Вы находитесь на странице: 1из 9

International Journal of Fatigue 23 (2001) 4553

www.elsevier.com/locate/ijfatigue
Fatigue analysis and life prediction of bridges with structural
health monitoring data Part I: methodology and strategy
Z.X. Li
1,*
, T.H.T. Chan, J.M. Ko
Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, Kowloon, Hong Kong
Received 20 September 1999; received in revised form 4 April 2000; accepted 9 June 2000
Abstract
This paper is aimed at developing a methodology and strategy for fatigue damage assessment and life prediction of bridge-deck
sections of existing bridges with online structural health monitoring data. A fatigue damage model based on the continuum damage
mechanics (CDM) is developed for evaluating accumulative fatigue damage of existing bridges. A structural model for the fatigue
stress analysis of bridge-deck structures is proposed, in which structures are modeled by elastic members and welded connections
with possible accumulative damage. Based on the proposed model, an analytical approach for evaluating the fatigue damage and
service life of bridge-deck sections based on strain history data from an online structural health monitoring system and the CDM
fatigue model are suggested. The updating of the representative block of cycles of the local stress by online monitoring data in the
future is included in the computational approach. In order to compare results of fatigue damage and service life prediction evaluated
by the CDM fatigue damage model, a modied PalmgrenMiner rule is developed for the same fatigue problem. 2001 Elsevier
Science Ltd. All rights reserved.
Keywords: Structural health monitoring; Fatigue damage; Life prediction; Bridge-deck section; Representative block of cycles; Continuum
damage mechanics
1. Introduction
Structural health monitoring has been accepted as a
justied effort for long-span bridges which are critical
to a regions economic vitality, such as the Tsing Ma
Bridge, the Kap Shui Mun Bridge and the Ting Kau
Bridge in Hong Kong and others. While instrumentation
systems, Wind And Structural Health Monitoring Sys-
tem (WASHMS) [1,2], are being installed on these
bridges, a common concerned problem is how to take
full advantage of the real-time monitoring data for effec-
tive and reliable health assessment of the structures. As
an important part of structural health, durability of the
bridge structures is mainly dominated by the fatigue
behavior of the critical elements of the bridge. On the
other hand, fatigue design of bridges according to the
* Corresponding author. College of Civil Engineering, Southeast
University, 2 Si Pai Lou, Nanjing, 210018, Peoples Republic of
China. Tel.: +86-25-3793384; fax: +86-25-7712719.
E-mail address: zhxli@seu.edu.cn (Z.X. Li).
1
On leave from Southeast University, Nanjing, 210018, P.R. China.
0142-1123/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S0142- 1123( 00) 00068- 2
code [3] is limited by design loading which may have a
signicant effect on the service of bridges. All of these
suggest that the fatigue analysis is signicant for the
safety of these bridges.
There have been a lot of researches on fatigue damage
analysis and life prediction theories [46]. The basic
concept and relative law of cumulative fatigue damage
were proposed by Miner [7] in 1945. His linear rule now
is applied to almost all of the fatigue design. However,
the life prediction based on this rule is often unsatisfac-
tory for fatigue under variable-amplitude loading [5],
since the following effects on fatigue damage accumu-
lation are not considered in Miners rule: (1) load cycles
below the fatigue limit which can propagate micro-
cracks if the cracks are initiated by load cycles with
amplitude higher than the fatigue limit; (2) the load
sequence effect. After the work by Miner, many different
fatigue damage models have been developed, which
include: (1) the works as an improvement of Miners
rule, such as the double linear damage rule [8], the dam-
age curve approach, rened double linear damage rule
[9] and the double damage curve approach [10]; (2)
46 Z.X. Li et al. / International Journal of Fatigue 23 (2001) 4553
fatigue damage theories based on plastic strain energy
or total strain energy [1115]; and (3) fatigue damage
theory based on the new subject of mechanics, con-
tinuum damage mechanics [1621], which is specially
suitable for the period of fatigue crack initiation,
although it is still at the primary stage of laboratory
study. A little work on bridge fatigue damage evaluation
has been made in recent years. For example, Zhao [22]
used a linear elastic fracture mechanics-based reliability
model which was updated through non-destructive
inspections to evaluate the fatigue damage of steel
bridges; Zheng et al. [23] experimentally investigated the
fatigue life of the steel of an old bridge by using speci-
mens cut from the chords of the old riveted steel bridge;
Agerskov and Nielsen [24] studied the fatigue damage
accumulation in steel highway bridges under random
loading by experiments and the analysis based on frac-
ture mechanics; Enright and Frangopol [25] applied the
time-variant reliability methods [26] and Monte Carlo
simulation to reliably predict the service life of deterior-
ating highway bridges. There has not been any relevant
work found on fatigue analysis for large cable-supported
bridges with structural health monitoring systems. It has
become a signicant problem to be studied with the
development of the structural health monitoring system
for large complicated structures.
Fatigue life can be considered to consist of two per-
iods: the fatigue crack initiation and the crack growth
period. The fatigue crack initiation covers the initiation
and growth of the small cracks in the micro-range size,
which cannot be detected by eye or non-destructive
inspection. For the health of a new bridge to be moni-
tored in order to determine the need for reinforcing or
