A BRIDGE MANAGEMENT STRATEGY BASED ON FUTURE

RELIABILITY AND SEMI-MARKOV DETERIORATION MODELS

F. Bortot, D. Zonta & R. Zandonini, University of Trento, Italy

ABSTRACT
This paper introduces the outline of a bridge management strategy based on the prediction of future
bridge reliability using a semi-Markov deterioration model. This approach applies equally well to single
bridges and to a whole network, and is expected to be implemented in the Bridge Management System
(BMS) of the Autonomous Province of Trento (APT). More in detail, the model assumes the lifespan of a
bridge to be divided into five conventional condition states, and waiting times in each state are random
variables with known distributions. Mean waiting times and probability distribution parameters are
currently estimated based on the information stored in the APT's BMS. Monte Carlo simulations are used
to calculate the cumulative waiting time distributions, from which the semi-Markov transition probability
matrices are derived. The transition probabilities are age dependent: older bridges have a higher
probability of deteriorating to the next condition state in the given time interval. Once calibrated, the
deterioration model allows calculation of the time variant capacity function, in terms of probabilistic
initial capacity and degradation function. The prioritization is based on the principle whereby priority is
given to those actions that, within a certain budget, will minimize the risk of occurrence of an
unacceptable event in the whole network during the considered time interval. Sample results of the
prioritization as applied to the APT stock are then discussed.
1. INTRODUCTION step procedures that follow the BRIME research
project guidelines [2]. These procedures are of
Bridge Management Systems (BMSs) are tools increasing refinement and include explicit
designed to help bridge managers keep track of the probabilistic analyses at the higher evaluation
bridge stock characteristics, conditions and steps. The prioritization is based on the principle
serviceability. In addition to this, the most recent whereby priority is given to those actions that,
BMSs feature structural reliability assessment and within a certain budget, will minimize the risk of
decision making tools capable of analyzing occurrence of an unacceptable event in the whole
maintenance plans and total life-cycle costs. The network during the considered time interval,
fundamental goal of these instruments is to allow usually assumed equal to 50 years. The
the owner to establish an effective operation optimization module for the choice of an effective
strategy for the stock as a compromise between maintenance strategy has been recently developed
many conflicting technical and social objectives, and is now undergoing a test program, but is not
such as: minimization of life-cycle cost, yet included in the APT BMS.
minimization of probability of failure, In this paper, we discuss the operation of this
maximization of network performance. system, we present the models used to predict
The Autonomous Province of Trento (APT), future bridge performance and we illustrate some
Italy, has recently adopted a Bridge Management sample results obtained from the prioritization
System entirely based on these concepts. The process.
system operates on the web, and includes modules
for condition state evaluation, safety assessment 2. THE APT BRIDGE STOCK AND
and prioritization. Condition appraisal is based on MANAGEMENT SYSTEM
visual inspection, and acknowledges the general
rules of the AASHTO Commonly Recognized The APT is a mountainous region in the Italian
Standard Element system [1]. Alps. Currently, APT owns and manages
A conservative reliability assessment procedure approximately 2340 kilometres of roadways and
is carried out for each bridge in the stock and is 936 bridges. As to range of bridge types and ages,
based on the sole inspection data. When the the stock may be considered quite similar to other
condition of the bridge calls for a more detailed European stocks: most of the bridges were
evaluation, its reliability is evaluated using multi- constructed or reconstructed after the Second
