Content uploaded by Franziska Schmidt
Author content
All content in this area was uploaded by Franziska Schmidt on Oct 05, 2015
Content may be subject to copyright.
WIM data to assess consequences of new traffic
regulations
Franziska Schmidt, Bernard Jacob
To cite this version:
Franziska Schmidt, Bernard Jacob. WIM data to assess consequences of new traffic regulations.
The Fifth International Conference on Bridge Maintenance, Safety and Management, IABMAS
2010, Jul 2010, France. 6p, 2010. <hal-00949085>
HAL Id: hal-00949085
https://hal.archives-ouvertes.fr/hal-00949085
Submitted on 19 Feb 2014
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-
entific research documents, whether they are pub-
lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destin´ee au d´epˆot et `a la diffusion de documents
scientifiques de niveau recherche, publi´es ou non,
´emanant des ´etablissements d’enseignement et de
recherche fran¸cais ou ´etrangers, des laboratoires
publics ou priv´es.
WIM data to assess consequences of new traffic regulations
F. Schmidt & B. Jacob
Laboratoire Central des Ponts et Chaussées, France
ABSTRACT: Bridge code calibration and assessment of existing bridges require an accurate knowledge of
traffic load patterns. WIM data are useful for these applications, either to account for the current traffic loads
or to assess the potential impact of future loads. Traffic load effects were calculated on two composite
bridges, using WIM data of a French motorway, and the effect of an increase of the gross weight limit from
40 to 44 tons was assessed by a micro-simulation. Extreme values and fatigue lifetimes are calculated and
compared for the current traffic and a traffic with the increased weight limit. The lifetime of details sensitive
in fatigue could be decreased by 20% if the weight regulation is changed.
1 INTRODUCTION
Bridge load effects, stresses or damage and lifetime
in fatigue are assessed either by measurements, us-
ing stain gauges, extensometers and other tech-
niques, or by calculation, using traffic load data and
bridge models such as influence line or surfaces, if
the dynamic interactions are neglected (Bailey &
Bez 1996, Nassif & Nowak 1996, Jacob 1998, Jacob
& Labry 2002). The development and implementa-
tion of WIM techniques allowed to collect traffic
load data on a large scale and to use them for bridge
assessment (Jacob et al. 2002), instead of using
simulated data (Hwang & Nowak 1991).
In the late 1980’s and early 1990’s, some exten-
sive works were developed in Europe using traffic
data of various EU member states (Vrouwenvelder
& Waarts 1993, Bruls et al. 1996, O’Connor et al.
1998) to design the Eurocode “Traffic loads on
bridges” (CEN EN1991-2, 2003). Extreme loads and
load effects were considered to calibrate the general
load models (Flint & Jacob 1996) and fatigue calcu-
lation were performed to design the fatigue load
models (Jacob & Kretz 1996).
All these models and assessment methods assume
that the traffic loads and models (Ditlevsen 1994)
are stationary. But the traffic loads and volume in-
crease with time because of the development of the
road freight transport, and the increases of the truck
legal weight limits. Thus it is necessary to periodi-
cally re-calibrate the bridge code load models and to
re-assess the reliability of existing bridges against
the traffic loads. That is one of the goal of the WIM
systems installed all around the world on highways
and motorways. The methodology and some tools
are briefly reported here, and an application to the
assessment of possible consequences of an increase
of the truck gross weight limit in France from 40 to
44 tons is presented.
2 WIM DATA
Traffic data and loads are collected in France for
bridge engineering by WIM systems since the early
1980’s. However, with the recent development of a
national WIM network for overload screening and
enforcement (Marchadour & Jacob 2008), more reli-
able and accurate data are collected continuously on
the main highways and motorways. The traffic data
used in this study were collected on two adjacent
traffic lanes of the French motorway A9 near Mont-
pellier in South France, over a month in June 2009.
Figure 1 shows the gross vehicle weight distribu-
tions on the slow lane over one and two weeks.
Figure 1. Gross weight distributions on the slow lane of the A9
motorway, 1 & 2 weeks in June 2009.
