FATIGUE-LOAD MODELS FOR GIRDER BRIDGES

By Jeffrey A. Laman1 and Andrzej S. Nowak/ Members, ASeE

ABSTRACT: A fatigue-live-load model is developed for steel girder bridges. The database for the model is
obtai~ed by ~eigh-in-motion (WIM) measurements. Five bridge structures were selected for testing to determine
the site-specific truck parameters and component-specific stress spectrum. The database includes 22,000 truck
files, each con~isting of gross ~ehicl~ weight (GVW), axle weights, and axle spacing. Stress cycles were mea-
sured at the mtdspan of all bndge gtrders and results are presented as cumulative distribution functions. The
WI~ .me~surements confirm that truc~ loads are strongly site-specific. The results also indicate a significant
v.anatton 10 stress spectrum between girders. A three-axle truck is proposed to represent the truck traffic. For
sites w~th l~- and ~ 1-axl~ trucks, an additional four-axle truck is proposed. The developed fatigue live load
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/25/15. Copyright ASCE. For personal use only; all rights reserved.

model IS venfied us10g fatigue-damage analysis to compare the model with measured results.

INTRODUCTION SELECTED BRIDGES AND TEST PROCEDURE

Fatigue is an important consideration in the design and anal- Field measurements were conducted on five steel-girder
ysis of steel bridge structures. Multiple passages of heavy bridges. The important parameters of the selected bridges are
trucks can lead to cracking and premature failure (Khaleel and summarized in Table 1. Strain transducers were attached to all
Anditani 1993). Analysis of fatigue performance involves the girders at the lower, midspan flanges for each of the five
determination of loads and material strength. Material response bridges. Dynamic strain cycles were measured under normal
has been studied by many researchers. Fisher (1977) devel- traffic using the rainflow algorithm. The data were collected
oped SoN curves for categories of details in steel structures. and recorded using two sets of equipment: Truck Weighing
Fisher's work demonstrated the importance of live-load level, System (TWS) with the main unit manufactured by Bridge
particularly magnitude of the live load and frequency of oc- Weighing Systems, Inc. and Stress Measuring System (SMS)
currence. The present paper focuses on fatigue-load models with the main unit manufactured by the SoMat Corp. The
for girder bridges. TWS provided information about truck-traffic parameters, and
The objective of fatigue-load research has largely been to the SMS compiled stress histograms for the girders and other
establish an equivalent fatigue truck that will cause the same components.
cumulative fatigue damage as the normal traffic distribution. The TWS calculates truck gross vehicle weight (GVW) and
A single, equivalent fatigue truck is a very attractive and useful axle weights from the strain history using the bridge influence
tool for the practicing engineer. Schilling (1984), Raju et al. lines. The truck configuration (number of axles and axle spac-
(1990), and The American Association of State Highway and ing) is determined using two parallel sensors attached to the
Transportation Officials (AASHTO) (Guide 1989) suggest that pavement in each traffic lane. For each bridge, the TWS is
the accuracy of the fatigue-truck model is improved by ad- calibrated using a truck with known axle weights. The accuracy
justing the fatigue truck axle weights in proportion to an of measurement varied depending on the number of axles on
equivalent total weight, calculated from the specific site load the vehicles. For vehicles with up to five axles, GVW was de-
distribution. In addition to the equivalent total weight, the termined within 5%, and for ll-axle trucks GVW was deter-
equivalent lane moment has been calculated for each bridge mined within 10%. Axle weights were determined within 20%.
in the study, which may be a more accurate indication of Mi- The SMS collects the strain history under normal traffic and
ner's equivalent stress for use in fatigue calculations, partic- assembles the stress-cycle histogram by the rainflow method
ularly for shorter «20 m) spans. The equivalent lane moment of cycle counting, and other counting methods. The data are
does not, however, include the effects of intermediate smaller then stored to memory and downloaded at the conclusion of
cycles caused by long vehicles crossing the bridge or dynamic the test period. The rainflow method, proposed by Matsuishi
effects. Greater accuracy is achieved when these intermediate and Bndo (1968), counts the number, n, of cycles in each pre-
cycles are included in the fatigue analysis. determined stress range, S/, for a given stress history. The SMS
In the present paper, a fatigue-load model is developed from is capable of recording up to 4 billion cycles per channel for
weigh-in-motion (WIM) measurements (Laman 1995). Statis- extended periods in an unattended mode. The batteries used
tical parameters of stress are calculated for girder bridges. The for the present study enabled testing to continue uninterrupted
for as long as three weeks. .
results indicate that magnitude and frequency of truck loading
are strongly site-specific and component-specific. Based on the
WIM data, a design fatigue truck is developed (Gagarine and TABLE 1. Parameter. of Selected Bridge.
Albrecht 1992). The model is calibrated against measured dy-
No. of Girder Girder Cover
namic strains to achieve uniform reliability against fatigue fail-
No. Location ADTT Span girders spacing size plate
ure. (1 ) (2) (3) (4) (5) (6) (7) (8)
I US-23 over 2,000 23.9 m 6 1.9 m W36xl94 yes
I Assi. Prof. of Civ. Engrg., Dept. of Civ. and Envir. Engrg., The Penn-
Huron River
sylvania State Univ., University Park, PA 16802. 2 US-23 over Sa- 2,500 9.9 m 10 1.5 m W27XI02 no
2Prof. of Civ. Engrg., Dept. of Civ. and Envir. Engrg., Univ. of Mich- line River
igan. Ann Arbor, Ml 48109-2125. 3 1-94 over 5,000 10.5 m 10 1.5 m W24X68 no
Note. Associate Editor: Prof. Dennis R. Mertz. Discussion open until Pierce Road
December I, 1996. To extend the closing date one month, a written re- 4 1-94 over Jack- 5,000 15.8 m 9 1.7 m W36XI50 no
quest must be filed with the ASCE Manager of Journals. The manuscript son Road
for this paper was submitted for review and possible publication on Jan- 5 US-23/M-14 1.500 16.0m 8 1.8 m W30x99 yes
over New
uary 19, 1995. This paper is part of the JolU7llJl of Structural Engi- York Central
neering, Vol. 122, No.7, July, 1996. @ASCE, ISSN 0733-9445/96/0007- Railroad
0726-0733/$4.00 + $.50 per page. Paper No. 9921.

