Florianópolis - Santa Catarina - Brazil April 3rd to 7th - 2011

Assessment of a Bridge WIM System on Integral Concrete Bridges and on Steel Orthotropic Decks Sio-Song Ieng [email protected] French Institute of Science and Technology for Transport, Development and Networks (IFSTTAR).

Franziska Schmidt [email protected] French Institute of Science and Technology for Transport, Development and Networks (IFSTTAR).

Frédéric Romboni [email protected] French Institute of Science and Technology for Transport, Development and Networks (IFSTTAR).

Bernard Jacob [email protected]

Abstract Bridge-Weigh-In-Motion uses bridges as a scale to weigh vehicles. Practically, this is done by measuring the strains in that bridge, and relating them to the weight and dimensions of a truck called “calibration trucks” whose shape and axle weights are well known. This article summarizes different B-WIM experiments, which the institute IFSTTAR (formerly LCPC) carried out, and the lessons drawn from this experience. First, the system has been tested on integral concrete slab and frame bridges. Different parameters have been investigated such as the pavement evenness and the bridge skewness. Then, orthotropic deck bridges have been instrumented and it was shown that for such bridges several points need to be further improved. Keywords: bridge, WIM, B-WIM, free of axle detectors, strain, gauges, transducers, loads, weights, trucks

Avaliação de um sistema de pesagem em movimento em pontes (BWIM) de concreto e em pontes de aço com decks ortotrópicos Resumo A tecnologia Pesagem em Movimento em Pontes (Bridge-Weigh-in-Motion – B-WIM) faz uso das pontes como uma plataforma para pesar os veículos. Na prática, isso é realizado por meio da medição das tensões na ponte e fazendo-se uma relação com o peso e as dimensões de um caminhão chamado “caminhão de calibração”, cujas formas e pesos por eixo são conhecidos. Este artigo resume diferentes experimentos de B-WIM realizados pelo IFSTTAR (outrora LCPC) e os ensinamentos depreendidos da experiência. De início, o sistema foi testado em pontes integrais de estrutura e laje de concreto. Investigaram-se diferentes parâmetros, como a irregularidade do pavimento e a assimetria da ponte. Depois, as pontes de tabuleiro ortotrópico foram equipadas e verificou-se a necessidade de implementação de melhorias no sistema. Palavras-chave: Ponte, WIM, B-WIM, Detector de eixo livre, Tensão, Medidor, Transdutor, Carga, Peso, Caminhão.

1. Introduction Bridge Weigh-in-Motion (B-WIM) systems use strain sensors, gauges or transducers, attached underneath the bridge deck, and an algorithm to assess axle loads and gross vehicle weights when a vehicle passes over the bridge at regular speed. B-WIM was first introduced in the USA by (Moses, 1979) as an alternative to the traditional road WIM sensors. But the main advances on B-WIM were done in the European research project Weigh-in-motion of Axles and Vehicles for Europe 1996-99 (WAVE) (Jacob, 1999 and 2002). Today, the only commercial marketed system is the SiWIM by

1

Florianópolis - Santa Catarina - Brazil April 3rd to 7th - 2011

Cestel, a Slovenian company, supported by ZAG. The system has advantages compared to the road sensor WIM systems: (i) the installation does not need civil engineering works, (ii) the installation does not disrupt the traffic flow, (iii) as the sensors are underneath the bridge, they should be more durable than road sensors directly exposed to the traffic loads and to the rain. The earlier B-WIM systems required road sensors to detect and count axles. But since the WAVE project and the concept of FAD (Free-of-Axle-Detector), strain sensors are able to detect axles and to calculate vehicle speed (Znidaric et al., 1999). However, to achieve an accurate weighing, a reliable B-WIM system shall be installed on a suitable bridge which must be seen as a part of the weighing system. Beside the sensors, which are only one part of the whole system, the bridge structure is involved in the measurement process. Indeed, the behavior of the bridge is taken into account by the algorithm through the concept of influence line (IL) (Ieng, 2010) or influence surface (IS). For this reason, a B-WIM system can be suitable for some kind of bridge and may be improved or adapted to other kinds. The first experimentations in France were carried out on integral slab bridges with a SiWIM system in 2005. The quality of the bridge and the pavement on it, are very important for a good accuracy. Today, the tests carried out on integral slab bridges in France provide accurate weights. The results are presented in the section 2. The SiWIM system is currently experimented in France on orthotropic steel deck bridges. The design and the behavior of such bridges are very different from the integral concrete slab bridges, with a 2dimensional structure made of longitudinal stiffeners supporting the deck plate, and the cross beams giving the transverse stiffness. For steel durability purpose, the extensometers cannot be screwed directly on the stiffeners, but on some steel plates glued on the bottom of the stiffeners. The experimentation is carried on the Millau viaduct (Jacob et al., 2010). The first results were promising, in class D+(20) or C(15) according to the European Specifications on WIM (Jacob et al., 2002), depending on the vehicle sample considered. However, it is shown, in section 3, that the current BWIM algorithm should be adapted and improved for the orthotropic deck bridges, to get a higher accuracy. The B-WIM applications are numerous. The most important are overloaded truck preselection for enforcement and bridge assessment and monitoring.

