Management Infrastructure

maintenance and repair of transport infrastructure
TECHNICAL GUIDE
t
iNe
rat
Du
t
iNe
rat
Du
Copyright LABORATRIO NACIONAL DE ENGENHARIA CIVIL, I. P.

Diviso de Divulgao Cientfica e Tcnica
AV DO BRASIL 101 1700-066 LISBOA
e-e: livraria@lnec.pt
www.lnec.pt
Publisher: LNEC
Collection: Manuals
Series: MN 13
1st. edition: 2012
Printing: 100 copies
Descriptors: Transport infrastructures / Steel structure / Reinforced concrete structure /

/ Maintenance of structures / Repair of structures / Durability of structures /
/ Structural testing / Guide / Europe
Descritores: Infraestruturas de transportes / Estrutura metlica / Estrutura de beto armado /
/ Conservao de estruturas / Reparao de estruturas / Durabilidade de estruturas /
/ Ensaio de estruturas / Guia / Europa
CDU 624.05[012.4]9:625(026)(4)
624.05[014]9:625(026)(4)
ISBN 978-972-49-2237-9 (paperback)

STRUCTURES MANAGEMENT
t
PART II
iNe
maintenance and structural
assessment
VOL 2
rat
Du
Alan OConnor
Franck Schoefs
Denys Breysse
Sidi Mohammed Elachachi
Vikram Pakrashik
Authors
Alan OConnor
Associate Professor, Trinity College Dublin
Franck Schoefs
Professor, University of Nantes
Denys Breysse
Professor, University of Bordeaux
Sidi Mohammed Elachachi
Professor, University of Bordeaux
Vikram Pakrashi
Lecturer, University College Cork
t
Other contributions
Hugo Patrcio
Structural civil engineer, REFER
Anne Audouin-Dubreuil iNe
Principal engineer, Conseil gnral de la Charente-Maritime
Reviewer
Afonso Pvoa
rat
Senior civil engineer, EP
Final revision by Editorial Commission member

Albert Daly
Research manager, NRA
Du
PREFACE
This is Part II, Volume 2 of the DURATINET Technical Guide - Maintenance
and Management of Transport Infrastructure, containing the maintenance and
structural assessment planning.
The content of this volume was prepared and reviewed within DURATINET
working group WG A2 Structures management.
The aim of WG A2 was to discuss and identify knowledge gaps as well as to
harmonize needs on structures management. Within this WG, technical
information on the management of steel and concrete infrastructure was
collated and compiled. All these subjects are considered in the two volumes of
Part II of the DURATINET Technical Guide.
t
iNe
WG A2 Structures Management
WG Leader:
rat
Allan OConnor
Trinity College Dublin, Ireland
Partners active members
Country Institution Members
EP Afonso Pvoa,
Portugal
REFER Hugo Patrcio
University of
Denys Breysse, Sidi Mohammed Elachachi
Bordeaux
Du
University of
Franck Schoefs
France Nantes
Conseil gnral de
la Charente- Anne Audouin-Dubreuil
Maritime
TCD Alan OConnor
Ireland
NRA Albert Daly
i
DURATINET project approved by the Atlantic Area Programme and co-financed
by ERDF
CONTRACT N: 2008-1/049
PROJECT TITLE: Durable Transport Infrastructure in the Atlantic
Area Network
ACRONYM: DURATINET
LEADER: Manuela Salta
Laboratrio Nacional de Engenharia Civil (LNEC)
Materials Department
Portugal
t
iNe
rat
Du
ii
GENERAL INDEX
Part II Structures Management
Vol. 1 General guidelines
Structural inspection
Structural testing
Structural condition rating
Archiving and reporting
Vol. 2 Maintenance and structural assessment
Structural assessment
t
Structural response modelling
Load modelling
iNe
Resistance modelling
Reliability analysis (assessment and optimising maintenance)
rat
Du
iii
CONTENTS
Part II - Structures Management: Vol.2
1 Executive summary ......................................................................................1
2 Introduction ...................................................................................................4
3 Structural assessment ................................................................................23
3.1 Global safety factor format .................................................................27
3.2 Partial safety factor format ..................................................................28
3.3 Reliability formats ...............................................................................29
3.4 Socio-economic format .......................................................................30
t
4 Structural response modelling ....................................................................31
4.1 Methods of structural analysis ............................................................31
iNe
4.2 Integration of field data and structural models ...................................32
5 Load modelling ...........................................................................................35
5.1 Time invariant loads............................................................................35
5.2 Time variant loads ..............................................................................35
5.2.1 Statistical modelling ....................................................................... 35
rat
5.2.2 Load monitoring data required for assessment ............................. 37
5.2.3 Assessment of characteristic traffic load effects ........................... 38
5.2.4 Direct simulation ............................................................................ 38
5.2.5 Poisson models ............................................................................. 40
6 Resistance modelling .................................................................................42
Du
6.1 Uncertainty in material properties .......................................................42

6.2 Modelling of materials in assessment.................................................45
6.3 Deterioration models...........................................................................45
6.3.1 Mechanistic models ....................................................................... 47
6.3.2 Deterministic models ..................................................................... 50
6.3.2.1 Straight-line extrapolation .......................................................... 50
6.3.2.2 Regression model ...................................................................... 50
6.3.3 Stochastic models.......................................................................... 50
6.3.4 Artificial intelligence deterioration models ..................................... 52
6.4 After-repair deterioration model ..........................................................53
7 Reliability analysis (assessment and optimising maintenance) .................54
iv
7.1 Probabilistic analysis ..........................................................................55
7.1.1 Calculation of the probability of failure (Pf) .................................... 57
7.1.2 Target reliability T ......................................................................... 57
7.2 Probabilistic load modelling ................................................................61
7.2.1 Dead & superimposed load ........................................................... 62
7.2.2 Traffic load ..................................................................................... 62
7.2.3 Probabilistic resistance modelling ................................................. 63
7.3 Incorporation of model uncertainty .....................................................64
7.4 Incorporation of supplementary information .......................................65
t
8 Examples ....................................................................................................68
8.1 Ferrycarrig Bridge, Co. Wexford, Ireland ............................................68
iNe
8.2 Barra Bridge, Aveiro, Portugal ............................................................71
8.3 Steel Wharf, Nantes, France ..............................................................72
9 Conclusions ................................................................................................78
10 References .................................................................................................80
rat
Du
v
Du
rat
iNe
t
MAINTENACE AND STRUCTURAL ASSESSMENT
1 Executive summary
In infrastructure management, decisions regarding maintenance, repairs and
rehabilitation are based on the assessment of the current and future condition of
the structure with respect to certain performance indicators.
Sometimes, additional socio-political or economic markers affect such decisions
over which there is usually little engineering control. The key idea is to ensure
the safety of a network of infrastructure in its entirety above minimum
prescribed levels of service at any given time within its lifespan.
This engineering philosophy is best captured by infrastructure management
through managing risks expressed probabilistically as the reliability of a
structure evolving in time. The probabilistic description of evolving risks takes
t
account of uncertainties in
material properties;
iNe
actions on the structure;
lack of knowledge regarding structural behaviour;
errors related to measurements;
human fallibility.
Within an infrastructure network, risk may be quantified as a sum of the
respective probabilities of failure of each individual structure multiplied by the
rat
consequences of this failure. A failure is defined as a non-compliance with one

or more performance criterion/criteria and may or may not correspond to
collapse or significant visual distress. The quantification of consequences of a
failure can be variable and subjective at a given time based on the interpretation
of the consequences. As a result, engineering recommendations focus on
managing risks by managing the probability of failure. The consequences are
broadly used for classification of the performance markers and the structure.
Du
The marker of reliability of a structure under this prevailing uncertainty and

specified performance criteria is known as reliability index (). The objective is
to maintain all structures within the infrastructure network above a minimum
acceptable level of performance (min) for a set of specified performance criteria.
Under such a framework, the failure probability of structure is equal to the
mapping of the negative value of the performance index to an equivalent
standardised correlated joint normal distribution or Nataf Distribution [1]
comprised of all stochastic variables affecting the system. Fig.1 depicts the
broad philosophy behind infrastructure management.
1
t
Fig. 1. Infrastructure management.
iNe
Any structure is associated with a reliability index in its as-built condition since
its inception. With time, the reliability index drops closer to the minimum
allowable value due to structural deterioration and the evolution of loads on the
structure. If no intervention takes place, such a degrading structure falls
unacceptably below the minimum allowable safety level after a certain time.
This time is quite often smaller than its intended service life. To increase the
service life, interventions of various degrees and at various frequencies can be
carried out to the structure as illustrated in Fig. 1.
rat
Such interventions improve the reliability index of the structure to various
degrees and the service life is extended as a function of these interventions
along with the ensuing rate of degradation and evolution of load. The
interventions on structure throughout its service life and the related extension of
service life as compared to a non-intervened situation are associated with
various levels of cost. Infrastructure management attempts to recommend the
best intervention option in the form of maximum improvement in reliability index
Du
against unit investment looking at whole life safety. To summarise,

infrastructure management provides with the best cost-beneficial solution at a
network level without compromising with the safety of the structure at any point
of time within any element of the infrastructure.
Since the performance of elements in an infrastructure network is probabilistic,
the effectiveness of infrastructure management decisions is dependent on the
quality of data and deterioration models estimating the future condition. The
quality of data relates to the condition assessments of a structure, while
deterioration models attempt to extrapolate the condition of the structure in
future (or sometimes in past) as accurately as possible. Of course, the
assessments and models can be formulated at various degrees and levels of
detail, accuracy and sophistication all being a function of requirement, time
and investment. Even with very good quality data, an amount of uncertainty
relating to human errors, modelling errors and testing errors remain which may
impact management decisions. This uncertainty should be modelled
probabilistically and quantified to assess the impact of this uncertainty on final
2
decisions. Not all of the participating probabilistic variables in infrastructure

elements can be measured or have well calibrated probabilistic descriptions.
Experienced engineering judgements should be exercised while constructing
probabilistic descriptions of such variables. Reliability analysis and probabilistic
assessments of structures are usually accompanied with sensitivity analyses
and participation factors of the variables in the final assessment values.
Consequently, the variables affecting the final results the maximum can be
revisited, re-measured, reformulated or monitored to ensure the correctness or
realistic description of it in a probabilistic format.
The main questions asked during infrastructure management can be
summarised in a 10 point format:
1. Why assess structures?
t
2. How to assess structures?
3. When to assess structures?
iNe
4. What to assess in a structure?
5. Who is eligible to assess the structures?
6. Where to intervene?
7. Which structures to manage with priority?
8. How to decide on indicators of performance for this prioritisation?
rat
9. How to intervene?
10. How can we cost-optimise the planning of maintenance?
These questions form the background and the basis to this volume. This volume
attempts to provide some general guidelines to address the issues raised above.
With time, the structural engineering sector has shifted away from design of
macro-scale infrastructure elements to a more multidisciplinary diverse context
and the shift of its centroid towards maintenance and assessments can be felt
Du
significantly. This trend is only to gain stronger with time.
3
2 Introduction
Our built infrastructure forms lifeline systems for our communities and needs to
cater for an increasing demand in transport capacity. While this demand is
constantly on the rise, a significant number of these built structures have
deteriorated since their inception and are degrading with time [2]. Corrosion,
spalling, excessive deformation, cracks and other signs of visual distress are
often associated with such degradation. As a result, in-service structures
require assessment and a prioritisation of the intervention measures is
necessary to ensure their structural integrity and safety. It is generally accepted
that the use of design standards for assessment is over-conservative and can
lead to unnecessary replacement or strengthening of existing structures with all
the attendant costs of traffic delays. In the design phase, loading conditions are
t
over-estimated and structural strength under-estimated to cater for the inherent
uncertainties associated with in-service conditions. While the cost of providing
iNe
this enhanced level of safety is marginal at the design stage, the same cannot
be said for assessment. Over-conservatism in assessment can lead to
considerable unnecessary expenditure by wrongly identifying a structure to be
underperforming where in reality the performance is within acceptable limits.
The principle of assessment thus lies in establishing the true level of safety of a
structure as faithfully as possible without compromising with acceptable safety
levels. The objective is to associate the performance of a structure with
indicators that faithfully capture the level of safety of a structure in its as-built
condition, allow comparing the safety levels of different structures and lend to
rat
further analyses and interpretations in interpolating or extrapolating the level of
safety based on the behaviour of the structure.
It is important to note here that the indicator/s of the performance of a structure
is/are dependent of what is/are defined as performance criterion/criteria. Usually,
structures are supposed to cater for safety and serviceability criteria. For all
these criteria, it is possible to define limiting conditions related to performance
called limit states. Structural performances are deemed unacceptable when a
Du
limit state is negative while they are acceptable when the limit states are
positive (zero being the limiting condition). Safety limit states relate to very
onerous consequences of the structure and are related directly to an actual
collapse or significant structural distress related to safety. A non-compliance of
these limit states is thus related to great risk. A non-compliance of serviceability
limit states significantly hampers the performance of a structure and often tend
to shorten the service life, but is not usually related directly to collapse or life
risks. Consequently, a structure that has failed under serviceability criteria may
not collapse and still be deemed acceptable under safety criteria. In many
cases, the minimum acceptable performance of an element in built
infrastructure is guided by safety limit states when assessing while a both safety
and serviceability guides a design. Consequently, an appropriate assessment of
a structure allows re-evaluation of its performance and increasing its service life.
There remains a possibility of conflict between the performance criteria
specified by the owner, the manager or the end-user of built infrastructure
networks and that specified by the engineers from safety and serviceability
4
requirements emanating purely from an engineering/scientific viewpoint. It is

recommended to adopt the engineering criteria under such circumstances to
ensure that the minimum safety standards are rigorously and unambiguously
established. Table 1 presents some bridge failures as an illustrative example of
some of the consequences of failure. As defined in the previous section, the
failure is a non-compliance with one or more performance criterion/criteria and
may or may not correspond to collapse or significant visual distress.
Consequently, consequences more benign than those listed in Table 1 may
have a direct and significant impact on structural management. This includes
the failure of railings, or the joints of a bridge deck. Damage from settlements
may or may not be benign, but is extremely important for maintenance
management.
t
Table 1. Some bridge failures in the last 40 Years. [4]
Bridge Cause of Failure
Silver Bridge, USA
iNe
An undetected crack formed in an eye bar during
manufacture had deepened due to corrosion and eventually
1967
the eye bar broke at the eye. Loss of 38 cars and 46 dead.
Milford Haven, UK Buckling of box girder diaphragm during construction. Four
1970 fatalities.
Kempten, West Road bridge centre span collapse as concrete being poured.
Germany 1974 Nine dead. Failure of A Frame falsework supports.
Ship hit unprotected twin tube pier. Arch deck fell into fjord.
Almo, Sweden 1980
Led to tighter regulations for pier protection.
rat
Single span, 32 year old segmental prestressed concrete

Ynys y Gwas Bridge,
bridge over the River Afan collapsed without warning. Cause:
South Wales, UK
corrosion of the longitudinal and transverse prestressing
1985
wires.
Schoharie Creek, Foundation collapse leads to mains span falling into creek.
New York State, USA Ten motorists dead. Scour undermined concrete spread
1987 footings.
Changson Viaduct, Pier failure shortly after opening causing collapse of main
Du
Seoul 1992 span. One dead, seven injured.

Haeng Ju cable Collapse of a temporary pier during construction. Only half of
stayed, Seoul, Korea the required volume of concrete had been placed in the piers
1992 foundation.
Songsu, Seoul, Korea Collapse of the midspan truss members due to fatigue
1994 exacerbated by traffic overloading. Thirty two dead.
Scour undermining one of the six masonry piers during flood
Entre os Rios bridge,
conditions, exacerbated by illegal sand extraction around the
Portugal 2001
piers, and hitting of piers by local ships. Seventy dead.
NB: The technical guide (TG) focuses mainly on bridges because this field was fruitful in terms of
development of methods. It can be extended to all infrastructures. One such example, on wharves,
is presented in the last section.
A true descriptor of the performance of a structure allows any number of built

infrastructure assessments to be compared relative to each other and ranked
according to their current, past or projected performances. Investment available
for built infrastructure is limited and gets more challenging [3] under difficult
economic climate. The prioritisation of investment over time within a built
5
infrastructure network is thus a key global challenge. An assessment relating to

the true safety of a structure tends to rank and compare the various elements
within a built infrastructure accurately and ensures correct prioritisation of
investment. A system optimal budget planning and minimisation thus directly
relates to the level of success achieved in assessing the built infrastructure on a
network level.
Bridge Management Systems (BMS) have assisted in striving towards the goal
of optimised maintenance management at a network level. A BMS is typically
characterised a set of rules, protocols, guidelines, and procedures to identify a
management strategy. The systems often refer to the computer programs or
software that organize and automate these rules, guidelines, and often include
storage and organization of inventory and inspection information, maintenance
scheduling, and work-program optimization. The key tasks in a BMS usually
t
involve collection of inventory data, inspection, assessment of condition and
strength, repair, strengthening or replacement and prioritizing the allocation of
iNe
funds. The systems are usually made of four significant components relating to
data storage, cost and deterioration models, optimization and analysis models
and updating functions.
[5]
A representative structure of a BMS includes registration and description of
the individual structures, initial determination of their actual condition,
judgement of structures in terms of safety, prediction of their future behaviour
resulting from ageing, definition of specific maintenance strategies (estimation
of maintenance-related costs such as inspection, repair, replacement, etc.),
allocation of the limited budget to specific structures, execution of maintenance
rat
strategies with adequate manpower and means, registration and evaluation of
funding, back coupling with maintenance strategies and design and scenarios
for the replacement of structures for economical, technical or other reasons
(increased traffic, new demands).
Typical modules of BMS are related to administration, inventory, inspection,
maintenance (including cost estimation) and prioritization (optimisation) related
to deterioration models and prioritisation functions. An administration module
Du
attempts to define the users (public, bridge inspectors, bridge administrators,

system administrator) and set up user privileges by defining data access, data
editing rights, use of analytical tools and rights of decision making. The module
is also capable of clustering bridges into different groups according to varying
criteria while allowing for data archiving.
An inventory module contains fundamental information on each structural
member included in the BMS. It generally includes the description of the bridge
and its sub-elements, documentation in all stages, photographs and drawings,
information about the storage of other documents, summary of condition
assessment and summary of load carrying capacity or structural safety
assessment.
The inspection module typically contains information about planned and
completed surveys. Condition data are based on the results of inspection,
[6]
testing and monitoring (according, Part IV Volume 3 of the Technical guide ).
Common methods of data acquisition involve the study of documents,
6
inspections and material testing, performance testing and monitoring (structural

health monitoring, system identification by static and dynamic measures, proof
load tests) and monitoring of live loads and environmental conditions.
A maintenance module is sometimes available where information about planned
and realized maintenance, repair and other types of activities in hierarchical
order are available. Under the maintenance module it is possible to enter
information about the activities required for maintenance, repair, etc. and to
register presumed costs. For this purpose a cost catalogue is used which can
be incorporated in the BMS. Once the activity is complete, the module allows
the real data to be entered, including actual costs incurred. The module enables
some statistical evaluations to be made. The type of activity, element of the
bridge involved, cost of activity (presumed/real), date of realisation (planned
term based on the degree of urgency/date of finishing), and personnel involved,
t
establishment of the nature of improvement and the warranty period are
typically recorded.
INVENTORY INSPECTION
iNe MAINTENANCE PRIORITISATION
MODULE MODULE MODULE MODULE
bridge: - different types - required activities; predictions:

- description; of inspections, - carried out - deterioration
- elements; testing, activities; (ageing);
-documentation; monitoring and - cost of activities. - failure risk;
- current modeling; - traffic analysis:
condition; - results of - life cycle cost
- current load survey; analysis (LCCA);
rat
capacity; - suggestion for - cost benefit

- current change of analysis (CBA)
serviceability. condition and prioritization;
structural safety - choosing the best
state. Strategy;
- deterioration
Models;
-budget
Du
bridge (project) level bridge (project) level

network level
Fig. 2. Structure and information flow scheme of Bridge Management System. [7]
The prioritisation module is based on information which was collected within the
previously mentioned modules and carried out predictions and analysis. The
module may operate at a project (e.g. bridge) level or at a network level. The
project-level module deals with optimisation and selection of the optimal
strategy for a certain bridge and is often based on life cycle cost analysis. The
network-level module optimizes bridge actions for a group of bridges and
various criteria are taken into consideration rather than only condition and
structural safety.
The planning is carried out in two time horizons at the least (short-term and
long-term). The most common criteria for making decisions include the
condition of the bridge or the network, safety and serviceability, budget limits,
7
restrictions to users, state, regional or other programs and priorities and social
and environmental consequences. Prioritisation modules use tools related to
deterioration models, cost analysis during the bridge lifetime, effectiveness
analysis of previous interventions and prediction of traffic growth, etc. Standard
follow-up activities from the output of BMS systems include routine maintenance,
repair, rehabilitation, strengthening or upgrading and reconstruction.
A number of European and American projects have investigated the decision
making processes associated with maintenance and reconstruction of bridges.
BRIME, a project funded by the European Commission under the Transport
RTD program of the 4th Framework program, focused on bridge management
processes. The project considered models attributed to Frangopol et al [8],[9] and
Branco and Brito [10]. The Frangopol [8],[9] model proposes a general methodology
for determining the optimum inspection/repair program for new and also for
t
existing bridges by minimizing the expected life-cycle costs while maintaining an
acceptable level of reliability. The method determines the optimum inspection
iNe
technique, number of lifetime inspections, number of lifetime repairs, and the
timing of these inspections/repairs. The expected total life-cycle cost CET
includes the initial cost CT and the cost of preventive maintenance CPM,
inspection CINS, repair CREP and failure CF:
CET =CT +CPM +CINS+CREP +CF (1)
All of them are expected costs and are computed from probability measure of
events like the detection, the false alarm and the failure. In some cases, not
only the expectation but fractiles of the total cost are considered [11]. The
rat
objective is to minimize CET while keeping lifetime reliability of the structure
above a minimum allowable value leading to the optimization problem under
constraint.
Minimize CET subject to Pf,life P*f,life , or (2)
Minimize CET subject to f,life *f,life , where (3)
= -1 (1 Pf) (4)
Du
where P*f,life is the maximum acceptable lifetime failure probability, f,life and
*f,life are the lifetime reliability index and the lifetime target reliability index, is
the standard normal distribution function. The Branco and Brito [10] model also
propose methodology developed to support decision-making on bridge repair
strategies. The decision criteria of this methodology are part of a global
management system which includes a periodic inspection strategy and the
selection of repair works which are performed with a knowledge-based
interactive system. The repair decision module is based on a cost/value
economic analysis which compares repair cost and their subsequent benefits
for the expected remaining service life of the structure for each repair alternative,
helping authorities to reach rational decisions. A global cost function C is
represented here as:
C=C0+CI+CM+CR+CFB (5)
where C0 are the initial costs, CI are the inspection costs, CM are the
maintenance costs, CR are the repair costs, CF are the failure costs and B are
8
the benefits. It is to be noted here that even before a preselected performance

