You are on page 1of 6

Fatigue evaluation of highway bridges

George Tsiatas
*
, Shane M. Palmquist
Department of Civil Engineering, University of Rhode Island, 1 Lippitt Road, Kingston, RI 02881, USA
Abstract
The present work provides a comparison of current AASHTO bridge fatigue guidelines with fracture mechanics procedures. Specically,
the remaining fatigue life of several actual bridges with welded cover plate ends is estimated using the AASHTO specications. The
estimates are compared with predictions made using linear elastic fracture mechanics principles where the time required for an initial
aw to propagate to a critical depth is calculated. It is found that fatigue lives of actual steel highway bridges as determined using fracture
mechanics far exceed the remaining safe fatigue life predictions made with current AASHTO guide specications. For the case of redundant
bridges, an adjustment factor is introduced which, at various probability levels, can produce closer estimates of bridge fatigue lives between
the AASHTO specications and fracture mechanics. q1998 Elsevier Science Ltd. All rights reserved.
Keywords: Fatigue life prediction; Bridge fatigue; Remaining safe fatigue life
1. Introduction
In 1990 the AASHTO Guide Specications for Fatigue
Evaluation of Existing Steel Bridges (AASHTO, 1990)
adopted the NCHRP Report 299 [1, 2] which represents a
probabilistic limit-state approach for fatigue evaluation of
steel bridges. A major component of this is the use of a
special fatigue truck which is based on actual truck trafc
spectra from 30 sites nationwide and more than 27,000
observed trucks [3]. The special fatigue truck brings the
calculated stress range to a more realistic level than earlier
specications which typically used a design truck such as
the HS20.
Two levels of fatigue performance may be estimated
using the AASHTO guide specications. They are the
remaining mean fatigue life and the remaining safe fatigue
life. The rst corresponds to a 50% probability of excee-
dance. The second corresponds to a 97.7% probability for
bridge details with redundant members that the actual fati-
gue life will exceed the predicted value. For bridge details
with nonredundant members, the probability is 99.9%.
These probabilities are comparable to the safety levels in
the fatigue provisions of the 1983 AASHTO Standard
Specications for Highway Bridges [2].
Clarity as to the redundancy of bridges with three girders
is lacking in the specications [1]. To avoid this dilemma
and to remain conservative, bridges with three girders are
considered herein as nonredundant for purposes of fatigue
evaluations.
In practice, the remaining safe fatigue life is customarily
used because of the inherent conservatism. While this is
appropriate for bridges with nonredundant members in
order to minimize the possibility of collapse, this strict
criterion may not be appropriate for bridges with redundant
members. In fact, it may result in premature monitoring and
expenses as demonstrated by welded cover plate ends from
four actual steel highway bridges studied in the present
work.
2. Theoretical procedure
The AASHTO guide specications equation for the
remaining fatigue life, Y
f
, in years which is consistent
with Miner's linear damage accumulation law is [2]
Y
f

