You are on page 1of 9

Reliability Index for Structural Health

Monitoring of Aging Bridges

P. B. R. Dissanayake* and P. A. K. Karunananda

Department of Civil Engineering, University of Peradeniya


Peradeniya 20400, Sri Lanka

This paper presents a reliability-based methodology for health monitoring of aging bridges. Initially,
critical failure criteria are proposed depending on the type of bridge. Safety margins are then defined
based on these criteria. It is assumed that quantities in the failure criteria are random variables that are
normally distributed. Based on this assumption, elementary reliability indices and thereby elementary
failure probabilities are estimated for each critical failure mode. Having found the elementary failure
probability values, system failure probability of the bridge is calculated for the period under
consideration. Finally, the system failure probability is used to obtain a system reliability index of the
bridge. The system reliability index is a quantative way to express the condition of the bridge for the
period under consideration. As a case study to illustrate the proposed reliability procedure, a wrought
iron railway bridge was selected from the national railway bridge network of Sri Lanka. For this bridge,
fatigue and corrosion were selected as the most critical forms of failure. Application of the proposed
methodology showed that the present condition of bridge was satisfactory for current loading
conditions. Predicting reliability, as opposed to damage, is a positive outlook approach to health
monitoring that is useful for infrastructure management.

Keywords structural health monitoring  condition estimation  maintenance  reliability


index  failure probability

1 Introduction infrastructure that directly affects the industrial


productivity of a country.
In today’s rapidly changing world, the quality Most current bridge maintenance strategies
of human life and the economic progress of a are based on visual inspections made by bridge
country depend on the quantity, quality, and inspectors at varying time intervals [3]. Since
efficiency of its infrastructure [1]. Civil infrastruc- human inspection depends on the individual ins-
ture is often in need of maintenance, rehabili- pector, there exists a degree of uncertainty in the
tation, or replacement [2], irrespective of the results obtained. Therefore, absolute dependence
location around the world. In particular, high- on inspection reports is not reliable and may lead
ways and associated ground transportation to incorrect decision-making, which results in
systems constitute a considerable investment in high maintenance costs for bridge owners.

Copyright ß SAGE Publications 2008


*Author to whom correspondence should be addressed. Los Angeles, London, New Delhi and Singapore
E-mail: ranjith@civil.pdn.ac.lk Vol 7(2): 0175–9
Figure 3 appears in color online: http://shm.sagepub.com [1475-9217 (200806) 7:2;175–9 10.1177/1475921708090555]

175

Downloaded from shm.sagepub.com at UNIV PRINCE EDWARD ISLAND on February 22, 2015
176 Structural Health Monitoring 7(2)

