COST 323 ”Weigh-in-Motion of Road Vehicles”

 

COST 323
”Weigh-in-Motion of Road Vehicles”

Final Report

APPENDIX 1

European WIM Specification

Version 3.0, August 1999

 

  1. FOREWORD

The COST 323 Management Committee and the contributors to this document have carefully collected the information and data published herein, in accordance with the latest scientific and technical principles. Nevertheless the editorial staff disclaim all liability on their part for any injury which may result from use of the data and information published herein.

This document does not constitute an official standard but provides a reference upon which standardisation committees can draw if they choose, and technical specifications for WIM us­ers and manufacturers. They may both refer to these specifications. The document specifies WIM systems in general, but does not specify products. In its final stage, this specification will constitute a pre-standard to be submitted to the CEN to assist in the preparation of an European Standard on WIM.

This document combines requirements or general clauses (numbered in bold), with some more informative explanations and examples, particularly in the field of statistics, to clarify or help with the implementation of the specifications. In order to distinguish them, the informative paragraphs are marked with a bar in the margin. They may be considered as parts of a “Hand­book on WIM”.

The contributions and remarks of the European WIM manufacturers were taken into account, and this specification was used for the evaluation of WIM systems by testing carried out by the COST 323 action.

These recommendations were developed such as to be widely independent on the technology and products (e.g. the type of sensor or electronics of the WIM system considered). They are expected to evolve with time, and future revisions may be done if needed.

The appendix I provides the simplified requirements of practical use for common users. Only the main clauses are presented; references are made to the detailed specification. It may be read before the detailed specification. That is recommended for the practitioners.

The scientific background used in this specification is presented in (B. Jacob, 1997).

Keywords

Traffic loads, Pavement conditions, Vehicle loads, Gross weight, Axle loads, Weigh-In­Motion, WIM Sensors, WIM Systems, Calibration, WIM data/system acceptance, Traffic data, WIM Specification, WIM Standard.

  1. CONTEXT, SCOPE AND OBJECTIVES

This document has been produced by the COST 323 Management Committee, as part of the COST Transport Action “WIM-LOAD”. It gives general and detailed recommendations for site selection, installation, operation, calibration and assessment by testing of WIM systems. It is based on COST 323 member countries and US experience ((NIST Handbook 44, 1995), (R. Gillmann, 1992), (TRL, 1994)), and existing national specifications (METT-LCPC, 1993) and (NWML, 1995)). However there are currently only a few specification documents and no offi­cial standard on WIM in Europe. Moreover, the existing US standard on WIM (ASTM, 1994) is mainly designed for model approval, or to indicate the potential upper limit of performance which can be achieved by the particular type of system as the road surface conditions shall be the best available for conducting the acceptance test. The main objective of this document is to cover the need for a complete specification, covering both aspects: (1) model approval and (2) on site acceptance test and accuracy assessment, pending the publication of an official European Standard produced by the CEN. It also provides a technical basis for such a stan­dard. Therefore this is a “pre-standardisation document”.

Even if in some situations, particularly for legal purposes, lorry weighing is currently limited to the use of static scales, in many European countries and for multiple applications, Weigh­In-Motion (WIM) systems are routinely or experimentally used. Therefore common specifica­tions are useful to check the real performance of WIM systems and to organise such trials . Moreover, the use of WIM systems for legal enforcement purposes is expected to become a main challenge in a near future, and will require a strong legal and standardised basis.

 

 

  1. TERMINOLOGY

The main terms used in this document are listed here. Some additional terms used in this document are defined in the Glossary of terms, containing a common multilingual terminol­ogy, and published by the COST 323 Management Committee (COST 323, 1998b). Some ad­ditional detailed definitions on the accuracy of WIM systems as well as mathematical and sta­tistical principles used, are presented in (B. Jacob, 1997).

In addition to the force of gravity, this force can include the dynamic effects of influences such as road surface roughness, vehicle acceleration, out-of-round tyres, dynamically unbalanced wheels or tyres, tyre inflation pressure, vehicle suspension and aerodynamic features and wind. For purposes of this specification, the WIM system shall be adjusted or calibrated to in­dicate the magnitude of the vertically downward, measured dynamic vehicle tyre force in units of mass (kilograms, kg or megagrams, Mg). The indicated mass can be converted to units of force by multiplying it by the local value of acceleration of free fall, if it is known.

The force of gravity – thus, the acceleration of free fall – is different at various locations on or near the surface of Earth; therefore, weighing devices in commercial use or in official use by government agencies for enforcement of traffic and highway laws or collecting statistical in­formation are usually used in one locality and are adjusted or calibrated to indicate mass at that locality. The indicated mass can be converted to weight (in units of force) by multiplying by the local value of acceleration of free fall, if it is known. For purposes of this specification, – and in accordance with common weighing practice – the WIM system shall be adjusted or calibrated to indicate the magnitude of estimated weight and load in units of mass (kilograms, kg or megagrams, Mg), and the direction of the associated force vector will always be down – wards toward the approximate centre of Earth.

Units

The SI recommends to express forces in N and kN (for large values), and masses in kg and Mg (for large values). 1 Mg=1000 kg (also a metric ton). The use of ton is depreciated in sci­entific use, while still mostly used by engineers, police, lawyers, road authorities and in most of the laws.

Because this specification is mainly for practical use, only the mass units will be used, either kg pr ton (1t=1Mg). When forces are considered, the ratio to the corresponding mass is 9.81 N/kg.

 

  1. USER AND PERFORMANCE REQUIREMENTS
    • The clauses of this document should be applied to specify and check the performance and accuracy of any WIM system in its environment. It contains definitions and criteria of accep­tance.
    • The WIM systems are classified in six accuracy classes, each of them corresponding to a range of applications or requirements. Additional classes are given for systems which do not meet the main classes.
    • The accuracy is mostly referred to the weights and static loads, i.e; for weighing purposes, and rarely to the real tyre impact forces applied by the wheels/axles on the pavement and on the WIM sensors, such as for technical studies on pavement and vehicles. The distinction must be clearly specified in writing, case by case. In the first alternative, it is recommended to spec­ify how the static loads and weights are obtained, and especially the issue of the static axle loads. In the second alternative, the means to obtain the reference values of the impact forces must be specified.

For practical reasons but also according to the most frequent requirement , reference to the static loads/weights may be assumed unless another reference value is specified.

Both of these references raise some difficult questions and issues, as mentioned in (B. Jacob, 1997).

The accuracy of a WIM system in its conditions of use, i.e., under moving traffic tyre loads, may only be defined in a statistical way (B. Jacob, 1997), by a confidence interval of the rela­tive error of a unit (an axle, an axle group or a gross weight), defined by: (Wd–Ws)/Ws, where Wd is the impact force or dynamic load measured by the WIM system and Ws the correspond­ing static load/weight (or any other specified reference value) of the same unit. Such a confi­dence interval centred on the static load/weight, is noted: [-6;+6 ], where 6 is the tolerance for a confidence level n (for example 90 or 95%).

Even for systems supporting the traditional definition of accuracy (OIML, 1996), weighing statically is not representative of real conditions of WIM system use.

  1. Statistics: Economical and technical studies of freight transport, general traffic evaluation on roads and bridges, collecting statistical data, etc..

6 up to of 20 to 30% (class D+(20), or D (25))

  1. Infrastructure and preselection: Detailed analysis of traffic, design and maintenance of roads and bridges, accurate classification of vehicles, preselection for enforcement, etc..

8 up to of 10 to 15 – 20% (class B (10), or C (15))

  1. Legal purposes: Enforcement and industrial applications, but only if the legislation allows the use of WIM for that purpose. Currently static weighing or LS-WIM are required for these applications; but some development is going on to increase the possibilities of HS-WIM for legal purposes.

8 up to of5 to 10% (class A (5), orB+ (7))

These figures are only given here as an indication; each user can define his own requirements for his particular application. Moreover the requirements depend on the environmental and road conditions. The chapter 8 specifies which figure apply to each entity (gross, axle, etc.).

Any level of accuracy not only refers to the performance of the WIM system used (i.e., the sensor(s) and electronic station with its software), but also to the calibration procedure and frequency,, to pavement/road quality and evenness and vehicle behaviour.

The confidence n in the accuracy level 8 (the confidence interval width) of a WIM system depends greatly on the conditions of measurement, that means principally the repeatability or reproducibility conditions of the sample measured, the environmental repeatability or repro­ducibility conditions and on the sample size and content (types of vehicles).

Different needs may lead to different accuracy requirements with respect to the weights. The following requirements are given unless otherwise stated by the customer:

Class A (5): legal purposes such as enforcement of legal weight limits and other particular needs; to provide reference weight values for in-service checks, if the classes B(10), C(15), D+(20) or D(25) are required for all the traffic flow vehicles (as­suming that it is not possible weigh in static such a large population);

Class B+ (7): enforcement of legal weight limits in particular cases, if the class A require­ments may not be satisfied, and with a special agreement of the legal authori­ties; efficient preselection of overloaded axles or vehicles; to provide refer­ence values for in-service checks, if the classes C(15), D+(20) or D(25) are required for all the traffic flow vehicles (assuming that it is not possible weigh in static such a large population);

Class B (10): Accurate knowledge of weights by axles or axle groups, and gross weights, for:

Classes C (15) or D+(20): Detailed statistical studies, determination of load histograms with class width of one or two tonnes, and accurate classification of vehicles based on the loads; infrastructure studies and fatigue as­sessments.

Class D (25): Weight indications required for statistical purposes, economical and technical studies, standard classification of vehicles according to wide weight classes (e.g. by 5 t).

Additional classes E(30), E(35), etc., are defined for WIM systems which do not meet the class D(25) requirements. These classes are specified in the chapter 8, to assess the accuracy of rough systems or of systems installed on poor WIM sites. However, they may be useful to give indications about the traffic composition and the load distribution and frequency.

 

  1. CRITERIA FOR THE CHOICE OF WIM-SITES

The WIM site characteristics have some influence on the in-motion vehicle behaviour and may lead to large discrepancies between the axle impact forces and the corresponding static loads. Therefore the specified criteria about the road geometry and the pavement characteris­tics are given in order to reduce these discrepancies and to keep them within some limits in accordance with the required accuracy levels.

The accuracy of a bridge WIM system also depends highly on the selection of the weighing site, particularly on the type of the superstructure and the evenness of the approach.

However these criteria, and above all those relating to the pavement profile, are mainly given as indicative, because only the specified WIM system performance (e.g. accuracy and durabil­ity) is mandatory. If some systems, as a result of their principal or intrinsic nature, may toler­ate weaker criteria and meet the accuracy and durability requirements – that should be proven by testing -, then they may be installed on other sites than those hereafter specified.

The pavement characteristics directly influence the signal recorded by any WIM sensor, be­cause of:

Thus not only the longitudinal evenness but also deterioration (such as rutting, deformation, etc.) limit the accuracy of the measurements, while cracking may reduce the WIM sensor du­rability or affect its response. The deflection and the transverse evenness may also affect the reliability and durability of the sensors.

