Structural Safety 25 (2003) 343362

www.elsevier.com/locate/strusafe

Life-cycle cost analysis of reinforced concrete structures

in marine environments
Dimitri V. Vala,*, Mark G. Stewartb
a

Department of Structural Engineering and Construction Management, Faculty of Civil and Environmental Engineering,
TechnionIsrael Institute of Technology, Haifa 32000, Israel
b
Centre for Infrastructure Performance and Reliability, School of Engineering, The University of Newcastle,
New South Wales 2308, Australia

Abstract
Chloride-induced corrosion of carbon steel reinforcement is the main cause of deterioration of reinforced
concrete (RC) structures in marine environments. One of the ways to protect RC structures from corrosion
is to use corrosion-resistant stainless steel reinforcing bars. However, stainless steel is between six and nine
times more expensive than carbon steel. Thus, its use can only be justied on a life-cycle cost basis. In the
paper a time-variant probabilistic model was presented to predict expected costs of repair and replacement
which was then used to calculate life-cycle costs for RC structures in marine environments under dierent
exposure conditions. Results of the life-cycle cost analysis can be applied to select optimal strategies improving durability of RC structures in marine environments, including the use of stainless steel reinforcement.
# 2003 Elsevier Ltd. All rights reserved.
Keywords: Chlorides; Corrosion; Deterioration; Life-cycle cost; Reliability; Reinforced concrete

1. Introduction
Reinforced concrete (RC) structures deteriorate with time if subject to an aggressive environment. As a result, as civil infrastructure is aging owners have to spend an increasing percentage of
their budgets on rehabilitation (or replacement) of existing RC structures. Thus, there is
obviously a strong nancial incentive to extend the service life of existing structures, and to design
new RC structures which will require less maintenance and repair over their lifetime.
Chloride-induced corrosion of reinforcing bars is the primary cause of deterioration of RC
structures in onshore and oshore marine environments. Corrosion is initiated by chloride contamination, often in conjunction with inadequate cover or poor quality concrete. It then leads to
* Corresponding author. Tel.: +972-4-829-2276; fax: +972-4-832-3433.
E-mail address: val@tx.technion.ac.il (D.V. Val).
0167-4730/03/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0167-4730(03)00014-6

344

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

cracking and spalling that indicates the need for an assessment of existing safety, repair or
replacement of damaged structural elements, or the need for more frequent inspections. All these
cases will require the allocation of additional nancial resources.
To improve durability, reduce maintenance costs and extend service life special design requirements for RC structures built in marine environments are usually specied. These requirements
include the use of high performance concrete (with low water/cement ratio), increased concrete
cover (e.g., AS 3600 [1]), and the use of admixtures (i.e., corrosion inhibitors). These measures
tend to increase the initial cost of structures, but do not completely eliminate the risk of corrosion
because of poor detailing, variable construction quality, etc. while high-performance concrete is
more susceptible to cracking.
An alternative approach is to use corrosion-resistant stainless steel reinforcing bars. These bars are
about six to nine times more expensive than carbon steel reinforcing bars [2]. However, it has been
suggested that expected reductions in maintenance and repair costs and the extension of service life
can justify their use on a life-cycle cost basis (e.g., [25]). It is also worth noting that the cost of reinforcing steel is only a small fraction of overall initial construction costs, thus, the replacement of
carbon steel reinforcement by stainless steel should result in a rather moderate increase in the overall
initial costs (normally, less than 20% [57]). Moreover, stainless steel reinforcement can be used
selectively by replacing carbon steel only in the most vulnerable areas (i.e., where cracking might
occur, locations of reduced cover or other areas where direct access of chlorides, water and oxygen
might have place) that should further reduce an increase in the overall initial cost.
The present paper will focus on a life-cycle cost analysis of RC structures in marine environments. A
structural deterioration life-cycle probabilistic model is used to calculate probabilities of cracking and
spalling for RC structures with carbon steel reinforcement in splash zones and coastal regions. The
model includes the random variability of surface chloride concentration, chloride diusion, threshold
chloride concentration, corrosion rates, concrete material properties, element dimensions and environmental conditions. If excessive cracking and spalling are observed across a structure then this would
constitute the end of service life if repairs were not conducted [810]. It is therefore assumed herein
that the incidence of widespread cracking and spalling will result in repair in order to extend the service life of RC structures. Time-dependent probabilities of spalling are calculated for annual increments over the lifetime of the structure (100 years) and the probability that multiple repairs will be
needed during the life of the structure is then calculated. Life-cycle costs considering initial construction
cost, costs of improved durability and expected maintenance and repair costs can then be estimated.
For RC structures with stainless steel reinforcement it is assumed that the probability of corrosion is negligible, and subsequently expected costs of structural repairs necessitated by corrosion can be neglected (e.g., [4]). Thus, expected life-cycle costs of RC structures with stainless steel
reinforcement can be assumed equal to their initial costs. This means that in order to make
decision about the use of stainless steel reinforcement in a RC structure only results of life-cycle
cost analysis of the structure with carbon steel reinforcement are needed.