2. Experimentations in France on Integral Bridges 2.1. Integral Slab Bridge Instrumentation In the European project WAVE, B-WIM techniques were investigated and the algorithms were improved (O'Brien et al., 2001). It was proven that the best results were obtained on spans of 5 to 10 m. The recommendations for the choice of a suitable bridge, are aimed to ease the installation and the experimentation: •

the underside of the bridge should be easy accessible (location, height...),

•

the single span bridge length is less than 10 to 15 m,

•

the bridge should be as straight as possible, with no or very low slope and skewness (< 10°),

•

the pavement should be smooth without bumps and ruts,

•

the electricity and telecommunication networks should be available on the site.

While the earliest B-WIM systems used both strain sensors installed under the bridge and axles detectors on the pavement, the SiWIM system only uses removable transducers fixed on the bottom of each longitudinal stiffeners (orthotropic bridge), or under the bridge's slab (integral bridges). This principle is called “Free of Axle Detectors” (FAD) (Znidaric et al., 1999). The transducers installation may be different for different kinds of bridges. For integral bridges, transducers can be screwed directly under the deck (Figure 1).

2

Florianópolis - Santa Catarina - Brazil April 3rd to 7th - 2011

The SiWIM system used in France consists in 16 transducers. Four sensors are dedicated to vehicle detection and velocity measurement. The 12 others sensors are used to weigh vehicle’s axles. They are placed laterally across the two traffic lanes of the bridge. Figure 1 shows a typical configuration of the sensors. However, depending on the bridge characteristics, the sensors configuration can be changed.

Figure 1: Transducers installation on an integral slab concrete bridge

2.2. Accuracy Assessment Several experiments were conducted across France on integral slab bridges, with the main objective of overloaded truck pre-selection for enforcement. The bridges were chosen following the criteria presented in section 2. But some characteristics of the bridge might not meet entirely the criteria. For the calibration, one or two calibration trucks are needed. The choice of the trucks suspension is very important. One should avoid steel suspensions, which are not anymore common in Europe and induce high dynamic effects, above all in the bogies. Air suspensions are highly recommended. Experimentations are conducted over one to three days, as follows: •

first day: one or two calibration trucks representative of the traffic flow (e.g. in France a rigid 3-axle truck (C3) and an articulated 5-axle semi-trailer) were used to calibrate the WIM system by performing repeated runs at different velocities and transverse locations,

•

following days: trucks are picked up from the traffic flow, either if they are detected as overloaded by the SiWIM system or not, and they are leaded by policemen to a static scale for weighing and other parameters (e.g. axle spacing) measurements,

•

accuracy assessment: according to the COST323 European specifications of WIM (Jacob et al. 2002), the relative errors er are calculated for each axle and group of axles load, and gross

⎛ Wd − Ws ⎞ ⎟⎟ , where Wd and Ws are the in-motion loads measured by the ⎝ Ws ⎠

vehicle weight: er = ⎜⎜

SiWIM and the static loads. Four sub-populations were considered: (i) gross vehicle weight, (ii) single axles, (iii) axle group (i.e. tandem and tridem), and (iv) axles of a group. The statistics of the relative errors are calculated for each sub-population. The main results are presented in the following sub-sections for each instrumented bridge

2.3. Nogent-sur-Seine Bridge (RN19) The bridge is located on the highway RN19, near the town of Nogent-sur-Seine, 90 km east of Paris. It has one span of 10 m in length and 11 m in width, an a slab thickness of 0.6 m. The bridge is perfectly straight (Figure 2). A static weighing area is located 3 km up-stream of the bridge. The traffic comprises almost 1500 trucks/day. In the direction from Paris to province (L1), the pavement on the

3

Florianópolis - Santa Catarina - Brazil April 3rd to 7th - 2011

bridge is smooth without bump, rut or crack whereas, in the opposite direction (L2), there is a quite important bump that produces oscillation when a vehicle passes over it. Thus, the test of the SiWIM system on this bridge allows to check the influence of the pavement profile on the accuracy. The system was installed in both directions. The selection of this bridge was done with the advice of Cestel and the installation was performed in September 2005.