Table 1: Classification and criteria of WIM sites

WIM site classes
I Excellent II

Good

 III

Acceptable
Rutting

(3 m – beam)

 Rut depth max. (mm) < 4 < 7 < 10
Semi-rigid Mean deflection (10-2 mm) < 15 < 20 < 30
Deflection Pavements Left/Right difference (10-2 mm) ± 3 ± 5 ± 10
(quasi-static) All bitumen Mean deflection (10-2 mm) < 20 < 35 < 50
(13 t – axle) Pavements Left/Right difference (10-2 mm) ± 4 ± 8 ± 12
Flexible Mean deflection (10-2 mm) < 30 < 50 < 75
Pavements Left/Right difference (10-2 mm) ± 7 ± 10 ± 15
Semi-rigid Deflection (10-2 mm) < 10 < 15 < 20
Deflection Pavements Left/Right difference (10-2 mm) ± 2 ± 4 ± 7
(dynamic) All bitumen Mean deflection (10-2 mm) < 15 < 25 < 35
(5 t – load) Pavements Left/Right difference (10-2 mm) ± 3 ± 6 ± 9
Flexible Mean Deflection (10-2 mm) < 20 < 35 < 55
Pavements Left/Right difference (10-2 mm) ± 5 ± 7 ± 10
Evenness IRI index Index (m/km) 0-1.3 1.3 -2.6 2.6-4
APL[I] Rating* (SW, MW, LW) 9-10 7-8 5-6

 

The rutting and deflection values are given for a temperature below or equal to 20°C and suitable drainage conditions.

* the rating quantifies the logarithm of the energy dissipated in one of the wavelength ranges: SW = Small Wave-lengths (0.7-2.8 m), MW = Medium Wavelengths (2.8-11.3 m), LW = Large Wavelengths (11.3-45.2 m). The scale is from 10 (lowest energy, excellent evenness) to 1 (highest energy, poorest pavement surface).

Comments about the deflection:

Finally, it should be recalled that the deflection affects the durability of the sensors, while the left/right difference may limit the accuracy of the measurements.

Comment about the evenness:

The measured evenness in terms of ratings at 200 m intervals is sufficient for screening sites; it is however necessary to consider more carefully the exact area of installation within the 200 m so as to avoid a single point having poor evenness:

Table 2: Choice of a WIM site according to the accuracy required

Accuracy site I (Excellent) site II (Good) site III (Acceptable)
Class A (5) + ^- –
Class B+ (7) + ^- –
Class B(10) + + –
Class C (15) (+) + +
Class D+ (20) (+) (+) +
Class D (25) (+) (+) +

 

legend: ‘-’ means insufficient, ‘+’ means sufficient, ‘(+)’ means sufficient but not necessary

Comment: This table does not give a strict relationship between the accuracy classes and the test site: some types of WIM systems – depending on the type of sensor and the measurement principle – may require higher or lower site classes to meet the same accuracy level. For example, large scales or large-based sensors (i.e. longer than the tyre imprint in the direction of the traffic flow) are less sensitive to the pave­ment evenness than are narrow-based sensors. Moreover multiple-sensor WIM systems may be installed in pavements with poorer evenness, if a suitable algo­rithm performs calculations to reduce the dynamic effects.

5.3 Particular Requirements for Bridges

Accuracy of the bridge WIM results is strongly related to the number of trucks (axles) which drive over those parts of the bridge which influence the structure at the same time (one truck at a time gives best results). Therefore the length of the structure and the traffic density have to be judged together (the more dense the traffic, the shorter is the optimal length of the struc­ture).

If the influence line is used in the weight assessment algorithm, an influence line based on ac­tual strain readings can improve the accuracy of calculation. This is particularly important

when a continuous bridge is instrumented. With this type of structure it is also essential that all the spans which considerably influence the behaviour of the instrumented span (where the strains of the superstructure are measured) are taken into account.

Table 3: Bridge selection criteria

Criteria Optimal Acceptable
bridge type steel girders, prestressed concrete girders, reinforced concrete girders, culvert, steel orthotropic decks (1) concrete slab
span length (2) (3)(m) 5 – 15 8-35
traffic density free traffic – no congestion (traffic jam)
evenness of the pavement before and on the bridge class I or II (Table 1) class III (Table 1)
skew (°) < 10 < 25

< 45(*)

 

(*) after inspection of calibration data

 

  1. ENVIRONMENTAL REQUIREMENTS

Most of the suppliers of WIM devices specify some environmental requirements for the use of their equipment. These requirements usually meet some existing standardised criteria, either for civil or military electronic devices. The following criteria are given to provide a common framework or to detail some requirements more specific to WIM sensors. They may be adapted by each customer with respect to the particular conditions of the WIM site chosen.

These requirements mainly concern the climatic conditions, but also deal with the traffic con­ditions and the facilities needed to install and operate the WIM systems.

For sensors which are supported by the pavement (such as strip sensors), the pavement modulus may have a strong influence on the sensor response; this is especially the case for bi­tuminous pavements. Bituminous pavement modulus varies by orders of magnitude with the temperature. Some indicative figures are given in Table 4.

Table 4: Variation of the pavement modulus (bituminous material) with temperature

Temperature – 15 °C 0 °C 15 °C 30 °C
Scale factor of the pavement modulus 10 8 5 1

 

This phenomenon may affect both the accuracy of the WIM system and the durability of the sensors. The system should take it into account.

 

  1. ON-SITE SYSTEM CHECKS AND CALIBRATION
    • General Recommendations
      • After installation and general checking, an initial calibration must be performed before an operational use of any WIM system. The accuracy of WIM data depends greatly on the calibration procedure of the WIM system.

A general statistical procedure for calibration and further checking of WIM systems, with re­spect to the statistical accuracy and classes is described in (B. Jacob, 1997).

It is important to note that a WIM system measures instantaneous impact forces, and only es­timates weights. WIM data deviations from weights could be considered both as measurement errors and as those resulting from dynamic effects.

These “true impact forces” are generally not easy to measure accurately with a perfect syn­chronisation with the WIM; however, some techniques were developed, using either shock or pressure devices (see 7.2.2), or instrumented vehicles (see 7.2.4).

The calibration is assumed to be over during a short time period, such as one or two consecu­tive days, except for automatic self-calibration (see 7.2.5).

Different calibration methods are commonly used, which depend on the sensor type, the appli­cation and requirements of the user and the time and means available.

The sensors which may be calibrated in static are: strain gauge and load cell scales, piezo­quartz crystal bars, capacitive strips or fibre optic sensors, but not piezo-ceramic or piezo­polymer cables. Even for the strip sensors (piezo-quartz, capacitive strips and fibre optic), the static calibration is not easy to perform because of the small area of the sensor (and thus the difficulty to apply a mass of several tons), and the loading condition differs from that under traffic flow, because the integration of the signal may not be performed during a static test.

This calibration method is especially convenient if the weight is to be estimated with low speed WIM systems on excellent pavement sites.

three axles must be used with static loads uniformly distributed within the scale range of the loads to be weighed, and three repetitions for each axle weighing shall be done.

The principle of such calibration methods is to apply to the sensor some repeatable calibrated shocks or pressure variations. It may be done using a DYNAPLAQUE, a FWD (Falling Weight Deflectometer), a Piezodyn (M. Huhtala and B. Jacob, 1995), or any other similar de­vices.

The advantage of such a method is that it is almost independent of the pavement profile and of the calibration vehicle characteristics and speed or load (see 7.2.3). However the tests per­formed have shown that most of the devices used give results scattered along a WIM sensor, not only because of an eventual heterogeneity of the sensor itself, but also because of the im­pact conditions around the sensor. Moreover, the impact conditions are very different from a tyre imprint and the force applied by an instantaneous vertical force. This method also re­quires the closure of the traffic lane during the calibration, which may be difficult for busy highways or motorways.

This method is mainly devoted to calibration with respect to impact forces, but not to the weights. It could be of interest if the WIM system is used for impact force measurements (7.1.3.2) , as in (M. Huhtala and B. Jacob, 1995), but until now this method has not yet been proven to be effective.

It is the most commonly used method because of its relative simplicity and directness, and be­cause it is suitable for all kinds of WIM systems. This method partially eliminates the repeat­able pavement dynamic effects (bias), but is sensitive to the calibration (test) vehicle charac­teristics, such as suspension type and parameters, dry friction, etc..

(r1) full repeatability conditions: if only one vehicle passes several times at the same speed, the same load and the same lateral position;

(r2) extended repeatability conditions: if only one vehicle passes several times at different speeds (according to the traffic lane conditions), different loads (e.g. fully loaded, half-loaded and empty), and with small lateral position variations (according to the real traffic paths);

(R1) limited reproducibility conditions: if a small set of vehicles (typically 2 to 10), represen­tative of the whole traffic composition expected on the site (silhouettes and gross weights), is used, each of them passing several times, at different speeds, different loads, and with small lateral position variations;

(R2) full reproducibility conditions: if a large sample of vehicles (i.e. some tens to a few hun­dred) taken from the traffic flow and representative of it, pass on the WIM system and are statically weighed before or after it.

If possible, the last two vehicles will be used fully loaded and half loaded. The tandem or tridem axles should be better equipped with air suspension. However, if mechanical suspen­sions are used to be representative of the common vehicles on the site, some care should be taken to measure the static reference axle loads (see 8.3).

It is also recommended to use one of the standard test plans described in the Appendix I, as they were designed to optimise the number of runs and vehicles versus the confidence level. Moreover, that would allow to use the graphs given in this Appendix I to facilitate the initial verification, after calibration.

The conditions must be specified before the calibration, and the results (in terms of accuracy class) must be analysed according to them (see 11.), for the level of confidence being used.

The higher the conditions (from (r 1) to (R2)) the more representative the calibration sample of the real traffic conditions, but the procedure becomes longer and more costly! Nevertheless this calibration procedure may be performed without traffic stopping ((r 1) to (R1)).

must be of the most common vehicle type to be weighed, and with three loading cases: empty, half loaded and fully loaded.

In such a case, the methods described in 7.2.3 (and in Appendix III) introduce some bias by partially eliminating the dynamic effects being sought. This is the case for some research pur­poses such as spatial repeatability investigations or pavement/vehicle interaction and pave­ment damage studies. For multiple-sensor WIM systems, the spatial repeatability is used to improve the accuracy of the static load estimator.

The advantage of this method is to make a “true” calibration on the parameter actually meas­ured by a WIM system, i.e. the wheel or axle impact force. Its disadvantage comes from the cost and difficulties of getting and operating such instrumented lorries, which also require

specialised technicians. Also there are only very few such instrumented vehicles available ac­tually, and the information and documentation about them are very poor.

The quality of the calibration greatly depends on the accuracy of the lorry instrumentation, which measures continuously each wheel impact force on the pavement as the vehicle travels. But these measurements are indirect, by the mean of accelerations and strain records, and gen­erally require a lot of computation afterwards.

This kind of method, introduced in France in the early 1980’s, has the great advantage of pro­viding a permanent automatic recalibration of a WIM system installed on a trafficked road, and therefore to correct any trend or bias due to sensor, electronics or pavement changes or due to external effects, such as temperature variations. However, it was shown that this proce­dure requires a prior knowledge of the traffic pattern and may be worst than nothing in some particular circumstances.

