composition Compressive Compressive composition Coefficient Coefficient
strength (fcka) strength (fcka) of absorption of absorption
(MPa) (MPa) (kg/(m2 Hh)) (kg/(m2 Hh))

CEM II/A-D 55.7 68.9 CEM II/A–D 0.146 0.176


CEM III/A 46.4 50.8 CEM III/A 0.154 0.402
CEM IV/A 48.9 54.5 CEM IV/A 0.278 0.176
CEM IV/B–V 48.8 54.4 CEM IV/B–V 0.214 0.205
CEM V/A 37.5 47.1 CEM V/A 0.550 0.226
s Standard deviation, fcm mean value of compressive strength
a
fck = fcm—1.645s, being classes X1 and X2 and at least C35/45 for the
exposure class XS3 (Table 2).
Regarding chloride migration (Fig. 3), Table 18
of diameter and 50 mm of thickness. Before being in shows that for the compositions designed for expo-
a low pressure hermetic recipient and immersed in a sure classes XS1 and XS2, the compositions with
solution of calcium hydroxide (Table 18), even silica fume (CEM II/A-D) and blast furnace slag (III/
though not specified by NT Build 492, the specimens A) present the lowest values with, 4.5 9 10-12 and
were subjected to 14 days of drying at 20°C and 50% 3.9 9 10-12 m2/s, respectively. The worst result
of RH. As regards capillary absorption, the test (highest) corresponds to the composition CEM IV/
procedure respected the Portuguese specification B–V (40% of fly ash) with 12.7 9 10-12 m2/s.
LNEC E393 in which cylindrical specimens with Regarding the compositions designed for exposure
150 mm of diameter and 50 mm of thickness were class XS3, the diffusion values are significantly lower
used. The conditioning consisted of 14 days of drying than those of the compositions designed for classes
at 20°C and 50% of RH before the test (Table 19). XS1 and XS2. However, the compositions showing a
Concerning the compression strength results lower migration coefficient are still II/A-D and III/A
(Table 17), it is possible to verify that the concrete with 1.8 9 10-12 and 2.0 9 10-12 m2/s, respectively,
composition with silica fume—CEM II/A-D—pre- while the worst performance (higher migration coef-
sents not only the highest values but also a significant ficient: 6.9 9 10-12 m2/s) is obtained for the compo-
difference when compared to the other compositions. sition with blast furnace slag and fly ash (CEM V/A).
On the other hand, composition CEM V/A, with blast Although with a lesser amount of fly ash, concrete
furnace slag and fly ashes, presents the lowest values composition CEM IV/A shows a lower migration
of all compositions and a little margin when compared coefficient than CEM IV/B–V. With more fly ash a
to the established limits in the prescriptive specifica- better performance is to be expected for CEM IV/B–
tion LNEC E464. In this document the compressive V. However, this type of cement (binder) was
strength classes must be at least C30/37 for exposure supplied as a pre-mixed, while the first one was

Table 18 Concrete
Concrete composition XS1; XS2 XS3
migration coefficient D0:
tests results Age (t0) Migration Age (t0) Migration
coefficient (D0) coefficient (D0)
(days) (910-12 m2/s) (days) (910-12 m2/s)

CEM II/A-D 36 4.5 36 1.8


CEM III/A 40 3.9 36 2.0
CEM IV/A 28 6.8 35 4.2
CEM IV/B–V 35 12.7 36 5.4
CEM V/A 41 8.9 40 6.9
Materials and Structures

reinforced concrete elements (Table 8) and a distance


to shore of 0 km (Table 6) were considered. For the
class XS2—submerged elements—a depth of 25 m
was considered (Table 5) (LNEC E465). Further-
more, the propagation period was only considered for
the exposure class XS2 following the premises of the
performance-based specification (LNEC E465).