Axle loads and spacing, speed and time of pas-
sage in 1/100 s are also recorded for each truck in
both lanes. The accuracy of these data are at least in
class C(15), but mostly in class B(10) of the Euro-
pean specifications of WIM (Jacob et al. 2002), i.e.
app. 95% of the gross weights are within ±10% and
of axle loads within ±15% of the static values taken
as the reference.
3 CALCULATION TOOLS
3.1 Influence lines
The effects of traffic on a bridge are usually calcu-
lated by applying traffic loads, either measured by
WIM systems or simulated by software, on influence
lines or surfaces. For local effects sensitive to the
wheel transverse location, WIM data shall contain
this information and an accurate influence surface,
i.e. a two-dimensional calculation, must be used. For
most of the global or semi-global and longitudinal
effects, such as bending moments at mid-span or on
pier, shear forces or pier reactions, one or a set of
parallel longitudinal influence lines (one per traffic
lane) are sufficient, and the traffic data lane by lane
are used. That is the case in this study.
The influence lines or surfaces are obtained nu-
merically, through 1-D bridge model or 2-D finite
elements calculations, or on site through experimen-
tation (by strain measurement on a bridge under a
known truck). In this later case, several parameters
are taken into account, such as the stiffness of the
pavement surface, sidewalks and safety barriers, that
are neglected in the numerical method.
The effects considered in this study are, for two
bridges:
− The bending moment stress at mid-span of a sim-
ple supported 40 m span bridge, near Auxerre.
This is a composite bridge with two main steel
girders and a concrete deck.
− The bending moment stress at mid-spans and on
piers of the Libourne 4-span bridge (respectively
48, 60, 60 and 48 meters-long). This is also a 2-
steel girder composite bridge.
The influence lines are given in MPa/MN in Fig-
ures 2 and 3. The dominant effects are in the lower
flange at mid-span and in the upper flange on pier.
These influence lines are those calculated for a load
applied along a steel girder. To obtain the influence
surface, they have to be combined with the trans-
verse influence line, given by Courbon’s coefficient.
Figure 2. Influence line of the bending moment stress at mid-
span of the lower flange of a girder, Auxerre composite bridge.
Figure 3. Bending moment stress influence lines at mid-span
and on piers of the upper and lower flanges, Libourne bridge.
3.2 Traffic data adaptation
The WIM data presented in section 2 contain two
adjacent lanes in the same direction, i.e. one slow
lane and one fast (passing) lane. Auxerre bridge
supports 3 traffic lanes (each 3.50m in width), in-
cluding 2 slow lanes (one in each direction) and one
fast passing lane. Libourne bridge has 2 slow lanes
(one in each direction).
Thus, the recorded data are adapted to fit to the
bridge operation conditions. The natural data of the
slow lane are applied on one slow lane of each
bridge, and the data of the fast lane on the fast lane
of Auxerre. The slow lanes in the other direction are
filled with the slow lane data shifted by 10 minutes,
which simulate a traffic correlated (same peak and
low hours) but with independent trucks.
3.3 CASTOR and POLLUX softwares
The POLLUX software, developed by the LCPC, is
a new and extended version of the former CASTOR
software (Eymard & Jacob 1989). It uses WIM data
and load effect or stress influence lines of any
bridge, and calculates the effects of the whole traffic
on this bridge (Figure 4). The traffic record may be
on one lane or on multiple lanes, and the software
may use one influence line per lane, or an influence
surface for the whole deck.
Figure 4. Software POLLUX: main screen and parameters.
Different modules of POLLUX perform successive
tasks:
− Traffic data cleaner: the traffic data files, pro-
vided by WIM systems, are first checked and
cleaned if necessary. If some anomalies are de-
tected, the involved vehicles are deleted. For ex-
ample, vehicles with a length below 1.5 m or over
25 m, or with only one axle or more than 8 axles,
are eliminated. If the gap between two successive
vehicles is greater than 90 minutes, the system is
assumed to have failed and the time period is de-
leted. All these limits are defined by the user.
− CASTOR, which became a sub-routine of POL-
LUX, calculates the effects of the traffic loads.
