CIB W080 / RILEM 175 - SLM: Service Life Methodologies

Factor Methods: State of the Art 21

4. GENERAL REQUIREMENTS FOR

SERVICE LIFE PREDICTION METHODS

Service life prediction of buildings or building elements, components or products can be both
a complex and time-consuming process. To date, the methods have not been developed into
an exact science given the many different factors that must be considered that thereby make a
thorough service life prediction an interdisciplinary activity. Service life prediction can be
based on two different principal approaches:

Deterministic approach
Probabilistic approach

This gives the basis for development of service life prediction methods of various
complexities and with different requirements of applicability and needs for input information.
Three levels of service life prediction methods can be described as shown in Figure A4.1.

Research methods
Probabilistic

Engineering
methods

Simple estimation
methods
Deterministic

Figure A4.1: Relation between different types of service life prediction methods.

The factor methods that are discussed in this report are based on the deterministic approach.

In 1987, Masters provided some general requirements to a service life prediction system
[Masters 1987].

1. "You should define the problem explicitly before attempting to solve it.
2. You should define service life such that a) it can be measured (quantitatively) and
b) it can be related to in-service performance.

NTNU – Norwegian University of Science and Technology

Department of Civil and Transport Engineering March 2004
CIB W080 / RILEM 175 - SLM: Service Life Methodologies
Factor Methods: State of the Art 22

3. You should be open to new approaches and methods rather than blindly accepting
those of tradition.
4. You should use simple and systematic procedures having a basis in logic, common
sense, and material science.
5. You should be aware that unsystematic, qualitative accelerated ageing test data
can be used to make anything look good, bad, or indifferent.
6. You should recognise that a) it is impossible to simulate all possible weathering
stresses in the laboratory, and b) it is not necessary to do it anyway.
7. You should ensure that degradation processes induced by accelerated tests are the
same as those encountered in-service.
8. You should measure the degradation factors.
9. You should be wary of the correlation trap.
10. You should recognise that, by using systematic, quantitative procedure, valid
acceleration tests can be developed."

In chapter 2 a brief overview was offered of the development over recent decades of service
life prediction methods. Based on these general requirements, a recommended procedure was
developed in which a systematic approach to methodology for service life prediction of
building materials and components is outlined [RILEM 1989]. The methodology is said to
include the identification of needed information, the selection or development of tests, the
interpretation of data and the reporting of results. It uses an iterative research approach,
thereby permitting improved predictions to be made as the knowledge base grows.
Mathematical analyses needed for prediction of service life are not described in detail, but
either deterministic or probabilistic analyses may be used.

The RILEM Recommendation is intended to be generic, and therefore applicable to all types
of building materials and components. Specific test methods and equipment vary depending
on the materials or components being evaluated as well as user requirements and for this
reason, this information is not included in the Recommendation document. The
Recommendation has been used as a basis for the development of the ISO Standard 15686
Part 2 [ISO 2001] The principle of this generic service life prediction method is given in
Figure A4.2.

In the Japanese Principal Guide [AIJ 1993] it is mentioned that various principles for the
prediction of physical service life have been proposed. Reference is given to some national
development and standards (US, France, Australia, Japan), along with the work of joint
CIB/RILEM committees. It is stated that the system for service life prediction used in the
RILEM Recommendation [RILEM 1989] is based upon the same principle as used in the
Japanese Principal Guide. Some more information from the Guide is given in chapter 5.

The problems of prediction of durability and service life are discussed in the British Standard
BS 7543:1992 [BSI 1992]. It is stated that a designer needs to have information on durability
to meet the building owner’s requirements and to develop a rational policy for the durability
of the entire construction system. The necessary information can be obtained from

a) experience in the use of traditional materials

b) certificates assessing the performance of products
c) research publications
d) predictions of the service life of products provided by their manufacturers

NTNU – Norwegian University of Science and Technology

Department of Civil and Transport Engineering March 2004
CIB W080 / RILEM 175 - SLM: Service Life Methodologies
Factor Methods: State of the Art 23

Definition
User needs, building context, type and
range of agents, performance requirements
Materials characterisation

Preparation
Identification of degradation agents, mechanisms and
effects, choice of performance characteristics and
evaluation techniques, feedback from other studies

Pretesting
Checking mechanisms and loads, and verifying choice
of characteristics and techniques by short-term
exposure
In-use-condition
(non-acc.) exposure
Exposure and evaluation
Accelerated Short-term Long-term
Field exposure
exposure exposure exposure

No Similar Inspection of
buildings
degradation
?
Experimental
Yes buildings

Dose - response
(degradation indicator)
Response classes

In-use exposure
Analysis/interpretation
Dose - environmental classes
Process performance-over-time
or dose-response functions to
establish prediction models

Service life prediction

Critical review, Reporting

Figure A4.2: Systematic methodology for service life prediction of building components.
(From ISO 15686 Part 2 [ISO 2001]).

