Pergamon
Compwrsd SrructwesVol. 49, No. 5, pp.837842, 1993
0 1994Elwier ScienceLtd Printed in Great Britain. 004s7949/93 s6.00 + 0.00
FINITE
ELEMENT
ANALYSIS
OF HIGHWAY
SUBJECTED
TO MOVING
LOADS
BRIDGES
Department
M. A. SAADEGIWAZIRI
of Civil Engineering,
New Jersey Institute
of Technology,
Newark,
NJ 07102, U.S.A.
(Received
I
July
1992)
AbstractA proper estimate of the dynamic effect of traffic load on the response of highway bridges is becoming increasingly important. Reduction in the ratio of dead load to total load which makes the effect of the live load more pronounced, and the adoption of ultimatestrength design method by many design codes are among those factors that make an accurate evaluation of the live load very important. This paper illustrates how a general purpose finite element package can be used to consider the dynamic effect of a moving load traversing a highway bridge. This objective is achieved by employing the load arrival time
option in ADINA. It is shown that, if the problem is modelled properly, the results are as good as the exact solution. Note that the exact solution is practically possible only for a simply supported bridge considering only the effect of the fundamental mode. Thus, using finite element a practical research study
can be performed to develop graphical design aides
for more accurate evaluation
of the impact factor.
The versatility and power of tinite element technique will also make it possible to easily investigate the
spatial nature of response to traffic load.
INTRODUCTION
Although 
the behavior 
of highway bridges has been 

the subject of numerous 
investigations in the past, the 

dynamic 
effect of a moving load crossing a bridge 
is 

still accounted for through 
a relationship adopted 
in 

1927 [l]. Literature surveys of works on the dynamic 

effect of 
a traffic 
load 
point to two factors 
as 
the 

reasons 
for the lack 
of any change since late 
1920s: 

(i) need 
for a simple relationship to account 
for 
a 

complex 
phenomenon, 
and (ii) adequacy 
of the 

adopted 
equation 
in light of the fact that 
few (if any) 

failures of highway bridges can be attributed 
directly 

to the dynamic effect of a moving load. 

However, now the U.S.A. is faced with an essential 

and 
urgent 
need 
to 
rehabilitate and maintain 
the 

infrastructure to provide 
the required level of service 

and to cope with growing demand on the highway 

system. Resources 
available to address this 
issue are 

limited and to optimize 
the allocation of these fund implement new 

ing 
it is essential 
to 
technologies. 

Finite element technology 
can play a significant role 

in management of highway bridges as it relates 
to 

maintenance, 
rehabilitation 
as well as replacement 

activities. 
The advancement in this area has resulted 

in 
the ability to rationally 
analyze and evaluate the 

load capacity of structural 
systems with great 
accu 

racy. Saving from accurate load ratings 
of highway 

bridges 
can be hundreds 
For 

example, 
FHWA 
estimates of millions annually. that ‘the additional 
user 

costs associated 
with 
commercial traffic 
detouring 

around loadrestricted bridges can cost from 

$150,000 annually 
on 
the low side, to $200,000 a day 

on 
high 
capacity 
bridges.’ Proper estimate 
of the 

dynamic 
effect of a moving load on the response 
of 
highway bridges is an essential component to accu
rate evaluation of load capacity of existing bridges. 

Furthermore, 
due to recent advancement 
in the devel 

opment 
of high strength materials bridges are becom 

ing more efficient, while the live load carried by these 

bridges 
is increasing. 
Consequently, the ratio 
of the 

dead load to total load is decreasing effect of the live load is becoming and the dynamic especially as it relates to evaluation 

more important, 

of the fatigue life 

of the structure. 

Previous 
analytical 
studies on the 
dynamic 

response 
of highway bridges subjected 
to moving 

load have implemented special mathematical models 

(analytical 
programs) 
developed for this purpose that 

are not 
widely available 
to design engineers 
[24]. 