fK 10
6
T
a
CR
S
S
r

3
2a (1)
where f is the safe or mean fatigue life factor; K is the detail
constant relating its susceptibility to fatigue damage and
depends on the AASHTO fatigue category of the bridge
detail; T
a
is the estimated lifetime average daily truck
volume, ADTT, in the outer lane; C is the number of stress
cycles per truck passage; R
S
is the reliability factor; S
r
is the
nominal stress range resulting from the live load of truck
trafc in units of ksi; and a is the present age of the bridge in
years. This equation was developed using traditional,
English units.
Probabilistic Engineering Mechanics 14 (1999) 189194
0266-8920/99/$ - see front matter q 1998 Elsevier Science Ltd. All rights reserved.
PII S0266-8920(98)00030-7
* Corresponding author. Tel.: 1 1-401-874-5117; Fax: 1 1-401-874-
2786; e-mail: tsiatasg@egr.uri.edu.
From linear elastic fracture mechanics (LEFM), the fati-
gue life in cycles may be found from the following crack
growth relationship [4]:
da
dN
ADK
m
(2)
where da=dN is the crack growth rate in units of mm/cycle;
DK is the stress intensity factor range which depends on the
properties of the material, the crack orientation, and the size
and shape of the crack; and A and m are constants which
depend on material variables, environment, load frequency,
temperature, and the applied stress range. For structural
steels, the crack growth exponent, m, is commonly assumed
as 3.0 [4]. For welded structural steels (excluding A514
steel), the upper bound value of the crack growth constant,
A, is 2.18 10
213
using units of mm for crack size and
MPa
p
m for the stress intensity factor range, DK [4]. For
this system of units, the average crack growth constant is
1.24 10
213
[5]. Fisher [6] has used the upper bound value
for the crack growth constant, whereas Zhao et al. [7] used
the average value for the crack growth constant. In the
present work, the average value is used as this more accu-
rately reects the actual fatigue crack growth.
The LEFM approach only considers macrocrack growth,
region II of Fig. 1. It does not consider crack nucleation or
microcrack growth, region I of Fig. 1. In fact, crack initia-
tion can be a substantial portion of the total fatigue life [8].
In addition to this, LEFM does not consider the fatigue life
beyond the critical crack size, region III of Fig. 1. Fatigue
life estimates based on LEFM are inherently conservative so
using the upper bound value for the crack growth constant is
not justied.
For an innite plate subjected to cyclic, tensile stress with
a central through crack, the stress intensity factor range is
dened as
DK S
r

pa
p
(3)
where S
r
is the applied range in live load stress which in this
study is determined in accordance with the AASHTO guide
specications, and a is half the crack length. Factors are
then used to modify this for other types where crack and
specimen geometry are different. These factors account for
the stress gradient acting near the crack tip, F
g
; nite width
or thickness of the member, F
w
; a free surface, F
s
; and a
three-dimensional crack, F
e
.
For the case of an assumed semielliptical crack propagat-
ing at the toe of a weld of a cover plate end, these factors are
as follows [6]:
F
g

23:539 ln
Z
t
f
11:981 ln
t
cp
t
f
15:798
1 16:789
a
t
f

0:4348
(4)
F
w

sec
pa
2b
r
(5)
F
s
1:211 20:186

a
ca
r
(6)
F
e

1
Zp=2
0

1 2
ca
2
2a
2
ca
2
sin
2
u
s
du
(7)
where Z is the weld leg size, t
f
is the ange thickness, t
cp
is
the cover plate thickness, a is the crack depth, b is half the
ange width of the girder, c is half the crack length as a
function of crack depth, and u is the angle for an elliptical
crack. The relation between c and a is given by [6]:
ca 3:549a
1:133
(8)
Using the technique of separable equations from differ-
ential equations and including the above-mentioned factors,
the fatigue life in cycles, N, found from Eq. (2) is
N
1
A
Za
cr
a
i
1
F
g
F
w
F
s
F
e
S
r

pa
p

m
da (9)
where a
i
and a
cr
are the initial and critical crack depths,
respectively. For AASHTO fatigue category E details,
such as welded cover plate ends, the mean initial crack
depth is approximately 0.51 mm for a lognormal distribu-
tion [9]. In service, girders with cracks at welded cover plate
ends of approximately 10.2 mm in depth would soon fail as
a result of fatigue [4]. These values are used herein as the
initial crack depth, a
i
, and the critical crack depth, a
cr
,
respectively.
Having estimated the fatigue life in cycles, N, the remain-
ing fatigue life in years, R, can be calculated from the
following equation which was proposed by the Pennsylva-
nia Department of Transportation [10, 11]:
R
ln
N 2MGF
2
365 ADTTn
2
1 1GF
2

11

ln1 1GF
2

(10)
where N is the fatigue life in cycles of the detail, which
herein is determined in accordance with LEFM; M is the
accumulated number of cycles based on past truck volume
and growth factors; ADTT(n
2
) which was dened
previously is based on the most recent year available, n
2
;
G. Tsiatas, S.M. Palmquist / Probabilistic Engineering Mechanics 14 (1999) 189194 190
Fig. 1. Fatigue crack growth versus stress intensity factor range.
and GF
2
is the future growth factor in percent. The accumu-
lated number of cycles M is estimated by the following
expression [10]:
M 365 ADTTn
1

