Service Life Estimation and Comparison with Present Conditions of Existing Reinforced Concrete Structures

Service Life Estimation and Comparison with Present Conditions of Existing Reinforced Concrete Structures Bertolini, L. Dipartimento di Chimica, Materiali e Ingegneria Chimica, Politecnico di Milano (email: luca.bertolini@polimi.it) Lollini, F. Dipartimento di Chimica, Materiali e Ingegneria Chimica, Politecnico di Milano (email: federica.lollini@polimi.it) Redaelli, E. Dipartimento di Chimica, Materiali e Ingegneria Chimica, Politecnico di Milano (email: elena.redaelli@polimi.it) Abstract Corrosion is a frequent cause of degradation of reinforced concrete (RC) structures. In order to design a RC structure subjected to reinforcement corrosion, a model which describes the initiation and subsequent propagation of corrosion is required. Several models, which vary from simple equations to quite complex procedures, have been proposed in the literature for a service-life oriented design of reinforced concrete structures. Among them, probabilistic approaches have been proposed since the end of the 1990s, which allow to consider the stochastic nature of the variables affecting the service life of a RC structure. These models, being only recently proposed, need to be validated on the long term. In this regard, the application of the design procedure to structures built in the past and the comparison of its output with the actual conditions arisen from inspection may be useful in understanding the reliability of the models. This paper deals with the initiation of carbonation-induced corrosion. The FIB probabilistic model for service life design was applied to an existing structure that was built in the 70s. This model allows to evaluate the probability of reaching a limit state, which indicates the boundary between the desired and the adverse behaviour of the structure. Data on composition, curing and environmental exposure conditions were available, while data on real concrete cover thickness were obtained from on-site measurements carried out during a recent inspection. The predictions of the model were then compared with the present condition of the structure that was assessed from the results of the inspection. A significant difference was observed between results of the model and results of field inspection. Keywords: reinforced concrete structures, performance-based methods, carbonation, inspection, assessment 239

1. Introduction Reinforcement corrosion is one of the main causes of degradation of a reinforced concrete (RC) structures (Bertolini et al, 2004a). A thin protective oxide film (passive film) spontaneously forms on the surface of reinforcing steel embedded in alkaline concrete. In this condition, corrosion rate of steel is negligible. However, in time, carbonation or chloride penetration can lead to the destruction of the passive film and allow reinforcement corrosion. In a RC structure, the service life can be defined as the sum of the initiation time, which ends when the steel is depassivated, and the propagation time, which finishes when a given limit state takes place, such as cracking or spalling of the concrete cover, beyond which consequences of corrosion cannot be further tolerated. In order to design the service life of a RC structure subjected to reinforcement corrosion, the initiation and subsequent propagation of corrosion should be modelled. Several models, which vary from simple equations to quite complex procedures, have been proposed in the literature for a service-life oriented design of reinforced concrete structures. Among them, probabilistic approaches have been proposed since the end of the 1990s, which allow considering the stochastic nature of the variables affecting the service life of a RC structure. These models, being only recently proposed, need to be validated on the long term. This paper deals with the initiation stage of carbonation-induced corrosion and focuses on the application of a probabilistic model to an existing structure whose present conditions where evaluated by inspection after about 30 years from the construction. The predictions of the model at the age of 30 years are compared with the actual condition of the structure, in order to give a contribution to the evaluation of the reliability of the model. Figure 1 shows a portion of a RC structure, with a steel reinforcement at a depth a, after a certain time of exposure to the environment; the concrete is partially carbonated and the carbonation depth, which is a time dependant variable, is x c. The initiation time is achieved when depassivation of steel takes place. The depassivation limit state function g(t), defined as the difference between the concrete cover thickness, a, and the carbonation depth, x c, is equal to zero: g(t) = a - x c (t) = 0. Figure 1: Portion of a reinforced concrete structure subjected to carbonation penetration showing the concrete cover thickness, a, and carbonation depth, x c 240