726/ JOURNAL OF STRUCTURAL ENGINEERING / JULY 1996

J. Struct. Eng. 1996.122:726-733.

 TRUCK WEIGHT DATA 4;----------------------,
The results of truck-weight measurements are presented by
Laman (1995). The histograms of GVW are summarized in
Fig. I for the five bridges selected for the present study. The
median values of GVW are similar for all locations, varying
from 150 leN to 210 leN. However, the maximum values of
GVW recorded at the five sites vary from 600 leN to 1,200
leN. The most important factors that influence the distribution
of GVW are traffic density [average daily truck traffic
(ADTI)] and the proximity of stationary-truck weigh stations.
The heaviest trucks were observed on busy interstate high-
ways, with large ADTI, and far away from the stationary-
truck weigh stations.
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/25/15. Copyright ASCE. For personal use only; all rights reserved.

For improved interpretation of the truck-weight results, par-

ticularly for the upper tail of the GVW, the GVW cumulative
distribution functions (CDFs) are plotted on a normal-proba-
bility scale in Fig. 2. The construction and use of normal-
probability paper is explained in probability theory textbooks
[e.g., (Benjamin and Cornell 1970)]. The vertical axis in Fig.
2 is the inverse normal distribution function, <1>-1, where <1> is
the standard normal-distribution function. In Fig. 2, the hori-
zontal axis is the truck GVW. For any GVW = x, there is a
probability, p, that x will be exceeded. In Fig. 2, the vertical
axis is <1>-I(p); therefore, 0 on the vertical scale corresponds
to a probability p = 0.5, 1 corresponds to p = 0.159, 2 cor-
responds to p = 0.0228, -1 corresponds to p = 0.841, and so
on.
The histograms in Fig. 1 and CDFs of GVW in Fig. 2 con- .4 '----'----'_....L.----I._...L...---L.._"---'-_'--....I...---o----J

firm that the truck traffic is site-specific. The variation in- o 200 400 600 800 1000 1200
creases for larger percentiles of GVW. The difference between GROSS VEmCLE WEIGHT, kN
the highest and the lowest median value is approximately 40%, FIG. 2. CDFs of GVW on Normal Probability Paper
but the variation is 100% for the extreme values of GVW.
STRAIN DATA
Strain histories were collected continuously for one-week
periods and were reduced using the rainflow algorithm. Data
were collected for each girder in the bridge. The data are pre-
sented in the form of CDFs and represent strain cycles due to
seven days of normal traffic. For an easier interpretation of
results, the CDFs are plotted on normal-probability paper.
For each bridge, the CDFs are shown for strains in girders
numbered from 1 (exterior, on the right side looking in the
direction of traffic). The number of girders in each bridge var-
ies from 6 to 10, as seen in Table 1. The average strain is less
than 50 X 10- 6 for all girders and all bridges; however, the
largest strains were observed in girders supporting the right
traffic lane (girder numbers G3, G4, and G5) and those nearest
the left wheel of traffic in the right lane. As expected, the
exterior girders of each bridge experience the lowest strain
extremes in the spectrum.
As a means of comparison of fatigue live load, the equiv-
alent stress, Seq, is calculated for each girder using the follow-
ing root-mean-cube (RMC) formula:
(1)