World War, the age distribution diagram showing 74% of the entire stock. As to types, 65,1% of
a peak in the 70's. Also, reinforced concrete, APT bridges are RC or PRC simply-supported or
regular and pre-stressed, is by far the most widely continuous beams, 10% are RC arches, 19,9% are
utilized construction material, covering more than simple concrete or masonry arches, while only the
remaining 6% includes steel and steel-concrete in its original conditions. According to [1], each
composite bridges. element of the bridge is associated with a
The main characteristics of the Bridge condition state. The number of reference states
Management System are: foreseen by the procedures varies from 3 to 5,
 the system gives the owner not only a clear depending on the element. Each state is labelled
indication of the condition of each bridge, with a number, whereby 1 always indicates the
but also of its safety level, expressed in terms undamaged situation while the maximum value
of a reliability index; refers to the most severe damage situation. Since
 for each bridge a priority index is calculated for the condition state is univocally stated in the BMS
scheduling maintenance activity; procedures, to which all inspectors must conform,
 all information is provided in real-time; this can be seen as an objective and quantitative
 the system is fully operative on the web; measure of damage. An overall condition index of
inspectors and evaluators upload the results the bridge is also calculated by weighting all the
of the conditions and safety assessments elements condition states according to their
through a web-based interface; the manager importance, as better explained in the next Section.
can access the results of the analysis through Reliability Data refers directly to the load
the same web interface; carrying capacity of the bridge and consists of a
 all the subjects involved in management set of reliability indices, each associated with an
operations can directly interact with the system: ultimate limit state and a specific Structural Unit
DoT managers, DoT inspectors, professional or substructure.
engineers involved in the assessment procedure, Network level data includes all the information
external consultants; that is not related to a specific bridge, but is
 maintenance and upgrade of the system are relevant to the whole stock or to a group of
continuous and transparent to the users. bridges having the same structural system and/or
material. Network level data are for example the
The BMS is based on an SQL database that global scheduling of maintenance actions, the
collects the data for the whole stock of bridges. price list of interventions that defines the cost
The data is organized in project level data and model and the deterioration rates of the Markovian
network level data. Project level data are: matrices that define the deterioration model of
 Inventory Data; each Standard Element.
 Condition State Data; Inventory Data, Condition State Data and
 Reliability Data. Reliability Data are used as input for the
applications that perform network level analysis
Inventory Data includes all the information related operations and the results themselves are recorded
to bridge identification, geographical location and into the database.
features, administrative issues, construction and
previous retrofits. 3. DETERIORATION MODEL
Condition State data give a measure of the type
and severity of the deterioration of the structure. There have been many attempts to develop
The aim of the condition assessment of bridge analytical models representing the degradation of
structures is to detect whether a deterioration bridge structures starting from the knowledge of
process is going on, and, if so, to evaluate the the deterioration process of materials [3]. These
degree of deterioration, with respect to the bridge models are typically very complex, as they attempt
to represent physical and chemical processes using
specific models for each different degradation
cause. Moreover, they have to account that the
actual condition of a bridge is the consequence of
many concurring degradation processes. For these
reasons, to obtain an effective and reliable
mathematical degradation model is a very difficult
task.
Condition State, C S