In most countries and road networks, there are some “characteristic vehicles” which have some axle(s) and/or gross weight with a low coefficient of variation and a quite constant mean (the target value). In such a case, the moving average of a certain number of these axle loads or gross weights becomes almost constant for a large enough sample size, and may be fitted to the target value. This provides a new coefficient of calibration after the passage of the required number of characteristic vehicles.

Nevertheless it must be noted that such a procedure introduces a statistical error due to the sample size of the considered “characteristic vehicles”. Therefore the time interval between two recalibrations (calculation of a new calibration coefficient) must be a compromise be­tween the reduction in statistical variance (by increasing the sample size) and the delay in re­calibration. If the temperature influence is to be eliminated, it is recommended to have such an interval in the range of 1 hour to a few hours. If only some long-term trends are to be elimi­nated, this time interval may be longer (e.g., 1 day to a few weeks).

Finally it should be noted that even if this type of calibration is very easy and inexpensive to implement after performing the appropriate preliminary studies and after the development of the proper software, it may also introduce some uncontrolled bias or variance.

 

  1. ACCURACY CLASS TOLERANCES WITH RESPECT TO THE WEIGHT
    • General
      • A WIM system must be checked following a well defined procedure or test programme and can then be classified into one of several accuracy classes according to the test results. These accuracy classes are defined with respect to the weight estimation; but in some particu­lar cases another reference may be adopted, such as independently measured impact forces.
      • The principle adopted for this classification consists of fixing the tolerance 6, i.e. the width of the confidence interval for an individual WIM measurement, and for a given level of confidence The requirement for this level depends on the calibration or test conditions, if this is performed using pre-weighed vehicles in motion (as described in section 7.2.3). For static calibration with calibrated masses or pre-weighed vehicles, this level of confidence must be 100 %.
      • In the statistical approach, adapted to most of the existing WIM systems, any individual measurement (of an axle, axle group or gross weight) must have a probability n higher than a required value n0 of being within the interval [Ws(1-6);Ws(1+6)] centred on the static load, where Ws is the corresponding static load. It also means that statistically a proportion n of a large sample of WIM data should be within the previous interval. Or the customer risk on an individual measurement is lower than (1-n0 ) under some specified conditions. The mathe­matical and statistical background is developed in (B. Jacob, 1997).

The principle used is that the tolerance 6 only depends on the accuracy class and on the entity considered, which may be the:

and additionally:

One criterion is considered for each of these entities.

The level of confidence n of any sample of data only depends on the test conditions ((r 1) to (R2)), on the environmental test conditions ((I) to (III) (see 11.1.4)) and on the sample size (number of runs and of test vehicles), and must be higher than a specified value n0 which also depends on the test conditions and sample size.

The test plan may depend on the WIM system type, accuracy class required and application.

This specification only considers individual measurements, while it is almost impossible to assess the accuracy of a WIM system using only aggregated data. If some WIM systems de­liver only statistics during operational period of use, detailed data should be provided for calibration and accuracy tests.

Table 5: Tolerances of the accuracy classes (5 in %)

Criteria (type of measurement) Domain of use Accuracy Classes:

Confidence interval width 8 (%)
A (5) B+(7) B (10) C (15) D+(20) D (25) E
1. Gross weight Gross weight > 3.5 t 5 7 10 15 20 25 > 25
Axle load: Axle load > 11
2. group of axles 7 10 13 18 23 28 > 28
3. single axle 8 11 15 20 25 30 > 30
4. axle of a group 10 14 20 25 30 35 > 35
Speed V > 30 km/h(1) 2 3 4 6 8 10 > 10
Inter-axle distance 2 3 4 6 8 10 > 10
Total flow 1 1 1 3 4 5 > 5

(1) This condition applies only for the sensors/systems which do not work statically or at very low speed.

The class designation by numbers 8c = 5, 7, 10, 15, 20, 25, and so on (tolerances for the gross weights) is in agreement with the OIML recommendation, and allows for use of any classes before A(5) or interpolated classes between the specified classes (e.g., class(13)).

Table 6: Tolerances of the accuracy classes E

Criteria (type of measurement) Accuracy Classes E Confidence interval width 8 (%)
E(30) E(35) E(40) E(45) E(50) etc.
1. Gross weight 30 35 40 45 50 …
2. Group of axles 33 39 44 49 55 …
3. Single axle 36 42 48 54 60 …
4. Axle of a group 41 47 53 59 65 …

 

Figure 1: Graphical representation of the accuracy class tolerances

 

trapolated by:

Group of Axes (GA)                   8=1.0467 Sc + 2.1556                For Sc > 50

Single Axles (SA)                       8=1.1333 Sc + 2.6667                For Sc > 50

Axles of a Group (AoG)             8=1.1333 Sc + 7.6667                For Sc > 50

Where 8c is the tolerance for the gross weight, and the accuracy class name is E(8c). These values must be incremented by steps of 5%. The 8 values obtained with the above formula must be rounded up/down to the closest integer.

Group of Axles (GA): 5 = 5c / 0.7 For 5c <7
5 = 5c + 3 For 7 < 5c <30
5 = 1.2 5c -3 For 30 < 5c < 35
5 = 5c +4 For 35 <5c<50
Single Axles (SA): 5 = 5c (85 – 5c) / 50 For 5c <10
5 = 5c +5 For 10 <5c <25
5 = 1.2 5c For 25 <5c<50
Axles of a Group (AoG): 5 =25c For 5c <10
5 = 5c +10 For 10 <5c <25
5 = 1.2 5c +5 For 25 <5c<50

 

If the reference values used for calibration or accuracy assessment are weights and static loads, the following rules and clauses should be applied.

Static axle loads should be measured by axle or wheel scales, which are approved for en­forcement and commercial applications, either mounted in grooves and carefully levelled to the road surface, or laid on the road surface. The road surface on the weighing area should be flat and horizontal. In the latter case, it is recommended to:

The level difference between axles of a same group should not exceed 2 mm. The level differ­ence between single axles or groups of axles should not lead to more than 0.5% slope (i.e. 1.5 cm for 3m spacing).

n

Ws. =                           £ Ws. .                                                     (1)

i q n                                       i, j

£ £ Ws j = 1

i=1j=1i,j

where i is the axle rank, q is the number of axles of the vehicle, Ws is the reference gross weight measured on a weigh-bridge, and Wsi,j is the measured load of axle i during the jth weighing.

It s recommended to take n=10, but any value may be accepted. Even for n=1, it is recom­mended to use this equation to get the axle reference static loads, if the gross weight was measured on a weigh-bridge.

If n is large enough (i.e. n > 8 to 10), it is recommended to eliminate the weighings which could have provided statistical outliers, identified by any statistical test.

 

  1. TYPE (MODEL) APPROVAL OF A WIM SYSTEM

A type (or model) approval is a complete standardised procedure to be applied once to any newly manufactured measuring system, before to market it, in order to deliver a quality label and some target performance under known conditions of use. This chapter only deals with the on-site accuracy assessment of a WIM system by testing, as part of a type approval procedure to be developed in the future. The same approach and tools as for initial or in service verifica­tions and acceptance tests (see chapter 10 and 11) are used. However, the site characteristics and the test plan are fully described in this chapter, while they are left to the user’s decision in chapter 11.

Before to be marketed with a quality label and a specified accuracy performance, any WIM system should pass the test procedure described in this chapter. The test must be organised under the responsibility of an official agreed organisation, to ensure the neutrality and the reli­ability of the conclusions. An official report should be written and published giving an account of the test results. The Appendix IV gives some indication about the result format and presen­tation.

The type approval test intends to assess the accuracy performances of a WIM system under fully specified conditions, and over a short time period. Therefore, it does not give any infor­mation about the durability or trend of the system and its parts, which are highly dependent on the environmental and traffic conditions.

The site conditions are chosen as representative of the best quality site for WIM, in order not to introduce too much site effect. Therefore, the real performance on common sites may sig­nificantly differ (being below) from those assessed through the type approval.

For a high-speed WIM system, Vm will be taken equal to 75km/h. For a low-speed WIM sys­tem, Vm will be the recommended operation speed.

on the sensor(s) (speed variation from 90 km/h to 60 km/h, or 12 to 5 km/h for a low­speed WIM system).

For both pre-calibration (see 9.3) and the test (see 9.4), the test vehicles will be weighed on an approved weigh-bridge and on wheel/axle scales. The rules of section 8.3 will be applied, with n > 6 (see 8.3.3). It will be checked that the standard deviations of the static axle loads are less than 1/3 of those measured in motion.

  1. INITIAL AND IN-SERVICE VERIFICATIONS

A verification of a WIM system may be done either:

or

In most cases (unless if an automatic self-calibration procedure is used) the calibration proce­dure provides data which may be used for an accuracy evaluation. In such a case, the same sample is used for calibration and for accuracy assessment. This is an initial verification.

This section only applies to bending plates, with strain gauge or load cell scales, and for weigh-bridges, which are able to measure static loads. In particular cases it may also be appli­cable to some strip sensors, and may be extended with caution to sensors calibrated with shock devices.

The required level of confidence of this interval [- k.6; k.6] is given in chapter 11.

 

In such a verification, the data used for the accuracy assessment must not have been used for any calibration or recalibration of the system.

The required level of confidence of this interval [- 8; 8] is given in chapter 10.

Such a test made under moving vehicles, such as in normal traffic conditions is often more re­alistic than the check mentioned in 10.2.2.

 

  1. PROCEDURE TO CHECK THE ACCURACY
    • General Rules
      • The assessment of the accuracy of a WIM system requires a test. This chapter deals with tests carried out using either repeated runs of pre-weighed vehicles (test vehicles), and/or the use of single runs of pre- or post-weighed vehicles from the traffic flow.
      • The more extensive the test plan means the longer the test period, a higher number of vehicle types and runs and ultimately a higher confidence in the conclusion. This means that the customer risk (i.e., the risk of accepting a system in a higher class than it is) decreases as the test becomes more extensive. In this analysis, the supplier risk, linked to the statistical es­timation of the mean bias, is fixed at 5%.
      • In this procedure, the customer risk is governed by the probability of an individual er­ror (with respect to the static load or weight) lying outside of the specified confidence interval (tolerance). An upper bound of this risk is fixed by specified values (1-n0), where n0 is the minimum required confidence level. This risk (1-n0), or the confidence level n0, may be cho­sen by the customer (see section 11.3).

Lower this risk, longer and more extensive (and expensive) the test. Then the customer should adapt it to its requirements, taking into account the manufacturer specification and the output of other extensive and detailed tests.

It should be emphasised that this risk is only assessed under the conditions of the acceptance test; it means that the farther the test conditions from the real traffic conditions, the lower the confidence and higher the customer risk.

a day or spread over a few consecutive days, such that the temperature, climatic and environmental conditions do not vary significantly during the measurements;

ture, climatic and environmental conditions vary during the measurements and all the site seasonal conditions are encountered.

If both types of vehicles are used, the data of each population should not be merged in the analysis.

Depending on the sensor type, temperature variations can cause bias because of sensor sensi­tivity or indirectly because of pavement modulus or behaviour changes.