where SI = midpoint of stress interval i; and PI = relative fre-

FIG. 1. Gross Vehicle Weight Histograms: (a) Bridge 1: (b)
Bridge 2: (c) Bridge 3; (d) Bridge 4: (e) Bridge 5
quency of cycle counts for interval i. The stress, SI' is calcu-
lated as a product of strain and modulus of elasticity of steel.
The equivalent stress values are plotted for each bridge in Figs.
3-7.
The CDFs of strain cycles are shown in Fig. 8 for the bridge
carrying US-23 over the Huron River (bridge 1). The distri-
butions of strain cycles and the extreme response vary less for
bridge 1 than for other bridges in the study. The extreme re-
sponse of the exterior girder, Gl, to the live load is 30% lower
than that for girder 3, which is the highest-stressed girder. This
can be attributed to the low number of girders and to the width
JOURNAL OF STRUCTURAL ENGINEERING / JULY 1996/727

J. Struct. Eng. 1996.122:726-733.

 20 20 r---------------,

j 15

f 10

! ~
0
1 2 3 4 4 .5 1 8
GIRDER GIRDER

FIG. 7.
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/25/15. Copyright ASCE. For personal use only; all rights reserved.

FIG. 3. Bridge 1-Equlvalent Streases for Girders 1-6 Bridge 5-Equlvalent Stresses for Girders 1-8
20,----------------, 5

j 15 ~ 4

i 10
i 3

I
~
!
:z

1
0

I
1 2 3 4 5 6 1 9 10
GIRDER
0
FIG. 4. Bridge 2- Equivalent Stresses for Girders 1-10

, 20,..----------------,

15
FIG. 8.
-I
0 50 100 150
STRAIN X 10"'
200 250

Bridge 1-Straln Cycle CDFs for Girders 1-6 (G1-G6)

300

f 10 girder 3 increases. This observation is important for focusing

I
inspection on potential fatigue-prone details and fatigue design
of components near the location of maximum equivalent
5
stresses.
CDFs of girder-strain cycles are shown in Fig. 9 for the
0
bridge carrying US-23 over the Saline River (bridge 2). The
1 2 3 4 5 6 8 9 10
response of each girder to the live load varies considerably for
GIRDER
this bridge. Bridge 2 is wider (13.7 m) than it is long (9.9 m),
FIG. 5. Bridge 3-Equlvalent Stresses for Girders 1-10 which accounts for the wide variation in girder response. The
20,..----------------,
maximum extreme value of strain measured was 205 X 10- 6
in girder 5 and the minimum extreme value of strain was 50
X 10- 6 in girder 2, which is located near the extreme right of
~ 15 the bridge. Equivalent stresses for the girder strains are pre-

i 10
~ cles due to the predominance of truck traffic in the right lane
~ stress decreases as girders are farther from girder 5 and is
0
1 2 3 4 5 6
GIRDER
FIG. 6. Bridge 4-Equlvalent Streases for Girders 1-9
(9.5 m) of the bridge with respect to the span length (23.9 m).
Equivalent stresses for the girder strain distributions are pre-
sented in Fig. 3 as a convenient method of comparison be-
tween girders. Girder 3, which is located directly below the
left wheel of vehicles in the right lane, is the most highly
stressed. This location is subject to both the greatest number
of loading cycles due to the predominance of truck traffic in
the right lane, and the effects of multiple presence. The level
of equivalent stress declines for girders as the distance from
728/ JOURNAL OF STRUCTURAL ENGINEERING / JULY 1996
sented in Fig. 4. Girder 5, located directly below the left wheel
of vehicles in the right lane, is the most highly stressed. This
location is subject to both the greatest number of loading cy-
and the effects of multiple presence. The level of equivalent
generally higher under the right lane. Again, this observation
is important for focusing inspection on fatigue-prone compo-
8 9
nents.
The CDFs of girder strain cycles in the bridge carrying 1-
94 over Pierce Road (bridge 3) are shown in Fig. 10. This
bridge is 13.7 m wide by 10.5 m long, which results in the
wide variation in girder response. The maximum extreme
girder strain measured was 310 X 10- 6 in girder 5 and the
minimum was 100 X 10- 6 in girder 10 located at the extreme
left of the bridge. The resulting unusually high stresses in the
girders were investigated and the design of the bridge was
compared with other bridges having similar span and girder
spacing. In particular, bridge 3 is similar to bridge 2. The spans
are 10.5 m and 9.9 m, respectively, and both structures have
girder spacing of 1.5 m. Despite the similarities in configura-

J. Struct. Eng. 1996.122:726-733.

 tion, the recorded strains are considerably larger in bridge 3. 5

~
Girder sizes of the two bridges are substantially different, with
a wide flange section of W24 X 68 in bridge 3 versus ..
W27 Xl 02 in bridge 2. The major cause of the higher stresses
is likely due to the difference in girder size. Equivalent stresses
for the girder strain cycles are presented in Fig. 5. Girder 5,
located directly below the left wheel of vehicles in the right i 3

!
lane, is the most highly stressed. This location, as in previously 2
discussed bridges, is subject to both the greatest number of
loading cycles due to the predominance of truck traffic in the 1
right lane and effects of multiple presence. The level of equiv-

I
alent stress decreases as girders are farther from girder 5.
The results of measurements taken on the bridge carrying 0
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/25/15. Copyright ASCE. For personal use only; all rights reserved.

