Truck Loading and Fatigue Damage Analysis for Girder

Bridges Based on Weigh-in-Motion Data

Ton-Lo Wang, M.ASCE1; Chunhua Liu, M.ASCE2; Dongzhou Huang, M.ASCE3; and Mohsen Shahawy,
M.ASCE4

Abstract: Based on data collected by weigh-in-motion (WIM) measurements, truck traffic is synthesized by type and loading condition.
Downloaded from ascelibrary.org by Florida Atlantic University on 09/28/13. Copyright ASCE. For personal use only; all rights reserved.

Three-dimensional nonlinear models for the trucks with significant counts are developed from the measured data. Six simply supported
multigirder steel bridges with spans ranging from 10.67 m (35 ft) to 42.67 m (140 ft) are analyzed using the proposed method. Road
surface roughness is generated as transversely correlated random processes using the autoregressive and moving average model. The
dynamic impact factor is taken as the average of 20 simulations of good road roughness. Live-load spectra are obtained by combining
static responses with the calculated impact factors. A case study of the normal traffic from a specific site on the interstate highway I-75
is illustrated. Static loading of the heaviest in each truck type is compared with that of the American Association of State Highway and
Transportation Officials standard design truck HS20-44. Several important trucks causing fatigue damage are found.
DOI: 10.1061/(ASCE)1084-0702(2005)10:1(12)
CE Database subject headings: Trucks; Data processing; Simulation; Dynamic loads; Fatigue; Bridges, steel; Bridges, girder;
Surface roughness.

Introduction measured data, Hwang and Nowak (1991) performed numerical

simulations of truck loading. Laman and Nowak (1996) devel-
Heavy truck traffic affects the service life of highway bridge su- oped three- and four-axle fatigue truck models to represent actual
perstructures. Damage typically occurs in the bridge deck and in trucks with axle number ranging from three to 11. Moreover, the
the main superstructure elements. In the evaluation of fatigue live load model is an important issue in the study of policy and
damage, it is necessary to obtain accurate live load spectra. Truck checking of heavy permit trucks ( Dicleli and Bruneau 1995; Fu
loading on a particular bridge is highly related to axle weight and and Hag-Elsafi 2000). Most of the previous analytical studies
used relatively simplified bridge and/or truck models. To further
configuration. To monitor gross vehicle weight (GVW) of passing
study dynamic impacts of multigirder bridges, Wang et al. (1992)
heavy trucks, stationary weight scales have been established over
and Huang et al. (1993) developed a three-dimensional nonlinear
major highways. However, this conventional scale measurement
truck model for the American Association of State Highway and
has several drawbacks, such as drivers’ awareness (they may
Transportation Officials (AASHTO) standard design truck
avoid it on purpose) and delays to traffic. Recently, weigh-in-
HS20-44 and used the grillage bridge model. Based on these stud-
motion (WIM) has been developed as an extensive measurement ies, a more detailed scientific investigation of impact and loading
device throughout the nation. The use of WIM databases can of normal truck traffic on bridge structures is available.
achieve more accurate truck loading, because it can overcome the The objective of this study is to synthesize truck traffic data
shortcomings inherent in stationary weight scales. Nowak et al. collected through WIM measurements, to establish live-load spec-
(1993) measured truck count per lane and multiple presence fre- tra, and to perform fatigue damage analysis for typical bridges.
quency in-lane and side-by-side on US-23 and I-94. Based on the Three-dimensional nonlinear mathematical models of typical
trucks with significant counts are developed based on the mea-
1
Professor, Dept. of Civil and Environmental Engineering, Florida sured axle weights and configurations. Road surface roughness is
International Univ., Miami, FL 33199. simulated as transversely correlated random processes. The mul-
2
District Two Bridge Structural Engineer, Florida Dept. of tigirder bridges are treated as a grillage beam system. Fatigue
Transportation, 710 NW Lake Jeffrey Rd., Lake City, FL 32055. damage analysis is performed according to the Miner’s linear
3
Senior Research Scientist, Stractural Research Center, Florida Dept.
damage rule. Several important observations are summarized.
of Transportation, 2007 E. Dirac Dr., Tallahassee, FL 32310; and
Professor, Dept. of Civil and Architectural Engineering, Fuzhou Univ.,
350002, Fozhou, China.
4
President, SDR Engineering Inc., 2434 Oakdale St., Tallahassee, FL Idealization of Bridge Structures
32312.
Note. Discussion open until June 1, 2005. Separate discussions must
In order to study the effects of truck traffic, six simply supported
be submitted for individual papers. To extend the closing date by one
month, a written request must be filed with the ASCE Managing Editor.
steel girder highway bridges were designed according to the
The manuscript for this paper was submitted for review and possible AASHTO (1996) Standard Specifications and the Federal High-
publication on July 10, 2000; approved on April 30, 2003. This paper is way Administration (FHWA 1990) Standard Plans for Highway
part of the Journal of Bridge Engineering, Vol. 10, No. 1, January 1, Bridge Superstructures. The span lengths are 10.67, 16.76, 22.86,
2005. ©ASCE, ISSN 1084-0702/2005/1-12–20/$25.00. 30.48, 36.58, and 42.67 m, respectively. The design is based on

