Bamforth (1998)

Prévia do material em texto

CONCRETE DURABILIITY BY DESIGN: Limitations of the current prescriptive approach and alternative methods for durability design Phillip B. Bamforth¹ INTRODUCTION Concrete durability, or more correctly its lack of durability in some of the more severe exposure conditions, is still a world-wide problem. Many structures exhibit deterioration in too short a time-scale, resulting in unplanned expenditure for the owner/ occupier. Despite many years of research, these problems continue to arise. This lends doubt to the ability to design structures with acceptable and predictable service lives. One of the principal mechanisms of deterioration is reinforcement corrosion caused either by the ingress of chlorides in coastal structures and structures exposed to de-icing salts, or by carbonation. The amount of research in this area has been extensive and many hundreds, if not thousands of papers are available in the public domain. But we still continue, in general, to use a relatively crude, deemed-to-satisfy, prescriptive approach to durability design when it should be obvious, simply by observing existing structures, that a more rigorous approach is needed. This is especially important for structures that require a long service life. A project completed by Taylor Woodrow, with support from UK industry and Government, has collected performance data from the public domain and developed models for predicting the onset of corrosion caused either by chloride ingress or by carbonation, or by a combination of the two mechanisms. Time to cracking may also be estimated. The models use for their input conventional concrete mix design criteria, familiar to all designers. These include the water/binder ratio, the binder content and type (including the quantities of pfa, ggbs, microsilica or metakaolin) and the use of admixtures for waterproofing and corrosion inhibition. Other performance enhancing measures such as the use of Controlled Permeability Formwork (CPF) and surface treatments and coatings may also be evaluated. The predictive models are deterministic, but provide a means for rapidly evaluating different options. More recently, probabilistic approaches have been adopted. These are more closely allied to conventional structural design, which are based on achieving a defined level of reliability. Such approaches are more honest in that they acknowledge that there can be no guarantee that corrosion will be avoided completely. The paper will demonstrate how such an approach can also be used to provide a basis for Whole Life Costing of reinforced concrete structures. The paper offers a justification for the need for more rigorous, performance-based methods of durability design. It describes the development of the models (and their advantages and limitations), provides some of the background data and give examples of their application to achieve long life structures. The paper concentrates on chloride-induced corrosion, as it is the author's experience that if current code requirements are met in practice, carbonation induced corrosion need not be a problem. BACKGROUND TO DURABILITY DESIGN minimum binder content, binder type, minimum AND JUSTIFICATION FOR A MORE grade) and the cover to meet a range of exposure RIGOROUS APPROACH conditions. If these limits are met, the designer is led to believe that the structure will be "durable". National and International codes offer extensive While this would appear to be a sensible approach, guidance on how to achieve durable structures with the extent of premature deterioration continues to regard to reinforcement corrosion, so why do we be excessive (Fig 1) and one only has to consider need a more rigorous approach to design for the the size of the concrete repair market and the prevention of corrosion of steel in concrete? extent of continuing research to realise that the Current codes provide prescriptive limits for the system is not working effectively. concrete mix, (maximum water/binder ratio, 1 Principal Consultant, Engineering Division, Taylor Woodrow Visiting Professor, Dept of Civil and Structural Engineering, Sheffield UniversityA review of the current 'deemed-to-satisfy' In accordance with the above definitions adequate approach to durability design leads very quickly to design for durability demands the ability to be able an understanding of why a more rigorous method to predict performance. An obvious approach for durability design is needed if we are, as an would be as follows; industry, to avoid continuing problems and costly repairs. a) Define the actions (environmental loads) on the structure. b) Define what constitutes the end of the design life, i.e. the condition limit state that defines the end of serviceability. c) Define the required performance of the cover concrete to achieve a low risk of the defined condition limit-state being reached. d) Establish models (scientific or empirical or pseudo-empirical) that define how the actions cause changes in the condition of the structure and use these models to estimate Fig 1 Corrosion of reinforcement causing cracking, the time until the structure ceases to be spalling and delamination of concrete cover serviceable in accordance with the defined condition limit state. The codes adopt one of two approaches to durability design, prevention or retardation. In this respect, current codes and standards Prevention of a deterioration process is achieved generally fall a long way short. by the selection of appropriate mix constituents, such that one of the elements required for the The (environmental) loading on the structure is deleterious reaction is eliminated and the reaction not quantified. cannot, therefore, take place. Examples of this approach include, Codes currently define whether or not a particular action must be considered, e.g. the presence of a) The use of sulfate resisting cement, which has salt, but do not quantify the extent. Consider, for a a low content of tricalcium aluminate, C3A, the moment, trying to design a structure without compound that is attacked by sulfates; knowing the mechanical loads that will be applied to it. This sounds like a ridiculous suggestion, but b) The selection of non-reactive aggregates or it's what we have been doing for many years with low alkali cement to avoid ASR; or respect to design to resist corrosion of reinforcement, the single largest cause of c) The use of entrained air to prevent the build deterioration of RC structures world-wide. up of internal stresses that lead to Exposure classes generally indicate whether or nor deterioration under conditions of freezing and salt may be present, but never how much. So we thawing. start the durability design process without knowing, in quantitative terms, what we are designing Retardation involves a process that against. For other actions the approach is more accommodates the reaction by delaying it's onset rigorous. For example, for sulfate-bearing ground or slowing down the rate at which it subsequently in the UK, there are five classes of exposure based occurs such that the consequences are of little on the actual level of sulphate present in the significance over the design life. This approach is ground [1]. So there is already a long-standing adopted for minimising the risk of reinforcement precedent for such an approach to durability corrosion. Carbonation or chloride ingress is not design. prevented, but the cover zone is designed (through achievement of adequate quality and depth) to prevent the action reaching the steel within the design life.DAMAGE LEVEL Loss of structural integrity Loss of steel section Cracking and spalling Rate of corrosion Threshold level INITIATION PROPAGATION PERIOD OF EXPOSURE Fig 3 Suggested 'condition' limit-states for deterioration caused by reinforcement corrosion. The working design life is rarely specified For durability design to have any value, not only must the condition limit-states be specified but also Fig 2 Differences in the performance of two 4-year the minimum period before these states are old pile caps on the same jetty in a hot wet marine reached. The latter is not an easy task because environment, but subjected to different local we know that not all of the structure reaches the microclimates. same limit-state at the same time. The inherent variability in both the actions on the structure and in the resistance of the concrete to these actions Condition limit states are not defined (cover and concrete quality) will result in some parts of the structure deteriorating more quickly Further limitations in the approach adopted by the than others. An extreme example is shown in Fig 2. codes for durability design can be demonstrated by To specify a single period for the design life is, developing the example for mechanical loading. therefore, on its own meaningless. How much of The structural designer must ensure that certain the structure should be allowed to reach the limit-states are avoided. These include condition limit-state before the design life can be serviceability limit states such as maximum assumed to have ended? For example, if cracking deflections or crack widths and the ultimate limit- is the condition limit-state is it assumed that the state of collapse. In considering the risk of these design life has ended when the first crack is limit states being exceeded, the design life is observed? This does not seem to be a reasonable defined in relation to the return period of the assumption. So how much cracking can be applied loads. But no limit-states are defined in tolerated? This presents a dilemma because there relation to the changing condition of the structure. is no single answer. Some structures may be able Statements are made about periods before the to perform perfectly satisfactorily with cracks over a cost of maintenance and repair becomes significant proportion of the steel. If a structure is 'unacceptable'. But what is unacceptable? required to be aesthetically pleasing no cracking may be acceptable. With regard to corrosion of reinforcement, condition limit-states may include the onset of A way to deal with this dilemma is to use a corrosion, achievement of a defined corrosion rate, probabilistic approach to durability design [2]. cracking, spalling, loss of concrete or steel section Knowing the variabilities associated with concrete and ultimately loss of structural capacity (Fig 3). we cannot predict precisely the time to defined events. We can, however, make an assessment of the risk of certain events occurring and how this risk changes with time, as shown in Fig 4.PERFORMANCE BASED DURABILITY DESIGN - 100 WHAT ARE THE BARRIERS? 