For each effect (set of influence lines for each
traffic lane), the time history of the effect is cal-
culated with a constant time step, chosen by the
user with respect to the influence line shape. All
the vehicles are assumed to be at a constant speed
of 20 m/s. Thus for a 100 m bridge length, to
sample the influence line by steps of 1 m, the
time step will be 0.05 s. The output of CASTOR
is a set of three histograms, which represent the
variations of the effects: (i) the mini-max histo-
gram (successive local minima and maxima), (ii)
the level crossing histogram, and (iii) the rain-
flow histogram (AFNOR 1993), only used for fa-
tigue assessment. The class width of these histo-
grams are chosen by the user.
− The sub-routine “Extrap” extrapolates the traffic
load effects over any time period longer than the
traffic record period, assuming that the traffic is
stationary, as explained in section 4.1.
− The sub-routine “Durvie” calculates the damage
and lifetime in fatigue, for a detail submitted to
the stress history resulting of the traffic loads ap-
plied to a stress influence line, as explained in
section 4.2.
4 EXTREME EFFECTS AND FATIGUE
This section describes the methods used by CAS-
TOR and POLLUX to extrapolate the extreme ef-
fects and to calculate bridge lifetime in fatigue.
4.1 Extreme extrapolated effects
The extreme load effects or stresses over the whole
planned or conventional lifetime of a bridge, are ne-
cessary to calibrate the bridge code load models or
to assess the reliability of a bridge submitted to traf-
fic loads. Depending on the considered limit state -
ultimate or serviceability limit state - and the human
or economical consequences of the bridge or part of
a bridge failure, and on the status of the traffic ac-
tion – main action or secondary action in a combina-
tion of actions -, an adapted return period is chosen.
The return period Rx of x for a random variable X,
which is assumed to represent successive independ-
ent occurrences of a random phenomenon, is the
mean time period between two occurrences of x. It
may be proved that if P[Max1≤i≤N(Xi)<x
α
]=
α
,
αα
α
NN
Rx≈
−
−
≈
)1ln( (1)
if N→ +∞, and then if N→ +∞ and
α
→ 0.
The return period for the traffic load characteris-
tic value is 2000 years in the Eurocode, which corre-
sponds to a probability
α
=5% to be exceeded in a
lifetime N=100 years.
The traffic load effect extrapolation method was
initially proposed by (Jacob et al. 1989). It consists
in fitting the upper part of a Gaussian distribution to
the upper tail of the level crossing histogram and
then using the Rice formula (Rice 1954). The traffic
load effect is assumed to be stationary and Gaussian
(Ditlevsen 1994), and thus the mean rate ν(x) of up-
crossings the level x>0 is given by:.
(Rice formula) (2)
where , and are respectively the standard var-
iation of , and the mean of the process. Then it
may be shown that:
(3)
The user chooses a return period and a confidence
level for the Kolmogorov test applied to check the
fit of the Gaussian distribution on the upper tail of
the level crossing histogram (Cremona 2001). Then
the value x with the chosen return period Rt is given
by Equation 3.
4.2 Fatigue
The fatigue damage D of a detail is assessed using
Miner’s rule:
(4)
where is the number of cycles in the class of
the stress rain-flow histogram; is the number of
cycles at failure corresponding to in the S-N
(Wöhler) two slope curve defined by:
(5)
in the Eurocode 1993 Part 1-9 (CEN EN 1993-1-9,
2005), corresponds to cycles, so
. The conventional lifetime (in years)
of the detail is 1/TD, where T is the length of the
time period which corresponds to the rain-flow his-
togram.
5 COMPARISON OF TRAFFIC LOAD EFFECTS
DEPENDING ON THE WEIGHT LIMIT
5.1 Weight limits in the current French legislation
and future plans
In France, and in Europe for international transport,
the standard gross vehicle weight limit is 40 tons.
Some countries allow higher weights, such as 44
tons in the UK for 6-axle articulated trucks. In
France, there are a couple of derogations which al-
low 44 tons for 5-axle articulated trucks if: (i) 40 ft
containers are transported in a multi-modal journey,
(ii) the freight comes from or goes to a maritime or
inland harbor, where it is carried by ship, (iii) some
specified agricultural goods are transported during
specific seasons. Moreover, log trucks are allowed
up to 48 tons on 5 axles and 57 tons on 6 axles by
decree.
