2. literature review and state of the artcib w080 / rilem 175 - slm: service life methodologies...

CIB W080 / RILEM 175 - SLM: Service Life Methodologies Engineering Design Methods: State of the Art 59

2. LITERATURE REVIEW AND STATE OF THE ART

2.1 General A literature review was performed, mainly concentrating on conference proceedings (see references for details). Reviewing the available literature, it was noted that, although titles of papers in earlier conferences indicated quite specific topics on service life prediction, their contents appear to be fairly general. Earlier papers mainly give outlines and point at areas of work to be done [Masters and Brandt 1989].

As a consequence, this state of the art review concentrates on publications as far back as about 1996. In the following text, the relevant and in general the most recent references to the topics dealt with are reviewed.

This report is limited to techniques for the prediction of service life. The term service life often appears within life cycle analyses (LCA). Service life can be part of an LCA, but an LCA is rather more comprehensive and comprises at least a calculation of all costs from cradle to grave inclusive of all investments over the entire lifetime and an assessment of the environmental impacts. LCAs as such are not dealt with in this report.

2.2 Need for Service Life Design In 1996 the need for service life design was identified and standardisation was suggested [Frohnsdorff 1996, Frohnsdorff and Martin 1996]. Nireki [1996] showed several approaches to solve the durability and service life issues respectively and identified needs for further research. In 1997 the need for a state of the art report to service life design was stated [Jernberg et al. 1997].

In studies to the cost benefit analysis with regard to road bridges, several authors give estimations of the functional service life of these bridges [De Brito and Branco 1998, Thoft-Christensen 1997].

Aikivuori in “Critical loss of performance – what fails before durability” [1999] points out, that service life limited by durability is seldom reached, as components are refurbished earlier due to other reasons:

Empirical research has been carried out to find out the actual reasons for initiation of repair projects on buildings. This research has shown that the owners of the buildings actually experienced the user requirements predominantly outside of the range of durability failures. Only 17 % of the repair projects were initiated primarily because of deterioration. The critical loss of performance seems to primarily be in the range of a subjective perception of the building. Very little technical or economical rationality can be seen in the actual decisions made on building refurbishment. In most cases the limiting factor for service life is not durability.

EMPA – Swiss Federal Laboratories for Materials Testing and Research Laboratory for Concrete and Construction Chemistry March 2004

CIB W080 / RILEM 175 - SLM: Service Life Methodologies Engineering Design Methods: State of the Art 60 Life cycle economics of the buildings have (according to the empirical findings) been evaluated on a technically/economically irrational basis. Decision-makers pay little attention to the condition and remaining potential for service life of the building components. They pay very limited attention to economical expectations. Optimisation of economical factors of the buildings is the primary goal in less than 10 % of repair projects.

If service life is seen as the actual time in service of the building components, the basis of service life prediction models should not be based on durability or economics of the building components only. Durability is of course the limiting factor for service life in the sense that service life can not exceed the limitations set by durability, but in fact the actual service life seldom reaches the full potential life time of the components limited by durability. The forecasting of the refurbishment requirements should therefore not rely on the durability-based concepts only. Asset and maintenance management should pay more attention to the more critical perception of the perceived quality of the buildings.

The problem of service life design has attracted more and more attention, mainly due to pressure from owners requiring such a design, supported by the CPD [1988] and EOTA requiring the topic to be addressed properly (see [Sjöström et al. [2002]).

2.3 End of Service Life

2.3.1 General definitions All design methods require clear definitions of the end of the service life. This is however not a universal and easily defined value. In general terms it is the point in time, when the foreseen function is no longer fulfilled. The properties of a building part can be split up into several sub-properties, e.g.

Safety: The integrity of the building part is maintained at the standard level of safety, • •

•

Function: The required function is fulfilled, (i.e. deflections are still within limits, a window can easily be opened and closed, etc.), Appearance: The expected appearance is given (i.e. the surface of the building part is of acceptable appearance, the glazing of windows has not deteriorated or turned opaque, etc.).

2.3.2 Definitions in structural engineering In structural engineering, depending on the function of the part of a building or structure, engineering criteria for the end of the service life are often used, in order to permit a meaningful calculation of the service life as such.

Typical engineering criteria are:

• A minimum concrete cover for a given environment. This definition represents standard practice based on experience, but is not based on a specific clearly defined service life.

• The arrival of the carbonisation front to the outside face of the reinforcing steel is considered to be the end of service life,

• The arrival of the front of the chloride ingress to the outside face of the reinforcing steel is considered to be the end of service life. These two definitions are rather conservative.



• The onset of spaulling is considered to be the end of service life. This definition assumes, that either spaulling is impairing the appearance beyond acceptance, or that shortly after the onset of spaulling, the required level of safety is not maintained any more due to corrosion of the reinforcement now lacking the protective concrete cover.