2. Life-cycle cost analysis

If all attributes and consequences of a decision concerning a structure can be expressed in
monetary terms then an optimal decision will be the one that minimises the life-cycle cost of the

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

345

structure. Generally, if the benets of each alternative are the same, then the expected life-cycle
cost up to time T, LCC(T), may be represented as
LCCT CD CC CQA CIN T CM T

M
X
Pfi TCSFi

i1

where CD is the design cost, CC the construction cost (materials and labour), CQA the expected
cost of quality assurance/control, CM(T) the expected cost of maintenance, CIN(T) the cost of
inspections, M the number of independent failure limit states (e.g., exure, shear, spalling), P(T)
the cumulative probability of failure for each limit state (i.e., probability that failure will occur
anytime up to time T), and CSFi the failure cost (i.e., damages, cost of life, injury, user delay, etc.)
associated with the occurrence of each limit state. Costs and benets may occur at dierent times
so in order to obtain consistent results it is necessary for all costs and benets to be discounted to
a present value. Discount rates are inuenced by a number of economic, social and political factors and thus can be quite variable. For example, discount rates used by various government
agencies are: Australia 7%, US 23% [11], UK Department of Transport 8% [12], Sweden 4%
[13] and Finland 6% [14]. Note that a high discount rate favours a short service life whereas a low
discount rate encourages longer service lives.
As already mentioned, chloride-induced corrosion of reinforcing steel is the main cause of
deterioration and subsequent reduction of service life for RC structures in marine environments.
Ingress of chloride ions into concrete leads to corrosion initiation followed by cracking and,
eventually, spalling of the concrete cover. At the rst sign of spalling the associated loss of
structural capacity (and safety) for a RC structure will not exceed 20% [15]. For load-bearing
elements this will undoubtedly lead to higher probabilities of failure for ultimate limit states (i.e.,
exure, shear). Nonetheless, the probabilities of failure for ultimate limit states will still be very
low compared to probabilities of serviceability failure. Hence, as long as corrective action
(repairs, replacement) is made to the structure almost immediately after spalling then in the following analysis serviceability failure caused by spalling is considered as the most inuential mode
of failure for the estimation of life-cycle costs for RC structures in marine environments.
It is assumed that spalling is always detected when the structure is inspected. The expected lifecycle cost given by Eq. (1) may then be re-expressed as
LCCT CD CC CQA CIN T CM T ESF T

where ESF(T) is the expected cost of spalling during service life T and CD, CC, CIN and CM are all
present value costs.

3. Evaluation of expected cost of spalling

3.1. Stainless steel reinforcement
For RC structures with stainless steel reinforcement it is assumed that the probability of spalling during the service life of a structure T can be neglected and, hence, the expected cost of
spalling is ESF(T)=0.

346

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

The types of stainless steel that have been most commonly used as reinforcement are austenitic
types 304 and 316 and duplex alloy 2205. Note that the type 304 is the cheapest and while it has
much better corrosion resistance than carbon steel it may not be sucient for highly aggressive
environments such as the splash zone, or right on the coast. For such environments it is recommended to use 316 or alloy 2205. For recommendations on the use of stainless steel see, for
example Ref. [3].
3.2. Carbon steel reinforcement
For RC structures with carbon steel reinforcement the expected cost of spalling can be estimated as
ESF T

T=Dt
X T=Dt
X

CSF

n1 in

1 riDt



Pf;n iDt  Pf;n iDt  Dt

where t is the time between inspections, n the number of spalling incident, i the number of
inspection, Pf,n(t) the probability of the n-th spalling incident before time t, CSF the cost associated with the occurrence of spalling (i.e., repair/replacement, user losses, etc.), and r the discount rate. Eq. (3) can also be written in a dierent form as
ESF T

T=Dt
X

CSF

i1

1 riDt

DPf;i

where Pf,i, is the probability of a spalling incident between the (i1)-th and ith inspections. This
probability can be calculated by the following recursive formula
i1

 X


DPf;i Pf iDt  Pf i  1Dt
DPf ;j Pf i  jDt  Pf i  j  1Dt

j1

where Pf(t) is the cumulative distribution function for the time of rst spalling (its calculation will
be described further in the paper). The number of spalling incidents n depends on the repair
strategy. Two repair strategies as proposed by Stewart [16] are considered in the paper. These two
repair strategies are illustrative only and provide some bounds on poor (Repair Strategy I) and
best (Repair Strategy II) scenarios. Both scenarios assume 100% repair eciency (i.e., element
returned to as new condition).
3.2.1. Repair Strategy I
. Repair is carried out immediately after spalling has been discovered.
. Repair provides no improvement in durability performance of the repaired structure (i.e., it
is repaired with the same cover and concrete quality as the original design specication).
. Spalling may re-occur during the remaining service life of the structure, i.e., multiple repairs
may be needed. The maximum possible number of spalling incidents is then equal to the
number of inspections during service life of the structure, i.e., nmax=T/t.

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

347

3.2.2. Repair Strategy II

. Repair is carried out immediately after spalling has been discovered.
. Repair improves durability performance of the repaired structure (e.g., by increasing concrete cover and/or concrete grade) such that the probability of spalling re-occurrence during
the remaining service life of the structure can be neglected. In this case the number of spalling incidents is limited to n=1, so that the second term in Eq. (5) disappears.
Of course a number of other maintenance and repair strategies dependent on structure type,
current (local) practices, availability of resources, regulatory requirements, etc. are also possible.
The procedure presented above can then be easily modied and used to predict and update extent
of damage, need for repair, cost of repair, eectiveness of repair, performance of a repaired
structure, etc. until the design life of the structure is reached.