However, commercial finite element packages with 

many capabilities 
are becoming increasingly 
avail 

able to even small firms. It is the purpose of this 

paper to demonstrate 
the use of the load 
arrival 
time 
option, available in programs such as ADINA
[5],
in modeling a moving load on a bridge. The accu racy of the method along with possible pitfalls are discussed.
BACKGROUND
The current 
design 
of highway bridges 
uses the 

following equation 
to account 
for the dynamic 
effect 

of a moving 
load 
(L 
in feet) 

I = 
5O/(L + 
125) < 0.3. 
_{(}_{1}_{)} 

This equation 
was adopted 
in 1927 by a joint 
com 

mittee of American and AASHTO. 
Railway Engineering Association 
837
838 M.A.
SAADEGHVAZIRI
Obviously, 
this simple equation 
can 
not account 
getting 
heavier. Furthermore, 
most design codes are 

for many parameters 
that influence the vehiclebridge 
adopting 
ultimatestrength 
criteria 
and 
a more 

dynamic interaction. 
Among those parameters 
which 
realistic 
estimate 
of the dynamic 
effect of 
live load 

influence 
the vehiclebridge interaction 
the 
most 
is 
essential 
to 
the 
application 
of 
this 
design 

important 
ones 
are: dynamic characteristics of 
the 
methodology. 
Indeed, 
the Ontario 
Highway 
Bridge 

bridge (frequencies, 
damping, etc.) dynamic charac 
Design 
Code has taken 
the 
first step by relating 
the 

teristics of the vehicle, initial condition 
of the vehicle 
‘Dynamic 
Load 
Allowance’ 
factor 
to the 
natural 

and the bridge, roughness of the bridge deck and 
frequency 
of 
the 
structure 
rather 
than 
the 
span 

irregularities 
in the approach. Extensive analytical 
length [6]. In a report by the ASCE committee on 

as well as experimental 
work conducted 
at the 
Uni 
loads and forces on bridges 
[7] the need for research 

versity of Illinois 
[24] has addressed 
many of these 
in areas 
of bridge 
loading 
is examined. 
As it relates 

parameters 
in details. 
As a result 
of 
this work, 
it 
to impact, 
it 
is stated 
that: 
‘there 
is 
as much 
need 

was determined that for a smoothly rolling load 
to bring 
uniform 
practical 
application 
in a coherent 

crossing a span the speed parameter, 
a, controls 
the 
way 
to 
the 
large 
body 
of 
fundamental knowledge 

dynamic amplification. 
That is, the dynamic deflec 
that 
has 
already 
been 
acquired 
on 
the subject 
of 

tion 
is increased 
over the static value 
by an amount 
bridge impact, 
as there is to further 
expand 
the field 

that 
depends on a. The speed parameter 
is defined as 
of knowledge 
in 
this 
area.’ One 
of the research 

follows: 
problems 
is identified 
as: Preparation 
of Design Aids 

ci = V/ZLf, 
Based on Rational 
Methods 
of Analysis 
for Dynamic 

Live 
Load 
Allowances 
due 
to 
Traffic 
Loading 
on 

where V is the speed 
of the load 
crossing the bridge 
Highway 
Bridges. 
The finite element 
technique 
can 

(ft/sec), L 
is the 
span 
length (ft), 
and f is the natural 
provide 
the means to convert the rational 
method 

frequency 
of the bridge (Hz). 
ologies 
for calculating 
dynamic load increments 
to 

Of course 
other 
parameters such 
as spacing 
of 
usable charts and 
other 
graphical 
design aids. 