1-94 over Jackson Road (bridge 4) are presented in Fig. 11.

For this bridge, the response of each girder to the live load ·1
varies considerably. This bridge is 13.7 m by 15.7 m long. 0 50 100 150 200 250 300
The maximum strain cycle measured was 240 X 10- 6 in girder STRAIN X 10"6
5 and the minimum was 85 X 10- 6 in girder 9, located at the FIG. 11. Bridge 4-Straln Cycle CDFa for Glrdera1-9 (G1-
extreme left of the bridge. Equivalent stresses for the girder G9)
strains are presented in Fig. 6. Girder 5, located nearest the
left wheel of vehicles in the right lane, is the most highly 5....--------------------.,
stressed. This location, as in previously discussed bridges, is
~
subject to both the greatest number of loading cycles due to
the predominance of truck traffic in the right lane and effects 5 ..
I
of multiple presence. The level of equivalent stress decreases
as girders are farther from girder 5 and is generally higher 3
under the right lane.

!
5 2
Z

~ . 1

~
=
3
! 0

.1

~
2 0 50 100 150 200 250 300
Sl'RAIN X 10.6
~ 1 FIG.12. Bridge 5-Straln Cycle CDFa for Glrdera 1-8 (G1-
G8)
~ 0

~
The CDFs of strain cycles for the bridge that carries M-14/
US-23 over the New York Central Railroad (bridge 5) are
-I shown in Fig. 12. The response of each girder to the live load
0 50 100 150 200 250 300
varies considerably, as is the case with other bridges. Bridge
STRAIN X 10"'
5 is 12.8 m wide by 16 m long, which accounts for the vari-
FIG. 9. Bridge 2-Straln Cycle CDFa for Glrdera 1-10 (G1- ation in girder response. The maximum girder strain cycle
G10) measured was 235 X 10-6 in girder 4 and the minimum was
105 X 10- 6 in girder 8, located at the extreme left of the
z 5.-----------------------., bridge. Equivalent stresses for the girder strains are presented
in Fig. 7. Girder 3, located directly below the right wheel of
~ .. vehicles in the right lane, is the most highly stressed. This
location differs from previously discussed bridges, however.

i 3
Girder 4 experiences the highest extreme stress cycle as ob-
served from the CDFs in Fig. 12. The level of equivalent stress
decreases as girders are farther from girders 3 and 4.

i
~
2

1
FATIGUE DAMAGE ANALYSIS USING WIM DATA
Fatigue damage is most commonly estimated using Miner's
rule of linear damage accumulation (Zwememan and Frank
~ 0 1988). Applying the stress-life approach, the level of damage

~
induced by a single stress cycle increases significantly with
higher levels of stress. Therefore, fatigue behavior of bridges
-I is predominantly a function of truck traffic and its parameters
0 50 100 150 200 250 300 350
(Tallin and Petreshock 1990). The histograms of truck GYW
STRAIN X 10.6 obtained in the present study are shown in Fig. 1 for each
FIG. 10. Bridge 3-Straln Cycle CDFa for Girders 1-10 (G1- bridge. Only vehicles with GYW > 65 kN were included in
G10) the graphs.
JOURNAL OF STRUCTURAL ENGINEERING / JULY 1996 / 729

J. Struct. Eng. 1996.122:726-733.

 The AASHTO Guide (1989) suggests that the equivalent
weight (Weq) of trucks be calculated and the GVW distributed
to the three axles of the fatigue truck in the same proportion
as the AASHTO design vehicle. This is expected to result in
a more accurate fatigue model for the particular sight than the
standard AASHTO fatigue truck. The Weq values as well as
the equivalent moments, calculated by the root-mean-cube
method, are presented in Table 2 for the five tested sites.
The percent of truck types (by number of axles) is given in
Table 3. The data indicate that between 40% and 80% of the
truck population is 5-axle vehicles, depending on the bridge
location. Vehicles with 3 and 4 axles are often configured sim-
ilar to 5-axle vehicles, and when included with 5-axle vehicles,
account for between 55% and 95% of the truck population.
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/25/15. Copyright ASCE. For personal use only; all rights reserved.
acquisition equipment over a one week time period are a mea-
sure of the average number of cYGles induced per vehicle. Us-
ing an estimate of ADTT for each bridge site and calculating
the minimum strain level of a 70 kN truck to exclude small
cycles attributable to light trucks, cars, and noise, Fig. 14 was
constructed. The dynamic cycles induced due to truck and
bridge interaction do not appear to be significant, since the
average number of cycles calculated from the static WIM and
the dynamic strain data collection are nearly the same. As a
comparison, the number of average stress cycles per vehicle
specified by AASHTO is also plotted. Because the accumu-
lated fatigue damage is directly related to the number of stress
cycles, it can be seen that the AASHTO criteria are not ade-
quate for spans in the range of 5- 20 m, other than 10m. A