4. Models for evaluation of probability of spalling

4.1. Corrosion initiation and propagation models
The penetration of chlorides into concrete is usually considered as a diusion process and thus
can be described by Ficks second law of diusion. If the surface chloride content, C0, and the
diusion coecient, D, are assumed constant, the chloride content, C(x,t), at a distance x from
the concrete surface at time t can then be calculated as




x
6
Cx; t C0 1  erf p
2 Dt
where erf is the error function. It should be noted that the movement of chlorides in concrete is
not a pure diusion process from a physical point of view. However, Eq. (6) is still used due to its
computational convenience with C0 and D calculated by tting this equation to measured chloride proles. The values of C0 and D obtained from such a curve-tting exercise are usually
referred to as the notional surface chloride content and the apparent diusion coecient. In
the following, whenever C0 and D are mentioned it will always imply that these are the notional
surface chloride content and the apparent diusion coecient, respectively.
It is assumed that corrosion of reinforcement starts when the chloride concentration at the level
of reinforcement reaches a threshold concentration, Cr. After the initiation, the corrosion rate is
estimated in terms of a corrosion current density, icorr. Major factors aecting the corrosion rate
are the concrete quality (mainly dependent on water/cement ratio) and cover. In the paper the
corrosion current density is estimated as proposed by Vu and Stewart [17]

0:3781  w=c 
7
mA=mm2
icorr
cover
where w/c is a water/cement ratio and cover is given in mm. It is also assumed that corrosion will
lead to a uniform reduction in the bar diameter of the reinforcing steel. In this case, the corrosion
current density can be directly transformed into the loss of steel by the use of Faradays law of

348

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

electrochemical equivalence, which indicates that icorr=0.01 mA/mm2 corresponds to a uniform

corrosion penetration of 11.6 mm per year. It is assumed that the mean and coecient of variation
(COV) of the model error of Eq. (7) are 1.0 and 0.2, respectively, and that the model error follows
a normal distribution [17].
4.1.1. Surface chloride content
For marine structures the surface chloride content depends mainly on the proximity to seawater. In this context the following environmental exposure zones are usually considered (e.g.,
[1]):
. Submerged zoneconcrete is always under water, oxygen is not readily available and the
risk of corrosion is low, especially when the concrete cover is uncracked.
. Splash and tidal zonesconcrete is exposed to wetting and drying cycles so that chlorides
accumulates on the concrete surface by a process of wetting with seawater, evaporation and
salt crystallisation. It is now generally accepted that the highest risk of corrosion is in the
splash zone.
. Atmospheric zoneconcrete is not exposed directly to seawater and sea spray carried by the
wind is the main source of chlorides. Within this zone AS 3600 [1] also distinguishes between
coastal category (up to 1 km from the coast) and near coastal category (between 1 km
and 50 km from the coast).
Accordingly, in the paper three possible locations of RC structures are considered: (1) in the
splash zone, and in the atmospheric zone (2) adjacent to the coast and (3) 1 km from the coast.
The mean and COV of the surface chloride content for these locations have been determined
using data on oshore and onshore RC structures along the Victorian and Tasmanian coast of
Australia reported by Collins and Grace [18] and McGee [19]:
(1) for splash zonemean=7.35 kg/m3 and COV=0.70;
for atmospheric zone on the coastmean=2.95 kg/m3 and COV=0.70; and
for atmospheric zone 1 km from the coastmean=1.15 kg/m3 and COV=0.50.
It was also found that the surface chloride content is well described by a lognormal distribution.
It is expected that surface chloride levels increase with time. However, as observations show this
take place only during the rst few years of exposure, eventually the surface chloride content
stabilise and remains almost constant over the rest of the service life of a structure (e.g., [20]).
Thus, in this study the surface chloride content is treated as a time-invariant random variable.
4.1.2. Diusion coecient
The chloride diusion coecient depends mainly on properties of concrete (e.g., water/cement
ratio, composition, degree of hydration and aggregate/cement ratio) and environmental conditions (e.g., temperature and relative humidity). It also depends on the concentration of chloride
ions in the concrete pore solution and even on the testing condition [21]. Due to the complexity of
the problem no reliable model is currently available that would account for all these factors and
simple empirical models, which consider only the inuence of mix proportions (especially, water/
cement ratio), are usually used. Among these models, the model proposed by Papadakis et al. [22]
appears to be the best t to available data [17] and is represented as

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

0
1
w
w
1 c
  0:85
B cc
C
c
D 0:15
@
w c a
w ADH20
1 c
1 c
c a c
c

m2 =s

349

where a/c is the aggregate/cement ratio, c and a are the mass densities of cement and aggregates, respectively, and DH2O is the diusion coecient in an innite solution (=1.6 109 m2/s).
There are no sucient data (e.g., a/c ratio, c and a are usually not provided with test results) to
estimate accurately statistical parameters for the model error. It is assumed that the mean and COV
for the model error are 1.0 and 0.2, respectively. It is also assumed that the model error follows a
normal distribution [17]. The water/cement ratio is estimated from Bolomeys formula [23]
w=c