axles relative 
to 
the 
span length 
were also identified 

to 
be important. 
The response of threespan con 
FINITE 
ELEMENT MODEL 

tinuous bridges 
was 
also examined. 
Consequently, 

an alternative 
impact factor 
was suggested 
as 
Load 
arrival 
time 
option 

follows: 
The arrival 
time flag in ADINA 
allows 
for acti 

I=O.i5+cc. 
vation of the load 
at a time 
that 
may be different 
than 

the solution 
time [5]. That 
is, with a nonzero 
arrival 

The constant 
term represents the the speed parameter 
initial 
time 
(ARTM) 
the 
loading 
time 
function 
will 
be 

condition 
and 
effect of the accounts for the 
shifted equal 
to 
the arrival 
time specified. Thus, 
the 

dynamic interaction 
between the 
vehicle and 
the 
load 
will be zero for those solution 
time t ,< ARTM, 

bridge. However, 
this proposed formula 
which 
has 
a 
and 
it is active 
for solution 
time 
t 
> ARTM. 
Note 

much greater 
relation 
than eqn (1) 
to 
the observed 
that 
for 
t = ARTM 
the loading 
is not 
active. There 

behavior of highway 
bridges under 
traffic load was 
fore, 
for a solution time starting 
from 
zero and 
an 

not 
adopted 
in 
any 
design code. 
ARTM 
equal 
to 
twice 
of 
the 
solution 
time 
step 

The need 
for change 
is essential 
for the fact that 
(i.e. 2At) 
the corresponding 
time function 
and 
load 

bridges are 
becoming 
more slender 
and trucks 
are 
values are shown 
in 
Fig. 
I. 

I,(?) 

T 
I;““‘ 

1 
7At 
i 
0.0 
1 

At 
= time step 

No. of Steps = 8 

' 
SOLUTION PER100 
^{c} 

Fig. 
1. The arrival time 
option: associated 
time 
function 
and 
resulting 
load values. 
Finite element analysis of highway bridges
839
#
#
of
of
elements
nodes
=
T
IS
varyng
lo
Fig. 2. A typical mesh and corresponding
Thus,
by defining
time functions
and
concentrate
loads with appropriate duration, 7p, and arrival times
one can easily simulate the effect of smoothly moving
truck with multiple axles.
Finite 
element 
mesh 

Although 
one can 
use 3D elements 
to model 
the 

bridge and investigate 
the spatial effect of traffic load, 

in this work the bridge is modelled using beam 

elements. The number 
of elements is controlled by the 

slowest velocity 
of the moving load to be simulated 

and the period of the bridge. 

The length 
of each element, AL, is equal to L/N, 

where 
L 
is 
the 
span 
length 
and 
N 
is 
the 
number 
of 

elements. 
Then, 
the 
time 
to traverse 
one element 

(i.e. to 
go from 
node 
i to 
node 
i + 
l), 
TV, 
is equal 
to 

AL/V, 
where 
V 
is 
the velocity 
of the moving 
load. 

The ratio 
of 7c, to 
I”, period 
of the highest mode with 

significant 
contribution to 
the response 
of the 
sys 

tem, must 
be 
as 
small as possible 
in order to avoid 

unrealistic 
impact 
amplification. 
Furthermore, 
to 

avoid causing 
a random type of loading 
the value 
of 

7p must 
be always equal to 
7,. That 
is, the 
load 
must 

be always acting on the system, although 
it will be at 

different 
locations 
at any 
given 
time. Other 
truck 

speeds, faster than the speed used in evaluation 
of 7,) 

can be modelled 
by applying 
the load to every several 

other node depending on 
the desired 
speed to 
be 

represented. 
These points 
are discussed 
further 
in 
a 

following 
section under modeling 
considerations. 