80 Cumulative percent failure Having developed an understanding of the 60 mechanisms causing corrosion of reinforcement and the critical influencing factors, then it should be 40 relatively simple to develop a performance-based Acceptable risk method of durability design. 20 Design life Performance based durability design relies on the 0 designer not only understanding the deterioration 0 20 40 60 80 100 120 processes involved but also being able to model Time (years) these processes mathematically. This is a design process that engineers should be very familiar with. Fig 4 A probability curve for a structure with a 60- A structural designer will assess the loads on an year design life and an acceptable risk of element, and design the element to resist these exceeding the defined serviceability limit state of loads using models that have been validated by 20%. testing and by experience of successful applications. To design a structure with a quantifiable service life with regard to durability it is Presenting the results in this way provides a necessary therefore to develop mathematical rational means for quantifying the design life with models for the mechanisms of deterioration that respect to durability. The design life can be can be used with same degree of confidence. The defined as "the time at which the acceptable process is virtually the same. Why then, with the risk of exceeding a specified condition limit information currently available, has this not state is reached". It also provides a means for happened? There are many possible reasons. defining high performance concrete in relation to durability, i.e. concrete that achieves a lower risk of There are no universally accepted methods of deterioration than achieved using current code test requirements. This probabilistic approach to durability design is discussed later. Performance based design requires test methods that provide results that can be related to long-term Deterioration processes are not modelled insitu performance, but none are universally accepted. The intent of the Eurocodes currently in Having defined the actions on the structure, the place, or being developed, was to adopt a working design life and the condition limit state that performance-based approach but the inability to defines when the working design life has ended, a agree on European standard test methods has process is needed to establish the performance of been one of the main barriers to achieving this the cover concrete in order that the design life will intent. be achieved. While models exists, there appears to be a reluctance among engineers to undertake It is my own view that the interests of researchers calculations for durability design that are no more have had a significant impact in slowing down the complex than used in structural design. Very consensus process. All researchers, and I include clearly there is a double standard, there being less myself here, like to develop their own models and willingness to exercise the same rigour to durability associated test methods. The classical engineering design as is applied to structural design [3]. doctorate almost always involves experimental research and developing a new piece of test Based on the above it is clear that significant equipment often forms part of this process. Many changes are needed if a more rigorous approach to predictive models are now available and new tests durability design is to be adopted. However, the for durability are proposed on a regular basis. potential benefits are great. In addition to avoiding These are a little like religions, everyone believes premature deterioration and the associated costs their own to be the right one but no one really of repair, a performance-based approach to knows until it's too late. Also, like religion, durability design will enable the introduction of new conversion is rare. But as long as the academics materials and innovative construction methods and researchers continue to disagree, then there is more readily. A prescriptive approach, by definition, makes it difficult to introduce innovation, little hope of a way forward. and when current code requirements are not sufficiently robust, this may often be to the detriment of the client. So how should we move forward?We try to use physical test to predict performance that may be dominated by chemical processes 1.2E-11 performance test methods may be that we try to Chloride diffusion at 6 months 1.0E-11 A possible reason for the lack of acceptable 8.0E-12 100% pc make short cuts to avoid dealing with difficult (m²/s) 6.0E-12 70% ggbs technical issues and therefore fail to achieve tests 4.0E-12 30% pfa that are universally applicable. For example, many 8% sf 2.0E-12 tests have been developed to measure the ways in which water is transported through concrete. We 0.0E+00 measure absorption under capillary action by depth 0.00 0.05 0.10 0.15 of penetration or weight gain. We measure Sorptivity through surface at 28 permeability under pressure by depth of penetration or flow rate; and from these results we try to predict durability. This seems reasonable, as Fig 5 The relationship between sorptivity at 28 almost all deterioration mechanisms in concrete days and apparent chloride diffusion coefficients involve water, either to transport deleterious agents measured after 6 months coastal, splash zone, or to provide the solution in which chemical exposure reactions occur. Unfortunately, in many cases these physical tests are little better than using strength as an indicator of durability. For concrete Information that is often not reported with data on with a particular set of mix constituents, each of insitu chloride ingress is the cement composition. these transport properties will be related to This can have a major influence on chloride strength. A higher strength concrete, again with diffusion, especially under transient conditions, due particular mix constituents, would be expected to to the effect of chemical binding of chloride. For result in a higher resistance to chloride penetration. example, it is widely reported that the higher the However, because chloride ingress is also C3A content of the cement, the greater the chloride influenced by its chemical reaction with the cement binding capacity [5]. As an example, Frey et al [6] binder and the extent of this reaction will depend have reported results for concretes produced at different w/c ratios using 6 Portland cements and on the cement composition, none of these subjected to chloride exposure for 5 years. These transport properties exhibits a relationship with are shown in Table 1. At a given w/c ratio the insitu chloride penetration that is applicable across change from a low-grade cement (PC35) with low the range of mix types commonly used in contents of C3A and alumina, to a high-grade construction. cement (PC55) with a high content of C3A and The author has measured water sorptivity through C4AF was almost two orders of magnitude. A the cast face of core samples and related the change of two orders of magnitude in diffusion results to the apparent diffusion coefficients coefficient is greater than can normally be achieved measured after 6 months UK coastal exposure [4] by adjusting the w/c ratio over the range for normal (Fig 5). While there is a reasonable relationship for and high performance concrete mixes. Increasing Portland Cement (PC) concretes, a change in the strength from 20 MPa to 80 MPa normally achieves a reduction in diffusion coefficient of cement type resulted in a significant deviation, about 10 times [5]. requiring new relationships to be developed for each of the different mixes investigated. For Table 1 Diffusion coefficients for Portland example, for specific values of sorptivity, concrete cements of different composition [6] with 30% pulverised fuel ash (pfa) exhibited consistently lower values of diffusion coefficient. Cement Dca X 10⁻¹² at w/c of: C3A C4AF Grade 0.4 0.5 0.6 0.7 PC35In a detailed study of literature [7] differences in for extensive and expensive repairs is equally chemical composition was also cited as a reason large. Is it unreasonable, therefore, to expect new for the significant difference in performance of old materials to undergo rigorous and long term type (1930) and modern concretes. testing before they are permitted for use in structures? This demonstrates that the durability, as determined by resistance to chloride ingress, is More work needs to be done to fully validate likely to be influenced to a greater extent by the models chemical characteristics of the cement than by the physical properties of the resulting Many of our difficulties in developing a concrete. performance based approach to design stems from a lack of reliable field data against which to validate This understanding of the influence of the cement either test methods or predictive models. chemistry is important in the development of high (durability) performance mixes, especially when Validation should be relatively simple, but durability blended cements are employed, and could explain differs in one essential respect from structural some of the currently unexplained differences performance - TIME. Models and equations for between otherwise similar concretes. predicting structural behaviour can be tested immediately in a laboratory. Models for durability There is a demand for short-term tests to necessarily require testing over long periods, or predict long term performance under shorter periods using methods of acceleration. Cement may continue to hydrate for As well as trying to simplify tests to measure many years, and the properties of the concrete will be influenced by the exposure condition. As an physical properties only, there is also a tendency to example, a sample of 8 year old concrete that was try to develop tests that offer rapid results. This exposed to chloride for the first time offered less desire to develop short-term tests arises for two resistance to the diffusion of chlorides than the reasons. Firstly, there is often confusion between same concrete that had been exposed