27
fcyl 13:5
0

where fcyl is the concrete compressive strength from a standard cylinder test in MPa.
Available data indicate that the diusion coecient decreases with time. However, there is
evidence that this decrease occurs mainly during the rst year of exposure, so that D can be
considered constant for mature structures [20,21].
4.1.3. Chloride threshold concentration
A review of reported values of the chloride threshold concentration prepared by Bamforth et al.
[20] show a very wide scatter of results from 0.125 to 2.2% by weight of cement. Moreover, there
is still no general agreement if the threshold concentration should be presented as a total or free
chloride content (e.g., [24]). As such, it may be better to consider the chloride threshold concentration in terms of corrosion risk [20]. Vassie [25] published data for UK bridges on the percentage of cases of corrosion in relation to the total chloride concentration at the level of
reinforcement. According to the data, the chloride threshold concentration can be presented by a
normal distribution with a mean of 0.95% by weight of cement ( 3.35 kg/m3) and COV=0.375.
The use of these parameters is also broadly supported by data presented by Matsushima et al.
[26], where it was suggested to use a mean of 3.07 kg/m3 and COV=0.41 for the threshold
chloride concentration. Since the threshold concentration cannot be negative and values below
0.125% by weight of cement have not been reported, the distribution is truncated at 0.1% by
weight of cement ( 0.35 kg/m3).
4.2. Modelling cracking and spalling
The occurrence of cracking, delamination and, eventually, spalling of the concrete cover is
referred to herein as a serviceability failure, and if not repaired may often precursors to more
critical and dangerous strength limit state problems. Liu and Weyers [27] developed a model to
predict the time to rst cracking (hairline crack of width less than 0.05 mm) since corrosion
initiationreferred to herein as Tcr1. In the Liu and Weyers model [27] the time to crack initiation is the time when stresses resulting from the expansion of corrosion products exceed the tensile strength of concrete. This is based on general corrosion (i.e., loss of area is uniform along
rebar). The critical amount of corrosion products needed to cause rst cracking equals the

350

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

amount of corrosion products needed to generate the critical tensile stresses. The time to cracking
is inuenced by corrosion rate, cover, bar spacing, concrete quality, and material properties.
However, it is generally accepted that the service life of a structure is reduced considerably only if
crack widths exceeding 0.30.5 mm are not repaired. Preliminary results from accelerated corrosion tests of RC slabs at The University of Newcastle conrm that the Liu and Weyers model [27]
reasonably predicts the time for rst cracking and suggest that crack widths in the range 0.51.0
mm occur at approximately 10Tcr1 for uniform corrosion rates [16]. A crack width of 0.51.0 mm
represents severe cracking. Thus, for the present study it is assumed that the time to severe
cracking and spalling since corrosion initiation is Tcr=10Tcr1.
The probability that cracking and spalling Pf(t) will occur before the time t is dened as
Pf t PrTi Tcr 4 t PrTi 10Tcr1 4 t

10

where Ti is the time to corrosion initiation and Tcr1 is the time to rst cracking from the Liu and
Weyers model. In the study this probability is calculated by Monte Carlo simulation. In the
probabilistic analysis concrete cover is treated as a normal random variable with a mean equal to
a specied value +6 mm and standard deviation of 11.5 mm [19]. Statistical parameters of random variables related to the models of corrosion initiation, propagation and spalling are summarised in Table 1. Statistical parameters of other random variables representing relevant
concrete properties (compressive and tensile strengths, and modulus of elasticity) are assigned
according to Vu and Stewart [17].

5. Example: LCC of RC structures exposed to marine environment

A common approach to protect carbon steel reinforcement against corrosion is to use better
quality concrete and thicker concrete cover. The approach is employed, for example, in AS 3600
[1], which species a combination of concrete strength (as an indirect measure of concrete quality)
and thickness of the cover depending on an exposure category. For RC members in a splash zone
AS 3600 requires the strength grade of 50-MPa and 50-mm cover; if only one surface of a RC
member is exposed to seawater 40 MPa concrete and 70-mm cover are permitted. For less severe
exposure the requirements are relaxed and, for example, for near-coastal category (150 km
Table 1
Statistical parameters of random variables
Parameter

Mean

COV

Distribution

Surface chloride content, C0

Splash zone
On coast
1 km from coast

7.35 kg/m3
2.95 kg/m3
1.15 kg/m3

0.70
0.70
0.50

Lognormal

Model error (diusion coecient, D)

Chloride threshold concentration, Cr
Model error (corrosion current density, icorr)
Concrete cover

