METARAIL Project Final Report for Publication

Transcription

1 Project Authors: M. Wirnsberger (SC), M.G. Dittrich (TNO), J. Lub (NSTO), G. Pollone (IPSE), M. Kalivoda (Psi-A), P.v. Buchem (ERRI), W. Hanreich(ÖBB), P. Fodiman (SNCF) Date: December 1999 Methodologies and Actions for Rail Noise and Vibration Control Contract RA-97-SC.1080 Project Coordinator: Partners: Associated partners: SC IPSE NS PSI-A TNO ERRI OEBB SNCF Project funded by the European Commission under the Transport RTD programme of the 4 th programme framework

2 Publication data form 1 UR (1st authors) Michael WIRNSBERGER/Michael DITTRICH 2 Project No RA-97-SC report No D13 4 Title The METARAIL Project - 5 Subtitle Methodologies and Actions for Rail Noise and Vibration Control 6 Language English 7 Author(s) 8 Affiliation Michael WIRNSBERGER Michael DITTRICH Giuseppe POLLONE Jan LUB Manfred KALIVODA Werner HANREICH Paul van BUCHEM Pascal FODIMAN Schreiner Consulting TNO IPSE Srl; NSTO; Psi-A Consulting ÖBB ERRI 9 Sponsor, co-editor, name and address European Commission / DG VII SNCF 10 Contract RA-97-SC Publication date December Summary The overall results of the METARAIL project are presented in this report. In METARAIL, methodology, techniques and systems for measurement of railway exterior noise have been developed and improved for the purposes of type testing, monitoring and diagnostics. All methods and techniques developed in the METARAIL project are evaluated, including relevant procedures for type testing and monitoring of railway noise, some of which have been put forward as potential improvements to the pr EN ISO 3095 standard for railway exterior noise type testing. These methods were applied in a round robin test with a special test train at 4 sites: in Austria, The Netherlands, France and Italy. Measurement results are presented and assessed, and comparison is made between the sites. Differences are shown to be sufficiently small due to improved test procedures. In addition, some of the diagnostic techniques such as antenna techniques and track vibration analysis were applied to demonstrate and evaluate the separation of track noise and vehicle noise, track characterisation and assessment of wheel and rail roughness. 14 Key Words Railway noise, type testing, monitoring, diagnostic techniques, methodologies 16 No of pages 105incl. Front pages 17 Price On application 18 Declassification date - 15 Distribution statement PUBLIC 19 Bibliography - DISCLAIMER AND COPYRIGHT No rights can be derived from the contents of this report. Copyright METARAIL Consortium Deliverable D 13 Page 2

3 Contents 1 INTRODUCTION General introduction Partnership SYNTHESIS OF DEVELOPED METHODS Overall parameter sensitivity Direct roughness measurement Indirect roughness measurement Source location and quantification with an antenna Separation of track noise with spatial track vibration Separation of track noise by using a reference vehicle Train speed measurement and vehicle identification Characterisation of track dynamics Site sound transmission New concepts in relation to type testing Overall review IMPROVEMENTS TO THE TYPE TESTING PROCEDURE Introduction Current standard and proposed improvements Measured quantities Speed correction Site description Deliverable D 13 Page 3

4 3.6 Measurement set-up Track and vehicle noise Track conditions - roughness Track conditions - dynamic properties MONITORING METHODOLOGY, SYSTEMS AND APPLICATION General aspects The METARAIL monitoring system System configurations Application examples THE ROUND ROBIN TEST Introduction Description of the Test Train and Test Sites Measurement Procedure Results Wheel roughness vs. Rail roughness Wheel roughness during the round robin test Track behaviour and quality Site sound transmission CONFERENCES AND PUBLICATIONS METARAIL-Workshops Publications about the project METARAIL CONCLUSIONS AND RECOMMENDATIONS REFERENCES Deliverable D 13 Page 4

5 1 INTRODUCTION 1.1 General introduction The METARAIL project has been undertaken as a DG VII-funded project for the European Commission, running from 1997 to METARAIL stands for Methodologies and Actions for Rail Noise and Vibration Control. The main objectives of METARAIL were to develop and improve methodology and techniques for measurement of railway exterior noise for the purposes of type testing, monitoring and diagnostics. The overall results of the METARAIL project are presented in this report, which is based primarily on the deliverable reports D10 and D11. A full list of deliverable reports and publications is given in the reference list. All methods, techniques and systems developed in the METARAIL project are presented and evaluated, including relevant procedures for type testing and monitoring of railway noise, some of which have been put forward as potential improvements to the pr EN ISO 3095 standard for railway exterior noise type testing. These methods were applied in a round robin test with a special test train at 4 sites in Austria, The Netherlands, France and Italy. Measurement results are presented and assessed, and comparison is made between the sites. Differences are shown to be sufficiently small due to improved test procedures. In addition, some of the diagnostic techniques such as antenna techniques and track vibration analysis were applied to demonstrate and evaluate the separation of track noise and vehicle noise, track characterisation and assessment of wheel and rail roughness. In chapter 2 developed methods are presented and some examples of measurement results are shown. Chapter 3 deals with potential improvements to the type testing procedure. Monitoring methodology, developed systems and their application are covered in chapter 4. The round robin test and its results are presented in chapter 5, followed by conclusions and recommendations in chapter 6. This work is partly funded by the Transport Programme of the European Commission, DGVII, whose support is gratefully acknowledged. Deliverable D 13 Page 5

6 1.2 Partnership The following partners worked under the Actual Cost Contract No RA-97-SC.1080 METARAIL: Schreiner Consulting GmbH Status Coordinator Adress A-4020 Linz, Derfflingerstr. 14, Austria Telephone Fax metarail@sc-linz.co.at TNO Institute of Applied Physics Status Full partner Adress NL 2600 AD DELFT Stieltjesweg 1, The Netherlands Telephone Fax dittrich@tpd.tno.nl Ipse S.r.l. Status Full Partner Adress I TORINO, Via Pietro Nenni, 67, SETTIMO TORINESE, Italy Telephone Fax ipselab@tin.it Psi-A Consulting Status Full partner Adress A 2380 Perchtoldsdorf, Wiener Gasse 146/3, Austria Telephone Fax psia-consult@eunet.at Deliverable D 13 Page 6

7 1.2.5 NS Technical Research Status Full partner Adress NL 3551 EM UTRECHT, Concordiastraat 67, The Netherlands Telephone Fax ÖBB Österreichische Bundesbahnen Status Associated partner Adress A-1070 Wien, Zieglergasse 6, Austria Telephone Fax werner.hanreich@fw.oebb.at ERRI - European Rail Research Institute Status Associated partner Adress NL-3511 MK UTRECHT, Arthur van Schendelstraat 750, The Netherlands Telephone Fax pvbuchem@erri.nl SNCF Centre d'essais de Vitry Status Additional Partner for final phase of project Adress F VITRY-SUR-SEINE CEDEX, 21 avenue du président Allende, France Telephone Fax philippe.pinconnat@sncf.fr Deliverable D 13 Page 7

8 2 SYNTHESIS OF DEVELOPED METHODS Within METARAIL, progress has been made in developing a number of measurement techniques, some of which can provide improvements to type testing and monitoring methods. In this chapter some of the main findings and conclusions relating to each method are discussed. The major questions in relation to type testing and monitoring are 1. how to increase reproducibility; 2. which quantities and measurement procedures can be used to obtain accurate data, and to eliminate variation factors (e.g. along the track ); 3. how to specify the track conditions, especially rail roughness and track dynamics; 4. how to separate vehicle and track contributions. Separation of vehicle and track contributions is in fact a way of increasing reproducibility, as often the track contribution may be predominant and vary between individual sites. Also, clear specification of track conditions is a prerequisite to improve measurement reproducibility. The methods developed in METARAIL vary in their range of application, accuracy, maturity level and required expertise. Some techniques currently remain mainly suited for diagnostic purposes in their own right, such as the antenna technique and the spatial vibration method. As will be seen in the following sections, Some simplifications have also been put forward which can have direct impact on the type testing methodology dealt with in the following chapter. 2.1 Overall parameter sensitivity The parameter sensitivity study performed in work package II (see [2]) revealed the importance of some parameters for the measurement of pass-by rolling noise. The analysis was based on both calculations with the TWINS package, acoustics theory, literature and measurement data acquired during METARAIL measurement campaigns in Austria and the Netherlands. Deliverable D 13 Page 8

9 The following parameters influence the noise radiated by the track, thereby affecting the total noise level, but also the ratio between track and vehicle noise. The values given in table [2.I] below are indicative, and valid for conventional track systems. Parameter Parameter Parameter value Level difference for value for for maximum min. and max. minimum noise noise level parameter value level Rail type UIC 54 E UIC db(a) Static Pad Stiffness 5 e 9 [N/m] 1e8 [N/m] 5.9 db(a) Pad Loss Factor db(a) Sleeper type Bi-bloc Wooden 3.1 db(a) Sleeper distance 0.4 [m] 0.8 [m] 1.2 db(a) Ballast stiffness 1 e 8 [N/m] 3e7 [N/m] 0.2 db(a) Ballast Loss Factor db(a) Wheel offset 0 [m] 0.01 [m] 0.2 db(a) Rail offset 0 [m] 0.01 [m] 1.3 db(a) Wheel Roughness Smoothest Roughest 8.5 db(a) Roughness of Smoothest Roughest db(a) uncorrugated rails Train Speed 80 [km/h] 160 [km/h] 9.4 db(a) Wheel load [kg] 5000 [kg] 1.1 db(a) Air temperature 10 C 30 C 0.2 db(a) Table 2.I Indicative parameter sensitivity on total rolling noise for conventional track systems. Rail temperature, which can differ from air temperature due to heat radiation, can influence pad temperature and thereby pad stiffness and damping. Deliverable D 13 Page 9