In 2008, the French parliament voted a law called
“Grenelle de l’environnement” to reduce the CO2
emissions and the fossil energy consumption. An ar-
ticle stated that road transport regulation shall be
adapted for that. In February 2009, the parliament
asked to the government to study the positive and
negative impacts of increasing the gross weight limit
to 44 tons. The main criteria to consider were road
safety, CO2 emission and energy consumption, traf-
fic congestion, and infrastructure (bridges and
pavement...) lifetime and maintenance.
5.2 Impact of 44 t trucks on traffic and bridges
For bridges, the first task was to compare the ef-
fects of a single 44 t truck to those of a 40 t truck.
Because the axle load limit would not be increased,
only medium span (15 to 40 m) bridges could be af-
fected, or 10 m continuous span bridge for bending
moment on pier. Shorter spans could not support a
whole truck at once. Longer spans are not so much
sensitive to a single truck. For medium span bridges,
the load effects would be increased by less than
10%, which is not too much critical for the extreme
loads and load effects on healthy bridges. However,
the impact of multiple presences (truck crossing or
overtaking) of 44 t trucks, or the effect on the fa-
tigue lifetime was to be investigated more in depth.
Therefore, it is necessary to make some assump-
tion on how the whole traffic would change with
new regulations and to assess the effects of this
whole new traffic. The very simple and crude as-
sumptions done here are: (i) the gross weight of
every heavy vehicle laying between 36 and 44 tons
would be increased by 10%, the additional weight
being uniformly distributed on all axes; (ii) all the
other vehicles would have the same load as before
(volume limitation or not enough freight to carry).
With these assumptions, a micro-simulation gen-
erated a “modified” traffic file named “44 t” derived
from the natural traffic record of the WIM system on
the A9 motorway, named “40 t”. Both traffics were
used with the influence lines introduced in the sec-
tion 3.1, to obtain the effects and their histograms.
5.3 Extreme load effects
The minimum and maximum bending moment
stresses induced in the lower flange of a main girder
of Auxerre bridge under the two week traffic of the
A9 motorway are given in Table 1, for the natural
traffic and for the modified traffic with 44 ton
trucks. The minimum values are negative because of
the Courbon’s coefficient of the opposite lane of the
considered girder. With the 44 ton truck modified
traffic, the extreme values computed are increased
by 6.5% (minimum) and 8.5% (maximum), i.e. less
then the gross weight increase of the fully loaded
trucks, because these extreme values are obtained
while two trucks are crossing each other, one of
them may exceed the legal load.
Using the level crossing histogram of the same
effect, and the Rice formula as explained in section
4.1, the minimum and maximum values are extrapo-
lated for return periods of 25, 50, 500, 1000 and
5000 years (Table 2). For the three chosen Kolmo-
gorov thresholds, β=0.9, β=0.95 and β=0.99, the ex-
trapolated values are the same; that means that the
simulations has converged (Cremona 2001). The
maximum stresses for a 2000 year return period, as
specified in the Eurocode 0 (CEN EN1990, 2002)
for traffic load characteristic values, are close to 100
MPa, which seems to be coherent with the bridge
design. However, the real local traffic on this bridge
is much lower and lighter than the A9 motorway
traffic.
Table 1. Extrema of the computed stresses, Auxerre, main
girder mid-span bending moment, lower flange, A9 traffic and
modified traffic with 44 ton trucks (2 weeks).
______________________________________________
σmin (MPa) σmax (MPa)
______________________________________________
40 tons -6.1 35.3
44 tons -6.5 38.3
______________________________________________
Table 2. Extrapolated minimum and maximum stresses, Aux-
erre, mid-span bending moment, lower flange, A9 traffic, for
different return periods.
______________________________________
_____
Rt (yrs) σmin (MPa) σmax (MPa)
___________________________________________
25 -6.1 67.2
50 -6.7 74.1
500 -7.1 93.4
1000 -7.3 98.5
5000 -7.6 109.3
___________________________________________
5.4 Fatigue damage
To assess the fatigue damage and detail lifetime, the
rain-flow histograms are used, with the S-N curves
and the Miners’ law (section 4.2). Figure 5 gives the
rain-flow histograms of the bending moment stresses
in the lower flange at mid-span for both traffics (40
tons and 44 tons) on Auxerre bridge. An increase of
10% of the peak abscissa (18 to 20 MPa) is visible
when the 44 ton trucks are introduced. Reversely,
the distribution shows a drop at 18 MPa in this case,
because truck gross weights between 36 and 40 tons
were increased by 10%, and thus, there are no more
trucks in this weight interval.
