Figure 1 – Test site – Installation plan (scheme)
PORTABLE AND MULTIPLE-SENSOR
WIM SYSTEMS TRIAL
EXECUTIVE SUMMARY
(Report of the COST 323 Management Committee, published by TU-Wien, Vienna, AT)
The European COST 323 Action (Weigh In Motion of Road Vehicles), whose one objective is to assess the performance of commercially available WIM systems, a test was carried out in France, in June 1996, with four different portable WIM systems manufactured or distributed in Europe and used in some EC member countries. The behaviour and accuracy level of these systems are studied, together with those of a multiple-sensor WIM system (MS-WIM).
It is the first test in Europe involving several portable WIM systems on the same site. The test site near Trappes, between Paris and Chartres (France), was chosen as being representative of a current European expressway, in respect of both pavement and traffic conditions. Another major reason for this choice was that an MS-WIM array had been installed on the road in 1994 for a OECD project. This was therefore a unique opportunity to also assess the performance of such an emerging WIM technique with a proper test plan. In a second step, the MS-WIM is used to provide reference values for static axle loads and gross weights of all vehicles of the traffic flow passing this section during the test period, for an assessment of the portable WIM system accuracy on a wider scale.
The accuracy of all tested WIM systems is assessed according to the draft 2.2 (June 97) of the European Specification on WIM Systems.
The test site is located on the highway RN10 at La Verrière near Trappes, 35 km South - West of Paris. It is a heavily trafficked highway with 2 x 2 lanes and 30,000 vehicles/day (both directions together) including 25 % heavy trucks. The pavement is made of a 50 cm thick cement stabilised layer, a 15 to 18 cm thick asphalt concrete base and an 8 cm thick bituminous concrete top layer. The site meets the requirements of a ‘good’ WIM site (class II) according to the European Specification on WIM Systems. A static weighing area is located on the RN10, about 5 km upstream of the site. It is periodically used for enforcement purposes.
Under the OECD/DIVINE project a MS-WIM array with 24 piezoceramic bars was installed on this test site in May 1994 and October 1995, and was connected to three WIM stations sheltered in a concrete housing under a bridge crossing the highway. This WIM array consists of 24 non-uniformly spaced sensors, on a total length of 36 m, and of 7 magnetic loops.
A schematic map of the test site including the position of the four portable WIM systems is given in Figure 1.
Figure 1 – Test site – Installation plan (scheme)
Each of the four portable WIM systems consists of two inductive loops with one weighing sensor in-between and a data acquisition unit. One system uses a capacitive strip sensor (Golden River CS) while the three other systems use capacitive mats. The details are presented in Table 1.
Table 1: Portable WIM systems tested
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Different samples of recorded data are considered and analysed for each system, according to the conditions, which are defined in the European Specification on WIM Systems. Several runs of two different test vehicles, a 2-axle rigid lorry (test vehicle #1) and a 5-axle articulated lorry (test vehicle #2), with different speeds and loads were performed and analysed separately (full and extended repeatability conditions (r1) and (r2)). For limited reproducibility conditions (R1) all runs of the two test vehicles mentioned are taken as one sample for accuracy analysis. Under full reproducibility conditions (R2) two different vehicle populations are considered:
Environmental repeatability conditions (I) apply according to the European Specification on WIM Systems because the test was carried out over three consecutive days under the similar temperature and climatic conditions.
For each run of a test vehicle or pre-weighed vehicle, and for each entity (a) to (d), the relative error of the axle load or gross weight recorded by the WIM system to the measured static weight (reference weight) is calculated. Accuracy evaluation and system classification is based on these error calculations. Accuracy calculations and system classification are performed for limited reproducibility conditions (R1), where all test vehicle runs are considered as one sample, and for full reproducibility conditions (R2) using vehicle population 1 and population 2 data.
The results of the accuracy classification are not always consistent between the different entities considered. As best performing portable WIM system Truvelo system is in class D(25) for gross weight and in class E(30) for all other entities. Golden River CS system, which provided the most scattered results, is in class E(50) for gross weight, in class E(45) for single axles and in classes E(95) and E(100) for group of axles and axles belonging to an axle group. Mikros system classifies in E(40) for gross weight and single axles. Both systems, Golden River CS system and Mikros system, which were badly calibrated for axle groups and axles of group weighing, provide low levels of confidence for these two criteria. Golden River System CM is in class E(30) for gross weight and axles of a group but only in E(35) for group of axles and even in E(45) for single axles.
The MS-WIM system meets the requirement of the highest accuracy class A(5) for two criteria, i.e. for the axle groups and axles of group taken individually. It is in accuracy class B+(7) for the gross weight. Because of underestimating significantly front axle load and overestimating second axle load the MS WIM classifies only in C(15) for single axles.
Table 2 summarises the accepted accuracy classes for each entity considered and the final accepted accuracy class for each system under limited reproducibility conditions (R1, test vehicles).