2.0 5 3.6 11.1

3.6

1.9 10

Figure 2: Nogent-sur-Seine integral slab bridge, RN19

The accuracy of the SiWIM system is shown in tables 1 and 2 for both lanes. The test conditions are R2 (full reproducibility with trucks from the traffic flow) and EI (full environmental repeatability, test carried over one day). Table 1: Accuracy of the B-WIM system on the lane L1, Nogent-sur-Seine, RN19, test conditions R2/EI Criteria

Number

Mean

Std. Dev

Class

Gross Weight

11

-0.03%

3.1%

B+(7)

Group of Axle

12

-1.5%

3.9%

B+(7)

Single Axle

21

-1.9%

6.4%

B(10)

Axle of group

33

-0.7%

11.9%

C(15)

Global Accuracy Class

B(10)

No mandatory criterion

Table 2: Accuracy of the B-WIM system on the lane L2, Nogent-sur-Seine, RN19, test conditions R2/EI Criteria

Number

Mean

Std. Dev

Class

Gross Weight

28

-0.26%

4.64%

C(15)

Group of Axle

12

-1.898%

4.76%

C(15)

Single Axle

56

-0.45%

5.74%

C(15)

Axle of group

36

1.906%

9.67%%

D+(20)

Global Accuracy Class

C(15)

No mandatory criterion

The comparison of the table 1 and 2 results shows that the bump, inducing some dynamic motions of the trucks, leads to drop the accuracy by one class, B(10) on L1 and C(15) on lane L2. The criterion of axles of a group was not considered for the final accuracy, as suggested in the COST323 European Specifications, because it should not be the governing criterion for B-WIM. However, the accuracy for this criterion is only one class below the final accuracy class. In conclusion, it is important to choose bridges with a smooth pavement profile for B-WIM system implementation.

4

Florianópolis - Santa Catarina - Brazil April 3rd to 7th - 2011

2.4. Montpellier Bridge (A9) The bridge is located in the south of France near Montpellier. It carries the 3 traffic lanes of the south to north bound of the motorway A9 that connects Spain and France. The bridge has a span of 6.5 m in length and 16.3 m in width, and the slab thickness is 0.6 m. The characteristics of the bridge meet every criteria of a good bridge for B-WIM. The heavy traffic is very dense with app. 6,500 trucks/day and direction. However the calibration truck was provided with steel suspension. Moreover, because of the very high truck density, there are often two trucks side by side on the bridge. In this case the SiWIM system cannot distinguish well each truck which significantly affects the accuracy. Only the slow lane was considered during this experimentation. Table 3 shows the accuracy of the B-WIM system, in test conditions R2 and EI (two consecutive days). Table 3: Accuracy of the B-WIM system for the slow lane, Montpellier, A9, 2009, test conditions R2/EI Criterion

Number

Mean

Std. Dev

Class

Gross Weight

91

2.06%

4.35%

B(10)

Group of Axle

88

3.23%

5.28%

B(10)

Single Axle

182

0.64%

4.99%

B+(7)

Axle of group

265

3.51%

9.29%

B(10)

Global Accuracy Class

B(10)

The accuracy found is already good and the results are reliable with a large truck sample from the traffic flow (91 trucks, 3 suspicious vehicles were eliminated from the original sample of 94 trucks), but we can even expect a better accuracy if using a faster algorithm for such a dense traffic and a calibration truck with air suspensions. More experiments conducted in France on similar bridges gave the same accuracy.

3. Orthotropic Deck Bridge WIM In 1996 the IFSTTAR (ex LCPC) launched the idea of using steel orthotropic deck bridge (OB) instrumentation as a B-WIM option (OB-WIM). This type of structure is rather flexible and the longitudinal stiffeners’ bending between two cross beams, spaced by 4 to 4.5 m, are sensitive to wheel and axle loads (figure 3). These stiffeners are generally spaced by 0.6 m, and thus mainly responding to a single or twin wheel load. Strain gauges or extensometers are fixed on the bottom of each stiffener under the traffic lanes. However, for the orthotropic deck bridges, screws may damage the bridge. Therefore, the safest solution is either to use glued strain gauges or to glue steel mounting plates at the bottom of the stiffeners. The SiWIM transducers are then screwed on the mounting plates as shown in the figure 4. The experimentation is conducted as presented in the section 2. Two OBs were instrumented in France. The first instrumentation was performed on Autreville bridge over the Moselle river, on the A31 motorway, during the WAVE project, and more recently on the Millau viaduct.