12 / JOURNAL OF BRIDGE ENGINEERING © ASCE / JANUARY/FEBRUARY 2005

J. Bridge Eng. 2005.10:12-20.

 Fig. 1. Cross section of bridge with 16.76 m span length

Fig. 2. Truck count versus Federal Highway Administration (FHWA)

HS20-44 loading. The bridges have a roadway width of 8.53 m classification
and a concrete deck thickness of 0.20 m. All the bridges consist of
five identical girders. The five girders are evenly spaced at 2.13 m
processed and used as input data. In this study, multiple presence
Downloaded from ascelibrary.org by Florida Atlantic University on 09/28/13. Copyright ASCE. For personal use only; all rights reserved.

for the first three spans and 2.44 m for the last three spans. Also,
there are diaphragms transversely connecting these girders. The of trucks is neglected because the occurrence is small; for ex-
number of intermediate diaphragm(s) is 1, 2, 2, 3, 4, and 5, re- ample, it was shown to be less than 8% in the research by Nowak
spectively, increasing with span length. Except for the shortest et al. (1993). Fig. 3(a) shows the histogram of GVW for every
span, all the bridges have composite sections. A typical cross passing truck. Because there is a large diversity of truck weights
section of the bridge with a span of 16.76 m (55 ft) is shown in and configurations, one classification method is hereby developed
Fig. 1. to simplify the analysis. According to FHWA classification, there
The six multigirder bridges are treated as a grillage beam sys- are a total of 15 vehicle types ( FDOT 1998). In each type, trucks
tem. Static and dynamic analysis of these bridges is performed are grouped by loading condition—empty or loaded. The dividing
using the finite-element method ( Wang et al. 1992; Huang et al. line for loading conditions is selected by judgment to ensure an
1993). The mass per length and cross section of each girder are acceptable coefficient of variation (COV), which is generally con-
considered to be uniform along the longitudinal axis of the sidered less than 0.3. According to the established criteria, a com-
bridges. Table 1 presents the primary data of these bridges. puter program is written to synthesize the data. At lane 2 of this
The equation of motion of a specific bridge under a moving station, there are a total of 7,453 truck counts in a 1-week period.
vehicle can be written as These trucks are classified into 23 truck categories in the analysis.
The mean value (MV) and standard deviation (SD) of the axle
Mb␦¨ b + Cb␦˙ b + Kb␦b = Fb 共1兲 weights and configurations are calculated in each category. The
MVs are used to obtain average loading effects in the following
where Mb⫽global mass matrix of the bridge structure; static and dynamic impact study. Tables 2 and 3 present the syn-
Kb⫽global stiffness matrix of the bridge structure; Cb⫽global thesized results for empty and loaded truck Type 9. From Tables 2
damping matrix of the bridge structure; ␦b, ␦˙ b, and ␦¨ b⫽global and 3, it can be seen that COVs in the two categories are gener-
nodal displacement, velocity, and acceleration vectors; and ally less than 0.3. The processed histogram of GVW is shown in
Fb⫽global load vector due to the interaction between bridge and Fig. 3(b). By comparing Figs. 3(a and b), it is observed that the
vehicle. One percent of damping ratio is assumed for the first and two histograms are apparently different when the processing
second modes of steel bridges in this study. The consideration of based on truck type and loading condition criteria is taken into
the damping matrix refers to Clough and Penzien (1996). account.
According to the traffic counts in Fig. 2, three types of trucks
are predominant: Types 5, 8, and 9. Mathematical models of the
Three-Dimensional Nonlinear Truck Models three types of trucks are established based on the data of used
nationwide-truck Types H20-44, HS20-44, and 3S2. Masses of
Truck traffic count and axle weight are monitored by WIM at 20 the tractor and trailer are derived according to their static equilib-
stations in the state of Florida. Station #26, located on the inter- rium relationship with the measured axle weights. The three-
state highway I-75, approximately three miles south of I-4, dimensional mathematical models for Types 5, 8(2S1), 9, and 10
Tampa, is found to have heavy truck traffic. The truck count of are illustrated in Fig. 4. Truck Type 10 (developed from Type 3-3)
southbound lane 2 is shown in Fig. 2. Truck traffic at this lane is is of interest because it is the heaviest truck with a single trailer.