2.3.3 Design code and legal definitions The Swiss code of practice code SIA 160 [1989] requires that both the safety limit state and serviceability limit state be checked. As soon as properties of a structure are time dependent, this requirement would actually enforce a service life calculation.

The European Construction Products Directive [CPD 1993] and the Swiss Bauproduktegesetz (National law covering building products, 2001) both address service life as the period of time during which the essential requirements have to be fulfilled.

2.3.4 Project definitions The designer has to set up the relevant criteria for the building or structure considered in consultation with the owner, depending on the requirements of safety, function and appearance. For this purpose Quilling et al. [2002] propose a framework for service life design which concentrates on a holistic approach and defines the necessary steps throughout the design process in order to ensure the data exchange between the parties involved: BRE (Building Research Institute) are currently developing a generic service life design system which will provide specific guidance for concrete structures. This paper describes the work that has been carried out to date in the development of the overall framework of the design system. This framework will be expanded as the project progresses to provide targeted guidance and tools to assist the practising engineer in achieving a structure that is durable for its required service life and in optimising whole life costs.

2.4 Factor Method The factor method and it’s developments are covered comprehensively in the respective part of this state of the art report: Factor methods for service life prediction: A state of the art. A summary can be found in Hovde [2002], a recent application in Abu-Tair et al. [2002].

The factorial method according to ISO/CD 15686 identifies the main factors of influence with regard to service life and there from, a plain figure for the service life of the building or building component can be calculated. Knowing the main factors of influence and the overall behaviour of a component facilitates the understanding of the relevant issues, but does not reflect reality very closely.

Examples on the factor method can be found in many publications, e.g. Strand [1999]. The shortcomings of the method are discussed in some of them. These main shortcomings can be summarised as follows:

• • • •

The plain multiplication of factors, which in reality might have a different weight, The result being a single figure instead of a result to reflect variance of reality, The data still to be accumulated, The lack of a direct relation to data gathered e.g. on environment, climate, installation quality, in use conditions, etc. The factors are usually set basing directly on the behaviour of the component in a given set of conditions, rather than basing on the



influence of individual parameters such as regimes of rainfall, temperature, wetting time, type of use, etc.

Considering the efforts in gathering input data, the simple figure result of the factor method, when executed as set out in ISO 15686, seems not to be adequate.

2.5 General Aspects of the Probabilistic Methods Examples of service life predictions using probabilistic tools can be found in numerous publications. Most deal with a single material or exposition. The main fields of application are the service life of reinforced concrete, the service life of pavements (streets or airports), see McNerney et al. [1997], Flintsch et al. [1997] and the service life of (wooden) building envelopes such as windows, wall claddings and roofs.

Apparently concrete is a dominant material as far as durability under severe conditions is concerned. Examples of durability in view of chloride ingress into concrete are quite common. Most studies deal with steady states as far as exposition is concerned. Some of the authors however, have considered transient (humidity, wetting, drying, etc., see e.g. Vu and Stewart [2002]).

In many cases, data has been collected and variables fitted to them. Open databases to draw upon seem not to be available yet. Those methods being applied in projects that have paying clients may result in the preparation of reports that are not freely disseminated.

2.6 Application of Probabilistic Prediction Methods Degradation is generally regarded as a stochastic process and the main parameters are in most cases known. Variations or secondary parameters on the other hand are often not explicitly and numerically taken into account, but their influence results in a considerable scatter of the behaviour of the structure.

2.6.1 Markov model for the deterioration The Markov model assumes deterioration to be a stochastic process governed by random variables. The structure may be split into a number of components, which deteriorate randomly. The main parameters of the deterioration are established for each component, together with the deterioration variables versus time. In Figure B2.1 deterioration using seven stages for condition rating is shown.

Figure B2.1: Markov deterioration function Research projects and large engineering projects often rely on models like the Markov model:


CIB W080 / RILEM 175 - SLM: Service Life Methodologies Engineering Design Methods: State of the Art 63 • Abraham and Wirahadikusumh [1999] in “Development of prediction models for sewer

deterioration” treat service life of sewer lines: Due to their low visibility, rehabilitation of sanitary sewers is often neglected until catastrophic failures occur. Neglecting regular maintenance of these underground utilities adds to life-cycle costs and liabilities, and in extreme cases, stoppage or reduction of vital services. Incorporating condition data and deterioration patterns of the city's sewer system is pivotal for obtaining a realistic assessment of the city's infrastructure. The paper explores the probability-based Markovian approach for modelling deterioration. This approach is based on the assumption that since the behaviour of sewer lines (i.e. the rate of deterioration) is uncertain, the selection of an appropriate repair strategy is also an uncertain procedure. Probability-based prediction models enable the comparison of the expected proportions in given condition states with the actual proportions observed in the field, and in this way possible defects in construction, materials, quality control, etc., can be identified. Expert opinions from engineers, who have developed the sewer assessment surveys for the City of Indianapolis, Department of Capital Asset Management (DCAM), are used for validating the deterioration models developed in the research. More realistic deterioration models will assist asset managers in improved performance modelling of the sewer infrastructure and also in determining this infrastructure's rehabilitation costs based on improved estimates of deterioration.