10 The most important parameters to be taken into consideration to improve repeatability and reproducibility of noise measurements are: 1. The measured quantity, in terms of a clear definition of the type of noise level and measurement period, averaging time and the processing of level histories; the equivalent A-weighted sound pressure level is suitable. 2. Pad behaviour, both in terms of material behaviour, load and temperature effects, which can be taken into account by measuring vibration isolation as a site parameter. 3. the wheel-track combination and roughness excitation which can be taken into account by measuring A-weighted railhead vibration velocity, which is highly repeatable, and when corrected for rail roughness also well reproducible. 4. Site sound transmission effects, especially at 25 meters in combination with temperature and wind variation, especially wind direction; this can reduce repeatability, implying that measuring at 7.5 m is a preferable option; Site sound transmission at 25 meters must be characterised separately to eliminate local site effects. 5. Train speed, which at higher speeds can cause larger level variations, if it is not measured accurately. 2.2 Direct roughness measurement Direct roughness measurement refers to measurement procedures that scan the wheel or rail surface directly. The most frequently used systems employ a mechanical needlelike sensor, which takes samples at short intervals, for instance 0.5mm, along a measurement line in the direction of travel. Although the effect of the contact patch is not included in such measurements, they do provide representative data which can be processed for calculation purposes. The discretised roughness amplitude signal is conditioned and transformed into the frequency domain and presented in 1/3-octaves. Such systems are still the most accurate available, although for rail roughness measurements several rail sections have to be measured separately and processed to obtain an estimate for a section of track. Deliverable D 13 Page 10

11 Within task II.3.3 Direct wheel roughness estimation, a system for direct wheel roughness measurement was developed, which allows efficient in situ measurement and processing of wheel roughness. This was achieved with a system that simultaneously measures the wheel roughness along 3 parallel lines over the wheel running surface and provides 1/3-octave spectra of the roughness over the whole wheel circumference. The wheels have to accessible by jacking up and measurements are usually performed in a railway workshop. This system is described in more detail in [12]. The developed system was applied in the round robin test to measure the wheel roughness of the test train (see below). Disc braked Block braked sinter Block braked cast iron Figure 2.1 Left Wheel roughness measurement system. Right Wheel roughness of wagons in the METARAIL test train, amplified on an arbitrary scale, each on 3 parallel lines on the running surface. Although no new rail roughness measurement systems were developed within METARAIL, an existing system was applied to measure roughness at each of the test sites. In the ISO 3095 working (sub)group on Track conditions work has been done to specify the procedures for obtaining an overall site roughness level from discrete measurement locations. This is to be published in upcoming versions of the standard. Both rail and wheel roughness measurement systems require considerable user expertise and experience. The cost for wheel roughness measurement systems are estimated at around kecu and for rail roughness measurement systems around kecu, including positioning devices, sensors, measurement chains, signal conditioning, computers and software, and depending on options chosen. Deliverable D 13 Page 11

12 2.3 Indirect roughness measurement A new indirect measurement technique for estimating total roughness during pass-by has been developed in task III.3 (see [9] and [15]). The technique uses vertical railhead vibration, and can provide an estimate for roughness spectra of vehicle groups, bogies and under the right conditions, of individual wheels, as long as the wheel roughness exceeds the rail roughness. The method has been tested on existing data indirectly, and compared to actual direct measurement results. It was found that the spread in measured spectra is comparable to the spread found in directly measured roughness data. The measured parameters are vertical railhead vibration, train speed, and vertical spatial decay of the track. Current recommended train speeds during indirect roughness measurements are 80 km/h and lower, and 5 measurement positions per rail are recommended to fully characterise the roughness over the whole wheel circumference. According to TWINS calculations the indirect roughness measurement predicts the total roughness within an accuracy of 5 db assuming the following conditions: - the wheel radius is greater than 0.35 m, - the static wheel load is above 50 kn and, - the rail vibration decay is updated on-line. For smaller wheels or wheel loads, the method can be adjusted. The current level of accuracy (spectral) of ±5 db is comparable to the spread found in direct roughness measurements. A number of 3 to 5 measurement positions are recommended to obtain a representative average, as the distance covered by one accelerometer is only about 60cm. These positions can be adjacent or distributed over a greater distance than shown in figure 2.2. V 60cm Figure 2.2 Indirect roughness measurement with 5 transducers Deliverable D 13 Page 12

13 The total roughness for a single wheel passage is directly related to vertical railhead vibration, as described in report D8 [9]. A simplified formula for this relation is given here, with the factors C 2 and C 3 from report D8 combined to a tabulated factor C 23 (f), and vibration expressed in terms of velocity: L rtot VTD( f ) ( f ) Lveq( f ) + 10lg + C23( f ) 10lg( 2πf ) 8.68 = (2.1) with L rtot (f) = total roughness spectrum [db re 10-6 m], L veq (f) = equivalent velocity vibration spectrum [db re 10-6 m/s], V = train speed [m/s], T = passage interval [s], D(f) = vertical spatial decay spectrum [db/m], and C 23 (f) = conversion spectrum for contact filter and contact vibration to apparent roughness, shown in a graph and a table below for given speeds. Formula (2.1) gives an estimate the for absolute value of the total roughness from railhead vibration, train speed, spatial decay and contact transfer functions. As it is sensitive to spatial decay, pad stiffness and contact geometry, it is recommended to only use the relative formula if possible: 0 C 23 Correction factor -10 db re 1 [-] km/h 100 km/h 120 km/h 160 km/h , Frequency [Hz] Figure 2.3 Correction factor C 23 for contact filter and contact vibration to apparent roughness Deliverable D 13 Page 13

14 L rtot2( rtot1 v, rail,2 v, rail, 1 f f ) L ( f ) = L ( f ) L ( ) (2.2) with L rtot1, L rtot2 = total roughness spectra as function of frequency at one point for 2 different pass-bys (wheel, bogie or train), and L v,rail,1, L v,rail,2 = velocity vibration spectra, either both equivalent levels or L vmax levels, for the corresponding pass-bys. This relation eliminates the need to know other parameters, and can be used if an initial roughness level L rtot,1 is known. The current assessment of the indirect roughness technique is that it is very promising, time-saving and potentially inexpensive, as measurements can be performed during pass-by on large numbers of wheels, and no special equipment other than standard accelerometers and frequency analysers are needed. At present, however, expertise is still required to perform the analysis, but at a later stage, the post-processing may be automated or simplified. The method certainly has potential to be applied for type testing and monitoring purposes, as indicated below. In figure 2.4 a comparison of the average indirect roughness and the direct roughness measurements are given for the SGS vehicles of the round robin test train. Four lines are presented: the indirect roughness and the combined direct roughness of wheels and rail, as well as the average wheel and rail roughness separately. Not all data is available over the entire wavelength range presented (begin and end of valid data is indicated by vertical lines). The agreement between indirect and direct roughness is very reasonable in the region 10 cm to 1 cm. For wavelengths longer than 10 cm no data for rail roughness is available. The indirect roughness method produces results up to wavelengths of about 50 cm. Comparison with wheel roughness shows that in this region the indirect roughness is much higher than the wheel roughness, which indicates that the rail roughness is predominant in this region. Deliverable D 13 Page 14

15 Figure 2.4 Comparison of indirect and direct roughness levels for SGS vehicles In general, the agreement between indirect and direct roughness is quite reasonable, and indirect roughness measurement seems to be a quite promising technique. Furthermore, the indirect roughness procedure seems capable of measuring long wavelengths on the rail, although of course validation of these results is hampered by the fact that no direct measurement for this wavelength region is yet available Wheel roughness monitoring One of the obvious applications of indirect roughness measurement is to obtain fast online information on roughness levels on whole trains, without taking the train out of operation. By measuring the railhead vertical vibration level Lv over individual wheels, bogies, wagons or whole trains at particular speed, differences in roughness can be measured, as long as wheel roughness exceeds rail roughness by 6-10 db (10 db difference gives and error smaller than 0.5 db) Estimate for upper limit of rail roughness A less obvious, but very useful application of the indirect technique is to detect an upper limit for rail roughness at a particular site, without using a direct rail roughness measurement device. This is done by monitoring railhead vibration for varying traffic, and selecting the lowest vibration levels at a given train speed. As the total roughness consists of the energetic sum of wheel and rail roughness, the rail roughness has to be equal to or below the lowest peak in total roughness, caused by very smooth wheels. Deliverable D 13 Page 15

16 The formula for the upper limit of rail roughness L r,rail,max (λ) is given by Lr, rail,max( ) = min ( Lrtot, i ( λ)), i = 1... n λ (2.3) i where L rtot,i total roughness of pass-by i (wheel, bogie, wagon or train) and n number of pass-bys, which may be at various speeds Alternative for measuring rail roughness in type testing For type testing, it should be feasible to apply a simplified indirect roughness measurement to check the rail roughness. Instead of measuring the absolute rail roughness in detail it suffices to know whether wheel roughness exceeds rail roughness by more than about 10 db, in the A-relevant frequency region at given speed. So the condition L L 10 (2.4) r, wheel r, rail + can be satisfied if L, testvehicles ( f ) Lveq, smoothwheel ( f ) + 10 veq (2.5) for identical speed. So by comparing railhead vibration between a train with wheels 10 db smoother than the test train, direct rail roughness measurement can be avoided. This condition should primarily be valid in the A-relevant frequency region, which can be indicated by the part of the noise spectrum within 10 db of its maximum. The procedure can be performed specifically at the train speeds of the type test. Deliverable D 13 Page 16

17 2.4 Source location and quantification with an antenna Within task II.3.1 an existing antenna measurement system was improved specifically for railway noise applications (see [7]). Practical limitations of antenna systems in general and of the system developed were examined, leading to conclusions on the specific modes of application for railway noise. Some new measurement results were presented, illustrating the application potential: - Measurements performed with the improved system illustrated that the emission level of the superstructure of a freight train travelling at around 90 km/h was well below that of the wheels and track, by at least 15 db. - By means of averaging a pass-by the distinct radiation pattern of the rail was visualised. - Relative source strengths of wagons in complete goods trains were visualised. - For shrouded vehicles, the rail contribution could be separated from that of the wheel by measuring at close distance to the train (see figure 2.5, p. 17). In order to get an indication of the noise reduction of the shrouds, the acoustical images are split at a height of 0.2 meter. The shrouds begin at this height. The parts above and below 0.2 meter are integrated separately for all wagon types. An estimate for the noise reduction of the shrouds could be made by assessment of the antenna data. The shrouds will reduce the sound radiation of the vehicle by 2.5 db at 500 Hz 6 db at 1 khz 6.5 db at 2 khz. This reduction does not depend on the speed of the train. Antenna systems remain a specialised diagnostic tool, requiring special expertise and multi-channel data acquisition and processing systems. Costs for antenna systems are estimated between kecu, depending on the configuration. Deliverable D 13 Page 17