Table 1. Properties and Masses of Bridges

Girder
Mass Intermediate diaphragm Diaphragm at ends
Span
length I ⫻ 10−2 J ⫻ 10−2 Exterior Interior I ⫻ 10−2 J ⫻ 10−2 Mass I ⫻ 10−2 J ⫻ 10−2 Mass
(m) 共m4兲 共m4兲 (kN/m) (kN/m) 共m4兲 共m4兲 (kN/m) 共m4兲 共m4兲 (kN/m)
10.67 0.19 2.08 16.44 11.79 1.31 0.004 0.49 0.13 0.004 0.49
16.76 0.99 3.98 17.06 12.40 12.03 1.723 0.62 0.85 1.723 0.49
22.86 1.58 4.11 18.67 14.01 12.03 1.723 0.62 0.85 1.723 0.49
30.48 4.00 4.53 18.98 14.32 40.83 2.789 1.00 7.71 2.789 0.54
36.58 6.74 4.55 19.84 15.18 50.86 3.351 1.12 11.66 3.351 0.67
42.67 10.26 4.56 20.61 15.95 73.47 3.908 1.12 16.50 3.908 0.67
Note: Bridges with a span less than 30.48 m have rolled girders, while the others have welded girders.

JOURNAL OF BRIDGE ENGINEERING © ASCE / JANUARY/FEBRUARY 2005 / 13

J. Bridge Eng. 2005.10:12-20.

 displacement between the ith wheel and bridge, where the super-
script dot of Utzi denotes its differential with respect to time;
zwi⫽vertical displacement of the ith wheel; usri⫽road surface
roughness under the ith wheel (positive upwards); and zbi⫽bridge
vertical displacement under the ith wheel (positive upwards),
which can be determined by the nodal displacement ␦e and the
displacement interpolation function of the element.
In the present study, the fourth-order Runge-Kutta integration
algorithm is employed to solve the nonlinear equations of motion
of a vehicle. The dynamic equations of the bridge are solved by
the modal superposition procedure based on the subspace itera-
tion method.
Downloaded from ascelibrary.org by Florida Atlantic University on 09/28/13. Copyright ASCE. For personal use only; all rights reserved.

Road Surface Roughness

The power spectral density (PSD) function proposed by Dodds

and Robson (1973) for highway surface roughness is adopted in
this study:

S共␾兲 = 再 A共␾/␾0兲−w1 ␾ 艋 ␾0
A共␾/␾0兲−w2 ␾ ⬎ ␾0
共3兲

where S共␾兲 = PSD 共m2 / cycle/ m兲; ␾⫽wave number (cycle/m);