to chloride type testing and control testing. The latter, from an age of one month over the same 8 year performed as a contract proceeds to ensure that period (authors results as yet unpublished [8]). the specified mix is achieved and maintained within acceptable limits, necessarily needs to provide A PROPOSED PERFOMANCE MODEL FOR rapid results and be relatively inexpensive because DETERIORATION CAUSED BY CORROSION OF of the many tests that may need to be undertaken. REINFORCEMENT The short time scale and low cost is not necessary, however, for type testing that is used to gain For concrete, all of the deterioration mechanisms approval for a particular constituent material or mix are complex interactions of physical and chemical combination. There are already precedents in the processes. A scientifically based model to predict UK for long term type testing of construction such complex and varying phenomena - even if materials, e.g. a 15 month programme of test is achievable conceptually - would be extremely required to gain approval by the Highways Authority difficult to define mathematically and to execute for waterproof membranes for bridges in the UK. quantitatively. At the opposite end of the spectrum, Such testing is especially important for materials simple empirical models based on a large number such as concrete for which there may be a of observations may not be sufficiently flexible to significant, and sometimes, dominant effect of deal with conditions outside the scope of the data ageing. The industry therefore requires two testing used in developing the model. regimes; one for evaluating the performance of From an engineering point of view, it is desirable new mix constituents, mix designs and other that models, whilst being sufficiently representative supplementary protective measures; and one for of the processes involved, are relatively easy to checking that once an approved mix has been use. Ideally, they should rely on mathematical selected, it is produced consistently throughout a formulations that do not require complex methods construction contract. of solution and, probably most importantly, they The second reason for the desire for rapid testing should include only parameters for which reliable comes from material producers who, not data can be obtained from laboratory or field tests. unreasonably, want to be able to demonstrate that At the very least they should include terms that their material offers advantages and hence be able represent the (environmental) loading and the to achieve rapid commercial benefits. Concrete is resistance offered by the concrete. still the largest volume construction material world- wide. The potential commercial benefits for the While it is acknowledged that chloride ingress material suppliers are enormous, but the potential involves a complex interaction of mechanisms, it iscommonly assumed that it can be approximated to of exposure, chlorides penetrate the near surface a diffusion process. This is because, in many zone primarily by a process of absorption. While conditions, the shape of the observed chloride the chloride profile has a shape that is consistent profile can be fitted using diffusion theory. The with a diffusion process, this shape also represents most common approach is to apply the error the variation of sorptivity with depth as shown in Fig function solution to Fick's second law of diffusion to 8 and it is most likely, during these early stages of derive values of an "apparent diffusion coefficient" exposure, that this is the dominant mechanism. and a "notional surface chloride level" from a measured chloride profile, as shown in Fig 6. The error function equation is of the form: X ) (1) where, Dca(t) is the apparent diffusion coefficient (m²/s) at time t(s) Cx is the chloride content at depth, X(m), after exposure, time, t(s) Csn is the notional surface level of chloride Fig 7 The coastal exposure site at Folkestone erf is the error function 0.5 0.4 Csn Senbetta & Scholar 0.4 Chloride (% concrete) Exposed cure 0.3 Membrane cure 0.3 0.2 Sorptivity 3d wet cure 0.2 0.1 0.1 0 0 1 2 3 4 5 Depth (cm) 0.0 0 20 40 60 Depth (mm) Fig 6 A typical chloride profile obtained from a coastal structure 0 to 15mm Numerous models have been proposed [9] for vary in levels of complexity from very simple analytical models assuming uniaxial diffusion into a homogeneous concrete, to much more sophisticated numerical models that take account Chloride at 6 months (% conc) 0.3 17 to 32mm predicting chloride ingress into concrete. These 0.2 0.1 of variations with depth, the time dependent changes in concrete properties, and chloride binding and leaching. 0 0.00 0.05 0.10 0.15 0.20 In developing the model proposed herein, the author has undertaken exposure trials on a range Sorptivity at 28 days of concrete mixes in marine splash zone conditions in the UK (Fig 7) and observed rates of chloride Fig 8 The relationship between sorptivity and penetration over a 10 year period. Results have depth (top), and sorptivity and chloride content been published for exposure up to 8 years [10]. (authors data and [12]) The results have indicated that in conditions of In the longer term the movement of chlorides is wetting and drying, chloride ingress occurs as a determined by diffusion in the sub-surface zone two-stage process [4, 11]. During the early period that is less affected by changes in surfaceconditions and maintains approximately uniform metakaolin (mk). The model is described in detail and constant moisture content. The proposed in reference [13]. model therefore assumes that the diffusion coefficient is time dependent, with high early life Modelling using a time dependent chloride values representing the absorption component of diffusion coefficient chloride ingress and the lower late life values representing the diffusion controlled part of the While the error function approach has been used process at greater depth. The form of the for many tears for predicting chloride ingress, the relationship between Dca and period of exposure identification and quantification of the age factor can be expressed by an equation of the form: has been a major advance in dealing with time dependent effects. The change in Dca with age is Dca = Dca(tm). shown in Fig 9 for three generic mix types, Portland (2) cement, blends of PC with ground granulated blastfurnace slag, and blends of PC with pfa. While the data are extensive the intent here is not to Dca(tm) is the value of the apparent diffusion confuse the reader. It is to demonstrate that there coefficient derived at time tm. Combining is indeed a very large body of data and that very equations (1) and (2) leads to the following definite trends in behaviour can be detected expression for predicting chloride levels based on a without sophisticated analysis. The scatter is not time dependent apparent diffusion coefficient:- surprising, as no account has been taken of, for example, the composition of the PC, the grade of x concrete, the w/c ratio or the levels of ggbs or pfa. (3) However, despite the scatter a clear trend emerges, particularly at later ages. Beyond 10 years both the ggbs and pfa mixes exhibit values that are largely below 1 X 10⁻¹² m²/s, while values Thus, to predict the rate of chloride ingress and the for the PC concretes are largely greater than this. time to the onset of corrosion the following values On average the difference between the PC and the must be quantified: blended cement mixes is about 10 times at ages that represent the expected design life for many The notional surface chloride level, Csn structures. The apparent chloride diffusion coefficient, Dca(tm) at time, tm 1E-10 The age factor, n The chloride threshold level, Ct 1E-11 It should be noted that Dca is not the actual value of the diffusion coefficient at time tm, but represents the value that, had it been constant over the life of the concrete up to the time of Apparent diffusion coefficient (m²/s) 1E-12 pc ggbs test, would have resulted in the chloride profile measured. 1E-13 pfa The actual value of diffusion coefficient at the time of measurement will therefore be lower than the 1E-14 value of Dca. 0.1 1 10 100 Age (years) This approach was first presented by the author in graphical form [7] and has now been developed Fig 9 Apparent chloride diffusion coefficients into a spreadsheet model on Microsoft Excel. In derived form chloride profiles addition to the author's results, substantial data (several hundred results) have been obtained from the literature, representing structures world-wide Obtaining input parameters [5]. These have been used to derive values for apparent diffusion coefficients and their To predict the rate of chloride ingress and the time relationship with time of exposure and with water- to the onset of corrosion the following values must binder ratio for a range of concrete mix types. be quantified: These include Portland cement, PC, and blends of PC with pulverised fuel ash (pfa, also called fly The notional surface chloride level, Csn ash), ground granulated blastfurnace slag (ggbs), The apparent chloride diffusion coefficient, micro-silica, (ms, also called silica fume, sf) and Dca(tm) at time, tm The age factor, nThe chloride threshold level, Ct wash-out from the surface and the low diffusion slows down the rate at which the chlorides can Values for these input parameters can be derived migrate into the surface. either from field data or from laboratory tests. In either case care has to be taken in obtaining and These differences must be taken into account interpreting the results. When collecting field data when modelling chloride ingress. assumptions may have to be made about the concrete mix and its constituents and about the Apparent diffusion coefficients and age factors exposure conditions over the life of the structure. Fig 2 illustrates very clearly how microclimates can Two of the principal factors influencing the influence performance. Laboratory testing provides apparent chloride diffusion coefficient, Dca, are the greater control over the concrete mix and the chemistry of the cementing materials (binder) and exposure conditions but rarely offers the real time the water-binder ratio. The chemistry influences exposure to take account of aging effects. Tests not only the initial value of Dca, but also how it may, therefore, be accelerated by increasing