• Leira et al. [1999] in “Degradation analysis by statistical methods” treat various methods: Several utilities experience a great future challenge due to deterioration of properties, this being both buildings and infrastructure: As service life ends, there will be an increasing need for rehabilitation (i.e. renewal and maintenance). Most maintenance decisions up to now have been based on the so-called fire brigade strategy, i.e. to make spot repairs after the failure has occurred, or based on rules of thumb. In order to enhance maintenance and rehabilitation decision-making, it is essential to improve our understandings about the deterioration processes. A set of tools should be developed for decision support. These should be based on, or take into account, existing knowledge of failures. Statistical methods can be regarded as a way to organise this knowledge. This paper describes how statistical methods can be applied for forecasting rehabilitation needs. Examples of trend plots, survival methods, condition class transition and stochastic model parameter analyses from concrete structures, roads and water networks are shown. It should be emphasised that there are similarities in the way various construction outputs can be analysed.

• Ansell et al. [2002] report a Markov approach in estimating the service life of bridge elements in Sweden: The service life of Swedish road bridges has previously been studied by collecting inspection reports and other significant information from 353 bridges. A total of 3747 bridge inspection remarks were gathered and the type and cause of damage were stated and each element was given a condition class. This information was then inputted into a relational database. ... Deterioration of bridge elements can be analysed numerically using the Markov chain theory. The deterioration of a particular structural member must be defined by a number of states, in this case given by the assessed condition classes. A state vector that gives rise to a new state after multiplication by a transition probability matrix defines the states of a population of elements. It is demonstrated how a transition probability matrix can be numerically determined to describe the deterioration process of a bridge element from data in the relational database. The numerical method used, is based on an iterative stepwise combination of the matrix elements until the error between a known deterioration average



curve and a curve given by the Markov chain is minimised. The method is computationally demanding for small steps but will, with larger steps, quickly converge towards an approximation close to the transition probability matrix. It is also demonstrated how the remaining service life of a bridge element can be estimated by studying the variation of the state vector with bridge age.

• Kaempfer et al. [2002] have applied a somewhat simplified deterioration model to sewer lines: The condition of sewer pipes and pipe joints are evaluated according to the scale and the effects of damage. The determined damages are assigned to one of five different damage classes. The damage classes range from very serious to negligible. In a second stage, the status of sewer sections is evaluated according to the greatest damage. These evaluation data are installed in a sewer database according to belonging functionality and stability variables such as significance of sewer section, hydraulic capacity, overflow frequency, material, construction year, geometry, size of covering and traffic load situation. In a third stage the correlation is graphically described between the network sections and the year of construction and different functionality and stability variables. The aging curves were derived from the available inspection data and the construction year for each status class (see Figure B2.2). The average residual service life of the sewer section is represented by a vertical line between the real age of the sewer section and the point of intersection with the aging curve of intervention status class. The different intersections on the horizontal line with the aging curves of different status classes indicate the ages at which the section is likely to drop to the next class or, going back in time, came from the previous class. The example of a small town shows how the acquisition of data and the evaluation of damaged sewers in a municipality is carried out and illustrates which priorities have to be established during the maintenance of sewer networks. The model city of 8,000 inhabitants is situated in the middle of Germany. The sewer network comprises around 25 kilometres with 700 individual sewer reaches. In 1998 the total sewer system was optically inspected. The results of the inspection serve as the basis for a database.



Figure B2.2: Status transition functions for concrete and stoneware sewers in Stadtilm, Germany

2.6.2 Variables defined as distributions The probabilistic methods often quantify uncertainties in the form of density distributions. Some examples are looked at in this section, as this concept could prove to be worth wile for application in the engineering design methods. For the assessment of service life using formulas with several variables, distributions can be used instead of plain values.

• Enright and Frangopol [1998] studied the deterioration of highway bridges using time-variant series reliability approach where both load and resistance are time dependent. The minimum level of safety governs the end of service life. The purpose of the analysis lies in the development of a reliability-based maintenance strategy: Experience has demonstrated that highway bridges are vulnerable to damage from



environmental attack, such as alkali-silica reaction, corrosion, and freeze-thaw. To make rational decisions in a life cycle cost perspective, reliable prediction of the service-life of deteriorating highway bridges is necessary. To obtain an accurate insight into this problem, time-variant reliability methods have to be used. The application of these methods in the performance and safety assessment of deteriorating structures is relatively new. In this study, the reliability of reinforced concrete highway girder bridges under aggressive conditions is investigated using a time-variant series system reliability approach in which both load and resistance are time-dependent. Monte Carlo simulation is used to find the cumulative-time system failure probability. An existing reinforced concrete T-beam bridge located near Pueblo, Colorado is investigated. The effects of various parameters such as variability in dead and live loads, live load occurrence rate, strength loss rate, degradation initiation time, resistance correlation, and number of girders under attack on the time-variant bridge reliability are studied. The results can be used to better predict the service life of deteriorating reinforced concrete bridges, and to develop optimal lifetime reliability-based maintenance strategies for these bridges.