criterion or a limit state is reached some cost may be accrued due to the
deterioration and loss of efficiency of the structure. Consequently, it may
sometimes be necessary to consider the idea of cost of non-performance
separately and appropriately modify the cost format. The benefits from a bridge
also take into account the road of which it is a part. In the decision-making,
every repair decision is made according to the Costs Effectiveness Index (CEI)
of each option that indicates how well the proposed repair compares to the no
action option and is defined as and may be employed to compare different
solutions for the repair of the same problem or to obtain the priority of action
among the repair of different anomalies within a bridge or different bridges in a
network.
(6)
t
Bastidas et al [12] suggest a multi-objective optimization with a specific function
iNe
when considering waste generation and CO2 emission. There are two
advantages in this approach. Firstly, the owner keeps in mind the origin of the
cost until the end of the analysis. Secondly, it avoids adding new assumptions
and assessments of the costs.
A multi-objective optimization for bridge management systems [13] report
considers network - and project-level optimization of multiple, user-specified
performance criteria. The network-level model provides a decision making tool
that optimizes bridge actions for multiple performance criteria. The bridge-level
rat
model evaluates the effect of bridge action alternatives on life-cycle cost and
other performance criteria for the purpose of selecting projects that are
consistent with the network goals. Both models use the AASHTO Bridge Ware
database, supplemented with additional data as needed. The bridge-level
model considers recommendations from the network-level model. In addition,
the network-level model can consider projects selected within the bridge-level
model. These models also can operate independently. Both models explicitly
consider the inherent uncertainties of estimated costs and outcomes. The
Du
models are implemented in graphical design software that helps bridge

managers visualize the life cycle of individual bridges and bridge inventories. A
comprehensive, yet minimal and operational set of goals and performance
criteria were described as preservation of bridge condition, traffic safety
enhancement, protection from extreme events (scour, fatigue/fracture,
earthquake, collision, overload, and other human-made hazards), agency cost
minimization and user cost minimization. The solution is based on utility theory
and the utility function involves weighting, scaling and amalgamation to yield an
appropriate objective function for optimisation. The bridge-level methodology
includes a life-cycle cost framework, preservation and functionality models,
candidate definitions, and their evaluation. The methodology also includes
optimization at the bridge-level in a bid to maximize utility of bridge actions in
the long term by selecting from an array of scoping and timing alternatives. The
method includes a recursive approach that is consistent with input data
available from existing bridge management. Intervention options involve do
nothing, Maintenance, Repair, Rehabilitation, and Improvement (MRR&I) and
9
total replacement. The MRR&I approach is generated through an investigation

of individual possible actions defined for each condition state of each bridge
element.
From a user perspective, software can aid bridge maintenance planners and
bridge programme managers where the planner typically gathers a set of
relevant information, manipulates it, and records any decision made from the
data analysis. The decision maker typically seeks information about the effect of
current decisions on the future performance of the bridge. Subsequently, the
decision maker defines a project, forecasts the future outcome, and then
adjusts the project and re-estimates the impact. These may be addressed at a
BMS level through user-case definitions where the tasks of the bridge
maintenance planner at bridge-level use-cases may be preparing background
inputs, selecting bridges for attention, analyzing candidates (treatment, types
t
and at network-level use-cases may be involved in prepare background inputs,
define program, analyze trade-offs and adjust candidates.
iNe
The assumptions and the assessment methods for each of the procedures
depend on one of two scenarios: the certainty scenario and the risk scenario.
The certainty scenario refers to the case where the consequence of each
alternative in terms of multiple criteria is known with certainty. The risk scenario
refers to the case where the trade-off issue remains (as in a certainty scenario),
but difficulties are compounded because it is not clear what will be the exact
consequences of each of the alternatives. A bridge-level optimization is
conducted by generating and comparing candidates that are alternative life-
cycle activity profiles and resulting performance predictions for a bridge. Each
rat
life-cycle activity profile is modelled as an infinite time series of cash flows
representing various types of annual agency and user costs. Because all of the
costs occur at various times in the future, they are processed in a standard
engineering procedure called the net present value analysis. Each cost item is
discounted by an amount that depends on how far in the future it occurs. The
discount factor represents how much less it matters for each year that it is able
to delay the cost. The essential decision to be optimized in the bridge-level
Du
analysis is the scope and timing of the first intervention. When multiple
candidates are defined for a bridge, they differ in terms of the first intervention.
Timing of the first intervention determines the length of the first waiting period,
which may vary from zero to the full length of the program horizon. Consequent
interventions are forecast for programming and for life-cycle cost analysis, but
are not the subject of the decision making by the bridge maintenance planner.
The preservation model framework in the Bridge-level optimization consists of a
collection of sub-models that work together to serve the life-cycle cost
framework. Some of the models are used in more than one stage of life-cycle
costing. Certain models are adapted directly from Pontis, BMS software, and
are intended to give the same results as Pontis, while others are developed
specifically for project-level analysis. The main models are deterioration
(prediction of the future conditions of elements on a bridge on the basis of the
most recent inspection and possible candidate implemented during the program
horizon), action effectiveness (prediction of the outcome of an intervention on
each element or on the bridge as a whole), cost estimation (computation of the
direct cost of preservation work on the bridge by aggregating over all elements
10
using their predicted quantities and units costs, indirect costs are calculated
separately), economic failure (description of the limiting behaviour of element if
deteriorated conditions are not corrected, in the form of unplanned actions) and
long-term cost (estimation of life-cycle costs beyond the end of the program
horizon based on conditions predicted at the end of the horizon). As the primary
cause of preservation needs in a bridge inventory, deterioration is the main
driver of the life-cycle cost analysis in the bridge-level model.
Markovian models are popular in this regard. A Markovian model assumes that
the probability of making a transition from one condition state to another
depends only on the initial state and not on past conditions or any other
information about the element. Thus, the model is expressed as a simple matrix
of probabilities, which can be manipulated by matrix multiplication. Markovian
transition probabilities are used to express the condition immediately following
t
the action, assumed to be at the beginning of the implementation year. These
probabilities are then multiplied by the do-nothing transition probability matrix to
iNe
forecast farther into the future. Because scope items are expressed in terms of
action types applied to multiple elements, action outcomes are expressed in the
same way when performance measures are calculated. Each candidate in the
cost estimation model has an initial cost that is assumed to occur at the
beginning of the implementation year. For the do-nothing candidate, the initial
cost is always zero.
In the Pontis network optimization, the role of the failure concept is to help
develop policies that generally do not permit failure. At the bridge level in the
economic failure model, factors such as constrained funding, the typical rule of
rat
10 years of deliberate inaction following a project, and the normal uncertainty in

deterioration prediction present the possibility that failure could occur in isolated
cases. So, the model needs a valid way of defining what we realistically mean
by failure and quantifying the effect of failure on life-cycle costs, particularly
the effect on road users and the cost of emergency repairs or replacement. In
the bridge-level life-cycle cost analysis, failure is a part of the probabilistic
outcome prediction. The model assumes that failure, though improbable under
Du
normal conditions, can happen on specific bridges. It is not necessary to predict

the specific bridge elements that would fail. However, it is necessary to
recognize the possibility of failure and include the expected value of the life-
cycle cost impact in the analysis for every bridge.
If a program is well funded and conditions remain relatively good, the failure
cost should be small and insignificant. However, if poorly funded or the
deterioration proceeds more rapidly than normal, or if the bridge structure is
particularly vulnerable to natural or human-made disasters, the failure becomes
more likely, and the failure cost is high. After the outputs and outcomes of a
candidate are fully analysed, the remaining step is to establish a basis for
comparing the candidate with other candidates.
Performance measures and a utility function are used to do this. There is a
table, where all of the performance measures used in the system are listed and
the data needed for evaluation and optimization in the system are organized.
The optimization can operate in a mode where the timing of interventions is
governed by the worst tolerable performance thresholds, or it can select optimal
11
timing based on a utility function (or a combination of both). Optimal scoping is

always based on the utility function. Each performance measure may participate
in a utility function for selecting the optimal candidates. Utility functions are
always oriented so that higher values are considered good and lower values
bad. If a performance measure is given a nonzero weight, then it is considered
in the bridge-level optimization. It is permissible for different groups of bridges to
use different sets of weights. An agency can define its performance measures
based on any available data about the bridge. In each pass through the bridge-
level optimization loop, a worst tolerable performance calculation and utility
function calculation occur using the performance measure table.
The outcome prediction occurs in various ways appropriate to each
performance measure. The NBI (National Bridge Inspection) serviceability
ratings and vulnerability assessments are taken directly from the bridge
t
inventory, subject to any changes made by earlier interventions and the
condition and sufficiency assessments are forecast to the end of the
iNe
intervention year using the element-level Markovian deterioration model and
converted to NBI indicators using the FHWA translator program. Life-cycle costs
are calculated, starting at the beginning of the intervention year and discounted
to that point, using the procedures. Custom performance measures are
envisioned to be defined as formulas with access to any available bridge-related
data so that they are calculated by updating the formulas. A base case
prediction is also made for each performance measure, assuming that no action
is taken that year, or in any future year, until the end of the program horizon.
Interventions taken in any earlier, higher-level recursions are assumed to be the
rat
same for both the intervention outcome and the base case. Benefits are
calculated by subtracting the base results from the intervention outcome for
each performance measure. After calculation of the final utility function value,
the candidate is compared first to other candidates (having different scoping
approaches) for the same year. If it turns out to be the highest-utility candidate
for the year, it is then compared to the best candidates for other years to
determine the optimal candidate.
Du
The NBI condition ratings are based on the evaluation of the materials and the
physical condition of the deck, superstructure, and substructure. These ratings
vary from 0-9 under the current standard where a value equal to 0 indicates a
failed condition and a value equal to 9 indicates excellent condition. Condition
ratings for sub - element like Deck Condition Rating (NBI Item 58),
Superstructure Condition Rating (NBI Item 59) and Substructure Condition
Rating (NBI Item 60) are also applicable.
[14]
Another European project, SAMARIS , funded by the European Commission
under the Transport RTD program of the 5th Framework program investigated
optimised condition assessment and structural safety assessment of highway
structures. The situation of built infrastructure was indicated for 12 countries
through questionnaire surveys. The analysis of the condition assessment
methods used shows that every bridge administrator in every country has
created or adopted a system for managing bridge stock. All systems have
common roots and similar rules. However, the systems are incoherent; they
take similar factors into consideration, but present different outcomes. European
12
countries with a huge stock of highway bridges, such as France, Germany, Italy,
Poland and Spain, do not use specific procedures for the safety assessment of
existing highway bridges and, in general, the basis for the assessment is the
same as for the design of new bridges. Some other countries, like the UK,
Denmark, Sweden and Slovenia, but also Canada and USA, have specific
codes or recommendations for the safety assessment of bridges, but the
procedures used, despite being based on the same principles, vary
considerably. A review of condition assessment methods of bridges used in
Europe and the United States showed that there are two main approaches to
the evaluation of the condition of the whole structure based on the condition
assessment of its elements. The first approach is related to cumulative
condition rating, where the most severe damage on each element is summed
for each span of the superstructure, each part of the substructure, the
carriageway and bridge equipment and the second approach is concerned with
t
the highest condition rating of the bridge components as the representative
condition rating for the structure itself.
iNe
Several BMS software are present, of which PONTIS, developed by FHWA and
[15]
six DOTs is a representative candidate. Approximately forty of fifty U.S.
states use or currently own Pontis. Pontis assist an agency with bridge
management from inspection to scheduling MR&R or replacement. "Pontis
supports the complete bridge management cycle, including bridge inspection
and inventory data collection and analysis, recommending an optimal
preservation policy, predicting need and performance measures for bridges,
and developing projects to include in an agency's capital plan" [16]. The
rat
deterioration rate is determined by condition data such as the inspection rating

of given components. The ratings are from one to five. The deterioration model
takes into consideration not only deterministic models, which rely on using
previously and currently stored ratings for historical data, but also probabilistic
models or Markovian model. In addition, Pontis has the capability to perform
network-level analysis with various budget constraints [17]. Pontis data illustrate
the potential deterioration of the bridge and the expected failure time so that an
agency can schedule repair. It also exemplifies the deterioration that will occur
Du
by neglecting maintenance or replacement measures. However, using Pontis

data will only assist in an approximation. Failure may occur earlier or later than
the expected time [15]. Pontis considers a do-nothing, a repair and a
replacement option in terms of intervention. Additionally, an engineer can
generate three more plans over a period of time longer than one year. All
suggestions are compared to the "do-nothing" proposal, which has an elevated
failure risk and a long-term cost. Pontis offers flexibility to customize the
database for easier use for a given agency. It can assist in inspection of asset
management, project planning, and programming such as developing applets
and applications [16]. With customization, Pontis can also aid to some extent in
project-level analysis. However, Pontis requires that it must be installed in every
computer within the agency for use. Therefore, the task of installing Pontis is
tedious. It is further complicated by the fact that Pontis creates updates and
thus, the new version of Pontis has to be installed onto all the computers again.
In addition, for every update, the data would have to be migrated over into the
new version. Pontis also causes the inspection data to be delayed since the
13
Pontis Data Interchange files (PDI) are sent from the counties to the districts
and then to the DOT. The data cannot be entered directly into the main server
from either the inspection site or the county.
BRIDGIT, developed under an AASHTO-sponsored National Cooperative
Highway Research Program, makes use of Markovian deterioration. BRIDGIT
inventories up to several thousand bridges and provides specific actions for all
bridges keeping in mind the network-level strategy, costs, benefits, and budget.
Unlike Pontis, it gives project-level MR&R for each bridge [18]. Since BRIDGIT
analyzes from project-level to network level instead of the reverse as Pontis
does, BRIDGIT runs slower for large bridge populations. Nonetheless, this
makes BRIDGIT ideal for smaller bridge populations [17]. BRIDGIT considers
replacement and "do-nothing" actions in addition to MR&R [18]. The main uses of
BRIDGIT include scheduling and tracking MR&R, keeping history of MR&R,
t
estimating the cost of MR&R and creating and maintaining a list of MR&R
actions [18]. BRIDGIT can work in parallel with Pontis to give independent
iNe
analyses and recommended actions. BRIDGIT also has the capability of using
and understanding Pontis inspection data and CoRe (Commonly Recognized)
elements. Further, BRIDGIT can convert NBI, agency defined, and Pontis
element condition ratings to BRIDGIT element condition states [18].
Life Cycle Maintenance and Management Planning System LIFECON LMS, an
European BMS, organizes "planning, construction, maintaining, repairing,
rehabilitation, and replacing structures" while considering "safety, serviceability,
economy, ecology, and other aspects of life-cycle planning [19]. LIFECON LMS
consists of three separate systems: the object-level system (also referred to as
rat
project-level system), the network level system, and the network- and object-
level system. The object-level system targets small organizations with
ownership of "limited number of concrete infrastructures", the network-level
system targets national road administrations and other organizations that are
responsible for upkeep of a large network of concrete infrastructure" and the
network- and object- level system is guided towards collaboration of object-level
work with network-level work. LIFECON LMS does not take into consideration
Du
just the "condition of the structures and the observed urgency of repair" but also
the "life-cycle costs, user costs, delay costs, minimum requirements of structural
performance, structural risks, traffic and other operational requirements,
aesthetics, environmental risks, and ecological pressures. Each need is
assigned a coefficient weight to aid in the decision making process and to
assess priority. The life-cycle and risk analyses "are performed to uncover
possible risks associated with MR&R actions" and "to study the cost-
effectiveness. The object of risk and structural analyses is to find "deficiencies
in the structural load bearing capacity or functional performance". The agency's
administration using LIFECON LMS assigns weights or level of importance
ratings to factors that correlate to the budget categories related to preventive
[19]
maintenance, repair, restoration, rehabilitation and new construction . The
factors or components of a bridge are inspected every 3 to 8 years depending
on the condition and age of the bridge. Although the observed data or state of
the components may not represent the current state of the structure a few years
later, by applying a deterioration model, present-day condition can be estimated
without re-inspection. Thus, within the framework of sustainable development
14
the manager may have wider concerns than the single safety level in the form of
resource consumption, environmental factors etc. and these wider issues then
become a part of the generalised user costs or benefits. The minimisation of
costs may be carried out either by minimising an objective function or through a
multi-objective optimisation.
At present, management of networks is more often object oriented, indicating
that the amount of money allocated to each network asset is determined by the
use of object oriented management systems (pavement management systems
(PMS), bridge management systems (BMS), etc.), working independently to
optimise the use of the money allocated. During the last ten years [20] a new
network management philosophy, Asset Management (AM) has emerged,
which puts the customer - the traffic user - into the centre, and, at the same time,
attempts to allocate the optimal socio-economic amount of money to each
t
specific asset (roads, bridges, tunnels, lighting, signs, guard rails, etc.). The
AM-system is to assist the road network administration in the process of
iNe
planning and optimizing the operation, maintenance, repair, rehabilitation and
replacement of the network and its assets (pavement, bridges, tunnels,
equipment, etc.) in the most cost-effective way in the long run, while minimizing
the consequences of traffic disturbances during road works. The AM-system
combines engineering principles with a sound business practice and economic
[20]
realities . The idea of cross-asset management is not a one-stop-shop
solution, but rather a best practice, robust methodology through which the entire
road transportation network may be maintained and operated in a safe and
efficient fashion with an emphasis on cost minimisation. The term cost does
rat
not necessarily mean the liquidity at any point of time and covers a broader
financial aspect. Some specific characteristics of cross-asset management
include high-quality information on asset inventory, the condition of such assets,
the management strategies of such assets and customer perceptions.
Additionally, unlike a standard management system of a certain type of sub-
asset (e.g. a Bridge Management System), cross-asset management explicitly
encourages cost-effective data collection, monitoring and target oriented asset
appraisal. When the complete information on an entire network of assets,
Du
including its sub-assets is not shared or the information are retained within
different clusters, the decision making may either have independent
components, (leading to sub-optimal and non-unique final results) or may be
unreasonable. The equilibrium or the minimisation, in such cases does not
consider all the stakeholders and consequently their expectations and
requirements are not reflected appropriately. Even when some information is
shared and the assessment of assets ranges wider than traditionally considered
objectives, the asset optimization may become a speculators optimisation
problem with different speculators having different requirements, expectations
and possessed information. The stakeholders can be broadly classified as
operators, users, neighbours, society, financing body and owners. The owners
naturally tend to strive for a system optimal equilibrium for which the cross-
asset optimisation may be lucrative as long as the impacts are assessed in
terms of direct investment. On the other hand, the users requirement may often
encompass intangible costs which not only includes the coast to the road user,
but also the cost of the environment, noise pollution or even the cost of comfort.
15
Such markers directly affect the perceived level of service and are thus valid
weighing parameters in a cross-asset optimisation process. The neighbours are
expected to have a mix of the previous two requirements. The societal
expectation very strongly reflects the perceived safety and the perceived level
of service of the asset as a whole. The financing bodys expectation, ideally,
forms a long term cost minimised solution where a cost prioritisation is expected
in line with the available cash flow. However, the expectations of the financing
body have increasingly acknowledged the importance of perception of the users
towards the network as a whole. The expectations are usually qualified by
quantitative and qualitative terms and such juxtaposition is acknowledged and
taken into account for cross-asset optimisation. The benefit may be viewed as a
qualitative term with multiple interpretations but in reality, within the asset
management framework it can become a strictly defined output once the
influencing variable, including an agreed definition of the value of the network is
t
accepted. The mobility is another expectation having a strong relationship with
travel time and accessibility of the users within the network. This variable
iNe
implicitly affects the weights of the other variables in this optimisation problem.
It is to be noted here that although each optimisation problem will prove certain
valid solutions the final decision is not unique and may either be multiple or
dependent on the definition and representation of qualitative variables (or both).
Consequently, the planning and conceptual design concepts remain exactly the
same while considering multiple objectives of requirements of different groups
of stakeholders. The value of the asset network comprises of a number of
tangible and intangible combinations of factors. There is no single definition of
this combination, but the perceived level of safety and service, the real cost to
rat
the network, the life-cycle-cost, the environmental cost, the impact on the
society and the economy as a whole. There seems to be a unanimous
agreement on using life cycle costs when considering the cost and the cost to
the road user has gained a significant acceptance in this aspect. The
environmental impact is mostly guided through legislations and is directly
reflected through the specific contractual conditions associated with the lifetime
maintenance management of the network. The benefit, as obtained from cross-
Du
asset management targets all different types of stakeholders and consequently

the final benefit is rarely expressed as a unique value of investment required or
saved. Rather, the benefit in terms of financial savings can be translated
contextually and related perfectly to the expectations of different stakeholders.
Such a targeted interpretation and quantification of benefit does not contradict
with the traditional idea of financial benefit since the benefit of each class of
stakeholder is represented in their specified and customised definition allowing
them the maximum control over an optimised asset of network.
It is apparent, that the guiding philosophy behind the infrastructure management
systems involves cost-optimised management of true safety. This leads to
Fig. 3, which is an annotated extension of Fig. 1 and attempts to embody the
concepts of the principles and objectives of structural assessment, definition of
limit state and budget planning a minimisation through reliability management.
The evolution of uncertainty and performances related to intervention are clear
from this figure. Reliability index tends to be the chosen metric for structural
performance since it accommodates any limit state, models all uncertainties
16
related to the system, is unique and amenable to interpolation, extrapolation or

further mathematical treatment and can compare performances of structures
with disparate limit states of performance very easily. This level of accuracy and
flexibility cannot be achieved through deterministic measures of structural
assessment where the measure are often non-unique, do not map to the true
safety of the structure and are not comparable under variable limit state
conditions. Currently, a reliability based infrastructure management can be
carried out within the remit of codes of practice (EN 1990:2002 [21]) providing
practising engineers with some legal indemnity in choosing such a full
probabilistic format. It is interesting to note in Fig. 3 how the level of uncertainty,
typically expressed as Coefficient of Variation (CoV) increases with the
deterioration of structure.
t
iNe
rat
Du
Fig. 3.Infrastructure management through reliability management.
Traditional approaches of assessing structures are based upon deterministic

and semi-probabilistic approaches. Such approaches usually consider a certain
limit state and attempt to model the actions on and the capacity of the structure
against the prescribed actions. The ratio of the prescribed actions to the
capacity is usually taken as a representative measure of safety. A value greater
than 1 indicates a non-compliance with the limit state under consideration. The
effects of uncertainties are usually catered for by appropriately overestimating
the load effects and underestimating the capacity effects based of equivalent
factors replacing the distributions related to uncertainties. The factors can be
appropriately scaled or updated to be less conservative when information based
on testing or monitoring is available. Fundamentally, a lower fractile of capacity
is compared against an upper fractile of load effects and the distance between
them is taken as a measure of safety. For a significant number of structures
such an approach is appropriate. However, it is possible that the structure is
17
indeed within limits of safety even when it is deemed unacceptable from a

deterministic or semi-probabilistic assessment. Such a situation occurs due to
significant residual conservatism within this assessment method. Since the
factors of assessment are based on codes of practice, catering to all structures
in general, the factors do tend to be somewhat more conservative.
Consequently, site-specific true measure of safety can often be much higher
than that established by the traditional approaches. The risks are two-fold.
Some structures can wrongly be designated as underperforming, ultimately
leading to the replacement of it and adding to the total investment on the
infrastructure. On the other hand, it is possible that a structure, probably on the
borderline of acceptance using a traditional approach is in fact underperforming
due some site-specific onerous effect. Under such circumstances, safety is
compromised. To summarise, traditional approaches are good for general use
but costly in case where safety cannot be documented.
t
Structures may have several limit states with very different values of reliability
iNe
indices associated to those limit states. Under such a condition, the
maintenance management may become a function of the decision of what is
intended to be achieved. Consequently, a structure violating serviceability limit
state may be important for prioritising decisions on intervention but such a
violation may not be important when the same prioritisation is being considered
for a network of significantly affected structures where the concern is the
avoidance of risk of actual collapse.
Reliability based or full probabilistic approach towards structural assessment
does not associate these risks since it considers uncertainty in its entirety and
rat
does not surrogate the uncertainty with deterministic or semi-deterministic
factors. The safety is a true reflection of the knowledge on the structures, is
unique, comparable and accurate in accepting or rejecting the performance of a
structure. The development of probabilistic and reliability-based assessment
approaches has contributed towards the establishment of more rational
assessment procedures. A number of EU funded research projects in this
regard have focused work packages in this area producing useful deliverables
Du
for infrastructure owners/managers tasked with optimising their resource.