18 Figure 2.5: Acoustical antenna image of the round robin test train at 80 km/h for the 500 Hz, 1 khz and 2 khz octave bands, using a T-shaped microphone antenna at 2 meters from the track. The effect of the shrouds on four of the wagons ( meter) is clearly visible. Red = highest level, blue = lowest level, white = everything lower than blue. Deliverable D 13 Page 18

19 The system performance achieved in the important fundamental case of uncorrelated sources in the far field cannot be fully achieved in situations where focusing, ground reflections and correlated sources seem to determine the limit to resolution and dynamic range. Effectively, the system resolution in its current configuration, when applied to moving trains at a distance of 7.5 m, is 0.80 m at 2000 Hz, 1.60 m at 1000 Hz, and 3.20 m at 500 Hz. The effective dynamic range taken across the entire image is around 15 db. However, some parts of the image usually show a dynamic range of 25 db or more. If these parts are known to be representative of the whole train, stronger conclusions can be drawn from the image. Aspects of the application to train noise that currently limit the system's performance beyond the fundamental antenna performance, are: - ground reflections; - distributed sources on the upper part of the train in the presence of strong concentrated sources on wheels and track; - the short distance to the source which gives higher spatial resolution, but makes focusing errors outside the focal point larger. Feasible applications of antenna systems for railway noise are currently the following: 1. Determining the sound emission of separate wheels (or rather the combination of wheel and track since these cannot be separated). This allows verifying sound reductions due to design changes by fitting only a limited number of prototype wheels. 2. Determining the total emission of the superstructure, and the total emission of the wheels (of an entire carriage). 3. Locating source distributions on the superstructure when their emission level per square meter is no more than 10 to 15 db below that of the wheels. The following applications are currently considered unfeasible: 1. Separating track and wheel noise in a single antenna image. The resolution in the desired frequency range is insufficient. It could be possible to separate track and wheel noise by combining antenna results with results from other types of measurement, or by using the radiation pattern of the track. Deliverable D 13 Page 19

20 2. In general, locating sources on the superstructure that are more than 15 db per square meter weaker than the strongest sources. Although sections of the train usually show a larger dynamic range. Also, sources more than 15 db weaker than the strongest source in the entire image may of course still be identified if they are not affected by the focusing effects of a nearby stronger source. 3. Locating very low frequency sources. The location of a 200 Hz aerodynamic source can be determined to a precision not higher than 6 meters, regardless of its strength. This is due to a fundamental limitation that applies to every antenna system at 7.5 meters from the track. There are ways to further improve the performance of the antenna system and to extend the range of tasks for which the antenna can be used (see also [7]), but these are mostly beyond the scope of the METARAIL project: 1. Other sources on the train well below the wheel-rail noise could be detected by screening the wheel-rail area along the track. 2. The effects of ground reflection could be reduced by applying sound absorbing material between antenna and track or by putting an absorbing noise barrier at a strategic location such that only the sound reflected by the ground is eliminated. 3. Ground effects could be taken into account by including more information on the source and ground characteristics in the processing software. 4. The dynamic range of the antenna could be doubled by applying a full 2-dimensional uniform array. This requires many more channels and microphones and would therefore be more costly. The signal processing software would be almost identical to the current software. 5. A solution to the limited resolution is to measure closer to the train at distances of for instance at 1.5 meters. Multi-channel antenna systems are primarily a diagnostic tool most suitable for source detection and quantification, not intended for actual type testing or long term monitoring of railway noise, which are defined as single microphone techniques according to the standards. There are nevertheless links to type testing and monitoring to the extent that data obtained from antenna measurements can provide additional insight into local sources on the train. Although it is difficult in principal to separate wheel and track noise with an antenna, some techniques have been developed which provide better resolution, such as swept focus and pass-by averaging. The spatial track vibration method also developed in METARAIL Deliverable D 13 Page 20

21 supplements this. Both techniques showed for the first time the radiation angle of sound from the rail. 2.5 Separation of track noise with spatial track vibration A new measurement technique was developed in task III.3 (see [8]), to obtain the noise contribution from the track from multi-channel rail vibration, in combination with measured vibro-acoustic transfer functions determined under static conditions. An advantage of this approach is that the track noise can totally and explicitly be separated from the wheel/vehicle noise. No contribution of wheel noise is present in the final result. A disadvantage of the method could be that a part of the measurements are performed without a train present on the track, which could affect track dynamics. Nevertheless, the technique is expected to be rather robust and the impacts of these effects are assumed to be limited. The method is an application of the so-called equivalent forces method, specially adapted and implemented for track noise analysis. One important new development is to take the moving excitation of the train into account. Although this indirect technique can be characterised as an advanced tool for train noise analysis, the measurement equipment needed and measurements themselves all rely on standard equipment and standard functions. Measurements can be performed using ordinary two-(or more channel) FFT-analysers, in combination with digital recorders which are readily available on the market. Specific expertise and software is however required for the data processing and interpretation. The method has been demonstrated during field measurements, at the time of the 2nd METARAIL measurement campaign near Deurne, The Netherlands, September Deliverable D 13 Page 21

22 The results from this demonstration show that the new tool can give a thorough analysis of the track sound emission during a train pass-by. This enables a quantification of the track source strength in terms of sound power level per metre track. Furthermore, details on track radiation and 3-dimensional directivity can be obtained and visualised. It can therefore be expected that the method is able to assess in detail the effect of particular measures applied to a train or track to reduce noise emission levels. Quantifying track noise and at the same time measuring total noise can also give information on sound radiated by the passing vehicles, based on the difference between total noise and noise from the track (of course the accuracy on the final results must be taken into account before making such a subtraction ). Results from the spatial track vibration method are illustrated in the figure below. Figure 2.6 : Example of reconstructed sound pressure distribution caused by the track only over the measurement surface. In this example one bogie is present in the measurement section. The wheel positions are indicated by the dotted circles. The reader is referred to [8] for more detailed information. Deliverable D 13 Page 22

23 The technique presented was applied for one case. It is recommended to increase experience with the method, especially for different kinds of track. In [8] the method was applied extensively, resulting in much (visual) detail on track noise. All kinds of simplifications of the method can be made, which can decrease the measurement effort significantly. In fact, based on this method a rather simple measurement method can be derived to measure the relation between rail vibration and total sound radiation of the track only. Such a measurement technique could serve in type testing measurements in order to separate track and vehicle noise. That technique could be an alternative for the method using a reference vehicle described in the following section. Further research on this point is required. 2.6 Separation of track noise by using a reference vehicle During the evaluation work performed on measurement techniques in task III.4, a relatively simple, operational method for separating track and vehicle noise contributions was conceived. The method is described here, as it is not included in earlier METARAIL reports. Sound pressure and vertical railhead vibration are measured simultaneously. First, the track behaviour is determined in a separate reference measurement using specially prepared reference vehicles which radiate substantially less noise than the track, irrespective of the wheel and rail roughness levels. Then the vehicle and track noise levels can be derived for arbitrary vehicles. Applicability and accuracy The method is applicable for any track-vehicle combination, as long as the 1/3-octave SPL spectrum of the reference vehicle is at least 10 db below that of the track. If the 10 db difference cannot be achieved, either lower accuracy should be accepted or alternative measurement procedures should be applied. There is no restriction on roughness levels of the vehicle or the track, although very poor surface conditions and unwelded rail joints must be avoided. The method is most suited for microphone distances of 7.5 m or less, as at larger distances the low vehicle contribution may be contaminated by adjacent noisier vehicles. Consequently, at larger distances such as 25m a much longer set of reference vehicles is required. The accuracy of the vehicle noise level is 0.5 db(a) if the reference vehicle is at least 10 db(a) quieter than the track. Deliverable D 13 Page 23

24 Variation in track behaviour can occur, depending on position along the track, and changes in track (pad) behaviour due to temperature and loading. L tot Reference quiet vehicle : Small, damped and shielded wheels, no L veh L track L = L L Href ptr, ref vr, ref L tot Normal test vehicle L veh L track Figure 2.7: Schematic illustration of reference vehicle in comparison with normal test vehicle Measurement of track reference function A track reference function L Href is measured, defined as: L Href ( ptr, ref vr, ref f f ) = L ( f ) L ( ) (2.6) measured during passage of several reference vehicles, where L ptr,ref = track sound pressure spectrum at 7.5m and L v,ref = railhead vertical vibration velocity spectrum with both spectra in 1/3-octaves, either both linear or both A-weighted. Deliverable D 13 Page 24

25 The rail vibration velocity spectrum can be obtained from acceleration by integration: L ( f ) L ( f ) 20lg 2πf v = (2.7) a with f= 1/3-octave centre frequency. There should be at least 3 reference vehicles with a total length of at least 45m. These vehicles must be at the end of the train, without any trailing vehicles or locomotive, unless it can be shown that no contamination of the noise from adjacent vehicles occurs. Data is acquired from the second reference vehicle to the last, at 80 km/h and optionally at other speeds. Figure 2.8: Train configuration for track reference function measurement; at least 3 reference vehicles with at total length of at least 45 m are used, positioned at the end of a train. The first reference vehicle is not included in the measurement. The track reference function is characteristic of the track type and the local conditions. Optionally, the effect of local conditions can be reduced by repeating the measurement at 2 or 3 track cross-sections. Estimates for the track reference function are shown in the figure below for the round robin campaign, which were obtained using a quiet (shrouded) reference vehicle. Deliverable D 13 Page 25

26 LpHref(est.) db re 20 Pa/m/s Estimated reference function for each country (based on SGS wagon) Austria NL France Italy Frequency [Hz] Figure 2.9 Measured track reference functions for the round robin test sites An alternative method to obtain a maximum level for the track reference function is by monitoring the transfer function from rail vibration to sound pressure at the measurement position, over a substantial number of train (or bogie) types and speeds; the minimum value (taken per 1/3-octave frequency band) gives an upper limit for the track reference function. [ L ( f ) L ( f )], i 1 n LHref ( f ) ptot vr =,... (2.8) i where n is a number of vehicle pass-bys containing some vehicles with wheels that produce less noise than the track. Data to be reported The following data should be registered and reported. - the track reference function L Href in 1/3-octaves, in db re 20 Pa/m/s - the equivalent rail vibration velocity spectrum, in db(a) re 10-6 m/s - the speed(s) and description of the reference vehicle applied for the reference measurement, or the procedure used to obtain the function estimate. Deliverable D 13 Page 26