A different approach defines the degradation e erimental manifold concurrent

model by analyzing the condition state variation x data. In this degradation causes
with time, on the basis of historical and p way, the are implicitly taken
into account and there is no need to define specific
mathematical models. This allows calculation of 5

an appropriate degradation curve for each bridge, 4.5

reflecting all the specific factors that have caused

4 Bad environmental
its degradation. conditions
In the APT-BMS, there is currently no 3.5

2.5

2 Optimal
Optimal
environmental
Environmental conditions
conditions
1.5

historical data available. To define the condition 1

0 10 20 30 40 50 60 70 80

variation over time, all the BMS bridges have been Age (years)

grouped as to their structural type and material Figure 1. Simply supported concrete bridges’ CS over age.
used. An overall condition index for each bridge in
has been calculated according to the following With this approach, bridge inspection data is used
four-step procedure: to develop probabilistic models that predict the
 Step 1: the numerical value associated with the future bridge condition. It should be noted that
condition state of each element is normalized to even though the bridge condition ratings are
a condition index (CI) spanning from 1 to 5; assigned following a standard procedure, the
 Step 2: the CI of each group of elements is subjective judgments of bridge inspectors is not
calculated as the maximum normalized value completely ruled out and thus the trend may
obtained in Step 1; reflect inherent human bias. Accurate predictions
 Step 3: a specific normalized weighting is are essential for effective MR&R decision making,
assigned to each group of bridge elements (such therefore probabilistic deterioration methods are
as: superstructure, substructure, equipment...), used to characterize bridge deterioration.
according to their structural importance;
 Step 4: the overall bridge CI is calculated as a 4. MODELLING BRIDGE
sum of the weighed CI. DETERIORATION AS A SEMI-
MARKOV PROCESS
The resulting CI is a real number in the interval
[1,4], 1 representing the design state and 5 the Markov chains models have been used extensively
most severely damaged situation. It should be to predict future bridge conditions. Several
noted that this procedure associates with each methods for estimating transition probabilities
element a condition index proportional to the from the available bridge inspection data have
value of the numerical label of its condition state. been suggested in the literature [5,6].
Therefore, this method implicitly assumes that the A common assumption is that transition
damage level is proportional to the discrete value probabilities do not depend on the facility's age,
of CS. It is understood that a more refined i.e., that transition probabilities are homogeneous.
approach should consider weighting these values In the APT-BMS the degradation process has been
according to the actual damage severity. developed and is currently modelled with Static
By plotting the overall condition state as a Markov Chains. Non-stationary Markov chains, in
function of the bridge age, two degradation curves which transition probabilities depend on the time
can be defined for optimal environmental spent in an initial condition state, appear to be
conditions and bad environmental conditions. For more realistic [7,8,9].
example, Figure 1 shows the results for simply A semi-Markov process is defined as a process
supported concrete bridges. The degradation curve that changes states in accordance with a Markov
of each bridge is obtained through a set of chain but remains in a state for a definite period of
explanatory variables that represent its specific time. The time needed for a facility to pass from
environmental conditions and will be between the an initial condition state to its next specified
two extremes. condition state is defined as transition time. This
Amount of time, also called waiting time or time- 2,…,n} respectively, where n is the number of
in-state, is assumed to be a random variable. condition states. Ti is a random variable with
Let T1,T2,…, Tn be random variables denoting continuous and positive density function fTi (t),
the waiting times for each condition state {1, cumulative density function FTi (t) and survival
function STi (t). Let T1k be a random variable
denoting the cumulative time of permanence in
states {1, 2,…,k}, i.e., the time spent by the
process to go from state 1 to state k.
k

Figure 2. Progression of the condition state vector over

time for a simply supported concrete bridge.
Tik  k  1,2,..., n − Once the deterioration process is modelled, a
1
∑T j j 1
comparison can be made between the mean
The corresponding probability distributions are estimated values of condition state over time and
denoted as fT1k (t), FT1k (t), ST1k (t). As introduced the observed values (Figure 3). Under the
by Kleiner [10], it can be shown that if the process hypothesis that no maintenance is performed, the
is in state i at time t, the conditional probability estimated mean condition state values show good
that it will pass to the next state in the time step ∆t correspondence with the observed degradation
can be expressed as follows: curve when the bridge is in optimal environmental
f T1i (t ) conditions. The time variant condition state vector
p i,i1 (t)  i  2,3..., n A(t) for a simply supported concrete bridge under
S T 1i (t ) − S1(i −1) −1 non-optimal environmental conditions is
(t )
This equation provides all the transition calculated as an empirically defined function of
probabilities needed to obtain the transition the condition state vector under optimal
probability matrix P(t) for the semi-Markov environmental conditions, and the explanatory
process. These transition probabilities are time- variables that mostly affect the deterioration rate
dependent and the process is non-stationary. This of a bridge, i.e. daily heavy vehicles occurrence,
means that the older the asset is, the higher the number of days with freezing temperature in a
likelihood is of deterioration to the next state in a year, and so on. The actual curve used for each
given period of time. bridge will be between the optimal- and the worst-
Given the transition probability matrix, the conditions curve, and will be defined on the basis
condition state vector collecting the probabilities of specific values of the explanatory variables.
of being in a given state, can be obtained as:
5.0
k
−1
t i,t i1 4.5
A(t  k)  A(t )∏ P observed
observedCSCS
i 0 4.0 estimated
estimated
CSCS
Condition State, CS