For common checks, some standard simplified test plans are given in the Appendix I.

The mean error estimation is affected by a statistical uncertainty, which depends on the sample size n (the uncertainty is removed for an infinite sample size !). This uncertainty is taken into account in the specified values of the following tables and in the formulas of section 11.4, as­suming that the samples have normal distributions (this may be checked by testing if re­quired).

Table 7: Minimum levels of confidence n0 , of the centred confidence intervals (in %) – case of a test under “environmental repeatability” (I)

^^^^^^Sample size (n) Test conditions^^^^^^^^ 10 20 30 60 120 TO
Full repeatability (r1) 95 97.2 97.9 98.4 98.7 99.2
Extended repeatability (r2) 90 94.1 95.3 96.4 97.1 98.2
Limited reproducibility (R1) 85 90.8 92.5 94.2 95.2 97.0
Full reproducibility (R2) 80 87.4 89.6 91.8 93.1 95.4

 

For sample size n not mentioned in this table, the figures may be interpolated using Figure 2, or a linear interpolation, or they are calculated in the Excel sheet presented in the Appendix IV.

Table 8: Minimum levels of confidence no , of the centred confidence intervals (in %) – case of a test under “limited environmental reproducibility” (II)

^^^^^^Sample size (n) Test conditions^^^^^^^^ 10 20 30 60 120 TO
Full repeatability (r1) 93.3 96.2 97.0 97.8 98.2 98.9
Extended repeatability (r2) 87.5 92.5 93.9 95.3 96.1 97.5
Limited reproducibility (R1) 81.9 88.7 90.7 92.7 93.9 96.0
Full reproducibility (R2) 76.6 84.9 87.4 90.0 91.5 94.3

 

For sample size n not mentioned in this table, the figures may be interpolated using Figure 2, or a linear interpolation, or they are calculated in the Excel sheet presented in the Appendix IV.

Table 9: Minimum levels of confidence n0 , of the centred confidence intervals (in %) – case of a test under “limited environmental reproducibility” (II)

^^^^^^Sample size (n) Test conditions^^^^^^^^ 10 20 30 60 120 TO
Full repeatability (r1) 91.4 95.0 96.0 97.0 97.6 98.5
Extended repeatability (r2) 84.7 90.7 92.4 94.1 95.1 96.8
Limited reproducibility (R1) 78.6 86.4 88.7 91.1 92.5 95.0
Full reproducibility (R2) 73.0 82.3 85.1 88.1 89.8 93.1

 

For sample size n not mentioned in this table, the figures may be interpolated using Figure 2, or a linear interpolation, or they are calculated in the Excel sheet presented in the Appendix IV.

Figure 2: Graphical representation of the minimum confidence level with the number of data

 

After the end of the data collection, the detailed analysis of the test results will be done through the following steps:

In this step, the percentage of missing vehicles (not including the vehicles recorded with an er­ror code) must be lower than the values indicated in the last line of Table 5.

The percentage of vehicles recorded with an error code may be higher (without any specified upper limit), but only in so far as it concerns the traffic conditions: vehicle passing partially off-scale, braking or accelerating over the specified limits of the system, etc..

(Wdi–Wsi)

X- =                   

xi          Wsi

where Wdi and Wsi are the in-motion measured value and the reference (static) value respec­tively of the same entity.

Then the mean m and the standard deviation s of the relative errors in each sub-population sample are calculated.

Remarks:

  1. While the calibration method applied does not provide individual coefficients by lorry type or axle rank (such as in the methods 1.a. to 1.d. of the Appendix III), the samples considered

must include all the gross weight or single axle, axle of group or group of axles loads results together. For example all the single axles of any rank must be considered together, but not the front axles and the rear/drive axles separately.

If different calibration coefficients are defined by vehicle type or axle rank (or type), then the samples considered may distinguish each sub-population.

  1. In case of a test in conditions (r1), the data collected for all the speed levels must be merged and analysed in only one sample, even if the full repeatability is not satisfied anymore.
    • Calculation of the confidence level

The confidence level n may be either estimated by a theoretical method (11.4.6.1) using the sample statistics of the test, or, in some cases, by a sample proportion (11.4.6.2). Both meth­ods are presented:

A lower bound n, of the probability for an individual value of a relative error, taken randomly from a normally distributed sample of size n, with a sample mean m and standard deviation s, to be in the centred confidence interval [-5; 8 ], is given at the confidence level (1-«) by (B. Jacob, 1997):

n = O(u1 )-O(u2), with u1=(8 -m) Is – tv,i-a/2 In112 and u2=(-8 -m) Is + tv,1–0/2 In112                      (2)

where O is the cumulative distribution function of a Student variable, and tv,1–o/2 is a Student variable with v = n-1 degrees of freedom. a is taken equal to 0.05.

Remark: If n is greater than 60, the cumulative distribution function O may be approximated by the cumulative distribution function of a standardised Normal variable. But this approximation is not of a very high interest in practice, and should only be used if the Student distribution function is not available.

Then the estimated level of confidence n, for each sample (and criterion) is calculated.

If the sample size n is greater than 10/(1-n0), where no is the minimum required level of confi­dence read in Table 7, Table 8 and Table 9 (according to the test plan), n may be statistically estimated by the proportion n‘ of the sample test data found within the confidence interval [- 8;+8].

This estimation may be eventually used while n > 5/(1-n0), but the statistical uncertainty in­creases as n decreases.

The sample proportion may only be used with the user’s or customer’s agreement, and if there is no possibility to calculate the n value.

At this stage, they are two ways to assess the accuracy level of a WIM system by testing:

If the sample proportion n‘ is used (11.4.6.2), the smallest value 6min of 6 which ensures that the centred confidence interval contains a sample proportion n‘ = no, is chosen, and the same check as above is done.

This last approach may allow to classify a system in any accuracy class, defined by the lowest accepted 6-value (6min).

  1. DATA STORAGE, PROCESSING AND TRANSMISSION

It is out of the scope of this document to specify in too much detail the content, structure and format, of the data files containing the output from WIM systems. It is mainly the responsibil­ity of the WIM system manufacturers or service suppliers to develop and implement software and data files, adapted to the requirements of each type of customer and user. Moreover, an excessively detailed specification could limit the progress and evolution in this domain, and prevent adaptation to the most advanced WIM technology.

These general guidelines are given to ensure user-friendless and facilitate the exchange of data between users. Some of these requirements may evolve with the WIM technology.

 

It is highly recommended to record and deliver the time of passage in hh:mm:ss:cc, up to hun­dreds of second, because at current highway speed (e.g., 20 m/s) the inaccuracy on the vehicle spacing may be too high for many applications if this time is rounded up to the second.

In both cases, the criteria for wrong result detection must be clearly indicated not only in the technical brochure of the WIM system, but also in any document presenting the data.

and on the right side (last columns) the data which only concerns some vehicles:

In such a way, the size of the files may be reduced, avoiding having many partially empty col­umns for the smallest vehicles (only the carriage return symbol – end of line – will be mixed with other data in the same column). If this principle is not applied, the number of columns must be the largest to be used for the longest vehicles.

Appendix IV gives an example of a standard detailed data file, designed for the accuracy as­sessment of a WIM system according to this specification.

  1. COST 323 VEHICLE CLASSIFICATION (NOT mandatory)

There are many vehicle classifications in a few or large numbers of vehicle categories used in Europe. It is not the scope of this specification to require a unique classification, while de­pending on the application, the regional traffic patterns, etc., one or the other may be better adapted. However, in order to facilitate some comparison between general patterns from one road to another, or to analyse in details the performance of WIM systems with respect of the type of vehicle to be weighed, a simple classification was agreed.

13.2. The COST 323 classification is given by:

Category Silhouette Description
Category 1 Cars, vans (< 35 kN) Cars, cars+light trailers or caravans
Category 2 Two axle rigid lorry
Category 3 “ö——– oo More than 2-axle rigid lorry
Category 4 —o o .                           °° Tractor with semi-trailer supported by single or tandem axles
Category 5 ‘VW ^5 ov w©

o o

 Tractor with semi-trailer supported by tridem axles
Category 6 —t>——– OO——– O “O———– ÖO——– ÖÖ Lorry with trailer
Category 7 ‘LJTJkJTl^ ’o O— Busses
Category 8 Other vehicles

Figure 3: COST 323 vehicle classification

APPENDIX I. SIMPLIFIED REQUIREMENTS

I-1 Criteria for the Choice of WIM Sites

The choice of a WIM site has a great influence on the accuracy, the reliability and the durabil­ity of any WIM system. Therefore sites are classified according to the road geometry and the pavement characteristics. Table 10 indicates the recommended choice of site depending on the required accuracy level. The widths of the accuracy classes are given in I-6.

Table 10: Choice of WIM site according to the accuracy required

Accuracy site I (Excellent) site II (Good) site III (Acceptable)
Class A (5) + ^— –
Class B+ (7) + ^— –
Class B(10) + + –
Class C(15) (+) + +
Class D+ (20) (+) (+) +
Class D (25) (+) (+) +

 

legend: ‘-’ means insufficient, ‘+’ means sufficient, ‘(+)’ means sufficient but not necessary

Comment: This table does not give a strict relationship between the accuracy classes and the test site: some types of WIM systems – depending on the type of sensor and the measurement principle – may require higher or lower site classes to meet the same accuracy level. For ex­ample, large scales or large-based sensors (i.e. longer than the tyre imprint in the direction of the traffic flow) are less sensitive to the pavement evenness than are narrow-based sensors. Moreover multiple-sensor WIM systems may be installed in pavements with poorer evenness, if a suitable algorithm performs calculations to reduce the dynamic effects.

The requirements for bridge WIM systems are given in the chapter 4.3.

 

I-1.1 Road Geometry

I-1.1.1. It is strongly recommended that road section between 50 m upstream and 25 m down­stream of the system meets the following geometrical characteristics:

I-1.1.2. The WIM systems should be installed away from any area of acceleration or decelera­tion, (i.e. close to a traffic light, toll station, etc.), in order to weigh vehicles travelling at uni­form speed. It is also desirable to avoid the area where drivers make gear changes, such as slip-roads, etc.

I-1.1.3. It is also desirable to avoid areas where the number of lanes changes as this can lead to vehicles changing lane at the site.

I-1.2 Pavement Characteristics

I-1.2.1. The pavements should meet the following criteria:

Table 11: Classification and criteria of WIM sites

WIM site classes
I Excellent II

Good

 III

Acceptable
Rutting

(3 m – beam)

 Rut depth max. (mm) < 4 < 7 < 10
Semi-rigid Mean deflection (10-2 mm) < 15 < 20 < 30
Deflection Pavements Left/Right difference (10-2 mm) ± 3 ± 5 ± 10
(quasi-static) All bitumen Mean deflection (10-2 mm) < 20 < 35 < 50
Pavements Left/Right difference (10-2 mm) ± 4 ± 8 ± 12
(13 t – axle) Flexible Mean deflection (10-2 mm) < 30 < 50 < 75
Pavements Left/Right difference (10-2 mm) ± 7 ± 10 ± 15
Semi-rigid Deflection (10-2 mm) < 10 < 15 < 20
Deflection Pavements Left/Right difference (10-2 mm) ± 2 ± 4 ± 7
(dynamic) All bitumen Mean deflection (10-2 mm) < 15 < 25 < 35
Pavements Left/Right difference (10-2 mm) ± 3 ± 6 ± 9
(5 t – load) Flexible Mean Deflection (10-2 mm) < 20 < 35 < 55
Pavements Left/Right difference (10-2 mm) ± 5 ± 7 ± 10
Evenness IRIindex Index (m/km) 0-1.3 1.3 -2.6 2.6-4
APL(1) Rating* (SW, MW, LW) 9-10 7-8 5-6

The rutting and deflection values are given for a temperature below or equal to 20°C and suit­able drainage conditions.

