(Zurich, 1993-1995)

Summary of the Final Report of May 1997
Copies of the whole report may be ordered from ETH at the cost of 30 ECU. Please, fill in the order form or contact:

Dr. Markus Caprez
ETH Zürich

Table of contents with abstracts in English and French

1. Introduction

The Swiss Federal Institute of Technology (ETH) and the COST 323 Management Committee (MC) organised the first European test of WIM systems/sensors in Hagenholz, nearby Zurich, from July 1993 until November 1995. The WIM test was planned on an urban road in the City of Zurich. Various systems and sensors currently available on the commercial market were installed together with a recently developed sensor. The aim of the project was to test the different systems and sensors under identical conditions over a long period.

Primary as part of the project "Dynamic Axle Load Measurements" by the "Commission for the Promotion of Scientific Research (KWF)", this large scale test became part of the second objective of the COST 323 action (Collection of information and evaluation of existing WIM systems), and was placed under the auspices of its Management Committee. This investigation was carried out in order to allow an objective evaluation of the systems and sensors. The results of this test also provide an opportunity to partially apply for the first time the draft European specification, and to give an idea of the accuracy classes of the WIM devices.

Some preliminary results were presented in the first European conference on WIM (Zurich, March 8-10, 1995). Then the main results were discussed with manufacturer's representatives. Afterwards a preliminary but uncompleted draft report was issued, including manufacturer's comments, and submitted to a panel of COST 323 MC's members.

During the year 1996, the European Specification on WIM were drafted. This Specification provides much stronger basis and theoretical background for any WIM test analysis than any of the former documents. Therefore the report of this WIM test was deeply revised late 1996. The analysis presented is in agreement with this Specification.

The results presented in this report only concern the gross weights of some limited types of lorries and vans, measured in motion on an urban road with a speed limit of 50 km/h. The results will therefore need to be supplemented by results from other tests on more trafficked highways and at higher speeds.

Very useful information was collected about the reliability and durability of the sensors and systems, including the electronics and software. Most of the devices encountered some problems, which were generally solved by the suppliers after some delay.

Some differences appear in the mean errors, coefficients of variation, repeatability of measurements and final accuracy between the different systems and sensors. However, it is not sufficient to describe and compare the systems and sensors using such criteria. There were differences between the sensor sizes and numbers (large scales, single or multiple strips); between sensors on their own and full systems; and in the investment costs of the systems and sensors.

2. The test site

The selected test site at Hagenholz in Zurich-Oerlikon was chosen close to the entrance to the garbage incinerator of the city of Zurich. Up to 200 vehicles per day, transporting rubbish to be incinerated are weighed in static on an approved weigh-bridge (precision ± 20 kg). Half of them are lorries and the other half are vans and cars. That allowed to perform a large number of static checks and to compare them with the WIM measurements of the corresponding vehicles, without disturbing the traffic flow.

On working days the daily traffic flow in the measuring direction was 7000 vehicles. The percentage of lorries is approximately 10 to 15 %, the average traffic speed 45 to 50 km/h.

The test site length is 150 m, providing space for all the proposed systems. The pavement is flexible and the road was build in 1975. The wearing course consists of a 40 mm thick bituminous layer (40 mm asphalt concrete, while the base course has a thickness of 200 mm with bituminous bound material and the sub-base consists of a 250 mm gravel layer. The bearing capacity of the capping layer is : E = 40 MN/m2. The site characteristics are given in Table 1 and compared to the tolerances of the European specification on WIM. The values are acceptable for WIM measurements, according to the limits of the European specification (site in class III).

Table 1: European Specification (COST 323), site criteria (average, class III)

Criteria Value Limit (class III)
longitudinal slope - ≤ 2%
slope in the transverse direction 2% ≤ 3%
radius of curvature 1000 m > 1000 m
Maximum Rutting 12 mm (*) ≤ 10 mm
Mean deflection (flexible pavement) 0.83 mm (**) ≤ 0.75 mm
Deflection difference (left/right) -- ≤ 0.15 mm
Evenness (IRI) 3.7 ≤ 4
(*) measured with a 4 m beam instead of a 3 m beam; may be considered equivalent to 10 mm.
(**) the measured value with a 100 kN axle is 0.643 mm at 4C; the standardised value is corrected for 130 kN.