7.1 Partial safety factor approach

Tables 20, 21, and 22 show the obtained values of the


design lifetime for reinforced concrete structures
based on the semi-probabilistic calculus using a
safety factor c = 2.3 (Eq. 12), according to in
Table 14. As mentioned, the initiation period was
Fig. 3 Visible chloride penetration front (NT Build 492 test
calculated using a characteristic value—cmin,dur—for
procedure). Cross section of a cylindrical specimen the concrete cover.
According to the established parameters of LNEC
obtained from simply mixing CEM I with fly ashes E465, for exposure class XS1 only the composition
along with aggregates, water and plasticizer. based on CEM IV/B–V did not achieve the target
Globally, the values of the coefficient of capillary period of 50 years (Table 20). For the 100 years
respect the upper limit defined by the most recent target period, all the compositions surpassed the
reference documents [42]—0.5 kg/(m2 Hh). The target life value. In both cases the concrete mix with
only concrete composition that surpassed this value blast furnace slag presented the highest values of
was CEM V/A, designed for XS1 and XS2 classes. design lifetime (CEM III/A).
For exposure class XS3, CEM III/A (blast furnace Regarding exposure class XS2 only the composi-
slag) was the concrete composition closest to the tions with cement IV/A and IIIA reached both target
referred limit, with 0.402 kg/(m2 Hh). The remaining periods (Table 21). Despite the minor difference
results are far from the prescribed limit, with values concrete with CEM II/A-D (8% silica fume) did not
between 0.146 and 0.278 kg/(m2 Hh). reach the target period of 50 years (tL = 49 years) but
In view of the distinct w/c ratios between compo- it managed to achieve the target period of 100 years.
sitions designed for XS1/XS2 (w/c = 0.55) and XS3 The compositions CEM IV/B–V (40% fly ash) and
(w/c = 0.45), the differences among all absorption CEM V/A (25% fly ash; 25% slag) did not get to both
values are considered to be insignificant. target periods although for a slight difference.
For the most aggressive exposure class—XS3—
compositions CEM IV/B–V and CEM V/A did not
7 Results of service life analysis reach both target periods. In the particular case of
CEM V/A the design lifetimes were less than half of
Service life design for exposure to chlorides was the required ones—11 years for tg = 50 years and
carried out in order to respect the requisites of the 36 years for tg = 100 years (Table 22).
exposure classes XS1, XS2 and XS3, for the target
periods of 50 and 100 years and the RC2 reliability 7.2 Probabilistic approach
class, which means a maximum failure probability of
6.7% or a minimum reliability index of 1.5 (Table 1). Like the safety factor approach, the probabilistic
This means that for the partial safety factor approach, approach is based on the modelling equations and
the initiation period is divided by c = 2.3 (Table 14). parameters defined in the performance-based specifi-
In the present analysis, in addition to the inherent cation LNEC E465. Yet, the lifetime results tL are
conditions of each exposure class, a mean temperature calculated by considering the required probability of
value of 15°C (Tables 7, 10), normal curing of the failure (Table 1). This means that the summation of ti
Materials and Structures

Table 20 Semi-probabilistic calculus of the design lifetime for class XS1


Composition D0a (10-12 m2/s) nb (-) Target period: 50 years cmin,dur = 35 mm Target period: 100 years cmin,dur = 45 mm
tpc (years) d
ti (years) e
tL (years) tpc (years) tid (years) tLe (years)

CEM II/A-D 4.5 0.55 0 183 79 0 558 243


CEM III/A 3.9 0.65 0 1931 839 0 8117 3529
CEM IV/A 6.8 0.65 0 806 351 0 3389 1474
CEM IV/B–V 12.7 0.65 0 88 38 0 372 162
CEM V/A 8.9 0.65 0 181 79 0 761 331
a
Coefficient of diffusion (Table 18)
b
Concrete’s ageing factor (Table 11)
c
Propagation period (Table 15)
d
Initiation period obtained from Eq. 5
e
Design service life obtained from Eq. 12

Table 21 Semi-probabilistic calculus of the design lifetime for class XS2


Composition D0 (10-12 m2/s) n (-) Target period: 50 years cmin,dur = 40 mm Target period: 100 years cmin,dur = 50 mm
tp (years) ti (years) tL (years) tp (years) ti (years) tL (years)

CEM II/A-D 4.5 0.45 40 22 49 80 49 101


CEM III/A 3.9 0.55 40 87 78 80 233 181
CEM IV/A 6.8 0.55 40 41 58 80 109 128
CEM IV/B–V 12.7 0.55 40 8 43 80 21 89
CEM V/A 8.9 0.55 40 14 46 80 37 96

Table 22 Semi-probabilistic calculus of the design lifetime for class XS3


Composition D0 (10-12 m2/s) n (-) Target period: 50 years cmin,dur = 45 mm Target period: 100 years cmin,dur = 55 mm
tp (years) ti (years) tL (years) tp (years) ti (years) tL (years)

CEM II/A-D 1.8 0.55 0 165 72 0 402 175


CEM III/A 2.0 0.65 0 1167 507 0 3672 1597
CEM IV/A 4.2 0.65 0 142 62 0 447 194
CEM IV/B–V 5.4 0.65 0 64 28 0 201 87
CEM V/A 6.9 0.65 0 26 11 0 82 36