[22]
COST 345 Procedures required for assessing highway structures ;
MEDACHS (Marine Environment Damage to Atlantic Coast Heritage
Structures) [23];
ARCHES [24];
FP6 - SUSTAINABLE BRIDGES [25];
SAMCO [26].
The COST 345 report emphasizes the benefits of appropriately applied and
interpreted results from testing (particularly non-destructive), monitoring and
condition assessment including visual inspections with representative markers.
The requirement of improved markers, training of inspectors and detailed
testing was felt. Similar points have been echoed by the MEDACHS project,
where the relationship among inspection, testing, probabilistic assessment and
repair was explored in detail within a wide group of experts consisting of
18
engineers, material scientists, end-users and policy makers. In fact, two main
conclusions of the newly published final report of ARCHES state the importance
of developing appropriate tools and procedures to avoid unnecessary
interventions (repair/replacements) in structures and faster, more cost-
effective and longer lasting rehabilitation techniques (repair and strengthening)
of sub-standard and unsafe bridges.
Another important project, FP6 Sustainable Bridges has looked directly into
assessment for future traffic demands and longer lives for bridges. Here, they
have concluded that although bridge assessment is similar to design in many
aspects in terms of basic principles, the philosophy of assessment is in fact
quite different. The conservatism in design has been singled out and
probabilistic and reliabilistic analyses including Bayesian updating has been
recommended.
t
The SAMCO report takes up this approach of reliability based infrastructure
management and attempts to address the issue in detail. The SAMCO report
iNe
presents a multi-level hierarchical assessment of structures, the philosophy of
which is embedded in almost all assessment systems. As per SAMCO report, 5
levels of assessment are defined. A Level 0 assessment is a nor-formative
qualitative assessment. Measurement based determination of load effect and
partial factor based method employing document review form Level 1 and Level
2 assessments respectively. The use of partial factors based on supplementary
investigations form a Level 3 assessment. Level 4 assessment involves
modified target reliability achieving using modification of partial factors. A full
probabilistic assessment qualifies for a Level 5.
rat
This approach can be considered to be representative and is taken up in a

modified form later in this TG. On a simpler level, a structure should be visually
inspected, deterministically assessed and re-assessed after carrying out
targeted tests before considering a full reliability based approach of assessment.
From a schematic viewpoint, it might be said that infrastructure maintenance
management is probabilistic by nature, interactive by requirement and
Du
hierarchical by resource and information limitations.

Fig. 4 presents a Hierarchical Mandala of an Interactive Infrastructure
Maintenance Management schema. As seen in Fig. 4, three main zones
inspection and testing, computational and decision making interact with each
other and within themselves to reach to an optimised decision. Often, selected
parts of this schema are used for actual decision making based on available
cost and/ or time resources.
The schema in Fig. 4 can be directly related to the bridge life-cycle as shown in
[27]
Fig. 5 and adapted from Khan . This reference also establishes the global
similarity towards the modern approach to infrastructure maintenance
management by providing a detailed and practical discussion in North American
and Asian context. The importance of model updating through testing [28],
employment of degradation and cost-benefit analysis directly to reliability based
decisions [29],[30] and the effects of load ratings on policies towards infrastructure
networks [31],[32] are regularly featuring in peer-reviewed publications and policy
documents.
19
The complexity and performance requirements of infrastructure maintenance

are fundamentally defined but evolving rapidly with the expansion of
computational and experimental boundaries within the commercial engineering
environment. Infrastructure maintenance, in the modern and evolving context
has become an interdisciplinary venture, requiring expertise in a number of
disciplines but the various disciplines are yet to fully understand or agree on a
quantified integrated risk assessment in the presence of various uncertainties.
Consequently, along with the development of appropriately identified and
agreed markers of performance [33] for a centralised maintenance management,
the formal establishment of the necessary skills and training requirements for
the assessment personnel as an employers representative is expected to gain
more importance with time.
Development of site-specific empirical relationships among the various testing
t
methods, levels of condition rating and assessment of structures are very
important for a rapid, practical but dependable decision making. Significant
works on such aspects are not yet present [31],[34]-[36], possibly due to the lack of
iNe
full-scale and realistic examples. Additionally, there are certain important
consequences of loading, like fatigue, that are yet to be taken up to their full
potential within the context of maintenance management. In terms of condition
updating, Bayesian [37] or Markov [38] based models are becoming popular while
managing risks by limiting load rating on bridges still remain a preferred option.
However, this latter solution tends not to be feasible based on modern minimum
requirements of carrying capacity of a bridge to allow the maintenance and
development of intra and inter-country trade routes. Another identified gap in
rat
successful commercial infrastructure maintenance management is the lack of
real-life examples or case studies for the practising engineers and can act as
benchmarks. There exists no other practical, technical or knowledge based
difficulty in employing a probabilistic maintenance management regime at a
network level. As an example, a probabilistic assessment on eleven bridges in
Denmark consistently increased the maximum allowable load on the bridges
thereby extending their serviceable lifetimes resulted in savings in excess of
[39]
15 million .
Du
This volume details numerical methods for assessing the safety and
serviceability of highway structures. These subjects are targeted at Engineers,
network managers and authorities to assist them in selecting the appropriate
methodology for assessing the safety and serviceability of highway structures.
In line with best practice the approach recommended here is that the level of
assessment should vary primarily according to the requirement of the structure
based on the possible consequences in future based on its as-built performance
indicators and secondarily on the degree of sophistication of the analysis, the
data available from the structure being assessed and the time available to the
Engineer. Higher levels of assessment often require measurements of material
strength and in-service loading conditions, and complex modelling and analysis.
However, a particular emphasis is given on the use of site-specific load and
resistance data.
20
t
iNe
rat
Fig. 4. Schema of infrastructure maintenance Management.
Construction
Du
Bridge
Demolition Construction Maintenance
Cycle
Rehabilitation
Fig. 5. Life-cycle of a bridge, adapted from Khan. [27]
21
As part of the work performed 3 case studies are presented at the end of this
volume:
Ferrycarrig Bridge;
Barra Bridge;
a wharf in Nantes.
The recommendations contained herein provide considerable potential for
savings in structure repair/replacement costs
The manual is divided into a number of chapters to detail the methodology
employed and focus on the specific aspects of assessment and maintenance
planning of built infrastructure. The organisation of the chapters in this volume
of the TG are summarised in this paragraph. Chapter 1 provides an executive
t
summary to this volume, emphasizing on the requirement, background and the
importance of appropriate maintenance and planning of built infrastructure.
iNe
Chapter 2, introductory chapter, chiefly deals with the broad philosophical
methodology, background and schema in assessing and managing structures
with a definite emphasis on probabilistic and reliability approaches. Chapter 3
formally introduces the various levels of assessment, current practices of
assessment and the relationship of the various forms against the true and
evolving safety of structures in the presence of uncertainty. Response models
of key infrastructure elements are presented in Chapter 4. This chapter, apart
from the standard response models, specifically discusses the conglomeration
of field data and the updating of structural model employed. Chapter 5 presents
rat
load modelling in the presence of uncertainties and acknowledges the evolution
of load patterns with time. Specifically, emphasis has been given on the
importance of extreme value statistics on the oncoming loads and possibility of
stochastically modelling the same through simulations techniques. This chapter
is followed by modelling resistance in the presence of uncertainty in Chapter 6.
The use of tests data and synthesis of such data with existing models for
updating and the handing of uncertainties and model errors are within the scope
of this chapter. The background for reliability analysis, created in Chapters 5
Du
and 6 is formally realised in Chapter 7. This chapter takes up the overview,

basic theory and methodology for a full probabilistic analysis. The concepts
presented in these chapters are illustrated and validated through case studies
from a number of partner organisations in Chapter 8. Finally, conclusions and
recommendations are presented in Chapter 9.
22
3 Structural assessment
The purpose of structural assessment is to define the condition of the structure
under consideration with respect to its ability to safely carry the loads to which it
is subjected. Fig. 6, an illustrative variation of the fundamental concepts
covered in Fig.1 and Fig. 2 elaborates on this purpose. Even at a design stage
the structure targets to be well above the minimum safety level prescribed by
current and applied codes, acknowledging that such a change in code
specification is quite natural. In fact, while designing the design loads and
resistances are kept more conservative than the target for the structure in its as-
built state. This conservatism allows for the uncertainty involved during the
actual building process. During the design and construction stage the actual
level of safety cannot be verified and the representative level of safety can be
t
considered to be equal to what is available from design. Once the structure is
built and becomes operational, under usual circumstances, the true level of
iNe
safety of the structure is increased. This happens because the actual load and
resistance combination happens to be less onerous than what is conservatively
considered during design. The discussion related to the safety levels in Fig. 6 is
in fact independent of assessment method and is absolutely general. With time,
due to the evolution in loads and the deterioration of the structure the level of
safety starts to drop away. The target level of safety is usually constant unless
there are specific changes in requirements of codes. The time from the
inception to the point where the true safety of the structure reaches the
allowable limiting level of safety determines the service life of the structure. If
rat
deterioration is more severe than what is planned, the service life of the
structure is compromised from what was posited during design. However, within
the framework of infrastructure maintenance, a reduction of safety level over
time can be pre-designed, reaching which, intervention is to be carried out on
the structure. Such a timely intervention can help improve the safety level of a
structure to an a significantly improved degree in a cost-effective fashion and
under the assumption of a certain range of further rate of degradation, the
Du
structure can extend its service life to a maximum with minimum investment.
This cost-optimised maintenance cannot take place is no intervention is carried
out until the structure reaches or is extremely close to the allowable minimum
safety limit. The same applies for making an intervention too early. It is
observed, that the success of optimisation significantly depends on an accurate
description of the evolution of damage or deterioration rate.
23
t
iNe
Fig. 6. Cost-effective infrastructure maintenance.
rat
The failure, defined is a non-compliance with one or more performance
criterion/criteria may or may not correspond to collapse or significant visual
distress. The limit of compliance may be related to the minimum acceptable
level illustrated in Fig. 6.
Visual inspections or structural monitoring (e.g., measurement of
physical/chemical properties) can often provide an indication of the degree of
Du
deterioration of a structure. Such preliminary assessments attempt to isolate

existing structures that need to be repaired. The visual inspections rely on
accurate interpretation and correlation with other indicators of visual distress.
Visual inspections can be more powerfully used in the presence of numerical
modelling results and/or results from structural testing (non-intrusive, intrusive
and semi-intrusive).
However, in many cases, even if the structure does not exhibit deterioration, it
may need to be assessed for more onerous in-service loads in which case a
numerical assessment will be undertaken to judge whether the level of risk is
acceptable or not. A simple analysis is cost-effective where it can demonstrated
that the level of risk was acceptable but, where it was unacceptable, the
Engineer should introduce more advanced methods of assessment. The
Engineer may proceed to a more elaborate assessment in anticipation of future
deterioration or increases in loading, to facilitate the bridge management
process.
24
The specific methodology and exact hierarchical nature of inspection based

maintenance management vary from country to country and are dependent on
the decision of the authorities that own the network. However, the fundamental
process seems to remain unchanged. Most of the bridges are inspected on a
regular basis. These regular inspections are visual and photographic in nature.
The interpretations are qualitative and measurements are carried out only when
required. The frequency of these inspections is the highest due to their ease
and efficiency. The next level of inspections follows a well-defined and detailed
protocol of data collection. The data is comprised of the inventory and the
assessment of the various components of the bridge in the form of discrete
ranking or continuous markers. Data collected at this stage can feed into
deterministic, semi-probabilistic and probabilistic assessments. The assessment
is chiefly visual and photographic but detailed in nature. Under certain
circumstances geometric details are noted as well. A hammer tapping survey is
t
also carried out at this stage, if required. The condition rating and ranking at this
stage are often amenable to further statistical or mathematical updating. Some
iNe
variants of inspection of this stage also include limited but invasive investigation
of the condition of the structure, especially in the case of possible corrosion. At
this stage, most assessments consider the safety and durability aspects of the
super and substructure along with possible stability conditions like evidence of
scour and undermining. Environmental factors are also taken into consideration.
Due to problems in access and the requirement of specialised training,
underwater inspections are considered separately within most bridge
management systems, although this inspection is placed at an approximately
similar hierarchical level along with the detailed inspection. The frequency of
rat
such detailed inspections is usually a function of the as-built rating of the most
current inspection on a structure. Of course, in the presence of further
information from any hierarchical level of the system, the time of revisiting the
structure can be immediately revised at any given time. Within the inspection
and maintenance hierarchy, the next level of inspection involves detailed testing
of the structure and the recording or monitoring of its mechanical, chemical and
electromagnetic properties. These tests can be intrusive, non-intrusive or semi-
Du
intrusive and are guided by the already available information from existing or
archived documents and inspections or ratings from previous or current visits to
the structure. Archived information on the analysis/assessment of the structure
and/or information of the condition of the structure at a previous time are also
invaluable in establishing a realistic evolution of the degradation of the structure.
Usually, the detailed tests attempt to establish the cause, the nature and the
extent or severity of damage in the structure. Some typical tests in this regards
are visual assessment, hammer tapping survey, cover-meter survey, Schmidt
Hammer technique (used sparingly and with caution), half-cell potential testing,
resistivity testing, chemical analyses of concrete dust to establish the
composition, chemical analysis to establish sulphate content, chemical analysis
to form the chloride profiles, carbonation test using phenolphthalein indicator,
petrographic and/or spectroscopic survey of extracted concrete cores,
compressive strength test of concrete cores and scanning electron microscopy.
All of these tests are correlated with each other and with available background
information including calculations and previous tests. Deterministic, semi-
25
probabilistic or full probabilistic assessments can be carried out independent to

this information and can be updated based on such information. The level of
involvement, expertise and the width of knowledge required in this level is
higher than routine maintenance or condition rating. These tests, along with
computational assessments (and sometimes without the assessments) can help
establishing the feasible intervention options on the structure. These
intervention options are usually considered in detail along with actual costs and
costs to the road user. It is attempted to optimise the infrastructure maintenance
on a network level through a cost-benefit analysis. The current practices and
possible new methods or markers of this concept have already been discussed
in guidelines developed by the European BRIME [40] project and SAMCO [26].
DURATINET classifies assessment in five distinct levels, numbered I to V, with
Level I being the simplest and Level V the most sophisticated, as follows:
t
Level I Assessment: Only simple analysis methods are necessary and
full partial safety factors from the assessment standards are used to
iNe
give a conservative estimate of load capacity.
Level II Assessment: More refined analysis and better structural
idealisation than in Level I is employed. Characteristic strengths for the
structural materials are determined based on existing available data.
Full partial safety factors are again used.
Level III Assessment: Unlike Level II, site-specific loading and/or
material properties are determined from new tests on the structure. Full
rat
partial safety factors are again used.
Level IV Assessment: Lower levels of assessment depend on implicit
levels of safety where reliability is based on the majority of structures of
the type concerned. On the other hand, Level IV takes into account any
additional safety characteristic of the structure being assessed by
allowing modified safety criteria determined through rigorous reliability
analysis or by judgmental changes to the partial safety factors.
Du
Level V Assessment: Reliability analysis is applied to a particular

structure. This analysis requires probabilistic data for all the variables
defined in the loading and strength equations in addition to specialist
knowledge and expertise.
The advantage of this approach is in the flexibility of using existing data
obtained at different levels of detailed and complexity along with the
independence of updating and reclassification of a structure including the
decision on it based on the as-built condition of the structure in a network level
without ensuring total compliance with safety standards at all points and all
times throughout the service life.
Fig. 7 embodies this philosophy through a schematic adopted by SAMCO. This
approach distinguishes between a qualitative and a quantitative assessment
while allowing the flexibility of comparing quantitative assessments at multiple
levels of data acquisition. It has been observed that time and resource are best
utilised by assessing a structure through each level judiciously and moving to
the next level based on the assessment conditions obtained from a previous
26
level. Some alternative levels of assessment can be considered within the

report.
There are four main formats of reliability analysis: global safety factor format,
partial safety factor format, reliability format and socio-economical formats.
3.1 Global safety factor format

The traditional method to define structural safety is through the general safety
factor format, which may be selected on the basis of experiments, practical
experience, and economic and political considerations. Global safety factor
formats were the basis for most of the former codes and standards used
throughout Europe. The general safety factor format is often associated with
t
elastic stress analysis and requires:
iNe
rat
Du
Fig. 7. Structural assessment levels. [26]
Sa = Ra = Rf / g (7)
where S is the applied stress, Ra is the allowable stress, Rf is the failure stress
of the material, g is the global safety factor.
27
Thus, the safety principle consists of verifying that the maximum stresses,
calculated in any section of any part of the structure under the worst case of
loading, remain lower than the allowable stress. The global safety factor format,
also referred to as permissible stress design, is not recommended for use in
assessment. The other three methods are reviewed in the following.
3.2 Partial safety factor format

The partial safety factor format is the basis of many codes and standards, such
as the Eurocodes, currently in use. This format is claimed to be semi-
probabilistic, considering the application of statistics and probability in the
evaluation of input data, the formulation of assessment criteria and the
determination of load and resistance factors. The semi-probabilistic partial
t
safety factor format can be described by the following formal limit state:
iNe
Sk S =Rk / R (8)
where Rk is the resistance, Sk is the stress, S, R are partial safety factors. The
reliability of a given structure is ensured by certain requirements for the limit
state, the characteristic values and the partial safety factors. Partial safety
factors are designed to cover a large number of uncertainties and may therefore
not be very representative for evaluating the reliability of a particular structure.
Partial safety factors should be calibrated using probabilistic methods and
idealised reliability formats, but in most of the countries where semi-probabilistic
rat
codes are used, the actual values for the partial safety factors are still
influenced by experience and economic and political considerations. Generically,
the partial factors attempt to describe the uncertainties related to the material
characteristics, load characteristics and model error or human error by
specifying conservative scaling coefficients as single numbers as applied to the
nominal or experimental material strengths and oncoming loads. The factors of
model error or human error are usually present as a multiplicative term along
with the material or load factors. The value of thr factors change dependent on
Du
the importance of the consequences of failure, level of control during the

manufacture of materials, experimental evidence of the condition of the
materials in their as-built condition and calibrated site-specific descriptions of
loading conditions in the current state of the structure.
Partial safety factors should reflect the knowledge of the uncertain parameters
at the specified level of assessment, allowing for factors such as the quality of
inspection, the extent and quality of on-site measurements, potential failure
modes and possible consequences of failure. They can be calibrated using
probabilistic methods and idealised reliability formats, although, in most of the
countries where a semi-probabilistic approach is applied, experiences as well
as economic and political considerations have an influence on the values used.
The partial safety factor format is the core of any modern design code and it is
strongly recommended in this manual for a general code of assessment of
existing structures and for assessment up to Level IV.
28
3.3 Reliability formats

Using reliability formats the stress S applied and the resistance R describing the
strength of the structural element are described by stochastic variables because
their values are not perfectly known. If the verification of the criterion related to
the limit-state results in the inequality S R, the structure is considered safe.
The difference, R - S, is called the safety margin M.
The safety margin M distinguishes three states:
the safe state or safety domain with M > 0;
the limit state with M = 0;
the unsafe state or failure domain with M < 0.
The probability of failure Pf related to R S characterises the reliability level of a
t
structure with regard to the limit state considered:
Pf = P (R-S 0) = P (M 0)
iNe (9)
Reliability methods taking into consideration the uncertainties of variables are

the main criteria for a realistic safety assessment. Thus, reliability formats using
probabilistic methods are an important alternative to semi-probabilistic
approaches.
The clear advantage of employing such an approach in the assessment of
structures is that it facilitates the derivation of a structure specific safety rating
rat
taking into account site specific measurements of loading and resistance

variables. In this way, it represents a significant departure from the safety factor
approach where safety factors are derived in a general sense to be applicable
to a large range of structures.
Applied procedures in Europe, USA and Canada are mainly based on the
existing design codes (levels 1 and 2) and only a few countries have specific
codes for bridge assessment.
Du
Partial safety factors are designed to cover a large number of uncertainties and
may therefore not be representative for evaluating the reliability of a particular
structure. The reliability format is also based on Limit States, but unlike the
semi-probabilistic partial safety factor format, it requires the calculation of the
probability of failure with a specified reference period. This calculation involves
the identification of all variables influencing the limit-state criteria, the statistical
description of these variables, the derivation of the probability density and its
moments for each basic variable, the calculation of the probability that the limit-
state criterion is not satisfied, and the comparison of the calculated probability
with a target level. Load demand, S, and capacity to resist load, R, are modelled
as random variables in the formulation of the Limit State. The difference
between demand and capacity is known as the safety margin, M = (R-S). The
safety margin is Normally distributed when the stress, R, and strength, S, are
Normally distributed. The reliability index, , is given by the ratio of the mean to
the standard deviation of the safety margin. In many cases alternative
29
distributions may be found to be appropriate to model the load and resistance

variables.
The determination of the probability of failure for the Limit State is a difficult task,
except for linear Limit States and Gaussian variables. The reliability index
methods (e.g., First Order Reliability Method and Second Order Reliability
Method) and the simulation methods (e.g., Monte-Carlo sampling and
Importance or Directional sampling) are two techniques which allow the
calculation of probability of failure for complicated functions. In most of the
cases, appropriate software tools are necessary for structural reliability analysis.
A structure that can be proven to have a reliability index higher than the
corresponding minimum value, can be considered to be sufficiently safe. It is
strongly recommended to use code calibration (partial safety factor format) to
base Levels I to IV assessment methods on a clear and documented reliability
t
format. Thus, preliminary investigation at lower levels will open the way for a full
reliability analysis at Level V, in cases where the results of a partial safety factor
iNe
format are believed to be too conservative.
3.4 Socio-economic format

Socio-economical formats are reliability formats where failure costs are
introduced to determine the required probabilities of failure or reliability indices.
This format can be based on decision-theory methods or on life cycle cost
rat
methods. Thus, for example, the assessment procedure can take into
consideration the fact that the failure of a retaining wall adjacent to a minor road
has much lower consequences than the failure of a major bridge.
The methodology for computation of reliability index is presented in Chapter 7 in
details. This manual focuses on a probabilistic assessment and other partial or
deterministic descriptions are not considered.
Du
30
4 Structural response modelling

The assessment of an infrastructural element requires the calculation of the
response of a mathematical model of the structure to a complete range of
loading conditions. The method of analysis to be used will depend on the
behaviour of the structural material, structural geometry and boundary
conditions, and nature of the applied load.
4.1 Methods of structural analysis