repairing by the administration, the rst period should
be a major part of the research. The Miners rule used
in many fatigue design guides has the advantage of sim-
plicity with mathematical elegance, making them attract-
ive to practising structural engineers. However, this
approach does not directly associate fatigue damage with
its physical mechanism such as fatigue crack initiation
and growth. Fatigue tests on different materials
[17,20,21] have shown that a nonlinear fatigue model
based on continuum damage mechanics (CDM) is more
reliable than the linear Miners model. Now further stud-
ies are needed for developing a new approach based on
a CDM fatigue model for accurately evaluating the state
of fatigue for existing structures. This paper aims to
develop a methodology and strategy for fatigue damage
assessment and life prediction of the bridge-deck section
based on online structural health monitoring data and a
fatigue damage model based on the continuum damage
mechanics. A structural model for the fatigue analysis
of steel bridge decks is proposed, in which the structure
is modeled by elastic members and welded connections
with possible initiated cracks described by a damage
variable. The model allows the process of fatigue
initiation at the connections of the bridge-deck section
to be analyzed by a CDM fatigue damage model. The
service life prediction for steel bridge decks based on
the online straintime history of the bridge under the
current volume of trafc could be modied by updating
the representative block of the local stress history by the
future online data, which will be included in the compu-
tational approach.
As mentioned above, the Miners rule has been popu-
lar in fatigue design of steel bridges. However, it can be
shown that a fatigue damage model based on CDM has
clear physical meaning and precise modeling for a dam-
aged structure. In order to compare results of fatigue
damage evaluated by the CDM fatigue damage model, a
modied PalmgrenMiner rule is developed for the same
fatigue problem.
2. Models for fatigue analysis of bridge-deck
sections
2.1. Structural model for fatigue stress analysis
Bridge members usually react to service stress far
below design critical stress in the way of elastic defor-
mation. Still there exist local failures, usually brittle, in
the presence of the notches or some type of geometric
discontinuities such as holes, grooves and welded con-
nections where the stresses are locally elevated. Bridge-
deck sections are constructed by connecting many plates,
beams and bars by welding, rivets or some other way.
Fatigue cracks in a bridge are likely to appear at the
location of these connections. Therefore, the bridge-deck
section should be one of the main parts of the fatigue
analysis for a bridge. Structurally the deck section of the
Tsing Ma Bridge is a hybrid arrangement combining
both truss and box form [2]. As shown in Fig. 1, two
longitudinal trusses to the full depth of the deck at
26 m centers act in conjunction with the steel orthotropic
decks of the upper and lower carriageways to provide
the vertical bending stiffness. Plan diagonal bracings at
the upper and lower levels enable the trusses to provide
lateral bending stiffness. Cross frames of Vierendeel
form are provided at 4.5 m centers with every fourth
frame being supported from suspenders. A stainless steel
cladding along the outer edges of the deck is provided
in order to control air ow across the deck. The nal
external appearance of the Tsing Ma Bridge deck is
therefore of a box with faired edges having continuous
gaps along the top and bottom surface.
Bridge fatigue is a high-cycle fatigue problem where
stress uctuations are low so that the deformation in the
structure is elastic except at notches and welds where
local stresses are concentrated. Therefore the material in
the vicinity of welds would be locally yielded and plastic
or micro-plastic deformation generates there. For accur-
47 Z.X. Li et al. / International Journal of Fatigue 23 (2001) 4553
Fig. 1. A deck section of the Tsing Ma Bridge.
ate evaluation of fatigue stress, fatigue crack initiation
and growth in the vicinity of welds should be properly
modeled and coupled into the structural model of a
bridge-deck section. Fatigue damage in the region of
fatigue crack initiation and growth of cracks in micro-
scale can be well modeled by the CDM concept and
theory. The structural model for fatigue damage analysis
therefore consists of elastic deformed members in which
stresses are below the elastic limit of the material, and
welded connections where fatigue cracks in micro-scale
initiate and grow. The deformation of elastic members
will follow the constitutive law of elasticity, while the
deformation in the vicinity of welds follows the consti-
tutive equations of damaged material. As an example of
such a structural model, a part of the bridge-deck section
considered, a typical longitudinal truss, is shown in Fig.
2. All vertical posts, diagonal bracings, top and bottom
chords are non-damage members, and welds at connec-
tions of these members are critical locations of fatigue
damage.
Suppose the macro-plastic strain does not exist in
members of a steel bridge on the preliminary stage of
service, and the only irreversible strain is micro-plastic
which will be considered later in the fatigue damage
model. The fatigue damage in the vicinity of welds is
coupled into the material constitutive law using the con-
cept of effective stress [18]. Therefore, the constitutive
equation for the proposed model of bridge structure can
be written as:
Fig. 2. A typical longitudinal truss.
e
ij
C
ijkl
s
kl
for non-damage members (1)
where e
ij
and s
kl
are the rate of strain and stress tensor,
respectively, C
ijkl
is a tensor of elastic compliance. If
damage in the material is described by introducing a sca-
lar variable, D, the deformation behavior of damaged
material is modeled as follows:
e
ij
C
ijkl
s