Table 3 shows that between 1.7% and 7.4% of the population
is II-axle vehicles. In states where vehicles with more than
five axles are permitted, this may be of importance. While
these vehicles do not account for a significant proportion of
the truck population, the more important determination is the
extent that these vehicles contribute to the total fatigue dam-
age.
In addition to the magnitude of the stress cycles, the number
of load cycles, n, is an important parameter in the fatigue-life
prediction. Static load cycles have been determined both an-
alytically by counting the stress cycles for different spans,· and
experimentally for the tested bridges. E~ch vehicle types (by
number of axles) in the WIM database was analyzed and av-
erage stress cycles per vehicle type were calculated for spans
of 6-60 m. The resulting average number of load cycles is
shown in Fig. 13 for different vehicle types by axle number.
The results demonstrate that there is a relationship between
the number of axles and average significant stress cycles in-
duced per vehicle for simple spans less than 24 m. Also evi-
dent from Fig. 13 is the decrease in the average number of
stress cycles per vehicle to 1.0 as the span increases. There-
fore, the parameter of stress cycles in relationship to vehicle
type becomes less important as the span of the bridge under
consideration exceeds 24 m, depending on the distribution of
smooth curve has been fit to the data of Fig. 14, relating av-
erage number of cycles per truck, na., and span length.
An analysis was performed using WIM results as a database,
to determine the contribution to the total fatigue damage by
each vehicle type. A program was developed that simulates
truck-traffic flow over an analytical, simple span, bridge model
of 6-60 m spans to calculate the static-stress history of each
~k. The analytical model was calibrated using captured
strain files of weighed trucks enabling accurate distribution of
load to the girders in the model and accurate calculation of
stresses. Following the stress-history calculation, a rainflow
analysis was performed by the program for each vehicle stress-
time history. Damage was then calculated using the stress-life
approach (Bannatine et al. 1990). The damage was accumu-
lated by Miner's rule in a matrix of vehicle types and span
lengths. Results of this damage-accumulation analysis dem-
onstrates that vehicles with 10 and 11 axles dominate the cal-
culation. The corresponding percentages of accumulated fa-
4
11 axle
IOaxl
9 axle

~
vehicle types in the normal traffic flow.
The experimentally measured stress cycles by the strain data

TABLE 2. Equivalent GVW and Moments e3

U

~
Equivalent Moment for Each Span (kN· m) 3
Bridge GVWoq
no. (kN) 6m 9m 18 m 27 m 36 m 60 m
(1 ) (2) (3) (4) (5) (6) (7) (8) ......
r.f.l
1
2
4
3
278
343
341
284
164
177
170
151
289
307
308
259
770
838
863
687
1,361
1,563
1,586
1,269
1,989
2,346
2,362
1,916
3,676
4,434
4,439
3,651
ati
347 174 307 874 1,622 2,417 4,537 r.f.l
5

TABLE 3. Distribution of Truck Types by Number of Axles (>15

e
r.f.l
2 3 axle
kips GVW) and Bridge
~ 2 axle
No.
truck
axles
(1 )
%
1

(2)
2
0/0

(3)
Bridge
3
0/0

(4)
%
4

(5)
5
%
(6)
e
<
2 9.5 9.5 7.7 8.0 8.3
3 11.0 6.0 5.2 6.2 7.7
4 14.5 6.9 5.6 5.7 11.6
5 54.3 63.5 72.6 69.1 61.2
6 3.4 3.6 3.8 2.6 4.6 1 L-_-..I.-_----l._ _~~iiiiI;;a;;...;;;;;;;:::::=:::~_....J
7 2.0 0.8 0.9 0.8 1.3
8 1.0 1.0 1.3 1.4 2.2 OSlO 15 20 2S 30
9 0.0 0.8 0.7 1.2 0.7 SPAN,m
10 1.1 1.3 0.4 0.8 0.5
11 3.2 6.6 1.7 4.1 2.0 FIG. 13. Average CyclesNehlcle Type (by Number of Axles)
versus Span

730/ JOURNAL OF STRUCTURAL ENGINEERING / JULY 1996

J. Struct. Eng. 1996.122:726-733.

 3,--------------------, 100
--CurvePit
o CelaJ1atcd Cycles
x Mcuurcd Cycles
§ 90

~ 80

~
< 70
Bridge 4
1

!
Bridge I

60 Bridge 2

'C Bridge 5
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/25/15. Copyright ASCE. For personal use only; all rights reserved.

