Bridge Maintenance, Safety, Management, Life-Cycle Sustainability and Innovations –

Yokota & Frangopol (eds)

© 2021 Taylor & Francis Group, London, ISBN 978-0-367-23278-8

Estimation of axle loads by bridge-weigh-in-motion using strain

measurement of transverse stiffeners
E. Yamaguchi, K. Tsuzuki & K. Oda
Kyushu Institute of Technology, Kitakyushu, Japan

ABSTRACT: A method of estimating traffic loads on a highway by the deformation of a bridge is called
bridge-weigh-in-motion (BWIM). The conventional BWIM is based on the strain measurement of main gir-
ders. To obtain supplemental information, strains of transverse stiffeners are also measured. Previously, it has
been shown that the estimation of the truck loads by the strains of transverse stiffeners solely was possible.
The method is competitive from the viewpoint of cost-effectiveness. Involving the integration of time-history
response of strain, the method is called BWIM by Integration Method with Transverse Stiffeners. The present
study tries to modify it so as to estimate the axle loads, concluding that it works fine for the axle load as well.
Yet it is noticed that when two trucks run side by side, accuracy may be deteriorated. The way to improve the
accuracy in such a case is also presented.

1 INTRODUCTION estimation of truck weight is carried out by the strains

of main girders. But those strains are insufficient to
The information on actual traffic loads travelling on conduct BWIM. The supplemental information such
highways is valuable to determine maintenance as truck velocity is required and needs be acquired by
scheme of bridges. Because of this, the techniques that the strains of transverse stiffeners.
weigh trucks in motion without disturbing traffic flow After the project, the possibility of estimating
have attracted many highway engineers and truck loads solely by the strains of transverse stiff-
researchers. Those techniques are called collectively eners was explored (Yamaguchi et al. 2010). It has
weigh-in-motion. been shown that the approach can give a good esti-
One of the weigh-in-motion techniques is based on mation of the truck load, provided that the transverse
the deformation of a bridge and is called bridge-weigh stiffeners are chosen carefully. The method is named
-in-motion (BWIM). The technique was first explored BWIM by Integration Method with Transverse Stiff-
by Moses (1979). Because of its lower cost than con- eners (BWIM-IT). BWIM-IT can lower the cost
ventional pavement scales, much attention has been since the required measurements are much less and
drawn to BWIM in Japanese bridge engineering com- the data manipulation is simpler. The objective of
munity (Matsui & El-Hkim 1989, Ojio et al. 2001, the present study is to modify the method to estimate
Miki et al. 1987, 2001, Ishio et al 2002), as mainten- axle loads. It is noted that the axle loads are import-
ance work is getting very important in recent years in ant information for bridge maintenance, as they are
Japan. closely related to the durability of the floor slab of
The authors have studied BWIM since 2002, too a bridge.
(Yamaguchi et al. 2004). It was related to a project of
investigating actual truck weights on National High-
way 201, Japan. A problem encountered in this pro- 2 ESTIMATION OF AXLE LOADS
ject was that a two-span continuous bridge with skew
was the only bridge available to the project while 2.1 Fundamentals
a short, simply-supported steel bridge with no skew is
ideal for BWIM. Therefore, a careful preliminary field BWIM-IT is based on the technique by Ojio et al.
test with trucks of known weights was conducted, (2001). The fundamental idea is briefly explained.
having shown that even this type of bridge could be When a truck with n axles runs, the strain at the
used for BWIM. The method for this BWIM is based point of interest g(x) can be given generally as
on that employed by Miki et al. (1987, 2001). The follows:

DOI: 10.1201/9780429279119-560
https://doi.org/10.1201/9780429279119-560

4089
 line, the number of axles, the distances between the
X
n
axles and the measured strain are given.
gðxÞ ¼ Wk  f ðx  Lk Þ ð1Þ
k¼1
In the present study, the strain influence line is
obtained by the three-dimensional finite element ana-
lysis of a bridge using MSC Nastran (2014). The
where x = position of the first axle; Ｗk = load due number of axles and the distances between the axles
to the k-th axle; f(x) = function of an influence line; can be determined from the measured strain. Then
and Lk = distance between the first axle and the k-th by using the influence line, the measured strain are
axle. In this approach, the following integration is to be simulated as closely as possible. The influence
carried out: area surrounded by the strain caused by each axle
load thus obtained gives the axle load sought in the
Z∞ Z∞ Z∞ present study.
X
n
A¼ gðxÞdx ¼ Wk  f ðxÞdx ¼ W  f ðxÞdx
k¼1
∞ ∞ ∞
3 VALIDITY
ð2Þ
3.1 Bridge and field test
where W = gross weight of the truck. The above The bridge used in the project of National High-
integration corresponds to the area A surrounded by way 201 is Sasaguri Bridge, which actually con-
the x-axis and g(x) in the x-g(x) graph, which is sists of two parallel bridges, the up-lane bridge
called the influence area in the present study. Equa- and the down-lane bridge. Each bridge has two
tion (2) indicates that the area A is proportional to lanes, and each bridge is for traffic in one spe-
the gross weight of the truck. Hence, if the influence cific direction, up or down. The up-lane bridge is
area Ac due to a truck of known weight Wc is about 70m long while the down-lane bridge is
obtained, the weight of a truck W can be evaluated about 80m long. Otherwise, they are very similar
from the corresponding influence area A as follows: to each other. Both are a two-span continuous
plate-girder steel bridge with five main girders.
A The cross sections and the plans are presented in
W¼  Wc ð3Þ Figures 1 and 2. The solid lines and the dotted
Ac
lines in the direction perpendicular to the bridge
axis in Figure 2 indicate cross girders and cross
Strains are usually obtained as the time-history frames, respectively. Each bridge has a side walk:
response. Therefore, when a truck runs with constant it is located above the G5 girder in the up-lane
velocity V, the influence area A can be computed bridge while it is above the G1 girder in the
alternatively as

Z∞
A¼V rðtÞdt ð4Þ
∞

where r (t) is the time-history response of the

strain.
The truck velocity is obtained by measuring the
strains of two transverse stiffeners in each traffic
lane. To that end, the identification of the strain
responses due to the same truck is required. This is
done by using the auto-correlation function. More
details are available in Yamaguchi et al. (2004).

2.2 Estimation of axle loads

As Equation (1) indicates, with the strain influence
line, the number of axles in a truck, the axle loads
and the distances between the axles, the strain
caused by the running truck can be computed. The
estimation of axle loads is its inverse problem: the
axle loads are obtained, while the strain influence Figure 1. Cross section of Sasaguri bridge.

4090
Figure 2. Plan of Sasaguri bridge.

down-lane bridge. The bridges cross a river and

the pier is inclined severely to the bridge axis,
making the bridges quite skew. Figure 3. Strain influence line at D3A.
While the axes of the girders are composed of line
segments, the floor slabs and thus the traffic lanes
are curved, the distance between a running truck and
the girders vary along the bridge axis.
At the initial stage of the project, the accuracy of
BWIM was the major concern, since Sasaguri Bridge
is by no means ideal for BWIM. Therefore, prelimin-
ary field test was carefully conducted, in which the
strains of transverse stiffeners were measured for
obtaining supplemental information (Yamaguchi 2004,
2010). Three trucks of known weights were used in
this test. The strains of the transverse stiffeners
recorded in the test are utilized in the present study.
The circles in Figure 2 show the locations of the
transverse stiffeners whose strains were measured in
the preliminary field test. There are two measured
transverse stiffeners in each lane so as to evaluate
the truck velocity. The strains measured at U1A,
U4B, D3A and D5A are made use of for the axle-
load estimation. Therefore, the strain influence line
at each of these four transverse stiffeners is obtained
by the three-dimensional finite element analysis of
Sasaguri Bridge with the single axle load, travelling
on the lane above the transverse stiffener. Nastran
(MSC 2014) is used for the analysis. As an example,
Figure 3 shows the strain influence line at D3A.
The preliminary field test was conducted in May,
2002. Using three calibration trucks, the following
four patterns of traffic situation were considered in
the preliminary field test:
Pattern 1: Only one truck runs
Pattern 2: Two trucks run in the same lane
Pattern 3: Two trucks run side by side
Pattern 4: Two trucks run in the same lane while
the other runs in the other lane
Figure 4 Illustrates the four patterns. Within
each pattern, various combinations of trucks were
considered and the same truck-running test was
conducted three times since variation in the Figure 4. Traffic situations in field test.