Fig. 3. Histogram of truck gross vehicle weight (GVW) at Station ␾0⫽discontinuity frequency⫽1 / 2␲ (cycle/m); A⫽roughness co-
#26: (a) Preprocessing; and (b) postprocessing efficient 共m3 / cycle兲; and w1 , w2⫽roughness exponent, taken
herein as 2.050 and 1.440, respectively, for the principal road.
In reality, these longitudinal random processes vary in the
These models simplify the trucks into several rigid masses con-
transverse direction (Fenves et al. 1962; Dodds and Robson 1973;
nected by springs and dampers. The total numbers of degrees of
Law et al. 1975; Honda et al.. 1982). However, existing studies
freedom for each model are, respectively, 7, 11, 16, and 18. The
model road surface roughness as either completely independent or
equations of motion of the vehicular systems were derived using
the same transverse profile. To reflect the nature in the transverse
Lagrange’s formulation. For details of derivation refer to Wang
direction, an approach suggested by Samaras et al. (1985) is em-
and Huang (1993).
ployed to simulate road roughness based on a given spatial cor-
relation relationship:
Interaction between Truck and Bridge q q

The interaction force between the ith wheel of a truck and a

Yr = 兺
i=1
BiXr−i − 兺 AiYr−i + B0Xr
i=1
共4兲
bridge is given as the following:
where Yr⫽two-variate discrete time series 共r = 1 , 2 , . . .兲; Ai and
Fibt = KtziUtzi + CtziU̇tzi 共2兲 Bi = 2 ⫻ 2 autoregressive and moving average (ARMA) coefficient
matrices, respectively, and these matrices can be derived from the
where Ktzi⫽tire stiffness of the ith wheel; Ctzi⫽tire damping co- prescribed 共p + 1兲 ⫻ 共p + 1兲 correlation function matrix; and
efficient of the ith wheel; Utzi = zwi − 共−usri兲 − 共−zbi兲, the relative Xr⫽two-variate Gaussian white noise series with mean zero and
satisfying:
Table 2. Processed Axle Weights of Vehicle Type 9 at Station 26
(Unit: kN)
Axle Mean value Standard deviation Coefficient of variation Table 3. Processed Axle Configuration of Vehicle Type 9 at Station 26
(Unit: m)
Empty
1 43.10 5.78 0.13 Axle spacing Mean value Standard deviation Coefficient of variation
2 35.67 9.87 0.28 Empty
3 31.85 8.90 0.28 A-B 4.74 0.98 0.21
4 23.97 8.27 0.35 B-C 1.37 0.10 0.08
5 23.89 8.63 0.36 C-D 9.74 1.20 0.12
D-E 1.38 0.42 0.31
Loaded
1 44.48 5.47 0.12 Loaded
2 65.25 12.37 0.19 A-B 4.50 0.93 0.21
3 61.20 13.66 0.22 B-C 1.37 0.11 0.08
4 62.14 15.26 0.25 C-D 9.76 1.10 0.11
5 61.29 15.92 0.26 D-E 1.42 0.47 0.33

14 / JOURNAL OF BRIDGE ENGINEERING © ASCE / JANUARY/FEBRUARY 2005

J. Bridge Eng. 2005.10:12-20.

Downloaded from ascelibrary.org by Florida Atlantic University on 09/28/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. Three-dimensional models of typical trucks: (a) Side view of Type 5; (b) side view of Type 8 (2S1); (c) side view of Type 9 (3S2); (d)
side view of Type 10; and (e) front view of Types 5 and 8 (2S1)

E关XrXsT兴 = Im␦rs 共5兲 static stresses at midspan and the shears at the entrance end for
the six bridges are listed in Fig. 7. From Fig. 7, it is observed that
where Im = 2 ⫻ 2 identity matrix; and ␦rs⫽Kronecker’s delta. the CDFs of these flexural stresses and shears for the six span
In this study, the two parameters p and q are chosen as 49 and lengths are different. Fig. 8 demonstrates the static moment and
40, respectively. The correlation coefficient a is assumed to be a shear due to moving loaded Type 9 loading.
constant. The coefficient a can be derived from available mea- To investigate the effects of overloaded trucks, the heaviest
sured data. Based on Honda et al.’s (1982) study, the value of GVW in each truck type is searched from the surveyed data at
correlation function coh2共␰ , ␾兲 in the range of ␾ = 0.01 to
1.0 cycle/ m is roughly 0.4. Thus, a can be obtained as a
= coh共␰ , ␾兲 = 0.63. Fig. 5 shows one set of correlated left and right
lines of good road surface roughness 共A = 20.0⫻ 10−6兲 with an
interval of ⌬ = 0.125 m. Fig. 6 indicates the simulated auto- and
cross-correlation functions and the targets. The simulated func-
tions are computed based on a total road length of 900 m. From
Fig. 6, it can be seen that the simulated results are of good accu-
racy. There are a total of 20 sets of good surface roughness gen-
erated in the following study.