The cost of maintenance should be optimized Therefore, depending on the bridge type, critical
based on the remaining service life of a bridge. failure modes are proposed. These failure modes
On the other hand, the capability of bridge define the mathematical procedure to evaluate
inspectors to find structural and nonstructural the structural health of bridges. The safety margin
defects, damages, and deterioration processes Mi for the ith mode of failure of the bridge is
should not be neglected. In fact, human experi- defined as:
ence should always be incorporated into main-
tenance decisions, but with sufficient subjective Mi ¼ ZRi  ZSi i ¼ 1; 2;. . . ; n ð1Þ
knowledge. Only then, can any strategy have
acceptable performance. Current procedures in where ZRi is the strength variable and ZSi is the
structural health monitoring of bridges have a load variable. When ZRi 4ZSi , the bridge is in a
number of shortcomings and prevailing uncer- safe state and when ZRi 5ZSi , the bridge is in a
tainties. The results of insufficient knowledge failure state. If the means and standard devia-
about structural health monitoring and bridge tions of the resistance and load variables are
maintenance are causing significant financial loss known, then a reliability index can be found for
due to frequent maintenance activities, unforeseen the ith failure mode as [6,7]:
bridge damages, and accidental bridge failures.  
These losses in terms of financial, physical, and ZRi  ZSi
other resources are unacceptable, irrespective of i ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  : i ¼ 1; 2;. . . ; n: ð2Þ
the wealth of a country. The complexity of 2
Z R þ Z S 2
i i
structural life prediction and general health mon-
itoring involving bridge maintenance has ener-
In (2), mZRi and  ZRi are the mean and the
gized researchers into formulating general
standard deviation of the strength variable, while
methods of life prediction and cost optimized
mZSi and  ZSi are the mean and the standard
maintenance strategies for bridges.
deviations of the load variable. The reliability
Thus, any proposed methodology for health
index expresses the condition of a bridge as the
monitoring and maintenance should be based on
mean value of the safety margin divided the
results of field studies as well as subjective under-
standard deviation of the safety margin.
standings accepted by practicing engineers. Since
Generally, the higher the value of reliability
there are many uncertainties existing in current
index, the better the condition of the bridge. It is
procedures, a probabilistic approach is advanta-
assumed here that both strength and load vari-
geous over deterministic approaches. In this
ables are normally distributed. The assumption of
situation, reliability-based methods can provide a
normality in variables simplifies calculation of
rational approach to use scarce resources effi-
failure probability while maintaining a satisfac-
ciently while maintaining a prescribed level of
tory accuracy.
reliability of a structure throughout its designated
If the reliability index is known, failure
service life [4].
probability Pfi can be calculated as:
Pfi ¼ ði Þ i ¼ 1; 2 . . . ; n ð3Þ
2 Methodology where / is the standard unit normal
distribution. Substituting Equation (2) into
2.1 Mathematical Procedure Equation (3) yields [6,7]:
Bridges can fail due to a number of critical 2 3
failure modes, depending on the type of bridge. 6 ZRi  ZSi 7
For steel bridges, fatigue is the dominant failure Pfi ¼ 1  6
4rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

7
5 : i ¼ 1; 2; . . . ; n:
mode. But corrosion also has a considerable Z2 R þ Z2 S
i i
impact on the life of steel bridges. Reinforced con-
crete bridges are influenced by moment and shear. ð4Þ

Downloaded from shm.sagepub.com at UNIV PRINCE EDWARD ISLAND on February 22, 2015
Dissanayake & Karunananda Structural Health Monitoring of Aging Bridges 177

The physical meaning of the failure probabil- This system reliability index can be used to
ity is that it conveys the idea of how close the express the present condition of the bridge.
failure state is. Hence, higher values for failure As outlined in the above equations, the system
probabilities imply greater chance of failure. reliability index of the bridge can be calculated at
According to Equation (4), for each failure any given time.
mode (i ¼ 1, 2, . . . , n), the elementary reliability
index and the elementary failure probability can
be calculated. The next step is to calculate the 2.2 Graphical Representation of the
system failure probability. The actual failure of a Reliability Procedure
bridge can be attributed to several failure modes.
Thus, a system model has to be built up. In this The graphical representation of the reliability
context, it is assumed that all failure modes are procedure is given in Figure 1.
combined with a series system. Hence, it is The procedure in Figure 1 can be followed
possible to calculate the system failure probability at suitable time intervals to calculate the
from a simple bound as [6,7], system reliability index. This value is compared
with the target reliability index of the bridge.
n Y
n  
Max Pfi  PF  1  1  Pfi : ð5Þ The target reliability index is the lowest permis-
i¼1;2
i¼1 sible value that a bridge can fail. If the current
value of the index is higher than the target value,
The lower bound of (5) represents the situa-
no major maintenance is required. However,
tion where all n failure modes are uncorrelated
routine maintenance is advisable at this stage.
and the upper bound represents the case when all
When the current system reliability index is
n failure modes are correlated. Having found the
higher than the target reliability index, major
system failure probability, it is possible to convert
maintenance should be carried out to increase the
it to get a system reliability index as follows [6,7],
current reliability index to a value greater than
S ¼ 1 ðPF Þ: ð6Þ target reliability index.