1.0
3.35 kg/m3
1.0
Specied+6 mm

0.20
0.375
0.20
=11.5 mm

Normal
Truncated normal
Normal
Normal

351

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

from coastline) the strength grade of 32-MPa with 40-mm cover or the 40 MPa grade with 30-mm
cover can be used. In this study combinations of 32, 40 and 50-MPa strength grades with 30, 50
and 70-mm covers are considered for three locations(1) splash zone; (2) atmospheric zone
adjacent to the coast; and (3) atmospheric zone I km from coastline. The example will consider
cracking and spalling of exposed surfaces of RC structures subject to chloride contamination
caused by seawater or sea spray. Note that the example concerns individual elements (e.g., beams,
columns, slabs or walls) and not a whole structure where spalling may occur at dierent places at
dierent points in time.
5.1. Cost data
The cost of construction is the sum of material and labour costs. Labour costs have been estimated to be 60% of material costs for RC members [14] and material costs are the sum of concrete and carbon steel reinforcement costs. The baseline case for costs is 40 MPa concrete (i.e.,
nominal or characteristic compressive strength is 40 MPa) and 50-mm cover and it is assumed
that for this case construction cost is CC=1.0. Costs of durability design specications (or
qualityCQA) are estimated from the literature (e.g., [13,28]) and are given in terms of construction cost for the baseline case, see Table 2. Note that the costs in Table 2 are estimated based
on the labour costs for precast construction of RC members and thus may be dierent for cast-insitu members.
The cost associated with of repair/replacement including user losses, etc. CSF, is considerable
and for some structures user losses are often much greater than direct repair, replacement and
maintenance costs. Repair, replacement and user losses are structure and site specic and so it is
dicult to make generalisations about these costs. It is assumed that the cost associated with repair/
replacement without improving durability performance is equal to double the construction cost and
the cost of quality assurance, i.e., CSF=2CC+CQA. This assumption is in line with data reported by
Ehlen and Marshall [29] on removal (CC) and replacement (CC+CQA) of RC bridge decks.
Replacement/repair with improved durability performance implies a maximum concrete
strength grade and cover (i.e., 50-MPa concrete and 70-mm cover). As it will be seen from results
of analyses presented further in the paper, for this design specication the associated probability
of spalling is small for all the three locations considered herein. Thus, this design specication is
used in Repair Strategy II, since the probability that spalling will re-occur in a structure after it
has been repaired can be neglected. Additional costs associated with this design specication can
be estimated from Table 2 (in this case CQA=0.075CC and so CSF=2.075CC). It is assumed that
design, inspection and maintenance costs (CD, CIN and CM) are similar for dierent design
specications and so are not needed for this comparative analysis.
Table 2
Additional or reduced () cost of durability design specications (CQA)
0

Cover (mm)

FC=32 Mpa

fC=40 Mpa

fC=50 MPa

30
50
70

0.072CC
0.025CC
0.026CC

0.048CC
0.0
0.052CC

0.027CC
0.022CC
0.075CC

352

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

5.2. Probabilities of spalling and mean LCC

Life-cycle costs are determined using a 4% discount rate and a service life of 100 years. It is
assumed that inspections are conducted every 2 years (i.e., t=2 years). Results of the analysis
cumulative probabilities of spalling Pf(t) and mean life-cycle costs (normalised to CC), are presented in Figs. 16.
The results show that the lowest mean life-cycle costs for the splash zone and the atmospheric
zones on the coast occur for 50-MPa concrete and 50-mm cover (see Figs. 2 and 4). For the splash
zone this corresponds to the durability design specication in AS 3600 [1]. For a coastal category AS 3600 permits to use of either 35-mm cover with 50-MPa concrete or 45-mm cover with
40 MPa concrete that, as the results demonstrate, reduces initial costs but slightly increases lifecycle costs. For the atmospheric zone 1 km from the coast the lowest mean life-cycle costs occur
for 50-MPa concrete and 30-mm cover (see Fig. 6). while AS 3600 permits slightly less stringent
specications. However, the dierence in mean life-cycle costs between the optimal specication
and the others is not signicant. Thus, the specications in AS 3600 appear to be quite satisfactory.
The results also show that Repair Strategy II ensures lower mean life-cycle costs compared with
Repair Strategy 1. However, the dierence between the two repair strategies decreases with the
decrease in probabilities of spalling and eventually disappears as these probabilities become
suciently low. This result is quite obvious since for low probabilities of spalling the risk of the
re-occurrence of spalling after repair is relatively low.

Fig. 1. Probabilities of spalling in splash zone.

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

Fig. 2. Mean life-cycle costs in splash zone.

Fig. 3. Probabilities of spalling on coast.

353

354

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

Fig. 4. Mean life-cycle costs on coast.

Fig. 5. Probabilities of spalling 1 km from coast.

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

355

Fig. 6. Mean life-cycle costs 1 km from coast.

5.3. Distribution of LCC and its 95% fractiles

In the previous section, the results for life-cycle costs were presented in terms of their mean
values. However, the present value cost of spalling and, subsequently, life-cycle cost are discrete
random variables, since spalling detection and subsequent repairs take place at discrete points in
time (when inspections are carried out) that due to the eect of discounting leads to dierent cost
values, which correspond to dierent probabilities of occurrence. Their cumulative distributions
can be readily inferred from Monte Carlo simulation instead of Eq. (4). In this analysis the
cumulative distribution function for the time of rst spalling, Pf(t), is employed to generate times
of spalling incidents, including consecutive incidents (i.e., after the rst spalling has occurred and
the structure has been repaired Pf(t) is used once again to generate the time of second spalling and
so on). Figs. 7 and 8 show cumulative distributions of the cost of spalling (normalised to CC) for
the two repair strategies. It can be seen that the results are more variable for Repair Strategy I,
when multiple repairs are possible. The distribution curves also show that the cost of spalling is a
discrete random variable. The distributions for Repair Strategy I are smoother (cf. Figs. 7 and 8)
due to the larger number of values which can assume the present value cost of spalling in the case
of multiple repairs.
Since life-cycle cost is a random variable it is may be preferable to use in decision making its
upper fractiles rather than mean values. The mean values and 95% fractiles of the life-cycle cost
for dierent structural specications in the splash zone are presented in Table 3. The minimum
means and 95% fractiles are highlighted for each repair strategy. The 95% fractile obtained for
50-MPa concrete with 70-mm cover is less than the corresponding mean life-cycle cost because