t,t+1
where P denotes the transition probability 3.5

matrix from time t to time t+1. Figure 2 shows an 3.0

example of the condition state progression over
2.5
time for a simply supported concrete bridge, i.e.,
the prediction over time of the probability that the 2.0

bridge will be in a given condition state, assuming 1.5

that at time t=0 its condition state is equal to 1. In 1.0

0 10 20 30 40 50 60 70 80
this example Weibull probability distributions are
Time (years)
assumed to represent the duration in time in each
condition state. Figure 3. Progression of the bridge condition state vector
over time.
Condition State, CS

The condition state vector A(t) is used for bridge in the stock, as explained in
estimating the total life-cycle costs and the Section 7.
cumulative-time probability of failure for each
 Profile under
no maintenance

5. ACTION MODEL Profile under

maintenance
∆CS
The mathematical model developed to represent
actions and their effects defines four actions: ts
Routine Maintenance, Repair, Rehabilitation and
Reconstruction. Routine Maintenance actions
comprehend protection measures, ordinary
maintenance operations and minor repairs. Repair Time
Figure 4. Maintenance actions model.
actions comprehend structural repair operations,
equipment repair and substitution. Rehabilitation
actions comprehend major repair operations, 6. MAINTENANCE STRATEGY AND
performed on the whole structure both on PRIORITIZATION ANALISYS
structural and on equipment elements.
Actions are meant to extend the life-cycle of a In the field of bridge management, several studies
bridge by positively affecting its degradation have investigated the maintenance strategy
process. In the current implementation of the optimization (e.g. Ref. [12,13,14]), but have a
system the effects of maintenance actions are number of limitations (e.g. they work with
modelled by the Markov approach: for each deterministic damage models or with failure
actions a modified transition matrix is defined. criteria that are not reliability based). In recent
When a maintenance action is performed, the years, probabilistic optimization methods have
corresponding transition matrix is applied to the been increasingly used in maintenance models (e.g.
known transition states to model the effects of the Ref [15,16]). In several recent studies, (e.g. Ref.
action. Guigner and Madanat [9] proposed a [17,18]) maintenance models have been developed
method to estimate the transition probability for steel and concrete bridges based on reliability
matrix when a rehabilitation action is performed. and minimum life-cycle cost.
In this method though, the transition matrix is In this paper a simple cost benefit analysis is
obtained without considering the effects of the presented for the project-level choice of a
particular maintenance strategy and for the
actual actions performed and there is no
network-level prioritization of the selected
differentiation between repair and rehabilitation
strategies. The results obtained for three different
actions.
structures are illustrated and discussed.
In the method proposed in this paper, which is
A set of possible maintenance actions is defined
explained in detail in [11], the effects of the
(e.g. routine maintenance, repair, rehabilitation,
maintenance actions are taken into account by
replacement) and then a given number of
introducing a variation in the degradation curve,
maintenance programs is considered. For each
obtained by calculating the mean value of the
maintenance program and for each bridge a
condition state vector A(t) (Figure 4). In the
priority index is calculated as follows:
method proposed, Routine Maintenance and
Repair actions stop the degradation process for a ∆P t  PX t L  −PX |a
given number of years, Rehabilitation takes the   t L 
∆C
∆C
X L