3. The systems and sensors, and collected data

In spite of all the preparatory work it was not possible to install all the WIM sensors and systems at the same time. The main reasons were delivery delays of some companies and the delay of some participants in deciding whether to participate. Between the end of June and October 1993, the road had to be closed four times for individual additional installations.

The air temperature during installation was normally between 15 and 25C. Only while installing the later systems the temperature was lower than 10C. The relative humidity during installation was between 70 and 95 %. The installation procedure was guided and performed by the system supplier or by his representative with the help of ETH staff.

The systems and sensors installed are briefly presented in Table 2 and the general layout of the test site is shown in Figure 1. The sensors were connected Table 2 also give indications for each system/sensor about the date of installation, the periods of presence (TP) and operation (TO) in the test, and the percentage of recorded vehicles. The differences between TP and TO correspond to the breakdowns and failures. The percentage of vehicle recorded are given with respect to the maximum sample size of post-weighed vehicles during TP, i.e. 2340 for the systems and 1470 for the sensors, for the longest TPs. For the sensors by Focas and Thermocoax, the low percentages are not due to any failure, but are explained because they were not always connected to an electronics.

Table 2: Systems, combined systems and sensors tested

Company Sensor(s) Electronics Date of installation TO/TP
% of
Golden River 4 Capacitive strips Marksman 600
Marksman 660
25.5/31 84%
Peek Traffic 2 Piezo-ceramic strips AWACS 6000 29/6/93 24.5/31 68%
Mikros 1 Capacitive scale TEL-2CM 4/8/93 13/29 91%
ECM 2 Piezo-ceramic strips Hestia P 19/10/93 24.5 / 26.5 76 %
PAT 2 Bending plates DAW 100 20/10/94 10 / 15 7 %
Kistler / PAT 2 Piezoquartz strips QEX DAW 100 20/10/94 10 / 15 6 %
Kistler / GR 1 Piezoquartz strip QEX Marksman 660 20/10/94 1 / 14 0 %
Atochem 1 Piezopolymer strip - 16/7/93 5 / 31 4 %
Focas-Vibetek 2 Piezopolymer strips - 16/7/93 30 / 30 9 %
Kistler 1 Piezoquartz strip QEX - 16/7/93 31 / 31 100 %
Thermocoax 1 Piezo-ceramic Vibracoax - 16/7/93 30 / 30 27 %
TO = time of operation (outside breakdowns), TP = time of presence in the test.
% of vehicles = percentage of recorded vehicles, with respect to the maximum possible during TP.

Figure 1: Schematic view of the site at Zurich/Hagenholz

4. Results, and classification according to the European specification

The tested WIM systems are analysed using the method proposed in the European WIM Specification (draft 2.2, June 97) produced by the COST 323 Management Committee. It should be emphasised that this draft specification only appeared in mid-1996 after the end of the test, and therefore neither the organisers nor the participants were informed about it. It was not taken into account in the test plan nor in the calibration procedure. But it may nevertheless apply to the results. Only one among the four accuracy criteria proposed in the specification is applicable to this test, because only the gross weights was considered.

It should also be emphasised that the test was performed on an urban road with a low speed traffic which does not correspond to a highway traffic. The test site was just acceptable (class III) according to the specification.