* The rating quantifies the logarithm of the energy dissipated in one of the wavelength ranges: SW =Small Wavelengths (0.7-2.8 m), MW =Medium Wavelengths (2.8-11.3 m), LW = Large Wavelengths (11.3-45.2 m). The scale is from 10 (lowest energy, excellent evenness) to 1 (highest energy, poorest pavement surface).

Comments about deflection and evenness are given in the detailed specification, section 5.2.1.

I-2 Environmental Requirements

I-2.1. For a standard check, sufficient in a common application, it is recommended to compare the conditions of use given and guaranteed by the manufacturer, and the detailed specifica­tions about climatic conditions, traffic conditions and mechanical resistance given in the main document, sections 6.1 and 6.2.

I-2.2. It is recommended that the site should have power supply and communication facilities; the cost of power installation can be high. The preferable site should have a static weighing

area or a static scale close to the WIM site, and a reasonable time for a calibration or test vehi­cle to perform a complete loop across the WIM site.

I-2.3. For safety reasons it is recommended that there should be sufficient space and conven­ient access to install a road side cabinet. A cabinet should protect the WIM system against climatic actions and vandalism.

I-2.4. It is recommended that WIM systems should not be installed under high voltage power lines.

I-2.5. It is important to avoid any overpass (aerodynamic effects) or bridge approach (poor evenness); it is also not recommended to install road sensors on a bridge or on any structure subject to dynamic effects.

I-3    On-Site System Checks and Calibration

The following three cases should be distinguished:

I-3.1 The same Company Supplies the Sensors, Electronics and Installation

In this case, no particular checks are required; the general guarantee of the supplier is suffi­cient.

I-3.2 A Manufacturer Supplies Sensors and Electronics, and Another installs them

In this case, it is necessary to carry out a quick acceptance test of the sensors and electronics before installation; such a test can be done either by the client or by the installer, according to the rules specified by the contract and following the recommendations given in sections II.2 and II.3 of the appendix II of the main document.

I-3.3 Different Suppliers Provide the Sensors, the Electronics and the Installation

The client should first check the compatibility between the sensors and the electronics, or ask the electronics manufacturer or the installer to do that. Then the rules in I-3.2. should be ap­plied.

In all cases the manufacturer’s specifications or guidelines for sensor installation should be applied. Summarised specifications are given in the main document, chapter 7.

I-4    On-Site System Calibration

I-4.1. After installation, aWIM system must be calibrated (initial calibration).

I-4.2. The procedure given by the manufacturer should be applied. Its compliance with one of the detailed procedures described in the main document, chapter 8 should be checked.

If the manufacturer or supplier does not propose such a procedure, or in case of a new system or a combined system supplied by several companies (I-3.3), one of the calibration methods given in the detailed specification, sections 7.2 and Appendix II should be applied.

I-4.3. An initial verification of the installed system should be performed to assess its accuracy before acceptance. The procedure is described in chapters 10 and 11.

I-4.4. Periodical calibration and system behaviour checks should be performed, at least once per year (level 2), but preferably every month if possible (level 1).

Level 1 check

An easy and economical calibration and system behaviour check may be performed either:

and/or

In case of incoherence or obvious wrong results, a complete accuracy check and eventually a recalibration should be performed (see chapters 10 and 11).

Level 2 check

It is recommended than an in-service test be performed periodically (e.g., once per year), fol­lowing the rules given in chapters 10 and 11. If a significant bias is found, a recalibration should be carried out.

I-5.1 Accuracy Class Tolerances

Several accuracy classes for individual measurements have been defined (see Table 10). Four main criteria are considered (depending on the system, some or all of them must be consid­ered). These classes are defined by the confidence intervals of the relative errors with respect to the static loads or weights (or in some particular cases with respect to a specified accepted reference).

Some indications about the customer requirements and the classes recommended for each type of applications are given in the detailed specification, chapter 4.

Additional classes E(30) to E(50) are defined in the detailed specification, chapter 8, for sys­tems which do not comply with class D(25).

Table 12: Width of the accuracy classes (confidence interval, 5 in %)

Criteria (type of measurement) Domain of use Accuracy Classes: Confidence interval width S (%)
A(5) B+(7) B(10) C(15) D+(20) D(25) E
1. Gross weight Gross weight > 3.5 t 5 7 10 15 20 25 > 25
Axle load: Axle load > 11
2. group of axles 7 10 13 18 23 28 > 28
3. single axle 8 11 15 20 25 30 > 30
4.axle of a group 10 14 20 25 30 35 > 35
Speed V > 30 km/h(1) 2 3 4 6 10 10 > 10
Inter-axle distance 2 3 4 6 10 10 > 10
Total flow 1 1 1 3 5 5 > 5

(1) For sensors which do not work statically or at very low speed.

 

The criteria on speed, axle spacing and counting are not mandatory in this specification. The accuracy class accepted for any WIM system is the best class for which all the criteria are sat­isfied, or the relevant criteria if some are excluded.

The levels of confidence required within the specified intervals depends on the test conditions (test period length, repeatability and reproducibility of the test and number of measurements). These levels are specified in chapters I-7 and I-8.

I-5.2 Reference Gross Weights and Static Axle Loads

I-5.2.1. The static weighing operations must be done either axle by axle, or by group of axles, by wheel/axle weighers approved for enforcement or commercial applications, or on a weigh­bridge in order to weigh a whole vehicle at once. It is strongly recommended to weigh the gross weights on an approved weigh-bridge to get a reliable reference. In this case, the proce­dure described in the detailed specification (see 8.3.3) should be applied to derive the refer­ence static axle loads.

I-5.2.2. The road surface on the weighing area should be flat and horizontal. In the latter case, it is recommended to:

The level difference between axles of a same group should not exceed 2 mm. The level differ­ence between single axles or groups of axles should not lead to more than 0.5% slope (i.e. 1.5 cm for 3 m spacing).

I-5.2.3. During wheel or axle static weighing operation, the vehicle brakes must be fully re­leased.

The results of lorry weighing during an enforcement operation by the police may introduce a bias. Therefore it is recommended that lorries (rented for example) specially allocated to the test for one or two consecutive days are used.

I-6 Type (model) Approval

I-6.1. A type (model) approval test should be organised once to label a newly marketed WIM system.

I-6.2. The test should be carried out on a site in class I (excellent), with a radius of curvature longer than 2500 m, but a straight road is highly recommended.

I-6.3. After the system installation by the manufacturer or the vendor, a pre-calibration should be done using two test lorries (chosen according to 7.2.3.3 of the detailed specification), and one load per lorry, with the following test plan (8 runs):

For a high-speed WIM system, Vm will be taken equal to 75 km/h. For a low-speed WIM sys­tem, Vm will be the recommended operation speed.

The static reference axle loads and gross weights should be given to the manufacturer or ven­dor, in order to adjust the system calibration.

I-6.4. After the pre-calibration, the test is carried out using the standard test plan N°3 (I-8.2). In addition to the 110 runs specified in I-8.2, two of the test vehicles will make 6 additional abnormal runs in order to check the ability of the system to detect such situation, and eventu­ally to mark the wrong measurements with a violation code. These runs will be done as:

I-6.5. During the test, the manufacturer or vendor will not be allowed to access to the system.

I-6.6. The reference weights and static loads will be assessed according to the rules given in 9.6 of the detailed specification.

I-6.7. All the recorded data, except those marked with a violation code by the system, will be considered. The data analysis will be done as for an in-service verification (see I-7.2) accord­ing to the clauses of section I-8.3. The test conditions will be (I) (environmental repeatability) and (R1) (limited reproducibility). The results will be reported according to the Appendix IV format.

A second analysis will be carried out as for an initial verification (see I-7.1), by removing the mean bias on the gross weight for all the runs, applying by software a constant multiplicative factor on all the recorded axle loads. Then, the factor 0.8 mentioned in I-7.1.3 will be used to assess the accuracy classes for each criterion. The test report will present both analyses.

I-7 Initial and in-Service Verifications of a WIM System

I-7.1 Initial Verification

I-7.1.1. After installation and calibration of a WIM system, or during recalibration, a test may be carried out to assess its accuracy, using the same data used for the (re)calibration. This is an initial verification.

I-7.1.2. If the WIM system is calibrated using static calibration masses, all the results must be found within the interval [-5 /2; 8 /2 ] of the relevant accuracy class and criterion (single axle, axle group or gross weight according to the measuring scale length).

I-7.1.3. If the WIM system is calibrated using repeated runs of pre-weighed vehicles or in­strumented vehicles, the confidence intervals given in Table 12 are modified using a width re­duction such as [- 0.88; 0.88] for each relevant accuracy class and criterion. The required con­fidence level of this interval is given in chapter I-8.

I-7.2 In-Service Verification

I-7.2.1. An in-service verification may be done at any time during the lifetime of a WIM sys­tem. In this case, the sample of data used for the accuracy assessment must not have been used for any calibration or recalibration of the system.

I-7.2.2. If the WIM system is checked using static calibration masses, all the results must be found within the interval [-5; 6 ] of the relevant accuracy class and criterion (single axle, axle group or gross weight according to the measuring scale length).

I-7.2.3. If the WIM system is checked using repeated runs of pre-weighed vehicles or instru­mented vehicles, or using single runs of pre- or post-weighed vehicles from the traffic flow, the confidence intervals given in Table 12 are used for each relevant accuracy class and crite­rion. The required confidence level of this interval is given in chapter I-8.

I-8 Procedures to Check the Accuracy of a WIM System

I-8.1 General Rules

I-8.1.1. The assessment of the accuracy of a WIM system requires a test. This chapter deals with tests carried out using either repeated runs of pre-weighed vehicles. The use of single runs of pre- or post-weighed vehicles from the traffic flow is considered in the main docu­ment, chapter 11.

I-8.1.2. The more extensive the test plan (longer test period, higher number of vehicle types and runs), the higher the confidence of the conclusion. This means that the client risk (i.e. the risk of accepting a system in a higher class than it is) decreases as the test becomes more ex­tensive, while the test cost increases. In this analysis the supplier risk linked to the statistical estimation of the mean bias is fixed at 5%.

I-8.1.3. In this specification, the client risk is governed by the probability ofan individual er­ror (with respect to the static load or weight) lying outside of the specified confidence interval. An upper bound of this risk is fixed by the specified values (1-no), where n0 is the required confidence level.

In this appendix, this risk (1-n0) is fixed at 5%, or 10% in one case (see I-8.5).