kl
in the vicinity of welds (2)
where s

kl
is a tensor of effective stresses that can be
expressed as follows [27]:
s

ij

s
ij

1D

s
ij

1hD
(3)
in which the McCauley bracket x yields x for x0
and 0 for x0, and 0h1 is associated with closing
of micro-cracking when the local stress state is com-
pressive.
The above model can be used not only to analyze
fatigue damage of steel structures, but also for nite
element modeling of damaged structures to calculate the
stressstrain response and the state of damage if the
evolution law of the damage variable is provided.
Further explanation of the structural model and strategy
for nite element modeling of a bridge-deck section will
be reported in a separate paper on the fatigue analysis
of steel bridges by the nite element method.
2.2. Local model for high-cycle fatigue damage
The damage evolution model for the high-cycle
fatigue damage problem must be written in terms of
stress since the irreversible strain involved in the fatigue
deformation is only micro-plastic, which is not measur-
able and difcult to calculate. Based on the theory of
thermodynamics and potential of dissipation, a general
damage model can be written as a function of the
accumulated plastic or micro-plastic strain, the strain
energy density release rate and current state of damage
[17]. The micro-plastic strain, usually neglected in a low
cycle fatigue problem, and its accumulation must be con-
sidered when high-cycle fatigue damage occurred in the
elastic range, even if macro-plastic strain does not exist.
48 Z.X. Li et al. / International Journal of Fatigue 23 (2001) 4553
Therefore, the damage evolution equation here is writ-
ten as:
D

=
R
v
s
2
eq
|s
eq
s
eq
|
b
B(1D)
a
s
eq

=0
if
s

s
f
s

s
f
(4)
where B, a and b are constants of material, s
f
is the
stress limit to fatigue, s
eq
is Von Mises equivalent stress
dened by:
s
eq

3
2
s
D
ij
s
D
ij
1/2
(5)
in which s
D
ij
is the deviatoric stress tensor. In Eq. (4),
s
*
is the damage equivalent stress which, for damage,
acts as the Von Mises equivalent stress used in plasticity,
and in the case of a one-dimension problem, it is
given by:
s

s if s0; s

1D
1hD

1/2
|s| if s0 (6)
and R
v
is a triaxial function to express the inuence of
the triaxial ratio of the stress state, and in pure bilateral
conditions (h=1) this function reduces to:
R
v

s
eq

2
3
(1n)3(12n)