Static Analysis Results

The static moments and shears are calculated for each synthesized
truck category. Under the synthesized truck categories and one-
truck loading, the cumulative distribution function (CDF) of the Fig. 5. Simulated left and right profiles

JOURNAL OF BRIDGE ENGINEERING © ASCE / JANUARY/FEBRUARY 2005 / 15

J. Bridge Eng. 2005.10:12-20.

Downloaded from ascelibrary.org by Florida Atlantic University on 09/28/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Simulated auto- and cross-correlation functions and targets:

(a) Autocorrelation functions; and (b) cross-correlation functions

Fig. 8. Static moment and shear due to moving loaded Type 9:

(a) Moment at midspan; and (b) shear at entrance end (note: moment
and shear increase with span length)

station 26, as shown in Fig. 9. It can be seen that the heaviest

GVW (in truck Type 13) is approximately twice that of the
AASHTO standard design truck HS20-44 (0.32 MN or 72 kips).
Comparison of the effects of these heaviest trucks with HS20-44
is shown in Fig. 10. In Fig. 10, the moment and shear of girder 2
are computed. It can be seen that the effects of several heaviest
truck Types, 6, 7, 8, 10, and 13, exceed those caused by HS20-44.
The “overloading” can reach as high as 42%. Because these re-
sults are based on single truck loading, this “overloading” does
not mean that the ultimate strength of the subject girder is vio-
lated. Two of the heaviest truck Types, 11 and 12, produce less
stress than the design truck. This indicates that, in addition to
GVW, the truck loading is closely related to axle configuration.
To further examine local effects, all the axle weights of these

Fig. 7. Cumulative distribution function (CDF) of calculated moment

and shear Fig. 9. Heaviest gross vehicle weight (GVW) in each truck type

16 / JOURNAL OF BRIDGE ENGINEERING © ASCE / JANUARY/FEBRUARY 2005

J. Bridge Eng. 2005.10:12-20.

Downloaded from ascelibrary.org by Florida Atlantic University on 09/28/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 10. Comparison of effects between heaviest trucks and

HS20-44: (a) Moment; and (b) shear (note: columns in each truck
type denote longest to shortest span) Fig. 11. Axle weight of various truck types: (a) Single axles; and
(b) single, tandem, and group axles

Dynamic Impact Analysis

heaviest trucks are shown in Fig. 11(a). It is found that all of them
are less than 0.14 MN (32 kips)—the heavy axle weight of HS20- To study the dynamic effects of these realistic trucks, the varia-
44. Nevertheless, the distance between tandem axles is about 1.2 tion of impact factors, Imp, with span lengths is shown in Fig. 13.
m, which is much less than bridge span length. Fig. 11(b) indi- The truck loading is transversely positioned symmetrically about
cates the weights of the tandem axles and those of HS20-44. It is girder 2, shown in Fig. 1. Truck models include Types 5, 8(2S1),
seen that the tandem axle weights might significantly exceed that 9, 10, and HS20-44. The actual truck models are established
of HS20-44. Such a heavy tandem axle weight may cause severe based on the MVs of measured axle weights. Traveling speed is
local damage in the bridge deck and secondary members. taken as 88 km/h (55 m/h), close to the speed limit of most high-
According to the AASHTO (1996) Standard Specifications, ways. Each impact factor is taken as the mean values of 20 simu-
the live load moment in slab is lations on good road roughness. There are approximately 2,000
simulations carried out in this study. From Fig. 13, it is observed