Determine types of failure modes

Estimate elementary reliability index

Estimate elementary failure probability

Determine system failure probability

If b current < b target


Estimate system reliability index Do routine maintenance

If b current < b target

Do major maintenance activity

Figure 1 Flow chart of the reliability algorithm.

Downloaded from shm.sagepub.com at UNIV PRINCE EDWARD ISLAND on February 22, 2015
178 Structural Health Monitoring 7(2)

2.3 Target Reliability Index the reliability index starts to reduce. During this
period, the important aspect of study is the rate
The value of the target reliability index used
reduction of the reliability index (). The value of
by researchers differs. The American Association
 is dependent on the how degradation process
of State and Highway Transportation Officials
have affected the particular bridge. For example,
specifications for the remaining life of steel bridges
a steel truss bridge located on a coastal environ-
is based on the use of a reliability-oriented
ment has higher  than a steel bridge located
approach. In this guide, two levels of risk are
inland as the salt environment triggers early
considered. For a redundant bridge, a reliability
corrosion of steel. When the reliability index
index of 2.0 is used. For a nonredundant bridge,
reaches the target reliability index (target), some
a reliability index of 3.0 is used [5]. In this study,
form of maintenance should be performed to
the target reliability index value used is 3.0.
maintain the safety of the bridge. This will
increase the reliability index by an amount .
The increase of  is accompanied with a certain
2.4 Variation of the System
cost. Hence, it is generally essential to optimize
Reliability Profile
the increase of reliability index with the cost of the
The nature of the system reliability index of a maintenance activity. The increase of the reliabil-
bridge is that it reduces with time. However, it ity index is higher for an essential maintenance
has an initial time period of constant value. The activity which accounts for a major repair than
length of this period depends on the loading that of a preventive maintenance activity that
conditions of the particular bridge and environ- accounts for routine maintenance procedure. The
mental factors at the site. With proper mainte- increase of the reliability index is constant for a
nance activities, the service life of the bridge can period of time T3. After that, the reliability index
be extended well beyond the design life of the goes down again with a reduced rate of degrada-
bridge. The variation of the system reliability tion of    where  is called the reduction
index with time referenced to maintenance activ- of deterioration rate due to maintenance activity.
ities is illustrated in Figure 2. Its effect continues for a time period T4. Then,
Initially, the value of the reliability index (0) the degradation rate becomes  again. When it
after construction is a high value (46.0) and for a reaches target again, maintenance should be
time period T1 it remains constant. After time T1, performed to improve the safety of the structure.

System reliability index


(bcurrent)

With maintenance
T1

T3 T4
b0

a–d

a Db1

btarget
T2 T5
Without maintenance

Time

Figure 2 System reliability variation of a typical bridge.

Downloaded from shm.sagepub.com at UNIV PRINCE EDWARD ISLAND on February 22, 2015
Dissanayake & Karunananda Structural Health Monitoring of Aging Bridges 179

3 Case Study Application Corrosion is another form of deterioration in


steel bridges. For corrosion, the reliability model
As a case study, a steel truss bridge was can be proposed based on steel area as:
selected and reliability modeling was implemented.
In this case, fatigue and corrosion are the most M2 ¼ ðAs Þcurrent ðAs Þrequired ð10Þ
dominant forms of failure. When a member in a
bridge undergoes cyclic loading in a repetitive where M2 is the safety margin for the corrosion
manner, it is liable to fail by fatigue. Generally, mode of failure, and (As)current is the steel area
fatigue is considered to be the one of the most present at the time of consideration, and
significant failure modes of steel bridges. (As)required is the steel area required to carry the
For a member as mentioned above, the load on the bridge. When (As)current  (As)required,
failure can be modeled using the reliability M2  0 the bridge is in a safe condition. When
approach as: (As)current 5 (As)required, M250 and the bridge is
in a failure condition. Both variables are assumed
M1 ¼ Nf  Nn ð7Þ to behave as normally distributed variables.
Based on the mean and standard deviation of the
two variables, the reliability index of corrosion
where M1 is the safety margin for fatigue failure failure can be expressed as:
of the member and it is the criteria used to
h i
express how close the element is to failure. Nf is
ðAs Þcurrent  ðAs Þrequired
the total number of stress cycles that can be 2 ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
h i: ð11Þ
applied to the bridge material. Nn is the number 2 2
ðAs Þcurrent þ ðAs Þrequired
of stress cycles applied up to now. Both variables
are assumed to follow normal distributions.
When Nf  Nn, M1  0 and the member is in the Having converted M to the standard nor-
safe region. If Nf5Nn, M150, the member is in mally distributed variable, the relationship
the failure region. between failure probability and reliability index
Based on these values, the reliability index can be found as:
for the first mode of failure can be found as: 8 h i 9
>
> >
  < ðAs Þcurrent  ðAs Þrequired > =
Nf  Nn Pf2 ¼ 1   rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
h i>: ð12Þ
1 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : ð8Þ >
> >
: 2 2 ;
ðN 2 þ 2 Þ ðAs Þcurrent , þ ðAs Þrequired
f Nn