3.1. Autreville bridge Experiments on Autreville bridge were conducted in August 1997 and July 1998 with a prototype of B-WIM developed in the WAVE project by (Dempsey et al., 1999). 1-D and 2-D models were used with the same measurements. The results are reported respectively in tables 4 and 5.

5

Florianópolis - Santa Catarina - Brazil April 3rd to 7th - 2011

Table 4: Autreville, accuracy of the weighing system (1-D model), test conditions R2/EI Criterion

Number

Mean

Std. Dev

Class

Gross Weight

44

1.12%

7.61%

D+(20)

Group of Axles

44

-0.5%

10.12%

D+(20)

Single Axle

89

3.79%

10.10%

D+(20)

Global Accuracy Class

D+(20)

Table 5: Autreville, accuracy of the weighing system (2-D model), test conditions R2/EI Criterion

Number

Mean

Std. Dev

Class

Gross Weight

28

0.55%

5.55%

C(15)

Group of Axles

27

0.17%

8.22%

C(15)

Single Axle

55

0.79%

8.95%

C(15)

Global Accuracy Class

C(15)

With the 1-D bridge model, the accuracy classes for all criteria are in the class D+(20). This accuracy is lower than that of common road sensor WIM systems installed on good or excellent WIM sites. The standard deviations of the relative errors are rather high, which cannot be fixed by calibration. The 1-D model, i.e. the longitudinal influence lines of the stiffener bending moments, is not accurate enough to account for the variations of the lateral wheel location and of the tire types, single or twin wheel and regular. With the 2-D algorithm, the mean bias are very low and consistent, and the standard deviations are reduced by almost 2%. Therefore the accuracy classes are C(15) for all criteria, comparable to most of the good road sensor WIM systems installed on good and excellent WIM sites. The advantage of the 2-D algorithm is clear.

3.2. Millau viaduct The Millau viaduct is a 2,460 m in length cable stayed multiple span bridge, in the south of France. It carries the motorway A75 (Clermont-Ferrand to Béziers and Montpellier). The deck is a steel orthotropic box, 32 m in width and 4.20 m in height, which carries 2 lanes and an emergency lane in each direction. The viaduct has an up-hill slope of 3% in the north-south direction and a radius of curvature of 20 km. The average daily traffic is app. 12,000 vehicles incl. 12 to 15% of trucks (i.e. 1500 trucks/day). The instrumented section is located in the first span, close to the north end of the viaduct, under the slow lane in the south-north direction. The trapezoidal stiffeners 2 to 13 were instrumented by extensometers, which covered the lane 1 and part of the lane 2 and the emergency lane, at mid-span between two cross beams (figure 5). Four extensometers were installed on the stiffeners 4 to 7, 4 m upstream, as part of the FAD B-WIM system.

Figure 5: Instrumentation of the Millau viaduct orthotropic steel box

6

Florianópolis - Santa Catarina - Brazil April 3rd to 7th - 2011

The measuring location was chosen upstream of the expansion join, in a current bridge section where the traffic speed is rather constant (the truck speed limit is 80 km/h on the viaduct) in order to avoid dynamic effects which could affect the WIM system. The SiWIM was installed inside the box, well protected against rain and wind. The experimentation was conducted as it is presented in section 2. However, communication means were more important because the SiWIM operator can not see the vehicles on the bridge. Only the vehicles passing on the slow lane were recorded. The results are given in Table 6. Table 6: Millau, accuracy of the weighing system, test conditions R2/EI Criterion

Number

Mean

Std. Dev

Class

Gross Weight

43

-3.24%

5.76%

C(15)

Group of Axle

39

-8.01%

5.32%

C(15)

Single Axle

86

1.09%

11.18%

D+(20)

Axle of group

115

-7.93%

9..46%

C(15)

Global Accuracy Class

D+(20)

For all criteria but the single axles which are in the class D+(20), the system is in the class C(15). However, the axles of a group are under-weighed by 8%, while the single axles are over-weighed by 1.1% and are rather scattered with a standard deviation of the relative errors larger than 11%. This is due to some lack of the SiWIM algorithm for such type of bridge, which uses a 1-D model (see section 3.1). More precisely, the current SiWIM algorithm uses a weighed average of the measured strains on all the instrumented longitudinal stiffeners to calculate the loads and weights. The coefficients of the weighed average are adjusted in the calibration process. The single axles are either front axles with single standard tires or drive axles with twin tires. Thus, the front axles have a much narrower lateral imprint (≈0.25 m in width) than the drive axles (≈0.70 m in width), which may explain the high scattering of the errors. The difference between the mean bias of the single axles and axles of a group, which mainly consist of wide single tires (≈0.35 m in width), is less obvious. It may be a cumulative effect of the longitudinal and transverse wheel load distribution. Most of the groups of axles are tridem with a wheelbase of 2 x 1.3 = 2.6 m, on which the total load is applied in the same short span. Thus, the maximum stiffener bending strain is less than for a single axle of the same load, which may explain the rather large negative bias.