Figure 5. Rain-flow distributions of the stresses in Auxerre
bridges at mid-span, lower flange, both traffics.
Then the fatigue damage is calculated for the S-N
curves 56, 63 and 71 MPa (Table 3). The detail is
classified in class 71 but the flange is 70 mm in
thick, and therefore a reduction factor of:
77.070/25
4= leads to the class 56.
The annual damage is obtained by multiplying
the weekly damage by 50. The conventional lifetime
is the inverse of the annual damage. The lifetimes
for Auxerre bridge are given in Table 3 for both traf-
fics. The values are very short, but the traffic of the
A9 motorway, one of the busiest in France, is much
more severe than the local traffic around Auxerre.
The most important finding is the reduction of the
lifetime by almost 20% when the 44 ton trucks are
introduced. That gives an estimation of the effect of
such a change in the regulation for bridges.
Table 3. Auxerre bridge lifetime for A9 traffic, 40 or 44 t gross
weight limit.
______________________________________________
Lifetimes (in years)
____________ _____________
Fatigue class 40 tons 44 tons
______________________________________________
Class 56 10.1 8.5
Class 63 14.3 12.0
Class 71 20.1 16.9
_____________________________________________
Similar results are given for Libourne bridge, and
more precisely for the weld of the vertical stiffener
on the girder lower flange at mid-span of the first
span. The fatigue class is 60 after the reduction due
to the flange thickness. The calculated lifetimes are
given in Table 4 for both traffics. The lifetime with
the natural traffic is app. 55 years, and reduced to 45
years with the 44 ton trucks.
Table 4. Libourne bridge lifetime, 1st span, bending moment of
it.
the lower flange, for A9 traffic, 40 or 44 t gross weight lim
______________________________________________
Lifetimes (in years)
____________ _____________
Fatigue class 40 tons 44 ton
______________________________________________
Class 50 34.6 28.9
Class 56 48.9 40.8
Class 63 69.0 58.8
_____________________________________________
These calculations show a decrease of the life-
time by app. 20% with the 44 ton trucks, as with
Auxerre bridge.
6 CONCLUSIONS
Bridge code calibration as well as assessment or re-
assessment of existing bridges under traffic loads
require an accurate knowledge of the load patterns.
WIM data are very useful for these applications, ei-
ther to give an account of the current traffic loads or
to forecast the potential impact of future loads.
With the computer software POLLUX developed
in the LCPC, the stresses induced in a fatigue sensi-
tive detail of two composite bridges, the weld be-
tween a vertical stiffener and the lower flange of a
main girder, were calculated. The traffic of a heavy
trafficked motorway in south of France, measured
over two weeks by an accurate WIM system, was
used. The same traffic was then modified to account
for a possible change in the gross weight limit in the
French law, being increased from 40 to 44 tons for
articulated 5-axle trucks and road trains. Extrapo-
lated extreme values and lifetime in fatigue were
calculated for both traffics and bridges.
The increase of the extreme stresses was limited
to 6.5 or 8.5%, on a 40 m simple supported span
(bending moment at mid span effect), carrying three
traffic lanes. The multiple presence of more than
one truck in adjacent lanes and the passage of excep-
tional or overloaded trucks explain that this increase
is less than 10% (40 to 44 t).
The computed lifetimes in fatigue for both
bridges are rather short, because the real traffics on
these bridges are local traffics and much lighter than
the motorway A9 one. However, the reduction of the
lifetime, whatever the fatigue resistance (S-N class)
and the bridge, is close to 20%. This value seems
reasonable because of the non linearity of the fatigue
damage with the load, while not all of the trucks are
affected by the weight increase. Moreover, the carri-
ers claim that increasing the weight limit by 10%
would slightly reduce the number of trucks on the
roads. Thus, we can estimate the lifetime reduction
for fatigue sensitive bridges between 10 and 20%, if
the weight regulation is changed.
7 REFERENCES
AFNOR (1993). Fatigue sous sollicitations d'amplitude varia-
ble. Méthode rain-flow de comptage des cycles. A03-406.