the varies with age. The relationship defining the severity of exposure (typically by increasing the change in Dca with time is given as equation 2. concentration of chloride in the test solution or by Age factors, n, have been derived from a increasing the temperature) and the results must comprehensive analysis of published data for a then be factored to reflect insitu behaviour. The range of mix types. This analysis has been widely complex aging process, which itself appears to be published [13-16] and proposed design values are influenced by exposure to chloride, makes this as follows:- difficult. Hence, reliance must be placed on correlation with field data, recognising the Portland cement concretes -0.264 uncertainties discussed above. pfa concretes -0.699 ggbs concretes -0.621 Surface chloride levels For these mix types the value of n has been The surface chloride level is determined primarily assumed to be constant for a particular mix type by the proximity to the source of the chlorides. For and unaffected by w/binder ratio. For ms example, surface chloride levels tend to reduce concretes, however, this assumption has been with height above sea level and with distance from found to be incorrect. the coast [5]. It is also a function of the type of concrete used and tends to be higher for mixes Results obtained in the UK [10, 14, 15] suggested that have a high chloride binding capacity and a little ageing for ms concrete and a consistently high high resistance to chloride penetration. For such value of n (-0.141). Concretes investigated by concretes the chlorides captured on the surface Gjorv et al [17], Maage & Helland [18] and the find it more difficult to escape, for example, under Lettkon project in Norway (Maage et al [19]) all wash-down conditions. A comprehensive review of produced much lower values of diffusion coefficient surface chloride levels has been carried out by the and a significant effect of age. The most notable author [14] and values have been recommended difference between these sets of results was the for predictive and design purposes. These are w/binder ratios employed. A study by Gautefall [20] given in Table 2. The designer can select which investigated the effect of w/binder ratio, as did values to use in combination with assumptions Berke [21] and Hansen et al [22]. These results about the other input parameters and the are shown in Fig 10. acceptable level of risk. An important point to note here is that none of the Table 2 Recommended surface chloride levels (% authors could have developed the relationship weight of concrete) for use in design [14] between the age factor, n, and the w/c ratio in Fig 10 using their data alone. Indeed, there had been Mix type Typical Upper 90% Upper 95% vigorous discussion' between some of the authors confidence confidence about the performance of ms under aging limit limit conditions. It was only by a rigorous analysis of all PC concrete 0.36% 0.70% 0.79% of the available information that it was found that all of the results were part of the same population and blended cement 0.51% 0.85% 0.94% mixes (pfa, ggbs) obeying the same relationship. It should be noted that the values are higher for the An important lesson is to be learned here. Too pfa and ggbs concrete. This is partly due to the often, differences between the results of different higher chloride binding capacity of these mixes, researchers are used as a reason to do more and partly due to their lower diffusion experimental research. Perhaps we should characteristics. The binding capacity reduces properly re-search the information already availablebefore embarking upon further experimental become increasingly clear that there is no single studies. Perhaps we should consider initiating value that represents the wide range of concreting doctorate studies that involve only re-search of materials and exposure conditions. However, in a existing knowledge. Other disciplines are forced to comprehensive review of published data Glass [23] do this. For example, history cannot be generated concluded that "At present, the chloride in a laboratory, but we can learn a lot from it threshold level is best considered in terms of nevertheless; and the effort needed to undertake corrosion risk". This approach was suggested such research is no less demanding. earlier by Browne [24] who proposed the risk classification in Table 3. These recommendations 1E-11 are broadly consistent with data from UK bridges w/binder [25], shown in Fig 12. Over 450 results were obtained which demonstrated that below 0.2% CI Apparent diffusion coefficient (m²/s) 0.7 (% wt of cement) the risk was very low. Above this 0.6 level the proportion of corroding steel associated with each level of chloride increased in a way that 1E-12 0.5 was consistent with a normal distribution of threshold levels. 0.4 Table 3 Risk of corrosion in relation to chloride 0.3 content 1E-13 0.01 0.1 1 10 100 Chloride Risk of Corrosion (% wt of cement) Period of exposure (years) 2.0 Certain ms concretes Relationships between Dca and water-binder ratio are shown in Fig 11. The results have been 120% normalized using the above age factors to 100% represent values expected after 20 years of Percent of bars corroding Results exposure. In the model, equations defining the 80% Best fit curves in Fig 11 are used to derive 20-year values 60% for Dca for the different mix types. Having defined one point on the time axis, the age factor is then 40% used to generate the full curve. 20% 1E-10 0% pc 0.0 0.5 1.0 1.5 2.0 ms Apparent diffusion coefficient (m2/s) 1E-11 Chloride content (% cement wt) pfa & ggbs 1E-12 Fig 12 The risk of corrosion at different levels of chloride [25] 8 8 1E-13 Many factors appear to influence the threshold level. Some of these are environmental. For example Sandberg et al [26] proposed that 1E-14 threshold levels are lower under conditions of 0.2 0.3 0.4 0.5 0.6 0.7 0.8 wet/dry cycling. The mix proportions may also be Water/binder ratio influential with lower threshold levels being achieved in mixes with high w/c ratios. The Fig 11 The relationship between Dca (normalised commonly used value of 0.4% CI (wt of cement) for 20 years of exposure) and water-binder ratio appears to be most applicable in conditions of [15] wet/dry cycling of high (>0.6) w/c ratio mixes [26], Chloride threshold levels while in mixes of low w/c ratios, under more stable moisture conditions, higher values may be While considerable research has been carried out tolerated. Blended cements tend to exhibit in an attempt to define threshold levels, it has threshold levels that are similar to, or lower than,that of PC concrete and corrosion inhibitors will cement. In particular, the general climatic increase the tolerance to chlorides. To determine conditions (i.e., temperature, humidity, time of the time to corrosion activation based on achieving wetness) and the microclimate (proximity to a chloride threshold level, the appropriate threshold moisture, orientation) will influence the internal level must be defined. moisture state of the concrete and the temperature will influence both the rate of chloride migration and Corrosion rate the threshold level. Andrade et al [28] have measured corrosion rates and have suggested While the prediction of chloride ingress is upper limits for different exposure classes as important, it is generally the time to cracking that follows: concerns most clients. This is the point at which problems are first noted and when interventions are Wet, rarely dry 0.5 µA/m² = 6 microns/year first considered. Measurements on the author's exposure blocks [15] indicated that there is a Airborne seawater and cyclic wet/dry relationship between chloride content, and 5.0 µA/m² = 60 microns/year corrosion rate, CR (Fig 13). Within the constraints of the study the corrosion rate appeared to be Tidal zone 10.0 µA/m² = 120 microns/year independent of the cover depth and can be expressed by an exponential equation of the form:- Assuming the experimental relationship between corrosion rate and chloride content, exponential CR = (4) equations have been derived for the above three conditions which pass through the points (0.5% CI, Assuming that corrosion is negligible for values of 1.2 microns/year and 3.0% CI, maximum corrosion CRcracking reduces as the tensile strength of the Age factor vs. w/binder ratio (for ms concrete increases. The reason for this is that the concretes) resistance of the concrete to cracking is strain, and Corrosion rate chloride content, for not strength, limited. As corrosion products are different exposure conditions generated they attempt to diffuse into the Threshold level percentage addition for surrounding concrete. In poor quality concrete with different binder combinations a relatively high porosity the corrosion products are Threshold level vs. temperature accommodated much more easily than in high quality concrete with relatively low porosity. As the If chloride profiles have been obtained from a forces that can be generated are greatly in excess structure, derived values of Dca and n can be used of the tensile strength of even high quality concrete to predict future behaviour. However, in such a it is the porosity, therefore, which is the critical case, where there are data at only one age, factor. In equation 5, the tensile strength is an knowledge of the mix is required to estimate the indirect measure of the porosity, hence as fct age factor. increases, the porosity reduces and reduces also. The tensile strength is not commonly Validation of the model measured but well-defined relationships have been developed between the tensile strength and the As the equations used for predicting chloride cube compressive strength of concrete and these ingress have been derived from data from can be used if direct measurements are structures and exposure trials the prediction of the unavailable. rate of chloride ingress, necessarily, represents physical observations. An example is shown in Fig THE SPREADSHEET MODEL 14, comparing results from the author's exposure trials with those predicted by the model. Based on the foregoing theory and relationships Water cured pc concrete derived from extensive data, a spreadsheet model 4 has been produced which runs on Microsoft Excel The spreadsheet model first predicts the level of 5mm chloride at different cover depths and