a) b) Figure 2: Example of probability density function (a) and frequency distribution (b) of the concrete cover thickness, a, and carbonation depth, xc This simple equation can be solved when the concrete cover thickness and the carbonation depth are constant throughout the structure, so that they can be treated as deterministic variables. However, in a real structure these variables are affected by many factors and are usually rather variable and characterized by uncertainty. In order to take into account this uncertainty, each parameter should be described through a probability density function (pdf). An example of the probability density functions of carbonation depth and the concrete cover thickness is shown in Figure 2a. The curves are partially overlapped, showing that there is probability that somewhere the carbonation depth exceeds the cover thickness. The extent of the overlapped area gives an indication of the amount of the reinforcement already depassivated. In order to estimate the probability that the limit state of depassivation is reached (P dep ), a probabilistic approach is required that takes into consideration the pdf curves. Therefore the initiation time, t i, is achieved when P dep, evaluated as the probability that the limit state function reaches negative values, is equal to a preset target probability, P 0 : P dep P g 0 P a x t 0 P c i 0 As far as the meaning of P dep is concerned, although there is no general agreement (Gulikers, 2007), it can be assumed that it represents the probability that steel is depassivated (i.e. corrosion initiates) in any given point of the structure. Extending the concept to the whole structure it appears reasonable to consider that it represents the relative surface of the steel being depassivated. As a result P dep = 0% means that no reinforcement has been depassivated; on the contrary a P dep of 100% implies that everywhere the limit state is reached and, thus, all the bars are depassivated. Probabilities in between 0-100% indicate the percentage of reinforcement which is expected to be depassivated at the time considered. In the last years probabilistic models were proposed to describe the carbonation propagation and evaluate the initiation time. Among these, the Model Code for Service Life Design published by the International Federation of Concrete (FIB) in 2006, is now being used for the design of structures with special requirements in terms of service life, e.g. for structures exposed to aggressive 241

environments and with the request of design service life equal to or exceeding 100 years. Nevertheless, the reliability of these probabilistic models should be verified on structures which were designed according to them. However, the assessment of the reliability of probabilistic models only be possible when the structures recently designed by the application of these models will be aged, i.e. in rather long lengths time. Existing structures could provide useful data for a first validation of the models. As a matter of facts, a preliminary understanding could be achieved by means of their application to existing structures and comparison of the results to the real corrosion conditions of these structures. As far as existing structures are concerned, during an inspection carbonation depth and concrete cover thickness can be directly measured in selected points of the structure and their statistical distribution can be calculated. For instance Figure 2b shows an example of the frequency analyses of the depth of carbonation and the thickness of the concrete cover. Similarly to the probabilistic approach, also in this case the frequency distributions can be used to estimate the percentage of reinforcement likely to be depassivated, R dep. To evaluate R dep a statistical approach is required as, for instance, the method used by J. Mattila in COST Action 521 (2003). This paper describes the simulation of a durability design of existing RC structures that were built in the 1970s, for which data on concrete composition, curing, environmental exposure conditions and design concrete cover thickness were available. Results of the simulation were compared with the actual corrosion conditions of the steel reinforcement that were obtained from an inspection recently carried out on these structures. 2. Description of the case study The case study concerns eight industrial buildings, with different construction details, of a nuclear power plant located in Northern Italy and built in the early 1970s. In 2002, after about 30 years from the construction, an inspection was carried out on the buildings to evaluate the reinforcement corrosion conditions of the structures (Bertolini et al, 2004b). During the survey, the presence of cracks and detachment of the concrete cover due to corrosion of bars was assessed. Areas for nondestructive tests and sampling were selected in order to be representative of the different exposure conditions. Table 1 shows an overview of the main findings obtained during the inspection and from the archive of the plant, such as the type of cement, the water/binder ratio, the concrete cover and the carbonation depth, assessed during the survey. Further details are given elsewhere (Bertolini et al, 2004b). Each building has been conventionally labelled with a capital letter. Only measurements carried out on the outdoor surface of the RC elements will be considered in this paper. As documented in the archive of the plant, the buildings were built with different cement types and water/binder ratios. Ground granulated blast furnace slag cement (GGBS) was used for buildings A (Reactor), B (Auxiliary), C (Turbine) and D (Diesel), whilst a pozzolanic cement (PZ) was used for buildings E (Off-gas) and F (ERSMA); a portland cement (OPC) was utilized for buildings G (Cooling Tower) and H (Water Intake). For each building the actual water/binder ratio of concrete (which was reported in quality control files) was rather variable, as shown in Table 1; for instance, 242