Separate or interdependent mathematical models of the structure and the soil
can be established to determine the structural response. Hence, a particular
t
model for a structure will be influenced by the assumptions adopted for the
foundation and the soil. If it is ensured that the ground can sustain the loading
iNe
with acceptable displacements or provide adequate stiffness, soil-structure
interaction can be ignored in low-level studies (in bridges, piled foundations
have often been employed to provide relatively rigid foundations and allow an
analysis of the structure in isolation). Reports such as COST 345 review a
number of available techniques to model the structural response according to
Ultimate and Serviceability Limit States. Perspectives are given on a reliability-
based design/assessment approach and on empirical, algebraic and numerical
(e.g., finite element) methods of analysis, the number of dimensions of the
rat
structural model (frame and spatial analysis), the behaviour of the structural
material (elastic or plastic), the magnitude of the displacements with respect to
the original geometry (linear or non-linear), the characteristics of the section
(cracked or uncracked reinforced concrete), the nature of the applied load
(static, dynamic, impact, fatigue) and the definition of the structure (in
deterministic or probabilistic terms).
An assessment at Level I is carried out with traditional methods of structural
Du
analysis while assessment at higher levels involves more refined methods of

analysis. Traditional methods of structural analysis are based on one- or two-
dimensional models with elastic materials, geometric linearity and static loads.
Other more rigorous techniques allow for three-dimensional modelling, a variety
of non-linear response actions and dynamics. The number of variables involved
in the modelling process increases with the level of assessment. Ideally in
higher levels of assessment, the method of analysis should take account of all
the significant aspects of the structural response to loads and imposed
displacements. The same methods of analysis can be used for Level II and
higher levels, but specific material properties and loading can be included in
higher levels. All categories are summarised for bridges and culverts in Table 2.
An assessment associated with complex mathematical modelling without field
validation, should be used with considerable caution.
31
4.2 Integration of field data and structural models

The original structure might have been altered not only due to ageing and the
application of loads, but also to grouting, saddling, guniteing or post-tensioning
in maintenance and for upgrading programmes. It is always necessary to carry
out a visual inspection of the structure. This inspection might reveal scouring of
piers and/or abutment supports, cracks in a section of the structure, the quality
and condition of the structural material, deformations of the profile, condition of
the joints, damping devices, etc. The approach used, and the input values, to
calculations can vary as a result of observation. Additionally, a number of
reduction factors relating to the condition of the bridge can be defined according
to observation. In the case of significant deformations, the cause should be
determined and the new profile may need to be considered in the analysis.
t
Table 2. Levels of assessment for various types of structures.
iNe
Level of Assessment
Structure Type
1 2 3 4 5
1-, 2- or 3-D
FEM analysis of specific details of the structure being assessed not considered in previous levels
linear or non-
linear;
Not skew
elastic or
Beam plastic;
1-D or 2-D allowing for 2- or 3-D;
linear elastic cracking linear or non-linear;
Not skew (beam theory or
rat
elastic or plastic;
Slab plane frame grillage or FEM
analysis) 2- or 3-D
Not skew (upstand model if
Reliability analysis based on probabilistic models

linear or non-
Beam & necessary); allowing
linear;
Slab for soil-structure
elastic or
interaction, cracking,
Not skew plastic;
and site-specific live
Cellular allowing for
loading & material
cracking;
properties
1-D or 2-D grillage or
Skew,
Du
simple grillage, FEM (upstand

Bridges
tapered linear elastic model if

and curved allowing for necessary)
torsion
2- or 3-D
linear or non-
Empirical or 2-D linear;
2- or 3-D;
Arch linear elastic elastic or
linear or non-linear;
arch frame plastic;
elastic or plastic;
allowing for
grillage or FEM
cracking
(upstand model if
2- or 3-D
necessary); allowing
linear or non-
2-D linear for soil-structure
linear;
elastic with interaction, cracking,
elastic or
Cable modified and site-specific live
plastic;
Stayed modulus of loading & material
modelling
elasticity for the properties
cable sag
cables
more
accurately
32
Table 2 (cont.). Levels of assessment for various types of structures.

Level of Assessment
Structure Type
1 2 3 4 5
Frame linear 2- or 3-D FEM,
Rigid
FEM analysis of specific details of the structure being assessed not

elastic 2- or 3-D FEM linear or non-
Frame linear linear or non-linear; linear; elastic or
Culverts
elastic elastic or plastic; plastic; allowing for

allowing for allowing for soil- soil-structure
Reliability analysis based on probabilistic models

Flexible soil-structure structure interaction,
interaction interaction, cracking and site-
(beam & cracking specific loading &
spring) material properties
considered in previous levels

3-D non-linear
Beam, 2- or 3-D
Simple FEM, allowing for
Wharves
Earth- non-linear FEM on

equilibrium soil constitutive
retaining elastic foundation
t
method of models and site-
walls or elasto-plastic
analysis specific loading &
continuum
material properties
iNe 2- or 3-D FEM of
Empirical soil in combination
models or 2- or 3-D FEM of with existing
Reinforced soil 1-D linear soil structure and site-
elastic specific loading &
material properties
Empirical 3-D non-linear
models or FEM with bedding,
2- or 3-D FEM;
beam-and- fracture planes,
Tunnels spring models
linear or non-linear;
and site-specific
elasto-plastic
rat
(non-cohesive loading & material

soil) properties
The assessment of a structure might require a better idealisation of the

structural response than one based on the observation of the visible portion of
the structure. However, complex analytical tools can only be justified if a
realistic assessment of the material properties and overall condition of the
existing structure can be made through accurate field measurements. Tests on
Du
concrete include cover depth, rebound hammer, ultrasonics, impact echo,

permeability, carbonation, thermography, radar, slot cutting, instrumented
coring. Clearly, only some of these tests will serve to improve the assessment
of load carrying capacity. Testing for reinforcement corrosion includes the
measurement of half-cell potentials, resistivity, chloride concentration [41] and
monitoring. Post-tensioning tendons can be tested with exploratory hole drilling,
radiography, ultrasonics or through on site monitoring. Other tests are related to
the determination of in situ stress [42]-[44] or monitoring while repair [45].
Load testing can be used to improve the reliability of structural models for the
Serviceability Limit State through measurement of static/dynamic effects and
other performance measures, including the generation of cracks and the
distribution of load. Such testing must be carried out with caution so as not to
inflict damage to the structure. Thus, the passage of heavily loaded trucks can
be used to determine the in-situ live-load behaviour of the structure and, by
extrapolation, to predict the maximum stresses due to the traffic load. Typically,
forced vibration or ambient vibration methods [46],[47] are used to determine the
33
frequencies and mode shapes of vibration of a bridge. As tests at full scale are
expensive and limited [48],[49] scaled physical models using measurements from
tests on the real structure, can also be used for assessment purposes [50],[51].
Measurements can provide more realistic values for support stiffness, joint
condition, restraints, behaviour of the cross-section, elastic properties of the
structural material, behaviour of the foundation, fill and structural material
density and road profile among many other aspects. These characteristics can
then be incorporated into the structural model. Optimisation techniques are
commonly used for adjusting parameters of the structural models to fit with field
measurements. The updated models can be used to more accurately predict
and assess the behaviour of the structure under different static or dynamic
loading conditions. In a structural reliability model, the uncertainties in the
design parameters will be modelled probabilistically. The process of identifying
t
the behaviour of a given structure is described by the ASCE Committee on
structural identification of constructed facilities [52]. It is important to note that
iNe
variabilities and uncertainties are present in inspection and testing and such
effects may have to be taken into account [53]-[55].
rat
Du
34
5 Load modelling
An existing bridge has to carry the same type of loads as a new one unless very
specific and new forms of actions or risks (e.g. an explosion) are considered.
From an assessment point of view, these loads can be classified into time
invariant (e.g., dead and superimposed dead load) and time variant (e.g., traffic,
wind, earthquake and temperature loading). Some other loads can be initially
time variant, but asymptotically time invariant at some point (e.g., differential
settlement, earth pressures, creep and shrinkage effects, etc.). From a generic
viewpoint, these can be interpreted as actions on the structure.
5.1 Time invariant loads
t
Time invariant loads are those that may reasonably be expected to remain the
iNe
same for the lifetime of the structure. Compared to the design stage, at
assessment stage these loads may have altered due to the effects of the
construction process and subsequent life of the structure. As a result,
calculations can often be assisted by on-site measurements which allow them
to be determined more accurately (e.g., measurement of the actual thickness of
the asphalt layer). As a result, the load and resistance models can be updated
whilst maintaining the required safety for the structure. Measurement of density
and element size may also justify the use of a lower value of a particular partial
rat
factor for assessment than adopted for design. Additionally, these loads can
also be appropriately used during intervention to improve the reliability of the
structure. For example, a deck improvement might be associated with a
compliant asphalt thickness which is less deep than the existing layer.
5.2 Time variant loads

Du
Time variant loads are those that can be modelled as stochastic variables.
When assessing a structure, representative records of traffic, wind, earthquake
and tide (if relevant) and temperature records may be available. Hence, the
characteristic load effects (e.g., values of bending moment, shear force, etc.
with specified probability of exceedance), may be predicted more accurately.
For time variant loads, establishment of an appropriate return period of a certain
severity of the action of the load is critical. The stochastic approach, in
combination with monitoring records requires modification where effects of
climate change are to be taken into account.
5.2.1 Statistical modelling

Time variant live loads, such as traffic, wind and temperature effects represent
random phenomena and as such, require statistical modelling to determine the
magnitude of their characteristic effects. The levels of extreme values of these
loads are associated with a return period against a certain intensity taking place
35
with a given degree of confidence in terms of statistical significance. This

degree of statistical significance and consequently the return period or the
intensity of the action at a given intensity vary when climate change is taken into
account independent of the fact whether the roots of such change are
anthropogenic or not.
Consequently, it becomes important to establish the description of such load in
terms of extreme value distributions and concentrate on the statistical
description of the tails, or the extremes of such distributions. Typically, the
Gumbel family of distributions (Gumbel, Frechet and Weibull) are fitted to
measured data, recorded over a period of time. Some amount of post-
processing on the descriptors of the distribution can be linked to engineering
judement as well. Subsequent extrapolation of these fitted distributions for a
specified return period yields the characteristic value to be used for time variant
t
loads. The difference in distributions contained within the Gumbel family lies in
their tail behaviour, e.g., whether they are bounded or unbounded in the
iNe
extreme. In modelling measured data for prediction of characteristic extremes,
care should be taken to ensure that the most appropriate distribution is selected.
[56],[57]
The principle of tail equivalence is employed in determining an
appropriate extreme value distribution. The extreme value and parent
distributions, G(x) and F(x) respectively, are considered tail equivalent if:
lim 1 (10)
Where the extreme value distribution G(x) is modelled by either the Gumbel I or
rat
Weibull distribution, given by Equations 11 and 12 respectively.
, 0 (11)
(12)
The threshold, and scaling parameters, and of the Gumbel and Weibull
Du
law, G(x), are estimated by the maximum likelihood approach [58].

The suitability of either the Gumbel or Weibull distribution is assessed by
plotting the extreme data on probability paper where linearity indicates the
appropriateness of the mathematical model [56],[59],[60]. Standard transformation
functions are employed to determine relative plotting positions for the recorded
data on the probability paper. The degree of linearity of the plotted data and the
closeness of fit of the chosen distribution to the data, reflects the accuracy of
the distribution. For example, Fig. 8 shows the probability paper plots for the Gumbel and
Weibull extreme value distributions respectively, for bending moment at the centre of a
single span, two-lane, 20 m long bridge. In the figure the parameters and are the
limit and dimensionless scaling parameters respectively relating to the distributions.
Differences in the mathematical formulation of the Gumbel and Weibull distributions are
apparent in the relationship between the transformed data, plotted on probability paper,
and the best fit extreme value distribution. Care should be taken in selecting the
appropriate distribution for extrapolation.
36
6 7
5 6
4 5
3
4
2
3
1
2
0
1
-1
-2 0
4000 4500 5000 5500 6000 4500 5000 5500 6000 6500
kN m)
kNm)
t
Fig. 8. Extreme value approximations for simply supported bridge and
iNe
bending moment at centre of span (total span 20 M). Left: Gumbel
I. Right: Weibull.
For short span lengths e.g., less than 20 m the Weibull distribution is
considered more appropriate as it implicitly recognises a physical upper limit for
the maximum load effect on short spans as a function of the maximum possible
axle and group of axle load [61],[62]. For longer spans e.g., in excess of 20 m span it is
found that either distribution is appropriate. However for medium to long span bridges e.g.,
rat
in excess of 50 m, a convex trend is observed in the right hand tail of the Weibull
distribution. As the tail region is of prime importance in extrapolation, the Gumbel
distribution, which is unbounded in the extreme, is more appropriate.
It is also important to consider the characteristic load effect being determined. It is
found that the choice of either extreme value distribution is dependent not only
upon the span length but also upon the load effect being considered [63].
Du
5.2.2 Load monitoring data required for assessment

The duration of the measurements used for extreme value modelling depends
upon the effect being determined. For wind and temperature data, maximum
and minimum values of the particular effect over a representative period of time
(e.g., 50 years) and for a specified sampling frequency (e.g., monthly) should be
collected. Traffic Weigh-in-Motion (WIM) data can be obtained by mounting
sensors in the road pavement or on an existing bridge structure and estimating
the corresponding static loads using appropriate algorithms. There is
considerable variation in the distributions of Gross Vehicle Weight (GVW)
between sites, as can be seen in Fig. 9. It is therefore important to collect data
appropriate to the bridge being assessed.
37
0.008
0.007
A1
0.006
A2
0.005 A31
fx(x)
0.004 A6
0.003
0.002
0.001
t
0 100 200 300 400 500 600 700 800
GVW [kN]
iNe
Fig. 9. Comparison of gross vehicle weight distributions at various sites on
the French motorway network.
It is clearly desirable to collect as much data as possible, but one or two weeks
of continuously recorded data may be sufficient for the purposes of assessment
[64]
. It is important to attempt to ensure that this data is representative and so, in
scheduling a measurement period, consideration should be given to seasonal
rat
variation patterns.
References do not specify the required accuracy of WIM data. However, some
guidance is given by Jacob [65]. These authors specify the required accuracy
with reference to the COST 323 WIM specification [66] Bridge loading is not
sensitive to WIM system accuracy and a system with accuracy as low as C(15)
is deemed to be sufficient. This corresponds to about 95 % of gross vehicle
weights (the exact percentage depends on test conditions) being within 15 % of
the exact static value [67].
Du
5.2.3 Assessment of characteristic traffic load effects

Like time variant models, the traffic loads operate under extreme value
distributions whose tails should be faithfully simulated or predicted.
Consequently, either a direct simulation of traffic-structure interaction or an
indirect description of the extreme value statistics of the load effect is generated.
Within design or assessment codes, values against certain characteristic points
on the extreme value distributions are provided with the possibility of updating
such information in the light of relevant available information
5.2.4 Direct simulation

The characteristics of the vehicles that traverse a bridge structure vary widely
with respect to their gross weight, axle spacing, distribution of load to axles,
location in lane, velocity and in the likelihood of multiple presence of vehicles on
38
the structure both longitudinally and transversely. In Levels I and II assessment,

extremes of Normal traffic are represented by notional load models. Some
countries have assessment standards with notional load models that are
generally less onerous than those specified for new structures. In countries
which do not have specific standards for assessment, the notional load model
for new structures may be used. Such codes are generally conservative in the
interests of simplicity and given the cost insensitivity of new structures to
loading. Assessment load models can be developed in European countries
where the Eurocode has been calibrated for national conditions and where the
data used for that process are available. It has been established in such studies
that the notional load model is considerably more conservative for some load
[68]
effects and some road classes than others , a phenomenon that could be
exploited to provide reduction factors for the Eurocode model.
t
For Levels III to V assessment, site-specific load modelling is allowed. A
number of approaches can be used to convert basic WIM statistics into
estimates of characteristic load effects [69]. This approach has the advantage
iNe
that few assumptions are necessary about the nature of the data. It is applicable
to short span bridges but, given the relative paucity of data in the tail of the truck
GVW distribution, it requires a considerable amount of data to give repeatable
results. The direct simulation approach is unlikely to include enough examples
of traffic congestion to accurately simulate the loading events that govern long-
span bridges.
An alternative to direct simulation is to form a histogram of GVW and to fit a
distribution to this. The distribution is subsequently used with Monte Carlo
rat
simulation, to generate "typical" weights for vehicles in a simulated traffic

stream. The histogram can be used directly to generate weights but this, like the
direct approach described above, suffers from a paucity of data in the tail region.
Many authors [70] have fitted bi- or tri-Normal statistical distributions to GVW
histogram; the histograms of Fig. 9, for example, fit well to a bi-Normal
distribution. This is easy to implement but Getachew and OBrien [71] have
suggested that a good fit to the distribution overall can result in a poor fit to the
Du
tail region which has significant consequences. They recommend a combination

of direct simulation from the histogram where there is sufficient data and the
fitting of a Normal distribution to the data in the tail region. For short-span
bridges, arch bridges and culverts, the weight of the individual axle or axle
group can be more important than the gross vehicle weight and these needs to
be represented accurately in simulations.
A major shortcoming of existing codes lies in the manner in which they have
taken account of the dynamic amplification of the static effect resulting from the
interaction between the moving vehicle and the structure. In the Eurocode,
dynamic multipliers are applied to characteristic extremes (e.g., load effect
values with a specified probability of exceedance) determined from simulation.
Such an approach ignores the combination of probabilities from two
independent phenomena, the meeting of heavy vehicles on a bridge and the
dynamic interaction between the vehicles and the bridge. As the worst static
and dynamic effects do not generally occur for the same loading case, site-
specific maximum design load effects determined from measured traffic data
39
and experimental bridge dynamic characteristics could lead to significant

savings when assessing a structure.
Additionally, the structural resistance models can be updated through an
accumulation of knowledge of the loads that the structure has successfully
carried to date. If a bridge has survived a number of years of service, its
resistance is higher than any of the prior imposed loads. The improvement of
bridge reliability with proven service can be taken into account in a Level IV
assessment. However, this can be difficult to achieve in practice without a
comprehensive knowledge of the load history of the structure.
5.2.5 Poisson models

Typically the critical loading events for bridges with influence length up to
t
approximately 50 m in the extreme occur due to (1) meeting events between
ordinary trucks and (2) meeting events involving heavy transports with ordinary
iNe
trucks. In both cases the extreme distribution function of the load effects can be
obtained from the so-called thinned Poisson process [72] e.g. only arrival and
meeting events including the heaviest groups of trucks in the various traffic
situations are considered. Fig. 10 indicates a typical meeting event.
For a two-lane bridge the extreme distribution Fmax of the considered load effect
q can be obtained from:

(13)
rat
L o ad ed
V eh icle
L an e 2
B rid g e
Du
L o ad ed
V eh icle
L an e 2 L an e 1 L an e 1
Fig. 10. Typical meeting event - normal passage situation, heavy transport
+ standard vehicle.
Where 1 and 2 are the intensities of the considered traffic in lane 1 and 2,
respectively. 12 is the intensity of the meeting, e.g. of the considered traffic in
both lanes at the same time. T is the considered reference period for the
extreme distribution (one traffic year). Further the distribution for the load effect
in lane 1, F1(q) lane 2, F2(q), and the distribution of load effects due to
simultaneous traffic load in both lanes, F12(q), must be determined. These three
distributions do in general include modelling of:
40
the number, configuration and weight of trucks;

the longitudinal and transverse appearance in bridge lanes;
the dynamic amplification of the static truck load.
It is extremely useful to generate structure specific live loading where extreme
load events can be correlated to their return periods since the design or
assessment codes are generic and tend to be conservative in estimating the
extreme loads. Structure specific live loading can be obtained through
experiments involving technologies like Weigh In Motion (WIM) [73] or can be
simulated based on information obtained from similar bridges (in terms of
geographic location, geometry etc.). The extreme loadings can then be
accurately estimated against the required design life of the structure. Individual
scenarios, typically useful for long span bridges can also be handled through
advanced micro simulation techniques [74]. Many long span bridges would be
t
exceptionally difficult and sometimes impossible to design using traditional
iNe
generalised design codes. Reliabilistic design in conjunction with advanced live
loading techniques can ensure optimal structural performance, without
compromising the required level of safety as prescribed by codes, under these
circumstances [75].
The extreme value of traffic loading considered for design is usually more
onerous than extreme value distribution considered for assessment. Extremely
abnormal loads, similar to HB loading within the British Code (BS 5400-4) is
usually not considered during assessment stage (BD 44/95 [76], BD 21/01 [77])
rat
(as opposed to an HA load in the same code). Additionally, under assessment

conditions, particular effects like collision, longitudinal effects or effects of
temperature are often not taken into account unless explicitly requested to. The
condition of the road and the Annual Average Daily Traffic (AADT) tend to affect
the extreme values as well. On the other hand, in the recent times,
exceptionally abnormally loaded vehicle might have to pass a certain bridge
typically known as Special Types General Order (STGO) and Special Order
(SO) Vehicles. Assessments are required for such situations. A typical example
Du
can be the crossing of a 234 ton truck carrying a transformer. These vehicles
cross a bridge under very controlled condition of path and speed and are
escorted appropriately by the police or other authorities. Often, the axle loads of
these trucks can be adjusted and ensured by hydraulic mechanisms.
41
6 Resistance modelling
6.1 Uncertainty in material properties

Material properties vary both spatially and temporally. There are spatial
variations due to the different combinations of material components within each
location (e.g., different combinations of aggregate, cement and water in the
case of concrete) and temporal variations due to the loading and physical
processes in the materials (e.g., hydration process in concrete that increase
stiffness and strength). The difference between the material properties of the
test specimens and of that forming the structure must be considered both in
terms of value and in terms of uncertainty of measurement. Table 3 [78],[79] gives
t
examples (from buildings) of systematic (bias) and random (coefficient of
variation) uncertainties found in some engineering material properties.
iNe
In fact, similar descriptions of uncertainty have also been considered for soil as
a structural material. Uncertainties in the form of Table [80] can be extended to
further properties and materials. Uncertainty in soil can be pronounced due to
high spatial variability and due to time-dependent uncertainties like scour.
Additionally, it is mentioned that loads and materials are treated separately
when the soil-structure interaction requires specific attention [81].
Statistical descriptions of material strength through data have been shown to be
rat
achieved for quite some time [82]. Weighing factors have also been specified for
loads on structural members [83]. Probabilistic framework, in terms of load
factors is available for geotechnical engineering [84]. Other works have
successfully incorporated uncertainties in loads [85] and resistance for a wide
range of response [86],[87]. It might be concluded that there are essentially no
realistic technical, computational or knowledge barriers to preclude full
probabilistic assessments within a commercial consulting environment. Even
then, the lack of availability of validated datasets, which may be used for
Du
generic problems or specific cases remains to be one of the challenges.