Ol--'---'---'-----.J~~~~__L__'___'__'_~_'_~~-'-"'""__' i 50
o
~
10 20 30 40
Bridge 3
SPAN,m
FIG. 14. Calculated and Measured Average Stress Cycles per ~ 40
Vehicle tS
tigue damage caused by 10- and II-axle trucks are shown in
~ 30

Fig. 15 for each bridge site. ~

~
To further analyze the effect of the heaviest vehicles on the 20
accumulation of fatigue damage, a heavy-axle truck vehicle
study was performed to evaluate the axle configuration and
load distribution for vehicles with an axle weight greater than
110 kN. It was observed that the heaviest axle on vehicles of
E 10

extreme GVW is a single axle separate from all other axles. O'--_-l..._--li--_....l..._---I._ _...l-_----l._---l
From a bridge strength and fatigue perspective, this is a more o 10 20 30 40 50 60 70
damaging configuration of load since it concentrates a higher
SPAN,m
percentage of the load at a single location rather than distribute
the load more evenly to an axle-group of axles. The majority FIG. 15. Percent Damage by Vehicle Type versus Span Length
of vehicles recorded with a single axle loaded significantly
more than other axles was 10- and ll-axle vehicles.

FATIGUE-LOAD MODEL
A fatigue-load model has been developed for each site based
on the analysis of gross vehicle weight, damage accumulation,
stress cycles, axle spacing, and moment peaks. For the devel-
opment of a fatigue truck-load model, vehicle characteristics
have been analyzed to determine the statistics of axle spacing, 44.5·102 kN 102 - 129 kN 102 - 129 kN
axle load, and significant stress-cycle locations. Analysis of FIG. 16. Three-Axle Fatigue Truck Axle Spacing and Weight
the distribution of single-axle locations and closely spaced axle Configuration
group locations reveals that vehicles with 3-7 axles are ac-
curately modeled by a 3-axle vehicle, while 10- and II-axle
vehicles can be better modeled by a 4-axle vehicle. Other ve-
hicle types in the distribution are not well represented by either
the 3- or 4-axle model; however, the damage caused by 2-,
8-, and 9-axle vehicles was found to be very low for the study
database. Therefore, for the present study, 2-, 8-, and 9-axle
vehicles may be excluded from consideration in the model } 3.4-4.3 m,
development without a significant loss of accuracy.
44.5 - 98.0 kN 191- 267 kN 165 - 267 kN 89.0 - 267 kN
To determine the appropriate number of axles and axle spac-
ing for the fatigue-load model, locations of axles and axle FIG. 17. Four-Axle Fatigue Truck Axle Spacing and Weight
groups were studied. Axles on 10- and ll-axle vehicles tend Configuration
to be arranged in groups of two, three, and four. Four definite
peaks were observed in the distributions of axle and axle- cific nature of the distribution of vehicle types by axle, a sin-
group locations indicating that a total of four single axles can gle-truck model for fatigue loading is not the most accurate
be used to approximate the 10- and ll-axle vehicles. This approach. For sites with traffic consisting of 2-6 axle trucks,
corresponds well with the analysis results presented in Fig. 13, the fatigue truck is shown in Fig. 16. For locations with sig-
whereby these vehicles induce, on the average, between three nificant fatigue damage caused by trucks with more than 6
and four significant stress cycles per vehicle for short (6 m) axles, as in the state of Michigan, a 4-axle truck, shown in
spans. As an additional verification of appropriate axle config- Fig. 17, establishes an improved fatigue-load model. Damage
uration for the fatigue-load model, the distribution of moment caused by the passage of all trucks with 2-9 axles is equiv-
peaks of each vehicle static-stress time history was studied. alent to the damage caused by an equal number of passes of
As expected, the results are similar to the axle-location study. the 3-axle fatigue truck over the bridge. Likewise, the damage
The proposed fatigue truck model is developed to envelope caused by the passage of 10- and II-axle trucks is equivalent
the damage caused by measured vehicles. Due to the site-spe- to the damage caused by an equal number of passes of the 4-
JOURNAL OF STRUCTURAL ENGINEERING / JULY 1996/731

J. Struct. Eng. 1996.122:726-733.

 5 10" 2 10"

Z Z
4 10"

iU
-<
ts
310"

210"
WIM TIlue
IDEALIZED TRUeD
iU
-<
ts
1 10"

~
BTO (LIlI'J)

Q
1 10"
AASBTO (Gllldo) ~
010 0 010 0
0 10 20 30 40 50 60 70
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/25/15. Copyright ASCE. For personal use only; all rights reserved.