4091
measurement is expected. The number of the 4 IMPROVEMENT
truck-running tests amounted to 60. It is noted
that in an effort to enhance the reliability of the The discussion in the previous section indicates the
field test data, traffic was controlled so that no necessity of including the influence of the truck load
public vehicles would run on the bridge during on the adjacent lane so as to improve the accuracy of
each truck-running test. axle-load estimation for Patterns 3 and 4 of the down-
For the estimation of truck weight, AC in Equation lane bridge.
(3) needs to be evaluated first. The measurements in Including the influence of a truck load on the
Pattern 1 are used, to this end. Since multiple truck- adjacent lane, the strains in Lanes 1 and 2, two lanes
running tests were conducted within Pattern 1 and dif- adjacent to each other, are to be evaluated as
ferent tests yield different values of AC, the medium is follows:
taken for the value of AC to be used in the present
study. It is however noted that the variation of AC is
quite small. X
N1 X
N2
ε1 ðtÞ ¼ f11 ðxn1 ðtÞÞWn1 þ f12 ðxn2 ðtÞÞWn2 ð5Þ
n¼1 n¼1
3.2 Axle loads
X
N1 X
N2
The results of the axle loads estimated by the present ε2 ðtÞ ¼ f21 ðxn1 ðtÞÞWn1 þ f22 ðxn2 ðtÞÞWn2 ð6Þ
method are shown in Table 1 in terms of percentage n¼1 n¼1
error. The method yields satisfactory accuracy in the
case of all the patterns of the up-lane bridge and Pat-
terns 1 and 2 of the down-lane bridge: even the worst where Np is the number of axles of a truck in
maximum error is only 10.8%. However, the error is Lane p and fpq is the influence line of the strain, i.e.
much greater in the case of Patterns 3 and 4 of the the strain of the transverse stiffener, in Lane p due to
down-lane bridge, the maximum error being up the axle load in Lane q. Wnq is the nth axle load in
to 88.9%. Lane q. The 2nd term on the r.h.s. of Equation (5)
By closely observing the strain measurements, the and the 1st term on the r.h.s. of Equation (6) are the
reason for the low accuracy in the down-lane bridge influences of the track running in the adjacent lane.
has been identified as the influence of the adjacent The influence line of the strain of the transverse stiff-
truck on the strains of transverse stiffeners. For ener due to the truck in the adjacent lane is obtained
instance, when a truck runs in the cruising lane of the also by the three-dimensional finite element analysis
up-lane bridge, the strain at U4B is much larger than of the bridge.
that at U1A. On the other hand, when a truck runs in The axle loads that give strains in good agreement
the cruising lane of the down-lane bridge, transverse with the measured strains would be the estimated
stiffeners in both of the cruising and passing lanes axle loads in the present study. To be specific, the
deform: the strain at D5A is not negligible in compari- axle loads that minimize the following two quan-
son with that at D3A. The influence of a truck running tities, the differences between the measured strain
on the adjacent lane is not expected in the present and the strain computed by Equations (5) and (6),
method and is considered a source of the error in the would be the present axle loads:
down-lane bridge.
X
M
I1 ¼ ½ε1  ðtm Þ  ε1 ðtm Þ2 ð7Þ
m¼1

Table 1. Average and maximum percentage errors in esti- X

M
mation of axle loads. I2 ¼ ½ε2  ðtm Þ  ε2 ðtm Þ2 ð8Þ
m¼1
Bridge Pattern 1st axle 2nd axle 3rd axle 4th axle

Up- 1 4.3 (6.3) 4.1 (8.6) 3.0 (5.7) 1.4 (3.1) where εi* is the measured strain, M is the number
lane of strain measurements and tm is the time of the
2 4.4 (7.0) 4.3 (9.0) 3.6 (9.4) 1.2 (3.3)
measurement.
3 2.8 (7.4) 4.7 (7.8) 3.8 (6.4) 1.7 (4.1)
By this modified method, the axle loads of
4 2.3 (7.2) 3.9 (6.4) 3.5 (6.2) 1.8 (4.7)
Patterns 3 and 4 of the down-lane bridge are esti-
Down- 1 4.1 (7.3) 6.1 (8.8) 2.9 (4.9) 2.0 (3.7) mated. The errors are summarized in Table 2.
lane 2 4.2 (9.6) 6.4 (10.8) 2.6 (4.9) 2.9 (6.5) The accuracy is improved significantly and is
3 44.2 (79.7) 58.5 (88.9) 49.2 (70.9) 51.1 (79.5) now considered acceptable for practical purposes.
4 32.5 (58.6) 38.5 (79.6) 30.9 (69.3) 34.1 (70.2) Once the axle loads are obtained, the truck
weight is computed simply by summing up the
(): maximum error axle loads.