Table 2: Accuracy classes accepted in limited reproducibility conditions (R1) – test vehicles
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Among the portable WIM systems only Truvelo system meets the requirements of class D(25) for gross weight and single axles, while it is in class E(30) for the other criteria. All the other portable systems are only accepted in one of the E classes. The Golden River CM system is in class E(30) for gross weight and groups of axles, E(35) for axles of a group and E(40) for single axles. The two other systems (Golden River CS and Mikros) are in classes E(45) to E(70) for all criteria, except Mikros system for single axles in class E(35).
The MS-WIM system meets requirements of accuracy class B+(7) for three criteria, and misses it for the axle groups criterion by 1.1%. That is deemed to be substantially in compliance with class B+(7) and therefore this system was approved to provide an acceptable estimate of the static loads as reference to verify the accuracy of the portable systems using all trucks of the whole traffic flow (population 2).
Table 3 summarises the accepted accuracy of each system under full reproducibility conditions (pre–weighed vehicles/population 1). The results are very consistent with those of the initial accuracy check performed with the two test vehicles used for system calibration (limited reproducibility conditions – R1). The differences are less than one accuracy class for all criteria and systems. Therefore these results confirm the consistency of the figures proposed by the European Specification concerning the confidence limits for different test conditions.
Table 3: Accuracy classes accepted under full reproducibility conditions (R2) population 1 - pre-weighed lorries
All systems are in accuracy classes E. Truvelo system is the most accurate. For gross weights it is in class D(25) and for the three other entities in class E(30). For Golden River CS system only gross weight measurements can be considered because no information about vehicle type and axle number is provided by the used data processing software. The system classifies in E(40) for this criterion. Mikros system meets the requirements of class D(25) for gross weights and single axles. For axles of a group this system is in class E(35) and for axle groups in class E(40). Golden River CM system meets the class E(30) for gross weights and E(35) for groups of axles and axles of a group but for single axles again only the class E(40) can be accepted.
Table 4 summarises the finally accepted accuracy of each system under full reproducibility conditions (traffic flow vehicles/population 2). In comparison with the results of the accuracy evaluation with the pre-weighed vehicles (population 1) Truvelo system and Golden River CM system remain in the same accuracy class E(30). Mikros system shows a higher accuracy by taking into account all vehicles of the traffic flow (population 2) and is classified in class E(40) instead of E(70). The higher scoring of Mikros system for population 2 vehicles may be explained by the many unloaded or half loaded vehicles in population 1 (pre –weighed vehicles), which loads are particularly poor estimated by this system. Also Golden River CS system has an increased accuracy class E(40) instead of E(45) for gross weights.
Table 4: Accuracy classes accepted under full reproducibility conditions (R2)
population 2 – traffic flow vehicles
However, such systems show some weaknesses in properly assessing axle and vehicle weights and therefore classify only in lower accuracy classes ranking between D(25) and higher. All given results are to be seen in the light of a rather poor calibration of all portable WIM systems tested in this trial. The calibration may likely be improved which certainly will enhance their field performance. Nevertheless, the evident dynamic effects artificially induced by measurements with such type of surface mounted sensors limit their accuracy, especially for weighing light, unloaded or partially loaded vehicles. Better calibrated, such portable systems may be used for traffic statistics collection, traffic survey and other engineering applications which do not require a high accurate estimation of the actual load distribution. In some cases, they may be helpful to detect overloaded vehicles in the traffic flow, while the risk of wrong pre-selection or of non detection remains rather high according to the large individual errors encountered.
Regarding the two different sensor types tested in this trial clearly the capacitive mat type sensor is much more reliable and accurate than the capacitive strip sensor. This simply is caused by their different geometry. Because of the poor performance a strip type sensor mounted on the road surface is not recommended to be used for portable WIM systems.
As further objective of the test accuracy analysis proves that the existing MS-WIM system, which consists of several piezoceramic bar sensors and which were installed at the site two years before the test, provided an accurate estimate of static loads. The static loads were calculated by a simply computing the average of the loads measured by each of the 13 sensors after careful calibration. The MS–WIM system meets for 3 criteria the accuracy class B+(7) requirements; it fails this class slightly for one criterion.
B+(7)Results in full reproducibility conditions (R2) – population 2 (traffic flow vehicles)
The total number of vehicles recorded from the traffic flow (population 2) and consequently also the size of the data sample used for accuracy calculation is different for each portable WIM system, because of the different times of operation of each system during the test.
Conclusions
In summary the experiment provided appealing information on the behaviour and accuracy of portable WIM systems currently on the market and in operation in some European countries. The main advantage of these systems is that they only require almost one or two hours for installation. They may be removed even within a shorter time. As such systems may be used on many different sites and do not require any huge infrastructure they are advantageous flexible and have therefore an economical mean to collect statistical traffic data over short time periods.