27 Specification of the reference vehicle The reference vehicle should be at least 10 db quieter then the track over the whole frequency range. On conventional track this can be achieved by fitting the wheels with absorbent partial enclosures down to 10cm above the rail, and fitting ring dampers on the wheels. A better alternative for this is to use of wagons with small massive wheels, such as those used in low floor truck transport wagons (e.g. type Saadkms; wheel diameter = 36cm, wheel mass = 115 kg ). Below 125 Hz low frequency vehicle noise may occur on vehicles with superstructure, so flat vehicles with soft suspension bogies are preferable. Wheel roughness should not differ substantially on either side of the reference vehicle, so disc-braked or unbraked vehicles should be used. It is of course necessary to ensure that the reference vehicle is indeed quieter than the track and to quantify this. There are several ways of doing this: 1. static reciprocal measurements; a stationary wagon is irradiated by a powerful sound source; the transfer functions from the sound source to the vibration responses on the wheels, rails (and if necessary, wagon superstructure) provide the relative contributions in sound radiation under operational conditions, for given roughness excitation; 2. apply the pass-by spatial vibration method to derive the track contribution and obtain the vehicle contribution from the track contribution and the total noise level; 3. measure the railhead vibration and calculate the track contribution using TWINS; detailed input is required for this. Determining track and vehicle noise levels For any train passage at the same location, the track and vehicle noise spectra can now be derived as follows: and L L ( f ) L ( f ) L ( f ) = (2.9) ptr Href + pveh L vr ( f ) /10 L ( f ) /10 ptot ptr ( f ) = 10lg(10 10 ) (2.10) processed as unweighted 1/3-octave spectra, and afterwards applying A-weighting to L ptr and L pveh. Deliverable D 13 Page 27

28 2.7 Train speed measurement and vehicle identification Within task II.4, methodologies for speed and train detection were investigated. There are several principal ways of speed detection by radar, laser or measuring time for a defined length. Commercial products are available for all different principles which can be applied easily for type-testing and monitoring purposes as well. Train detection is a more sophisticated task. Almost all European railway operators uses their own train control and engine systems that contain useful information about trains and vehicles. There are differences in data format, type of information included and media used. In Austria, for example, the RZÜ and ARTIS systems can provide for all information necessary for noise monitoring electronically. There are attempts to introduce European centralised traffic control. However, it is not yet known when such a system will be available. Maybe the introduction of some other ways of automatic vehicle identification, which do not use traffic control data but bar-codes, chips, or other devices on the vehicle, will solve the problem earlier. There are some train noise and/or vibration monitoring systems on the market available or under development, which already often have different systems for train and speed detection. An automatic train noise monitoring station needs a lot of non-acoustic information. Therefore it is essential that there is a defined interface between the acoustic and the non-acoustic modules. A minimum and a recommended data interface were defined allowing a standard data transfer from the vehicle detection module to the acoustic module of a monitoring system. Information included is based on UIC guidelines and the HERMES train data exchange format which both are commonly used in Europe. Deliverable D 13 Page 28

29 # # # # # Content record # 0 $ $ $ $ $ $ $ $ Train descriptor record # : 0 0 reference pass-by time [hh:mm] record # 2 total train lenght (optionally; 0 if not available) record # 3 total number of axles (optionally; 0 if not available) record # 4 total number of vehicles (optionally; 0 if not available) record # _ UIC vehicle number of first vehicle (in the direction of traffic) record # _ UIC vehicle number of second vehicle (in the direction of traffic) : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : record # n _ UIC vehicle number of vehicle n-1 (in the direction of traffic) record # n _ UIC vehicle number of last (nth) vehicle (in the direction of traffic) Description digit not used digit optionally used Table 2.II: Record format proposed for minimum requirement interface record # 0 $ $ $ $ $ $ $ $ ; 0 0 : 0 0 ; 0 0 record # _ ; 0 0 ; ; ; record # _ ; 0 0 ; ; ; record # _ ; 0 0 ; ; ; : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : record # n _ ; 0 0 ; ; ; record # n _ ; 0 0 ; ; ; Description record # 0 col 1..8: train descriptor col 9: delimiter col : reference pass-by time [hh:mm] col 15: delimiter (optionally) col : total number of vehicles (optionally) col : not used record # 1..n col 1..14: UIC vehicle identification number col 15: delimiter col : number of axles col 18: delimiter col : vehicle length [dm] col 22: delimiter (optionally) col : empty vehicle weight [t] (optionally) col 26: delimiter (optionally) col :weight of loaded vehicle [t] (optionally) Table 2.III: Record format proposed for recommended interface Deliverable D 13 Page 29

30 2.8 Characterisation of track dynamics Track dynamic behaviour has major influence on the noise contribution of the track, and therefore influences reproducibility when the track noise exceeds vehicle noise. An example of the effect of track parameters on the total noise levels is given in table 2.I. For this reason, it is required to somehow characterise the track. The way in which this can be done may depend on the further use of the quantities measured. Modelling experience indicates that for the track dynamic behaviour the most relevant parameters are: rail geometry and material, pad static and dynamic stiffness and loss factor, fastener type, sleeper geometry, material and distance. If these parameters are known, track response calculations can be made with models such as TWINS, based on nominal input data. At a specific site, particularly the values for pad and fastener behaviour may vary from nominal values, depending on local variations due to alignment, age and maintenance. The real track can be characterised in situ by measuring under stationary conditions on unloaded track: - vertical and horizontal railhead impedance (contains pad stiffness and damping) - vertical and horizontal spatial decay (takes pad and fastener behaviour into account); Impedance (or mobility) and spatial decay can be measured by means of impact response measurements in the frequency region Hz. More representative characteristics can be obtained during pass-by by measuring vibration isolation and vertical and horizontal railhead vibration. Dynamic pad stiffness and damping can be estimated from vibration isolation, whilst spatial decay can be estimated from vertical and horizontal railhead vibration. It should however be mentioned, that measurement of spatial decay is still not a standardised procedure. One way of estimating spatial decay applied in METARAIL is to determine the slopes on either side of a wheel passage in the vibration signal, for each 1/3-octave frequency. If the individual wheel peaks cannot be distinguished in the vibration signal, then a longer distance should be used, for instance the approach of a vehicle, or the gap in between passing bogies. The reference function defined in section 2.6, could also be adopted as an overall track characteristic function, as it easy to measure and contains all relevant track properties, representative for operational conditions. Vibration isolation is also easy to measure under operational conditions and is a useful indicator for track vibro-acoustic behaviour. Deliverable D 13 Page 30

31 It should nevertheless be emphasised that the variation from point to point on the track can be substantial, and that when measuring mobility, vibration isolation or the track reference function, it is necessary to average at least 3-5 measurement points to obtain representative data (see [15]). 2.9 Site sound transmission When comparing measured SPL data between 2 sites, differences in measured levels may occur due to ground absorption and reflections, in particular at measurement distances of 25m or larger. A simple calibration measurement can be performed to measure the sound transmission by placing an artificial broadband sound source on the track measurement cross-section in the middle of the track at average axle height, 50 cm above the rail surface. If differences are found between 2 sites, corrections can be made if necessary. This method is illustrated in [2] and was performed during the round robin test ([15]) New concepts in relation to type testing One of the major questions arising from the methods described above, is how they can be used in the scope of type testing, so as to improve reproducibility. The figure below illustrates that 3 parameter groups for the track (left) and 3 for the vehicle (right) affect the overall sound pressure level. This already suggests why spread in measured data can be so large, if all these effects are not kept constant. Each of the arrows can be characterised either by a spectrum or a single number (e.g. vehicle speed). So theoretically it is possible to characterise the vehicle and track in terms of roughness and vibro-acoustic behaviour. Once these are available, the track and vehicle contributions can be derived. Deliverable D 13 Page 31

32 Rail roughness Wheel roughness Track vibroacoustics L paeq Vehicle vibroacoustics Site effects Vehicle speed Figure 2.10 Main influence factor categories for railway rolling noise Now it is possible to separate track and vehicle noise, the question arises which values to use when comparing noise emission of 2 different vehicle types. It would seem most logical to compare the vehicle contributions at given speeds, and to indicate whether the total noise level is dominated by track or vehicle noise, and to state whether wheel roughness exceeds rail roughness. When considering the concept of vehicle noise classification, useful indicators could be: - the vehicle noise level, and optionally the track noise level (speed and roughness dependent) - the vehicle roughness level (speed and track independent). It is also possible to establish a roughness-independent quantity by subtracting the total roughness level from the sound pressure level and from the rail vibration level. Noise type tests for railway vehicles are intended to characterise the vehicle and not the track, and to give an overall assessment of vehicle noise generation due to wheel roughness and vehicle vibro-acoustic transmission and radiation. Considering the above points, it may be worth taking them into account in future improvements to type testing standards. One issue in this respect is the possibility of defining a standard for acoustic track characterisation, which would include rail roughness, track characteristic function and track dynamics. Deliverable D 13 Page 32

33 2.11 Overall review Summarising all of the developments undertaken so far in METARAIL, it can be concluded that a major step forward has been made with respect to improving reproducibility, accuracy, specification of track conditions and separation of vehicle and track noise contributions. Deliverable D 13 Page 33