1 1GF
1

n
21
GF
1
(11)
where ADTT(n
1
) is based on the year the bridge was built,
GF
1
is the past growth factor in percent, and n is the number
of years the bridge has been in service.
3. Application
Fatigue cracks developing at the toe of welds of cover
plate ends represent the majority of fatigue-related problems
in Rhode Island. Originally, these welded cover plates were
used to decrease the depth of the steel girder while increas-
ing the moment capacity. As a result, cover plates did not
span the entire length of the girder.
Four steel girder bridges with fatigue-prone welded cover
plate ends having redundant members are examined using
AASHTO guide specications and LEFM. These details
have been selected from typical Rhode Island bridges
built between 1958 and 1966. All are simply supported as
is common in Rhode Island, and all girders have ange
thicknesses greater than 20 mm which results in an
AASHTO fatigue category of E
0
. This is the worst possible
AASHTO fatigue rating. Tables 1 and 2 list the pertinent
characteristics of each steel bridge and cover plate detail,
respectively. Figs 24 show a general view of bridge 2 as
well as specic details related to the covered plate ends.
This single span bridge built in 1966 is located on interstate
95 over Route 3 in Richmond, RI.
Remaining fatigue life predictions for the four bridge
details using the two methods are given in Table 3 along
with the nominal stress range, S
r
, and the ADTT, respec-
tively. Neither LEFM or the AASHTO guide specications
can accurately determine when fracture will actually occur.
Stress range, truck volumes, and growth factors are always
changing and some approximation is required. However,
both methods are rational and can be used for estimation
purposes.
From Table 3, the nominal stress range of all four
bridges is narrow as expected since the superstructure
geometries are similar. However, the ADTT has a far
greater range, from 2000 to 9580, inclusive. Differences
in the ADTT account for the majority of the variations in
the estimated remaining fatigue lives. Bridges 1 and 2
have moderate ADTT volumes of 2000 and 2240, respec-
tively. Their estimated remaining fatigue life is substan-
tial. Bridges 3 and 4 have higher ADTT volumes of 9580
and 7650, respectively. Their estimated remaining fatigue
life is exhausted.
More importantly, welded cover plate ends of steel
bridges 1 and 2 have a remaining safe fatigue life of 8.5
and 2 1.5 years, while the remaining mean fatigue life is
175 and 103 years, respectively. The difference between the
remaining safe fatigue life and remaining mean fatigue life
may appear unusually large, but this is typical. This can be
explained by examining the total safe fatigue life and the
total mean fatigue life.
According to Moses et al. [2], the ratio of mean to allow-
able stress range averages 1.243 for AASHTO fatigue cate-
G. Tsiatas, S.M. Palmquist / Probabilistic Engineering Mechanics 14 (1999) 189194 191
Table 1
Bridge characteristics
Bridge no. Year built Span length of interest Deck type
a
Girder type
b
and section Haunch (mm) Girder spacing (m) Slab thickness (m)
1 1959 16.5 Comp. Int. W840 193 0 1.96 0.178
2 1966 25.9 Comp. Ext. W920 342 50.8 2.31 0.191
3 1965 23.5 Comp. Int. W920 223 25.4 2.13 0.178
4 1958 25.3 Non. Int. W920 417 0 1.86 NA
a
Deck type refers to whether the deck is composite or noncomposite.
b
Girder type refers to whether the girder is an interior or exterior member.
Table 2
Cover plate end characteristics
Bridge no. Support to
cover plate
a
(m)
Cover plate
thickness (mm)
Weld leg
size (m)
1 3.23 9.5 8.0
2 5.49 22.2 15.9
3 3.20 31.8 15.9
4 5.33 15.9 12.7
a
Support to cover plate gives the horizontal distance from the support to
the nearest cover plate end. Fig. 2. Bridge 2.
gories B to E
0
. The average ratio of total mean fatigue life to
total safe fatigue life is
R
m-s
1:243R
S