for building A it ranged between 0.35 and 0.53. No information was available on the w/b ratio of building H. Table 1: Main characteristics of the buildings (# = number of measurements; m = mean value; σ = standard deviation) (Bertolini et al, 2004b) Building Type of w/b * Concrete cover (mm) ** Carbonation binder * Wall Grooves Depth (mm) ** # m σ # m σ # m σ A GGBS 0.35-0.53 121 40 13 108 28 14 31 16 7 B GGBS 0.39-0.50 70 38 10 233 18 11 15 14 8 C GGBS 0.45-0.59 74 44 17 129 28 17 23 23 12 D GGBS 0.45-0.50 42 37 4 42 19 6 5 17 6 E PZ 0.48 50 39 5 - - - 11 33 15 F PZ 0.48 62 58 17 - - - 12 34 7 G OPC 0.56 55 35 13 - - - 17 14 9 H OPC Un. 116 47 24 - - - 16 20 8 * = design values; ** = values from inspection; un. = unknown. The design concrete cover thickness was 50 mm for all buildings. However in some buildings the concrete cover thickness was locally reduced due to the presence of vertical grooves, with depth of about 25 mm and spacing of 1.2 m. From experimental measurements of the thickness of the concrete cover measured on the outermost reinforcement and of carbonation depth, a frequency analysis was carried out. Experimental data both of concrete cover thickness and carbonation depth were fitted with different types of probability distributions functions. According to the chi-square test, the most suitable distribution was found to be the normal distribution. Table 1 shows the moments of the normal distribution (mean value, m and standard deviation, σ) and the number of measurements (#) of the concrete cover thickness measured on the wall and on the grooves and the carbonation depth. It can be observed that, except for buildings F and H, the mean concrete cover was about 40 mm, but the average value decreased to 20-28 mm in the grooves. The carbonation depth varied from 14 mm, measured on building G, to 34 mm on building F. For building A a great number of measurements (Table 1) of concrete cover thickness and carbonation depth was available and from their frequency distributions R dep was determined using the method proposed in (Mattila et al., 2003). The variation interval of the two frequency distributions was divided in n smaller depth intervals, Δx i, of 5 mm each. The proportion of cover depth, p a, and carbonation depth, p xc, measurements falling within every interval was evaluated and R dep was calculated as: R dep i n R x dep i n 1 i 1 p xi a p ( x x c x ) i 1 2 p xi a p xi x c 243

R dep was estimated as 5.8% on the wall and this percentage increased to about 25% in the grooves, due to the lower concrete cover thickness (Table 2). R dep was evaluated also for the other buildings, although for some of them fewer measurements were available, and results are shown in Table 2. Table 2: R dep estimated from frequency distribution of measurements of carbonation depth and concrete cover thickness (Bertolini et al, 2004b) Building A B C D E F G H Wall 5.8% 4.9% 13.8% 0% 25.1% 10.1% 11.4% 16.7% Grooves 25.4% 34.6% 39.2% 49.8% - - - - 3. Evaluation of probability of failure by means of FIB Model Code In this work the probabilistic approach proposed in the FIB Model Code for Service Life Design (FIB, 2006) to describe carbonation induced corrosion was applied to evaluate P dep in the outside surface of the RC structures of the plant, after 30 years from the construction. The limit state equation, functions and parameters involved in this model cannot be reported in this paper, and reference to the FIB Model Code is made for a detailed description. 3.1 Selection of values for the design parameters Values of some variables involved in the limit state equation are provided by the FIB Model Code, while others, such as the concrete cover, a, the inverse effective carbonation resistance of concrete, 1 R ACC,0, the time of curing, t c, the relative humidity of the carbonated layer, RH real, the time of wetness, ToW, and the probability of driving rain, p SR, should be chosen in the design stage in relation with the construction materials and the environmental exposure conditions. As far as the concrete cover thickness is concerned, different scenarios where considered. Firstly calculations were performed using the design concrete cover of 50 mm and a standard deviation of 6 mm, assuming that strict controls were done at the construction site. Then data on concrete cover thickness collected during the inspection of 2002 were taken into account (Table 1). 244