(initiation) ? tp (propagation) \ tg (target period) are Trapper [35] and DuraCrete [4] and the variables
allowed only below a certain percentage corresponding were defined in function of their statistical nature and
to the probability of failure (Eq. 14), Pf = 6.7% related parameters presented in Tables 23, 24, and
(b = 1.5). 25. Exposure classes XS1 and XS3 do not account for
The determination of lifetime results was carried the variables of the propagation period since their
out using the Monte Carlo simulation. Hence, the corresponding corrosion rates (Table 12) calculate
modelling equations were outlined in a limit state minor results for the propagation period (0–2 years
function (Eq. 13) according to LNEC E465, Val and according to Eq. 10).
Materials and Structures

Table 23 Probabilistic
Variable Mean value—l Standard deviation—r Distribution
calculus of the design
lifetime for class XS1 Cover (cnom) 45 mm (55 mm—100 years) 8 mm Log-Normal
D0 Table 18 0.20 l Normal
CR 0.4% by cement weight (Table 4) 0.12 l Normal
Cs 2.31% by cement weight (Eq. 3) 0.10 l Normal
kvert = 0.7; khor = 1.0
n Table 11 0.05 Normal
t0 Table 18 – Deterministic
kD,T 0.8 (Table 10) 0.20 l Normal
kD,RH 0.4 (Table 9) – Deterministic
kD,c 2.4 (Table 8) – Deterministic
Model uncertainty k 1.0 0.15 l Normal

Table 24 Probabilistic
Variable Mean value—l Standard Distribution
calculus of the design
deviation—r
lifetime for class XS2
Cover (cnom) 50 mm (60 mm—100 years) 8 mm Log-normal
D0 Table 18 0.20 l Normal
CR 0.4% by cement weight (Table 4) 0.12 l Normal
Cs 6.93% by cement weight (Eq. 3) 0.10 l Normal
kvert = 1.4; khor = 1.0
n Table 11 0.05 Normal
t0 Table 18 – Deterministic
kD,T 0.8 (Table 10) 0.20 l Normal
kD,RH 1.0 (Table 9) – Deterministic
kD,c 1.0 (Table 8) – Deterministic
ftd 4 MPa 0.20 l Normal
/0 8 mm – Deterministic
a 10 – Deterministic
Icorr 0.01 lA/cm2 0.20 l Normal
Model uncertainty k 1.0 0.15 l Normal

Table 25 Probabilistic
Variable Mean value—l Standard Distribution
calculus of the design
deviation—r
lifetime for class XS3
Cover (cnom) 55 mm (65 mm—100 years) 8 mm Log-normal
D0 Table 18 0.20 l Normal
CR 0.3% by cement weight (Table 4) 0.12 l Normal
Cs 4.05% by cement weight (Eq. 3) 0.10 l Normal
kvert = 1.0; khor = 1.0
n Table 11 0.05 Normal
t0 Table 18 – Deterministic
kD,T 0.8 (Table 10) 0.20 l Normal
kD,RH 1.0 (Table 9) – Deterministic
kD,c 2.4 (Table 8) – Deterministic
Model uncertainty k 1.0 0.15 l Normal
Materials and Structures

Unlike the safety factor approach, in the probabi- Icorr represent the corrosion current density. For these
listic approach the input data of the variable c variables the mean values and the COV were based
representing the concrete cover were: (i) its mean on LNEC E465. Due to the lack of data, COV for the
(nominal) value, and (ii) the associated distribution model uncertainty k was simply assumed.
law with a standard deviation of 8 mm assigned The results of the probabilistic calculus are
according to Lindvall [38], Ferreira [36], fib [37] and presented in Figs. 4, 5, 6, 7, and 8 and Tables 26,
DuraCrete [4]. The mean values of the migration 27, and 28 considering both target periods of 50 and
coefficient D0 were obtained from the experimental 100 years. Figure 4 shows the results of design
program (Table 18) and the related COV assigned lifetime, in years, considering the Portuguese standard
according to Lindvall [38], Ferreira [36], fib [37] and limit for the probability of failure (Pf = 6.7%).
DuraCrete [4]. Additionally, Fig. 5 presents Pf values of each cement
Since in this approach the propagation period was type for the design working life periods most
also modelled for class XS2 (Eq. 10), the assigned commonly required for reinforced concrete structures.
diameter of the steel reinforcement is /0 = 8 mm, ftd For exposure class XS1 (aerial zone), Figs. 4 and 5
represent the concrete splitting tensile strength and show that CEM IV/B–V had the lowest predicted