This risk is only assessed under the conditions of the acceptance test (see I-8.5); this means that the farther the test conditions are from the traffic conditions, the lower the real confidence and the higher the customer risk.

I-8.1.4. Depending on the test period length, the test may be carried out in (see definitions in the main document, section 11.1.5):

I-8.1.5. Neither recalibration nor component exchange should be done during a test period (see the main document, section 11.1.5 for further details).

I-8.1.6. According to the number of test (pre-weighed) vehicles, and load and speed cases, the test may be carried in (see definitions in the main document, section 7.2.3.2):

(r1) full repeatability

(r2) extended repeatability

(R1) limited reproducibility

(R2) full reproducibility (not considered in this appendix)

Some standard simplified test plans are given in section I-8.2.

I-8.1.7. Test vehicles are vehicles which are pre-weighed on an approved static scale or weigh-bridge, and perform repeated runs over the system. The static weighing operation must be made carefully, according to I-5.2.

I-8.2 Test Plans

Three main test plans are proposed, two of these being divided into two sub-plans, in order to comply with the customer requirement and resources. They are described in order of increas­ing cost and reliability. They all are in environmental repeatability conditions (I). The choice of test vehicles should be based on the most common types in the traffic flow.

The tandem or tridem axles should be equipped with air suspensions, in order to avoid gross errors on the static reference axle loads.

Test plan N° 1.1. One load, 10 runs, full repeatability conditions (r 1) (Confidence Level n0 = 95%)

This very short test is mainly recommended for periodical checks, carried out several times per year, or if only one type of vehicles is to be weighed by the system.

The test is carried out within a single day, according to Table 13:

It is recommended that the test vehicle is loaded to close to the mean gross weight of the same type of vehicle in the traffic flow.

Table 13: Test plan N°1.1

Test vehicle Speed Number of runs
2-axle rigid lorry 1.2Vm 2 runs —
or Vm 6 runs 7 runs
5-axle semi-trailer 0.8Vm 2 runs 3 runs

 

Vm: mean lorry speed in the traffic flow – last column, only if 1.2Vm exceeds the speed limit.

Test plan N° 1.2. Two loads, 30 runs, extended repeatability conditions (r2) (Confidence Level n0 = 95%)

This short test is recommended for a yearly check of a WIM system.

The test is carried out within a single day, according to Table 14:

Table 14: Test plan N°1.2

Test vehicle Speed Load cases and number of runs
fully loaded half loaded
2-axle rigid lorry 1.2 Vm 3 runs — 3 runs —
or Vm 9 runs 10 runs 9 runs 10 runs
5-axle semi-trailer 0.8Vm 3 runs 5 runs 3 runs 5 runs

 

Vm: mean lorry speed in the traffic flow – in italic, only if 1.2Vm exceeds the speed limit.

2 lorries are used: a 2-axle rigid lorry, and a semi-trailer with tridem, (or a rigid lorry with a trailer and a tandem or tridem axle, if the traffic flow contains a high proportion of this type).

Test plan N° 2.1. (confidence level n0 = 90%)

One load per lorry and 2 x 10 = 20 runs, limited reproducibility conditions (R1).

This test can be conducted for a newly installed WIM system, or after repair or modification of the system, if customer resources and time are limited; the confidence level is lower than for the other tests.

The test is carried out within a single day, according to Table 15:

Table 15: Test plan N°2.1

Speed Test vehicles and number of runs
2-axle rigid lorry 5-axle semi-trailer (or road train)
1.2Vm 2 runs — 2 runs —
Vm 6 runs 7 runs 6 runs 7 runs
0.8Vm 2 runs 3 runs 2 runs 3 runs

 

Vm: mean lorry speed in the traffic flow – in italic, only if 1.2Vm exceeds the speed limit.

It is recommended that the test vehicles are loaded to close to the mean gross weight of the same type of vehicles in the traffic flow.

If a road train is used instead of a semi-trailer, it is recommended that this vehicle should have at least one tandem or tridem axle.

Test plan N° 2.2. (confidence level n0 = 95%)

Two loads per lorry and 110 runs, limited reproducibility conditions (R1).

This test is recommended for a newly installed WIM system, or after repair or modification of the system.

The test is carried out within one to three consecutive days but under the same climatic condi­tions, according to Table 16:

Table 16: Test plan N°2.2

Test vehicle Speed Loading and number of runs
fully loaded half loaded
2-axle rigid lorry 1.2Vm

Vm 0.8Vm

 8 runs

14 runs

8 runs

 —

20 runs

10 runs

 5 runs

10 runs

5 runs

 —

13 runs

7 runs
5-axle semi-trailer 1.2Vm 8 runs — 8 runs —
(or road train) Vm 14 runs 20 runs 14 runs 20 runs
0.8Vm 8 runs 10 runs 8 runs 10 runs

 

Vm: mean lorry speed in the traffic flow – in italic, only if 1.2Vm exceeds the speed limit.

Remark: the required number of runs is based on the gross weight criterion. If only the single axles criterion (and eventually the axles of a group criterion), is (are) to be checked, the total number of runs may be reduced proportionally, in order to get 110 single axle (axles of a group) passes. But in this case the other criteria should not be checked

Four lorries, and 110 runs, limited reproducibility conditions (R1), (confidence level n0 = 95%).

This test plan is the most representative, with a limited number of test vehicles, of the real traffic flow. Nevertheless it is a quite extensive test, which generally cannot be applied as a common acceptance test, but is recommended for new types of system or if several systems are tested together on the same test site.

The test is carried out within one or two consecutive days but under the same climatic condi­tions, according to Table 17:

Each test vehicle should be loaded to the mean gross weight of the same type of vehicles in the traffic flow.

The total number of runs of each type of lorry may be adapted to fit with the proportions in the traffic flow on the WIM site, with a total number equal to 110.

The test confidence may be increased (but also its cost) by splitting each set of runs (by vehi­cle type) into two different loads (full and half load, according to the expected proportions in the traffic flow).

Table 17: Test plan N°3

Test vehicle Speed Nb runs Number of runs
2-axle rigid lorry 1.2Vm

Vm 0.8Vm

 30 8 runs

14 runs

8 runs

 —

20 runs

10 runs
3-axle or 4-axle rigid lorry 1.2Vm Vm 0.8Vm 6 1 run

4 runs

1 run

 —

4 runs

2 runs
5-axle semi-trailer 1.2Vm Vm 0.8Vm 60 15 runs

30 runs

15 runs

 —

40 runs

20 runs
road train 1.2Vm

Vm 0.8Vm

 14 4 runs

6 runs

4 runs

 —

9 runs

5 runs

Vm: mean lorry speed in the traffic flow – in italic, only if 1.2Vm exceeds the speed limit.

 

The road train (rigid lorry with trailer) should be of the type which is the most common at the site.

If the proportion of one of the proposed vehicle types is negligible (or much smaller than the other ones) in the traffic flow at the site, this type may be excluded, but the total number of runs should be kept equal to 110.

I-8.3 Test Results Analysis

I-8.3.1. The level of confidence is generally fixed at 95%; in one case (test plan 2.1) it is only 90% to reduce the number of runs by a factor 6.

I-8.3.2. Before making the analysis, the numbers n of recorded gross weights, groups of axles, single axles and axles of a group must be counted. The number of gross weights should be equal to the value specified in the test plan; if this number is less than 95% of the specified value, the test must be continued or started again, in order to comply with this requirement.

I-8.3.3. The simplified procedure for assessing WIM system accuracy is described step by step below:

  1. For each entity (gross weight, single axle, group of axles and axles of a group) the individ­ual relative errors with respect to the static load (weight) or the accepted reference values are calculated:

(Wdi–Wsi)

x, =        —         [†]100 (in %)

i           Wsi

where Wdi and Wsi are the in-motion measured value and the static (reference) value.

  1. The mean m and the standard deviation s of the relative errors in each sub-population of xi (same entity) are calculated.
  2. For the considered accuracy class, |m|/s and 8/s are calculated (8 is given in Table 12); in the case of an initial verification, 8 is replaced by 0.80.8 (see I-7.1.2).
  3. In the diagrams given below, for each test plan and each sub-population size n, one curve delimits the “acceptance domain” and the “rejection domain”.

For each entity (sub-population), if the point of coordinates (|m|/s ;8/s) in the relevant diagram is in the “acceptance domain”, then the considered accuracy class is accepted; if not, the con­sidered accuracy class is rejected; a lower class is considered and the process is repeated from step 3.

Remark: For the test plans N°1.1 to N°2.1, all the possible number n are given in the relevant diagrams (one curve for each value).

For the test plans N°2.2 and N°3, the possible numbers of n becomes too high, and one curve corresponds to each possible interval for n. For any modified test plan which provides a number n outside of these intervals, either:

 

 

 

I-8.4 Graphs
Figure 4: Diagram of the test plan N°1.1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 6: Diagram of the test plan N°2.2

 

 

Figure 7: Diagram of the test plan N°3

APPENDIX II. SENSOR ACCEPTANCE AND INSTALLATION

II-1 Checking Before Installation

II-2 Mechanical Checking

II-2.2 Case of Mats (Capacitive or other)

 

II-2.3 Case of Strain Gauge or Load Cell Scales

to be completed later with contribution from Captels and Pietzsch

II-3.1 Case of Piezo-Electric Bars

For a class A(5), B+(7) or B(10) system, the homogeneity of the response along the bar may be tested by the use ofa repeatable shock device, such as a FWD (Falling Weight Deflectome­ter) or a DYNAPLAQUE (a French portable calibrated shock device used in various coun­tries), with an adapted plate to allow the application of the force to the sensor alone.

The procedure will then be as follows: n series of 3 successive shocks are applied at n points, distributed every 25 cm along the bar, and the means of the 3 measurements are calculated; the two extreme mean values are eliminated and the remaining series of (n-2) mean values are analysed (it produces 11 values for a 3.50 m bar); the bar can be regarded as satisfactory if the range of the mean values observed is within ± 10%. It is recommended to repeat these tests at another ambient temperature (difference of at least 10°C) to estimate the influence of temperature on the bar/pavement system and eventually to correct the signal for temperature effects.

II-3.2 Case of Piezo-Quartz Bars

to be completed later with contribution from Kistler

II-3.3 Case of Capacitive Strips

to be completed later with contribution from Golden River

II-3.4 Case of Fibre Optic Sensors

to be completed later with contribution from Alcatel

II-3.5 Case of Capacitive Mats

II-3.6 Case of Strain Gauge or Load Cell Scales

to be completed later with contribution from Captels and Pietzsch

II-4    Installation recommendations

A WIM system installation (sensors and electronics) requires special procedures depending for example on the type of sensor, the structure of the road, the environmental conditions and the application.

It is obvious that general information should be given concerning the criteria for the site choice, with respect to the embedding and installation procedure of the sensors, etc. (see (METT-LCPC, 1993) for piezo-electric bars).

Nevertheless, the responsibility for a fair installation is the manufacturer’s. He must have at his disposal the required information from the road manager, for example about the road struc­ture, the environment and the climatic conditions during the installation.