Further, in BS 7543 the following statements are made regarding prediction of durability and
service life:

"4. Predictions of durability

A designer needs to have information on durability to meet the building owner's

requirements and to develop a rational policy for the durability of the whole
construction.
-----
It is important to note the following:

(i) Prediction of durability is subject to many variables and cannot be an exact

science.

A new building is a unique design to meet a specific set of conditions on a specific site.
Unless these conditions are the same as those previously recorded for precisely the
NTNU – Norwegian University of Science and Technology
Department of Civil and Transport Engineering March 2004
CIB W080 / RILEM 175 - SLM: Service Life Methodologies
Factor Methods: State of the Art 24

same form of construction, the predicted life for the building and its parts can only be
an estimate.

(ii) Accelerated testing of components by itself can seldom be used to give an accurate
basis for predicting service life. Accelerated testing is not usually feasible for large
assemblies of components.

(iii) Relevant test certificates are not always available from manufacturers and may
have to be obtained by testing for a specific project."
-----

"9. Predicted service life

9.1 Method of assessment

The predicted service life of a building should be assessed in one or more of the
following ways:

(a) Assess by reference to previous experience with the same, or similar construction
and in similar occupation or climatic circumstances.

(b) Assess by measuring the natural rate of deterioration over a short period of use or
exposure and estimating from the measurement when the durability limit will be
reached.

(c) Assess by interpolation from accelerated tests that have been devised to shorten
the response time to the action of an agent. The science of accelerated testing is
complex: care should be taken not to produce different effects by changing the natural
intensity of agents."
-----
"Prediction of service life will normally apply to components and small scale
assemblies. Whole buildings and large assemblies are more often one-off designs that
make previous experiences of durability less relevant and because of their size it is
less easy to test their performance under controlled conditions.

Whatever method is used to assess it, the predicted service life is unlikely to be a
precise figure because the effect of an action in any building is not likely to be
accurately predictable. More reliable predictions can be made when there is a
correlation between the results of different assessments."
-----
"In cases where the prediction of service life cannot be very accurate it may
nevertheless serve as a useful purpose when items are being ranked in order of
durability. The interpretation of data from tests requires skill and experience and
knowledge of building maintenance. It is often necessary to rely on an informed
opinion for service life prediction.”

Martin et al. [1994] have carried out a comprehensive study on methodologies for predicting
the service lives of coating systems. They present a set of criteria for judging the adequacy of
any proposed service life prediction methodology. These criteria include the ability to:

1. Handle large variability in the times-to-failure for nominally identical specimens

NTNU – Norwegian University of Science and Technology

Department of Civil and Transport Engineering March 2004
CIB W080 / RILEM 175 - SLM: Service Life Methodologies
Factor Methods: State of the Art 25

2. Analyse multivariate data

3. Discriminate among these variables. That is, the service life prediction
methodology should be able to separate the few significant variables from the
many insignificant variables
4. Fit both empirical and mechanistic failure models to short-term laboratory-based
exposure results
5. Establish a connection between short-term laboratory-based and long-term in-
service results
6. Provide mathematical techniques to predict the service life of a coating system
exposed in its intended in-service environment

In the Canadian Standard CS 478-95 [CSA 1995] there is also a general description of
methods for service life prediction. The alternative methods are described in the following
way:

"7. Predicted Service Life of Components and Assemblies

7.1 General
It is understood that the predicted service life of any building component, including
repaired as well as new components, is approximate based on the assumed
environmental conditions and on installation, operating and maintenance procedures.

7.2 Methods to predict Service Life

7.2.1
The predicted service life of components or assemblies may be assessed by one or
more of the following three methods:
(a) demonstrated effectiveness, in accordance with Clause 7.3
(b) modelling of the deterioration process, in accordance with Clause 7.4; and
(c) testing, in accordance with Clause 7.5

7.2.2
All methods used to determine predicted service life should be based on a sound
understanding and application of the principles of building science, in accordance
with Clause 7.6.

7.2.3
For the prediction of service life or an assembly,
(a) demonstrated effectiveness may be applied where identical assemblies have been
used
(i) successfully; and
(ii) in the same environments
(b) modelling and demonstrated effectiveness should be applied where
(i) a similar component or assembly has been used successfully in the same
environments; or
(ii) proven components or assemblies have been used successfully, but in
moderately different environments; and
(c) modelling and testing should be applied where
(i) innovative components and assemblies are to be used; or
(ii) proven components or assemblies are to be used in significantly different
environments.