Fig. 4 Design lifetime in (a) Minimum required: (b) Minimum required:


years for each composition 50 years 100 years
VA
VA
and each exposure class for XS IVB-V X IVB-V
a probability of failure of 3 IVA S3
IVA
6.7% (limit defined by NP IIIA IIIA
EN 206-1): a tg = 50 years XS IIA-D X IIA-D
and b tg = 100 years 2 S2

XS X
1 S1

1 10 100 1000 1 10 100 1000


Design Lifetime -years Design Lifetime -years

Fig. 5 Probability failure (a) 100.0% IIA-D (b) 100.0%


Maximumallowed: Maximumallowed:
for chosen lifetime values IIA-D
for 100 years of Lifetime

6.7 % 6.7 %
for 50 years of Lifetime

IIIA
Probability of Failure
Probability of Failure

IIIA
of: a 50 years, b 100 years IVA
10.0% 10.0% IVA
IVB-V IVB-V
VA VA
1.0% 1.0%

0.1% 0.1%

0.0% 0.0%
XS1 XS2 XS3 XS1 XS2 XS3

Fig. 6 Probability of (a)14% (b) 14%


XS1 IIA-D XS1 IIA-D
failure versus time for
Probability of Failure
Probability of Failure

12% XS1 IIIA 12% XS1 IIIA


exposure class XS1: XS1 IVA XS1 IVA
a tg = 50 years and 10% XS1 IVB 10% XS1 IVB
XS1 VA XS1 VA
b tg = 100 years 8% 8%
6% 6%
4% 4%

2% 2%

0% 0%
0 5 10 15 20 25 30 35 40 45 50 0 10 20 30 40 50 60 70 80 90 100
Time -years Time -years
Materials and Structures

Fig. 7 Probability of (a) 14% (b) 14%


failure versus time for XS2 IIA-D XS2 IIA-D

Probability of Failure
Probability of Failure
12% XS2 IIIA 12% XS2 IIIA
exposure class XS2: XS2 IVA XS2 IVA
10% 10% XS2 IVB
a tg = 50 years, XS2 IVB
XS2 VA
b tg = 100 years 8% XS2 VA 8%
6% 6%
4% 4%
2% 2%
0% 0%
0 5 10 15 20 25 30 35 40 45 50 0 10 20 30 40 50 60 70 80 90 100
Time - years Time - years

Fig. 8 Probability of (a) 14% (b) 14%


XS3 IIA-D XS3 IIA-D
failure versus time for

Probability of Failure
XS3 IIIA XS3 IIIA
Probability of Failure

exposure class XS3: 12% 12%


XS3 IVA XS3 IVA
a tg = 50 years and 10% XS3 IVB 10% XS3 IVB
b tg = 100 years XS3 VA 8% XS3 VA
8%
6% 6%
4% 4%
2% 2%
0% 0%
0 5 10 15 20 25 30 35 40 45 50 0 10 20 30 40 50 60 70 80 90 100
Time - years Time - years

Table 26 Design lifetime tL


Composition Target period: 50 years Target period: 100 years
S. factor (years) Probab. (years) Ratio prob./SF S. factor (years) Probab. (years) Ratio prob./SF

CEM II/A-D 79 86 1.09 243 223 0.92


CEM III/A 839 590 0.70 3529 1729 0.49
CEM IV/A 351 252 0.72 1474 809 0.55
CEM IV/B–V 38 36 0.95 162 120 0.74
CEM V/A 79 70 0.89 331 227 0.69
Exposure class XS1. Partial safety factor versus probabilistic

Table 27 Design lifetime tL


Composition Target period: 50 years Target period: 100 years
S. factor (years) Probab. (years) Ratio prob./SF S. factor (years) Probab. (years) Ratio prob./SF

CEM II/A-D 49 70 1.43 101 103 1.02


CEM III/A 78 111 1.42 181 189 1.04
CEM IV/A 58 83 1.43 128 131 1.02
CEM IV/B–V 43 58 1.35 89 86 0.97
CEM V/A 46 70 1.52 96 106 1.10
Exposure class XS2. Partial safety factor versus probabilistic

performance amongst all compositions and did not compositions fulfilled the limit. The composition
reach the target period of 50 years with Pf \ 6.7% with blast furnace slag—CEM III/A—presented by
(Pf = 10.6). For a target period of 100 years all far the best probabilistic results.
Materials and Structures