structure to a lower CS (i.e. it takes the bridge

condition back in time for n years) and stops the probability of failure over the duration (0, tL]
degradation process for a given number of years, multiplied by an importance factor that in the
while Reconstruction takes the structure back to current implementation depends on failure mode,
CS = 1. The definition of the effects of these bridge dimensions and average daily traffic, a is
actions has been implemented through the use of the considered maintenance program, ∆C is the
parameters, to provide a more flexible model. actualized life-cycle cost associated with the
Where PX(tL) is the bridge cumulative- action implementation.
time
 For each maintenance program, the published yearly in the APT Construction Pricelist
prioritization module calculates the cumulative- Bulletin.
time probability of failure, the importance factor In the APT-BMS, the maintenance programs
and the life-cycle costs. Mori and Ellingwood [19] considered for the priority index calculation are:
first proposed a time-variant method for directly  0 (zero): do-nothing (routine maintenance
evaluating the cumulative-time failure probability only);
of the series system. In Mori and Ellingwood's  A: rehabilitation intervention at time 0;
formulation, the degradation function is not  B: reconstruction intervention at time 0.
statistically defined. A consequence of this
assumption is that the standard deviation of the For each maintenance program, the bridge
bridge capacity decreases with time, and this is deterioration process is evaluated and the
apparently against common experience. To corresponding time-cumulative probabilities, life-
overcome this limit, a probabilistic degradation cycle costs, rehabilitation and reconstruction
model is proposed. Assuming that the system priority indexes are calculated. At project level the
reliability is dominated by a single limit state, and best intervention program is the one with the
that this limit state can be formulated as the highest priority index, while network level
difference between a capacity R and a demand S, prioritization is made by ordering all the best
the cumulative-time probability of collapse can be project level programs: the first intervention
formulated as: performed will be the one with the highest index .
PF t L  
In the following we present the results obtained
for three sample bridges in the APT stock. Table 1
reports a summary of the outcome of the
∞   tL
   prioritization analysis, automatically carried out
1   
 
1− by the BMS for the three reference maintenance
− L −  FS r 1 −   f  d dt
∫ ∫ fR
0

∫ r dr
exp t 
0
 
0 0
 
 programs.
 
where  and FS are the mean occurrence rate and indirect costs and social costs. All future costs are
the cumulative distribution function of the demand actualized using the official financial discount rate
S,  is a probabilistic capacity degradation
function depending on the time variant vector A(t),
which collects the condition states of the bridge,
and fR0 is the probability density function of the
baseline capacity R0 assumed in design.
The life-cycle cost associated with a
maintenance program is estimated as the sum of
different costs: inspection cost, maintenance
actions cost and failure cost. The inspection cost
changes with type, size, accessibility of the
structure and it is assumed to be time-independent.
The maintenance actions cost (routine
maintenance and rehabilitation) is estimated on the
basis of routine maintenance or rehabilitation unit
costs of bridge elements and depend on their
condition states. Reconstruction cost is estimated
as euros per square meter of deck area, and it
depends on the typology of the elements
composing the bridge. The failure cost takes into
account all the structural and functional costs
associated with a potential failure, included
 The SP65 bridge on the Maso river has a
simply supported concrete structure and is a
common type of bridge in the APT stock. The
structure shows minor deterioration of the beams,
including localized concrete cover spalls, mostly
due to an inefficient drainage system. In this case,
strategy A (rehabilitation at time 0) has the
highest priority index, as shown in Table 1.
The Canova Viaduct carries a 4-lane highway
which represents one of the most critical road
connections in the region, with an average daily
traffic of over 15000 vehicles. The main
structure is 686m long and 17.70m wide, and has
34 simply supported spans of variable length. The
bridge dates from 1978 and shows signs of
advanced deterioration at the cross-beams,
resulting in some cases in the failure of the post-
tensioning system. These faults are due to poor
design in detail and execution. As shown in
Table 1, the risk associated with the do-
nothing maintenance program is relatively high.
Rehabilitation is ranked as the most cost-effective
action. However, reconstruction is also associated
with a high priority, higher than that calculated
for the rehabilitation of the SP65 bridge.
 1.E+08

A
Bridge on Maso River Bridge on Maso River B
Bridge on Vignola River B A Bridge on Vignola River
1.E+07
Canova Viaduct Canova Viaduct
Cumulative Cost ( € )

∆PX (tL)
1.E+06 A
A

A B
B A B

A
A
B 1.E+05

0
0
1.E+04
B
B
1.E+03
1.0E-09 1.0E-08 1.0E-07 1.0E-06 1.0E-05 1.0E-04 1.0E-03 1.0E-02 1.0E-01 1.0E-
00 ∆C
Cumulative Time Risk
Figure 6. Graphical representation of the priority index.
Figure 5. Cumulative time risk and total cost for three
bridges in the APT stock.