Norme française.
Bailey, S.F. & Bez, R. (1996), Site Specific Traffic Load Mod-
els, 7th International Conference on Applications of Statis-
tics and Probability in Civil Engineering, ICASP 7, vol. 3,
AA Balkema, Rotterdam.
Bruls, A., Croce, P., Sanpaolesi, L. & Sedlacek, G. (1996).
EVN1991-Part 3: Traffic Loads on Bridges; Calibration of
Load Models for Road Bridges, Proceedings of IABSE Col-
loquium, Delft, The Netherlands, IABSE, 439-453.
CEN EN1990 (2002). Eurocode 0: Basis of structural design.
CEN EN1991-2 (2003). Eurocode 1: Actions on structures -
Part 2: Traffic loads on bridges.
CEN EN1993-1-9 (2005). Eurocode 3: Design of steel struc-
tures - Part 1-9: Fatigue strength of steel structures.
Cremona, C. (2001). Optimal extrapolation of traffic load ef-
fects. Structural Safety (23): 31-46.
Ditlevsen, O. (1994). Traffic loads on large bridges modeled as
white noise fields. ASCE Journal of Engineering Mechan-
ics 120(4): 681-694.
Eymard, R. & B. Jacob (1989). Un nouveau logiciel : le pro-
gramme CASTOR pour le Calcul des Actions et Sollicita-
tions du Trafic dans les Ouvrages Routiers. Bulletin de liai-
son des laboratoires des Ponts et Chaussées (164): 64-78.
Flint, A. & Jacob, B., (1996), Extreme Traffic Loads on Road
Bridges and Target Values of their Effects for Code Cali-
bration, Proceedings of IABSE Colloquium Basis of Design
and Actions on Structures, IABSE, Delft, 469-477.
Hwang, E.S. & Nowak, A.S. (1991). Simulation of dynamic
load for bridges. Journal of structural engineering, ASCE,
vol 119, 6, 853-867.
Jacob, B., Maillard, J.B. & Gorse, J.F. (1989). Probabilistic
traffic load models and extreme loads on a bridge, Proceed-
ings of ICOSSAR'89, 1973-1980.
Jacob, B. (1998), Application of weigh-in-motion data to fa-
tigue of road bridges, Pre-Proceedings of the Second Euro-
pean Conference on WIM of Road Vehicles, eds O’Brien
and Jacob, Lisbon, European Commission, Luxembourg,
214-229.
Jacob, B. & Kretz, T. (1996). Calibration of bridge fatigue
loads under real traffic conditions. Proceedings of the
IABSE Colloquium on Eurocodes, Delft, March, 479-487.
Jacob, B., OBrien, E.J. & Jehaes, S. (eds.) 2002. Weigh-in-
Motion of Road Vehicles: Final Report of the COST 323
Action, LCPC Publications, Paris, 535 pp.
Jacob, B. & Labry, D. (2002). Evaluation of dynamic effects of
heavy vehicles on bridge fatigue, Proceedings of 7th Inter-
national Symposium on Heavy Vehicle Weights & Dimen-
sions, Delft, The Netherlands, June.
Marchadour, Y. & Jacob, B. (2008). Development and imple-
mentation of a WIM network for enforcement in France,
Proc. of the International Conference on Heavy Vehicles -
ICWIM5, eds. B. Jacob et al., ISTE/J. Wiley, 335-346.
Nassif, H.H. & Nowak, A.S. (1996). Dynamic load for girder
bridges under normal traffic. Archives of civil engineering,
XIII, 4.
O'Connor, A.J., Jacob, B., O'Brien, E.J. & Prat, M. (1998). Ef-
fects of traffic loads on road bridges - Preliminary studies
for the re-assessment of the traffic load model for Eurocode
1 Part 3. Pre-Proceedings of the Second European Confer-
ence on WIM of Road Vehicles, eds O’Brien and Jacob,
Lisbon, European Commission, Luxembourg, 231-242.
Rice, S., Ed. (1954). Mathematical analysis of random noise.
Selected Papers in Noise and Stochastic Processes New
York : Dover.
Vrouwenvelder, A. & Waarts, P.H. (1993). Traffic Load on
bridges, Structural Engineering International, 169-177.