R -1 ACC,0 [(mm 2 /year)/(kg/m 3 )] 14000 12000 10000 8000 6000 4000 2000 0 model data literature data (Bertolini et al, 2007) 0.35 0.4 0.45 0.5 0.55 0.6 0.65 w/b ratio Figure 3: Example of inverse carbonation resistance as a function of water/binder ratio for concrete made with GGBS cement 1 Another parameter which should be established is the resistance of concrete to carbonation, R ACC,0, which mainly depends on the water/binder ratio and the type of cement. In the design of a new structure this parameter should be determined through an experimental test, however no tests were carried out in the construction phase of the buildings considered in this paper. Hence, in order to estimate appropriate values for the mean value of the carbonation resistance of the concrete used for these buildings, data provided by model itself were considered. This, of course, led to the simplification of neglecting possible differences between the cement used for the construction of the building and today s available cements. Besides these data of R -1 ACC,0, values published in a previous work were also taken into account (Bertolini et al, 2007). Figure 3 shows, as an example, the inverse carbonation resistance as a function of the water/binder ratio for concrete made with GGBS cement. A significant difference can be observed between of the two series of data, showing that even considering modern cements different input data could be used. Since the two exponential lines are different especially for high values of w/b ratio, for a given water/binder ratio two mean values of the inverse carbonation resistance were determined; in particular one value was obtained from the fitting line of FIB data and the other from the fitting line of literature data. Moreover from the fitting line of all data also the average value was considered. The standard deviation was calculated from the mean value of R -1 ACC,0 through the equation defined in the model. No experimental data on pozzolanic cement are reported in the model and, in that case, only literature data were used. Because no information was available on the water/binder ratio of building H, a w/b of 0.48 was considered, since the compressive strength of concrete of this building was similar to that of buildings E and F. 245

Table 3: Values of the inverse carbonation resistance, R -1 acc,0, of concrete used as inputs of the model (m = mean value, σ = standard deviation) Building Type of binder w/c 1 R acc [(mm 2 /year)/(kg/ m 3 )] model data literature data (Bertolini et al, 2007) Average value * m σ m σ m σ A GGBS 0.35 1611 777 1694 807 1703 811 0.53 11648 3633 5521 2029 8770 2612 B GGBS 0.39 2500 1094 2202 991 2452 1077 0.50 8377 2809 4534 1740 6674 2353 C GGBS 0.45 4835 1830 3265 1347 4234 1650 0.59 22524 6076 8187 2759 15144 4458 D GGBS 0.45 4835 1830 3265 1347 4234 1650 0.50 8377 2809 4534 1740 6674 2353 E PZ 0.48 - - 3773 1508 - - F PZ 0.48 - - 3773 1508 - - G OPC 0.56 3281 1352 4835 1830 3842 1529 H OPC 0.48 1856 867 1813 851 1804 848 * obtained from the fitting line of all data (both the model and literature data) Table 3 shows the mean value and the standard deviation of R -1 ACC,0 coefficient for each building, determined from the model and the literature data. As reported in the archive data of the plant, concrete was cured 7 days and so the execution transfer parameter, k c, was equal to 1. In order to apply the model, the exposure conditions should be defined in terms of relative humidity, RH real, number of days with rainfall higher than 2.5 mm, which allows to calculate the time of wetness, ToW, and probability of driving rain, p SR. As suggested by the code itself, they were determined from data of the weather station nearest to the buildings (which is about 30 km far). Both for the relative humidity and the number of days with rainfall higher than 2.5 mm, data of different years (from 2002 to 2008) were collected. No information was available for the previous years, but it can be reasonably considered that data from 2002 to 2008 are representative of the real exposure conditions of the structures of the plant. As far as the relative humidity is concerned, daily collected data were available for each year and an average value was calculated. From data of the different years, for both parameters the minimum, maximum and average values were calculated and used as inputs in the model application. Hence, respectively, mean values of 63%, 83% and 75% were chosen for the relative humidity, whilst mean values of 48, 66 and 55 days per year with rainfall higher than 2.5 mm were taken into account, which correspond to 0.13, 0.18 and 0.15 for the time of wetness. Due to the height of the buildings and their complex geometry, three values of the probability of driving rain, p SR, were considered, 0.2, 0.4 and 0.8. They allowed considering different exposure 246