• Lounis et al. [1998] in “Further steps towards a quantitative approach to durability design” presents further steps in the development of reliability-based approaches for the durability design and service life prediction of building components which integrate the requirements of safety, serviceability and durability: In General, the load and resistance should be modelled as stochastic processes and the resulting durability problem is formulated in a time-dependent probabilistic format. Using the classical reliability approach, the resulting time-dependent reliability problem is transformed into a time-independent reliability problem through the adoption of an extreme-value probability distribution for the maximum lifetime load. The resistance degradation and its variability are included in the model, and the probabilistic design problem is transformed into a deterministic (or semi-probabilistic) problem using the first-order second moment theory. This semi-probabilistic integrated approach to durability design and prediction overcomes the shortcomings of the empirical factorial approach and the complexities of a fully time-dependent probabilistic method. An alternative approach using stochastic process theory is proposed to formulate the durability design problem as a crossing problem for which the probability of failure within the component lifetime is obtained from the first-passage probability for the stochastic process. In addition, a service life-based formulation of the durability design and prediction problems is presented in order to illustrate its equivalence with the perfor-mance-based formulation. It is shown that in principle the same probabilistic approaches used for the development of structural design approaches for safety and serviceability are also applicable for durability design. The durability design objective is to keep the probability of failure within a specified time interval (or service life) below a certain threshold value that depends on the consequences of failure of the component or system. It is expected that in the near future, further simplifications of the proposed approaches will be made leading to practical and reliability-based methods to durability design or service life prediction. These simplified methods will be implemented in the design of durable new structures and optimal life-cycle maintenance management of existing structures. The Markov model considers steadily degrading systems, where for each property, during each time period, a probability of deterioration is defined. This method thus requires fairly sophisticated inputs in the form of probabilities, which are not easily estimated, as they cannot be read directly off the real behaviour of the structure in the field. The Markov model requires an in depth knowledge of the system dealt with or on the other hand has to rely on significant simplifications.


CIB W080 / RILEM 175 - SLM: Service Life Methodologies Engineering Design Methods: State of the Art 67 • Fagerlund [1999] treats frost attack using this variant of probabilistic approach:

The main parameters: saturation, critical and effective, frost, are introduced as distributions. Significant frost damage will not occur until a certain critical moisture level is transgressed over a sufficiently big portion of the structure. The critical moisture level is a “fracture value” which can be compared with the load carrying capacity in structural design. It is a materials property that seems to be rather uninfluenced by normal variations in environmental properties, such as number of freeze-thaw cycles and minimum freezing temperature. The moisture content inside the structure depends on the outer moisture conditions; the more moist the environment, the larger the inner moisture content, and the larger the risk of frost damage. The actual moisture content in the structure can be compared with the actual load in structural design. The risk of frost damage can be calculated when the frequency functions of the two parameters, critical moisture content and actual moisture content are known (see Figure B2.3). Some hypothetical cases are treated in the paper showing that the probability of frost damage might actually decrease with increasing exposure time in moderately moist environments, but that it normally increases with increasing exposure time in continuously moist environments where the structure has no possibility to dry.

Figure B2.3: Distributions of critical and effective saturation in concrete versus time

• Flourentzou [1999] uses four typical degradation schemes to quantify the behaviour of an element. The choice of the respective degradation curve or combination thereof is somewhat theoretical: The service life of buildings is an important factor e.g. in life cycle assessment and the assessment of global costs. Based on experience much information is available regarding the service life of building elements. However, for existing buildings such information is of little use as the key question is the probable date of repair/replacement. MEDIC (“Prediction Method of probable Deterioration Scenarios and Refurbishment Investment Budgets”) is developed on the theories of conditional probabilities to help assess the residual service life and thereby the necessary investments in refurbishment.



When passing from working on general products, like the life span of wooden windows, to specific objects, for example 29-year old wooden windows, the current condition of the object must be taken into account. An evaluation of residual service life must for that reason be closely connected to a good diagnosis method. In the European project EPIQR (“Energy Performance Indoor Environment Quality Retrofit”, see also Brandt et al. [1999]) the deterioration of building materials and components is described by the use of a classification system with four classes for the qualitative condition (e.g. of a facade or window, see Figure B2.4). MEDIC calculates the probability to change from one class to another with time. The prediction is based on the combination of the a priori probability based on experience from a large number of previous investigations/refurbishments and the current state of the object under study.