356

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

Fig. 7. Distribution of cost of spalling in splash zone: Repair Strategy I.

Fig. 8. Distribution of cost of spalling in splash zone: Repair Strategy II.

357

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

Table 3
Mean and 95% fractiles of life-cycle cost in splash zone
0

FC (MPa)

Cover (mm)

Repair Strategy I

Repair Strategy II

Mean

95% Fractile

Mean

95% Fractile

32

30
50
70

2.011CC
1.391CC
1.180CC

4.054CC
2.471CC
1.766CC

1.677CC
1.339CC
1.175CC

2.704CC
2.270CC
1.766CC

40

30
50
70

l.533CC
1.176CC
1.101CC

2.978CC
1.916CC
1.370CC

1.432CC
1.169CC
1.101CC

2.584CC
1.875CC
1.368CC

50

30
50
70

1.168CC
1.057CC
1.080CC

2.106CC
1.252CC
1.075CC

1.l61CC
1.057CC
1.080CC

2.077CC
1.261CC
1.075CC

the probability of spalling during 100 years for this specication is less than 5% and so the 95%
fractile is simply equal to the initial construction cost of the structure.
5.4. Eects of time between inspections, discount rate, cost of repair and service life
Parameters aecting the life-cycle cost of a structure include the time between inspections, t,
and the discount rate, r. The results presented in the previous sections were obtained for t=2
years and r=4%. To examine the inuence of these parameters on the life-cycle cost additional
analyses with t=5 years (r=4%) and r=7% (t=2 years) have been carried out for structures
in the splash zone. Results are presented in Figs. 9 and 10 and Table 4. As expected, an increase
in the discount rate leads to reduction of the life-cycle cost. It is assumed in the analyses that the
cost of repair remains constant (i.e., does not depend on the time between the appearance of the
rst signs of damage caused by corrosion and the repair). Under this assumption an increase in
the time between inspections results in a decrease in the life-cycle cost (compare the results in
Tables 3 and 4).
Another parameter obviously having an inuence on life-cycle costs is the cost associated with
repair/replacement, CSF. In the previous analyses it was assumed that CSF is approximately
double the initial construction cost. In order to examine the eect of CSF on the life-cycle cost
and, subsequently, on the optimal durability design specications this parameter it can be presented in a more general form as CSF=aCC+CQA, where the multiplier a can be any non-negative number. Depending on a, the minimum mean life-cycle costs and the corresponding optimal
design specications have been found for RC structures in the splash zone. Results are presented
in Fig. 11 where it is seen that the optimal design specication becomes more stringent as the cost
of repair/replacement increases. Although for most practical situations (a=0.63.6), the optimal
specication is not very sensitive to the cost of repair/replacement, i.e., for the splash zone this
will be 50-MPa concrete with 50-mm cover.
Clearly, life-cycle costs are inuenced also by service life. However, results presented in Table 5
show that the mean life-cycle costs for RC structures in the splash zone increase only slightly as
the service life increases from 75 to 125 years. As a result, the minimum mean life-cycle costs for

358

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

Fig. 9. Distribution of cost of spalling in splash zone, cover=50 mm: Repair Strategy I.

Fig. 10. Distribution of cost of spalling in splash zone, cover=50 mm: Repair Strategy II.

359

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

Table 4
Means and 95% fractiles of life-cycle cost in splash zone, cover 50 mm
Time between inspections, t

Discount rate, r

fC (MPa)

Repair Strategy I

Repair Strategy II

Mean

95% fractile

Mean

95% fractile

5 years

4%

32
40
50

1.377CC
1.165CC
1.055CC

2.425CC
1.900CC
1.242CC

1.329CC
1.160CC
1.055CC

2.135CC
1.779CC
1.253CC

2 years

7%

32
40
50

1.195CC
1.084CC
1.037CC

1.988CC
1.517CC
1.080CC

1.183CC
1.084CC
1.037CC

1.905CC
1.533CC
1.088CC

Fig. 11. Mean life-cycle costs for optimal design specications in splash zone.