time 3 increments using equation 3. A second stage predicts the rate of corrosion within each defined Chlorides (%cement) 15mm 2 25mm time increment based on the predicted chloride 35mm content at the cover depth. The increments of 1 45mm corrosion are then summated to determine the total corrosion with time. Predictions are based on limited knowledge of the concrete and the 0 environment, but requires the following input: 0 2 4 6 8 10 Period of exposure (years) Cement type and content (including the percentage of mineral additions) Water-binder ratio Water cured pfa concrete 4 Temperature Surface chloride level, as percent weight of 5mm concrete 3 The values of apparent diffusion coefficient, Dca, and the age factor, n, can be computed within the Chlorides (%cement) 2 15mm model using a series of algorithms developed from background data. A spreadsheet database 1 containing about 2000 lines of data was used to 25mm develop relationships as follows, 35mm 0 45mm 0 2 4 6 8 10 Apparent diffusion coefficient water/binder ratio, for different binder Period of exposure (years) combinations Apparent diffusion coefficient percentage Fig 14 Predicted and measured rates of chloride mineral addition penetration in concretes with low and high Apparent diffusion coefficient resistance. Data from coastal exposure blocks [4] temperature In relation to corrosion, the only absolute measure Age factor vs. percentage mineral addition, is gravimetric and there is little data available in this for different binder combinations form. However, Thomas [27] has published the results of measurements from 4-year exposuretrials, in which weight loss measurements were that predictions are broadly consistent with related to chloride levels in order to derive observations for concrete alone. However, it is threshold levels for different mix types. Using the now common to enhance performance using published values for binder type and content and various additional measures, particularly in very water-binder ratio, the model was used to derive aggressive exposure conditions. These additional curves relating reinforcement weight loss to the measures are accommodated by varying one or chloride level at the cover depth for mixes based other of the parameters within the predictive model. on PC and a 70:30 blend of PC and pfa. The A summary of the parameters that are influenced predicted results are shown in Fig 15 and compare by a range of supplementary measures is given in well with Thomas's own data. Indeed, the model Table 4 predicts performance that is more consistent with the results than Thomas's linear relationship. It is Table 4 Supplementary measures for reducing apparent that at high levels of chloride there is a the risk of reinforcement corrosion and the way in very high degree of corrosion. The linear which they influence the parameters in the relationship between chloride content and weight corrosion model [30] loss grossly underestimates these high values, while the TE model predicts their occurrence. Supplementary Influence on measure Cs Dca Ct Cover Integral waterproofer reduction a) Predicted using the spreadsheet model Corrosion inhibitor INCREASE 3.5 Reinforcement INCREASE 3.0 pc concretes OR CPF INCREASE INCREASE pfa concretes 2.5 Weight loss (%) GRC formwork INCREASE 2.0 Coatings and surface reduction treatments 1.5 1.0 0.5 Integral Waterproofers 0.0 0 1 2 3 4 The principal effect of an integral waterproofer (WP) is on the surface chloride level. Within the Chloride content (% binder) model it is assumed that if a normal WP is used the value of Csn is reduced by 10%. For high range WP's the reduction is 20%. b) Reported by Thomas (scanned from ref. 27) Corrosion Inhibitors 3 OPC The principal effect of corrosion inhibitors is on the 15% P1 level of chloride that can be tolerated before Rebar mass loss (%) 2 Threshold chloride = 0.39% 30% P1 corrosion commences. Calcium nitrite is the most 50% P1 widely used inhibitor and the one for which there is the most data. The relationship between the 1 amount of inhibitor added (I in litres/m³) and the corrosion threshold level, Ct, can be represented Processing loss - 0.087% by an equation of the form: 0 0 1 2 3 4 Cti = Cto + 0.06. I. (6) Chloride content at bar location (mass % cement) Where Cto is the base threshold level for the combination of concrete mix type and steel used Fig 15 Comparison of measured and predicted values of chloride content and mass loss of steel Cti is the revised threshold with the inhibitor add f₁ Taking account of additional measures to is a safety factor to take account of uncertainties in enhance durability assuming a similar relationship applies to all situations. The model has been derived empirically, and has been validated against independent data, indicatingWithin the model, the rate of corrosion is related to A review of data indicated that, on average, the the difference between the actual chloride content value of DCPF is about 45% of concrete cast against and the adjusted threshold value. conventional forms and this value is used in the model. A factor of 0.9 is also applied to the surface Type of reinforcement chloride level. The type of reinforcement is dealt with in the same Permanent GRC formwork way as a corrosion inhibitor in that it changes the chloride threshold level. Three types of steel are The effect of permanent GRC formwork is also included and each has been allocated a baseline accommodated in the model through increased threshold value It is recognised that once cover. Two approaches can be adopted. If no galvanised steel becomes corrosively active it will testing is carried out on the GRC then the apparent corrode very quickly. In the model it is assumed, diffusion coefficient can be calculated with a therefore, that the time to cracking will be the same knowledge of the w/binder ratio. The equivalent as the time to corrosion initiation. Threshold values increase in cover is then obtained using equation 9. used in the model are shown below; Type of Chloride threshold = dGRC (9) reinforcement level (% cement) Conventional 0.4 If measurements are made on the GRC then these reinforcement can be used in the derivation of the equivalent Galvanised steel 1.0 cover. Stainless steel (316 or 3.0 Coatings and surface treatments better) Coatings and surface treatments can be dealt with The threshold level for stainless steel is set at a by making an adjustment to the surface chloride level that is about the same as a typical surface level at the concrete / coating interface. This chloride level in the splash zone condition. Hence, adjusted value may be obtained by measurement even with nominal cover, activation will not be or by calculation. In general the application of a predicted within the lifetime of even long life coating will significantly reduce the level of chloride structures at the interface and the enhanced performance will depend on maintenance of the coating. Each of these threshold values may be revised if more data becomes available. 0.40 Controlled permeability formwork 0.35 0.30 CPF changes the characteristics of the near uncoated surface zone through a reduction in w/binder ratio. In the model the enhanced performance of the Chloride level (%conc) 0.25 0.20 affected zone is accommodated by an effective increase in cover. The affected zone, (d, in mm) is 0.15 itself a function of the initial w/binder ratio. Based 0.10 on limited data a simple relationship has been derived to estimate the depth of the affected zone. 0.05 This is of the form; 0.00 0 10 20 30 40 50 = 1 + 40. (w/b) (7) Depth (mm) Having established the depth of the affected zone this is then adjusted to take account of the reduced Fig 16 Chloride profiles recorded in coated chloride diffusion coefficient, DCPF through this concrete blocks exposed to UK coastal splash zone. This is done using the equation; zone conditions for 5 years [author's unpublished data] = (8)Combining measures General procedure for durability design When combining measures to enhance durability it The model is simply a tool that aids the decision is generally advisable to use measures that making process in design [30]. The durability influence different parameters. If pfa or ggbs is design procedure is divided into eight steps: used to reduce the apparent diffusion coefficient, then additional benefit could be achieved by 1. Specify the service life, i.e. time to reach adopting a further measure to either reduce the defined serviceability and/or ultimate limit states. surface chloride level, such as an integral These could be: waterproofer, or to increase the threshold level, such as a corrosion inhibitor. It is much more Onset of corrosion difficult to achieve additional benefit by trying to Defined rate of corrosion enhance a property that is already improved. Extent of cracking Spalling Predicted times to corrosion and cracking Loss of steel and concrete section Excessive deflection Values of time to corrosion and time to cracking Collapse are given in Table 5 for a range of options that can be used to enhance resistance to reinforcement Different limit states are appropriate for different corrosion in splash zone conditions. Results structures and elements of structures and the obtained using supplementary protective measures clients must have an input here. are compared with a benchmark PC concrete, and with adjustments that can be made to cover and 2. Define the exposure/service environment. w/c ratio. Exposure classes in design codes provide guidance for specific actions e.g., chlorides or Table 5 Times to corrosion and cracking [T carbonation. To use the model it is necessary to indicates typical values, E indicates extreme define specific levels of action. For chloride values] exposure this will be the surface chloride level. Measure adopted 20-year Dca m/s Csn (% conc) Ct cement) Time to exceed Ct (years) Time to cracking 3. Identify the current minimum requirements (years) of the code with regard to cover and concrete mix details. PC concrete, w/c =0.42, T 0.82 0.35 0.40 33 53 4. Apply the models to predict the service life cover = 50mm E 1.39 0.51 0.40 11 20 in relation to the service environment and the T 0.59 0.35 0.40 51 79 Reduction in w/c ratio to 0.35 E 1.01 0.51 0.40 17 28 selected serviceability limit states. The chloride T 0.82 0.35 0.40 99 148 model requires input data for: Increase cover to 75 mm