In terms of concrete it is observed that properties related to permeability limits,
limits on diffusion rates etc. are difficult to be set on a country to country basis.
Durability parameters of steel structures have also been emphasized. In this
regard the use of highly durable materials like stainless steel rebars has been
emphasized. The applications of green materials like green concrete for repair
or recycled aggregates for repair have been encouraged.
Note that the condition assessment from on site measurements is still a
challenge in terms of optimization. It is costly and in some cases the result is
imperfect (non exact). Errors in decision can then occur. In terms of
consequences, unnecessary repair or failures are planned. All this can be
introduced in the risk analysis presented at the beginning of this manual. The
imperfect inspection is then characterized with the so called Probability of
Detection (PoD) and Probability of False Alarm (PFA). This global method is
called Risk Based Inspection.
42
Table 3.Systematic and random uncertainties of engineering materials.

Quality of Construction
Average (Cast in-place) Excellent (Factory precast)
Coefficient
Standard Coefficient Standard
Property Mean Mean of
Deviation of Variation Deviation
Variation
Concrete in
structure loaded to
failure
Compressive
Strength (psi)
fc=3000 psi 2768 487 2768 375
fc=4000 psi 3388 595 0.175 3388 460 0.135
fc=5000 psi 4013 705 4013 544
Modulus of rupture
t
(psi)
fc=3000 psi 402 88 402 85
fc=4000 psi 445 97 0.220 445 94 0.210
fc=5000 psi 485
iNe
106 485 102
Splitting Strength
(psi)
fc=3000 psi 306 54 306 52
fc=4000 psi 339 60 0.175 339 58 0.170
fc=5000 psi 369 65 369 63
Modulus of
Elasticity (ksi)
fc=3000 psi 2956 351 2956 310
fc=4000 psi 3260 387 0.120 3260 342 0.105
rat
fc=5000 psi 3537 420 3537 371

Static Yield
Strength (ksi)
Grade 60 66.8 6.2 0.093 66.8 6.2 0.093
Grade 40 44.8 4.8 0.107 44.8 4.8 0.107
Modulus of
29000 957 0.033 29000 957 0.033
Elasticity (ksi)
Effect of discrete
1.01 0.04 0.040 1.01 0.04 0.040
bar sizes
Prestressing Steel,
Du
Grade 270
Ultimate Tensile
281 7 0.025 281 7 0.025
Strength (ksi)
Ultimate Tensile
0.05 0.0035 0.070 0.05 0.0035 0.070
Strain
Modulus of
28400 568 0.020 28400 568 0.020
Elasticity (ksi)
Stress at Transfer
(ksi)
Pretensioned
189 2.8 0.015 189 2.8 0.015
Beams
Post-Tensioned
189 3.8 0.020 189 3.8 0.020
Beams
Post Transfer
Losses (ksi)
Stress-Relieved
36 5.8 0.160 36 5.8 0.160
Strands
Stabilized Strands 26.5 5.3 0.200 26.5 5.3 0.200
43
Table 3 (cont.). Systematic and random uncertainties of engineering materials.

Quality of Construction
Average (Cast in-place) Excellent (Factory precast)
Coefficient
Standard Coefficient Standard
Property Mean Mean of
Deviation of Variation Deviation
Variation
Deviation of Beam
Dimensions from
Nominal Values
Flange Width (in.) - - +0.16 0.25
Stem Width (in.) +0.09 0.19 0 0.19
Flange Depth (in.) +0.03 0.47 0 0.19
Overall Depth (in.) -0.12 0.25 +0.12 0.16
Depth of bottom
reinforcing steel -0.19 0.50 +0.12 0.34
(in.)
t
Depth of bottom
prestressing steel -0.19 0.25 +0.12 0.17
(in.)
iNe
Stirrup Spacing
0 0.53 0 0.27
(in.)
Beam spacing and
0 0.69 0 0.34
span (in.)
Deviation of
Column
Dimensions from
Nominal Values
Overall width and
+0.06 0.25 - - -
thickness (in.)
Concrete Cover for
18x18 in column
rat
(in.)
Exterior Bars +0.32 0.17 - - -
Interior Bars +0.04 0.79 - - -
Notes: Yield strength of reinforcing bars was represented by beta distribution, whereas effect of discrete bar
sizes was assumed to follow a modified lognormal distribution with a modification constant of 0.91. All other
variables were assumed to follow normal distributions.
1psi=6895 Pa; 1ksi=6.895 MPa; 1in =25.4 mm
Table 4. Uncertainties in soil as an engineering material.

Du
Standard
Parameter Distribution Mean
Deviation
Foundation E Lognormal 2*107 kN/m2 0.2 * Mean
2
G Lognormal 54450 kN/m 0.3 * Mean
Soil, Layer 1 2
max Lognormal 33 kN/m 0.25 * Mean
2
Soil, Layer 2 2
2
Soil, Layer 3 2
2
Soil, Layer 4 2
44
6.2 Modelling of materials in assessment

Partial safety factors from assessment standards (lower than at the design
stage) can be used for Level I, but characteristic strengths for materials can be
based on existing data (e. g., mill certificates and results of in situ testing from
the same or a similar structure) for Level II with allowance for the differences
between, for example, cores and characteristic cube strengths. Information from
load tests can be used for Level III or higher. Level IV uses modified partial
safety factors to account for any additional safety characteristics specific to the
structure being assessed and Level V uses structural reliability analysis instead
of partial safety factors. In the undeteriorated state the material parameters,
originally probabilistic distributions, are often related to a low underlying
uncertainty. In the case of deterioration, uncertainty at various levels enters the
modelling of materials. Apart from a significant change in the representative
t
value or the distribution itself the underlying uncertainty generally increases as
well with degradation unless identified and quantified clearly through testing.
iNe
Uncertainty related to the model of damage is also a challenge in modelling in
this regard. It is not possible to completely predict the future deterioration and
often conservative estimates are made when considering degradation for the
future. Such estimations can be revisited in the light of new information
available through modelling, testing or monitoring.
6.3 Deterioration models

rat
Deterioration models describe the deterioration and change in resistance and

behaviour of the bridge and its elements, which leads to reduced performance
and, ultimately, to the reduced reliability of a whole structure. Deterioration
models are used to predict the change in structural parameters due to the
expected structural loading, environmental conditions and maintenance
practices. Thus, deterioration models are fundamental in the development of
prediction models for bridge performance or condition. Average repair intervals
Du
may be determined on the basis of the average lifespan of individual bridge

components.
The deterioration process exhibits the complex phenomena of physical and
chemical changes that occur in different bridge components. What makes the
problem more complicated is that each element has its own and unique
deterioration rate [88]. Accurately predicting the rate of deterioration for each
bridge element is, therefore, crucial to the success for optimised maintenance
management. In the late 1980s, deterioration models for bridge components
were introduced in order to predict the future condition of infrastructure assets
as a function of their expected service condition [89]. Deterioration models in
Infrastructure Management Systems (IMS) were first developed for Pavement
Management Systems (PMS) [90]. Deterioration models in PMS differ from those
in BMS because of the differences in construction materials, structural
functionality, and the types of loads carried. In addition, safety is more important
in bridges than in pavements. Despite of the dissimilarities in the deterioration
models for pavement and bridges, the approaches to developing pavement
45
deterioration models for PMS have been employed in the development of bridge
deterioration models in BMS. In a study conducted at the transportation
systems centre (TSC), Busa et al [91] examined the factors affecting the
deterioration of a bridges condition. The study concluded that the top ranking
factors that affect deterioration include age, average daily traffic, the
environment, the bridge design parameters and the quality of the construction
and materials used. A realistic and useful deterioration model needs to use
information concerning the design, construction, materials, climate, environment,
and history of the MR&R of the bridge. Two types of data would need to be
accounted for: static and dynamic data. The static data includes relatively
constant information such as the type of the bridge, the length of the bridge, the
bridge materials, the traffic loads, the climate, the environment, construction of
the bridge, and any anomalies (for example the bridge is constructed on a
skew). The dynamic data information includes significantly changing over time
t
such as the inspection ratings and the history of MR&R of the bridge.
[92]
iNe
According to the FHWAs Bridge Management System report , most studies
of deterioration rates tend to predict slower declines in condition ratings after 15
years. The report included results from a regression analysis of NBI data for the
deterioration of structural conditions (the scale ranges from 9 (Excellent
condition) to 0 (Failed condition)). For example, the results suggest that the
average deck condition rating declines at the rate of 0.104 points per year for
approximately the first 10 years and 0.025 points per year for the remaining
years. In addition, the overall structural condition declines at a value of 0.094
per year for 10 years and 0.025 per year thereafter. These results suggest that
rat
the condition will not fall below 6 until after 60 years, which is not the case in
real life: bridges deteriorate at a much higher rate [92]. In another study [88] the
estimated average deterioration of decks was about 1 point in 8 years and 1
point in 10 years for the superstructure and substructure, respectively. A simple
description of the deterioration process over time for an imperfect bridge in a
hostile environment is given in Fig. 11. The curve from A to B represents the
deterioration of a material, a bridge element, a single bridge or indeed a stock of
bridges. This representation is very close to reliability based maintenance
Du
management of bridges and is easily absorbed in a full reliabilistic format.
Fig. 11. Bridge deterioration. [93]
46
The development of an accurate global model of deterioration is virtually

impossible. There are too many variables to consider, especially in the case of
reinforced or prestressed concrete or masonry. With the current state of
knowledge, however, examination of the deterioration mechanisms of structural
elements subject to a hostile environment is possible with a view to finding a
suitable local model. This leads to predictions related to the condition of a
bridge in the future, including possibilities of indicating necessary preventative
action at the design stage. This attempts to contain deterioration to a certain
specified degree, if it happens at all even after taking necessary precautions.
In general, deterioration models can be categorized into four main categories:
mechanistic models;
deterministic models;
stochastic models;
t

iNe
artificial intelligence (AI) models.
6.3.1 Mechanistic models

Mechanistic models are detailed models that describe the specific deterioration
mechanisms of particular bridge components. These models are usually
effective at the project level but not at the network level [94] Miyamoto et al [95]
used load-carrying capacity and durability to predict bridge deterioration. The
load-carrying capacity is defined as the bridges performance based on the
rat
load-carrying capacity of the bridge member, whereas durability is defined as

the ability of the bridge member to resist deterioration. The scores for the load-
carrying capacity and durability are ranked on a scale of 0 to 100 (a newly built
bridge). As the bridge deteriorates, the score decreases and finally drops to 0,
indicating that the bridge should no longer be in service and requires immediate
action. Yet, as reported by Stukhart et al [96], most of these models have not
been tested in practice and none of the Department of Transportation (DOT)
Du
uses such models. In addition of being unreliable for the development of BMS, it
is difficult with the use of these models to incorporate the various variables
affecting the deterioration process. Note that the power models given in the
following are very tractable but should be used with care. Their consistency is
more to predict a main trend than to transfer variability in a detailed probabilistic
analysis. They should be replaced as often as possible with models available
from the literature that are more complete and incorporate physical and
meaningful variables.
Notwithstanding the huge number of masonry arch bridges in service, the most
common building material for bridges and structures is, undoubtedly, reinforced
concrete. It is used to form structural elements such as beams, columns, slabs
(both ground and floor), beam and slab, composite beam and slab, foundations
and bridge parapets etc. The (basic) material constituents are cement, sand,
[97]
coarse and fine aggregate, water and steel reinforcing bars . The well-known
phenomena that cause deterioration of concrete are alkali-silicate reaction,
sulfate attack, freezethaw, acid attack, carbonation, and chloride ingress.
47
Taking all these phenomena into consideration would lead to a complex

deterioration model. For the time being, it is more useful to take the decisive
phenomenon into consideration - chloride-initiated corrosion. According to
Browne [98], McCarter et al [99], McCarter et al [100], Tang and Wang [101], the
attack (by carbonation and chloride contamination) is in two stages:
Stage 1: when chloride contamination or carbonation has reached the
reinforcement.
Stage 2: when corrosion due to oxidisation of the steel has expanded
the corrosion products from the steel sufficiently to cause cracking,
spalling and delamination (the corrosion products occupy about five
times that of uncorroded steel and tend to cause cracking, spalling and
delamination).
t
For corrosion damage to occur, carbonation or chlorides + water + oxygen +
low resistivity concrete would have to be present in sufficient quantity and depth.
iNe
According to Browne [102] the penetration rate for carbonation can be determined
from the simple diffusion law:
(14)
Where x is the distance penetrated after time t. (The diffusion constant k is
obtained either from experimental work or from a calibration procedure involving
a regression analysis of past projects.) Equation 14 is based on empirical data
only, and the more general form of the equation to take into account of the
rat
fractal growth of the calcium carbonate within the pores would be:
(15)
Where i is a fractional parameter between 0.5 and 1.0. Both k and i are
obtained from a calibration procedure based on site results.
Modelling the deterioration of a steel bridge involves a three-part process [103]:
The breakdown of the paint system, which has no immediate effect on
Du
the serviceability limit state.

The corrosion of the steel members. If they are secondary members
(for example diagonal ties or struts in a truss bridge), then if there is a
measure of built-in redundancy in the design, an alternative load path is
available there should be no cause for alarm. If they are principal
members such as the tension or compression boom of a truss bridge,
then the strength of the bridge is threatened.
Corrosion of riveted, bolted or welded connections. The likelihood of
failure would depend on whether the fastening/gusset is corroding as a
whole, or proportionately.
Although this may be true for the more dominant mechanisms, in practice there
are a multitude of mechanisms all taking place simultaneously, each one a
subset of a larger population [104]. According to Van der Toorn and Reij [105]
When there are more ageing elements in a structure, the maintenance strategy
for the structure as a whole will normally differ from the sum of the individual
48
maintenance strategies [106]. The mechanism of the corrosion process for the
superstructure of steel bridges has been developed by Sobanjo [107], Melchers
[108]
, Melchers et al [109], Melchers et al [110]. The following expression predicts
the deterioration:
(16)
where C = average corrosion penetration, t = time in years, and A, B = constants.
In steel and concrete bridges the dominating deteriorating process is corrosion.
For timber it is fungal decay and wood-boring insects which dominate. The main
aim of any timber treatment is to break the egg-laying cycle of the wood-boring
beetle. The species of beetle doing the damage depends greatly on:
The type and condition of the timber playing host to the beetle.
t
The density of the wood heartwood is less susceptible than sapwood.
The degree of existing decay: in decayed wood, the cycle development
iNe
may be faster and repeated more often. Some types of woodborer will
only attack timber which has started to decay.
Resistance to attack is greatly enhanced by the application of a preservative or
an insecticide or a mixture of both, which are capable of preventing infestation
and eradicating existing infestation. There are many commercial products on
the market and so it is advisable to enlist the services of a specialist so that the
most effective product is chosen. Modern preservatives are based on a soluble
borate and exhibit good penetration into damp timber. Application is by brush,
rat
sprays or pressure injection, and is best carried out in the Spring when the
insects come to the surface to mate. Thereafter treatment should continue
annually until attack is terminated.
Timber structures can suffer severe structural damage at the hands of wood
rotting fungi. The spawn (of which the fungus plant is composed) lives in the
wood and feeds on it by releasing enzymes which enable digestion of the wood.
The fungus absorbs the digested material to produce more spawn and
Du
continues until there is no more digestible material in the wood when it dies.
The spawn needs a minimum moisture content of 20 % and a low relative
humidity in which to germinate and thus initiate decay. Generally, if there is no
water present then there will be no decay. Brown Rot is the result of the fungi
feeding on the cellulose wall. It is characterised by wood that is browner than
normal and which cracks across the grain. The wood becomes spongy, lighter
in weight, and is severely weakened. In areas where brown rot has ceased and
become dry, it is often erroneously referred to as dry rot. White Rot feeds on
the wood directly, causing the wood to lose its natural colour and to appear
whiter than normal. The wood gradually loses its strength and, as in brown rot, it
becomes spongy to the touch. Timber bridges are, for the most part, in a totally
exposed environment and subject to mist, fog, rain and ground water. The
vulnerable areas are at connections, bearings and expansion joints which tend
to work lose after a period of time, due to vibration from traffic, leaving a gap (be
it ever so small), between the timber and steel or concrete. Wind-blown fungal
spores can then find their way into the gaps and decay begins. Treatment of
infected areas using preservatives and insecticides is similar to beetle
49
infestation, but in addition, the exposed wood surfaces should be coated with a
fungicide and a water repellant finish.
6.3.2 Deterministic models

Deterministic models are dependent on a mathematical or statistical formula for
the relationship between the factors affecting bridge deterioration and the
measure of a bridges condition. The output of such models is expressed by
deterministic values (e.g., there are no probabilities involved) that represent the
average predicted conditions. The models can be categorized as using straight-
line extrapolation, regression, and curve-fitting method [111],[112]. The two main
types of deterministic methods are: the straight-line extrapolation and
regression models.
t
6.3.2.1 Straight-line extrapolation
iNe
The simplest condition-prediction model is based on straight-line extrapolation;
this method can be used to predict the material condition rating (MCR) of a
bridge given the assumption that traffic loading and maintenance history follow
a straight line. The method requires only one condition measurement to be
carried out after construction; an initial condition can be assumed at the time of
construction and a second condition is determined at the time of the inspection.
The straight-line extrapolation is used because of its simplicity [90]. Although this
method is accurate enough for predicting short-term conditions, it is not
rat
accurate for long periods of time. In addition, the straight line method cannot
predict the rate of deterioration of a relatively new bridge, or of a bridge that has
undergone some repair or maintenance.
6.3.2.2 Regression model

Regression models are used to establish an empirical relationship between two
or more variables: one dependent variable and one or more independent
Du
variables. Each variable is described in terms of its mean and variance [90].
Several forms of regression models are presented in the literature, including
linear and non-linear ones. Even if these regression models are based on
statistics, they are, despite everything, regarded as deterministic.
6.3.3 Stochastic models

The theory of stochastic processes is now being increasingly used in
applications in engineering and other applied sciences. The general concepts of
stochastic processes can be found in Srinivasan and Mehta [113]. The use of
stochastic models has contributed significantly to the field of modelling
infrastructure deterioration because of the high uncertainty and randomness
involved in the deterioration process. The most commonly used stochastic
techniques for infrastructure deterioration are the Markov chain model for
project and network level and also reliability method for project level. One of the
most popular stochastic techniques obtained from operation research is the
50
Markov Decision Process (MDP). This process has been used to develop
stochastic deterioration models for different infrastructure facilities. Markovian
bridge deterioration models are based on the concept of defining states in terms
of bridge condition ratings and obtaining the probabilities of a bridge condition
changing from one state to another [114]-[116]. The inputs to the deterioration
model are the present condition assessment data for the bridge. Based on the
FHWA condition rating for the deterioration in the condition of the Indiana
Department of Highway bridges (IDOH) and Jiang et al [115] have developed a
performance prediction model by using the Markov chain. In this model, a
transition probability matrix was developed for three main bridge components:
the deck, superstructure, and substructure. The transition probability matrices
take into account the type of structure (steel or concrete), the effect of age
(assuming that the rate of deterioration differs with age), and the highway type
(interstate or other). The drawback of this study is that it does not consider other
t
factors affecting the deterioration process such as traffic density and climate.
iNe
In Pontis, Markov chain is utilized in the development of the element
deterioration model. The model incorporates five condition states for each
element. To include the factors that affect the deterioration, Pontis classifies
each element of a bridge into one of four categories of environment: benign, low,
moderate, or severe. Each environment represents a different level of the
impact of the external factors on the performance of the element over time, and
[88]
a deterioration matrix is assigned for each element in each environment . It
should be noted that the transition matrix (and accordingly, the deterioration
behaviour) is greatly affected by the service condition (or the environment) to
rat
which the bridge element is exposed. In an interesting study by Morcous et al

[112]
, they attempted to describe clearly the service conditions associated with
the four environmental categories described in the Pontis system: benign, low,
moderate, and severe. Genetic algorithms are used to arrive at the best of the
four environmental categories. Once the category is known, then the transition
matrix associated with it is used to predict the deterioration.
Although Markovian models have been employed in many advanced BMS such
Du
as Pontis and Bridigit, great advances in modelling bridge deteriorations have

been achieved with their use; they are still based on assumptions and have
some limitations:
Markovian models assume that past conditions have no effect on
predicted ones [117];
Markovian models assume discrete transition time intervals, a constant
bridge population, and stationary transition probabilities [118];
it is quite difficult for Markovian models to consider the interactive
effects among the deterioration mechanisms of different bridge
components [119];
the transition probabilities are estimated in terms of subjective
engineering judgement and require frequent updating when new data
are obtained [120].
51
6.3.4 Artificial intelligence deterioration models

The area of artificial intelligence (AI) is comprised of several different
techniques that have been utilized in a variety of applications during the last few
decades. Artificial neural networks (ANN), case based reasoning (CBR), and
machine learning (ML) are examples of AI techniques that have been
recognized as powerful tools for solving engineering problems [121]. The
feasibility of using ANN in modelling bridge deterioration has been investigated
by Sobanjo [107]. A multi-layer ANN was utilized to relate the age of the bridge
superstructure (in years) to its condition rating (a numeric value from 1 to 9).
The network configuration used in this study is depicted in Fig. 12. The
inspection records for 50 bridge superstructures were utilized to train and test
the network; 75 % of the data were used for training, while the remaining data
were used for testing. The use of this ANN resulted in 79 % of the predicted
t
values were with a 15 % prediction error.
iNe
rat
Fig. 12. Multi-layer neural networks. [107]
Du
In more detailed Artificial Intelligence model, Tokdemir et al [120], using age,

traffic, and geometrical and structural attributes as explanatory variables,
predicted a bridge sufficiency index (SI) ranging from 0 to 100. Testing the
performance of the developed ANN resulted in an average percentage of
correct solutions of 33.5 % and 62.5 % with a prediction error of 3 % and 6 %,
respectively. Two of the difficulties associated with using ANN models are:
the determination of an efficient ANN architecture is carried out in an ad
hoc manner and does not follow clear rules [122];
ANN work well when the input and output variables are numerical
values. The conversion from nominal data to numbers may lead to the
loss of information that was contained in the original representation [123].
52
6.4 After-repair deterioration model

It is very important that any BMS be able to estimate the future conditions of
bridges after a specific repair has been performed. It was recorded in the
literature that the rate of deterioration of rehabilitated bridges is greater than
that of newly constructed bridges. It was, also, noticed that rehabilitated bridges
do not revert back to their best condition [124]. However, currently, most BMS
assume that the rate of deterioration after repair is the same as that in effect
when the bridge was constructed. Bolukbasi et al [125] recently investigated the
rate of deterioration of reconstructed steel decks for highway bridges in Illinois.
The study concluded that a reconstructed deck has at least a 25 % shorter life
span than new decks. Although Bolukbasi et al [125] quantified the rate of
deterioration for bridge decks after improvement; they included only one type of
improvement: deck replacement. The study needs to be expanded to include
t
other improvement types as well as other bridge elements.
It is observed from the other reports in this project, that the type of repair, the
iNe
extent of repair and the location of repair are not independent elements of
infrastructure maintenance management but are a function of available
information and the evolution of safety of the structure. Consequently,
uncertainties related to degradation model, material strengths and other
epistemic and aleatory factors are always present affecting the improvement of
a structure. Consequently, a high degree of confidence on the performance of
maintenance management, testing of structures and sophisticated damage
models directly affects the performance of repair. Additionally, the performance
rat
a number of repair techniques may not be modelled and consequently access

to the data of their empirical performance under stringently defined testing
protocol may be extremely helpful is predicting the true improvement of a
structure. In this regard, encouragement should be provided for the use of
green materials and sustainable solutions like the use of recycled materials.
Du
53
7 Reliability analysis (assessment and optimising

maintenance)
Reliability analysis requires establishing the as-built reliability level at a higher
level than target reliability levels with regards to a number of performance
criteria. The schematic of this entire process is given in Fig. 13. The
performance criteria of the structure are obtained through deterministic analysis
and the target reliability index is established. The stochastic variables related to
both loading and resistance are modelled and the target classification of the as-
built structure is ascertained. The computed value of reliability index is
compared with target value and the structure is reclassified accordingly.
Sensitivity analysis, a by-product of this scheme, can directly enter into a
feedback loop affecting the uncertainty modelling and consequently the
t
reliability index. Consequently, reliability analysis is intrinsically an iterative
process.
iNe
rat
Du
Fig. 13. Schematic for process.
54
7.1 Probabilistic analysis