再
that the impact factors, for two loaded Types 9 and 10 (with
共S + 2兲/32 · PLL for continuous span GVWs of 0.29 and 0.36 MN) as well as HS20-44 (with a GVW of
M LL = 共6兲 0.32 MN), are in accordance with the AASHTO (1996) Standard
PLL · 共X/E兲 for cantilever span
Specification. The Commentary to the AASHTO (1990b) Guide
Specifications reports that an average of 10% of impact is ob-
where PLL⫽one rear wheel load of design truck HS20-44; served in the field measurements. The computed average impact
S⫽girder spacing in feet; X⫽distance in feet from wheel load to factor of loaded Types 9 and 10 is 10%, coinciding with the field
point of support; and E = 0.8X + 3.75 ft, distribution width of the observations. The reason for the impact factors being higher than
wheel load. In slabs continuous over three or more supports, a the specified value by AASHTO (1996) Standard Specifications is
continuity factor of 0.8 is applicable.
To check the effects of the heavy tandem axle weight, a 0.2 m
(8 in.) thick deck is modeled as a continuous plate supported by
five steel girders using GT STRUDL plate finite element. The
deformation of the girders is neglected in this analysis. The 28-
day compressive strength of the concrete is 31 MPa (4.5 ksi). Fig.
12 presents the position of tandem loads of Type 8 at cantilever
span. Each axle weight is 128 kN (28.8 kips). For continuous
span, the load will be moved transversely to achieve the maxi-
mum bending moment in the deck. Table 4 gives the calculated
moments and the specified values. The results indicate that the
moments induced by these heavy axle loads exceed the specified
value by AASHTO (1996) Standard Specifications at girder spac- Fig. 12. Position of tandem axle loads at cantilever span: (a) Front
ings of 1.83 m (6 ft) and 2.44 m (8 ft). view; and (b) elevation view (1 ft⫽0.3048 m)

JOURNAL OF BRIDGE ENGINEERING © ASCE / JANUARY/FEBRUARY 2005 / 17

J. Bridge Eng. 2005.10:12-20.

 Table 4. Comparison of Calculated Moment and Specified Values (kN-m)
Cantilever Continuous
Girder
spacing AASHTO AASHTO
(m) GT STRUDL standard Ratio GT STRUDL standard Ratio
1.83 1.38 1.57 0.61 4.56 4.34 0.95
2.44 4.53 4.57 1.01 5.87 5.42 0.92
3.05 7.17 7.38 1.03 6.60 6.51 0.99
3.66 9.10 9.56 1.05 7.01 7.59 1.08
4.27 10.43 11.31 1.08 7.27 8.68 1.19
Note: AASHTO⫽American Association of State Highway and Transportation Officials; ratio⫽AASHTO/GT STRUDL; and number in bold controls
design.
Downloaded from ascelibrary.org by Florida Atlantic University on 09/28/13. Copyright ASCE. For personal use only; all rights reserved.

that the corresponding trucks have GVWs less than 0.16 MN, values are intentionally selected to cover most of the calculated
which is half of HS20-44. For example, the impact factors for impact factors caused by Types 9 and 10. The purpose is to con-
empty truck Types 5 and 8 (2S1) is very high because they have sider mainly the heavy trucks that cause significant flexural
rather low GVWs of 0.06 and 0.10 MN. This confirms the ten- stresses. This consideration of dynamic impacts will not involve a
dency that a lighter truck weight generally leads to a higher im- significant loss of accuracy in fatigue analysis, because empty
pact factor ( Hwang and Nowak 1991; Huang et al. 1993). trucks cause only a low level of stress ranges. Combining the
dynamic impacts with the aforementioned static results, the CDF
of the dynamic stress ranges of the most highly stressed girder is
Fatigue Damage Accumulation given in Fig. 14. It can be seen that the stress ranges due to the
normal truck traffic are different for the six bridge spans. The
To evaluate fatigue damage caused by the surveyed normal traf- stress range is 6.89–43.43 MPa (1.0–6.3 ksi). According to the
fic, it is necessary to obtain the dynamic stress ranges of a specific AASHTO (1990a) Guide, most of the calculated stress ranges
girder. Based on the previously described impact study, the im- multiplied by the reliability factor (1.35 for redundant members)
pact factors can be approximately taken as 1.15 for loaded trucks are less than the limiting stress range. In this study, the maximum
and 1.20 for empty trucks, as shown in Figs. 13(a and b). The two chance of exceeding the limiting values by passages is only 1.2%.