Figure B2.4: Four representative resultant degradation curves

• Dotreppe [1999] uses a two stage degradation scheme for modelling the behaviour of reinforced concrete bridge decks as shown in Figure B2.5:

1. Initiation and then 2. Propagation (depassivation) Composite steel-concrete constructions are presently widely used, and certainly in the field of composite bridges where they appear quite competitive. Most of the structural problems related to this type of construction are presently solved, concerning particularly the design of the steel girders. Nowadays particularly in the northern temperate zone where de-icing salts and freeze thaw are a problem studies are focused on the durability of the concrete slab, which controls the performance of these bridges. The model commonly accepted for the description of the corrosion process in a reinforced concrete element is presented. The evolution regarding the problem of the influence of the crack width on the durability of concrete is discussed. The factors leading to cracking of the concrete slab are examined, with special attention to thermal and autogenous shrinkage involving early cracking, and the results of a practical example are presented. The most essential requirements regarding durability are mentioned. Concerning reinforcement of the slab the classical solution consists in using standard reinforcing steel. However, as the slab is cracked, durability will be controlled by the corrosion development, which leads to uncertainty regarding service life. Prestressing can ensure a satisfactory performance during a sufficiently long period. Several parameters have to be assessed carefully, such as the type of prestressing and the amount of prestress to be introduced in the slab.



Figure B2.5: Two stages of the degradation scheme used by Dotreppe

• Siemes [1999] presents a probabilistic method of predicting the behaviour of concrete structures. He defines four stages relevant to service life:

1. Depassivation, 2. Cracking, 3. Spaulling and 4. Collapse.

For chloride-induced corrosion, an equation giving the required concrete cover as a function of the chloride concentration has been derived for a chosen time of service. The variables are introduced as distributions: Due to the high construction costs and the social importance the durability demands for large infrastructures is becoming more and more important. Service life requirements of 100 years or even more are quite common. For the bored reinforced concrete tunnel under the Western Scheldt in the Netherlands the requirement was a service life of at least 100 years. No method had been specified to prove this service life. Since the concrete codes are only based on deem-to-satisfy rules for the durability, without any specification for the service life, it was not possible to base the design on existing codes. The service life design has been made on the basis of the methodology that has been developed in a research project of the European Community. This project with the name ‘DuraCrete’ has further improved the existing reliability and performance based structural design method by introducing the modelling of degradations and environmental actions. It is believed that the service life design of the Western Scheldt Tunnel is the first project were the DuraCrete approach has been applied in practice.

• In Siemes [2002] a further refined five stage model is shown for concrete from its initial state to collapse, see Figure B2.6:



Figure B2.6: Adverse events during corrosion of the reinforcement to collapse

• Lair et al. [1999] and [2001] predict the service life using two approaches. On the one hand, they perform a Failure Mode and Effects Analyses (FMEA). This method allows the identification of the failure modes, i.e. the failure to fulfil one of the functions for which the building part was designed. On the other hand, they collect service life information form all available sources (expert opinion, statistical studies, modelling, artificial and natural ageing, etc.), assess their quality, and, by means of a data fusion procedure, give a probability of failure, together with optimistic and pessimistic values of this probability (upper and lower bounds). These two approaches give a band of service life as shown in Figure B2.7.

In the past decades much effort has been put into the improvement of the durability of concrete structures. This has resulted in a reasonable understanding of the main degradation processes or experience with measures to prevent degradation. The results of this effort can be found in the present concrete codes and in manuals on durability design. The design rules are in general presented as deem-to-satisfy rules. If the rules are followed it may be assumed that the structure is durable. The present approach does not give direct insight into the service life, the necessary maintenance or the probability of premature failure.

Further it is clear that a lack of durability can have an influence on the structural behaviour. The direct relationship between durability and safety and serviceability of concrete structures has however not been made in the concrete codes. In the Brite-EuRam project ‘DuraCrete’ the durability design has been developed into a service life design based on performance and on reliability for reinforced concrete structures. This offers the possibility to present the design on the same level as the structural design, also based on performances and reliability. The structural and service life design can even be integrated. The ‘DuraCrete’ approach can be modified for the service life design of other structural materials and building materials.



Figure B2.7: Range of service life defined by the two approaches used by Lair et al.

Lair et al. [1999]: Assessing the service life of building products is relevant for all building actors (insurers, manufacturers, building owners and architects). Indeed, the knowledge of building product service lives leads to a reduction of maintenance costs and environmental impact, and an improvement of safety. This paper deals with a methodological approach for durability assessment. The major steps are:

• Research of available durability data and their organisation in a graph structure followed by the assessment of belief and plausibility distribution of service life.

• A Failure Mode and Effects Analysis, including a structural and a functional analysis in order to search all potential failures (weathering factors, product design and setting up).