If mean values and standard deviations of the


Here, Nf and Nf are the mean and the
two variables can be found, the probability of
standard deviation of variable Nf. Similarly Nn
corrosion failure can be calculated. Thus, elemen-
and Nn are the mean and the standard deviation
tary reliability indices of fatigue and corrosion
of variable Nn [6,7]. Having found the reliability
can be used to find the system failure probability
index as Equation (8), it is then possible to find
as in Equation 5 and the system reliability index
the failure probability of the element that is
as in Equation (6).
subjected to fatigue. The failure probability is
Nowak and Szerszen [8] have studied model-
given by [6,7]:
ing of fatigue and corrosion. In fatigue, they have
2 3 used the same reliability model in Section 2.1.
l  lN n For corrosion effects, three corrosion rates;
6 Nf 7
Pf1 ¼ 1  /4qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi5 ð9Þ low, medium and high have been studied. Their
2
ðrNf þ rNn Þ 2
effects on moment capacity and shear capacity
have been considered. However, the combined
where / is the standard unit normal distribution effects of fatigue and corrosion have not been
function. studied in detail.

Downloaded from shm.sagepub.com at UNIV PRINCE EDWARD ISLAND on February 22, 2015
180 Structural Health Monitoring 7(2)

The bridge selected for study was one of the sets so that the behavior of the bridge under
longest and the busiest railway bridges in Sri loading can be interpreted easily with the use of
Lanka spanning 160 m over two banks of the these different member sets. Figure 4 shows the
Kelani river situated close to the capital elevation of the truss and its different member
Colombo. It is an eight-span bridge with Warren sets [9]. Figure 5 shows the plan view of the bridge
type trusses, and semi through and double lane
rail trucks. Bridge piers are made of Cast Iron Table 1 Variation of number of trains over
caissons with infilled concrete. The bridge was time.
constructed in 1885 by British engineers and Period No of trains/day
some modifications were made later on. A pre-
1995–Present 80
sent view of the bridge is shown in Figure 3. 1985–1995 72
Table 1 shows the increase in the number of 1975–1985 64
trains over the bridge with time. 1970–1975 56
The selected bridge consists of members 1950–1970 42
having different shapes and sections, which belong 1930–1950 32
1910–1930 24
to different sets. In the structural analysis of the
1885–1910 20
bridge, it is important to identify these different

Figure 3 A view of the case study bridge (Dissanayake & Karunannanda).

SET2 SET2 SET2 SET2 SET3 SET3 SET4 SET4 SET4


SE

SE
SE

SE

N
SE

SE
N

IN

SE
IN

N
IN

7I
5I

9I
T9

T9

T9
T

T8
T8
T6

T9
T

T9
T
6O

7O

T
SE
SE

SE
SE

SE
O
SE

IN
U

U
U

T
T

T
SET1

Z
T
SE
SE

SE
SE

SE

SE
IN

U
IN

IN
IN
IN

IN

O
T7

T
T6

T
T9

T9
T

T8

T9

T
T5

T6

T9
8O

9O
7O

9O
SE
O

SE

SE
SE

O
SE

SE

SE
U
U

U
U

U
T
T

T
T

SET2 SET2 SET2 SET2 SET3 SET3 SET4 SET4 SET4


X

Figure 4 Member set numbering for the truss.