34 3 IMPROVEMENTS TO THE TYPE TESTING PROCEDURE 3.1 Introduction During the course of the METARAIL project, some members of the consortium became involved with the CEN/TC256/WG3 working group on Railway exterior noise measurement, which has now linked up to the ISO 3095 standard. Involved partners are ERRI, TNO, NSTO, and IPSE. In this way, results from METARAIL could be put forward to the standard committee. This resulted in proposed modifications and improvements to the ISO 3095 standard, of which the next European standard should be pr EN ISO The proposals put forward in this are not definitive, but are available to the standard working group to evaluate and adopt. Proposed improvements are realistic in terms of the current standards. It would be possible to propose more radical modifications of the standards, depending on future regulations. For instance, it could be considered to only measure wheel roughness and quantify track dynamic and acoustic properties. It should once again be emphasised here that the recommendations made in METARAIL pertain to the constant speed test and wheel-rail rolling noise. 3.2 Current standard and proposed improvements The reader is assumed to be familiar with the ISO 3095 standard and the CENTC256/WG3 document on measurement of exterior railway noise (see [1],[4], [5]). A flow diagram for the proposed improved standard pr EN ISO 3095 is given in figure 3.1. The proposed improvements with respect to the constant speed test are the following. 1. The precision is not indicated in the standard, but the existing standard is at best survey grade (worse than ±5 db) whereas engineering grade is required (±2 db or better). The precision should be indicated. Deliverable D 13 Page 34

35 2. Allow a measurement distance of 7.5m for wagons within a train instead of 25m, as at 25m, adjacent vehicles can strongly influence measured levels, introducing errors. 7.5m is current practice in many cases. This is not required, although possible, for fixed train configurations (such as e.g. 4-coach passenger unit). 3. Allow multiple cross-sections for consecutive pass-by measurements, for instance 3 sections at 25m intervals to reduce the number of pass-bys, to reduce spread in noise levels due to speed variation, and due to variation in site sound transmission. This is in fact not in contradiction with the existing standard. 4. Include a requirement for the number of measured vehicles within a mixed train, as a single vehicle is often insufficient. It is recommended to use at least 4 wagons to avoid masking from adjacent vehicles, and to achieve statistical relevance. 5. Clearly indicate for non-fixed train configurations the section of the group of vehicles measured for the test, which should be from centre to centre. If the last vehicle is at the end of the train, the measurement section should be from centre to end. 6. Introduce more appropriate measurement quantities such as L paeq,t instead of L pafmax for individual vehicles and vehicle groups within a train, T.E.L. (Transit Exposure Level) for fixed train configurations. It may be wise to still include L pafmax besides L paeq,t and TEL, as historically much data was measured in L pafmax. 7. Include a more accurate correction for train speed, as in practice it is often needed and the 30lgV correction is inaccurate. 8. Introduce the derived quantities vehicle noise and track noise, to increase reproducibility, as the total level is sometimes dominated by the vehicle, sometimes by the track. 9. Provide procedures by which these contributions can be determined (for example the reference vehicle method or others; see chapter 2). Deliverable D 13 Page 35

36 10. Add a procedure on how to measure rail roughness, and an upper limit that a test site should not exceed (currently undertaken by the ISO 3095 working group on track conditions, see section 3.7). 11. Describe how track parameters should be characterised and registered, not only the nominal characteristics, but also in situ parameters such as vibration isolation and spatial decay. 12. Include information described above in the type test report. 13. It would be worth considering a separate standard for acoustic characterisation of the track, including track roughness, dynamic and radiation properties. Dynamics and sound radiation could be included in one track characteristic function, a transfer function from rail vibration to sound pressure at the measurement point (such as the reference function, for example). 14. The acceleration test and stationary test could be put into a separate standard, and a brake noise test could be included. 15. Further work is required to analyse the standard specifically for application to urban rail vehicles such as trams, metros and light rail systems. Due to the lower speeds, different track geometries and operating conditions there may be substantial differences in the test conditions. This is not within the scope of the METARAIL project, but is recommended to the ISO working group. Some of the above recommendations are dealt with in the following sections. Deliverable D 13 Page 36

37 Flow diagram for pr EN ISO 3095 Define the train with vehicles to be tested Specify test speeds: 0.95 Vmax, 80, 120, Select measured quantity LpAeq,T, TEL, LpAFmax and measurement section of train Select (new) test site (Flat land and no reflective objects) Visual inspection Visible corrugation? Measure rail roughness (direct or indirect upper limit) Below threshold? no yes regrind and run in yes no Configure noise and vibration setup Multiple cross-sections? Vehicle - track separation? Calibration Choose separation method Configure required vibration channels Background noise check For reference vehicle method, For speed 1 to n acquire reference function Lref Measure Lp, also Lv for separation with passby of reference vehicle Passbys until 3 Lp values within 3 db(a) range Average these (Lp and Lv) Vibration isolation and Lv horizontal+vertical(decay) end Derive Lpveh and Lptr if Lref available Report Site description Measurement set-up Track dynamic properties (nominal, static or operational, e.g. vibration isolation and decay rates) Track roughness or confirmation that track conforms to limit (with reference to report) Average noise and vibration levels (overall and spectra) Level history of A-weighted noise and vibration levels Partial contributions of overall levels Procedure applied and result of separation function Figure 3.1 Flow diagram for the constant speed test in railway noise exterior noise type testing, including roughness measurement and separation of vehicle and track noise (Proposal for pr EN ISO 3095). Deliverable D 13 Page 37

38 3.3 Measured quantities The measurement quantities for sound pressure level should depend on the type of test and the kind of train. From discussions within the ISO 3095 working group it has been recommended to change the measured quantities as follows. Constant speed Stationary test Acceleration and test braking Parts of trains L paeq,t L paeq,t d=25m or d=7.5m d=7.5m Whole trains and TEL L paeq,t L pafmax single vehicles d=25m d=7.5m d=7.5m If L< 50m, d=7.5m Table 3.2 Recommended measured quantities for railway noise type testing Definitions and illustration of these quantities are given below. T L p(t) LpAFmax L paeq,t Figure 3.3 pafmax and L paeq for noise levels of part of a train The A-weighted equivalent sound pressure level over pass-by time T=t 2 -t 1 : L t lg ( pa ( t) p0) dt (3.1) paeq, T = / T t1 This is generally used for measuring from one part of the train to another, for example from centre to centre of the first and last wagon in a group. Deliverable D 13 Page 38

39 The sound exposure level (SEL): SEL = 1 lg T 0 t t ( p ( t)/ p ) dt 10 (3.2) SEL is directly related to L paeq,t by A 0 L paeq, T SEL 10 lgt = (3.3) The transit exposure level (TEL), which is used for fixed train configurations: ( TEL = 10lg pa( t)/ p0 ) dt T p (3.4) In practice, the infinity signs indicate a measurement period including the rise and fall of the pass-by level, which is longer than the transit time (front to end of train). TEL is related to SEL by TEL = (3.5) SEL 10lg( Tp ) Where p A (t) is A-weighted sound pressure, p 0 is reference sound pressure ; T p is the transit time for front to rear buffer of the train, T 0 is reference time = 1 second. L (t) p 7.5m Tp=L/V T=t2-t1 Figure 3.4 Normalisation time for TEL and L paeq Deliverable D 13 Page 39

40 For short vehicles such as locomotives, TEL is recommended, as L paeq,t can be unstable for short T at high speeds, as illustrated below. L (t) p 7.5m Tp=L/V t1 T = t2-t1 t2 Figure 3.5 Effect of short vehicles at high speeds For vibration measurement, the equivalent vertical velocity level L v,veq,t is: L t2 veq, T = / 1 lg T t 1 2 ( v( t) v ) dt 10 (3.6) 0 with v(t) = time signal of vibration velocity, measured for instance on the railhead in vertical direction. The A-weighted level history of sound pressure level L pa (t) and of vibration velocity level L va (t) are defined as: L and L 2 2 ( p ( t)/ ) ( t) 10lg p pa = (3.7) A ( v ( t) / ) ( t) 10lg v va = (3.8) A 0 with t = t 0 + i* t, i= 0,1,2,3,. and sample interval t=1/32 second. This can be obtained with analysers capable of 1/32 second sample time and linear averaging at 1/32 second. Deliverable D 13 Page 40

41 A spectral map of vibration velocity L v (t,f) is defined as the 1/3-octave linear (unweighted) spectrum of the velocity level: 2 2 ( v ( t, f ) / ) L ( t, f ) 10lg v v = (3.9) c c 0 with f c = 1/3-octave band centre frequency. The spectral map can be derived from the vibration time signal, and can be used to estimate the spatial decay in horizontal and vertical directions. The vibration isolation between rail and sleeper D tr is defined as the difference between the railhead vertical vibration spectrum (unweighted, equivalent) L v,rail,vert (f c ) and the sleeper vertical vibration spectrum (unweighted, equivalent) L v,sl,vert (f c ) near the fastener, both during pass-by: D tr( fc v, rail, vert c v, sl, vert c ) = L ( f ) L ( f ) (3.10) Velocity can also be substituted by acceleration here. 3.4 Speed correction The existing standard gives a correction in measured overall noise level for speed of 30 lg(v 1 /V 2 ) which can be inaccurate. In reality this correction term varies between 20 lg(v 1 /V 2 ) and 30 lg(v 1 /V 2 ) and can even be speed dependent. It is more reliable to give a spectral correction, which can take the total roughness into account. A simplified formula is given here which is more accurate than the 30 lg V correction or linear interpolation. The formula can be derived by using formula (11) or (12) for indirect roughness from report D8 [9]. The correction in noise level from one speed V 1 to another speed V 2, L p (V 1,V 2 ) can be expressed in terms of maximum vertical rail vibration level: V2 V2 Lp( f, V1, V2) = 20lg( ) + Lvmax, V1( f ) Lvmax, V1( f ) (3.11) V V 1 1 with L vmax,v1 (f) = maximum spectrum of vertical railhead velocity measured at speed V 1. Deliverable D 13 Page 41

42 This formula is applicable for noise levels and vibration levels for whole trains or vehicles or wheels. Note that a frequency shift within the same spectrum L vmax,v1 (f) is used, which indicates that for higher speed, a different part of the roughness spectrum produces a different level at the same frequency f. 3.5 Site description The site description should be comprehensive and include the information shown in section 5.2. Much of the data is nominal, for example pad stiffness is the value specified by the manufacturer and may vary at the actual site itself. 3.6 Measurement set-up In addition to the set-up with a microphone at 7.5m/1.2m or 25m/3.5m, data at extra cross-sections and vibration data can be acquired as illustrated below. B A D1 D2 C2 25m C0 25m C1 M2 7.5m, 1.2m M0 7.5m, 1.2m M1 7.5m, 1.2m M3 25m, 3.5m Figure 3.6 Comprehensive measurement set-up with 3 track cross-sections C0,C1, and C2, taking variations in track behaviour and sound transmission into account. This type of set-up allows pass-bys to be registered simultaneously instead of sequentially, with less speed control problems, thereby increasing accuracy. Deliverable D 13 Page 42