Table 28 Design lifetime tL


Composition Target period: 50 years Target period: 100 years
S. factor (years) Probab. (years) Ratio prob./SF S. factor (years) Probab. (years) Ratio prob./SF

CEM II/A-D 72 76 1.06 175 162 0.93


CEM III/A 507 351 0.69 1597 888 0.56
CEM IV/A 62 48 0.77 194 124 0.64
CEM IV/B–V 28 25 0.89 87 66 0.76
CEM V/A 11 14 1.27 36 35 0.97
Exposure class XS3. Partial safety factor versus probabilistic

For exposure class XS2 (submerged concrete 1.0 tg: 50 yrs


elements) the design lifetime results of all composi- 0.9 tg: 100 yrs

/ tL for class XS2


tions complied with the requirements for both target 0.8
periods, except CEM IV/B–V for a target period of 0.7
100 years (Pf = 11.9%). Once more, the results 0.6
based on probabilistic calculations show that the 0.5
concrete mix with slag has the best results of 0.4
predicted performance.
Ratio tp
0.3
As to exposure class XS3 (tidal and splash zones), 0.2
Figs. 4 and 5, and Table 28, show that only concrete 0.1
mixes with slag (CEM III/A) and silica fume (CEM 0.0
II/A-D), in this order, reached and surpassed the IIA-D IIIA IVA IVB-V VA

target period of 50 years, though CEM IV/A is not far


Fig. 9 Proportion of the propagation period tp on total
(tL = 48 years). For a target period of 100 years this structures lifetime for exposure class XS2
value was not achieved or surpassed by the compo-
sitions CEM IV/B and CEM VA.
Figures 6, 7, and 8 show the probability of failure Compared to prescriptive methods, the perfor-
through time for all compositions and exposure mance-based methods allow the consideration of
classes XS1, XS2 and XS3. It can be verified that, different variables although, in fact, some of these are
at this level of probability of failure (until Pf = 15%) very difficult to quantify. In the modelling equations
the relation between Pf and time is nearly linear for of the initiation period, beside environmental param-
exposure classes XS1 and XS3, while for class XS2 eters, the chloride threshold CR and the ageing factor
the evolution is similar to an exponential curve. This n are still roughly quantified and the Portuguese
means that, at an early stage of service life, for specification E465 chose to adopt the values pre-
classes XS1 and XS3 the limits of accepted perfor- sented in Tables 4 and 11 which, at some point, differ
mance—i.e. Pf—are reached sooner than for class from reference documents [4, 5]. The modelling
XS2, which can be explained by the influence of the equation of initiation is particularly sensitive to the
propagation period on the entire lifetime of the variation of the mean values of these variables [36].
structure, where the rising of the probability of failure As regards chloride threshold LNEC E465 [20]
is higher. defines 0.3–0.6% by weight of cement for all
Considering that for class XS2, in view of LNEC exposure classes and w/c ratios (Table 4), while
E465, the propagation period has an important role on according to Glass and Buenfeld [43] this variable
the total lifetime of reinforced concrete elements, might have a wider range from 0.2 to 2.5%. Despite
Fig. 9 represents the proportion of tp in relation with being generally assumed to be in the range of
tL. Since tp does not account for the influence of the 0.4–1.0% by weight of cement, it may change
cement type, differences between compositions are depending on the environment exposure. In sub-
due to their performance during the initiation period ti. merged elements in marine environment this value is
Materials and Structures