0 10 20 30 40 50 60 70
SPAN,Meten SPAN,Meten
FIG. 18. Bridge 1-Fatlgue Model Comparison FIG. 21. Bridge 4-Fatlgue Model Comparison

1.5 10" 410"

8
I
310-'

i
1.0 10"

210-'
~ ~ WIM
lol 5.0 10"
~ ts MODIlL

~ ~
1 10-'

0.0 10'0 010 0

10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
SPAN, Meters SPAN,Meten
FIG. 19. Bridge 2-Fatlgue Model Comparison FIG. 22. Bridge 5-Fatlgue Model Comparison

110" axle vehicles plus the 3-axle fatigue-truck damage times the
number of all other vehicles. Total fatigue damage using the
z 810" WIM database is calculated by accumulating the fatigue dam-

i
age for each truck.
Any new fatigue-load model must be accurate over a range
610" of bridge spans; therefore, the damage accumulation was cal-
. . - - AASHTO (L1lJI) culated for spans from 6 m through 60 m.· The damage ac-
~ 410" AASBTO (Gllldo) cumulation calculated from the WIM database is plotted in the
ts WIM TIlUeD
figures and is the best available information to use for the
calculation of fatigue damage. Using the WIM fatigue damage
~ 2 10" IDEALIZED TIlUeD
as a basis for comparison, Figs. 18-22 show that the calcu-
lation of fatigue damage is generally overestimated by a factor
0100
between 2-6 for shorter spans and underestimated by a factor
0 10 20 30 40 50 60 70 between 2-4 for longer spans.
SPAN,Meten Observe in Figs. 18-22 that the fatigue-load model induces
FIG. 20. Bridge 3-Fatlgue Model Comparl8on
damage very close to that caused by the simulation of the
actual truck-traffic flow based on the WIM measurements. The
damage due to the developed model is, in most cases, within
axle fatigue truck over the bridge. Combining the two trucks' 95% of the damage caused by the WIM database.
results is an accurate representation of the actual truck popu-
lation at the particular bridge. CONCLUSIONS
As a means of comparing the available models, accumulated
fatigue damage was calculated for four load models, including Currently available fatigue-analysis procedures and avail-
(1) the actual WIM database; (2) idealized fatigue trucks; (3) able fatigue-load models for girder bridges are reviewed. On
the fatigue truck specified in the AASHTO Guide (1989); and the basis of WIM measurements, site-specific load spectra are
(4) the fatigue truck specified in the AASHTO LRFD (1994). presented for five bridge sites. Component stress spectra are
The results of this analysis are presented for each bridge in presented for girders in the form of CDFs and the equivalent
Figs. 18-22. stresses are calculated for comparison.
The fatigue-load model has been compared with the avail- Site-specific statistics required for fatigue analysis based on
able models in Figs. 18-22 as a function of span length for current models are presented. Strains are measured to deter-
the most highly stressed girder using the stress-life analysis mine component-specific load spectra. The WIM data are an-
method and Miner's damage-accumulation rule. Total fatigue alyzed to determine the proportion of truck types in the total
damage using the fatigue-load model is calculated as the 4- population and the fatigue damage attributed to each of these
axle fatigue-truck damage times the number of 10- and 11- types. The WIM database allowed truck-axle location and mo-
732 f JOURNAL OF STRUCTURAL ENGINEERING f JULY 1996