E 1.39 0.51 0.40 32 50 Reduced w/c and increased T 0.59 0.35 0.40 153 >200 cover 1.01 0.51 0.40 50 74 surface chloride level - related to exposure T 0.2 0.48 0.32 >200 >200 class 30% pfa concrete E 0.33 0.62 0.32 196 >200 apparent diffusion coefficient - related to T 0.19 0.49 0.40 >200 >200 65% ggbs concrete mix type and surface treatments E 0.31 0.70 0.40 177 >200 T 0.32 0.50 0.30 124 >200 n, age factor - related to cement type 10% sf concrete E 0.54 0.67 0.30 28 49 threshold level - related to cement type, T 0.37 0.50 0.34 58 84 15% metakaolin concrete admixtures, w/binder ratio E 0.67 0.67 0.34 22 35 T 0.82 0.35 1.00 >200 >200 20 litres calcium nitrite E 1.39 0.51 1.00 52 73 For each of the supplementary measures, changes High range integral T 0.82 0.28 0.40 44 73 in these factors have been quantified, enabling the waterproofer E 1.39 0.41 0.40 13 25 benefits to be predicted in terms of an extension to Controlled permeability T 0.82 0.32 0.40 58 91 the service life. formwork E 1.39 0.46 0.40 18 31 GRC permanent formwork, T 0.82 0.35 0.40 94 140 15mm thick, w/c = 0.4 E 1.39 0.51 0.40 40 59 5. Compare the predicted service life (i.e., the Coating 1, 50% reduction in T 0.82 0.18 0.40 100 >200 Csn E 1.39 0.26 0.40 24 46 time to reach the selected limit state) with the Coating 2, 80% reduction in T 0.82 0.07 0.40 >200 >200 design life for first order selection. Csn E 1.39 0.10 0.40 >200 >200 T 0.82 0.35 1.00 178 178 Galvanised steel a) If the code requirements result in a service E 1.39 0.51 1.00 36 36 T 0.82 0.35 3.00 >200 >200 life that is greater than the required service Stainless steel E 1.39 0.51 3.00 >200 >200 life then no additional measures are required. However, there may be scope for economies through optimisation if the code grossly exceeds the required service life.b) If the code requirements result in a service distributions of load and resistance overlap. Partial life that is less than the required service life, safety factors are therefore applied to ensure that, then additional measures must be within the design life, the risk is kept to an considered. Also, the magnitude of the acceptable level by appropriate separation of the required improvement must be assessed. two distributions. 6. Select systems that offer the appropriate For structural design, the risk of failure must be level of enhancement. The selection process very low. For example, for ultimate limit state (i.e. begins using a "coarse filter" to reject systems collapse), Eurocode 1 is based on a probability of 7 which, for the particular service environment, are X 10⁻⁵ for T = 50 years. A higher level of risk may unlikely to achieve sufficient technical be tolerated in relation to serviceability limit states, improvement. As the systems have been selected as the consequences are much less severe. because they have demonstrated benefits, few Furthermore, there is generally a visual warning systems will be eliminated at this stage. This will long before the defect has serious safety provide a first level of selection based on implications, with the opportunity for intervention to improvement in durability alone. reinstate the structure and to prevent further damage. There are cost implications, however, 7. Short list by introducing other factors: and for this reason the risk of corrosion should still be designed at an acceptably low level. A i) Achieving design life (the coarse filter will only probability of the onset of corrosion of 10⁻² may be indicate improvement qualitatively) by more appropriate. application of the models ii) Constructability, other benefits, disbenefits Service Period Design iii) Maintenance requirements Distribution of R(t) iv) Appearance R(t) v) Cost vi) Health and safety requirement R,S 8. Optimise the Short-listed Options Use the model to predict design lives for each of S(t) Time the short-listed measures and to optimise each Distribution of S(t) solution (e.g. how much pfa or ggbs or ms, the Ptarget Mean lifetime level of corrosion inhibitor, which coating etc.) Service period Limitations and developments Lifetime distribution It is recognised that the model is currently limited in Lifetime Design many respects and it is still being developed. For example, there are still uncertainties about the Fig 17 Service period design and lifetime design effects of temperature on chloride penetration [31, [33] 32]. In addition, threshold levels are affected by many parameters, other than the cement type, e.g. The importance of taking account of variability can the water-cement ratio, the cover depth and the be demonstrated by a simple example. Consider state of the steel/concrete interface. A sub-routine an element in one of the most severe, splash zone to deal with these factors must be developed. The exposure conditions. UK codes require the use of prediction of time to cracking is also very simplistic a low w/c ratio concrete (400 kg/m³). The time to achieve a the model is very quick to use and yields results chloride threshold level of 0.4% (cement weight) that are not inconsistent with the observed has been calculated using both typical values of performance of structures. and and upper 80 percentile values. The results are given in Table 6, together with the PROBABILISTIC APPROACH TO DURABILITY assumptions used in the calculations. DESIGN The differences between the typical case and the In order to take account of the variabilities and (more extreme) design case are substantial and uncertainties in durability design, a probabilistic demonstrate the importance of selecting approach may be adopted. Again, this approach is appropriate values for the input parameters. similar to that used in structural design [33. 34]. The basic principle is very simple and is illustrated Table 6 Calculation of time to corrosion activation in Fig 17 [33]. The risk of failure, i.e. the loss of for Portland cement concrete with a w/c of 0.45 in a serviceability as defined by a particular condition marine environment limit state, is determined by the extent to which thehundreds of observations showed evidence of Typical (Case 1) Design (Case 2) corrosion even at chloride levels in the range 0.2% Cover Csn = 0.36 Csn = 0.58 to 0.35% wt of cement. However, at some depth (mm) Dca = 9.4 X 10⁻¹³ m²/s Dca = 1.47 X 10⁻¹² m²/s locations with much higher chloride levels n = -0.264 n = -0.264 (exceeding 1.5%) no corrosion was observed. The 25 4 years 1 year data are illustrated in Fig 12 and can be 50 26 years 9 years represented by a normal distribution with a mean of 75 77 years 26 years 1.1% chloride and a Standard Deviation of 0.6. It is interesting to note that the value of chloride that Fig 18 shows a cumulative frequency of chloride represents a 5% risk is 0.33% for the site contents at different rebar depths and how this conditions. This indicates the importance of changes with age. adopting characteristically low values for design. At the commonly assumed threshold of 0.4%, the risk 100% level was 12% and this increased to 42% at a chloride level of 1%. 80% Advantages of probabilistic analysis Frequency 60% 50 years Adopting the probabilistic approach explains why 40% 20 single observations of apparently similar structures 10 exposed in apparently similar environments can 20% vary widely. They are all part of the same 5 population but at different points within the overall 0% distribution. This also explains why the relative 0.0 0.5 1.0 1.5 2.0 2.5 performance of different materials or systems applied to concrete structures may differ between Chloride at the steel (% cem) structures and between researchers. Fig 18 Calculated cumulative distribution of Probabilistic analysis also enables a rational chloride levels at 50mm depth and its variation with definition of design life based on the time for the time of exposure risk of a defined serviceability limit state to be reached (Fig 4). The influence of changes in The data can also be presented to show the specification and design can be presented in terms probability of different threshold values being of the change in risk, enabling the Client and the exceeded and how this probability changes with Designer to understand the implications and to time (Fig 19). It can be seen that, regardless of make decisions accordingly. For example, the whether the threshold level is assumed to be 0.4% performance of different systems for enhancing or 1.0%, there is still a significant risk of these durability can be assessed either by comparing the values being exceeded after a relatively short relative levels of risk after a defined period, e.g. the period of exposure in relation to normally expected design life, or by comparing the relative times for design lives (after 20 years, the probabilities are the level of risk to exceed a defined limit, e.g. 5% about 75% and 20% respectively). of the reinforcement becoming active or 2% of the surface requiring repair. Ct 100% 0.40% USE OF PROBABILISTIC ANALYSIS IN LIFE 0.70% 80% CYCLE COSTING Probability of threshold value bieng exceeded 1.00% 60% There is a growing recognition of the need for Life 1.50% Cycle (or Whole Life) Costing (LCC) in order to 40% demonstrate to clients the value of designing for greater durability and hence the benefit to society 20% 2.00% in the long term by reducing the utilisation of raw 0% materials and the energy required in their conversion to building materials and components. 