Therefore, 
the mesh to model 
a single axle vehicle 

(say 40 kips 
weight) traversing 
a bridge 
with 
span 

equal 
to 
960” at 
a velocity 
of 1000 in/set 
(56.8 mph) 

will be as shown 
in Fig. 2. Faster vehicular 
speed such 

as 
2000 in/set can be modelled 
by applying the 
load 

to 
every other 
node 
(i.e. nodes 
1, 3, 
5, 
., 25 
with 

arrival 
times equal to 0, 0.04, 0.08, 
., 0.48 set). 
For 

slower velocity 
of the truck 
either finer mesh can 
be 

used or the load duration can be elongated 
to accom 

modate 
lower velocity. The later option can 
be 

employed 
as long as the ratio of the load duration, 
7,,, 
= 24 
_{I} 
_{3} 
_{I} 
0.06 

25 

give 
different 
0 

0.96 

ARTM corresponding 
to 
V = 1000 in/set 

to system period, 
T, 
is 
low 
enough for accuracy as 

discussed in the following 
section. 

RESULTS 

In the following 
examples effects of surface rough 

ness as 
well 
as 
the initial 
conditions of the bridge 

and the truck are not being considered. Furthermore, 

it is assumed 
that 
the weight of the moving 
load is 

negligible compared 
to the weight 
of the bridge, and 

the magnitude 
of the moving 
load is constant. 

Simplespan 
bridge 

The midspan 
deflection 
and moment 
of 
a simply 

supported 
beam traversed 
by a constant 
force ignor 

ing 
damping 
is determined 
using 
finite element and 

the results 
are compared 
to 
the 
exact solution. The 

exact solution 
for deflection, 
Y as given 
by Walker 

and Veletsos 
[2], is as follows: 

2PL3 
m 

^{Y}^{=} 
^{}^{1} 714EL.=, 
t 
t 

sin 2nm 
sin 2an 2 

T, 
Tb 
sin  , nnt L 

X 
1n2(n2cz’) 
Cl 
n3(n2cx2) 
1 

where P is the magnitude 
of the moving 
load, EI is 

the bridge Young’s 
modulus 
and moment 
of inertia, 

L 
is the 
span 
length, 
c( is the speed parameter, Tb is 

the fundamental 
period 
of the bridge, 
and 
n is the 

mode number. 

To simplify the exact solution 
only the fundamen 

tal 
mode is considered. 
Furthermore, 
the quantity 

2PL3/n4EI 
which 
represents 
the first 
term series 

approximation 
to 
the 
static 
midspan 
deflection is 

replaced by the exact value of midspan 
static 

deflection, 
y,, . 
Thus, 
the 
central deflection 
takes 

the following form 

t 

sin 27ca 
sin 2n L 

Tb 
T, 

Y, = .u2 
= YSI 
[ 
_{l}_{}_{u}_{2} 
a1_a2 
_{I} 
_{8}_{4}_{0} M. A. SAADEGHVAZIRI
Table 1. Dynamic 
amplification 
factors for exact and 
FE solutions 

Speed parameter, 
a 

0.1 
0.2 
0.3 
0.4 
0.5 
0.6 

DAD exact solution 
1.09 
1.08 
1.41 
1.62 
1.73 
1.76 

DAD finite element 
1.09 
1.07 
1.40 
1.60 
1.70 
1.74 

DAM/DAD FE 
0.822 
0.822 
0.822 
0.822 
0.822 
0.822 

DAD = Dynamic amplification factor 
for deflection; 
DAM = Dynamic 
amplification 

factor for moment. 

which 
is maximum 
when 
exactly the same 
as the exact value 
of rt r/l2 = 0.8224 

[2] for all values of speed parameter. 

cos 27ru + 
= cos 2n f 
An interesting 
point 
to consider is the contribution 

b 
b 
of higher modes to the value of the reactions. 
As seen 

For 
different 
values of 
CI the 
exact solution 
is 
from Figs 3 and 4, the midspan deflection and 

compared to 
the finite element 
solution employing 
moment 
are not affected by consideration of higher 

modal 
superposition 
method and 
considering 
only 
modes, however, the time history of the left reaction 

the fundamental mode (note that for typical high 
(Fig. 5) does indicate 
significant contribution 
by the 

way bridges tl = 0.06 to 0.2). The values of dynamic 
higher modes. 
Note 
that 
considering 
single 
mode 