43 B2 B3 Cross-section C0 A2 A3 B1 A1 Cross-sections C1,C2 A2 A3 B1 A1 Figure 3.7 Vibration measurement points for a comprehensive measurement setup. Rail vibration is measured close to the sleeper. Horizontal measurements are not strictly required. 3.7 Track and vehicle noise Directly measured sound pressure levels of any type (L paeq, L pamax or TEL) quantify the total noise level which consists of sound radiation of the vehicle and the track. The partial contributions of vehicle and track can be derived by various methods, described in chapter 2. These contributions are called the vehicle noise level and the track noise level, and have the symbols given below. Total noise level Track noise level Vehicle noise level Equivalent noise level L paeq,t L paeq,tr L paveh Maximum noise level L pamax L pamax,tr L pamax,veh Transit Exposure Level TEL TEL tr TEL veh Table 3.I Annotation for track noise and vehicle noise levels Deliverable D 13 Page 43

44 They provide additional information which is of use when comparing results from different sites, or from different pass-bys at the same site. The partial contributions are unique for a given microphone position, and may differ in ratio at different microphone positions and at different speeds. For a given microphone position, if the total noise level L ptot and one of the partial noise levels, e.g. L ptr are known, the other partial noise level can be derived according to: Lpveh L ptot / 10 Lptr /10 = 10lg(10 10 ) (3.12) where L ptot = total noise level at a given position, L ptr = track noise level at the same position, L pveh = vehicle noise level at the same position, all as either equivalent, maximum or TEL levels. The same type of level should be used in the formula. For conventional vehicles and tracks the partial noise levels can be similar in magnitude, which is why the effect of noise control measures on only the vehicle or only the track have a limited effect. Reporting of derived quantities When reporting partial levels, the total level Lptot, the vehicle level Lpveh and the track level Lptr should be stated as A-weighted equivalent, maximum or TEL levels, for each measured speed. Optionally, the partial contributions can additionally be presented in octave or 1/3-octave spectra. The procedure used to derive the partial levels should be indicated. Deliverable D 13 Page 44

45 3.8 Track conditions - roughness The text in this section is put forward as draft proposal for improvement of the next European standard pr EN ISO 3095 / The proposal specifies the protocol for measuring rail roughness in order to compare the data in a consistent manner to an upper limit. This upper limit guarantees that the overall change in rolling noise due to the variability of wheel/rail combined roughness will be less than 4 db. The rail roughness measurement protocol is one of the main improvements to the standard helping to increase the accuracy and reproducibility of type testing procedures to comply to engineering grade (± 2 db) instead of survey grade (± 5-10 db). Where the following text refers to indirect measurements compared to direct measurements, conventional indirect measurement techniques are referred to, not the new technique developed in Metarail as described in section Measurement location and protocol Lateral position on the railhead On a straight track the wheel runs on a clearly visible running band, usually situated near the centre of the rail head. The running band can be as wide as 60 mm (for old track) or as narrow as 10 mm (for new track). Rail roughness has to be measured on a line in the centre of this running band. If the running band is wide enough, two supplementary parallel, equidistant lines at either side of the centre line must be measured. The distance between the centre of the running band and the supplementary measurement lines depends on the width of the running band: running band width 10 mm: measurement of 1 line 10 mm < running band width 20 mm: measurement of 3 lines, 5 mm equidistant running band width > 20 mm: measurement of 3 lines, 10 mm equidistant Width and position of the running band must be checked on different cross sections of the test site to comply with variations along the test track. Deliverable D 13 Page 45

46 Position along the track Background information For exterior and interior noise measurements, the roughness of the track in the vicinity of the noise measurement location and the track section at which the interior noise measurements are performed, has to be determined. The objective of the track measurement protocol is to characterise the rail roughness at a certain track section without measuring the roughness of the entire track in detail. Therefore the test track is divided in cross sections located at specified intervals from the exterior noise measurement section. Roughness samples are taken at these cross sections. Figure TEST TRACK indirect roughness measurements direct roughness measurements REFERENCE section exterior noise microphone Figure 3.8 Track sections relevant for rail roughness measurement There are two ways to perform the roughness measurements: direct roughness measurements (aimed at exterior noise measurements); indirect roughness measurements, combined with direct roughness measurements (aimed at interior noise measurements). The second option provides an alternative for taking direct measurement samples along the entire test track, by combining direct roughness measurements at the exterior noise measurement location with indirect roughness measurements at the entire test track. In this procedure, the section where the direct roughness measurements are performed acts as a reference section. Deliverable D 13 Page 46

47 Direct roughness measurement Direct roughness measurements are performed with a standard roughness measurement instrument placed on the railhead. The instrument contains a measurement probe that is guided along the rail, measuring the relative height of the railhead in micrometers. The instrument provides data ranging from 0,1 cm to 10 cm wavelength and shall comply to a relevant standard (to be determined). The direct measurement protocol is illustrated in the figure below. REFERENCE section r r r r l l cross section r measurement length l exterior noise microphone 3 parallel, equidistant lines Figure 3.9 Measurement positions for direct rail roughness measurement The reference section is divided into 6 cross sections. The middle cross sections are positioned at the centre of the reference section, nearest the microphone (cross sections 3 and 4). The other cross sections are positioned at distances of r and 2r from the centre of the reference section. The length of the reference section to be measured is taken proportional to the microphone distance r from the track and varies from -2r to +2r relative to the centre of the reference section where the noise measurement microphone is positioned. For the nominal microphone distance of 25 m therefore, roughness is measured over a distance of 100 m. Deliverable D 13 Page 47

48 For each cross section and each rail 3 parallel lines (1 if running band width 10 mm) shall be measured, leading to a total of 36 lines. A description is given in the section on Lateral position on the railhead. The measurement length l of the line of which the roughness profile is measured shall be at least 1 m. Indirect measurement When exterior noise measurements are performed at cross sections of the track differing from the reference section, or if (simultaneously) interior noise measurements are performed, direct roughness measurements have to be performed also for the applicable track sections where these noise measurements are performed. Alternatively, if direct roughness measurements are performed only for the reference section, data may be collected for alternative parts of the test track by indirect measurements. Indirect roughness measurements can be performed by measuring noise or vibration with an axle-box accelerometer or a microphone located under the train or a microphone located in a passenger coach. The wheels of the train should be permanently smooth, disc or sinter-block braked or unbraked, to avoid the influence of wheel roughness. The indirect roughness signals shall be recorded along the entire track where the noise measurements are performed, including the reference section where the direct measurements have been performed. The train speed must be recorded to allow conversion from frequency spectra to wavelength spectra Processing of roughness data Direct measurement A third octave roughness wavelength spectrum is calculated from each measured roughness line. For each cross section the energy average of the roughness spectra is calculated. The average direct roughness spectrum valid for the reference track section is the energy average of all calculated roughness spectra. It is demonstrated that large differences in roughness level may result from different processing methods. The main problem is the correction for pits and spikes in the signal that occur due to e.g. rail burns in the measured roughness spectra. Beyond a certain Deliverable D 13 Page 48

49 depth and width, the wheels will not follow the spike profile, and hence will not vibrate accordingly. Neglecting these pits and spikes during processing will cause artificially high roughness levels resulting in non valid analysis values and possibly a rejection of the test track as a result of exceeding the limit roughness values. Methods that can be used for the removal of pits and spikes are not yet standardised. General information on processing and correction methods is given in [13]. Indirect measurement The indirect measured signal for each alternative measurement section is analysed separately and the energy average is calculated. The measurement time interval shall comply to section of pr EN ISO Indirect data measured for an alternative section shall be compared to the indirect data measured for the reference track section where the direct measurements have been performed. The difference is added to the average direct measured roughness spectrum of the reference section and compared to the limit roughness spectrum. To convert frequency spectra to wavelength spectra, the relation λ= V / f is used, with V the recorded average train speed. In order to cover a wavelength range of 1 to 10 cm for measurements in frequency bands upto and including 10 khz, the maximum allowable train speed for indirect measurements is 360 km/h. For assessment of wavelengths lower than 1 cm (downto and including 0.25 cm), train speed is restricted to a maximum of 90 km/h. Deliverable D 13 Page 49

50 A schematic overview of the procedure is given in the following figure. INDIRECT roughness measurements Reference section Alternative section DIRECT roughness measurements db direct average reference section db indirect average reference section db indirect average alternative section limit - λ f f indirect average correction delta (alternative-reference) λ = v / f - speed v db corrected average alternative section limit λ Figure 3.10 Procedure to derive rail roughness by comparing indirect data at one location to the direct roughness at the reference section Test section approval The direct measured average roughness spectrum is compared to the limit roughness spectrum as described in section of the main text of the standard. The demand of homogeneity of the test track implies that the roughness levels measured at the different (cross) sections along the track should not exceed the limit roughness spectrum. However, as small variations are inevitable, the following criterion is applicable when the limit roughness spectrum is partly exceeded. Deliverable D 13 Page 50

51 The test section is approved if the following criterion is met: For each section and third octave band, the level of the average roughness spectrum with third octave band centres at wavelengths between 1 and 8 cm shall not exceed the limit roughness spectrum with more than 6 db peak level in case of a single band or 3 db peak level in case of a maximum of 3 adjacent bands over the wavelength range, or the combination of both. Only one event of a single band, 3 adjacent bands or a combination of these bands exceeding the limit is allowed. max 6 db max 6 db db APPROVED db APPROVED db REJECTED max 3 db max 3 db max 3 db limit limit limit max 1 band, 3 bands max 3 bands 3 bands 2 bands λ λ λ Figure 3.11 Approval criteria for the rail roughness at the test section Data presentation Results shall be presented in third octave bands with rail roughness level (in db re 1 µm) as a function of wavelength (in cm), in decreasing order. The wavelength range shall at least contain the wavelengths between 8 cm and 1 cm, covering 10 third octave bands. The ratio of the horizontal and vertical axis shall be 3 : 4 (1 octave : 10 db), as specified in ISO 3740 and IEC Publication 263. Numbering of the wavelength labels should comply with the preferred frequencies of ISO 266. A sample graph is shown below. Deliverable D 13 Page 51