For some types of sensors, such as piezo-electric and piezo-quartz bars or capacitive strips, the choice of resin to fix the sensor in the pavement may have some influence on the accuracy and has a strong influence on the sensor lifetime and installation delay; some recommendations are given for piezo-electric bars in (METT-LCPC, 1993).

References to some sensor installation guidelines or specification may be given here; contri­butions from various manufacturers would be welcome.

APPENDIX III. CALIBRATION METHODS

The calibration methods most commonly used are briefly described below, from the simplest to the most sophisticated; other methods may be considered.

Description of common calibration methods

We note:

Wd, =M,

In the conditions (r2), it is recommended to consider the different configurations (loads and speeds) of the same vehicle as different vehicles for the data analysis.

Calibration coefficient: a calibration coefficient is defined as a multiplicative factor C to be applied to a raw recorded “dynamic” load Wd to get the final estimation of the static load (or the “calibrated” result) noted W: W = C.Wd .

A calibration coefficient is intended to eliminate as far as possible any systematic bias in the WIM system, which may partially be induced by the pavement profile (spatial repeatability ef­fect).

If the WIM system uses more than one sensor, at least one calibration coefficient must be computed for each of them.

In some “sophisticated” WIM systems, several calibration coefficients may be computed for each sensor, depending on the type of vehicle or on the axle rank (see 2. and 3. below).

For bridges, the calibration coefficient is replaced by a calibration curve, an influence line or surface.

Among the proposed methods outlined below, the first two (1.a and 1.b) are the most com­monly used, while the third one (1.c) is often recommended; they all provide only one calibra­tion coefficient by sensor.

1.a. Calibration on the mean bias: this method consists of calculating the calibration coeffi­cient C such that for the mean bias of the relative errors for the gross weights of all the test vehicles measured in motion (one measurement for each run) is re­moved, each of them being accounted as many times as the lorry passed:

Z ni

y ( Wd,L ‘

Ws
i,k \ Wsi 7

This method provides an unbiased estimator of the gross weight. It is recom­mended in (r1).

1.b. Calibration on the total weight: this method consists of calculating the calibration coeffi­cient C as the ratio of the total static gross weight of all the test vehicles (each of them being accounted for as many times as the lorry passed) to the total gross weight of these vehicles measured in motion (one measurement for each run):

ZniWsi

i                    

Zwdik
i,k

This method provides an unbiased estimator of the total weight of all the vehicles. It is only recommended if the WIM purpose is the estimation of the whole traffic tonnage, such as in economical surveys of goods transportation.

1.c. Calibration on the mean square error (1): this method consists of calculating the slope of a regression line which passes through the origin in an orthonormal diagram plotting the individual “dynamic” gross weights versus the individual static gross weights of the test vehicles for each passage. It is based on the fact that a WIM system should provide “dynamic” loads which are proportional to the static loads. The calibration coefficient C is given by:

ZniWsi2

C =

ZWsiWdik i,k

This method may be applied for conditions (r2) to (R2), with more than 3 lorries (or loading cases); it minimises the mean square error of the individual gross weight measurements with respect to the static gross weights for all the vehicles

passed, with the constraint that the “dynamic” gross weights are proportional to the static ones. It is recommended for most applications, when the purpose is the estimation of the individual lorry weights, because the estimator has a lower vari­ance than the two previous ones and a very small bias.

1.d. Calibration on the mean square error (2): this method consists of calculating the slope and the ordinate at the origin of the regression line in an orthonormal diagram plotting the individual “dynamic” gross weights versus the individual static gross weights of the test vehicles for each passage. The mean square error should be smaller than with the previous method, but the proportionality between the “dy­namic” loads and the static loads is no longer ensured, which is not in accordance with theory. The calibration procedure becomes: W = C.(Wd – b), with b and C given by:

.A                   A 2

X n IIX n Wsi I .—I X n Wsi I

c=7——— a( A i—Afi y                                        Ă                                 (6)

IX n II X WSi Wdik -IX n WSi || X Wdk

\ i             J\ i,k                / \ i                                 J\ i,k          ^

and
XniWsi2

i
y nWSi Yy WsWdik ii                          iik

i                     )\ i ,k               J

 

This method is not recommended in most cases because of the reason explained above. Furthermore, if applied, the b value should be rather small and independent of the calibration vehicle sample considered, which is not necessarily the case.

In both methods 1.c. and 1.d., the gross weights may be replaced by the axle loads and the formulas adapted. The calibration coefficients will then be slightly different. This is not highly recommended, because the individual axle loads are more significantly affected by the dy­namic motion of the vehicles than the gross weights, and because the static axle loads are not well defined.

  1. Calibration by lorry type: this method provides one calibration coefficient for each type (silhouette) of lorry from the test sample, or for each class of silhouette (e.g., rigid lorry, tractor + semi-trailer, lorry + trailer). It is only applicable for conditions (R1) and (R2), and of interest if the WIM station software is able to manage such a set of calibration coefficients according to each lorry type. The same formulas as in a. to 1.d may be applied, as many times as the number of lorry classes consid­ered. The same remarks apply to each formula and procedure.
  2. Calibration by axle rank: this method provides one calibration coefficient for each rank (and/or type) of axle within a lorry, taking into account the fact that the axle dy­namic behaviour depends on their rank in the vehicle. It is only of interest if the WIM station software is able to manage such a set of calibration coefficients ac­cording to each axle rank. It is recommended to consider the following sub­populations, some of which may be merged for simplification:

The formulas given above are again applied to each sub-population by replacing the gross weights by the axle loads. The same remarks apply to each formula and procedure.

Except for bridge WIM, all of these calibration methods are more efficient in cases (R1) and (R2) with a test lorry sample being representative of the expected traffic flow. In the case of (r1) or (r2) it is recommended to choose loads (gross weights and axle loads) which are repre­sentative of the load distribution encountered for the same type of vehicles as the test lorry in the traffic flow.

 

APPENDIX IV. STANDARD FORMAT, COMPUTER TOOLS

IV-1 Standard Results’ Format and Computer Tools for Accuracy Assessment

IV-1.1 Data Presentation and Statistics Calculation

Figure 10 and Figure 11 gives a standardised table, designed under Excel, which contains the most useful information in order to apply this specification and to assess the accuracy of a WM system. A sample of this Excel sheet is available on floppy disk, or by Internet on the European WIM Web site (http:/www.zag.si/wim/specification).

The first part of the table gives the standard data delivered by the WIM system, which should be easily extracted from the original data files. The vehicles affected by a violation (error) code were eliminated, but accounted for, to be reported in the test report. The general heading only contains a summary of the required information listed in 12.1.18 of the specification. The successive columns contain:

N.B. the axle spacing is used in the pre-processing of the raw data to identify the single axles and axles of group, but are not necessary for the further accuracy analysis.

All the weights and loads are given in kg, but with a scale division of 100 kg according to the sensitivity and accuracy of the system.

The second part of the table gives the relative errors, automatically calculated by formula in the Excel sheet, and the axle type (single axles or axle of group). Finally, the statistics of the relative errors, as required in 11.4.5 and 11.4.6 are automatically calculated by formula, com­bining the individual relative errors and the type of axle information.

The small table of these statistics are the sufficient information needed to perform the accu­racy calculation, using the test conditions (see IV-1-2).

IV-1.2 Accuracy Calculation

The accuracy calculation, according to the procedure detailed in the chapter 11, may be auto­matically done using the standardised Excel sheet given in Figure 12. The statistics calculated in IV-1.1 are introduced in the relevant cells, as well as the test conditions. The percentages of identified vehicles in the whole test sample are reported for information. If the system has a violation code, two percentages should be given, taking into account or not the vehicles iden­tified but wrongly measured.

For an initial verification, where the same data sample is used to recalibrate the system, and thus the mean bias on the gross weight is (almost) removed, 8 is automatically multiplied by k=0.8 (see 10.1.3).

The built-in formula calculates the values of n0, and using the solver with appropriate arbi­trary initial values of 8min automatically fulfils the table. The standardised graph is also pro­vided, which shows both the 8min and 8c values for all the criteria.

This Excel sheet is also available on floppy disk, or by Internet (see IV-1).

Figure 10, Figure 11 and Figure 12 are at the end of the section IV-2.2.

IV-2 Example of Implementation of the Checking Procedures

In order to illustrate the procedure explained in chapter 11, an example is given hereafter.

IV-2.1 Calibration Plan

A WIM system was installed and calibrated during one and half day (environmental repeat­ability conditions (I)), following the procedure described in 7.2.3. The calibration plan was the following:

According to this calibration plan, we are in the conditions of limited reproducibility (R1).

Table 18: Calibration plan with two pre-weighed test lorries

Test vehicle Speed (km/h) Loading and number of runs
Fully loaded Half loaded Empty
80 10 runs – –
2-axle rigid lorry 65 20 runs – –
50 10 runs – –
5-axle semi– 80 10 runs 10 runs 5 runs
trailer 65 10 runs 10 runs 5 runs
50 10 runs 10 runs 5 runs

 

IV-2.2 Initial Verification and Accuracy Check

The results of the initial verification using the calibration sample data are summarised in Table 19.

The values of 5 are taken from Table 5 for the classes retained, and multiplied by the reduction factor k = 0.8 (see 10.1.3). The theoretical probability n is used. The minimum required n0 are either taken from Table 7 (conditions (I) and (R1)) and interpolated, or automatically by the Excel sheet of Figure 2. Values of 8min obtained for n = n0 are also given to be compared to k.5.

It may be seen that the WIM system fulfils the requirements of class (C(15) in this initial veri­fication, and even B(10) for the group of axles criterion (tridem here). Figure 8 show the re­sults as well.

Table 19: Results of the initial verification

Statistics of relative errors Accuracy calculation
Number Mean Std deviat. no Class 0.8*S Smin Sc n Accepted class
Criterion n m (%) s (%) (%) (%) (%) (%) (%)
gross weight 115 -0.29 4.28 95.1 C(15) 12 9.3 11.7 99.0 C(15)
group of axles 75 0.23 6.01 94.6 C(15) 14.4 13.1 13.4 96.6
single axle 235 -0.62 7.31 95.7 C(15) 16 15.9 14.9 95.9
axle of group 225 0.26 6.96 95.7 B(10) 16 15.1 9.4 96.9

Figure 8: Results of the initial verification

 

IV-2.3 In-Service Check of the WIM System

After the initial calibration, a test was performed to check the accuracy of the system in more realistic conditions, i.e. in full reproducibility conditions (R2). For that in-service checking, the test plan was to take about one hundred lorries from the traffic flow, with the help of the police during an enforcement period over three consecutive days (environmental repeatability conditions (I)). These lorries were pre-weighed on an approved static scale installed 5 km up­stream of the WIM site. The axle loads were measured on this scale. Every pre-weighed lorry was then identified by its registration plate (and some visual description) and when passing on the WIM site (two operators were linked by radio).

The sample composition of these pre-weighed lorries was chosen in accordance with the traf­fic composition of this road, following a special agreement with the police. 86 lorries were weighed on the static scale and available for the test analysis.

The results of the test are summarised in Table 20 with the same presentation as in Table 19. The value of n0 are taken from Table 7 (interpolated) or calculated.