4. Conclusions 4.1. B-WIM Applications B-WIM weights trucks through the deformations of bridges. While providing both strain, axle and vehicle load measurements, a B-WIM is fully adapted for bridge assessment and monitoring applications. It is an efficient tool for fatigue damage assessment on steel bridges, because the calculated stresses under the measured traffic loads can be calibrated thanks to the strain measurements, to take into account the dynamic effects, and then the traffic data may be extrapolated over the whole bridge lifetime. The pre-selection of overloaded vehicles is another important application. Overloading of trucks leads to disrupt the fair competition between carriers. It has been calculated that 20% of overloading results in a gain of 26,406 euros per year for a 5-axle articulated truck of 40 t. The effects of overloads on road safety and infrastructure lifetime are significant. Pre-selection allows to focus the checks on the overloaded trucks, and not to stop partially or legally loaded trucks, and thus to increase the efficieny of enforcement.

4.2. Conclusions and Perspectives B-WIM system provides advantages in overloaded trucks control compared to current road sensor

7

Florianópolis - Santa Catarina - Brazil April 3rd to 7th - 2011

WIM systems: • • •

safe system installation, traffic flow is not disturbed or stopped during the installation, sensors are not visible by the road users.

However, it cannot fully replace road sensor WIM systems, because a bridge is needed. The bridge, as a part of the whole B-WIM system, should be chosen carefully. The French experiments provide interesting examples of validation and assessment of a B-WIM system. Due to the different behaviors of several bridges, the current system is efficient when used on appropriate kinds of bridges, such as integral slab bridges. But the algorithm should be improved or generalized to more types of bridges, such as the steel orthotropic decks. The investigation on B-WIM must now focus on: the installation and calibration techniques that can be used for other types of bridges, the axle weight estimation algorithms that take into account the behaviour of the bridge through influence lines or surfaces.

5. References Dempsey, A., Jacob, B. and Carracilli, J. (1999), Orthotropic Bridge WIM for determining Axle and Gross Vehicle Weights, Weigh-in-Motion of Road Vehicles, ed. B. Jacob, Hermes Science Publications Paris, pp 227-238,. Ieng, S.-S. (2010), Bridge Influence Line Estimation for Bridge Weigh-in-Motion System, Proceedings of IEEE 2010, Int. Conf. on Measurement and Control Engineering (ICMCE2010), November 16-18, Chengdu, China , pp 25-29. Jacob, B. (1999), “Proceedings of the Final Symposium of the project WAVE (1996-99)”, Paris, May 6-7, Hermes Science Publications, Paris, 352 pp. Jacob, B. (2002), Weigh-in-motion of Axles and Vehicles for Europe, Final Report of the Project WAVE, LCPC, Paris, 103 pp. Jacob, B. O’Brien, E.J. and Jehaes, S. (2002), Weigh-in-Motion of Road Vehicles - Final Report of the COST323 Action, LCPC, Paris, 538 pp., + French edition (2004). Jacob, B., Hannachi, M., and Ieng, S.S. (2010), Bridge weigh-in-motion on steel orthotropic decks Millau viaduct and Autreville bridge, in Proc. of the 5th Int. Conf. on Bridge Maintenance and Safety IABMAS2010, Philadelphia, July 11-15. Moses, F. (1979), Weigh-in-Motion System using Instrumented Bridge, ASCE Transportation Engineering Journal 105, pp 233-249. O’Brien, E.J., Znidaric, A., Baumgärtner, W. et al. (2001), Weigh-in-motion of Axles and Vehicles for Europe – Bridge WIM Systems, Report of the WP1.2, WAVE project, LCPC, Paris, 115 pp. Znidaric, A., Dempsey, A. , Lavric, I. and Baumgärtner, W. (1999), B-WIM Systems without Axle Detectors, Weigh in Motion of Road Vehicles, ed. B. Jacob, Hermes Science Publications Paris, pp 101-110.

8