SP12 bridge on Vignola river is a minor bridge In this case, replacement is ranked as the most
serving a local road with an average traffic of a cost-effective action. Figure 5 shows the
few vehicles per day and consisting of a single cumulative time risks and the total costs for the
simply supported 5m-wide and 10m-long span. three bridges considered. A bi-logarithmic scale is
Given the minor importance of the bridge, the used due to the great difference between the
cumulative-time risk evaluated by the system is Canova Viaduct and the other two bridges.
relatively low, as reported in Table 1. However, Figure 6 is a graphical representation of the
the cost of any potential action is also very low, priority indexes: each line corresponds to a
and for this reason the resulting priority indexes maintenance program and its gradient is its
are comparable with those of the Canova Viaduct. priority index. The maintenance programs are thus
ordered from left to right, from the one with the
highest priority index to the one with the lowest.
SP65 Bridge on Maso River
0 A B 7. CONCLUSIONS
-5 -5 -7
PX 4.17·10 1.02·10 5.96·10
Cost [€] 26300 83700 369280 In this paper an approach is presented to model the
∆PX - 3.15·10-5 4.11·10-5 bridges deterioration based on observational data.
- 57000 343000 The use of semi-Markov chains to characterize the
∆C [€]
-1 -10 -10 deterioration process in a probabilistic way has led
 [€ ] - 5.49·10 1.19·10
to a realistic prediction of the bridges condition
Canova Viaduct state variation in time, even though a longer data
0 A B history would be of great help for improving the
-3 -3 -5
PX 8.00·10 3.08·10 2.51·10 results accuracy. The environmental conditions
Cost [€] 2069000 6733000 11893000 that mostly affect the degradation rate of a bridge
∆PX - 4.92·10-3 7.97·10-3 are represented by a set of explanatory variables
∆C [€] - 4660000 9820000 that modify the deterioration process. In this way
-1 -9 -10
 [€ ] - 1.05·10 8.11·10 each bridge has a specific probabilistic degradation
SP12 Bridge on Vignola River curve. The effects of the maintenance actions are
0 A B taken into account by introducing a stop or a
PX 7.80·10
-5
3.31·10
-5
2.79·10
-7 negative jump in the degradation curve. For a
Cost [€] 20450 72300 98200 given maintenance program, the deterioration of
∆PX - 4.49·10 -5
7.77·10-5 a bridge in terms of condition state variation in
- 51900 77800 time is used to calculate the bridge cumulative-
∆C [€]
-1 -10 -10 time probability of failure and the life- cycle costs.
 [€ ] - 8.66·10 9.99·10
A cost-benefit analysis is performed in order to
Table 1. Outcome of the prioritization analysis performed on select the best project-level maintenance strategy
three bridges in the APT stock. (among those considered), and the highest
intervention priority at network level. The importance of a bridge in terms of geometrical
prioritization analysis also considers the dimensions and average daily traffic.
The results of the cost-benefit analysis
automatically carried out by the BMS have been
presented for three sample bridges of the stock. A 6. Cesare, M.A., Santamaria, C., Turkstra, C. and
drawback of this approach is that only a limited Vanmarcke, E.H., 1992 Modeling bridge
number of maintenance options is considered. In deterioration with Markov chains. Journal of
Transportation Engineering, 118(6), 820-833.
addition, the priority indices evaluated depend on 7. Madanat, S., Mishalani, R. and Wan Ibrahim,
the reference intervention scenario chosen. As a W.H., 1995 Estimation of infrastructure
consequence, for a given bridge the selected transition probabilities from condition rating