s
H
s
eq

2
(7)
In the one-dimensional case, Eq. (4) becomes:
D

=
s
2
|ss |
b
B(1D)
a
s
D

=0
if
s

s
f
s

s
f
(8)
in which s =s
m
is the mean stress.
Eq. (4) or Eq. (8) is a general constitutive model for
high cycle fatigue, which is valid for any kind of load-
ing. It has to be integrated over time for each cycle if
the cycles are different in magnitude of stress range and
mean stress. The identication of the coefcients B, a
and b in the model needs the Woehler curve for uniaxial
periodic fatigue and the measurement of damage
obtained by means of a strain controlled fatigue test.
3. Methodology and strategy of fatigue analysis
3.1. Fatigue analysis approach on the basis of online
monitoring data
With the above model for fatigue analysis, the assess-
ment procedure for the fatigue of a bridge can be
described as follows:
1. Global structural analysis by the nite element com-
putations to determine the components of nominal
stress corresponding to fatigue and extreme loads
generated by cyclic loading.
2. Local stress analysis to determine the hot-spot stress
concentration factors and the degree of the localiz-
ation, i.e. the proportion of the localized stress to total
stress relevant to the location where cracks or micro-
cracks may be generated.
3. Determination of stress spectrum to generate the hot-
spot stress range and mean stress histogram for the
location where fatigue cracks may occur.
4. Fatigue damage analysis for the crack initiation period
or fatigue crack growth analysis for the crack growth
period to determine the state of fatigue damage and
the remaining service life of the structure.
It can be seen from the above that the fatigue analysis
is rstly based on the stress analysis to get the distri-
bution of the stress in structures. The stress analysis is
known to be very difcult for a large complicated struc-
ture, such as large cable supported bridges, since there
are too many uncertain conditions remaining in the cal-
culation. The development of the structural health moni-
toring system provides an efcient way to get an accur-
ate evaluation of local stress history for bridge-deck
sections under actual trafc. As an example of the sys-
tem, the WASHMS [1,2] has been installed for the three
cable-supported bridges in Hong Kong, which is con-
sidered to be the most heavily instrumented bridge pro-
ject of the integrated online monitoring system in the
world. The bridge response is monitored by a total of
approximately 774 sensors, including accelerometers,
strain gauges, displacement transducers, level sensors,
anemometers, temperature sensors and weigh-in-motion
sensors installed permanently on the bridges, and the
data acquisition and processing system. The strain
gauges were installed to measure stresses at bridge-deck
sections. As shown in Fig. 3, the locations of strain
gauges installed for the Tsing Ma Bridge including rail
track sections at Chainage (CH) 24662.50, bridge-deck
trough section at CH 24664.75 and deck at tower and
rocker bearing links at CH 23623.00. The most critical
parts of the cross-frames for fatigue damage have been
identied during the design of the WASHMS by the
Flint and Neill Partnership [28]. Therefore, all these
strain gauges are supposed to be installed at critical parts
of the bridge-deck sections. Then, data measured by
these strain gauges can be used to obtain nominal
stresses corresponding to critical parts of the bridge-deck
section. The local stresses at hot-spots are then calcu-
lated by local stress analysis with the nite element
method or conveniently corrected using appropriate cor-
rection factors, which is described in the companion
paperPart II [31].
Straintime histories recorded by strain gauges
installed at bridge-deck sections now become the most
49 Z.X. Li et al. / International Journal of Fatigue 23 (2001) 4553
Fig. 3. Strain gauges layout of the Tsing Ma Bridge.
useful database for online fatigue assessment of the
bridge. Based on the measured data from the structural
health monitoring system, proposed procedures for the
fatigue analysis of bridge-deck sections are suggested
and the ow chart of the procedures is shown in Fig. 4.
Fig. 4. A ow chart of procedures for the fatigue analysis of a bridge-deck section.
3.2. The representative block of cycles of stress at
trafc loading
Fatigue damage is dependent on the nature of the
stresstime history that is generated by the live load on
50 Z.X. Li et al. / International Journal of Fatigue 23 (2001) 4553
the bridge due to trafc. The measured data of strain
time history at the gauge locations of the bridge is
observed to be very complicated. Fig. 5 shows part of
strain histories recorded by a strain gauge in the cross
frame at CH 24662.50. It is seen that the straintime
history consists of complicated cycles with variable
amplitude of strain range and varying values of mean
strain. If the original straintime history was directly
used to assess the fatigue by the rain-ow counting pro-
cedures, it would be almost impossible to record the
strain from the time the bridge started on service to the
end, and to put all of the recorded data into a computer
to get the rain-ow counting results. Therefore it is
necessary to analyze the character of recorded strain
time history.
The recorded strain is generated by the live load on
the bridge due to the trafc loading, and the trafc on
a bridge can be considered as approximately repeated
from day to day. The investigation on the particular fea-
ture of the straintime history [29] showed that the
straintime history curves are similar from day to day.
They have common characteristics in curve shape and
magnitude of cycles, which can be observed from
recorded strain cyclic histories shown in Fig. 5, three
days straintime history at (a) May 20, (b) May 15 and
(c) April 27, 1999.
Based on the above analysis, the real-time strain
cycles recorded by strain gauges will be modeled as a
block of repeated cycles in the fatigue analysis of bridge-
deck sections based on the monitoring data. The original
Fig. 5. Straintime history at (a) May 20, 1999; (b) May 15, 1999; (c) April 27, 1999.
strain history is therefore represented by the block-
repeated cycles in which a standard block, dened as a
representative block of daily cycles, is repeated every
day. The representative block of cycles can be obtained
by statistical analysis of many samples of daily strain
time history recorded at the same location. The detailed
procedures for obtaining the representative block of
cycles are shown in Fig. 4.
3.3. Fatigue damage rate generated by the
representative block of cycles
After obtaining the representative block of cycles of
strain history, the rain-ow counting method is used to
determine the stress spectrum as shown in the ow chart
of the procedures for the fatigue analysis (see Fig. 4). It
gives stress range, relevant mean stress and its cycle
counts in the representative block of cycles of stress
time history. Then the fatigue damage Eq. (4) or Eq. (8)
should be integrated for each cycle over time of the
block, from which the fatigue damage generated by the
representative block of the cycles can be derived.
Considering rstly the mean stress s
m
=0 for simplicity
of the calculation, and neglecting the variation of
(1D)
a
in the integration over one block of cycles, inte-
grating Eq. (8) over one block yields:

D
dD
dN
bl
D
dD

m
rb
i1

s
Mri
0
s
b+2
B(1D)
a
ds for s 0 (9)
51 Z.X. Li et al. / International Journal of Fatigue 23 (2001) 4553
where m
rb
is the number of cycles with the maximum
stress over the stress limit to fatigue in the representative
block, N
bl
the number of blocks, and s
Mri
the maximum
stress at the ith cycle and its value is greater than the
stress limit to fatigue.
Noticing the stress amplitude s
ar
when s
m
=0 is equal
to the maximum stress s
Mr
, the fatigue damage rate gen-
erated in one block of cycles is obtained from Eq. (9)
dD
dN
bl

m
rb
i1
s
b+3
ari
B(1D)
a
(b+3)
(10)
Now, consider the effect of mean stresses by using
the equation similar to that of Morrow, or that of Smith,
Watson and Topper [30]:
s
ar
(s
M
s
a
)
1/2
(11)
where s
a
represents stress amplitude when mean stress
s
m
0.
Substituting the above equation into Eq. (10), and
since s
M
=s
a
+s
m
, it gives:
dD
dN
bl

m
rb
i1
[(s
ai
+s
mi
)s
ai
]
(b+3)/2
B(1D)
a
(b+3)
for s
Mi
s
f
(12)
The above equation gives the fatigue damage rate gen-
erated by the representative block of cycles of stress. It
depends on the stress spectrum in the representative
block of cycles of the local stress history, the state of
damage and the material properties in elasticity, micro-
plasticity and damage.
3.4. Updating of the stress spectrum in the
representative block of cycles
The stress spectrum in the above analysis is based on
the online strain generated by the live load on the bridge
that is mainly due to the current volume of trafc load-
ing. It has to be updated from time to time since the
trafc volume on the bridge varies with the economic
development in Hong Kong. The updating method is
as follows:
(a) Record the online straintime history on the
updated time, for the period at least over one week.
For considering the effect of temperature in different
seasons, it is better to record online data for four
weeks in four seasons, respectively.
(b) Determine the representative block of cycles of
the local stress history by statistical analysis of the
daily samples of the local stress history at the same
location of interest.
(c) Determine the updated stress spectrum (s
ai
,s
mi
)
by use of the rain-ow counting program.
Then the updated fatigue damage rate generated by the
representative block of cycles of stress can be obtained
by replacing the updated stress spectrum into Eq. (12)
as follows:
dD
dN
bl