Downloaded from shm.sagepub.com at UNIV PRINCE EDWARD ISLAND on February 22, 2015
SET7VAR
SET7VAR

SET7VAR
SET2 SET2 SET2 SET2 SET3 SET3 SET3 SET4 SET4 SET4

L5
x3
,5x
5/8

VAR2
VAR2
VAR2
VAR2
VAR2
VAR2
VAR2
VAR2

SET9 SET9 SET9 SET9 SET9 SET9 SET9


L4

8
x3

/8

3/
x3

x3

x
/8

x3

x3
1/

L4

L4
, 5x
x3

SET8
SET8
SET8
SET8
SET8
SET8
SET8
SET8

,5
L3

SET9 SET9 SET9 SET9 SET9 SET9 SET9

L
L4
L3
8

4x
x3 3/

3
x

3x
x3

x3
/8 x3

/8
3/8
Dissanayake & Karunananda

L3

8 L3

SET8
SET8
SET8
SET8
SET8
SET8
SET8
SET8

3/ x3

/8
x x3

x3
x3 / 8
L3

x3

SET1
L4
SET9 SET9 SET9 SET9 SET9 SET9 SET9
L3
,5 8
x
8

3, x3/
3/

5x

L4
L4
3x

x3
1/
x

2
x

x3
L4
3x

SET8
SET8
SET8
SET8
SET8
SET8
SET8

L4

SET8

x3
3/

/8
8

SET9 SET9 SET9 SET9 SET9 SET9 SET9

/8 Y
x5
x 3,5
L5

Downloaded from shm.sagepub.com at UNIV PRINCE EDWARD ISLAND on February 22, 2015
VAR1
VAR1
VAR1
VAR1
VAR1
VAR1
VAR1

VAR1

SET2 SET2 SET2 SET2 SET3 SET3 SET3 SET4 SET4 SET4
X

SET7VAR
SET7VAR
SET7VAR
Structural Health Monitoring of Aging Bridges

Figure 5 Member set numbering for the bridge deck.


181
182 Structural Health Monitoring 7(2)

Table 2 Present elementary reliability indices of the different member sets for fatigue and
corrosion failures.

Reliability Reliability index


Member set index for fatigue for corrosion
Set 2 (Main girder) 9.60 8.00
Set 3 (Main girder) 7.30 8.00
Set 4 (Main girder) 4.33 7.50
Set 8 (Cross girder) 4.33 7.50
Set 9 (Rail bearers) 4.85 9.00
Set 6 out (Diagonal tension members) 7.09 9.00
Set 7 out (Diagonal tension members) 7.12 9.00
Set 8 out (Diagonal tension members) 6.84 9.00
Set 9 out (Diagonal tension members) 6.96 9.00