conditions in relation to the wind direction during rain events. A higher p SR means a higher probability that the structures are wet during the rainfall and it could be representative of the top of the building structures, whilst the lowest value could be characteristic of sheltered areas. Table 4 summarizes the mean value, the standard deviation and the pdf of each selected parameter involved in the calculations of P dep. It can be again observed that for the input parameter R -1 ACC,0, RH real, ToW and p SR, since the choice of one single value was rather uncertain and difficult, three values were proposed: in particular a minimum, a maximum and an intermediate one. Table 4: Values of the selected parameters and types of statistical distribution used as inputs in the limit state equations (m = mean value, σ = standard deviation) Parameter Description Unit Type of Value distribution m σ a Concrete cover mm normal Table 1 Table 1 t Time years deterministic 30 - RH real Relative humidity % beta 75/63/83 18 t c Period of curing day deterministic 7-1 Racc Inverse carbonation resistance (mm 2 /year)/(kg/m 3 ) normal Table 3 Table 3 ToW Time of wetness - deterministic 0.15/0.13/018 - p SR Probability of driving rain - deterministic 0.2/0.4/0.8-3.2 Results The results obtained on building A will be first described in details and then the results for the other buildings will be summarised. Initially calculations were carried out using the design concrete cover thickness of 50 mm and an extremely low P dep of about 10-10 was obtained at the age of 30 years (i.e. the time of the inspection). According to such a low value of P dep, the steel bars would be expected to be passive and therefore no visible signs of damage due to corrosion of steel reinforcement should be present on the structure. However this was not the case, as building A showed, during the inspection, the presence of cracks and of the detachment of the concrete cover which indicate that not only the initiation time but also the propagation time was already exceeded (Bertolini et al, 2004b). Calculations were then repeated taking into account the actual distribution of the concrete cover thickness measured during the inspection (Table 1). Since the actual water/binder ratio of the structures of building A ranged between 0.35 and 0.53 (Table 1), and it was not possible to clearly know the real value, P dep was evaluated for the two limit values of w/b and it was assumed that P dep varied between them. For each value of w/b, calculations were performed combining the minimum, the maximum and the intermediate values of the input parameters R -1 ACC,0, RH real, ToW and p SR (cases 1-9 shown in Table 5). Taking into account the intermediate values (case 1) P dep was 0.36% for w/b = 0.35 and 1.35% for w/b = 0.53 on the wall; values of 4.79% and 9.5% were evaluated on the grooves 247