maintenance strategy may not be optimal. data. Journal of Infrastructure Systems, 1(2),
120-125.
8. Ravirala, V. and Grivas, D.A., 1995 State
ACKNOWLEDGEMENTS increment method of lifecycle cost analysis for
highway management. Journal of
This research has been possible thanks to the Infrastructure Systems, 1(3), 151-159.
9. Guigner, F. and Madanat, S., 1999
financial support of the Autonomous Province of Optimization of infrastructure systems
Trento (contracts # 277/02 and #342/03). The maintenance and improvement policies.
authors wish to thank the APT department of Journal of Infrastructure Systems, 5(4), 124-
134.
Transportation, and specifically Raffaele De Col, 10. Kleiner, Y., 2001 Optimal scheduling of
Luciano Martorano, Paolo Nicolussi and Matteo rehabilitation and inspection/condition
Pravda, for their continuous support in the assessment in large buried pipes. In:
development and operation of the system. The Proceedings of the 4th International
Conference on Water Pipeline Systems,
authors also wish to thank all of the people who Managing Pipeline Assets in an Evolving
cooperated in the research, including David Market. 181-197.
Capraro, Giovanni Cortese and Alessandro Lanaro. 11. Bortot, F., 2006 Reliability-Based
Decision- Making in Infrastructure
Management. Ph.D.
REFERENCES thesis, University of Trento, Italy.
12. Vorster, M.C., Bafna, T. and Weyers, R.E.,
1991 Model for determining the optimum
rehabilitation cycle for concrete bridge decks.
Transportation Research Record, 1319, 62-71.
TRB, National Research Council, Washington,
D.C.
13. Jacobs, T.J., 1992 Optimal long-term
scheduling of bridge deck replacement and
rehabilitation. Journal of Transportation
Engineering, 118(2), 312-322.
1. AASHTO, 1997 Guide for Commonly 14. Markow, M.J., Madanat, S.M. and Gurenich,
Recognized Structural Elements, (AASHTO: D.I., 1994 Optimal rehabilitation time for
Washington, D.C.). concrete bridge decks. Transportation
2. Astudillo, R., 2002 BRIME: basis for an Research Record, 1392, 79-89. TRB, National
European bridge management system. In Proc. Research Council, Washington, D.C.
First Int. Conf. on Bridge Mainteinance, Safety 15. Thoft-Christensen, P. and Sørensen, J.D., 1987
and Management, IABMAS'02, Barcelona, 14- Optimal strategy for inspection and repair of
17 July, 2002, edited by J.A. Casas, D.M. structural systems. Civil Engineering Systems,
Frangopol and A.S. Nowak (CIMNE: 4, 94-100.
Barcelona). 16. Sommer, A.M., Nowak, A.S. and Thoft-
3. Thoft-Christensen, P., 2002 Deterioration of Christensen, P.,1993 Probability-based bridge
Concrete Structures. In Proc. First Int. Conf. inspection strategy, Journal of Structural
on Bridge Maintenance, Safety and Engineering, 119(12), 3520-3536.
Management, IABMAS'02, Barcelona, 14-17 17. Frangopol, D.M. and Estes, A.C., 1997
July, 2002, edited by J.A. Casas, D.M. Lifetime bridge maintenance strategies based
Frangopol and A.S. Nowak (CIMNE: on system reliability. Structural Engineering
Barcelona).
4. Enright, M.P. and Frangopol, D.M., 1999b International, 7(3), 193-198.
Condition predicting of deteriorating concrete 18. Enright, M.P. and Frangopol, D.M., 1999
Reliability-based condition assessment of
bridges using Bayesian updating, J. Struct. deteriorating concrete bridges considering load
Eng., ASCE, 125(10), 1118-1124. redistribution, Structural Safety, 21, 159-195.
5. Jiang, Y., Saito, M. and Sinha, K.C., 1988 19. Mori, Y. and Ellingwood, B., 1993b
Bridge performance prediction model using the Reliability-based service-life assessment of
Markov chain. Transportation Research aging concrete structures. Journal of
Record, 1180, 25-32. TRB, National Research Structural Engineering, 119(5), 1600-1621.
Council, Washington, D.C.