75 and 125 year service lives produce optimal durability design specications similar to those
obtained for a service life of 100 years.
5.5. Sensitivity of results to model uncertainties
Statistical parameters of the model errors in the corrosion current density, icorr, and diusion
coecient, D, i.e., Eqs. (7) and (8), used in the analysis were based on assumptions and so are
approximate, although Eq. (8) appears to provide a reasonable t to eld data [17]. Results presented herein are valid if they are relatively insensitive to these assumptions. To check this lifecycle costs have been calculated with COV of the model errors ranging from 0.1 to 0.5 (COV=0.2
in the main analysis) for dierent specications of RC structures in splash zone. The expected
life-cycle costs decrease slightly with an increase in COV of the model errors from 0.1 to 0.5. The

360

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

Table 5
Mean life-cycle costs for splash zone for 75 and 125 year service lives
0

fC (MPa)

Cover (mm)

Repair Strategy I

Repair Strategy II

75-year

l25-year

75-year

125-year

32

30
50
70

1.989CC
1.377CC
1.170CC

2.018CC
1.395CC
1.183CC

1.675CC
1.334CC
1.167CC

1.678CC
1.340CC
1.176CC

40

30
50
70

1.520CC
1.167CC
1.095CC

1.539CC
1.179CC
1.103CC

1.427CC
1.163CC
1.095CC

1.432CC
1.171CC
1.102CC

50

30
50
70

1.161CC
1.053CC
1.079CC

1.170CC
1.058CC
1.081CC

1.156CC
1.054CC
1.079CC

1.162CC
1.058CC
1.081CC

COV of the model error for D has a higher eect on the results than that of the model error for
icorr. The maximum dierence between the expected life-cycle costs was observed for 32-MPa
concrete with 30-mm cover, Repair Strategy I and 2-year interval between repairsfor COV=0.
1 (model error for D) the expected life-cycle cost was 2.024CC, while for COV=0.5 it was
1.950CC, i.e., the maximum dierence was less than 4%. An increase in COV of the model errors
also led to a slight increase in the variance of the life-cycle cost distributions that in some cases
resulted in an increase in 95% fractiles of the life-cycle cost. The maximum increase in the 95%
fractiles was only 3%. Thus, the assumptions about statistical parameters of the model errors
have an insignicant eect on the life-cycle cost analysis.
5.6. Decision making
The results of the life-cycle cost analysis can be used to select design solutions. For example,
when t=2 years and r=4% the optimal specication with carbon steel reinforcement for the
splash zone will be 50-MPa concrete with 50-mm cover if it is based on the minimum mean lifecycle cost (=1.057CC), or 50-MPa concrete with 70-mm cover if it is based on the minimum 95%
fractile (=1.075CC, see Table 3).
If the use of stainless steel reinforcement is considered as an alternative solution then it can be
assumed that stainless steel can be used with the cheapest specication; namely, 32-MPa concrete
with 30-mm cover without risk of corrosion (the initial cost using carbon steel reinforcement is
0.928CC). Thus, for the splash zone, it would be cost-eective to use stainless steel reinforcement
with 32-MPa concrete and 30-mm cover if the resulting increase in the initial cost does not exceed
the expected cost of repairs of the same structure with carbon steel. In this case, an increase in
initial costs of less than 0.129CC (or 13.9% based on 1.061CC0.928CC) when the mean life-cycle
cost is used for the decision, or 0.147CC (=1.075CC0.928CC) when the decision is based on the
95% fractile. Likewise, the use of stainless steel would be cost-eective if it resulted in an increase
in initial costs of no more than 12.6 and 10.6% for the coast and 1 km from the coast zones,
respectively.

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

361

6. Conclusions
The present paper focused on durability design specications and their inuence on serviceability (cracking and spalling) reliability. In conjunction with a probabilistic life-cycle cost analysis this provides a risk-based criterion for optimising repair strategies and various durability
design specications for RC structures in marine environments under dierent exposure conditions. It was found, for example, that the lowest mean life-cycle costs for the splash zone and the
atmospheric zones on the coast occur for 50-MPa concrete and 50-mm cover and that this is
similar to durability design requirements specied in Australia. Life-cycle costs were also presented in terms of 95% fractiles and the sensitivity of results to the discount rate, time between
inspections, service life and cost of failure shown. For the cost data used in the paper a comparison of life-cycle costs suggested that stainless steel reinforcement is cost-eective only if the
construction cost using stainless steel is no more than about 14% higher than the construction
cost of using carbon steel reinforcement, for the splash zone. It has also been demonstrated that
for the repair strategies considered in this study the expected costs of failure for ultimate limit
states are small compared to the expected costs of failure due to spalling and, thus, can be
neglected in the analysis.

Acknowledgements
The rst author acknowledges support for this research by the Fund for the Promotion of
Research at the Technion and Technion V.P.R. FundM. R. Saulson Research Fund.