m
rb
i1
[(s
ai
+s
mi
)s
ai
]
(b+3)/2
B(1D)
a
(b+3)
for s
Mi
s
f
(13)
4. Fatigue damage evaluation and service life
prediction
With the fatigue damage rate given by Eqs. (12) and
(13), the fatigue damage can be evaluated by the integral
of these equations over time for a blocked cycle. Inte-
grating Eq. (12) at the initiation condition: N
bl
=0, D=0,
it gives
D1

1
(a+1)N
bl
B(b+3)

m
rb
i1
[(s
ai
s
mi
)s
ai
]
(b+3)/2

1/(a+1)
(14)
If there are N
0
bl
blocks with the stress spectrum s
ai
,
s
mi
(i=1, 2, . . ., m
rb
) and then more blocks with updated
stress spectrum (s
ai
,s
mi
) (i=1, 2, . . ., m
rb
), the integral
of Eq. (13) at the initiation condition: N
bl
=N
0
bl
, D=D
0
gives:
D1

(1D
0
)
a+1

(a+1)(N
bl
N
0
bl
)
B(b+3)

m
rb
i1
[(s
ai
(15)
s
mi
)s
ai
]
(b+3)/2

1/(a+1)
in which D
0
is obtained by Eq. (14) when N
bl
=N
0
bl
.
For multiple updates made on the service period of
the bridge, the evaluation of the fatigue damage is an
iterative process. The iterative equation after the (k+1)th
updating of the stress spectrum is expressed as:
D
(k1)
1

(1D
(k1)
0
)
a1

(a1)(N
(k1)
bl
N
(k1)
bl(0)
)
B(b3)

m
(k1)
rb
i1
[(s
(k1)
ai
(16)
s
(k1)
mi
)s
(k1)
ai
]
(b3)/2

1/(a1)
D
(k1)
0
D
k
; N
(k1)
bl(0)
N
k
bl
in which, N
k
bl
, the number of blocks before the (k+1)th
updating, equals the sum of all increments of the block
in the updating period N
j
bl
(j=0, 1, . . ., k):
N
k
bl
= S
k
j0
N
j
bl
.
52 Z.X. Li et al. / International Journal of Fatigue 23 (2001) 4553
The prediction of the service life can be obtained from
Eq. (15) when the fatigue damage reaches its value at
failure D
f
:
N
f
bl

B(b+3)[(1D
0
)
a+1
(1D
f
)
a+1
]
(a+1)

m
rb
i1
([(s
ai
+s
mi
)s
ai
]
(b+3)/2
)
N
0
bl
(17)
where, if the KUth update is made on the service period
of the bridge before failure, N
0
bl
is the number of blocks
before the last updating, and D
0
=D
KU
0
=D
(KU1)
is calcu-
lated by using the iterative Eq. (16).
For reasons of comparison, the service life prediction
is calculated using a modied PalmgrenMiner rule. The
PalmgrenMiner rule simply states that fatigue failure is
expected when life fractions sum to unity. It is a rule
most commonly used because of its simplicity. However,
the summations of cycle ratios at fatigue failure were
observed not to be unity in many situations, especially
for structures under variable-amplitude loading [5]. In
many cases, the summation was found to be less than
unity and with the value in the region of 0.60.9. From
the point of view of continuum damage mechanics,
fatigue failure occurs when fatigue damage in the
material reaches a limit value of damage. The limit value
of damage, D
f
is an intrinsic material property to be
dependent on the durability of the material. It can be
determined from experiment for a given material and its
value ranges between 0.15 and 0.85, depending on the
material [22]. The parameter D
f
has been used in the
above CDM model, so it is necessary to use the same
parameter as a threshold value for fatigue failure in the
compared calculation by Miners rule.
The summation of cycle ratios in the left of the equ-
ation of the PalmgrenMiner rule is actually a kind of
expression of cumulative fatigue damage, and Miners
rule therefore implies that fatigue failure is expected
when the value of cumulative fatigue damage, expressed
as a life fractions sum, equals unity. A modied
PalmgrenMiner rule is proposed here, which states that
fatigue failure is expected when the value of cumulative
fatigue damage, expressed as a life fractions sum,
reaches the threshold value of the fatigue limit, i.e., the
value of fatigue damage at failure, D
f
. For the blocked
cycles of local stress history discussed here, the rule is
expressed as follows:
N
f
bl
i
N
i
N
fi