0 20 40 60 80 100 Time (years) Fig 19 Calculated probabilities of exceeding Life cycle costing involves estimating the total costs different chloride threshold levels associated with construction, operation, The threshold level is also subject to variability and maintenance, repair and demolition. To take uncertainty. For example, a comprehensive study account of the fact that different operations will of bridges in the UK [25] involving several take place at different times, incremental costs areconverted to current costs using a discounted cash flow system. This takes account of interest rates Input Data and inflation through the discounted cost. As described above, the performance of a At its simplest level, therefore, LCC is relatively particular structure or element (as affected by insitu straightforward. If the cost in year t is equal to Ct, environment can be defined by a distribution and the discount rate is r, then the life cycle cost for function (or histogram) and a probability function a structure with a design life of N years, expressed (or cumulative distribution function). The example as the cost at current value, is as follows: in Figure 20 shows the (normal) distribution functions for a structure with a design life of 60 t=N years at which time no more than 5% of the Present Cost = Σ (10) structure should have exceeded its serviceability r t=0 + 100 limit state. It is assumed that the Standard Deviation on the service life is 30 years and this To carry out a life cycle cost analysis it is results in the requirements for a mean service life necessary to make predictions about the long-term (when the risk of failure is 50%) of 109 years. performance of a building or structure. In particular values must be ascribed to the following; 20 1. The capital cost of construction 15 2. The cost of routine maintenance 3. The rate of deterioration 4. The level of deterioration at which intervention is Percent failures 10 5 required 5. The cost of repairs 0 6. The cost of lost production during the repair 0 50 100 150 200 process Time (years) 7. The rate of deterioration of the repairs 8. Any other costs resulting from the need to maintain and repair the structure 100 Cumulative percent failure 80 In addition, it is necessary to predict both the interest rate and the inflation rate in order to 60 calculate the discounted costs based on the future 40 value of money. The discounted cost rate, r, is calculated as follows; 20 0 (1 + interest rate) 0 50 100 150 200 Discounted cost rate, r = 1 (11) (1 + inflation rate) Time (years) The LCC calculations are, therefore, dependent on Figure 20 Distribution functions for service life the reliability of numerous assumptions, each of which is subject to a degree of uncertainty. Presenting the time to achieve the serviceability limit state as a cumulative distribution function Current LCC models tend to make simplifying enables the time increments to be determined at assumptions based on observations of which intervention is required. For a normal performance and engineering approximations of distribution, the rate of deterioration increases up to deterioration rates. The extent of damage likely to the time at which 50% has exceeded the defined have occurred after a specific period is, thus, limit state. As most structures will be designed estimated, e.g. after 20 years of exposure repairs such that the level of deterioration will be much will be required to (say) 5% of the surface. It is less than 50%, then this will apply to all properly also commonly assumed that repairs will be designed structures. The common assumption of required at fixed intervals, e.g. every 20 years. In a uniform deterioration rate in LCC calculations the absence of extensive performance data from must, therefore, be reconsidered. Using the structures or validated predictive models such example in Figure 20 and assuming intervention is assumptions cannot be avoided. The probabilistic required at time intervals representing 2% approach provides a means for quantifying the rate increments of failure, calculated periods to and extent of deterioration and hence a basis for intervention are as follows; estimating the cost of intervention. The output from the probabilistic model, i.e. the probability Percent failure 1% 2% 3% 4% curve (Figure 4), becomes the input to the LCC Years to intervention 40 48 53 56 model [2].Having calculated the capital cost and the While the first repairs are not required for 40 years, incremental costs of operation, repair, etc. the subsequent repairs, at the same magnitude, would costs are accumulated to produce a life cycle cost be required at reducing intervals. The significance curve and the total cost at the end of the design of this in relation to planned maintenance and life. repair is clear. Table 7 Calculated times to repair and the A Model For Life Cycle Costing associated costs based on intervention at 1% increments of (risk of) damage A model has been developed that uses the probabilistic approach to predict the rate of expenditure on repair by using as input data a REPAIR COSTS probability function to define the design life and its New repairs variability. Other input data are similar to those Constant costs yes required for existing LCC models as follows: Cost per square m £600 Current costs Disc. Financial data % Age Repaired per unit Average values of Interest Rate and Inflation Rate def. area area cost cost over the life of the structure. (Var) (fixed) 1.0 40 558 400 600 334950 51237 2.0 48 558 400 600 334615 34871 Capital costs 3.0 53 558 500 600 334615 27384 Calculated from the total volume of concrete and 4.0 56 558 500 600 334615 23088 reinforcement and its cost of placement (including 5.0 60 558 600 600 0 0 6.0 62 558 600 600 0 0 formwork). 7.0 64 558 600 600 0 0 8.0 66 558 600 600 0 0 Time dependent costs 9.0 69 558 600 600 0 0 The time dependent costs are broken down into 10.0 71 558 600 600 0 0 the following elements; 136581 Operating costs - these are the normal running costs of the building or structure and include planned maintenance. APPLICATIONS OF PERFORMANCE BASED Repair costs - these are the costs of completed DURABILITY DESIGN repairs. The time at which repairs are required are predicted using the probabilistic While there has been much discussion about analysis. performance based durability design, there has Lost production - these are the costs incurred as a been very little real application. Deemed-to-satisfy, result of the need to undertake repairs, for prescriptive specifications for durability are still example, the cost of decanting staff and used in almost every contract. Where a rental of alternative premises or the cost of a performance-based approach is used it generally road closure. involves a series of type tests prior to construction, Other costs - these are costs that are not covered or early in the contract period. Time is often short by any of the above. however, and difficulties often arise because of the need to demonstrate, in a very short timescale, that Costs relating to repairs are determined in time the concrete will perform satisfactorily for many increments that are derived from the probabilistic decades. This often leads the contractor into analysis of deterioration. From these input producing very low w/c concretes, which may be parameters the times are calculated at which difficult to place, when high performance in relation defined levels of intervention are needed and these to durability does not necessarily need high times are then fed into the cost model. If the first strength. Normal structural grades of concrete intervention is required when the risk level is 1% containing pfa or ggbs will offer as good a and the estimated time to reach this level is, say, resistance to chloride ingress and rebar corrosion 40 years then the model assumes that 1% of the as much higher grades of PC concrete. surface will be in need of repair after this time. An Unfortunately this does not show up in short term example from the spreadsheet is shown in Table 7. tests. In addition to predicting the rate of deterioration of Such an example was the Storebaelt West Bridge. the structure, a similar approach is use to predict This is a precast, prestressed (in part) reinforced the performance of the repairs which may concrete bridge comprising two independent post- themselves deteriorate within the design life. tensioned box girder platforms, carrying a twin- track high-speed railway and four-lane motorwayacross the Storebaelt. The girders span between develop a new air-entraining agent specifically 62 concrete caissons with spans of 82m and 110m. for this project. The design life is 120 years and considerable effort ii) Accommodating the delayed setting time, was made by the client, the Danish government, to resulting from the need to use very high develop a durable concrete though extensive admixture doses, within the slip formed parts of laboratory testing before the contract commenced. the structure. The focus was on resisting chloride induced reinforcement corrosion and preventing freeze- iii) Difficulty in finishing the deck slab surface due thaw damage. Having established a concrete that to the very low bleed resulting from the use of performed well under laboratory conditions, a microsilica. The surface tended to tear very largely prescriptive concrete specification was easily and the mechanical finisher could not be produced. It called for a three-component binder, used. comprising Portland cement, pfa and microsilica, as well as air entrainment. The latter was tested by The contractor had reasonably assumed that a measurement of the specific surface of the air void prescribed mix would meet all of the normal system in cores taken from the placed concrete. construction requirements and the difficulties in working with the concrete were recognised on site during the mix development stage. Changes were proposed for the use of a simpler, blended cement mix but this was not accepted by the client. The bridge was completed but not without difficulty. Other structures have used a similar approach of type testing followed by prescriptive requirements. The tunnel lining segments for the Channel tunnel were produced on this basis, with extensive chloride penetration testing being carried out in the mix development phase. Some contracts have adopted other types of testing. Occasionally water permeability, or water Fig 21 Precasting yard for the Storebaelt West absorption or gas diffusivity tests are adopted as Bridge part of a quality control regime. While the intent is to produce more durable concrete, and by simply A unique feature of the concrete specification, and having more rigorous QC the concrete is likely to one that created many problems, was the upper be produced to a higher quality, it is generally limit on water content of 140 litres/m³. This is the difficult to find the rational behind either the test only contract on which I have seen this rigorously method or the limiting values. Concrete Society applied. The water demand for the specified report 31 [35] gives typical values for a range of crushed rock, 20mm aggregate concrete is fluid transport properties of concrete and states normally about 190 litres/m³ in a mix with no very clearly that the values should not be used for admixtures. To achieve the 50 litre water reduction