According to the specification, the best WIM systems on such a site are expected to be only in class C(15). This classification only considers lorries (GW over 3.5 t). According to the test conditions (long term test), the environmental conditions are "full environment reproducibility (III)" and according to the number of weighed vehicles the test plan conditions are "full reproducibility (R2)". Therefore the minimum level of confidence required within centred confidence interval limits is 90% for about 150 lorries (the PAT system and PAT/Kistler combined system), and 92% for 1000 to 1500 lorries (the other systems). With these population sizes, the sample proportions within the tolerances may be considered to estimate the level of confidence. Table 3 gives the sample percentages of results within various centred confidence intervals for each system, and the accuracy class accepted (for gross weight only).

Table 3: Statistical accuracy of the systems compared with the static weights

Width of the confidence interval/class percentage of resultsin the confidence intervals
(centred on the static weight) Golden River Peek Traffic Mikros ECM PAT Kistler
± 5 % A(5) 21.0 % 25.1 % 42.7 % 34.8 % 71.0 % 37.2 %

± 10 %


38.0 % 45.5 % 73.7 % 61.1 % 86.5 % 67.9 %

± 15 %


53.5 % 59.4 % 87.7 % 75.9 % 93.5 % 89.1 %

± 20 %


68.4 % 68.9 % 95.3 % 82.7 % 98.7 % 97.8 %

± 25 %


79.2 % 76.0 % 97.9 % 86.3 % 98.7 % 98.5 %

± 30 %


86.2 % 81.8 % 99.0 % 90.8 % 100 % 99.3 %

± 35 %


88.9 % 87.3 % 99.5 % 94.0 % - 100 %

± 40 %


91.4 % 93.0 % 99.8 % 96.0 % - -

± 45 %


93.0 % 96.3 % 100 % 97.9 % - -
Accuracy class E(45) E(40) D+(20) E(35) C(15) D+(20)
The levels of confidence, which are underlined, are the smallest over the acceptance thresholds.

The main results of the test are:

Durability and reliability of the systems and sensors

Most of the systems have had some problems with their software and/or power supply. The Peek AWACS 6000 had some problems with the electronics and the power supply, the ECM Hestia station had some failures of its power supply and battery, and with the implementation of the automatic self-calibration procedure in the test conditions, and in the Mikros TEL-2CM some EPROM failed; some software failure also interrupted the measurements of the PAT DAW100 for many months. The Golden River Marksman 600 was replaced by a Marksman 660. Both combined systems (PAT and Golden River with Kistler) have had difficulties in reaching a working stage, because of the misadaptation of the software to the sensors. Most of these problems were solved by the manufacturers, but after long delays.

Some severe problems occurred with some sensors:

  • the Golden River capacitive strips all failed after 18 months;

  • the Mikros capacitive mat failed twice during the test period;

  • the Atochem piezopolymer strip failed after 3 months and was not replaced.

The other sensors behaved well throughout the test period (only the Peek piezoelectric strips induced some damage on the edge of the pavement at the end of the test period).


Among the four systems which took part in most of the test period (Golden River, Peek, Mikros and ECM), only one (Mikros) met the requirements of accuracy class D+(20). The other systems are far from being accepted in class D(25), being at least 7% too short considering the confidence level; nevertheless they may be classified into classes E(35) (ECM), E(40) (Peek Traffic) and E(45) (Golden River). With much less data (about 1/10), the PAT system just meets the accuracy level of class C(15), which is more or less the best class to be expected in such a pavement (class III), according to the European Specification COST 323. The combined system PAT/Kistler is comfortably in class D+(20).

Among the sensors, the piezoquartz strip by Kistler has very good performance, it provides results in the same accuracy range (but a bit more scattered) as the best bending plates; nevertheless these sensors are expected to be definitively more expensive than the other currently available strip sensors. The Vibracoax piezoceramic strip sensor provided results close to those of the ECM system (which uses similar sensors). The Vibetek piezopolymer strip is not adapted for WIM, with very high bias and scattering. It is difficult to appreciate the performance of the Atochem piezopolymer strip, because of its premature failure (the few data recorded seem to be good).