Reliability of an engineering system refers to the probability of survival or the
complement of probability of failure [126]. The concept of a Limit State (LS) is
used to define Failure in the context of structural reliability analysis. The term
Failure in the reliability analysis does not necessarily imply structural collapse,
but in most cases it refers to a situation when the performance of the structure
exceeds a predefined limit. For example; if the LS to be considered is the
initiation of reinforcement corrosion, then Failure in this case may be defined
as when the chloride content at the reinforcement depth exceeds a critical value.
Thus the LS is a boundary between desired and undesired performance of a
structure as indicated in Fig. 14 [127].
t
G(R, S)>0
Safe region
iNe
Load S
G(R, S)=0
G(R, S)<0
Unsafe region
rat
Resistance R
Fig. 14. Depiction of limit state. [127]
In a standard case, when the LS function is not time-dependent, it may be

written as [126]:
Du
Z= G(R,S) = R S (17)
where Z is the LS margin, G is the LS function, R is the random variable
representing the resistance and S is the random variable representing the
corresponding load effect or action.
R and S may be functions of other variables, deterministic or random. Both R
and/or S may also be functions of time as schematically shown in Fig 15.
Z(t) = R(t) S(t) (18)
55
R, S
R(t)
S(t)
t( years)
t
tL
Fig. 15. Schematic representation of time-variant reliability problem. [128]
0 0 iNe
If in the LS functions Z < 0, then the failure state is reached. The probability that
the LS has been violated (Z < 0) is referred to as the probability of failure (Pf),
and can be obtained by solving the following equation [126]:

Where, X is the vector of all relevant basic variables, G(X) is the LS function
(19)
rat
which expresses the relationship between the LS and the basic variables, fx(x)
is the joint PDF for the n-dimensional vector X of basic variables.
In structural reliability, the probability of failure Pf is also represented by the
reliability index (), which is expressed by [126]:
(20)
Where is the joint standard normal distribution function obtained from certain
Du
equivalence conditions starting with original non-normal distributions of

variables involved.
An appropriate figure (Fig. 16) has been shown by SAMCO (2006) to appreciate
the probabilistic joint description of the load and the resistance variables.
56
t
iNe
Fig. 16. Joint description of load and resistance variables.
7.1.1 Calculation of the probability of failure (Pf)

The region of integration G(X) 0 denotes the space of LS violation.
Theoretically, the solution to Equation can be obtained through one of the
following three methods [126]:
rat
direct analytical integration;

numerical Integration, such as simulation methods;
transforming the integrand into a multi-normal joint PDF for which some
special results are readily available.
Except for some special cases, the integration of Equation over the failure
Du
domain G(X) 0 cannot be performed analytically [126]. In addition, the LS

equation contains functions of the basic variables that are too complicated for
calculus to be used in the evaluation of their integrals. Methods 2 and 3 in this
case become very practical choices for the evaluation of the failure probability.
One common technique of type 2 methods is Monte Carlo (MC) simulation. As
for type 3 methods, there are several classical techniques such as the First
Order Second Moment method (FORM) and the Second Order Reliability
Method (SORM), etc. These methods have been described in detail in many
classic references [60],[126],[129].
7.1.2 Target reliability T

For the calculated probability of failure to be of significant engineering value it
has to be compared to criteria by which the performance of the structural
system under investigation can be rated. The criterion introduced to fulfil this
requirement is known as the Target Reliability (T). Values for the minimum
acceptable target performance level, T, for the design/assessment of different
57
types of structure components have been explicitly specified by various national

and international codes [21],[130]. These values were derived by a process of
probabilistic calibration to different design codes. The specified T values were
intended to be applicable to a wide range of structural components. In most
cases the selection of a target performance value depends on parameters such
as importance of the structure, possible failure mode, etc. Table 5 presents
values of the allowable maximum probability of failure specified by ISO
2394:1998 [130].
Table 5. Lifetime target probability of failure (Pf). [130]
Relative costs of Consequence of failure

safety measures
Small Some Moderate Great
t
High 0.5 10-1 10-2 10-3
iNe
Moderate 10-1 10-2 10-3 10-4
Low 10-2 10-3 10-4 10-5
The target reliability (T) indices specified by the Eurocode are given in Table
6. These values refer to a situation where the LS function is time-variant. Table
7 provides definitions for the reliability classes indicated in Table 6. The T
values are specified for structures with High, Moderate, and Low consequence
for loss of human life. Once the structural probability of failure (or ) is
rat
computed, it can be compared with the T value specified in Table 6 for the
considered LS and the consequence determines compliance or violation.
Table 6. Minimum acceptable safety levels. [21]

Minimum acceptable values
Reliabilty
1 year reference period 50 year reference period
Class
(=associated Pf) (=associated Pf)
Du
CC3(RC3) 5.2 (1.0*10-7) 4.3 (8.5*10-6)

CC2(RC2) 4.7 (1.3*10-6) 3.8 (7.2*10-5)
CC1(RC1) 4.2 (1.3*10-5) 3.3 (4.8*10-4)
Table 7. Reliability classes. [21]
Reliabilty Examples of buildings and civil

Description
Class engineering works
High consequences for loss of
CC3(RC3) Grandstands, public buildings
human life
Medium consequences for loss of
CC2(RC2) Residential and office buildings
human life
Low consequences for loss of Agricultural buildings (e.g.
CC1(RC1)
human life people do not normally enter)
58
The target reliability level is that level of reliability required ensuring the
acceptable safety and serviceability of a structure. The authorities or bridge
owners must specify the target reliability level. These levels can be explicitly or
implicitly specified in a code. The target reliability level for assessment is
different from that appropriate for the design stage due to:
economic considerations, that lead to the use of less conservative
criteria for assessment as the increment in cost of upgrading an existing
structure is much larger than that for increasing safety at the design
stage;
social considerations, such as heritage values and disruption of
occupants and activities caused by an intervention that do not affect the
design of new structures;
t
Sustainability considerations, e.g., reduction of waste and recycling,
more appropriate with the rehabilitation of existing structures.
iNe
Table 8 compares target lifetime reliability levels in various codes and standards
currently in use. The Engineer dealing with the assessment of an existing
structure must decide among the available tables which of the values are most
suited and best applied to the solution of the problem at hand as the estimated
probability of failure associated with a project is very much a function of the
costs as well as an understanding of the issues, modelling the data, etc. In the
ISO 13822:1999 [131] the target reliability levels depend on the type of Limit State
examined as well as on the consequences of failure, and they range from a
rat
reliability index, , of 2.3 for very low consequences of a structural failure to 4.3
for structures whose failure would have very severe consequences. In the
Ultimate Limit State, a value of 4.3 would be suitable for most cases. The value
recommended by ISO 2394:1998 [130] and the JCCS [132], depends on the
consequences of a structural failure as well as the costs of a safety
enhancement measure. The JCCS [132] target reliability index ranges from 3.1 to
4.7. In Eurocode 1 [133], only depends on the type of limit state examined and it
ranges from 1.5 to 4.7, while in the NKB report [134], the failure type and
Du
consequence is taken into account in the determination of . The Canadian

Standards Association (CSA) obtains through an equation allowing for
element and system behaviour, the inspectability and the traffic category, and it
can take values between 2.00 and 3.75 [135].
59
Table 8.Comparison on international safety requirements.

Code Target Reliability Index
Small consequences of failure = 0.0
Relatively high Some consequences of failure = 1.5
cost of safety
Moderate consequences of failure = 2.3
measure
Great consequences of failure = 3.1
Relatively
ISO Moderate cost Some consequences of failure = 2.3
2394:1999 of safety
[130] Moderate consequences of failure = 3.1
measure
Great consequences of failure = 3.8 (1)
Relatively low Some consequences of failure = 3.1
t
cost of safety
measure Moderate consequences of failure = 3.8
iNe
Great consequences of failure = 4.3
Relatively Minor consequences of failure = 3.1
large cost of Moderate consequences of failure = 3.3
safety
measure Large consequences of failure = 3.7
Relatively Minor consequences of failure = 3.7
JCSS [132]
Normal cost of
Ultimate Moderate consequences of failure = 4.2
safety
limit state Large consequences of failure = 4.4
measure
Relatively
rat
Minor consequences of failure = 4.2
small cost of
Moderate consequences of failure = 4.4
safety
measure Large consequences of failure = 4.7
Design working life: bridges 100 years = 1.5
Serviceability
1 year = 3.0
Eurocode
[133] Fatigue Design working life: bridges 100 years = 1.5 - 3.8
1:1993
Design working life: bridges 100 years = 3.8
Du
Ultimate
1 year = 4.7
Less serious failure consequences = 3.1
Ductile w/
extra carrying Serious failure consequences = 3.7
capacity failure
Very serious failure consequences = 4.2
NKB
Report
[134] Less serious failure consequences = 3.7
Ductile w/o
Ultimate extra carrying Serious failure consequences = 4.2
Limit State capacity failure
Brittle failure Serious failure consequences = 4.7
(1)
For Ultimate Limit State analysis revises values. See code.
60
Table 8 (cont.).Comparison on international safety requirements.

Code Target Reliability Index
Ductile w/ Less serious failure consequences = 3.1
extra carrying Serious failure consequences = 3.7
NKB capacity failure Very serious failure consequences = 4.2
Report Less serious failure consequences = 3.7
[134] Ductile w/o
Ultimate extra carrying Serious failure consequences = 4.2
Limit capacity failure Very serious failure consequences = 4.7
State
Brittle failure Serious failure consequences = 4.7
Sudden loss of capacity no warning E = 0.0
Adjustment for
t
Sudden failure no warning retention of
element E = 0.25
post failure capacity
behaviour, E
CSA [135]
iNe Gradual failure probable warning E = 0.5
Element failure leads to total collapse S = 0.0
Ultimate Adjustment of
limit state system Element failure probably does not lead to
S = 0.25
= 3.5 behaviour, S total collapse
(E + S Element failure leads to local failure only S = 0.5
+ I + Component not inspectable I = - 0.25
Adjustment for
PC) Component regularly inspectable I = 0.0
inspection
2.0 Critical component inspected by
level, I I = 0.25
evaluator
rat
Adjustment for All traffic categories except permit

PC= 0.0
traffic controlled
category, PC Traffic category permit controlled PC= 0.6
(1)
For Ultimate Limit State analysis revises values. See code.
7.2 Probabilistic load modelling

Du
In the following the principles for the creation of loading models are given.
In classification, loads are categorized as permanent loads (dead load,
prestressing, etc.) and variable loads (traffic, wind, etc.).
In a reliability analysis the loads are modelled as stochastic variables - a
modelling that for variable loads will often be a function of time.
In the reliability analysis it will often be necessary to combine several load
processes, e.g. load on two or more lanes. With variable loads and
combinations of variable loads a distinction is made between the distribution of
immediate values and extreme values, where the distribution of extreme values
is adjusted to the reference period (here one year) using the definition of the
required safety.
61
7.2.1 Dead & superimposed load

The permanent loads comprise the weight of the structure and the earth
pressure on the structure. Permanent loads are generally modelled by normally
distributed variables. In uncertainty modelling, a distinction is made between the
dead load of the structure itself, G, and the superimposed dead load, GW.
The following can be used as a starting point:
G is assumed to be normally distributed with a variation coefficient of
5 %.
GW is assumed to be normally distributed with a variation coefficient of
10 %.
Permanent loads from different sources are assumed to be
t
stochastically independent.
The above shall be evaluated in each case. It is possible to reduce the given
iNe
uncertainties by measurement.
7.2.2 Traffic load

The traffic load is dependent on the type of structure in use and correspondingly
the traffic load on that structure is defined. Standard operational vehicular
model are often represented as a combination of a Uniformly Distributed Load
(UDL) and a Knife Edge Load (KEL) for road bridges, the intensity of which is
rat
governed chiefly by the span of the load. Abnormal vehicles are also specified
in such format. In all formats, the presence of such models is generally confined
only to probable cases for which safety and serviceability are considered. Under
certain circumstances, specific abnormal vehicular passages or reduced and
contained vehicular limits are also in place that is required to be checked for
assessment. The location of the vehicles at certain lanes is often appropriately
weighed to allow for bunching. A microsimulation approach generally does not
require significant limitations but these models also consider distributions of
Du
axle weights of vehicles and the correlations (or the lack of it) of high axle loads
leading to the load extremes of a structure in its service life. Rail bridges are
specified with a number of specific rail loadings for a network and are used for
design or assessment. The passage cases producing the maximum effect on a
structure (for example, two heavy trucks crossing each other on two lanes at the
midpoint of a simply supported bridge may produce the maximum effects) are
the critical cases to be taken into consideration. The longitudinal and transverse
positioning of loadings is also important in this regard. Due to the wide number
of possibilities of situations that may give rise to extreme vehicular loading, finite
element based software have become popular nowadays for such modelling.
However, for a full probabilistic modelling or microsimulation modelling the axle
weights are all represented as distributions and the loading on the structure is
ascertained through a large number of simulations and represented as an
output distribution. For an appropriate representation of the load, the tails of the
distributions of the loadings are required to be modelled precisely since they
represent the extreme effect on a structure. Consequently, for high loads, apart
62
from the critical crossing locations, the likelihood of such occurrence and
considerations of correlations are required to be modelled with great care. The
dynamic amplification of a bridge may actually go down with increasing weight
on a bridge. However, in most models, the amplification factor is always
represented as a number greater than 1, the excess being modelled as a
probabilistic distribution. Some Coefficient of Variation (CoV) is typically
associated with such dynamic factors to account for local and unknown effects.
The dynamic factor often does not appropriately account for the effect of
lurching and consequently a parameter related to lurching is sometimes
considered and represented as a distribution. The model uncertainty is
represented as a multiplicative term with a distribution.
7.2.3 Probabilistic resistance modelling
t
Probabilistic resistance modelling significantly depends on a number of factors
(on their own or combined) including the control of manufacture of the materials,
iNe
assumed degradation models, experimental results at any time on the structure
and engineering judgement on the uncertainty of a number of factors. A wide
range of degradation modelling is possible and the scope of detailed discussion
is within WG A3 and WG A4 of this project. Repair of structures implying green
materials or recycled materials also have a direct impact on the cost over the
lifetime of a structure and have significant influence on the prioritisation and
optimisation of infrastructure maintenance management funds. These are
considered in other reports of this project. Within a full probabilistic format a
rat
resistance model is often associated with a number of different sources of

uncertainties that are separately accounted for. Such uncertainties tend to form
the basis of model uncertainty and are explained in the next section. However, it
should be noted that the choice of relatively constant parameters are often of
great importance since the incorporation of too many stochastic parameters
may lead to an over-parameterization of the probabilistic model and then it
becomes difficult to isolate the factors that the chiefly responsible for
significantly changing the reliability index. Factors associated with operations on
Du
built or manufactured materials, like prestressing and associated losses may be

modelled statistically. Experiments may be carried out to update the parameter
values. However, the number of experiments carried out for infrastructure
maintenance management is often not of statistical significance. This problem
may be addressed by encouraging data sharing among the various
stakeholders of an infrastructure system. However, considering practicalities
and the present condition where the access to data may be scarce,
representative values from the tests are most typically used to calibrate
parameters of assumed distributions related to capacity variables. These
assumed distributions are chosen based on previous data, engineering
judgement and existing body of information through research or engineering
practice. The capacity model may not always be algebraic and iterative
simulations may be necessary to be carried out.
63
7.3 Incorporation of model uncertainty

The model uncertainty takes account of: (1) the accuracy of the calculation
model, (2) possible deviations from the strength of material properties in the
structure involved as compared with that derived from control specimens and (3)
material identity. The model uncertainty is taken into account by introducing
judgement factors Im related to the material properties. The judgement factor Im,
which is assumed to be log-normally distributed with mean value equal to 1.0
and coefficient of variation VI, is introduced by multiplying the basic material
variables by Im. VI, is calculated as [134],[135]:
VI = [VI12 + VI22 + VI32 + 2(1V1 + 2V1 + 3V1)VM] (21)
where the variation and correlation coefficients, VI and i respectively, are
t
determined from Tables 9 to 11 and VM is the coefficient of variation of the basic
material variable.
VI 0.04
iNe
Table 9 Coefficient of variation and coefficient of correlation for the
accuracy of the computation model.
Good
Accuracy of the computation model, I1
Normal
0.06
Poor
0.09
rat
- 0.3 0.0 0.3
Table 10. Coefficient of variation and coefficient of correlation for the

uncertainty in determination of material parameters.
Uncertainty in determination of material parameter for the structure on the

basis of the vicarious material parameter, I1
Du
Good Normal Poor

VI 0.04 0.06 0.09
- 0.3 0.0 0.3
Table 11. Coefficient of variation and coefficient of correlation for

uncertainty in the identity of materials.
Material Identity, I2
Good Normal Poor
VI 0.04 0.06 0.09
- 0.3 0.0 0.3
64
7.4 Incorporation of supplementary information

Supplementary information provides a choice of updating the stochastic and
deterministic description of the variables participating in the limit state and
directly influences the as-built safety level of structure. Similar supplementary
information at various time instants can be extremely valuable is establishing
the true evolution of safety of the structure. Such information also provides the
basis of predicting future rates of degradation (and hence, the service life)
accurately. The updating happens in various levels, complexity and detail. It can
be visual, be based on updated condition rating or can be a detailed structural
health monitoring results. Consequently, even when the statistical description of
the load effects remain unchanged, updating the resistance distribution may
significantly change the overlap of the tails of the distribution, This situation is
illustrated in Fig. 17 where updated resistance distribution changes the overlap
t
area of the tails of the load and the resistance distribution, thereby affecting the
final safety rating.
iNe
rat
Fig. 17. Incorporation of additional information as an update of distribution

function.
For existing bridges the possibility of obtaining supplementary information on

e.g. material parameters, dimensions and loads in relation to the design is an
Du
important advantage, often used in practice. For a reliability-based classification

it is important that this information is used rationally. Supplementary information
can be in the form of observations of a given material/load variable or of an
event. Observations of a given variable can be used to estimate or update the
distribution parameters of that variable. Observation of an event can be used for
a direct updating of the reliability insofar as it is affected by that event.
An example where the observation of a given variable can be used to estimate
or update the distribution parameters is the measured compressive strengths.
Such an approach is in fact reflected in deterministic assessments when a worst
credible strength of concrete compression is used from statistically significant
and representative samples instead of factored design strength. A similar
conclusion can be drawn from measuring the tensile strength of reinforcements.
The update of strength parameters directly affects the final distribution of the
resistance variables. An example of observation of an event to directly update
the reliability can be illustrated by proof load tests on structures. Under certain
and very limited conditions (usually for Forensics or where a part of the
65
structure is defunct or can be used as a representative section for testing to

failure) a full failure test have been carried out on a limited part of a structure.
It is important to note here that the updating of reliability through additional
information is not a direct, but rather a feedback relationship. The sensitivity
studies of the affecting variables on the reliability index indicate the critical
variables that affect the safety of the structure in its as-built condition and the
evolution of the same in time. A quantitative ranking of the variables by their
participation percentages is also possible when the limit states can be
expanded in first or second order forms about the design point. Consequently,
the additional information to be sought can be guided by the numerical studies
on the reliability of a structure. A similar parallel is only possible under limited
and subjective circumstances for deterministic or semi-probabilistic
assessments. Additionally, a sensitivity study helps avoid over-parameterisation
t
of the limit state and accurate reformulation of the statistical variables can be
carried out.
iNe
To update the reliability of a structure by taking supplementary information into
account, sufficient documentation for the supplementary information must be
provided to assure appropriate quality control.
Whatever might be the format, a First Order Reliability Method (FORM) or a
Second Order Reliability Method (SORM) may be followed. The limit states are
generally convex within the domain of definition and consequently in most of
such reliability problems, the design point at which the minimum credible safety
measure is obtained remains unique. For Monte Carlo type simulations, often
rat
directional searches or Latin Hypercube simulation are carried out for
computational efficiency.
In the observation of a given material or load variable the distribution
parameters for the variable such as mean value and standard deviation can be
estimated and updated. In estimating and updating parameters the observations
shall be stated. The uncertainty in the measurement shall be stated and taken
into account in the estimation of distribution parameters. Estimation of the
Du
distribution parameters shall be carried out on the basis of a method in which

the statistical uncertainty of the parameters is determined. This means that the
distribution parameters can be estimated on the basis of maximum likelihood
and Bayesian Statistics. For inspection related studies, typically a random

variable is assumed to have a prior distribution of f () while the fitted

posterior density function is f () following an inspection with discrete number of
measurements. The relationship between the prior and the posterior [30] is
expressed as:
(22)
Where L () is the likelihood function and k is a normalizing constant.

Empirical Bayes model has also been successfully used to update reliability
where a number of deterioration scenarios are present related to the durability
of a structure [57].
Under certain circumstances, for example, simulation based establishment of
distribution parameters a combination of observations and numerical simulation
66
can be directly combined to form the tails of the extreme value distributions.
Here, the parameter updating becomes dependent on the underlying
assumptions and the nature of the simulation and care should be exercised to
avoid unreasonable or unrealistic statistical descriptions.
Additionally, it is to be remembered that the number of available tests or
information are often limited in structural assessment and at a certain level
some engineering judgement may be required to be applied with care.
The following section presents some examples of structures with respect to the
content of the TG in brief. Three structures are considered the Ferrycarrig
Bridge, Co. Wexford, Ireland, the Barra Bridge, Aveiro, Portugal and a Steel
Wharf, Nantes, France.
t
iNe
rat
Du
67
8 Examples
8.1 Ferrycarrig Bridge, Co. Wexford, Ireland

Ferrycarrig Bridge is located in a tidal estuary in County Wexford on the South
East coast of Ireland. It was constructed in 1980 and is managed by the Irish
National Roads Authority (NRA). The bridge is a 125.6 m long beam structure
consisting of eight, 15.7 m equal spans. A photograph of the bridge is shown in
Fig. 18.
t
iNe
rat
Fig. 18. A photograph of Ferrycarrig Bridge
Fig. 19 shows the Ferrycarrig Bridge cross section. As can be seen in this figure,
each bridge deck span consists of 28 precast, prestressed bridge beams with a
reinforced in-situ concrete infill.
Du
Fig. 19. Cross-section of Ferrycarrig Bridge.
At the intermediate supports the bridge deck is supported by a crosshead beam

which in turn sits on a pair of pier walls. These pier walls each encase two
groups of four steel hollow piles which are driven to bedrock. The pier walls
68
extend from the crosshead beams to the low water level. The bridge is
continuous over all piers except the middle pier where an expansion joint has
been provided in the deck. There are concrete abutments at both ends of the
bridge structure which are integral with the bridge deck.
In 2002 a routine inspection of Ferrycarrig Bridge highlighted a number of
problems with the bridge, namely:
staining and visible concrete cracking of the crosshead beams;
failure of expansion joints both at the middle support and at the south
abutment;
cracking and staining of South retaining wall and abutment.
These problems lead to a special inspection and structural investigation being
t
carried out in 2004. The special inspection and structural assessment of
Ferrycarrig Bridge, identified a number of issues that affected the long term
iNe
serviceability of the structure. These issues are summarised as follows:
The observed cracking in the pier crosshead beams, ranging from
hairline to 3.5 mm, was considered to be due to a lack of reinforcement
to resist both the applied ultimate limit state torsion moments and the
serviceability limit state stresses (e.g. shrinkage, thermal, creep).
It was anticipated that the cracks would continue to develop, thereby
ultimately compromising the integrity of the piers it was concluded that
pier strengthening works should be carried out.
rat
Other elements of the bridge needed to be replaced, namely, the bridge

deck waterproofing system, the bridge deck expansion joints and the
mechanical bearings at Pier 4.
Concrete dust samples revealed that chloride levels in the concrete
showed a distinct concentration gradient decreasing with depth into the
concrete. The levels of chloride ion concentration in the crosshead
Du
beams were considered moderately high.