significantly increased due to low oxygen content and Moreover, two important differences should be
low values of steel electrochemical potential [44]. outlined: the safety factor approach (Sect. 1, b-I)) of
In view of this, results of predicted lifetime for the performance-based specification LNEC E465
exposure class XS2 (submerged elements) could be (i) does not account for the model uncertainty and
considerably increased with higher input values of CR (ii) sets the minimum propagation period values; on
for the initiation period. the other hand in the probabilistic calculus (Sect. 1,
As to the ageing factor n, Duracrete [4] and fib [5] b-II) (i) the model uncertainty is considered through
recommend values that vary from 0.30 to 0.80 the coefficient k and (ii) the mathematical expression
regarding types of cement and exposure classes and of the propagation period is implemented (Eq. 12).
despite slight differences when using these values, in For class XS1, Table 26 presents the results of
some cases, results change considerably due to the predicted design lifetime for both approaches and
sensitiveness of the modelling equation to this their ratio—probabilistic approach/safety factor.
parameter. This can considerably affect results for There is a better convergence of results for the target
all exposure classes. period of 50 years whose ratios vary from 0.70 to
1.09, while for 100 years the values vary from 0.49 to
7.3 Partial safety factor versus probabilistic 0.92. In both cases the least convergence is obtained
approaches for CEM III/A and CEM IV/A, which happens to
have the highest predicted design lifetime.
Taking into account that the performance-based For the most aggressive exposure class—XS3—
method has been carried out using two different the analysis is similar with ratios between 0.69–1.27
approaches, it is useful to analyse the convergence of for tg = 50 years and 0.61–0.97 for tg = 100 years
their results. The lifetime results of both perfor- (Table 28). For these exposure classes—XS1 and
mance-based approaches are analysed and compared XS3—the majority of the ratios are under 1.0 which
in Tables 26, 27, and 28. means that generally the safety factor approach gives
It is however important to outline that in both higher lifetime values than the probabilistic approach.
approaches some results are far beyond what it can be As regards the exposure class XS2 (Table 27), the
called as realistic. Concerning durability Portuguese ratios vary from 1.35 to 1.52 for 50 years of target
codes consider the possibility of designing structures lifetime and from 0.97 to 1.10 for a target period of
for working lives of 50 or 100 years. It is therefore 100 years. Unlike classes XS1 and XS3, for class
acceptable not only to try to achieve such expected XS2 all cases, except one, present ratios above 1.0.
lifetimes without major interventions as to have This means that for this exposure class the probabi-
modelled results in the same order of magnitude. This listic approach led to higher results than the safety
is why some of the results obtained may be consid- factor approach.
ered as unrealistic (let us say [150 years), in view of In general, the differences between the perfor-
what structures show in situ concerning corrosion. mance-based approaches may be explained by the
In order to perform a sensible comparison, it is following considerations: (i) all safety factor values
important to outline some of the differences associ- are obtained by using a log-normal distribution for
ated with the statistical nature of these two the lifetime of structures and considering a pre-set
approaches. Even though the safety factor approach coefficient of variation (COV) of 0.50; (ii) the
uses a deterministic calculus taking into account a probabilistic approach considers uncertainties associ-
probabilistic translation through the introduction of a ated to the models while the safety factor approach
certain safety factor, it considers cmin,dur as the input doesn’t reflect it on the safety factor values; (iii) the
cover value in Eq. 12, which is a characteristic value, safety factor approach uses a characteristic value—
not a mean one, interpreted as obtained from a 6 mm cmin,dur—of the concrete cover for the deterministic
standard deviation of a normal distribution (Sect. calculus of the design lifetime while the probabilistic
4.2). On the opposite, the probabilistic approach approach uses a mean value (nominal—cnom).
considers the cover mean value (nominal cover cnom) In the particular case of exposure class XS2,
with and an 8 mm standard deviation which is besides the aspects previously outlined, the propaga-
adopted by different references [4, 36]. tion period is pre-established in the safety factor
Materials and Structures