The alternative method described in 11.4.7.2 is applied by comparing the value of 8min, for which n=n0, to the 5 of the required class. 8min gives the right accuracy level between two lev­els codified by letters. For two criteria (single axles and gross weight) the accuracy level is really in-between the conventional limits of classes B(10) and C(15). For the axle groups, it is closer to the class C(15) limit, while for the axles of group the value is just above the limit of the class B(10) (Figure 9).

Table 20: Results of in-service verification

Statistics of relative errors Accuracy calculation
Number Mean Std deviat. no Class S Smin Sc n Accepted class
Criterion n m (%) s (%) (%) (%) (%) (%) (%)
gross weight 86 -2.27 6.09 92.6 C(15) 15 13.0 13.0 96.3 C(15)
group of axles 66 0.30 8.44 92.1 C(15) 18 17.1 14.1 93.6
single axle 197 -3.92 7.66 93.7 C(15) 20 17.1 12.1 97.3
axle of group 169 -0.19 10.07 93.5 C(15) 25 20.3 10.3 97.9

 

Figure 9: Results of in-service verification

 

The system is accepted in accuracy class C(15) for all criteria.

In comparison with the initial verification, the bias on the single axle loads and on the gross weights are respectively increased by factors of 5 and 10 (but the last one was very small), while the standard deviations of the axle group and gross weight samples increased by more than 40 %.

This example, based on true data, shows a typical difference between the two steps (initial verification and in-service checking).

 

80
System: “Name or manufacturer”

Period of the test: “from date1 to date2”

Test conditions: (I to III) and (r1 to R2)
Location: “test site”                      Lane N°: kk

Number of test vehicles: nnnn
RECORDED DATA
N° Date/ time T

(°C)

 V (km/h) Type In motion loads/weights Wd (kg) Static loads/weights Ws (in kg)
GW A1 A2 A3 A4 A5 A6 GA1 GA2 GW A1 A2 A3 A4 A5 A6 GA1 GA2
1 5/4/98

8:10:25

 9,3 85 5 38000 6400 10600 7000 7000 7000 21000 39000 6500 10800 7300 7200 7200 21700
2 8:11:23 9,5 89 5 39600 6200 13500 6900 6300 6700 19900 40200 6600 12100 7300 7000 7200 21500
3 8:12:28 9,7 89 5 39400 5900 10300 9400 7300 6500 23200 38500 5900 9600 8700 7900 6400 23000
4 8:13:06 9,9 88 5 40900 7100 10100 8200 8100 7400 23700 40500 6700 10100 7900 7900 7900 23700
5 8:13:30 10,1 87 5 40000 8100 11700 7100 7000 6100 20200 39800 7500 11800 6900 6600 7000 20500
6 8:13:56 10,1 90 5 30700 6700 12500 3600 4200 3700 11500 28200 6700 9900 3900 4000 3700 11600
7 8:14:54 10,1 89 5 41300 6400 9800 8300 8600 8200 25100 42800 6100 10600 8700 8800 8600 26100
8 8:16:09 10,1 79 6 36600 6500 12400 9000 8700 35700 6400 11100 9200 9000
9 8:16:12 10,1 85 5 40500 6500 9200 8500 8100 8200 24800 39500 6500 9400 8100 8200 7300 23600
10 8:16:32 10,1 88 5 42200 8200 13800 6900 6500 6800 20200 42000 7800 13400 7000 6900 6900 20800
11 8:17:38 10,1 81 5 48300 8000 13000 8900 9200 9200 27300 48600 8100 13700 8900 8900 9000 26800
12 8:18:28 10,2 88 5 33500 6100 5600 7300 7400 7100 21800 32600 6000 5200 7000 7200 7200 21400
13 8:19:03 10,3 89 5 48700 7800 14200 9100 8500 9100 26700 46800 7200 14500 8400 8400 8300 25100
14 8:19:50 10,4 86 5 37400 6400 8600 7600 7500 7300 22400 38600 6200 8300 8100 8400 7600 24100
15 8:20:49 10,5 80 6 44900 6700 7200 8300 6400 7400 8900 16300 43900 6500 7500 8000 7100 7000 7800 14800
16 8:21:03 12 88 5 39000 6200 10100 7500 7900 7300 22700 38400 6000 9700 7400 8000 7300 22700
17 8:21:20 12 92 5 20200 5400 5300 3500 2900 3100 9500 20600 5200 5400 3400 3300 3300 10000
18 8:22:29 12,3 86 5 37000 6100 7900 8400 7200 7400 23000 35000 6000 7400 7800 6800 7000 21600
19 8:23:27 12,3 95 2 6800 2700 4100 6900 2900 4000
20 8:23:33 12,4 88 5 43300 6900 11400 8800 8200 8000 25000 45400 6700 12100 8900 8900 8800 26600
21 8:23:37 12,8 85 5 41500 6400 8800 7800 8200 10300 26300 40500 6200 9000 7300 8000 10000 25300
22 8:25:13 13,4 89 6 42800 8400 12200 6700 7500 8000 18900 42900 7800 12300 6900 8000 7900 19200
23 8:25:25 13,4 87 5 31300 6500 8800 5500 5000 5500 16000 29900 6300 8200 5200 5100 5100 15400
Load and weights may be expressed either in kg, 100 kg, tons or kN; the unit must be specified in the headings
European WIM Specification
Type : the classification is that recommended by this specification; otherwise the vehicle categories should be given apart

– axle loads are placed in columns A1 to A6, according to the axle rank; GA1 (and GA2 if needed) contains the group(s) of axle loads

– the type may be replaced by the number of axles, or this number may be added in an additional column

– for more than 6 axles, add columns after A6; for more than 2 axle groups, add columns after GA2

Figure 10: Standardised recorded data format and statistics – parti

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

81
Cont‘n
STATISTICS
N° Relative errors (%) Type of axle (1=SA, 0=AoG) Statistics of the relative errors (%)
GW A1 A2 A3 A4 A5 A6 GA1 GA2 A1 A2 A3 A4 A5 A6 GW SA AoG GA
1 -2,56 -1,54 -1,85 -4,11 -2,78 -2,78 -3,23 1 1 0 0 0 number 23 52 60 21
mean 0,97 1,52 0,17 0,06
2 -1,49 -6,06 11,57 -5,48 -10,00 -6,94 -7,44 1 1 0 0 0 st. dev 3,22 6,31 5,88 4,77
3 2,34 0,00 7,29 8,05 -7,59 1,56 0,87 1 1 0 0 0 GW= gross weight SA= single axle AoG= axle of a group GA= group of axles
4 0,99 5,97 0,00 3,80 2,53 -6,33 0,00 1 1 0 0 0
5 0,50 8,00 -0,85 2,90 6,06 -12,86 -1,46 1 1 0 0 0
6 8,87 0,00 26,26 -7,69 5,00 0,00 -0,86 1 1 0 0 0
7 -3,50 4,92 -7,55 -4,60 -2,27 -4,65 -3,83 1 1 0 0 0
8 2,52 1,56 11,71 -2,17 -3,33 1 1 1 1
9 2,53 0,00 -2,13 4,94 -1,22 12,33 5,08 1 1 0 0 0
10 0,48 5,13 2,99 -1,43 -5,80 -1,45 -2,88 1 1 0 0 0
11 -0,62 -1,23 -5,11 0,00 3,37 2,22 1,87 1 1 0 0 0
12 2,76 1,67 7,69 4,29 2,78 -1,39 1,87 1 1 0 0 0
13 4,06 8,33 -2,07 8,33 1,19 9,64 6,37 1 1 0 0 0
14 -3,11 3,23 3,61 -6,17 -10,71 -3,95 -7,05 1 1 0 0 0
15 2,28 3,08 -4,00 3,75 -9,86 5,71 14,10 10,14 1 0 0 1 0 0
16 1,56 3,33 4,12 1,35 -1,25 0,00 0,00 1 1 0 0 0
17 -1,94 3,85 -1,85 2,94 -12,12 -6,06 -5,00 1 1 1 1 1
18 5,71 1,67 6,76 7,69 5,88 5,71 6,48 1 1 0 0 0
19 -1,45 -6,90 2,50 1 1
20 -4,63 2,99 -5,79 -1,12 -7,87 -9,09 -6,02 1 1 0 0 0
21 2,47 3,23 -2,22 6,85 2,50 3,00 3,95 1 1 0 0 0
22 -0,23 7,69 -0,81 -2,90 -6,25 1,27 -1,56 1 0 0 1 1
23 4,68 3,17 7,32 5,77 -1,96 7,84 3,90 1 1 0 0 0
– the relative errors are calculated cell be cell from the previous part of the sheet using the formula: e=(Wd-Ws)/Ws

– the type of axle can be delivered by the WIM system, or derived from the axle spacing (AoG if the spacing is less than 2.2 m)

– the statistics of the relative errors are calculated using formula, which combine the relative errors and the type of axle cells
Figure 11: Standardised recorded data format and statistics – part 2
European WIM Specification                                                        Standard Format, Computer Tools

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

82
A B C D E F G H I J M O P
1 Conditions (1) Test plan Envt Initial verification (Yes=1, No=0): 0
2 rr1 I
4 SYSTEM Number Identified Mean Std deviat no Class 3 min 3c n Accepted class
5 Entity (%) (%) (%) (%) (%) (%) (%) (%)
6 gross weight 100 96,2 1,50 3,67 95,0 B(10) 10 8,5 8,5 98,0 C(15)
7 group of axles 50 96,2 2,00 5,78 93,9 C(15) 18 13,3 10,3 99,0
8 single axle 170 95,5 2,10 6,54 95,5 B(10) 15 14,8 9,9 95,7
9 axle of group 120 96,0 2,50 9,72 95,2 C(15) 25 21,8 11,8 97,8
10 11
Q
European WIM Specification
12

13 14

15

16

17

18 19

20

21 22

23

24

25

26

27

28

29

30

31 32

33

34

35

36

37
(1)

“r1″=full repeatability

“r2″=extended repeatability

“rr1″=limited reproducibility

“rr2″=full reproducibility

“I”=environmental repeatability

“II”=environmental limited reproducibility

“III”=environmental full reproducibility
Users’ instructions:

1.         Enter the test conditions in cells B2 and D2, and put “1” in cell M1 if the same data sample was used for calibration (initial verification)

2.        Enter the test statistics (on relative errors) in cells B6 to B9 and D6 to E9

3.         (option) Initialise the expected values of 8min in cells I6 to I9 (only if step 4. fails)

4.         Start the command “Tools/Solver/Solve/keep the results”

– “Outils/Solveur/Resoudre/Garder la solution du solveur”

, and then OK if successful

if the solver doesn’t find an accepted solution, return to step 3. and modify the initial values of 8min
gross weight
group of axles
Criterion
single axle
axl eof g rou p
38

39
Figure 12: Standardised accuracy calculation sheet and presentation

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

83

[I] The APL is a device developed in France and in use in various countries, which measures the longitudinal profile; it consists of two single wheel trailers operating at 72 km/h, towed by a car.

[†] the procedure described in the detailed specification, chapter 11 may be applied.

Company

Solutions

News

Support

Newsletter Sign-Up

©2024 Analytiqum

Linkedin Facebook Instagram Twitter