3
(12)
where the reliability factor, R
S
, corresponding to the remain-
ing safe fatigue life of bridge details with redundant
members is 1.35. Therefore, R
m-s
is approximately 5. For
bridge details with nonredundant members, this ratio is
approximately 10.
In the case of bridge 2 a low remaining safe fatigue life of
2 1.5 years prompted a complete retrot of the cover plate
ends. Every detail of this type on the bridge was bolt spliced,
Fig. 5. Was this retrot appropriate given that the remaining
mean fatigue life and LEFM estimations are 103.0 and 39.4
years, respectively?
4. Recommended fatigue criteria
As shown in Table 3, LEFM fatigue lives are substantially
longer than the remaining safe fatigue lives of the AASHTO
guide specications. Current practice uses the remaining
safe fatigue life as recommended by the AASHTO guide
specications as the governing fatigue criteria. Since the
remaining safe fatigue life is often only calculated corre-
sponding to major rehabilitation projects (usually 23
times during the life of the bridge), remaining safe fatigue
life values of 30 years or less are often assumed as nearing
the end of the fatigue life thereby prompting the engineer to
consider retrotting or replacement, as in bridge 2.
For the case of nonredundant bridges the conservatism
associated with the use of the safe remaining fatigue life
is needed but for the case of redundant bridges the criteria
could be relaxed. Most of the research involved with the
development of the fatigue specications concentrated on
the strength part of the problem. Current research on the
loading component may lead to improved predictions and
a more narrow margin between the mean and the safe fati-
gue life estimates. An adjustment factor is developed here
which can be introduced in the fatigue life equation for the
case of bridge details with redundant members and produce
closer agreement with the fracture mechanics-based
estimates for the four bridges considered.
As a starting point the equation for the remaining mean
fatigue life of a redundant bridge member is used. This is
derived from Eq. (1) with f 2 and R
s
1:
Y
f

2K 10
6
T
a
CS
3
r
2a (13)
G. Tsiatas, S.M. Palmquist / Probabilistic Engineering Mechanics 14 (1999) 189194 192
Fig. 3. Bolt spliced cover plate end, exterior girder, bridge 2.
Fig. 5. Section view, bridge 2.
Fig. 4. Plan view, bridge 2.
The total mean fatigue life is multiplied by a factor R
f
:
Y
f
R
f
2K 10
6
T
a
CS
3
r
2a (14)
where R
f
is here called the redundancy factor and is less than
unity. Different values of R
f
correspond to different prob-
abilities that the actual remaining fatigue life will exceed the
remaining calculated life. Fig. 6 shows the relation of R
f
and
the probability of exceedance. A value of R
f
0.2 corre-
sponds to the safe fatigue life whereas a value of R
f
1
corresponds to the remaining mean fatigue life. It is not the
intent here to establish appropriate levels of safety but to
demonstrate that various probability levels can be achieved
using the form of Eq. (14).
In the case of the four example bridges a value of R
f
0.5
is arbitrarily selected for better agreement with the fracture
mechanics results. This corresponds to a probability of 85%
that the actual remaining fatigue life will exceed the remain-
ing calculated life. Further simplication of Eq. (14) results
in the following:
Y
f