amplification 
factor 
(i.e. ~/JJ,,) for both methods 
are 
response 
there is a sudden 
drop in the reaction 
as the 

given in Table 
1. The finite element solution compares 
load passes the abutment 
and enters 
the span. 
This is 

very well to the exact solution. The ratio of moment 
due to low contribution 
of the first mode (sinusoidal 

amplification 
factor to that for deflection for the finite 
shape) to reaction when 
the load 
is close 
to 
the 

element solution is also given in 
this 
table which is 
supported 
end. 
Considering 
only the 
first mode 
the 
Fig. 3. Center span deflection.
_{f}
_{4}_{@}_{J}_{O}_{t}
_{.}_{4}
/Multi modt
Time,
WC
Fig. 4. Center span moment.
Finite element analysis of highway bridges 
841 

Time, 
set 

Fig. 5. Reaction at entering abutment. 

L 

a 
= 0.8: cu = VC?Lf, 

f, 
= first mode frequency 

Fig. 6. A threespan bridge. 

Table 2. Dynamic amplification factors for deflection (threespan bridge) 

Speed parameter, a 

0.1 
_{0}_{.}_{2} 
0.3 0.4 
0.5 
0.6 

‘Exact’ solution [3] 
1.06 
1.14 
1.30 1.20 
1.74 
2.02 

Finite element solution 
1.07 
1.15 
1.27 1.18 
1.74 
2.02 

maximum reaction 
is equal 
to 41.4 kips. With 
this 
Similar to the 
simplespan bridge 
the finite 

value the ratio of shear amplification 
factor to that 
of 
element results are 
in very 
good 
agreement 
with 

deflection is equal 
to 0.646 which is exactly equal 
to 
the 
‘exact’ solution. 
Some 
points to be considered 

the theoretical value 
of 7~‘/48 given 
by Walker 
and 
for a good balance 
between accuracy and practicality 

Veletsos [2]. For a multimode 
solution 
the maximum 
are discussed in the next section. 

reaction is 46.35 kips which 
is 
12% higher than 
the 

value obtained by single mode solution. Note that 
a 

similar observation has been made by Walker and 
~OD~LLING CONSIDERATIONS 

Veletsos 121. 

The effect of higher 
modes 
is included simply 
by 
Through two examples it was demonstrated 
that 

considering more than 
one mode in the modal super 
general purpose finite element 
packages 
can be used 

position analysis, or 
one can 
simply consider 
the 
to 
model the effect of moving loads traversing 
a 

effect of all modes by performing 
direct time history 
bridge. However, there 
are 
a couple 
of points 
that 

analysis rather than 
modal 
supe~osition. This ihus 
need to be considered 
in developing the finite element 

trates the power of the finite element method and the 
model. 

contribution that it can make 
to 
improvement 
of 
The first point 
to consider 
is the duration of the 

design guidelines. 
loading. This is determined 
by the velocity that 
is 

being represented, 
which in turn is related 
to the 
size 

Multipleswan bridge 
of the elements used. For example, if a simply 

supported bridge of length 960” is to be traversed 
by 

The threespan bridge 
shown 
in 
Fig. 6 was 
also 
a truck with lOOOin/sec velocity and 
the mesh con 

analyzed for different 
values 
of CL The results 
for 
sists of only four elements (i.e. five nodes), 
then 
the 

a 
= 0.8 are given in Table 
2 along 
with the ‘exact’ 
duration of loading 
must be 0.24 sec. However, 
for 

solution evaluated 
by NietoRamirez 
and Veletsos [3] 
most 
cases this will not give accurate 
results due 
to 

using a specialpurpose 
program. 
high ratio of Q”. 
In this example, 
for 
a = 0.2 
the 
_{8}_{4}_{2} M. A. SAADEGHVAZIRI
system period 
would 
be 0.384, which 
gives 
a 0.625 
of moving loads 
traversing 
highway 
bridges. With 