w/b = 0.35 w/b = 0.53 w/b = 0.39 w/b = 0.5 w/b = 0.45 w/b = 0.5 w/b = 0.48 w/b = 0.48 P dep (%) w/b = 0.45 w/b = 0.35 w/b = 0.45 w/b = 0.5 w/b = 0.56 P dep (%) w/b = 0.53 w/b = 0.45 w/b = 0.39 w/b = 0.59 w/b = 0.48 w/b = 0.5 w/b = 0.59 considering a w/b of respectively 0.35 and 0.53. Calculations performed with the different combinations showed that, for w/b ratio of 0.35, P dep ranged between 0.27% and 0.48% in the wall. Similarly in the grooves P dep ranged from 4.1% to 5.59%. Such low differences suggested that a small variation in each of these parameters had a limited effect, which can be reasonably neglected. Since negligible effects were obtained taking into account small variations of the input parameters, only results evaluated with the intermediate values of the parameters, i.e. RH real = 75%, p SR = 0.4 and ToW = 0.15 will be further considered. Table 5: P dep (%) (in bold type) calculated on the wall and on the grooves of building A. For each selected parameter involved in the calculations the mean value is indicated Case 1 R acc RH real p SR ToW w/b = 0.35 w/b = 0.53 [(mm 2 /year)/(kg/ m 3 )] w/b = 0.35 w/b = 0.53 % - - wall grooves wall grooves 1 1703 8770 75 0.4 0.15 0.36 4.79 1.35 9.59 2 1611 11684 75 0.4 0.15 0.35 4.72 1.89 11.4 3 1694 5521 75 0.4 0.15 0.36 4.78 0.84 7.47 4 1703 8770 75 0.2 0.15 0.48 5.59 2.25 12.65 5 1703 8770 75 0.8 0.15 0.27 4.1 0.72 6.88 6 1703 8770 75 0.4 0.13 0.39 4.95 1.52 10.2 7 1703 8770 75 0.4 0.18 0.34 4.58 1.16 8.82 8 1703 8770 63 0.4 0.15 0.46 5.6 1.9 12.28 9 1703 8770 83 0.4 0.15 0.28 3.93 0.97 7.14 5 25 4 20 3 15 2 10 1 5 0 0 a ) A B C D E F G H b ) A B C D Figure 4: P dep in the wall (a) and in the grooves (b) for the different buildings Figure 4 shows the results obtained for the different buildings. As for building A, also for buildings B, C and D the water/binder ratio ranged between two limit values and calculations were performed with both values. Hence, for building B, P dep ranged between 0.075% and 0.28% on the wall. A lower P dep 248

R dep (%) R dep (%) was calculated on the wall of buildings D (< 2 10-5 %) and E (< 5 10-7 %). For each building P dep strongly increased taking into account the concrete cover thickness in the grooves: for instance, for building B, P dep increased from 0.28% on the wall to 20% on the grooves. 4. Comparison between results from inspection and FIB application In order to assess the reliability of the model, its results were compared with results of field inspection. Figure 5 shows the comparison between P dep, which stands for the calculated probability that depassivation occurs after 30 years, and R dep, which represents the percentage of reinforcement expected to be depassivated at the moment of the inspection, carried out after 30 years from the construction. For all the buildings, considering both the wall and the grooves, R dep was higher than the values of P dep and often it was more than double. These results show that the application of the model made in this work led to underestimate the percentage of reinforcement which was no longer in passive state. For instance on the wall of building C, P dep was between about 1.5% and 3.5%, due to the ranging of the w/b ratio between 0.45 and 0.59, and R dep was about 14% (Figure 5a). Even a higher discrepancy was observed in the buildings made with pozzolanic cement (buildings E and F): P dep obtained by the application of the model was lower than 0.1%, whilst R dep calculated from the results of the field inspection ranged between 10 and 25% (Figure 5a). The disagreement between results of field inspection (i.e. R dep ) and results of the model (i.e. P dep ) may be due to various reasons, such as, for instance, the reliability of inspection results, uncertainty in the model parameters or in parameters chosen for the calculations (for instance, the carbonation resistance or the concrete cover thickness). 30 50 25 40 a ) 20 15 10 5 0 Buildings A C E G B D F H 0 5 10 15 20 25 30 P dep (%) b ) 30 20 10 0 Buildings A B C D 0 10 20 30 40 50 P dep (%) Figure 5: Pdep and Rdep for different buildings in the wall (a) e in the grooves (b) As far as the reliability of field inspection results is concerned, it depends on the number of experimental measurements of carbonation depth and concrete cover thickness. It appears reasonable 249