K 10
6
T
a
CS
3
r
2a (15)
Eq. (15) can be referred to as the remaining practical
fatigue life and results for the four bridges are given in
Table 3. Considering that the remaining safe fatigue life
has been exhausted for nearly all four of the welded cover
plate ends, a remedial decision under current AASHTO
guide specications would have to be made. Using the
remaining practical fatigue life, only two of the four bridges
need to be further investigated for fatigue crack potential.
Although the remaining practical fatigue life is not as
highly conservative as the remaining safe fatigue life, this
level of safety appears reasonable and cost effective. The
redundancy factor, R
f
, can be modied to reect future
research and trends like many of the variables in the
AASHTO guide specications. It should be pointed out
that for steel bridges with nonredundant members, the
remaining safe fatigue life as recommended by the
AASHTO guide specications should be strictly used as
the governing criterion.
5. Conclusions
A study of four redundant bridges with welded cover plate
ends has revealed large differences between the remaining
safe fatigue life and the remaining mean fatigue life as well
as predictions made using a fracture mechanics-based
procedure. Current practice relies primarily on the remain-
ing safe fatigue life as the governing criterion which is more
conservative than the remaining mean fatigue life.
However, it may result in premature monitoring and
expenses. Even worse, state agencies may not follow
through with monitoring or even load posting. For the
case of steel bridges with redundant members an adjustment
factor is introduced and a ``practical'' fatigue life is
G. Tsiatas, S.M. Palmquist / Probabilistic Engineering Mechanics 14 (1999) 189194 193
Table 3
Remaining fatigue life (in years)
Bridge no. S
r
(MPa) ADTT Fatigue life prediction methods
LEFM AASHTO guide specications
Safe Mean Practical
b
1
a
12.0 2000 40.6 8.5 175.0 70.4
2
a
13.5 2240 39.4 2 1.5 103.0 37.2
3
a
12.7 9580 23.1 2 21.5 7.9 10.6
4
a
13.4 7650 2 2.1 2 26.1 4.8 14.6
a
Bridge detail is a welded cover plate end with a redundant member. This detail has an AASHTO fatigue category E
0
.
b
Recommended remaining practical fatigue life for which to make cost effective decisions related to bridge inspection and rehabilitation.
Fig. 6. Redundancy factor versus probability of exceedance.
calculated. Using various values of this factor different
probability levels can be achieved. For a value of 0.2 the
practical and the safe fatigue life coincide. A value of 0.5
produces fatigue estimates which are in better agreement
with fracture mechanics-based estimations for four real
bridges.
Acknowledgements
Funding for this research has been provided by the Rhode
Island Department of Transportation (RIDOT) and the
Federal Highway Administration (FHWA). The authors
would like to thank especially Colin A. Franco, Francis
Manning, and Georgette Shahine with RIDOT and Eileen
Young with FHWA for their input and participation on this
project.
References
[1] AASHTO. Guide specications for fatigue evaluation of existing steel
bridges. Washington, DC: American Association of State Highway
and Transportation Ofcials, 1990.
[2] Moses F, Schilling CG, Raju KS. Fatigue evaluation procedures for
steel bridges, NCHRP Report 299, 1987.
[3] Moses F, Snyder RE, Likins GE. Loading spectrum experienced by
bridge structures in the United States, Report FHWA/RD-85/012.
Warrenville, OH: Bridge Weighing Systems, 1985.
[4] Barsom JM, Rolfe ST. Fracture and fatigue control in structures:
applications of fracture mechanics. Englewood Cliffs, NJ: Prentice-
Hall, 1987.
[5] Hirt MA, Fisher JW. Fatigue crack growth in welded beams.
Engineering Fracture Mechanics 1973;5:.
[6] Fisher JW. Fatigue and fracture in steel: case studies. New York:
Wiley, 1984.
[7] Zhao A, Haldar A, Breen FL. Fatigue-reliability evaluation of steel
bridges. Journal of Structural Engineering 1994;120(5):16081623.
[8] Shilling CG et al., Fatigue of welded steel bridge members under
variable-amplitude loadings, NCHRP Report 188. Washington, DC:
Transportation Research Board, 1978.
[9] Albrecht P, Yazdani N. Risk analysis extending the service life of the
steel bridges, Report FHWA/MD No. 84/01, 1986.
[10] Tsiatas GE, Palmquist SM. An evaluation of AASHTO fatigue
specications based on case histories using fracture mechanics,
RIDOT Report, 1996.
[11] Breon JA, Hasmukh ML. User's manual for computer program
P453110: bridge analysis and rating (BAR7). Pennsylvania
Department of Transportation, 1991.
G. Tsiatas, S.M. Palmquist / Probabilistic Engineering Mechanics 14 (1999) 189194 194

You might also like