J. Struct. Eng. 1996.122:726-733.

 ment-peak location analysis to aid in the development of a
fatigue-load model. Average cycles per truck as a function of
span are presented for each of the several truck types by num-
ber of axles. Site-specific fatigue-load models are developed
and calibrated against both the WIM data collected at each site
and the strain histories.
The truck-load spectra for bridges are strongly site-specific.
Bridges located on major routes between large industrial met-
ropolitan areas will experience the highest extreme loads.
Routes where vehicles are able to circumvent stationary-weigh
stations will have very high extreme loads. Bridges not on a
major route, that are very near a weigh station, or that are
within a metropolitan area, experience much lower extreme
Downloaded from ascelibrary.org by Indian Inst of Tech - Guwahati on 03/25/15. Copyright ASCE. For personal use only; all rights reserved.
loads.
Live-load-stress spectra are strongly component-specific.
Each component experiences a very different distribution of
strain-cycle ranges. The girder that is nearest the left wheel
track of vehicles traveling in the right lane experiences the
highest stresses in the stress spectra and decreases as a func-
tion of the distance from this location. This information can
be useful to target bridge-inspection efforts to the critical
members.
A vehicle type that dominates the distribution of vehicle
types does not necessarily dominate the fatigue damage of the
particular component. A vehicle type that dominates the dis-
tribution of lane moments will likely dominate the fatigue
analysis. This has been demonstrated in the present study by
the 10- and l1-axle vehicles at each bridge and for several
span lengths. Eleven-axle vehicles dominate the extreme val-
ues of the load spectra. Distributions of lane moment dem-
onstrate that II-axle vehicles, although longer and having
more even-load distribution, produce lane moments that are
much greater than 5-axle vehicles.
The fatigue-load models based on a 3-axle truck may over-
estimate the fatigue damage for bridges with a simple span
shorter than 12-18 m, and underestimate the fatigue damage
for longer spans. The proposed fatigue-load model more ac-
curately predicts the fatigue damage caused by normal truck
traffic passing over a bridge. The model is site-specific and is
characterized by the load spectra of the bridge.
ACKNOWLEDGMENTS
The research presented in the present paper has been partially spon-
sored by the Michigan Department of Transportation, the Great Lakes
Center for Truck Transportation Research, and the University of Michi-
gan, which is gratefully acknowledged. Thanks are also due to Tadeusz
Alberski and Kayleen Seaver for their extensive assistance during the
field investigations conducted for the present study.
The presented results and conclusions are those of the authors and not
necessarily those of the sponsors.
J. Struct. Eng. 1996.122:726-733.
APPENDIX I. BIBLIOGRAPHY
Fisher, J. W., Mertz, D. R., and Zhong, A. (1983). "Steel bridge members
under variable amplitude long life fatigue loading." Nat. Cooperative
Hwy. Res. Program (NCHRP) Rep. No. 267, Transp. Res. Board, Wash-
ington, D.C.
Frank, K. H. (1992). "Using measured stress histories to evaluate the
remaining fatigue life of bridges." 3rd Int. Workshop on Bridge Re-
habilitation, 529.
Moses, F., Schilling, C., and Raju, S. (1988). "Fatigue evaluation pro-
cedures for steel bridges." Nat. Cooperative Hwy. Res. Program
(NCHRP) Final Rep. No. 12-28(3), Transp. Res. Board, Washington,
D.C.
Nowak, A. S., Nassif, H., and Frank, K. H. (1992). "Fatigue load spectra
for steel girder bridge." Transp. Res. Rec. No. 1393, Transp. Res.
Board, Washington, D.C.
Nyman, W. E., and Moses, F. (1985). "Calibration of bridge fatigue de-
sign model." J. Struct. Engrg., ASCE, 111(6), 1251-1266.
APPENDIX II. REFERENCES
Bannatine, J. A., Comer, J. J., and Handrock, J. L. (1990). Fundamentals
of metal fatigue analysis. Prentice-Hall, Inc., Englewood Cliffs, N.J.
Benjamin, J. R., and Cornell, C. A. (1970). Probability, statistics and
decision for civil engineers. McGraw-Hill Book Co., Inc., New York,
N.Y.
Fisher, J. W. (1977). Bridgefatigue guide-design and details, Pub!. No.
T112-11n7, Am. Inst. of Steel Constr., Chicago, Ill.
Gagarine, N., and Albrecht, P. (1992). "Predicting fatigue life of highway
bridges with weigh-in-motion data." 3rd Int. Workshop on Bridge Re-
habilitation, Darmstadt, Germany, June 1992, 679-693.
Guide specifications for fatigue design of steel bridges. (1989). Am. As-
soc. of State Hwy. and Transp. Officials (AASHTO), Washington, D.C.
Khaleel, M. A., and Anditani, R. Y. (1993). "Effect of alternative truck
configurations and weights on the fatigue life of bridges." Transp. Res.
Rec. No. 1393, Transp. Res. Board, Washington, D.C.
Laman, J. A. (1995). "Fatigue load models for girder bridges," PhD
dissertation, Univ. of Michigan at Ann Arbor, Mich.
LRFD bridge design specifications. (1994). Am. Assoc. of State Hwy.
and Transp. Officials (AASHTO), Washington, D.C.
Raju, S. K., Moses, F., and Schilling, C. G. (1990). "Reliability calibra-
tion of fatigue evaluation and design procedures." J. Struct. Engrg.,
ASCE, 116(5), 1356-1369.
Schilling, C. G. (1984). "Stress cycles for fatigue design of steel
bridges." J. Struct. Engrg., ASCE, 110(6), 1222-1234.
Tallin, A. G., and Petreshock, T. (1990). "Modeling fatigue loads for
steel bridges." Transp. Res. Rec. 1275, Washington, D.C.
Zwerneman, F. J., and Frank, K. H. (1988). "Fatigue damage under var-
iable amplitude loads." J. Struct. Engrg., ASCE, 114(1),67-83.
APPENDIX III. NOTATION
The following symbols are used in this paper:
Soq = root mean cube equivalent stress calculated from stress
histogram;
S/ = midpoint of stress interval i;
Pi = relative frequency of cycle counts for interval i; and
Woq = equivalent truck gross vehicle weight calculated from
WIM distribution using root mean cube.
JOURNAL OF STRUCTURAL ENGINEERING / JULY 1996/733