In July 2007 refurbishment works on the Ferrycarrig Bridge commenced. These
works comprised of:
extensive concrete repair to all crosshead beams;
re-waterproofing of the existing bridge deck including footways;
replacement of the existing bearings at the central pier;
replacement of the expansion joint over the central pier;
repainting of the existing parapet system;
crack injection of the abutments.
The repair works were completed in January 2008. Although the deterioration of
the Ferrycarrig bridge itself was not due to reinforcement corrosion, the nature
of the bridge repairs afforded the NRA a unique opportunity to conduct site
based research into the efficiency of typical alternative concrete repair options
69
in Irish marine environments. The crosshead beam repairs involved removal of

the existing concrete to a depth of 50 mm beyond the reinforcement using
hydro-demolition. As a research initiative it was decided to employ five different
concrete cover repair options for the bridges seven crosshead beams. Six of
the crossheads were instrumented and are remotely monitored so that the
relative merits of the five repair methods can be studied. The five cover
concrete repair strategies employed were as follows:
Crosshead 1: Cem 1 Ordinary Portland Cement (Cem 1-OPC),
standard 50 mm cover.
Crosshead 2: Cem 1 OPC with 70 mm cover.
Crosshead 3: Cem 1 OPC with silane surface treatment.
Crosshead 4: Cem 1 OPC with ground granulated blastfurnace slag
t
(GGBS) as partial replacement (60 % GGBS).
iNe
Crosshead 5: Cem 1 OPC with mixed in corrosion inhibitors.
Crosshead 6: Cem 1 OPC with GGBS as partial replacement (60
percent GGBS, identical to crosshead 4).
Crosshead 7: Cem 1 OPC 50 mm cover (identical to crosshead 1).
The process of chloride-ion ingress and the resultant reinforcement corrosion is
very slow. Consequently, it will be a number of years before any conclusions
can be drawn on the relative efficiencies of the repair methods from Ferrycarrig
rat
Bridge. Investigations are currently on to investigate the repair methods utilised
at Ferrycarrig bridge in the short to medium term through accelerated laboratory
testing.
A probability based analysis of Ferrycarrig Bridge is presented in the Appendix
A of this report. The deterministic analysis, which was carried out on the
structure in 2002 according to BS 5400-4:1990 [137] is also presented in the
appendix. This analysis of Ferrycarrig Bridge is utilised as a practical example
Du
to compare the results and methods of a probability based assessment to the

more conventional deterministic assessment. The analysis focuses on the
ultimate limit state (ULS) torsional capacity of the crosshead beams in the
undeteriorated state. These beams failed the original deterministic torsional
capacity analysis in 2002. The serviceability limit state is not considered in the
Ferrycarrig bridge analysis example.
The appendix also presents the probabilistic deterioration of the torsional
capacity of the crosshead beams with time due to chloride induced corrosion.
This allows the reader not only to compare the deterministic analysis to a
probability type analysis, but also allows a reader to see how a marine
structures probabilistic safety factor, , might decrease over time due to the
ingress of chlorides and the resultant chloride induced corrosion.
70
8.2 Barra Bridge, Aveiro, Portugal

Barra Bridge, Portugal was designed by Edgar Cardoso in 1971 and has been
opened to the traffic in 1973. Situated in the north of the country, the reinforced
and pre-stressed structure crosses the Mira Canal in Aveiro on the national
road EN.109-7. The Barra Bridge has been subjected to a substantial number of
repairs in the last few years. Fig. 20 presents a plan view photograph of the
Barra Bridge.
t
iNe
Fig. 20. A plan view photograph of Barra Bridge.
rat
A model of the Barra Bridge was created. The actions on the structure and the
resistances were probabilistically treated in order to establish the reliability
before and after repair.
Studies completed by the Portuguese engineers highlighted structurally critical
zones to be at the middle of spans and under a pier designated as pier no8.
For the purpose estimating prestressing forces, as assumption of a uniform loss
Du
of 20 % in each tendon is considered. The bridge was strengthened employing

HEM280 sections as shown in Fig. 21 Ultimate Limit State (ULS) was
considered for the purposes of assessment.
CHI
C C
HEM280
Fig. 21.Repairing of Barra Bridge
71
Table 12 and Table 13 present the moment capacity of the bridge at the critical
locations before and after repair as calculated in this project and by a previous
investigation in Portugal respectively.
Table 12. Resistance capacity calculated in this project.
Resistance capacity before Resistance capacity after

reparation (kN.m) reparation (kN.m)
Pier 8 -319730 -364150
Mid-Span 20460 22100
Table 13. Resistance capacity calculated in an investigation in Portugal
t
previously.
iNe
Resistance capacity before Resistance capacity after
reparation (kN.m) reparation (kN.m)
Pier 8 -321437 -351667
Mid-Span 21001 24141
8.3 Steel Wharf, Nantes, France

rat
Instrumentation and analyses were carried out on a steel wharf in Nantes,
France.
Cases of instability related to instability of the soil mass along a rear line that
separates potential failure of the subgrade, instability of the front soil mass by
insufficient thrust sheet due to a too short strut (under-mining), breakdown of
elements of the curtain of insufficient strength, entrainment of the whole of the
propping and surrounding terrain with "great shift" whose fracture line is more or
Du
less distant from the contour support, inadequate bearing capacity of the curtain
under excessive vertical loads and excessive punching failure of the soil
beneath the base of the curtain and hydraulic instability due to water runoff
associated with height difference between the level of water in the basin and the
water table behind the screen that mainly depend on the length of the strut and
the permeability of soil layers crossed were considered.
During operation of the docks, the frequency of occurrence of such a danger of
instability may rise during an excavation at the foot of the curtain. The
schematic of the structure is presented below (Fig. 22).
Finite element models were performed through CAST3M and PLAXIS. The soil
was represented by a homogeneous material whose behaviour was
represented by a perfect Mohr-Coulomb elastoplastic model. Mindlin beam
theory was used for modelling beams. The modelling approach was validated
on a real instrumented structure (Hochstetten sheet piling) as shown in Fig. 23.
72
hc ht hw
hs
hl
hr
lt
H
f
t
iNe
Fig. 22. Schematic of a dock consisting of a sheet pile retaining wall built
and anchor-red at the top by perpendicular passive ties.
rat
Du
Fig. 23. Overview of the full-scale experience.
In general, the assumptions used to model the mechanical behaviour of

experimental Hochstetten sheet piles were satisfactory. The mechanical
behaviour of banks was satisfactorily modelled using finite elements.
The impact of corrosion of the mechanical behaviour of the structure was
studied. The spatial-temporal fields of corrosion were integrated to a
mechanical finite element model. The model helped predict loss of steel
thickness and the consequent effects of the response of the structure. The
influence of the variability of corrosion with depth was also modelled in this
regard.
Corrosion causes over time a decrease of geometric and mechanical
characteristics of the sheet piles making the retaining curtain more flexible. A
73
decrease in the bending moment is then observed in its free portion located in
the submerged area. In return, the normal force in the rods increases slightly,
together with the bending moment in the strut part and at the anchors.
Despite the influence of corrosion, the maximum bending moment is still in the
submerged area. Therefore, the area where corrosion is most severe is not
necessarily the area which is put the most under mechanical pressure.
The margin of safety of the structure is impacted when the structure is not or no
longer protected. Three probabilistic approaches were used for coupling with
the deterministic mechanical finite element model involving Monte Carlo
simulations, First Order Reliability Method (FORM) and the Method of Spectral
Stochastic Finite Element (SSFEM).
Corrosion in each exposure zone was modelled through a Gamma probability
t
law. Evolutions of statistical parameters (mean, 5 and 95 percent fractiles) of
the bending stress Sz (c) (absolute value) along quay T-S1 and after different
iNe
exposure times in port environment were obtained.
The limit state related to bending was formed and the safety margin was related
to the designed bending capacity and maximum oncoming bending moment.
The probability of exceeding the design bending stress Szd was then
calculated for the three exposure periods (t = 10, 25 and 50 years). Fig. 24
presents the change of statistical parameters of the bending stress over time.
0 Poutre de couronnement Szd t = 0 ans
rat
-2 ZT Moy. - t = 10 ans
ZL Q05%
-4 Q95%
Profondeur (m)
Moy. - t = 50 ans
-6 Q05%
Q95%
-8
ZI
-10
-12 ZM
-14
Du
ZG
-16
0 25 50 75 100 125 150
Contrainte (MPa)
Fig. 24 Change of statistical parameters of the bending stress Sz(C)

(absolute value) along the sheet pile wall for T-S1 setup Aafter 10,
25 and 50 years of exposure in port environment.
The time evolution of the calculated reliability indices at sensitive areas is

presented in Fig. 25.
After only a few years of exposure, the reliability index of the sheet piling in the
free part is less than the target reliability index (0 = 3.8). Therefore, the
structural safety becomes insufficient in the regulatory sense. This aspect
highlights the importance of planning and implementing at the outset effective
means (paint system, cathodic protection, sacrificial over thickness, etc.) to
protect the metal port works against corrosion.
74
5.5
Partie libre
5.0 Tirants d'ancrage

Indice de fiabilit
4.5
4.0
= 3,8
3.5
3.0
2.5
2.0
0 10 20 30 40 50 60
Dure d'exposition (annes)
t
Fig. 25. Time evolution of reliability index for the sheet pile wall in the
two areas identified as sensitive.
iNe
The parameter importance measures computed from these factors are
presented next in Fig.26.
Perte paisseur (ZT) Perte paisseur (ZL)
Perte paisseur (ZI) Perte paisseur (ZM)
Masse volumique sol sec Masse volumique sol djaug
Angle de frottement Cohsion
rat
Surcharge exploitation Limite lastique acier

100
90
Facteur d'importance (%)
80
70
60
50
40
Du
30
20
10
0
0 10 25 50
Dure d'exposition (annes)
Fig. 26. Important factors of random variables in the free part of the sheet
piling.
At t = 0 year, the most influential random variables are the friction angle
'(54 %), operating overloads q (22 %) and the elastic limit of steel e (11 %),
representing a total of 87 % of the sensitivity of reliability index . Then, for the
3 exposure times of sheet piling in port environment, the effects caused by
corrosion in the submerged zone (Z1) play a leading role on the sensitivity of
the reliability index. It is also interesting to note that the influence of thickness
loss in the submerged area decreases over time (about 70 % after 10 years and
75
only 40 % after 50 years of exposure). This decrease is explained by the fact

that corrosion acts favourably upon duress (stress) and adversely on resistance.
Corrosion of port metal structures leads to a loss of random steel thickness
which increases over time, the kinetics of which depending on the exposure
area in question. This loss of cohesion of the mechanical behaviour impacts on
the level of structural safety of these structures.
The most efficient methods in terms of accuracy and computing were selected
as:
The method of Monte Carlo where unreliability probabilities are higher
than the desired 10-2.
Method of Spectral Stochastic Finite Element (SSFEM) projection on
non-intrusive polynomial chaos when the number of random variables is
t
less than 5.
iNe
FORM method, especially when the problem involves a large number of
random variables (M = 10) and when the failure probabilities are very
small, of the order of 10-6. This method proved to be most effective
against the criteria listed above. The probabilities of failure obtained
using different methods are very close.
The results of finite elements simulations show that corrosion impacts both
stress and structural strength. The effects of resistance to corrosion are
unfavourable since they entail a reduction of the geometrical characteristics of
rat
the piles. In parallel, the loss of steel thickness increases the bending of the
sheet pile wall, causing a decrease in the bending moment in the free part and
a respective increase at the level of anchors.
However, whatever the depth of the retaining wall, the adverse effects
predominate, leading to an increase in stress due to bending moment. The two
areas that appear to be most sensitive to the coupling effect "pathology -
mechanical behaviour are located at the anchor and pulling in the submerged
Du
area, at the maximum bending moment. It is interesting to note that the area of
lowest water in which corrosion is most severe is not the most mechanically
stimulated area, which is beneficial in terms of margins of safety for the
damaged structures.
These trends are confirmed by the results of the mechanical reliability study.
When a building is not effectively protected against corrosion, it is only after
several years of exposure in port environment that its structural reliability no
longer meets regulatory requirements, particularly in the submerged area, at the
maximum bending moment. This area is particularly sensitive to the effects of
corrosion because the geometrical and mechanical characteristics of the sheet
pile wall are sized relative to the demands therein. At the anchors, the margin of
safety is more important, which explains that reliability is observed during the
first 20 years of exposure in the case treated. In areas less stressed, the effects
of corrosion on structural reliability are negligible given the large margin of
safety due to the homogeneity of the geometrical and mechanical
characteristics of sheet pile wall along its entire height.
76
This analysis, therefore, highlights two areas in which measurements of loss of

steel thickness must be made with respect to the criterion of limit areas studied.
Moreover, it is important to note that the optimal maintenance strategy must
take into account all the criteria of limit areas: sheet piling openings, horizontal
movement of the head curtain, etc.
Finally, recommendations are given regarding the parameters used for sizing
and/or re-evaluation of structural safety of port facilities. The influence of
corrosion effects, the geotechnical parameters (friction angle and cohesion) and
excessive operation, the sensitivity of reliability index are also evidenced
through the importance factors. Thus, given their variability, these parameters
must be modelled by random variables. It is also essential that managers define
their risk management processes with appropriate actions to reduce the
uncertainties associated with these parameters, including:

t
a better characterization of geotechnical
establishment of a measurement protocol;
parameters for the
iNe
tighter controls on conditions of use of wharves and sensitization of
managers;
characterization of the effects of corrosion over time, especially in the 2
areas identified as sensitive.
rat
Du
77
9 Conclusions
This volume addresses the core issues of assessment and maintenance
planning in infrastructure maintenance management. In this regard, a
probabilistic format has been put forward as the most appropriate marker for the
performance of the built infrastructure elements. Consequently, the
management of safety of the infrastructure has been illustrated through the
management of reliability index.
The condition assessment of an as-built infrastructure has been observed to be
an interrelated function of structural inspections, tests (destructive, semi-
destructive and non-destructive), deterministic analyses, semi-probabilistic
analyses and full probabilistic analyses. The importance of inspections and
tests has also emphasized the importance of training of personnel in relation to
t
these works. A computerised management system for the infrastructure
elements has been acknowledged to be a key component towards the
iNe
realisation of cost optimised maintenance management. In this connection, the
requirement of synthesising information from different levels has been observed.
The importance and the availability of more sophisticated cost formats have
also been noted. In this regard, the concept of cross asset management
considering the requirements of multiple stakeholders may be important. The
optimised maintenance management philosophy for infrastructure elements
seems to be moving towards a network optimised format that optimisation for
the sub-elements of the network.
rat
Assessment of structures may be done at a number of levels. The detail, time,
sophistication and cost of such levels become more complex at higher levels.
An extremely detailed assessment, like a full probabilistic assessment is
recommended only when the traditional methods and lower level assessments
methods do not work, but indicate that a higher level of re-analysis is required.
In this regard, the global safety factor, partial safety factor reliability and socio-
economic formats of assessments are discussed.
Du
The development of structural response models has been related to the

requirement of assessment, available data and results from continuous
monitoring and testing. In this regard, the modelling of load may be carried out
in a structure specific fashion, when necessary. The extreme loads may be
established from partially available data and an extreme value distribution fit or
even from direct simulation. The importance of modelling the extreme values in
the form of an appropriate and correct probabilistic description of the tail of such
distribution is emphasized on. The uncertainty related to materials has been
discussed. In particular, this TG has noted the importance of the damage
models or degradation models of materials in infrastructure ratings and
consequently, their final effect on the decisions taken for the entire
infrastructure. Various forms of degradation models have been discussed. The
difficulty and the impact of the uncertainty associated with post-repair
deterioration models of materials have been discussed.
78
An overview of reliability analysis has been presented in this technical guide.

The relationship of target reliability level with economic, social and sustainability
considerations are observed in this regard. The treatment of various
uncertainties, including uncertainties related with the model itself is presented in
this format. The incorporation of supplementary information through testing,
new data, inspection or assessments is also discussed leading to updating or
reliability indices of infrastructure elements as individuals or for the network as a
whole.
t
iNe
rat
Du
79
10 References
[1] DITLEVSEN, O. and H.O. MADSEN. Structural reliability methods.
[Chichester, UK]: John Wiley and Sons. 1996.
[2] ZNIDARIC, A. et al. A comparison of road structures of six EU countries.

ICE, Journal of urban planning and design. 2011.
[3] ENRIGHT, M.P. and D.M. FRANGOPOL. Maintenance Planning for

Deteriorating Concrete Bridges. ASCE. Journal of structural engineering,
1999. Vol. 125, Issue, 12, pp. 1407-1414.
[4] RYALL, M.J. Bridge management system. Elsevier, 2010.
t
[5] ERTEKIN, A.O., H.H. NASSIF and K. OZBAY. bAn LCCA algorithm for
bridges: a comprehensive approach. TRB (Transportation Research
iNe
Board), 87th annual meeting, Washington D.C., January 13-17, 2008.
[6] Maintenance and repair of transport infrastructures Technical Guide.

Part IV Concrete structures. Vol. 3 Testing techniques. DURATINET
project WG A3. Lisbon, LNEC, 2012.
[7] NCHRP (National Cooperative Highway Research Program). Report 483 -

Bridge Life-Cycle Cost Analysis. Washington D.C., 2003.
rat
[8] FRANGOPOL, D.M., J. KONG and E. GHARAIBEH. Reliability based life-
cycle management of highway bridges. ASCE. Journal of computing in
civil engineering, 2001. Vol. 15, Issue, 1, pp. 27-34.
[9] FRANGOPOL, D.M and M. LIU. Maintenance and management of civil

infrastructure based on condition, safety, optimization, and life-cycle cost.
Structure and infrastructure engineering. Taylor & Francis, 2007. Vol. 3,
Issue, 1, pp. 29-41.
Du
[10] BRANCO, F. A. and J. BRITO. Decision criteria for concrete bridge repair.
Structural engineering international, 1995, Number 2.
[11] SCHOEFS, F., et al. Comparison of additional costs for several

replacement strategies of randomly ageing reinforced concrete pipes.
Computer aided civil and infrastructure engineering, 2009. Vol. 24, Issue 7,
pp. 492-508.
[12] BASTIDAS-ARTEAGA, E. et al. Probabilistic evaluation of the

sustainability of maintenance strategies for RC structures exposed to
chloride ingress. International journal of engineering under uncertainty -
Hazards, assessment and mitigation (IJEUU: HAM). Special Issue, 2010.
[13] NCHRP (National Cooperative Highway Research Program). Report 590 -

Multi-objective optimization for bridge management systems. Washington
D.C., 2007.
80
[14] SAMARIS PROJECT. Sustainable and advanced materials for road

infrastructure - FP5. 2006.
[15] THOMPSON, P. D. The Pontis bridge management system. Structural

engineering international, 1998, pp. 303-308.
[16] ROBERT, W.E, et al. Pontis bridge management system state of the
practice in implementation and development. Transportation research
board of national academies: 9th International bridge management
conference. [Orlando, Florida, U.S.A], 2003, pp. 49-60.
[17] CZEPIEL, E. Bridge management systems literature review and search.