Fig. 13. Dynamic impact factors: (a) Loaded truck for moment; (b) empty truck for moment; (c) loaded truck for shear; and (d) empty truck for
shear

18 / JOURNAL OF BRIDGE ENGINEERING © ASCE / JANUARY/FEBRUARY 2005

J. Bridge Eng. 2005.10:12-20.

 Fig. 14. Cumulative distribution function (CDF) of calculated
dynamic stress range
Downloaded from ascelibrary.org by Florida Atlantic University on 09/28/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 16. Accumulated fatigue damage due to various categories

Therefore, the truck traffic will not cause severe fatigue problems Conclusions
on the girders.
The fatigue damage analysis is performed based on the Min- Utilizing 3D nonlinear truck models derived from WIM data,
er’s linear damage rule and the stress-life approach. According to transversely correlated road roughness, and the grillage bridge
this rule, the damage in just one stress range cycle is 1 / Ni if Ni model, this study has developed a more detailed analytical meth-
cycles of a specific stress range Si are needed to cause a structural odology. Truck traffic is classified into limited categories accord-
detail to fail. When the number of cycles, ni, at stress range Si is ing to vehicle type and loading condition. This simplification
applied, the damage fraction Di is ni / Ni. Failure is assumed to makes the analysis convenient and without significant loss of ac-
occur when the summation of the damage fraction, Di, equals 1.0. curacy.
Based on the passages in each truck category, the computed dam- Static analysis indicates that truck traffic-induced flexural
age accumulation in one week for the six span lengths is shown in stress at midspan and shear at the entrance end vary with bridge
Fig. 15. The stress range Si corresponding to each category is span length. The gross weight of the heaviest truck can be twice
used to compute its Ni. Each Fig. 15 also illustrates the damage that of the AASHTO standard design truck HS20-44. Several
accumulation for the fatigue truck specified in the AASHTO heaviest trucks types generate more loading on bridge structures
(1990a) Guide. This fatigue truck is implied in AASHTO (1998) than HS20-44. Based on single truck loading, the observed over-
LRFD by using HS20-44 with a 9.14 m (30 ft) spacing between loading can reach as high as 42%. Truck loading does not neces-
two heavy axles and a load factor of 0.75. Based on the surveyed sarily increase with GVW; therefore, it is closely related to axle
trucks, the equivalent GVW is 0.24 MN (54 kips), exactly the configuration. All the axle weights of these heaviest trucks are
same as that of the standard fatigue design truck in the AASHTO found less than the heavy one of HS20-44. However, if tandem
(1990a) Guide. From Fig. 15, it can be seen that the fatigue de- axles spaced at about 1.2 m are considered, the axle weight will
sign truck of the AASHTO (1990a) Guide causes damage close to significantly exceed that of HS20-44. The axle load can cause
that from the surveyed WIM data. In this analysis, Category A of slightly higher transverse bending moment in the deck slab than
the AASHTO (1996) Standard Specifications is used for rolled the specified value by the AASHTO (1996). Standard Specifica-
beams and Category B is used for welded built-up beams. The tions. The overweight may deteriorate bridge decks.
increase in future truck volume is not included. The stress cycles For heavy trucks, the average impact factors based on 20
per truck passage are taken in accordance with AASHTO (1998) simulations are lower than the specified values of the AASHTO
LRFD. To study the role of various trucks, the fatigue damage (1996) Standard Specifications. Also, the total average of the
accumulation is calculated for each truck category. The results for computed impact factors of moment for loaded Types 9 and 10 is
the six bridges are shown in Fig. 16. From Fig. 16, it can be seen 10%, which is in accordance with the Commentary to the
that the loaded truck Types 9, 8(2S2), 7, and 8(3S1) are of the AASHTO (1990b) Guide Specifications. Dynamic impact factors
most significance. These trucks are either four- or five-axle. under light truck loading are higher than those specified. These
light trucks have very low GVWs as compared to HS20-44.
Through the damage accumulation analysis for six bridge span
lengths, it is found that the fatigue design truck of the AASHTO
(1990a) Guide achieves close effects with the actual truck-traffic
flow based on the WIM measurements. The comparison of fatigue
damage accumulation demonstrates that the loaded truck Types 9,
8-1(2S2), 7, and 8(3S1) contribute the most to the fatigue dam-
age.