These results are not much surprising, even if they may be slightly disappointing for some manufacturers or users. On a class III site, it is not expected to achieve better than class C(15) accuracy with the current systems, even with large based sensors. The capacitive scale (Mikros) was affected by some sensor failures and electronic problem, that may explain its lower accuracy. The strip sensors are a bit more sensitive to some dynamic effects of the pavement/vehicle interaction which quickly increase with pavement roughness. But only the highest frequencies (app. 15 Hz) of the non suspended masses, which correspond to the shortest wavelengths of the signal on the road (1 m), are smoothed on the large scales; the low frequencies (from 0.5 to 4 Hz) which correspond to wavelengths between 3.75 and 30 m are not filtered by any WIM sensor. Therefore, less than half of the dynamic component may be eliminated by the large scales.

Conclusions and Recommendations

Extensive comparative measurements were carried out on a test site with various WIM systems and WIM sensors currently available on the commercial market. They are representative of most of the WIM technologies in use around the world.

The results from the Hagenholz test and the insights gained from this project were presented at the first European Conference on Weigh-in-Motion of Road Vehicles in March 1995. This European test was closed at the end of 1995.

The layout of the Hagenholz experimental site has proved to be useful in achieving the desired results. The layout allows an evaluation and a comparison between different WIM-systems and facilitates an objective assessment of the long-term behaviour of the different installations and systems.

Although the test site is only acceptable (class III) according to the European Specification currently drafted, especially with rather deep ruts and a slightly rough pavement, an accurate statement concerning the reproducibility and the accuracy of each WIM-system is possible. The values of the pavement evenness correspond to real conditions, such as found in potential WIM sites. However, it is true that this experimental track deals with only one road structure. Therefore, it would seem prudent to repeat the test and compare the results on different road structures.

The traffic composition was also typical of an urban road, with a limited number of vehicles types and speed range. 93.4% of the HGV's (over 3.5 t) were 2- to 4-axles rigid vehicles. Therefore additional tests would be desirable with other types of vehicles, especially tractor with semi-trailer and trailer, and an higher traffic flow, as well as at higher speed.

The results are as expected for WIM systems on such a site, even if we may expect better results with the classical strip sensors without the encountered problems with the electronics and software. The European Specification (COST 323) requires a good site (class II) to reach accuracy class B(10) and an excellent pavement (class I) to reach accuracy classes A(5) or B+(7), with currently marketed WIM systems. The differences between the four best systems using different sensor technologies (PAT, PAT/Kistler, Mikros and ECM) must be evaluated taking into account the approximate market costs: the value for money is likely to be rather constant for these four systems, if the sensor prices and installation costs are taken into account and the accuracy level taken as the criterion of quality. But for customers and users, some other criteria must also be considered, such as the reliability and durability, the user-friendliness, the supplier's service quality, etc.

The large amount of available data, collected over a long time period through various climatic conditions and seasons, provides a strong basis for classification of WIM-systems. But only the gross weight criterion may be checked using the recorded data, instead of the four criteria proposed in the Specification. It must also be noted that only a few types of lorries were weighed in this test, most of them being 2-, 3- and 4-axles rigid vehicles. These vehicles and small vans are characteristic of an urban road and traffic but are not really representative of most of the main European highways and motorways.

For that reason it was decided by the COST 323 Management Committee to carry out some additional tests on more heavily trafficked highways or motorways, with vehicle types representative of the main European roads. For such tests, a better pavement evenness, in class I or II, is chosen. Early 1997, a large scale long term test started on a busy motorway (A31, in Eastern France, between Metz and Nancy) linking North and South Europe. More than 4,500 lorries of all the types use this road per day and direction and it has a good pavement. This trial is part of the European Test Programme (1996-98), organised by the COST 323 Management Committee, and is called the Continental Motorway Test (CMT); a complementary test - Cold Environment Test (CET) -, which is more research oriented, started in Northern Sweden (Lulea) in June 1997, as part of the project 'WAVE' (WP3.1) - Weighing in motion of Axles and Vehicles for Europe - of the fourth framework programme (Transport).