Northwestern University. 1995.
[18] HAWK, H. BRIDGIT. User-friendly approach to bridge management.
t
Transportation research circular, 498 - Volume I. [Denver, Colorado
U.S.A], 1999.
iNe
[19] VESIKARI, E. and M. SDERQVIST. Life-cycle management of concrete
infrastructures, for improved sustainability. Transportation research board
of national academies: 9th international bridge management conference.
[Orlando, Florida, U.S.A], 2003, pp.15-28.
[20] PIARC. Technical committee on road bridges and other structures. 2004.
[21] EN 1990:2002/ A1:2005/ AC: 2010, Eurocode 0: Basis of structural design.

rat
[22] COST 345. Procedures required for assessing highway structures. Joint
report of working groups 2 and 3, methods used in European states to
inspect and assess the condition of highway structures. European
Cooperation in the Field of Scientific and Technical Research, 2002.
[23] MEDACHS (Marine Environment Damage to Atlantic Coast Historical and

Transport Structures and Buildings). Methods of assessment, repair,
Du
maintenance - Nr. 197, INTERREG IIIB. Atlantic Area, Final Report. 2008.
[24] ARCHES. Assessment and rehabilitation of central european highway

structures. Executive summary report, 2010.
[25] SUSTAINABLE BRIDGES PROJECT. Guideline for load and resistance

assessment of existing European railway bridges. Advices on the use of
advanced methods, 2007.
[26] RUCKER, W., F. HILLE and R. ROHRMANN. Guideline for the

assessment of existing structures. Final Report - SAMCO project. 2006.
[27] KHAN, M.A. Bridge and high structure rehabilitation and repair. The
McGraw-Hill Companies Inc., 2010.
[28] BROWNJOHN, J.M.W, et al. Assessment of highway bridge upgrading by

dynamic testing and finite-element model updating. ASCE. Journal of
bridge engineering, 2003. Vol. 8, Issue 3, pp. 162-172.
81
[29] ENRIGHT, M.P. and D.M. FRANGOPOL. Service-life prediction of

deteriorating concrete bridges. ASCE. Journal of structural engineering,
1998. Vol. 124, Issue, 3, pp. 309-317.
[30] ESTES, A.C and D.M. FRANGOPOL. Minimum expected cost-oriented

optimal maintenance planning for deteriorating structures: application to
concrete bridge decks. Reliability Engineering and System Safety, 2001.
Vol. 73, Issue, pp. 281-291.
[31] AKGUL, F. and D.M. FRANGOPOL. Rating and reliability of existing

bridges in a network. ASCE. Journal of bridge engineering, 2003. Vol. 8,
Issue 6, pp. 383-393.
[32] FU, G. and C. FU. Bridge rating practices and policies for overweight
vehicles. A synthesis of highway practice NCHRP synthesis 259.
t
Transportation research board, Washington D.C., 2006.
iNe
[33] STRAUB, D. and M.H FABER. Risk based inspection planning for
structural systems. Structural safety, 2005. Vol. 27, Issue 4, pp. 335-355.
[34] ESTES, A.C and D.M. FRANGOPOL. Updating bridge reliability based on
bridge management systems visual inspection results. ASCE. Journal of
bridge engineering, 2003. Vol. 8, Issue 6, pp. 374-382.
[35] LIU, M. and D.M. FRANGOPOL. Optimal bridge maintenance planning

based on probabilistic performance prediction. Engineering structures,
rat
2004. Vol. 26, Issue 7, pp. 991-1002.
[36] RAMNICEANU, A. Correlation of corrosion measurements and bridge

conditions with NBIS deck rating. Masters Thesis, Virginia Polytechnic
and State University, USA. 2004.
[37] MAES, M.A. Updating performance and reliability of concrete structures

using discrete empirical Bayes methods. ASME. Journal of offshore
Du
mechanics and arctic engineering, 2002. Vol. 124, pp. 239-244.
[38] SCHERER, W.T. and D.M. GLAGOLA. Markovian models for bridge
maintenance management. Journal of transportation engineering, 1994.
Vol. 120, Issue 1, pp. 37-51.
[39] OCONNOR, A. and I. ENEVOLDSEN. Probability based bridge

assessment. ICE. Journal of bridge engineering, 2007. Vol. 160, Issue 3,
pp. 129-137.
[40] BRIME. Final Report - DELIVERABLE D14. 2001. Contract Nr. RO-97-
SC.2220
[41] BASTIDAS-ARTEAGA, E. et al. A comprehensive probabilistic model of

chloride ingress in unsaturated concrete. Engineering Structures, 2011.
Vol.33, Issue 3, pp. 720730.
82
[42] CIVJAN, S.A. et al. Method to evaluate remining prestress in damaged

prestress girders. Research Report Nr. 1370-2, Research Project 3-5-
93/5-1370, Centre for Transportation Research, Bureau of Engineering
Research, University of Texas, Austin. 1995.
[43] ONYEMELUKWE, O.U, R.E. MOUSSA-ISSA and C.J. MILLS. Field

measured prestress concrete losses versus design code estimates.
Experimental mechanics, 2003. Vol. 43, Issue 2, pp. 201-215.
[44] SKELTON, S.B. and J.A. RICHARDSON. A transducer for measuring

tensile strains in concrete bridge girders. Experimental mechanics, 2006.
Vol. 46, pp. 325-332.
[45] PAKRASHI, V. et al. Monitoring and repair of an impact damaged

prestressed concrete bridge. Proceedings of the institute of civil engineers.
Journal of bridge engineering, 2011.
t
iNe
[46] FARRAR, C.R, S.W. DOEBLING and D.A. NIX. Vibration based structural
damage identification. Philosophical transactions of the royal society,
2001. Vol. 359, Issue, 1778, pp. 131-149.
[47] CARDEN, P.E. and P. FANNING. Vibration based condition monitoring: A

review. Structural health monitoring, 2004. Vol. 3, Issue 4, pp. 355-377.
[48] MOYO, P. and J.M.W. BROWNJOHN. Detection of anomalous structural

behaviours using wavelet analysis. Mechanical systems and signal
rat
processing, 2002. Vol.16, Issues 2-3, pp. 429-445.
[49] OMENZETTER, P., J.M.W. BROWNJOHN and P. MOYO. Identification of

unusual events in multi-channel bridge monitoring data. Mechanical
systems and signal processing, 2004. Vol. 18, pp. 409-430.
[50] BILELLO, C. and L.A. BERGMAN. Experimental vibration of damaged

beams under a moving mass: theory and experimental validation. Journal
Du
of sound and vibration, 2004. Vol. 274, Issues 3-5, pp. 567-582.
[51] BILELLO, C., L.A. BERGMAN and D. KUCHMA. Experimental

investigation of a small-scale bridge model under a moving mass. ASCE.
Journal of structural engineering, 2004. Vol. 130, pp. 799-804.
[52] DOEBLING, S.W. and C.R. FARRAR. The state of the art in structural
identification of constructed facilities. ASCE report - Committee on
Structural Identification of Constructed Facilities. 1999.
[53] SHEILS, E. et al. Investigation of the effect of the quality of inspection

techniques on the optimal inspection interval for structures. Structure and
infrastructure engineering. 2011.
[54] SHEILS, E. et al. Development of a two-stage inspection process for the

assessment of deteriorating bridge structures. Reliability engineering and
system safety, 2010. Vol. 95, Issue 3, pp. 182-194.
83
[55] BREYSSE, D. et al. Accounting for variability and uncertainties in NDT

condition assessment of corroded RC-structures. European journal of
environmental and civil engineering - Durability and maintenance in
marine environment, 2009. Vol. 13 Issue 5, pp. 573592.
[56] CASTILLO, E. Extreme value theory in engineering. Academic Press.

1991.
[57] MAES, M.A. Tail heaviness in structural reliability. Proceedings of ICASP

95. CERRA ICASP, Paris, 1995.
[58] CASTILLO, E. and J.M. SARABIA. Engineering analysis of extreme value

data: selection of models. Journal of Waterway, Port, Coastal and Ocean
Engineering, 1992. Vol. 118, Issue 2, pp. 129-146.
t
[59] GUMBEL, E.J. Statistics of extremes. Columbia University Press. 1958.
iNe
[60] ANG, A.H.S. and W.H. TANG. Probability concepts in engineering
planning and design - Vol. 1, Basic Principles. New York: John Wiley,
1975.
[61] BAILEY, S.F. Basic principles and load models for the structural safety
evaluation of existing road bridges. Ph.D. Thesis Nr. 1467, cole
Polytechnique Fdrale de Lausanne, Switzerland, 1996.
[62] OCONNOR, A. Probabilistic load modelling for highway bridges. PhD

rat
Thesis, Trinity College Dublin, Ireland, 2001.
[63] OCONNOR, A. et al. Report of current studies performed on normal load

model of EC1- Traffic loads on bridges. RFGC. Hermes Science
Publications, 2001. Vol.5, Issue 4, pp. 411-434.
[64] O BRIEN, E.J et al. Procedures for the assessment of highway structures.
Proceedings of the institute of civil engineers Journal of transport, 2005.
Du
Vol. 158, Issue 1, pp. 17-25.
[65] JACOB, B. Assessment of the accuracy and classification of weigh-in-

motion systems: part 1 - Statistical background. Heavy Vehicle systems,
International journal of vehicle design, 2000. Vol. 7, Issues 2-3, pp. 136-
152.
[66] COST 323. Weigh in motion of road-vehicles Final report, Appendix 1.

European WIM specification. 1999.
[67] JACOB, B., E. OBRIEN and W. NEWTON. Assessment of the accuracy

and classification of weigh-in-motion systems: part 2 - European
specification. Heavy Vehicle systems, International journal of vehicle
design, 2000. Vol. 7, Issues 2-3, pp.153-168.
[68] O BRIEN, E.J, A. OCONNOR and A. GONZALEZ. Eurocode for Traffic

Loads on Bridges (EC 1.3) Calibration for Irish Conditions. Department
84
of Civil, Structural and Environmental Engineering, Trinity College,

Research Report Nr. 98-001. 1998.
[69] EYMARD, R. and B. JACOB. Un nouveau logiciel: le programme castor

pour le calcul des actions et des solicitations du traffic dans les ouvrages
routiers. Bulletin Liaison des LCPC, 1989. Nr. 144, pp. 7180.
[70] NOWAK, A. S. Live load model for highway bridges. Journal of structural
safety, 1993. Vol. 13, Issues 1-2, pp. 5366.
[71] GETACHEW, A. and E.J. OBRIEN. Sensitivity of predicted bridge traffic

load effects to the tails of truck weight distributions. PARKE, G.A.R. and P.
DISNEY, eds. Bridge Management 5, proceedings of the 5th International
conference on bridge management. London: Thomas Telford, 2005, pp.
275-282.
t
[72] DITLEVSEN, O. and H.O. MADSEN. Structural reliability methods.
iNe
[Chichester, UK]: John Wiley and Sons, 1996.
[73] O BRIEN, E.J, A. HYRAPETOVA, and C. WALSH. Microsimulation

modelling of traffic loading on medium and long-span road bridges. KOH,
H.M. and D.M FRANGOPOL, eds. Proceedings of the 4th International
conference on bridge maintenance, safety and management, Seoul. 2008.
[74] O BRIEN, E.J, B. ENRIGHT and A. GETACHEW. Importance of the tail in

truck weight modelling for bridge assessment. ASCE. Journal of bridge
rat
engineering, 2010. Vol. 15, Issue 2, pp. 210-213.
[75] DEMPSEY, A., V. PAKRASHI, and A. OCONNOR. Probabilistic design

and assessment of transport infrastructure - TRA2010. [Brussels,
Belgium], 2010.
[76] DMRB 3.4.14/ BD 44/95, The assessment of concrete highway bridges

and structures, 1995.
Du
[77] DMRB 3.4.3/ BD 21/01, The assessment of highway bridges and

structures, 2001.
[78] ELLINGWOOD, B., et al. Development of a probability based load

criterion for American national standard A58. Building code requirements
for minimum design loads in buildings and other structures. US national
bureau of standards special publication 577. Washington, D.C., 1980.
[79] CEB-FIP. Reliability of concrete structures. CEB Bulletin - 202, Lausanne,

1991.
[80] TONG, C. Quantifying uncertainties of a soil-foundation structure-

interaction system under seismic excitation. Report LLNL-TR-402883,
Lawrence Livermore National Laboratory. 2008.
[81] BRESYSSE, D. et al. Generic approach of soil-structure interaction

considering the effects of soil heterogenity. Symposium in print - Risk and
85
variability in geotechnical engineering. Geotechnique, 2001. Vol. LV,

Issue 2, pp. 143-150.
[82] MIRZA, A.S., M. HATZINIKOLAS and J.G. MACGREGOR. Statistical

descriptions of strength of concrete. ASCE. Journal of the structural
division, 1979. Vol. 105, Issue 6, pp. 1021-1037.
[83] MIRZA, A.S. and J.G. MACGREGOR. Probabilistic study of strength of

reinforced concrete members. Canadian journal of civil engineering, 1982.
Vol.9, pp. 431-444.
[84] SCOTT, B., B.J. KIM and R. SALGADO. Assessment of current load
factors for use in geotechnical load and resistance factor design. ASCE.
Journal of geotechnical and geoenvironmental engineering, 2003. Vol.
129, Issue 4, pp. 287-295.
t
[85] NOWAK, A. S. Load model for bridge design code. Canadian journal of
iNe
civil engineering, 1994. Vol. 21, pp. 36-49.
[86] COLLINS, M.P., F.J. VECCHIO and G. MEHLHORN. An international

competition to predict the response of reinforced concrete panels.
Canadian Journal of Civil Engineering, 1985. Vol. 12, pp. 624-644.
[87] TURAN, O.T., C. HIGGINS and D. ROSOWSKY. Statistical modelling of

coupled shear-moment resistance for RC bridge girders. Journal of bridge
engineering, 2008. Vol.13, Issue 4, pp. 351-361.
rat
[88] THOMPSON, P.D., et al. The Pontis bridge management system.
Structural engineering international, 1998, pp. 303-308.
[89] CHEN, C.J. and D. JONSTON. Bridge management under a level of

service concept providing optimum improvement action, time and budget
prediction. FHWA/NC/88-004, Center for transportation studies, civil
engineering department, North Carolina state university. 1987.
Du
[90] SHAHIN, M. Pavement management for airports, roads and parking lots.
Chapman & hall, 1994.
[91] BUSA, G. A national bridge deterioration model. Report to the U.S.

Department of transportation, research and special programs
administration, transportation systems center kendall square. Cambridge,
MA. 1985.
[92] FHWA (Federal Highway Administration). Bridge management systems.

Report FHWA-DP-71-01R. U.S. Department of transportation.
Washington D.C., 2001.
[93] LOUNIS, Z. Network-level bridge management using a multi-objective

st
optimization decision model. 1 CSCE Specialty conference on
infrastructure technologies, management and policy, [Toronto, Ontario,
Canada], June 2-4, 2005, pp. 121-1 to 121-9.
86
[94] KAYSER, J. and A. NOWAK. Capacity loss due to corrosion in steel-

girder bridges. ASCE. Journal of structural engineering, 1989. Vol. 115,
Issue 6, pp. 1525-1537.
[95] MIYAMOTO, A., K. KAWAMURA and H. NAKAMURA. Practical

applications of a bridge management system in Japan. Transportation
research record, 1999. Vol.1696, pp. 14-25.
[96] STUKHART, G. et al. Study for a comprehensive bridge management

system for Texas. Research report 1212-1F, prepared for the Federal
Highway Administration (FHA), Austin, TX. 1991.
[97] COST 521. Corrosion of steel in reinforced concrete structures, final

report. CIGNA, R. et al., eds. European Communities, EUR 20599,
Luxembourg. 2003, pp.6072.
t
[98] BROWNE, R. D. Practical considerations in producing durable concrete.
iNe
Seminar on the improvement of concrete durability, ICE, 1986.
[99] MCCARTER, W.J, H. EZIRIM and M. EMERSON. Absorption of water

and chloride into concrete. Magazine of concrete research - 44, 1992. Nr.
158, pp. 31-37.
[100] MCCARTER, W.J, M. EMERSON and H. EZIRIM. Properties of concrete

in the cover zone: developments in monitoring techniques. Magazine of
concrete research - 47, 1995. Nr. 172, pp. 243-251.
rat
[101] TANG, J.L. and J.N. WANG. Measurement of chloride ion concentration
with long period grating technology. Smart materials and structures, 2007.
Vol. 16, pp. 665-672.
[102] BROWNE, R. D. Practical considerations in producing durable concrete.

Seminar on the improvement of concrete durability, ICE, 1986.
Du
[103] BERTOLINI, L. Corrosion of steel in concrete - Prevention, diagnosis,

repair. Wiley, VCH, 2004. Chapter 17, pp. 300-306.
[104] BROOMFIELD, J. P. Corrosion of steel in concrete, understanding,

investigation and repair. London: E & F.N Spon, 2007.
[105] VAN DE TOORN, A. and A.W.F. REIJ. A systematic approach to future

st
maintenance. 1 International conference on bridge management,
University of Surrey, Guildford, UK. London: Elsevier, 1990, pp. 215-222.
[106] COST 509. Corrosion and protection of metals in contact with concrete,
final report. COX R.N. et al, eds. European Commission, Directorate
General Science, Research and Development, Brussels, EUR 17608.
1997.
[107] SOBANJO, J. A neural network approach to modelling bridge

deterioration. ASCE. Proceedings of the 4th congress on computing in
Civil engineering. [Philadelphia, PA, U.S.A], 1997, pp. 623-626.
87
[108] MELCHERS, R. E. Examples of mathematical modelling of long term

general corrosion of structural steels in sea water. Corrosion engineering
science and technology, 2006. Vol. 41, Issue 1, pp. 38-44.
[109] MELCHERS, R. E. and R.J. JEFFREYR. Probablistic models for steel

corrosion loss and pitting of marine infrastructure. Reliability engineering
& system safety, 2008. Vol. 93, Issue 3, pp. 423-432.
[110] MELCHERS, R. E. and C.Q. LI. Reinforcement corrosion initiation and

activation times in concrete structures exposed to severe marine
environments. Cement and concrete research, 2009. Vol. 39, Issue 11, pp.
1068-1076.
[111] MORCOUS, G. Case-based reasoning for modelling bridge deterioration.

Ph.D. thesis, Civil and environmental engineering, Concordia University,
t
Canada. 2000.
iNe
[112] MORCOUS, G., Z. LOUNIS and M. MIRZA. Identification of environmental
categories for Markovian deterioration models of bridge decks. ASCE.
Journal of bridge engineering, 2003. Vol. 8, Issue 6, pp. 353-361.
[113] SRINIVASAN, S. and K. MEHTA. Stochastic process. McGraw-Hill Inc.

1978.
[114] ELACHACHI, S.M. and D. BREYSSE. Comparison of relative risk based

model and artificial neural networks for sewer assets maintenance
rat
management. International forum on engineering decision making. [Port
Stephens, Australia], 12-15 December 2007.
[115] JIANG, Y., M. SAITO and K. SINHA. Bridge performance prediction model
using Markov chain. Transportation research record, 1988. Vol. 1180, pp.
25-32.
[116] JIANG, Y. and K. SINHA. Bridge service life prediction model using the
Du
Markov chain. Transportation research record (TRB), National Research

Council, Washington, D.C., 1989. Vol. 1223, pp. 24-30.
[117] MADANAT, S., M. KARLAFTIS and P. MCCARTHY. Probabilistic

infrastructure deterioration models with panel data. ASCE. Journal of
infrastructure management systems, 1997.Vol. 3, Issue 1, pp. 120-125.
[118] COLLINS, L. An introduction to Markov chain analysis. CATMOG, Geo

Abstracts Ltd. University of East Anglia, Norwich. 1972.
[119] CESARE, M. et al. Modelling bridge deterioration with Markov chains.

ASCE. Journal of transportation engineering, 1992. Vol. 118, Issue 6, pp.
820-833.
[120] TOKDEMIR, O.B, C. AYVALIK and J. MOHAMMADI. Prediction of

highway bridge performance by artificial neural networks and genetic
88
algorithms. Proceeding of the 17th International symposium on automation

and robotics in construction (ISARC). [Taipei, Taiwan], 2000.
[121] LEE, J.H., et al. Improving the reliability of bridge management systems
using a backword prediction model. Journal of automation in construction,
2008. Vol. 17, Issue 6, pp. 758-772.
[122] BOUSSABINA, A. The use of artificial neural networks in construction

management: A Review. Construction Management and Economics.
International journal of construction management & technology, 1996. Vol.
14, pp. 427-436.
[123] ARDITI, D. and O. TOKDEMIR. Comparison of case based reasoning and

artificial neural networks. ASCE. Journal of computing in civil engineering,
1999. Vol. 13, Issue 3, pp. 162-169.
t
[124] YANEV, B. and C. XIAOMONG. Life cycle performance of the New York
iNe
city bridges. New York department of transportation: New York, NY. 1993.
[125] BOLUKBASI, M., D. ARDITI and J. MOHAMMADI. Deterioration of

reconstructed bridge decks. Structure and infrastructure engineering,
2006. Vol. 2, Issue 2, pp. 23-31.
[126] MELCHERS, R. E. Structural reliability analysis and prediction. New York:

John Wiley, 1999.
rat
[127] NOWAK, A. S. and K.R. COLLINS. Reliability of structures. New York:

McGraw-Hill, 2000.
[128] STEWART, M. G. and D.V. ROSOWSKY. Time dependent reliability

analysis of deteriorating reinforced concrete bridge decks. Structural
safety, 1998. Vol. 20, pp. 91-109.
[129] HALDAR, A. and M. SANKARAN. Probability, reliability, and statistical

Du
methods for engineering design. John Wiley & Sons, Inc., 2000.
[130] ISO 2394:1998, General principles on reliability for structures.
[131] ISO 13822:2010, Bases for Design of structures-assessment of existing

structures.
[132] JCSS. Probabilistic model code. The joint committee on structural safety.
2001.
[133] EN 1991-1:2002, Eurocode 1: Actions on structures - Part 1-1: General

actions - Densities, self-weight, imposed loads for buildings.
[134] NKB (Nordic Committee for Building Structures). Recommendation for

loading and safety regulations for structural design. NKB report Nr. 36,
1978.
89
[135] ALLEN, D. E. Canadian highway bridge evaluation: reliability index.

Canadian journal of civil engineering, 1997. Vol.19, Issue 6, pp. 987991.
[136] NKB (Nordic Committee for Building Structures). Recommendation for

loading and safety regulations for structural design. NKB report Nr. 55,
1987.
[137] BS 5400-4: 1990, Steel, concrete and composite bridges - Part 4: Code of
practice for design of concrete bridges.
t
iNe
rat
Du
90

Management Infrastructure

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Management Infrastructure

Uploaded by

Copyright:

Available Formats

maintenance and repair of transport infrastructure

Copyright LABORATRIO NACIONAL DE ENGENHARIA CIVIL, I. P.

1st. edition: 2012

Printing: 100 copies

Descriptors: Transport infrastructures / Steel structure / Reinforced concrete structure /

ISBN 978-972-49-2237-9 (paperback)

Final revision by Editorial Commission member

6.1 Uncertainty in material properties .......................................................42

consequences of this failure. A failure is defined as a non-compliance with one

The marker of reliability of a structure under this prevailing uncertainty and

against unit investment looking at whole life safety. To summarise,

decisions. Not all of the participating probabilistic variables in infrastructure

significantly. This trend is only to gain stronger with time.

requirements emanating purely from an engineering/scientific viewpoint. It is

Single span, 32 year old segmental prestressed concrete

Seoul 1992 span. One dead, seven injured.

A true descriptor of the performance of a structure allows any number of built

infrastructure network is thus a key global challenge. An assessment relating to

attempts to define the users (public, bridge inspectors, bridge administrators,

inspections and material testing, performance testing and monitoring (structural

bridge: - different types - required activities; predictions:

capacity; - suggestion for - cost benefit

bridge (project) level bridge (project) level

the benefits. It is to be noted here that even before a preselected performance

models are implemented in graphical design software that helps bridge

total replacement. The MRR&I approach is generated through an investigation

10 years of deliberate inaction following a project, and the normal uncertainty in

normal conditions, can happen on specific bridges. It is not necessary to predict

timing based on a utility function (or a combination of both). Optimal scoping is

deterioration rate is determined by condition data such as the inspection rating

by neglecting maintenance or replacement measures. However, using Pontis

asset management targets all different types of stakeholders and consequently

related to the system, is unique and amenable to interpolation, extrapolation or

Fig. 3.Infrastructure management through reliability management.

Traditional approaches of assessing structures are based upon deterministic

indeed within limits of safety even when it is deemed unacceptable from a

for infrastructure owners/managers tasked with optimising their resource.

This approach can be considered to be representative and is taken up in a

hierarchical by resource and information limitations.

The complexity and performance requirements of infrastructure maintenance

Fig. 4. Schema of infrastructure maintenance Management.

Fig. 5. Life-cycle of a bridge, adapted from Khan. [27]

and 6 is formally realised in Chapter 7. This chapter takes up the overview,

deterioration of a structure. Such preliminary assessments attempt to isolate

The specific methodology and exact hierarchical nature of inspection based

probabilistic or full probabilistic assessments can be carried out independent to

Level V Assessment: Reliability analysis is applied to a particular

level. Some alternative levels of assessment can be considered within the

3.1 Global safety factor format

Fig. 7. Structural assessment levels. [26]

3.2 Partial safety factor format

the importance of the consequences of failure, level of control during the

3.3 Reliability formats

Reliability methods taking into consideration the uncertainties of variables are

taking into account site specific measurements of loading and resistance

distributions may be found to be appropriate to model the load and resistance

3.4 Socio-economic format

4 Structural response modelling

4.1 Methods of structural analysis

analysis while assessment at higher levels involves more refined methods of

4.2 Integration of field data and structural models

Reliability analysis based on probabilistic models

simple grillage, FEM (upstand

tapered linear elastic model if

CC3(RC3) 5.2 (1.010-7) 4.3 (8.510-6)