52 20 db re 1 micron 10 ISO limit Roughness level 0-10 Austria Netherlands France Italy , , , Roughness wavelength [cm], Frequency [Hz] at 90 km/h Figure 3.12 Sample graph of roughness spectra presentation (speed indication is optional) Upper limit for rail roughness from monitoring In addition to the rail roughness measurement methods described above, it is also possible to estimate an upper limit for rail roughness at one site by monitoring smooth wheels of various trains and selecting the lowest recorded total roughness at a given speed. At least 3 vibration measurement points along the rail should be included to ensure sufficient accuracy. In addition, to obtain an estimate over a longer piece of track as illustrated in the previous sections, even more vibration measurement locations will be required. This method is based on the indirect roughness measurement technique described in [9] and still needs to be evaluated experimentally. It is therefore not yet proposed as an alternative method in the standard. 3.9 Track conditions - dynamic properties A discussion on track properties is given in section 2.8. It would be preferable to just measure the track reference function and the vertical vibration isolation between the rail and sleeper from pass-by vibration measurements. These quantities should be averaged over a number of positions on the track, and if possible also for various train types and speeds. Deliverable D 13 Page 52

53 4 MONITORING METHODOLOGY, SYSTEMS AND APPLICATION 4.1 General aspects The term railway noise monitoring has several different meanings: a) to repeat the type testing procedure periodically, e.g. after overhaul or maintenance to check actual noise emission of single vehicles. b) to measure noise emission of individual rail vehicles under normal operational conditions whenever passing by a monitoring site. c) to measure noise impact at a monitoring site to check the overall equivalent level of all rail traffic during a specified period (e.g. day/night). The term noise monitoring is used for all three of these activities. In this report, monitoring means measuring rail vehicles emission under normal operational conditions whenever passing by a monitoring site to check single vehicles noise emission. Monitoring of railway pass-by noise can be performed in different ways, depending on the purpose of the application. Either noise immission at receiver positions (e.g. near dwellings) or emission levels closer to the train are measured. Emission monitoring can be used for assessing daily noise exposure levels, noise levels of whole trains or individual vehicles or train parts. If microphones are put sufficiently close to the train or if vertical rail vibration is measured, individual wheels and bogies can be assessed. By measuring noise levels of individual trains or vehicles, it is possible to use a monitoring system for rail pricing, if clear assessment criteria have been set. Besides measuring noise levels from individual vehicles or wheels, vibration levels and spectra can be used for condition monitoring, once again only if clear criteria are given. Examples of monitoring applications are given in table 4.I. The use of vibration measurement depends on whether more detailed information is required, such as separation of vehicle and track contributions, or wheel roughness monitoring. Deliverable D 13 Page 53

54 Immission Emission Microphone Accelerometer Purpose Quantity Daily exposure Daily exposure 1 0 LC LpAeq Whole trains LC,DN LpAeq Parts of trains/wagons 1 1 LC,DN,RP,CM LpAeq Bogies/wheelsets/wheels 1 1 or more DN,CM LpA(t),LvA(t) Type test 1 or more 1 or more LC,AT LpAeq,LvAeq Table 4.I Examples of monitoring applications for railway pass-by noise. LC=Legal conformityto noise limits; DN= Detection of noisy trains, vehicles, bogies or wheelsets from a group; RP=Rail pricing; CM=Condition monitoring; AT=Acceptance testing. Besides acoustic quantities such as noise and vibration, most monitoring applications require measurement of train speed, train and vehicle identification, time and date, temperature, wind speed and direction (especially at 25 m distance or more). Humidity and air pressure have a negligible effect on measured noise at these distances. In contrast to type testing, monitoring is a continuous activity, usually performed on various passing traffic. Type testing can be regarded as a special form of monitoring, whereby a predefined train configuration is measured under more closely defined conditions. As monitoring is performed over a longer period and collects more data, it is preferable to restrict the number of measurement channels and limit the amount of data registered. In general, monitoring systems have to be robust and automated. Data can be processed locally and transmitted by telephone or by wire, or if it is limited, collected locally on a data storage unit. A flow diagram for a general approach to monitoring railway pass-by noise is given in figure 4.1. A separate flow diagram for type testing is also given in the previous chapter. Deliverable D 13 Page 54

55 Rail roughness Wheel roughness Track vibroacoustics L paeq Vehicle vibroacous Site effects Vehicle speed Figure 4.1 General procedure for monitoring of railway pass-by noise. 4.2 The METARAIL monitoring system The automatic noise monitoring system developed within the METARAIL project and described in report [10], has been designed as multi-purpose, low cost and easy-tohandle device, with a good reliability in measurement results. Deliverable D 13 Page 55

56 The main requirements of the system were to allow monitoring performance with minimal costs, both from the point of view of instrumentation and manpower, as the system is designed to be handled by nontechnical staff; to provide a set-up adaptable to the purpose of the measurements. Its field of application can therefore be wider than a conventional monitoring system, designed for detecting a well defined and limited number of parameters. For this reason it seems useful to provide some definitions, to give a clearer understanding of the purposes of the system and the performance it can supply. In previous reports a distinction was made between monitoring, type testing and diagnostics. There is sometimes an overlap between the three. For example, strictly speaking, the definition of monitoring should be the systematic detection of a number of parameters connected with a defined kind of event. This definition is still quite general, and it is clear that the monitoring method can vary according to the requirements and the parameters to be detected. Type testing and diagnostics can also be defined as a particular kind of monitoring. Therefore, in designing the METARAIL monitoring system some design limits were required. More specifically, "monitoring" has been restricted to the detection, within the ordinary railway circulation, of a limited number of parameters connected with the event "train pass-by", which do not require further post-processing and which can provide the following data: train length and speed; type recognition of individual vehicles (goods wagons and passenger coaches) the noise emission or immission of the whole train in terms of L paeq or L pamax ; noise emission of individual vehicles within a single train with the above parameters; the total railhead (vertical) vibration level for the whole train or individual vehicle; A-weighted sound pressure level or total railhead vertical vibration time history. Deliverable D 13 Page 56

57 The above data can be used for the following purposes: 1. Evaluation of daily exposure of the buildings surrounding the investigated track; 2. Separation of the railway noise contribution from other acoustic sources in the vicinity (e.g. roads, motorways, fixed noise sources); 3. Creation of a database which allows to couple single trains or single vehicles to their respective noise emission 4. Location, within a single train, of vehicles of which the noise emission exceeds limits fixed by local legislation, in order to define rail pricing. This type of application assumes, however, the possibility of correlating the acquired database with other databases in belonging to the railway companies, containing data identifying the vehicles in individual trains. If the monitoring is performed along test tracks and on particular vehicles in a particular condition (e.g. new vehicles during delivery) rather than on ordinary transport along normal tracks, and with methodologies which follow some standard (see Chapter 3) we can call it "Type Testing". As in such cases the purpose is, generally, to verify the correspondence of the noise emission of the tested vehicle, or better, of the vehicle type, to a noise limit in accordance with the international standards, it is generally necessary to acquire a greater amount of data than for a normal monitoring activity and to perform more post-processing (e.g. frequency analysis). For this purpose the monitoring system will therefore have to be adapted to the wider needs and to be equipped, for example, with a tape recorder which allows stored data post-processing and a greater number of acquisition channels. Besides the simple performance of an acoustic test and the subsequent acceptance of the vehicle, another important application of type testing is the creation of a database containing the acoustic and vibrational signature of the tested vehicle. Finally, measurements can be used for condition monitoring of vehicles in ordinary service, for maintenance purposes. This covers a large application field, which can be summarised as follows: Deliverable D 13 Page 57

58 location of the vehicles whose noise emission exceeds the fixed limits: this function is the same as described for simple monitoring; location, on those vehicles, of the sound sources that cause the increase in noise emission (wheel flats, superstructure resonances, etc.); monitoring of the noise emission increasing trend of a defined vehicle with its progressive wear (predictive maintenance): such application is mainly based on the databases created during type testing, that is to say on the noise emission characteristic of the newly built vehicles and furthermore on the possibility of exact identification of their railway company and serial number while within an ordinary train. While type testing can require some post-processing of the acquired data, but not necessarily in real-time, the procedure defined above implies the need to operate in real time on each train pass-by, as done for normal monitoring. The most critical aspect of permanent monitoring is to find a way to control all the meteorological, operational and site related conditions influencing the SPL measured. While type testing is performed under ideal conditions, permanent monitoring needs to be done also when raining, with snow on the ground or the train braking. This needs a lot of additional information to make the monitored SPL reliable by excluding all results contaminated by external effects. With respect to normal monitoring, both a more refined analysis of the train itself is required (e.g. evaluation of noise emission of each wheel) and an extremely efficient interface with the available databases of the different European railway companies. It is now clear that - at least for the present - the diagnostic activity can be carried out only partially, as a consequence of the nearly complete lack, with some exceptions, of the above cited databases. In the future those databases will be hopefully created, but before it will be necessary to define their standard structures, so that it will not be required to use various software. Furthermore, as stated in an earlier METARAIL report [2], studies must be performed in order to define a vehicle recognition system within a train that does not require the access to a database, that is to say that each vehicle should be independently recognised, by reading, a bar code for example (laser reading techniques). Deliverable D 13 Page 58

59 Besides the above described applications of the monitoring system, slight improvements to the software/hardware could allow gathering of data for: wheel roughness monitoring: this requires the 3 channels to be fully employed for vibration measurement. This kind of measurement requires also that the rails are smooth; rail roughness threshold detection from varying traffic: similarly to the previous point the measurement of vibration on 3 channels will be required. This will be possible with smooth wheels; vibration isolation monitoring: this will imply simultaneous vibration measurement on 2 channels. With the connection of an external data acquisition unit allowing data post-processing, other applications could be possible, like spatial decay estimation. 4.3 System configurations The monitoring system developed in the METARAIL project can be applied for simple monitoring purposes, with the possibility to be extended for type testing. Wheel or vehicle condition monitoring applications are not yet included as more complex software is required, and the required databases are not yet available or accessible through computer networks within European railway companies. Taking into account moreover that: one of the main objectives of the project is the realisation of a low cost and easy-touse system, the potential users can have different requirements, a non-modular system containing all the options, besides not fulfilling the above objectives, would be cumbersome, the best solution appeared to be the design of a modular system, so that the single user could obtain just those components which fulfil his needs. A more detailed description of the developed system given in an earlier METARAIL report [10]. An overview of its main units and component is shown below. Deliverable D 13 Page 59