ratio for rp/T. 
The analysis 
of the bridge results 
in 
a 
the 
power 
and 
versatility 
of 
finite 
element, 
an 

deflection amplification 
factor equal 
to 1.4, which is 
analyst can 
easily 
investigate 
the dynamic 
effect of 

significantly 
different 
than 
the exact 
value of 
1.08 as 
multiaxles 
trucks 
coming from 
both 
directions 
at 

shown in Table 1. If the mesh is refined to eight 
different intervals. 
Furthermore, 
using 3D elements, 

elements (i.e. 9 nodes) the load duration will decrease 
the 
spatial 
nature 
of 
response 
can 
be 
also 

to 
0.12 sec. With 
this 
mesh the deflection 
amplifica 
investigated. 

tion factor will be 1.089, which is a very good estimte 

of the actual 
value. Further 
refinements 
of the mesh 

changed the results 
slightly. 
The number 
given 
in 

Table 1 is for 
a mesh 
with 24 elements. 
This is taken 

as the finite element 
solution 
because 
further 
refine 

ment of the mesh 
(up 
to 192 elements) 
did not make 

any 
difference 
in 
the 
value 
of the dynamic 
amplifi 
REFERENCES 

cation factor. 
So, 
the 
results 
indicate 
that 
a mesh 
I. 
AASHTO, 
Standard specifications for bridges. 
Ameri 

must be used such 
that the ratio of to/T 
is less than 
can Association 
of State 
Highway 
and 
Transportation 

0.1, however, 
this 
ratio does 
not 
have 
to 
be 
very 
Officials, 
Washington, 
DC (1989). 

small. 
H. Walker 
and 
A. S. Veletsos, 
Response 
of simple 

is the fact that 
the load 
span highway 
bridges 
to moving 
vehicle. 
SRS 
No. 
272, 

Another point to consider 
Department 
of Civil 
Engineering, 
University 
of Illinois, 

must be always acting 
on the system. That 
is one 
can 
Urbana, 
Illinois, 
Sept 
(1963). 

not reduce the load duration 
rather than refinement 

and 
A. 
S. Veletsos, 
Response 
of three 

of the mesh. For example, for the problem 
discussed, 
span continuous 
highway bridges 
to 
moving 
vehicles. 

one may decide to use tP = 0.12 with a four elements 
Engflg Urbana, Experi. Station Bull. 489, IL (1966). 
University 
of Illinois, 

mesh. Therefore, 
in 
order 
to represent 
the 
same 
S. Veletsos 
and 
T. 
Huang, 
Analysis of 
dynamic 

velocity the arrival 
times are 0.0, 0.24, 0.48, 
., 0.96 
response of 
highway 
bridges. 
J. 
Engng Mech. 
Diu., 

at 
the first, 
second, 
third, 
. 
., and 
last node. 
This 
ASCE %, 
EM5 
(1970). 

gives a DAF 
equal to 0.57, which is grossly in error. 

Engineering, 
Automatic 
dynamic 
incremen 

tal nonlinear 
analysis. Report 
ARD 
871, 
ADINA 

So, the load 
duration 
must 
be always equal 
to 
the 
Engineering, 
December (1984). 

time required to traverse one element. 
Ontario 
Bridge 
Design 
Code, 
2nd 
Edn. 

Ontario Ministry 
of 
Transportation 
and 
Communi 

cation, Downsview, 
Ontario, 
Canada 
(1983). 

CONCLUSIONS 
loading: 
research 
needed. 
Committee 
on Loads 

and Forces 
on 
Bridges 
of the 
Committee 
on 
Bridges 
of 

It is shown 
that 
general 
purpose 
finite 
element 
the Structural 
Division. 
J. 
Strucf. 
Die., AXE 
108, 

packages can 
be used 
to model the dynamic 
effect 
ST5 (1982). 