approach (40 years for tg = 50 years and 80 years target life of 100 years. Furthermore, classes XS1 and
for tg = 100 years) while in the probabilistic XS3 show better convergence for compositions with
approach it is modelled according to Eq. 10 presented lower service life results as the opposite happens to
in the performance-based specification LNEC E465. XS2, where better convergence is verified for com-
positions with higher values. The different computa-
tion of the propagation period tp for class XS2 in the
8 Concluding remarks two approaches could have had a significant influence
on the results so to partially explain the differences
The present study has performed the analysis of the previously pointed out.
convergence of two distinct methods for the evalu- From the results analysis and due to the complex-
ation of structures’ service life in marine environ- ity of the chlorides’ penetration process into the
ment—chlorides action—using the experimental concrete, there is a need to carry out in-depth studies
results of concrete compositions with different to allow a better understanding of the chloride
cement types. The compared methodologies are the transport process. In fact, in order to have these
prescriptive method and the performance-based performance-based methods useful for practice, some
method. Moreover, the performance-based method of the involved parameters still need to be discussed,
has been carried out through two alternative to some extent, as regards their relation with the
approaches: the partial safety factor approach and environment. Taking into account the reduction of
the probabilistic approach. chloride rate ingress into concrete in time and the
With respect to chlorides migration the experi- difficulty of its quantification, more field and labo-
mental results have shown that the concrete compo- ratory tests should be carried out. An improved
sition with blast furnace slag cement (CEM III/A) had quantification of the ageing factor n and/or of the
the best performance (lowest results) followed by the quantification of the migration coefficient at ages
composition with cement type CEM II/A-D (with over 28 days, might be relevant to account for the
silica fume), while the compositions CEM IV/A, hydration properties of types of cement as those with
CEM IV/B–V and CEM V/A presented higher fly ash and slag and whose results were far from
diffusion values. The compressive and capillary expected (CEM V/B–V and CEM V/A) [5, 17, 45].
absorption results did not follow the same tendency As to the performance-based approaches for
as the chlorides migration test. chloride action, the statistical distribution of the
After the input of the experimental results into the service life should be analysed considering a calibra-
mathematical models for the service life evaluation, tion of safety factors so that both approaches—
for exposure classes XS1 and XS2 and both target probabilistic and safety factor—could converge.
periods of 50 and 100 years the composition with
blast furnace slag (CEM III/A) presented by far the
higher predicted design service life values, followed
by the composition with fly ashes and silica fume References
(CEM IV/A). For class XS3, CEM III/A was again
the composition showing the best predicted perfor- 1. EN 1990—Eurocode 0 (2002) Bases of structural design.
CEN, Brussels
mance and surpassing largely both target values, 2. EN 1992-1-1—Eurocode 2 (2004) Design of concrete
while compositions CEM IV/B–V (fly ash) and CEM structures. Part 1-1: general rules and rules for buildings.
V/A (fly ash and slag) have shown to have the lowest CEN, Brussels
predicted performances and to be far from the 3. CEB-FIP (1993) Model Code 1990. T. Thelford, London
4. DuraCrete (2000) Probabilistic performance based dura-
minimum required target periods of 50 and bility design of concrete structures. The European Union—
100 years. Brite EuRam III, DuraCrete, Final Technical Report of
As to the convergence between the two perfor- DuraCrete project, Document BE95-1347/R17, CUR,
mance-based approaches, classes XS1 and XS3 have Gouda, Nederland
5. fib (2006) Bulletin 34. Model code for service life design.
generally shown a better convergence for the target Lausanne, Switzerland
period of 50 years, while class XS2 has reached 6. RILEM (1996) Report 14—durability design of concrete
better agreement between the two approaches for the structures. E&FN Spon Press, London
Materials and Structures