that the available measurements may be insufficient and not fully representative of the whole structure. Limited measurements were available for buildings D, E and F and for these building the major disagreements between results of field inspection and model results were observed. However also for buildings A and B a significant discrepancy was detected nevertheless a great number of measurements was available and hence the discrepancy may not be due to only to this factor. Parameters chosen for the application of the probabilistic model influence P dep. For building A it was observed (Table 5) that the concrete composition (R -1 acc,0) and the environmental conditions (RH real, p SR and ToW) had a limited influence. A sensitivity analysis was carried out in order to better understand the role of each input parameter which should be chosen in the design phase and results will be presented in a paper under submission. The analysis proved that the concrete cover thickness was the most influential parameter. This strong influence of the concrete cover cannot explain the discrepancy between model and inspection results, since in both the evaluations the same set of data of cover depth were taken into account. Even possible differences between cement materials used in the past and those used nowadays, in the opinion of the Authors, is not able to explain the observed differences, since the concrete composition (R -1 acc,0) did not strongly influenced P dep. It can be assumed that disagreements between model and inspection results may depend mainly on other causes. This leads to the possibility that the parameters present in the probabilistic model to calculate the probability of failure, which for instance modify accelerated test results in any real environmental condition, are not fully reliable. Therefore, it seems reasonable to propose the collection of further data from real structures in order to assess the reliability of model parameters. 5. Conclusions In order to evaluate the probability of reaching the initiation time the FIB Model Code for Service Life Design was applied to existing buildings that were built in the 70s. For these structures data on concrete composition, curing and environmental exposure conditions were available. Moreover a recent inspection was carried out and the real concrete cover thickness and corrosion conditions of the steel reinforcement were assessed. From results of the model it was observed that the environmental exposure conditions and the concrete composition only slightly influenced the probability of depassivation due to carbonation. Results of the model were compared with results of the field inspection and a significant difference between them was observed. The percentage of reinforcement expected to be depassivated evaluated by means of inspection results was more than double than the probability of depassivation calculated with the model. The discrepancy between results of field inspection and results of the model may be due to various reasons, such as, for instance, the reliability of inspection results, uncertainty in the model parameters or in parameters chosen for the calculations (for instance, the carbonation resistance or the concrete cover thickness). Therefore, it seems necessary to propose further studies in order to clarify these issues. 250

References Bertolini L, Elsener B, Pedeferri P and Polder R B (2004a), Corrosion of Steel in Concrete: Prevention, Diagnosis, Repair, Weinheim, Wiley-VCH. Bertolini L, Manera M and Anselmi F (2004b), Investigation on reinforced concrete structures of Caorso nuclear power plant in Italy, CNSI/RILEM Workshop on Use and Performance of Concrete in NPP Fuel Cycle Facilities, Madrid. Bertolini L, Lollini F and Redaelli E (2007), Influence of concrete composition on parameters related to the durability of reinforced concrete structures, Proceedings of the International RILEM Workshop on Integral Service Life Modelling of Concrete Structures, Eds. Ferreira R.M., Gulikers J. and Andrade C., Guimarães, Portugal. COST Action 521 (2003), Corrosion of Steel in Reinforced Concrete Structures, Final Report, Eds. Cigna R, Andrade C, Nürnberger U, Polder R B, Weydert R, Seitz E, European Communities, Luxembourg, Publication EUR 20599. FIB (2006), Model code for service life design, Bulletin n 34. Gulikers J (2007), Critical issues in the interpretation of results of probabilistic service life calculations, Proceedings of the International RILEM Workshop on Integral Service Life Modelling of Concrete Structures, Eds. Ferreira R.M., Gulikers J. and Andrade C., Guimarães, Portugal. 251