deck with its member sets also defined accord- Table 3 Calculated final values of system reliability
ingly. These two figures give an idea of different indices of different member sets.
member sets and their respective locations in the
bridge. Out of these different member sets, there System reliability
Member set index
are nine sets of members liable for fatigue
failure [9]. These are set 2 (main girder), set 3 Set 2 (Main girder) 8.00
(main girder), set 4 (main girder), set 8 (cross Set 3 (Main girder) 7.30
Set 4 (Main girder) 4.33
girder), set 9 (rail bearers), set 6 out (diagonal
Set 8 (Cross girder) 4.33
tension members), set 7 out (diagonal tension Set 9 (Rail bearers) 4.85
members), set 8 out (diagonal tension members), Set 6 out (Diagonal tension members) 7.09
and set 9 out (diagonal tension members) [9]. Set 7 out (Diagonal tension members) 7.12
Set 8 out (Diagonal tension members) 6.84
Set 9 out (Diagonal tension members) 6.96
3.1 Results
Considering the loading history of the bridge
as shown in Table 1, for the first mode of failure
(fatigue), reliability indices at present were calcu- it is time to replace them with new members. The
lated and are summarized in Table 2. research findings have been applied to existing
Similarly, the bridge was studied for the bridges and based on results life estimation can be
current condition of corrosion. Based on this done satisfactorily.
inspection, elementary reliability indices of indivi- The proposed methodology can be applied to
dual members were found and are shown in life estimation of the bridge. Plotting the system
Table 2. reliability index over time will allow prediction of
By applying the methodology introduced, the remaining service lives of the different
system reliability indices were calculated and it is member sets of the bridge.
shown in Table 3.
From the above, it can be concluded that,
presently all member sets are safe since their 4 Conclusions
reliability values are above the target reliability
index value (3.0). However, reliability indices of This study was focused on how to estimate
some members are relatively low compared to the health of wrought iron bridges using a
others. The level of corrosion of these members reliability index. To formulate the index, fatigue
should be monitored and fatigue accumulation and corrosion were selected as the critical failure
should continuously be checked. Once these values modes. Reliability models were proposed based
reach the target value of the reliability index, on these failure modes. Using the index, failure

Downloaded from shm.sagepub.com at UNIV PRINCE EDWARD ISLAND on February 22, 2015
Dissanayake & Karunananda Structural Health Monitoring of Aging Bridges 183

probability of a wrought iron bridge can be 2. Kong, J.S. and Frangopol, D.M. (2004). Prediction of
found and used to express the condition of the reliability and cost profile of deteriorating bridges under
bridge. A case study was selected from the time and performance-controlled maintenance. Journal
national railway network of Sri Lanka. It was of Structural Engineering, 130(12), 1865–1874.
3. Sommer, A., Nowak, A.S. and Christensen, P.T. (1993).
found that the current state of the bridge is
Probability based bridge inspection strategy. Journal of
satisfactory even though signs of corrosion
Structural Engineering, 119(12), 3520–3536.
appeared. With the theoretical understanding as 4. Estes, A.C. and Frangopol, D.M. (2005). Load rating
illustrated in the paper, improved bridge main- versus reliability analysis. Journal of Structural
tenance and health monitoring can be carried out Engineering, 131(5), 843–847.
for wrought iron bridges. 5. Yang, J. and Dewolf, J.T. (2002). Reliability assessment
of highway truss sign supports. Journal of Structural
Engineering, 128(11), 1429–1438.
Acknowledgment 6. Christensen, P.T. and Baker, J. (1982). Structural
Reliability Theory and its Applications, Berlin,
Germany: Springer-Verlag.
The authors of this paper would like to express their sincere
7. Christensen, P.T. and Murotsu, J. (1986). Application of
acknowledgement to the National Science Foundation of
Structural Systems Reliability Theory, Berlin, Germany:
Sri Lanka for the financial support given through the
Springer-Verlag.
research grant no: NSF/SCH/2005/02 and RG/2002/E/01.
8. Nowak, A.S. and Szerszen, M.M. (2001). Life Cycle Cost
Analysis and Design of Civil Infrastructure Systems,
pp. 239–250, USA: American Society of Civil Engineers.
References 9. Ranaweera, M.P., Aberuwan, H., Mauroof, A.L.M.,
Herath, K.R.B., Dissanayake, P.B.R., Siriwardane,
1. Kong, J.S. and Frangopol, D.M. (2004). Cost–reliability S.A.S.C. and Adasooriya, A.M.N.D. (2002). Structural
interaction in life–cycle cost optimization of deteriorat- appraisal of the railway bridge at Colombo over Kelani
ing structures. Journal of Structural Engineering, 130(11), river, Engineering Design Centre, University of
1704–1712. Peradeniya, Peradeniya, Sri Lanka.

Downloaded from shm.sagepub.com at UNIV PRINCE EDWARD ISLAND on February 22, 2015

You might also like