The proposed method is a multi-model and multi-scale approach; multi-model in order to adjust the model with our knowledge and our aim (modelling real life of building, but not a too complex and unusable model), multi-scale to take into account the links between the three geometric scales materials/products/building. Finally, it gives

(1) Distribution of nominal service life, for normal weathering processes, with corresponding belief and plausibility degrees,

(2) Details on the design and setting up problems, on exceptional weathering phenomena, which could lead to a shorter service life.

• Faber and Gehlen [2002] describe the probabilistic concept for the assessment of the durability of existing reinforced concrete structures with special emphasis on the spatial variability of the parameter dominating deterioration. They use the fault tree and decision tree concept and four levels of damage (see also [Siemes 1999]). The method is illustrated on the problem of chloride diffusion where even the chloride concentration on the surface is treated probabilistic.

2.6.3 Practical examples of probabilistic methods From literature, a few examples are extracted and commented upon in the following paragraphs, showing the main lines of application.


CIB W080 / RILEM 175 - SLM: Service Life Methodologies Engineering Design Methods: State of the Art 72 Selected examples In practical applications of the probabilistic methods, the main parameters have to be identified and often, the theoretical models are modified and simplified, most often reducing the parameters included in the model to a minimum:

• Breitenbüchner et al. [1999] present the design for service life of the Western Scheldt tunnel, the governing factor being the concrete cover. Assuming the chloride attack being the main parameter for degradation the required cover was derived at. Inputs for this calculation such as chloride concentrations, diffusion factors, etc. were defined as stochastic variables (density distributions, see Table B2.1). As a limiting property the reliability index was chosen, for onset of corrosion 1.5-1.8 up, for onset of spalling 2.0-3.0 up to collapse 3.6-3.8. : Due to the high construction costs and the social importance the durability demands for large infrastructure are becoming more and more important. Service life requirements of 100 year or even more are usual. For the bored reinforced concrete tunnel under the Western Scheldt in the Netherlands the requirement was a service life of at least 100 years. No method has been specified to prove this service life. Since the concrete codes are only based on deem-to-satisfy rules for the durability, without any specification to the service life, it was not possible to base the design on existing codes. The service life design has been made on the basis of the methodology that has been developed in a research project for the European Community. This project with the name ‘DuraCrete’ has further improved the existing reliability and performance based structural design method by introducing the modelling of degradations and environmental actions. It is believed that the service life design of the Western Scheldt Tunnel is the first project were the DuraCrete approach has been applied in practice.

Table B2.1: Input distributions used in the design of Scheldt Tunnel by [Breitenbüchner et al. 1999]


CIB W080 / RILEM 175 - SLM: Service Life Methodologies Engineering Design Methods: State of the Art 73 • Helland [1999] shows how the remaining service life of existing concrete structures was

assessed based on the chloride ingress and taking into account a decreasing chloride diffusion over time (see Figure B2.8): After reviewing the present technology on performance-related durability criteria, CEN TC-104 concluded for the coming EN 206 and ENV (execution standard) that none of these were sufficiently mature to be incorporated in a technical standard. For this reason, all the clauses in these two standards will be of the traditional ”deemed to satisfy” type. A technical standard shall not deal with responsibilities. However, they will normally be the main references for contractual agreements. Hence the fulfilment of the technical requirements will heavily influence the producers’ liability. The present ”deemed to satisfy” requirements define the quality of the product purchased by the owner. Future performance-related requirements have to be suited both for the producer and the client as a reference to split the risk of liability for possible future shortcomings of the structure. This split must be acceptable for both parties, and the final evaluation of conformity should be concluded as early as possible after the construction. The paper describes a scenario where the evaluation of conformity might be done within a fairly short period concerning a structure’s ability to withstand the ingress of chlorides in its specific environment.

Figure B2.8: Service life solutions using different materials

• Wisemann [1999] also gives an example of structural service life prediction using distributions for the parameters to assess durability of designs using different materials: Parking garage configurations in Canada present one of the most adverse climates for reinforced concrete. The historic excessive use of de-icing salts on North American roadways has, by exposing the structural elements to a saline solution at a heightened ambient temperature, enhanced the rate of deterioration in parking structures. Various rehabilitation options may be considered for different circumstances, from simple stripping and repair of affected areas, to the chloride extraction or re-alkalinisation of structural elements, and ultimately to demolition and reconstruction. The emergence of "innovative" materials and methods promising extended or altered in-service performance has left designers without a clear view of the relative benefits of the more traditional approaches. This paper examines the degradation models available for service life prediction of



parking structures, in particular methods to evaluate systems using innovative materials (for results of service life prediction see Figure B2.8). This paper also presents the projected service life extensions provided by design, rehabilitation and material options and evaluates the life-cycle economics of their application to a specific parking garage scenario. The assessment considers the total life-cycle costs, including capital, operations and maintenance costs, re-capitalisation as well as the projected rehabilitation costs for each scenario.