7. Folić R (2009) Durability design of concrete structures— Concrete and Cement Composites, Woodhead Publishing,
part 1: analysis fundamentals. Sci J Facta Univ Ser Arch Abington Hall, pp 187–246
Civ Eng 7(1):1–18. doi:10.2298/FUACE0901001F 26. NP ENV 13670-1 (2007) Execution of concrete structures.
8. Mays G (2001) Durability of concrete structures: investi- Part 1: general rules. IPQ, Lisbon
gation repair protection. E&FN Spon Press, London 27. Narasimhan H, Chew MYL (2009) Integration of durabil-
9. ICDCS (2008) Advances in concrete structural durability ity with structural design: an optimal life cycle cost based
proc int conf on durability of concrete structures. Zhejiang design procedure for reinforced concrete structures. Constr
University Press, Hangzhou, China November 2008 Build Mater 23(2):918–929. doi:10.1016/j.conbuildmat.
10. RILEM (2009) Concrete durability and service life plan- 2008.04.016
ning. In: Kovler (ed) Proceedings of 2nd international 28. Baroghel-Bouny V, Nguyen TQ, Dangla P (2009)
RILEM workshop concrete life’09, Haifa, Israel, Septem- Assessment and prediction of RC structure service life by
ber 2009 means of durability indicators and physical/chemical
11. Costa A, Appleton J (2002) Case studies of concrete models. Cem Concr Comp 31:522–534. doi:10.1016/j.
deterioration in a marine environment in Portugal. Cem cemconcomp.2009.01.009
Concr Comp 24(1):169–179 29. Thiery M, Villain G, Baroghel-Bouny V, Dangla P (2006)
12. Costa A, Appleton J (1998) Inspecção e Reabilitação de 4 Modelling of concrete carbonation based on coupled mass
Pontes Cais. Jornadas Portuguesas de Engenharia de transport and chemical reactions. In: Proceedings interna-
Estruturas, LNEC, Lisboa tional RILEM workshop on performance-based evaluation
13. REBA (1967) Regulamento de Estruturas de Betão Arm- and indicators for concrete durability, Madrid, Spain,
ado. Decreto no 47723 de 20 de Maio March 2006
14. REBAP (1983) Regulamento de Estruturas de Betão 30. Aı̈t Mokhtar K, Loche J-M, Friedmann H, Amiri O, Am-
Armado e Pré-esforçado. Decreto-Lei no 349-C/83 de 30 mar A (2007) Steel corrosion in reinforced concrete. In:
de Julho Report no. 2-2—concrete in marine environment. ME-
15. Mitchell D, Frohnsdorff G (2004) Service-life modelling DACHS—Interreg IIIB Atlantic Space—Project no 197.
and design of concrete structures for durability. Concr Int Marine environment damage to Atlantic coast historical
26(12):57–63 and transport works or structures: methods of diagnosis,
16. Maekawa K, Ishida T, Chijiwa N (2007) Computational repair and of maintenance
life-cycle assessment of structural concrete subjected to 31. NT Build 492 (1999) Concrete, mortar and cement based
coupled severe environment and mechanistic actions. In: repair materials: chloride migration from non-steady state
Proc CONSEC’07, Tours, France, June 2007, 3–18 migration experiments. Nordtest, Espoo
17. Gjorv OE (2009) Durability design of concrete structures 32. Bentur A, Berke N, Diamond S (1998) Steel corrosion in
in severe environments. E&FN Spon Press, London concrete: fundamentals and civil engineering practice.
18. NP EN 206-1 (2005) Concrete—part 1: specification, E&FN Spon Press, London
performance, production and conformity. IPQ, Lisbon 33. Rodriguez J, Andrade C (1990) Load bearing capacity loss
19. LNEC E464 (2007) Concrete. Prescriptive methodology in corroding structures. In: Proceedings of ACI convention,
for a design working life of 50 and 100 years. LNEC, Toronto
Lisbon 34. Andrade C, Alonso C, Rodriguez J, Casal J, Diez JM
20. LNEC E465 (2007) Concrete. Methodology for estimating (1995) Relation between corrosion and cracking. Internal
the concrete performance properties allowing to comply report of Brite/Euram project BE-4062. DG XII. C.E.C
with the design working life of the reinforced or pre- 35. Val DV, Trapper PA (2008) Probabilistic evaluation of
stressed concrete structures under environmental exposures initiation time of chloride-induced corrosion. Reliab Eng
XC and XS. LNEC, Lisbon Syst Saf 93:364–372. doi:10.1016/j.ress.2006.12.010
21. Tuutti K (1982) Corrosion of steel in concrete. CBI 36. Ferreira RM (2004) probability based durability analysis of
research report no 4.82. Swedish Cement and Concrete concrete structures in marine environment. PhD Thesis,
Research Institute, Stockholm, Sweden University of Minho, School of Engineering, Guimarães,
22. Andrade C, Alonso C, Molina FJ (1993) Cover cracking as Portugal
a function of bar corrosion: part 1–experimental test. Mater 37. fib (2010) Bulletins 55 and 56. Model code 2010—first
Struct 26:453–464. doi:10.1007/BF02472805 complete draft, vol 1 and 2. Lausanne, Switzerland
23. Costa AJS (1997) Durabilidade de Estruturas de Betão 38. Lindvall A (2003) Environmental actions on concrete
Armado em Ambiente Marı́timo. PhD Thesis, Techni- exposed to marine and road environments and its response.
cal University of Lisbon, Instituto Superior Técnico, PhD Thesis, Chalmers University of Technology, Göte-
Lisbon borg, Sweden
24. Coppola L, Fratesi R, Monosi S, Zaffaroni P, Collepardi M 39. NP EN 12390-3 (2003) Testing hardened concrete—parte
(1996) Corrosion of reinforced concrete in sea water sub- 3: compressive strength test. IPQ, Lisbon
merged structures. In: Proceedings of 3rd international 40. LNEC E393 (1993) Concrete. Capillary absorption of
conference on performances of concrete in marine envi- water. LNEC, Lisbon
ronment, New Brunswick, Canada, August 1996, 41. NP EN 197-1 (2005) Cement. Composition, specification
pp 127–160 and conformity criteria. IPQ, Lisbon
25. Nürnberger U, Sawade G, Isecke B (2007) Degradation of 42. NP EN 1504-3 (2006) Products and systems for the pro-
pre-stressed concrete. In: Page CL&MM (ed) Durability of tection and repair of concrete structures. Definitions,
Materials and Structures

requirements, quality control and evaluation of conformity. 44. Manera M (2008) Chloride threshold for rebar corrosion in
Structural and non-structural repair. IPQ, Lisbon concrete with addition of silica fume. Corros Sci
43. Glass and Buenfeld (1997) The presentation of the chloride 50(2):554–560
threshold level for corrosion of steel in concrete. Corros 45. Bijen J (2003) Durability of engineering structures.
Sci 39(5):1001–1013 Woodhead Publishing, Cambridge

View publication stats

You might also like