specification purposes. Nevertheless, and to achieve a reasonable level of workability for specifications still emerge requiring tests that do transportation by pump and for placing in often, not always relate to the deterioration mechanisms complex sections, a high dose of superplasticing that are likely to be predominant. But it is at least admixture was needed. This influenced the air void good to see someone occasionally thinking beyond system and the setting time. A complicated compressive strength. juggling act was needed therefore to work within the prescriptive limits and to meet the demands of Some codes are also acknowledging the need for a the construction process. Particular issues were as more rigorous approach to durability design. The follows; UK codes for maritime structures {BS6389] was recently revised, and while it still includes default i) Achieving the correct air void system in the prescriptive limits, it encourages the designer to pumped and compacted concrete. To achieve think more broadly about the actions on the the specified values within the structure, the structure, the condition limit states and the design fresh concrete had to be adjusted to life. This is, in itself, a significant change. accommodate the changes that occurred during concreting. This could not be achieved with any combination of commercially available admixture combinations and the supplier had toCONCLUSIONS 3. Bamforth, P.B. (1998) Double standards in design, Concrete Magazine, March Despite an enormous body of information from 4. Bamforth, P. B. and Pocock, D. C, (1990) laboratory and field studies, and structures that Minimising the risk of chloride induced corrosion by continue to deteriorate prematurely, little change is selection of concreting materials, Proc. 3rd Int. being made in the way in which the construction Symp. "Corrosion of Reinforcement in Concrete industry at large specifies concrete for durability. A Construction, SCI, Wishaw, May, pp. 119-131. few large and prestigious structures have attempted to use a performance-based approach, 5. Bamforth, P. B, Price, W. F. And Emerson, and while this has not always been successful, it M. (1997) An international review of chloride must be applauded nevertheless. But we are a ingress into structural concrete, Transport code led industry and until codes themselves move Research Laboratory, Transport Research away from prescriptive specifications change will Laboratory, Contractor Report 359, be slow. And codes will not change until there is 6. consensus among those with expertise in the area. Frey, R, Balogh, T And Balazs, (1994) Kinetic method to analyse chloride diffusion in various concretes. Cement and Concrete As an interim measure, if a prescriptive approach is to be used, then more robust limits must be used. Research, Vol. 24, No. 5, pp 863-873. This does not mean lower and lower w/c ratios or 7. Bamforth, P. B. (1993) Concrete higher and higher covers. We can build in classification for r.c. structures exposed to marine chemical robustness through the use of and other salt laden environments, Structural appropriate cementing materials (pfa, ggbs, ms) or Faults and Repair 93, Vol II, Edinburgh, June/July, employ supplementary measures if we understand, pp. 31-40. and can prove their performance. 8. Taywood Engineering, (1998) Guide for The model proposed herein provides a tool that prevention of corrosion in reinforced concrete may be used in the design process to enable exposed to salt, Part 1; Performance of R.C Blocks materials and systems to be compared and Exposed for 10 Years in Marine Splash Zone selected to meet the needs of specific structures. Conditions, TE Report no 1304/98/10248, DETR As with all tools, the more it is used, the more skill Partners in Technology Programme Contract CI that is developed by the user. And the more 39/3/231. feedback that is obtained from structures, the more 9. Nilsson, L-O, Poulsen, E, Sandberg, P, the model can be refined to reflect real Sorensen, H. E and Klinghoffer, O. (1996) HETEK performance. Chloride penetration into concrete, State-of-the Art, Transport processes, corrosion initiation, test methods and prediction models, The Road ACKNOWLEDGEMENTS Directorate (Denmark), Report No. 53, Copenhagen The author wishes to thank the Directors of Taylor 10. Bamforth, P. B., (1997) The derivation of Woodrow Construction Ltd for permission to input data for modelling chloride ingress from eight- publish this paper. Partial funding from the EC and year UK coastal exposure trials, Magazine of the (then) UK Department of the Environment Concrete Research, 1999, Vol 51, No2, April, under the Partners in Technology Programme is pp87-96 also gratefully acknowledged. The author also thanks the Concrete Society of New Zealand for 11. Bamforth, P. B., (1995) A new approach to the invitation to present this paper. the analysis of time-dependent changes in chloride profiles to determine effective diffusion coefficients for use in modelling of chloride ingress, RILEM REFERENCES International Workshop on Chloride Penetration into Concrete, St Remy-les-Chevreuse, France, 1. Building Research Establishment, (1996), Oct 1995, Paper 21, pp. 195-205. Sulfate and acid resistance of concrete in the 12. Senbetta, E And Scholar, C F. (1985) A ground, BRE Digest 363, BRE, Garston, UK. new approach for testing concrete curing efficiency. ACI Journal, Jan-Feb, pp 82-86. 2. Bamforth, P.B. (1998) A new approach to 13. Bamforth, P. B. and Pocock, D. C, (2000), durability design using risk analysis, Australian Design for durability of reinforced concrete Concrete Institute Annual Technology Seminar - exposed to chlorides, Workshop on Structures with Durable Concrete for the next Millennium, (invited Service Life of 100 years - or more, Gulf Hotel - paper), Perth, Australia, November. Bahrain, 19th November 2000 14. Bamforth, P. B., (1996) Definition of exposure classes and concrete mix requirementsfor chloride-contaminated environments, SCI chloride environments, Durability of Building Conference on Corrosion of Reinforcement in Materials, Vol 3, Elsevier Scientific Publishing Co., Concrete Construction, Cambridge, pp. 176-190. Amsterdam. 15. Bamforth, P. B., (1996) Predicting the risk 25. Vassie, P R., (1984) Reinforcement of reinforcement corrosion in marine structures, O corrosion and the durability of concrete bridges. E Gjorv Symposium 'Concrete for Marine Proc. Instn. Civ. Engrs, Part 1, 76, pp 713-723. Structures, 3rd CANMET/ACI International 26. Conference on Performance of Concrete in Marine Sandberg, P., Pettersson, K., Sorensen, H. Environment, New Brunswick, pp. 207-233 E., and Arup, H. (1995) Critical chloride concentrations for the onset of active reinforcement 16. Bamforth, P B and Chapman-Andrews, J, corrosion, RILEM International Workshop on (1994) Long-term performance of r.c. elements Chloride Penetration into Concrete, St- Remy-les- under UK coastal exposure conditions, Corrosion Chevreuse, France, pp. 453-459. and Corrosion Protection of Steel in Concrete, 27. Proc. of Int. Conf., Sheffield University, July 1994, Thomas, M. D. A., (1996) Chloride thresholds in marine concrete, Cement and Ed R. N. Swamy, Sheffield Academic Press, Vol 1, pp 139-156. Concrete Research, Vol. 26, No. 4, April, pp. 513- 519. 17. Gjorv, O, Tan, K. and Zhang, M-H. (1994) 28. Chloride diffusivity of high strength lightweight Andrade, C. and Alonso, C. (1996) concrete, ACI Materials Journal, Sept./Oct. 1994, Corrosion rate monitoring in the laboratory and on pp 447-452. site, Construction and Building Materials, Vol. 10, No. 5, 1996, pp. 315-328. 18. Maage, M. And Helland, S. (1993) Chloride 29. penetration in concrete structures exposed to Diez, J M. (1997) Duracrete Task 2 - marine environments, Nordic Miniseminar on Modelling of degradation, Sub-task 2.2.3 Modelling Chloride Penetration into Concrete Structures. of the structural consequences of corrosion, Chalmers University, Goteberg, Sweden, January Geocisa Report, April 1997 1993, pp 125-137. 30. Bamforth, P B, (1999) Options for 19. Mange, M, Holland, S & Curliness, J E, reducing the risk of reinforcement corrosion in r.c (1998) Chloride penetration into concrete with structures and their incorporation into the design lightweight aggregate aggregates, DP 3.6, LITTON process, Anti-Corrosion Materials and Methods, Konstrukjonbetong, Project 3, Report 3.6, Draft for Vol 46, No 4, 1999, pp268-275 comment, May 1998. 31. Polder, R B, (1996) Laboratory testing of 20. Gautefall, O. (1993) Experience from nine five concrete types for durability in marine years exposure of concrete in the tidal/splash zone, environment, 4th Int. Symp. on Corrosion of Nordic Miniseminar on Chloride Penetration into Reinforcement in Concrete Construction, (ed Page, Concrete Structures. Chalmers University, Bamforth, Figg) Cambridge, UK, July 1996, pp 115-123. Goteberg, Sweden, January 1993, pp 148-158 21. Berke, N. S, (1989) Resistance of 32. Maage, M, Helland, S & Carlsen, J E, microsilica concrete to steel corrosion, erosion and (1998) Chloride penetration into concrete with chemical attack, Third Int. Conf. on Fly Ash, Silica lightweight aggregate aggregates, DP 3.6, Fume, Slag and Natural Pozzolans in Concrete, LETTKON Konstrukjonbetong, Project 3, Report ACI Special Publication SP-114, 1989, Vol 2. pp 3.6, Draft for comment, May 1998. 861-886. 33. Siemes, A. J. M., Vrouwenvelder, A. C. W. 22. Hansen, G K, Rindal, D and Horrigmoe. M. and van den Beukel, (1985) Durability of (1993) Diffusion chlorides in high strength buildings: a reliability analysis, HERON, Vol. 30, concrete, High Strength Concrete 1993, No. 3, 1985. Lillehammer, Norway, 1993, pp 671-677. 34. Siemes A. J. M. and Rostam, S., (1996) 23. Glass, G K. and Beunfeld, N. R., (1995) Durability, safety and serviceability - A performance Chloride threshold levels for corrosion-induced based design, TNO report no 96-BT-R0437-001, deterioration of steel in concrete, RILEM Feb 1996, (presented at the IABSE Colloquium International Workshop on Chloride Penetration 'Basis of Design and Actions on Structures' Delft, into Concrete, France, Netherlands, March 27-29, 1996) pp. 429-440. 35. Concrete Society, (1988) Permeability 24. Browne, R D., (1982) Design prediction of testing of site concrete; A review of methods and the life of reinforced concrete in marine and other experience, Technical Report 31.