• Teply [1999] deals with practically all variables influencing the service life of a reinforced concrete beam (variations in beam properties and degradation (corrosion) due to carbonation and chloride attack): For a specific RC beam, the depassivation of reinforcement is assessed and a consequent corrosion process in time is evaluated using numerical models. The cross-section of the beam is analysed using the “layer approach”. The efficient statistical modelling of the carbonation process and the consequent corrosion of reinforcement is updated utilising the in-situ measurements. The influence of deterioration on failure probability is assessed and sensitivity analysis is performed.

• Hong [2000] considers the degradation of reinforced concrete structures due to both aggressive environment and in-service loading for service life prediction: Aggressive environment and in-service loading can cause the resistance degradation of existing reinforced concrete structures. They must be considered for the safety evaluation and service-life prediction of deteriorating structures. This paper presents an integrated approach for time-dependent system reliability analysis considering the stochastic nature of load processes, uncertainties in strength, degradation initiation time, and strength degradation mode. The approach takes into account the partial correlation among the failures of structural elements, which arises from the fact that elements of a structural system are subjected to the same load processes and/or depend on some common basic random variables. The approach is efficient, since it does not require simulation. Analysis results obtained by using the proposed approach indicate that the reliability of series systems is relatively insensitive to the correlation between the failure of structural elements. However, the reliability of parallel system is highly sensitive to the correlation. The use of the time-dependent reliability obtained by neglecting possible correlation as a safety measure, therefore, should be avoided.

Further applications The topic of service life of concrete structures namely depending on the ingress of chlorides is widely covered in the papers of 8DBMC and 9 DBMC (see references).

2.7 Developments of the Factorial Method Towards Probabilistic Methods The momentum for of developing more realistic models based on the factorial method is gained on the one hand from the ease of understanding and on the other hand from the need for a more satisfying result of the service life prediction.


• From this background Moser [1999] uses the definitions of the factorial method, but employs variables with density-functions instead of plain figures. The variables are based on data given by the manufacturer, by tests, by experience, by expert opinion, and others. Reliable data from expert opinion can be derived by application of the so-called recursive Delphi method. Experts are required to estimate the minimum (say 5%), the average (50%) and the maximum (say 95%) fractals of the variable considered. These estimates are fitted into density distributions of any kind such as: standard, symmetric, asymmetric, custom-defined, or others (or even deterministic, here for design level) see Figure B2.9.


Figure B2.9: Input variables defined by various types of distributions

The result of the computation using existing simple software such as VaP [1996] is a density function of the expected service life of a similar lot of elements (see Figure B2.10). These resulting distributions indicate for e.g. the average or mean service life, but fractiles such as the first 5% or 70% of the elements at the end of their service life can also be read off the distribution curve at once.

These functions can be processed through to for e.g. a replacement cost versus time function of a building or other works. The experts’ estimates are reconsidered after calculation of the resulting service lives or replacement demand. The necessary fine-tuning based on the experts’ experience leads to a realistic model and powerful engineering tool.

This method is proposed for application in the current draft of ISO 15686-4 – Service Life prediction data requirements.



Fig. 5: Density distributions for the windows in the four faces

Figure B2.10: Resultant service life distributions for four sides of a building.

• Aarseth and Hovde [1999] discuss a similar principle in broader terms based on inputs in meetings in Gävle 1998. The distribution function is restricted to the Erlang-distribution and based on estimates of the 1% and 99% fractals, see Figure B2.11. (These fractals have according to experience of the author proved to be rather difficult to assess for experts.) Furthermore, the equation in ISO 15686 is modified from a multiplication of factors to a summation of delta years starting at the reference service life. It is doubtful, whether this principle adequately covers the real variation of the individual factors:

Figure B2.11: Erlang function showing estimated values

The ISO/CD 15686-1 “Service life planning” describes a deterministic method that allows an estimate of the service life to be made for a particular component or assembly in specific conditions. In “real life” the service life has a big scatter and should be treated as a stochastic quantity. In this paper we introduce the “step-by-step” principle as a stochastic approach to the ISO factor method. The “step-by-step” principle provides a more systematic approach to the estimating process and makes possible a stochastic handling of the factors. For each factor three estimates shall be made, the minimum, maximum and the most expected value of the factor. In this way the uncertainty is identified and estimated for each factor. The most uncertain factors should, if possible, be



divided into sub-elements and more information should be gathered in order to reduce the uncertainty. In this stochastic approach the “factors” are treated as elements that finally are summed up.

Also unlike the proposed ISO factor method, the estimates are expressed in years, instead of in numbers close to 1. These changes facilitate seeing the consequences of the estimates during the estimating process. After a statistical calculation the estimated service life is expressed as three figures, the expected value plus/minus one standard deviation. Two examples are shown where the service life is estimated for a window: first in a deterministic way according to the proposed ISO factor method, then in a stochastic way according to the proposed “step-by-step” principle, see Figure B2.12.