Acknowledgment

This research project was sponsored by the Florida Department of

Transportation (FDOT) State Project No. BC-379. This support is
Fig. 15. Accumulated fatigue damage acknowledged and greatly appreciated.

JOURNAL OF BRIDGE ENGINEERING © ASCE / JANUARY/FEBRUARY 2005 / 19

J. Bridge Eng. 2005.10:12-20.

 References Florida Department of Transportation (FDOT) (1998) Florida Annual Av-
erage Daily Traffic Rep.. Tallahassee, Fla.
American Association of State Highway and Transportation Officials Fu, G., and Hag-Elsafi, O. (2000). “Vehicular overloads: load model,
(AASHTO) (1990a). Guide for maximum dimensions and weights of bridge safety, and permit checking.” J. Bridge Eng., 5(1), 49–57.
motor vehicles and for the operation of the non-divisible load oversize Honda, H., Kajikawa, Y., and Kobori, T., (1982). “Spectra of road surface
and overweight vehicles, Washington, D.C. roughness on bridges.” J. Struct. Div. ASCE, 108(9), 1956–1966.
American Association of State Highway and Transportation Officials Huang, D. Z., Wang, T. L., and Shahawy, M. (1993). “Impact studies of
(AASHTO). (1990b). Guide specifications for fatigue evaluation of multigirder concrete bridges.” J. Struct. Eng., 119(8), 2387–2402.
existing steel bridges, Washington, D.C. Hwang, E.-S., and Nowak, A. S. (1991). “Simulation of dynamic load for
American Association of State Highway and Transportation Officials bridges.” J. Struct. Eng., 117(5), 1413–1434.
(AASHTO). (1996). Standard specifications for highway bridges, Laman, J. A., and Nowak, A. S. (1996). “Fatigue-load models for girder
16th Ed., Washington, D.C. bridges.” J. Struct. Eng., 122( 7), 726–733.
American Association of State Highway and Transportation Officials Law, D. B., Williamson, H. J., and Hudson, W. R. (1975). “The charac-
(AASHTO). (1998).LRFD bridge design specifications, Washington,
Downloaded from ascelibrary.org by Florida Atlantic University on 09/28/13. Copyright ASCE. For personal use only; all rights reserved.

terization of road roughness on bridge decks and the adjoining pave-

D.C. ment.” Res. Rep. No. FHWA-RD-75-SO409, Texas Highway Depart-
Clough, R. W., and Penzien, J. (1996). Dynamics of structures, 2nd Ed.,
ment. Planning and Research Div., Austin, Tex.
McGraw-Hill, New York.
Nowak, A. S., Nassif, H., and DeFrain, L. (1993). “Effect of truck loads
Dicleli, M., and Bruneau, M. (1995). “Fatigue-based methodology for
on bridges.” J. Transp. Eng., 119( 6), 853–867.
managing impact of heavy-permit trucks on steel highway bridges.” J.
Struct. Eng., 121(11), 1651–1659. Samaras, E., Shinozuka, M., and Tsurui, A. (1985). “ARMA representa-
Dodds, C. J., and Robson, J. D. (1973). “The description of road surface tion of random processes.” J. Eng. Mech., 111(3), 449–461.
roughness.” J. Sound Vib., 31( 2), 175–183. Wang, T. L., and Huang, D. Z., (1993). “Computer modeling analysis in
Federal Highway Administration (FHWA). (1990). Standard plans for bridge evaluation—Phase III.” Final Rep., Project No.FL/DOT/RMC/
highway bridge superstructures, Washington, D.C. 0542(3)-7851, Florida Department. of Transportation, Tallahassee,
Fenves, S. J., Veletsos, A. S., and Siess, C. P. (1962). “Dynamic studies of Fla.
bridges on the AASHTO test road.” Rep. No.71, National Academy of Wang, T. L., Huang, D. Z., and Shahaway, M. (1992). “Dynamic re-
Science, Washington, D.C. sponse of multigirder bridges.” J. Struct. Eng., 118(8), 2222–2238.

20 / JOURNAL OF BRIDGE ENGINEERING © ASCE / JANUARY/FEBRUARY 2005

J. Bridge Eng. 2005.10:12-20.