References
[1] AS 3600. Concrete structures code. Sydney: Standard Association of Australia; 1994.
[2] Knudsen A, Jensen FM, Klinghoer O, Skovsgaard T. Cost-eective enhancement of durability of concrete
structures by intelligent use of stainless steel reinforcement. In: Proc Int Conf on corrosion and rehabilitation of
reinforced concrete structures, Orlando, FL, 811 December 1998 [on CD-ROM]. Orlando (FL): Federal Highway Administration; 1998.
[3] The Concrete Society. Guidance on the use of stainless steel reinforcement (Technical Report No. 51). Berkshire,
(UK): The Concrete Society; 1998.
[4] Kilworth SR, Fallon, J. Stainless steel for reinforcement. In: Macmillan GL, editor. Proc 2nd Regional Conf on
Concrete Durability in the Arabian Gulf, Bahrain, 1921 March 1995. p. 22542.
[5] McDonald DB, Sherman MR, Pfeifer DW, Virmani YP. Stainless steel reinforcing as corrosion protection. Concrete International 1995;May:6570.
[6] Abbott CJ. Corrosion-free concrete structures with stainless steel. Concrete 1997;31(5):2832.
[7] Smith FN, Tulimin M. Using stainless steels as long-lasting rebar material. Materials Performance 1999;May:726.
[8] Stewart MG, Rosowsky DV. Structural safety and serviceability of concrete bridges subject to corrosion. Journal
of Infrastructure Systems, ASCE 1998;4(4):14655.
[9] Stewart MG. Ongoing issues in time-dependent reliability of deteriorating concrete bridges. In: Das P, editor.
Management of highway structures. London: Thomas Telford; 1999. p. 24153.
[10] Thoft-Christensen P. Corrosion crack based assessment of the life-cycle reliability of concrete structures. In: Corotis
RB, Schueller GI, Shinozuka M, editors. Proc ICOSSAR01Eighth International Conference on Structural Safety
and Reliability. Rotterdam: A.A. Balkema; 2001 [CD-ROM].

362

D.V. Val, M.G. Stewart / Structural Safety 25 (2003) 343362

[11] Annual supplement to NIST handbook 135 and NBS special publication 709. Energy price indices and discount
factors for life-cycle cost analysis 1995. Caithersburg, MD: National Institute of Standards and Technology, 1994.
[12] BD36/92. Evaluation of maintenance costs in comparing alternative designs for highway structures. UK:
Department of Transport; 1992.
[13] Troive S, Sundquist H. Optimisation of design parameters for lowest life-cycle cost of concrete bridges. In:
Dunaszegi L, editor. Proc 5th Int Conf on developments in short and medium span bridge engineering98, Calgary, July 1998. Montreal (Quebec): The Canadian Society for Civil Engineering; 1998. p. 110113.
[14] Koskisto O. Reliability-based optimization of plant precast concrete building systems. Acta Polytechnica Scandinavica civil engineering and building construction series no. 108. Espoo (Finland): The Finnish Academy of
Technology; 1997.
[15] Task Group Bond Models. Bond of reinforcement in concrete: State-of-art report. Lausanne (Switzerland): b;
2000.
[16] Stewart MG. Spalling risks, durability and life-cycle costs for RC buildings. In: Proc Int Conf on safety, risk and
reliabilitytrends in engineering, Malta, 2123 March. Zurich: IABSE; 2001. p. 53742.
[17] Vu K, Stewart MG. Structural reliability of concrete bridges including improved chloride-induced corrosion
models. Structural Safety 2000;22(4):31333.
[18] Collins FG, Grace WR. Specications and testing for corrosion durability of marine concrete: the Australian
perspective (ACI SP 170-39). In: Malhorta VM, editor. Proc 4th CANMET/ACI Int Conf on durability of concrete. Farmington Hills (MI): American Concrete Institute; 1997. p. 75776.
[19] McGee R. Modelling of durability performance of Tasmanian bridges. In: Melchers RE, Stewart MG, editors.
Applications of statistics and probability in civil engineering. Rotterdam: Balkema; 2000. p. 297306.
[20] Bamforth PB, Price WF. An international review of chloride ingress into structural concrete. Report no. 1303/96/
9092. Middlesex (UK): Taywood Engineering Ltd Technology Division; 1996.
[21] HETEK. Chloride penetration into concrete: state of the art [report no. 53]. The Danish Royal Directory; 1996.
[22] Papadakis VG, Roumeliotis AP, Fardis MN, Vagenas CG. Mathematical modelling of chloride eect on concrete
durability and protection measures. In: Dhir RK, Jones MR, editors. Concrete repair, rehabilitation and protection. London: E & FN Spon; 1996. p. 16574.
[23] Bolomey J. Granulation et prevision de la resistance probable des betons. Travaux 1935;19(30):22832.
[24] Glass GK, Buenfeld NR. The presentation of the chloride threshold level for corrosion of steel in concrete. Corrosion Science 1997;39(5):100113.
[25] Vassie PR. Reinforcement corrosion and the durability of concrete bridges. Proceedings of the Institution of Civil
Engineers: Part 1984;76(8):71323.
[26] Matsushima M, Tsutsumi T, Seki H, Matsui K. A study of the application of reliability theory to the design of
concrete cover. Magazine of Concrete Research 1998;50(1):516.
[27] Liu Y, Weyers RE. Time to cracking for chloride-induced corrosion in reinforced concrete. In: Page CL, Bamforth PB, Figg JW, editors. Corrosion of reinforcement in concrete construction. Cambridge (UK): SCI, Royal
Society of Chemistry; 1996. p. 88104.
[28] Koskisto O, Ellingwood BR. Reliability-based optimization of plant precast concrete structures. Journal of
Structural Engineering, ASCE 1997;123(3):298304.
[29] Ehlen MA, Marshall HE. The economics of new-technology materials: a case study of FRP bridge decking
[NISTIR 5864]. Gaithersburg (MD): National Institute of Standards and Technology; 1996.