NTNU – Norwegian University of Science and Technology

Department of Civil and Transport Engineering March 2004
CIB W080 / RILEM 175 - SLM: Service Life Methodologies
Factor Methods: State of the Art 26

The degree to which an assembly or its components are innovative or the service
environment is dissimilar to one previously experienced should be established by the
application of building science principles."

A RILEM workshop [RILEM 1995] on environmental aspects of building materials and

structures was held in Finland in September 1995. In a very brief summary of the workshop,
it is concluded that:

“Environmental aspects form a very complex problem area with many factors to be
included in the evaluation. This typically results in complicated evaluation and
assessment methodologies which are difficult to apply."
-----
"At present, there is no standard comprehensive methodology for assessing the
environmental issues of the entire building process and the life cycle of building
products. The methodologies in use today are incomplete and may give contradictory
results. Many of the applied assessment methods concentrate only on some aspects of
the total environmental performance."

Based on these conclusions, it is stated that there is a great need for design methodology and
methods that are capable of analysing, evaluating and optimising the environmental impacts
together with other multiple performance requirements.

Sarja and Vesikari [1996] have edited a RILEM Report on durability design of concrete
structures. They also present a discussion of what they call durability models. These models
may be:

Degradation models – mathematical presentations that show an increase in degradation

with time (or age) and with appropriate design parameters.

Performance models – mathematical presentations that show decreased performance

as a function of time and appropriate design parameters.

Service life models – mathematical presentations that show the service life of a
structure as a function of different design parameters.

The authors state that there may be durability models for different levels such as materials,
structural elements and buildings, and all of these can be used in durability design. Further,
they state that:

"7.1.2 Deterministic and stochastic durability models

Durability models can also be divided into deterministic or stochastic models.
Deterministic durability models are used in deterministic durability design where the
scatter of degradation (or performance or service life) is not taken into account. With
known values of parameters the model yield only one value (of degradation or
performance or service life) which is often the mean value. In some cases,
deterministic models are formulated to give an upper or lower fractile value instead of
the mean.

In many cases the information yielded by deterministic models is insufficient to

evaluate the risk of not reaching the target service life. Especially in the mechanical
NTNU – Norwegian University of Science and Technology
Department of Civil and Transport Engineering March 2004
CIB W080 / RILEM 175 - SLM: Service Life Methodologies
Factor Methods: State of the Art 27

design of structures, stochastic design methods are considered essential as the scatter
due to degradation is normally wide and the degree of risk may be great."

The authors have identified the following needs for durability models:

1. technical material development

2. ecological evaluation of materials
3. network level management systems for the maintenance, repair and rehabilitation of
structures
4. planning of project level repairs
5. risk analysis of important structures
6. design of a material mix and quality assurance at the construction site
7. structural durability design.

In the report the following evaluation is given regarding durability models:

“7.2.2 Quantification of degradation, performance and service life

The final step in the process of producing durability models is quantification and
formulation. Statistical methods and theoretical reasoning are the tools used for these
tasks. Simplifications, omitting irrelevant factors and limitation of relevant factors are
often necessary actions.

Durability models can be based on empirical or analytical grounds. Empirical models

are based on experience and test results. They are developed from results of field
surveys and laboratory tests by applying correlative and other statistical methods.

Analytical models are based on laws of nature and fundamental reasoning. They are
created as a thorough analysis of degradation mechanisms and kinetics. Before
models can be applied, tests are usually required for determining values for some
material properties.

Very often, empirical models represent a viewpoint of engineers, and analytical

models that of material scientists. A drawback of empirical models is that mechanisms
of influence are poorly understood in models in general. Consequently, any deviation
from the limits of the model may not be possible without entailing risk. Analytical
methods are based on a deeper understanding of the characteristic features of
damage, but their practical importance may be small if the parameters in the model
are not measurable or the models cannot otherwise be brought to a level of practical
utilization.

Both the empirical and analytical viewpoints should be considered when developing
durability models. Models can be considered good when based on an analysis of
mechanisms and factors leading to degradation, yet subjected also to laboratory and
field tests.”

In a discussion paper on service life prediction, Bourke and Davies [1997] present a list of
essential and/or desirable characteristics of a service life prediction system. They state that

NTNU – Norwegian University of Science and Technology

Department of Civil and Transport Engineering March 2004
CIB W080 / RILEM 175 - SLM: Service Life Methodologies
Factor Methods: State of the Art 28

"the relative importance of each is arguable, but important features may be

considered to include the following:

easy to learn
easy to use
quick to use
accurate
easy to update
easy to communicate
adaptable
supported by data
links with existing design methods and tools
free of excessive bureaucracy
recognises the importance of innovation
relevant to diverse environments
acceptable to practitioners and clients alike
reflects current knowledge
a flexible level of sophistication for either outline or detailed planning"

NTNU – Norwegian University of Science and Technology

Department of Civil and Transport Engineering March 2004