60 Unit Components 1. Main control unit (MCU) Light barrier radio connected, for system triggering Pair of transducers for wheel pass-by detecting Meteorological transducers for wind, air humidity and temperature Radio transmitter for controlling peripheral noise immission monitoring units Personal computer with data acquisition card and software for database 2. Noise emission and vibration Pair of microphones with preamplifier and monitoring unit digital/analogue converter Accelerometer with charge amplifier 3. Noise immission monitoring unit Radio receiver for event triggering Microphone with preamplifier Data acquisition and storage unit 4. External data acquisition unit Multichannel DAT tape recorder with remote control from main control unit Table 4.II Overview of units of the monitoring system Some examples of different configurations and applications are given in table 4.III below. Case Description MCU (n 1) Unit n 2 Unit n 3 Unit n 4 1 Noise emission monitoring X X 2 Noise emission and immission monitoring X X X 3 Type testing X X X 4 Noise immission monitoring X X Table 4.III Configurations of the monitoring system Deliverable D 13 Page 60

61 The above listed devices can be combined according to the purpose of the measurements. The general set-up is shown below. Unit n 2 Noise emission and vibration monitoring Unit n 4 External data acquisition Unit n 3 Noise immission monitoring Main Control Unit (n 1) Figure 4.2 Noise monitoring system: general connection scheme Deliverable D 13 Page 61

62 4.4 Application examples The developed monitoring system can be applied in a variety of ways. Two examples are shown below: daily monitoring and monitoring per train pass-by. Results produced by the monitoring system are presented. Daily monitoring Date Time Spee Train type Number Number Train Wind Temperat Wind Humidit LpAeq d of of Axles Length Speed ure [ C] Direction y [%] [db(a)] [km/h vehicles [m] [m/sec] ] 08/04/ Passenger Shuttle SSE /04/ Freight SE /04/ Passenger Shuttle E /04/ Railcars SSW /04/ Passenger Shuttle SW /04/ Railcars SW /04/ Railcars SE /04/ Freight NE /04/ Passenger Shuttle 16/10/ Locs and passenger cars for Test Train NE N /10/ Freight N /10/ Freight NNE /10/ Passenger Shuttle NE Table 4.IV Example of produced output from monitoring system for daily monitoring of rail traffic Deliverable D 13 Page 62

63 Monitoring per train Date 08/04/98 08/04/98 08/04/98 08/04/98 08/04/98 08/04/98 08/04/98 08/04/98 08/04/98 16/10/98 16/10/98 16/10/98 16/10/98 Time (Elapsed Time) Speed [km/h] Train type Passenger Freight Passenger Railcars Passenger Railcars Railcars Freight Passenger Locs and Freight Freight Passenger Shuttle Shuttle Shuttle Shuttle passenger Shuttle cars for Test Train Number of vehicles Number of Axles Train Length [m] Wind Speed [m/s] Temperature [ C] Wind Direction SSE SE E SSW SW SW SE NE NE N N NNE NE Humidity [%] LpAeq [db(a)] LpAeq [db(a)] per vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Vehicle Table 4.V Example of produced output from monitoring system for monitoring per train Deliverable D 13 Page 63

64 5 THE ROUND ROBIN TEST 5.1 Introduction A round robin test for exterior noise type testing was performed at four test sites in Austria, the Netherlands, France and Italy, using a specially composed test train with 2 groups of low noise freight wagons and one group of normal wagons. The round robin test was carried out to apply and demonstrate some of the proposed improvements to existing type testing methods. In addition, improved diagnostic techniques were demonstrated and the noise reductions due to low noise solutions were quantified. This was carried out within task IV: Application of Type Testing Methodologies - Round Robin Testing and Low Noise Solutions. The main objective of this task was to demonstrate the improved accuracy and reproducibility and the methods by which results obtained from pass-by tests at different sites can be compared. Achieving site-independence of type testing methods was a major issue in METARAIL. A detailed report of the procedure and results is given in report D11 ([15]). The sites were provided by ÖBB in Austria, NS in the Netherlands, SNCF in France and FS in Italy, whose assistance is gratefully acknowledged. The measurement and analysis work was performed by METARAIL partners IPSE Srl in Italy, Psi-A Consulting in Austria, SNCF in France and TNO and NSTO in the Netherlands. Wheel roughness measurements were performed by TNO in the Netherlands and rail roughness measurements by NSTO at all the sites. TNO was responsible for the scientific coordination and final compilation of results. The round robin test consisted primarily of a type test according to existing type testing standards such as ISO 3095, including some proposed extensions (pr EN ISO 3095, 1998). The measurements were carried out in the period of August to October The tasks performed by involved partners are shown in the table below. Deliverable D 13 Page 64

65 Task Austria Netherlands France Italy Provision of test vehicles Site provision and test train logistics Type test measurements Rail roughness measurements Wheel roughness measurements Impedance measurements Antenna and spatial track contribution ÖBB ÖBB ÖBB ÖBB ÖBB NS SNCF IPSE/FS Psi-A NSTO SNCF IPSE NSTO NSTO SNCF/NSTO NSTO Indirect, from Psi- A rail vibration TNO direct and indirect from Indirect, from SNCF rail vibration data NSTO rail vibration data data TNO TNO TNO TNO - TNO - - Indirect, from IPSE rail vibration data Table 5.I Task overview for the METARAIL round robin test. The tests were performed on the dates indicated below under good weather conditions. Country Date Number of test train pass-bys Temperature range deg. C Wind speed range m/s Humidity range % Austria 19/8/1998 (1 day) ( 3) Netherlands 17-18/9/ about (1 night) France 29/9-1/10/ (3 days) Italy 16/10/1998 (1 day) Table 5.II Date of the tests and weather conditions Deliverable D 13 Page 65

66 5.2 Description of the Test Train and Test Sites The Test Train A test train was composed especially for the METARAIL round robin test, consisting of 3 groups of 4 vehicles of the same type. The rest of the train consisted of disc-braked coaches at one end to distance the locomotive noise from the quietest wagons, and a locomotive at each end, at those sites where the train was running back and forth on the same track. All the test wagons were unbraked during transportation between the various sites, and were equipped with a GPS tracking system. Pictures of the test vehicles and the test train are shown in the figures below. TT1 Figure 5.1 Test train in full composition in Wiener Neustadt, Austria. Opposite to measurement side; Locomotive, Faccs gravel wagons, Rkqss/Sjgss car transport wagons, SGS container, passenger coaches and loco at end For 80km/h and 100 km/h Loco Disc Disc Sgs Sgs Sgs Sgs Sgjns Rkqss Sgjns Rkqss Faccs Faccs Faccs Faccs Loco TT2 For 140 km/h Loco Rkqss Rkqss Sgs Sgs Sgs Sgs or TT3 For 140 km/h Loco Disc Disc Sgs Sgs Sgs Sgs Rkqss Rkqss Loco Figure 5.2 Test train configuration for 80/100 km/h and 140 km/h Deliverable D 13 Page 66

67 Similar vehicles were used in groups of four to achieve sufficient averaging and to avoid any possible effects of masking by adjacent vehicles with higher noise levels. In each country, local (electric) locomotives and passenger coaches were used. The speeds chosen for the measurements were based on the maximum allowable speeds for the various wagon types. The hopper and Sgjss wagons were limited to 100 km/h, whilst the Sgs and Rkqss wagons had a maximum speed of 140 km/h. The wagon orientation was kept the same at all sites by tagging the vehicles with stickers on one side. During shipping from country to country the wagons were unbraked to avoid damage to the wheel surfaces. Consequently, they were shipped in groups of four and the test train was reassembled in each country. The wagons were followed by a GPS system so as to be able to locate them at all times. Figure 5.3 ÖBB Hopper wagon, type Faccs, Vmax = 100 km/h, cast iron block braked. 4 in test train Deliverable D 13 Page 67

68 Figure 5.4 ÖBB Flat wagons, type Sgjss with Vmax = 100 km/h, with metal sinter block brakes. 2 in test train. Figure 5.5 ÖBB Flat wagons, type Rkqss with Vmax = 140 km/h with metal sinter block brakes. 2 in test train. Deliverable D 13 Page 68

69 a) b) c) Figure 5.6:a/b/c DB Shrouded container wagons, disc braked, type Sgs, Vmax = 140 km/h. 4 in test train. Shrouds were only mounted on one side. Deliverable D 13 Page 69

70 5.2.2 Test Sites Four sites were selected with similar track components; ballasted track with either UIC54 or UIC60 rails, concrete sleepers (bi-block or monoblock). Railpads and fastener systems varied somewhat, from country to country. All sites were in service as frequently used lines for a sufficiently long time to be well run in, with the last rail grinding at least one year before the round robin tests. Details of the test sites are shown on the following pages. One of the main requirements was that each test site should have low rail roughness levels. This was checked beforehand by NSTO during a preselection procedure. The earlier requirement for type testing was that rails should be visibly smooth or that rail roughness should be measured. A visual check only is insufficient to avoid differences between sites. The 1998 improvements to the ISO 3095 standard require that the test site has rail roughness below a specified limit. Studies by ERRI have shown that if the rail roughness is below this limit, for most rolling stock the measurement spread will then be within a range of 4 db(a). Figure 5.7 Test site 1 in Wiener Neustadt, South of Vienna, Austria. Deliverable D 13 Page 70

71 Figure 5.8 Test site 2 in Schalkwijk near Utrecht, The Netherlands. Figure 5.9 Test site 3 at Meursault south of Beaune, France Deliverable D 13 Page 71

72 Figure 5.10 Test site 4 at Savigliano near Turin, Italy. Test train on return track. Average rail roughness at all sites 20.0 Roughness level (re 1 micrometer) ISO limit Austria Netherlands France Italy Roughness wavelength [cm], Frequency [Hz] at 90 km/h Figure 5.11 Averaged rail roughness at all round robin sites Deliverable D 13 Page 72