one block
D
f
(18)
where N
f
bl
is the number of blocks to failure, and N
i
, N
fi
are the number of cycles at the stress amplitude s
ai
and
relevant number of cycles to failure from the SN curve
for s
ai
, respectively, in the representative block of cycles
of local stress history. Again if the KUth update is made
before the fatigue failure, the Eq. (18) should be revised
as follows:

KU
k1

N
k
bl
i
N
ki
N
fki

for the kth updated block

D
f
(19)
where N
k
bl
is the increment in the number of blocks with
kth updated representative block, and S
KU
k1
N
k
bl
=N
f
bl
the
number of blocks to failure.
According to the standard [3], the number of cycles
to fatigue failure at the stress range s
i
used in the above
equation can be obtained from the following equation:
N
fi
K
2
s
m
i
(20)
where K
2
and m have values given in the standard [3]
for different class of details.
5. Concluding remarks
This paper studies how to evaluate fatigue damage and
to predict the service life of bridge-deck sections for a
large suspension bridge with a permanent structural
health monitoring system installed such as that for the
Tsing Ma Bridge. The systematic study has developed
the methodology and strategy for the fatigue analysis.
The fatigue damage models presented in the paper allow
calculation of the cumulative damage of high-cycle
fatigue in a bridge which is considered at a continuum
scale as a deteriorating process with its physical mech-
anism such as fatigue crack initiation and growth.
The developed methodology and strategy allows
fatigue assessment to be carried out by taking full advan-
tage of the online structural health monitoring data.
Therefore, for bridges with the structural health monitor-
ing system, accumulative fatigue damage can be
assessed by the use of the proposed approach for bridge-
deck sections under actual trafc loading.
The stress spectrum of the representative block of
cycles under normal trafc loading can be obtained by
rain-ow cycles counting and statistical analyzing of
daily samples of the stress spectrum. The location of
possible fatigue failure will be detected by comparing
values of fatigue damage at the toe of the welds near
the critical members, and therefore the service life of the
bridge is the predicted life of the detected critical mem-
ber. As an important part of the computation, the updat-
ing of the stress spectrum of a representative block of
cycles at the hot-spot by the real-time monitoring data
is included in the proposed approach to cope with the
change of the trafc volume in the future.
The method and strategy developed in this paper have
been applied to evaluate the fatigue life of the Tsing
Ma Bridge, and the calculated results have veried the
validity of the proposed approach, which will be
presented in the companion paperPart II [31].
53 Z.X. Li et al. / International Journal of Fatigue 23 (2001) 4553
Acknowledgements
Funding support to the project (Project Code: G-
YY15) by the Hong Kong Polytechnic University is
gratefully acknowledged. The writers wish to thank the
Highways Department of the Hong Kong SAR Govern-
ment for their support throughout the project.
References
[1] Lau CK, Wong KY. Lantau xed crossing and Ting Kau
Bridgethe bridge monitoring system. In: Proceedings of the
Second International Symposium on Civil Infrastructure Systems,
Hong Kong, 1996:8697.
[2] Lau CK, Wong KY. Design, construction and monitoring of the
three key cable-supported bridges in Hong Kong. In: Lee PKK,
editor. Structures in the new millennium. Rotterdam, Nether-
lands: A.A. Balkema, 1997:10515.
[3] BSI, BS5400: Part 10, Code of Practice for Fatigue, 1982
[4] Fatemi A, Yang L. Cumulative fatigue damage and life prediction
theories: a survey of the state of the art for homogeneous
materials. International Journal of Fatigue 1998;20(1):934.
[5] Schijve J. Predictions on fatigue life and crack growth as an
engineering problem, a state of the art survey. In: Fatigue 96,