Figure B2.12: Example service life calculation

This method, although somewhat similar to the one proposed by Moser [1999], deviates substantially from ISO 15686. The restriction to a set form of density distribution and the different basic equation are further drawbacks of this method.


CIB W080 / RILEM 175 - SLM: Service Life Methodologies Engineering Design Methods: State of the Art 78 2.8 Other Concepts Some authors have used other than the above concepts for more sophisticated investigations.

• Estes and Frangopol [1999] have set up a model for bridges as a series-parallel combination of failure modes. Using time-dependent deterioration models and live load models, lifetime repair can be optimised. The authors state however, that considerable research effort is required to develop accurate input data: As reliability based methods gain increased acceptance, there is greater opportunity to use scarce resources more efficiently while maintaining a prescribed level of reliability of a structure throughout its service life. The goal is to provide management decisions that will balance lifetime system reliability and expected life-cycle cost in an optimal manner. This study proposes a system reliability approach for optimising the lifetime repair strategy for highway bridges. The approach is demonstrated using an existing Colorado State highway bridge. The bridge is modelled as a series-parallel combination of failure modes, and the reliability of the overall bridge system is computed using time-dependent deterioration models and live load models. Based on an established repair criterion, available repair options, repair costs, and updating, the optimum lifetime repair strategy is developed. The sensitivity of the optimum strategy to changes in various problem parameters including the prescribed service life, system failure criterion, and net discount rate is studied. Finally, the conclusions reveal that the proposed approach demonstrates real potential for practical applications, needs frequent updates through inspection, and requires considerable research effort to develop accurate input data.

• Raj [2000] presents the framework of a method, which partitions the systems into sub-systems linked by multidimensional variables. System analysis is done in the linking variable space (LVS) yielding detail information on how the sub-fields influence the overall variability of the service life. The example works on a light bulb and serves to illustrate how design regime is created in the LVS by overlaying the results from engineering design and materials science sub-systems.

• Liang et al. [2001] use a multiple layer fuzzy method model for the assessment of the service life of bridges. Thereby the deterioration is modelled using the fuzzy theory and the end of the service life is defined by the minimum safety index: The principal objective of this paper is to set up an evaluation multiple layer fuzzy method model for evaluating the damage state of existing reinforced concrete bridges. After establishing the detailed evaluation supportability and anti-seismic ability of existing bridges items, the fuzzy mathematical theory is adopted to evaluate the damage state of any member of an existing bridge. The damage state of any member of an existing bridge is systematically and completely composed as an evaluation model of multiple layer fuzzy mathematics. The evaluated results may be used for the safety index and reference index for repair or reinforcement in existing bridges. In addition, the evaluated results may also be used as a design reference for service life in future bridges. The evaluated model may be divided into the degrees of grades I, II, III, IV, and V, which are described as non-damage, light damage, moderate damage, sever damage, and unfit for service, respectively. Using the proposed model, the Huey–tong bridge, Jzyh–Chyang Bridge, Ay–gwo west road viaduct, and old Hwan–Nan viaduct in Taipei was chosen for evaluation. The results of the present investigation indicate that the order of repair and reinforcement in the Ay–gwo west road viaduct, Huey–Tong bridge, old Hwan–Nan viaduct, and Jzyh–Chyang bridge. Thus, the multiple layer fuzzy method appears to be of advantage for evaluating the damage of existing reinforced concrete bridges.


CIB W080 / RILEM 175 - SLM: Service Life Methodologies Engineering Design Methods: State of the Art 79 • Vu and Stewart [2002] use a reliability model which includes the spatial and random

variability of both chloride diffusion, concrete cover and concrete strength: In this paper, the service life of structures exposed to aggressive environments is measured by the probability of cracking and spalling of concrete cover. The time to corrosion cracking/spalling is experimentally investigated from accelerated corrosion testing of RC slabs with the emphasis on trying to quantify the relationship between concrete quality (w/c ratio; or strength), concrete cover, crack propagation and time. The probability of cracking and spalling of concrete cover is calculated by using a structural deterioration life-cycle reliability model. The reliability model includes the random spatial variability of concrete compressive strength, concrete cover and the surface chloride concentration. The reliability model also includes a stochastic deterioration model that considers the random variability of chloride diffusion, threshold chloride concentration and corrosion rates. Therefore, the reliability model can be used to predict the proportion of a concrete surface likely to spaull for any reference period (see Figure B2.13). This is a useful criterion for predicting the service life of RC structures.

Figure B2.13: Input data and resulting effects of concrete cover and w/c ratio on concrete cracking and spalling

Some of these concepts are even more sophisticated and more elaborate than probabilistic methods. The concepts are, at the present time, only suitable for research or specific use in singular and unique projects, the latter case that would warrant the cost of undertaking elaborate investigations.


2. literature review and state of the artcib w080 / rilem 175 - slm: service life methodologies...

Documents