4092
Table 2. Average and maximum percentage errors in From the viewpoint of this background, the
improved estimation of axle loads. statistics are taken for the truck heavier than
245kN currently. Preliminary results are presented
Bridge Pattern 1st axle 2nd axle 3rd axle 4th axle in Figures 5 and 6. They are based on the meas-
urements for 229 days. Indeed, a large number of
Down- 3 6.9 (15.0) 3.7 (12.1) 6.0 (13.0) 4.4 (12.8) truck beyond the legal limit were running. Many
lane 4 8.9 (18.1) 5.8 (17.8) 5.2 (10.8) 5.4 (10.3)

(): maximum error

5 CONCLUDING REMARKS

The economical method to estimate axle loads were

presented. A good result is obtained for one bridge,
while large error is observed for the other bridge when
two trucks run side by side. The reason behind this has
been found the influence of a truck in the adjacent lane.
The proposed method was then modified. Its val-
idity has been demonstrated, confirming that an
acceptable accuracy was achieved for the estimation
of axle loads.
245kN and 98kN are the legal limits for the truck
weight and the axle load on this highway, respect-
ively. The truck that exceeds the limit must obtain
permission to run. In reality, however, quite a few
trucks are said to run without the permission; no offi-
cial statistics are available, though.

Figure 5. Distribution of truck weights heavier than

245kN. Figure 6. Distribution of axle loads.

4093
of them may not have permissions since the of steel plate girder bridge. Journal of Structural Engin-
number of issued permissions is much less than eering 47(A): 1083–1092.
the number of those trucks. Matsui, S. and El-Hkim, A. 1989. Estimation of axle loads
The figures show that heavy trucks run more of vehicles by crack opening of RC slab. Journal of
often on the up-lane bridge than on the down-lane Structural Engineering 35(A): 407–418.
bridge, and the cruising lane is under much severer Miki, C., Murakoshi, J., Yoneda, T. & Yosimura, H. 1987.
traffic-load condition than the passing lane. It is Weighing trucks in motion. Bridge and Foundation 87
(4): 41–45.
therefore concluded that the cruising lane of the up- Miki, C., Mizunoue, T. & Kobayashi, Y. 2001. Monitoring
lane bridge requires most attention in maintenance system of bridge performance with fiber-optic
work. communications. Journal of Construction Management
This class of statistics would be thus helpful for and Engineering 686/VI-52: 31–40.
constructing a good maintenance scheme, to which Moses, F. 1979. Weigh-in-motion system using instru-
the present research is believed to contribute. mented bridges. Transportation Engineering, Journal of
ASCE 105(TE3): 233–249.
MSC. Nastran 2014.
REFERENCES Yamaguchi, E., Matsuo, K., Kawamura, S., Kobayashi, Y.,
Mori, M., Momota, K. & Nishinohara, T. 2004. Accur-
Ishio, M., Nakamura, S., Tamakoshi, T. and Nakasu, K. acy of BWIM using two-span continuous steel girder
2002. About the WIM system using the influence line. bridge. Journal of Applied Mechanics 7: 1135–1140.
Proc. of 57th Annual Conference of The Japan Society Yamaguchi, E., Naitou, Y., Matsuo, K., Matsui, Y.,
of Civil Engineers, JSCE, 1447–1448. Takaki, Y. & Kawamura, S. 2010. BWIM by transverse
Ojio, T., Yamada, K. and Kobayashi, N. 2001. Develop- stiffeners of steel I-girder bridge. Journal of Civil Engin-
ment of bridge weigh-in-motion system using stringers eering, Division F 66 (2), JSCE: 251–260.

4094