Oxford: Pergamon, 1996:114964.
[6] Sjostrom C. Overview of methodologies for prediction of service
life. In: Masters LW, editor. Problems in service life prediction
of building and construction materials. Kluwer Academic Pub-
lishers, 1997:320.
[7] Miner MA. Cumulative damage in fatigue. Journal of Applied
Mechanics 1945;67:A15964.
[8] Manson SS. Interfaces between fatigue, creep, and fracture. Inter-
national Journal of Fracture Mechanics 1966;2:32863.
[9] Manson SS, Halford GR. Re-examination of cumulative fatigue
damage analysisan engineering perspective. Engineering Frac-
ture Mechanics 1986;25(56):53971.
[10] Manson SS, Halford GR. Practical implementation of the double
linear damage rule and damage curve approach for treating cumu-
lative fatigue damage. International Journal of Fracture
1981;17(2):16995.
[11] Halford GR. The energy required for fatigue. Journal of Materials
1966;1(1):318.
[12] Golos K, Ellyin F. A total strain energy density theory for cumu-
lative fatigue damage. ASME Journal of Pressure Vessel Tech-
nology 1988;110:3641.
[13] Ellyin F, Kujawski D. plastic strain energy in fatigue failure.
ASME Journal of Pressure Vessel Technology 1984;106:3427.
[14] Kliman V. Fatigue life prediction for a material under programm-
able loading using the cyclic stressstrain properties. Materials
Sciences and Engineering 1984;68(1):110.
[15] Kliman V, Bily M. Hysteresis energy of cyclic loading. Materials
Science and Engineering 1984;68:118.
[16] Kachanov LM. Introduction to continuum damage mechanics.
The Netherlands: Martinus Nijhoff, 1986.
[17] Krajcinovic D, Lemaitre J. Continuum damage mechanics: theory
and applications. Vienna: Springer, 1987.
[18] Lemaitre J. A course on damage mechanics. Springer Verlag,
1996.
[19] Lemaitre J, Chaboche JL. Mechanics of solid materials. Cam-
bridge, UK: Cambridge University Press, 1990.
[20] Chaboche JL, Lesne PM. A non-linear continuous fatigue damage
model. Fatigue and Fracture of Engineering Materials and Struc-
tures 1988;11(1):17.
[21] Bhattacharya B, Ellingwood B. Continuum damage mechanics
analysis of fatigue crack initiation. International Journal of
Fatigue 1998;20(9):6319.
[22] Zhao ZW, Haldar A. Bridge fatigue damage evaluation and
updating using non-destructive inspections. Engineering Fracture
Mechanics 1996;53(5):77588.
[23] Zheng XL, Li Z, Shi YJ, Yang YM, Shi ZJ. Fatigue performance
of old bridge steel and the procedures for life prediction with
given survivability. Engineering Fracture Mechanics
1996;53(2):25162.
[24] Agerskov H, Nielsen JA. Fatigue in steel highway bridge under
random loading. Journal of Structural Engineering, ASCE
1999;125:15262.
[25] Enright MP, Frangopol DM. Service-life prediction of deteriorat-
ing concrete bridges. Journal of Structural Engineering
1998;124(3):30917.
[26] Mori Y, Ellingwood BR. Reliability-based service-life assess-
ment of aging concrete structures. Journal of Structural Engineer-
ing 1993;119(5):160021.
[27] Li ZX, Xiao LG. Finite element analysis of local damage and
post-failure behavior for strain-softening solid. International Jour-
nal of Fracture 1996;80(1):8595.
[28] Flint and Neill Partnership. Lantau xed crossing and Ting Kau
Bridge, wind and structural health monitoring criticality and vul-
nerability ratings review. Highways Department, Government of
Hong Kong, January, 1998.
[29] Chan THT, Ko JM, Li ZX. Fatigue analysis for steel bridge-deck
under blocked cycles of trafc loading. In: SPIEs 5th Inter-
national Symposium on Nondestructive Evaluation and Health
Monitoring of Aging Infrastructure, Newport Beach, California,
USA, 79 March, 2000.
[30] Smith KN, Watson P, Topper TH. A stressstrain function for the
fatigue of metals. Journal of Materials, ASTM 1970;5(4):76778.
[31] Chan THT, Li ZX, Ko JM. Fatigue analysis and life prediction
of bridges with structural health monitoring dataPart II: appli-
cations. International Journal of Fatigue 2000;23(1):5563.

Вам также может понравиться