LOWNOISEPAD: Low-cost noise control by optimised rail pad: Feasibility study on the use of rail pad as noise mitigation measure

Eduard Verhelst 1 SD&M, Structural Dynamics & Monitoring, La Borderie, 87120 Domps, France Pinar Yilmazer 2 UIC International Union of Railways 16, rue Jean Rey, 75015 Paris, France Jakob Oertli 3 Swiss Federal Railways, Infrastructure Hilfikerstrasse 3, 3000 Bern 65, Switzerland Günter Dinhobl 4 ÖBB-Infrastruktur AG, Stab UE / team research & development 1020 Wien, Austria

ABSTRACT When aiming to minimise track noise, there is a significant conflict of aims that must be considered. Tracks with stiff rail pads are generally less noisy than ones with soft rail pads. Stiff rail pads, however, can decrease the long-term quality of the track. The situation with soft rail pads is exactly opposite: these cause more noise but offer better protection for the track. The International Union of Railways (UIC) works with its members to better understand the noise-generating mechanisms and to make tangible progress towards a being a better neighbour. The LOWNOISEPAD project brings together the European railway community in an effort to find an optimal rail pad for both noise and track quality issues and for the different situations encountered in 12 European railway Infrastructure Managers’ network. 1. INTRODUCTION

Based on today’s knowledge, in most cases at speeds between 80 and 160 km/h, the rail is identified as the dominant source. Besides the acoustic rail roughness, the stiffness and damping properties of rail fastening systems such as rail pads have a considerable influence on rail noise. In

1 wv1@telenet.be 2 YILMAZER@uic.org 3 jakob.oertli@sbb.ch 4 Guenter.Dinhobl@oebb.at

Jai. inter noise 21-24 AUGUST SCOTTISH EVENT CAMPUS ? O ? . GLASGOW

practice, rail pad design is always a compromise between track engineers’ demand for flexibility and acoustic engineers’ expectations for sufficiently high Track Decay Rate (TDR).

The project initiative, funded by 12 European railway infrastructure managers 5 (I.M.), aims to conduct a thorough feasibility analysis in order to achieve the above-mentioned ideal compromise. The goal is to optimise the rail pad so that noise emissions from the track are reduced as much as possible without increasing the cost of maintenance or lifecycle costs, e.g., due to maintenance. Each participant will independently conduct the same tests on their network, including TDR (compliant with ISO 15461) and pass-by (compliant with ISO 3095) measurements, using their own selection of a national track, rolling stock and speed. Running this approach at UIC will promote the deployment and use of the optimised rail pads in the near future by gaining consensus among acoustic and track engineers.

2. METHODOLOGY

The methodology is explained true the Work packages of the project. 2.1. WP1. Preparation of procedures

WP1 is the guide for project execution by the participating I.M. (Infrastructure Managers). It is based on literature, paper review, and similar successful projects (e.g. at Infrabel, SBB, ÖBB and DB). 2.2. WP2. Analysis of the existing situation at the railway network

The aim of WP2 is to know the link between the current track design and the associated acoustic quality. The acoustic quality can be described mainly by two parameters: the acoustical rail roughness, with dominant wavelengths between 0.5 cm and 50 cm, and the Track Decay Rate (TDR) with dominant frequencies between 500 Hz and 2.5 kHz. In this project the focus lies on the TDR.

Track database The goal of the database is to collect and compare information from 12 I.M. to link mechanical properties with know acoustical properties of their track.

Priority is given to the track and track components to be tested in WP5, for most I.M. this is the standard track in their network.

Track properties: Track (e.g. gauge), Track components (rail, rail fixation, sleeper, ballast, isolator). Rail pad properties: Composition, Mechanical stiffness (test bench under preload), Acoustical stiffness, Acoustical damping.

TDR single value parameter It was requested by some I.M. if it is possible to describe the TDR as a single parameter. A first definition is proposed, based on analysis of already available measured vertical TDR curves and their relationship with measured noise emission Insertion Loss. Since it is seen that in most cases the dominant effect of the TDR change on noise emission is limited in a specific bandwidth in a first step, it is proposed to take into account four 1/3 octave bands: 630 Hz, 800 Hz, 1000 Hz, 1250 Hz.

- The ISO 3095 limits for 630 Hz, 800 Hz, 1000 Hz, 1250 Hz are subtracted from the TDR values: respectively 6 dB/m, 2.19 dB/m, 0.8 dB/m and 0.8 dB/m. This leads to negative values if the TDR values are lower than the reference values. - To weight the contribution to the A-weighted noise emission the A-weighting values are applied. - The energetic mean is calculated for the four 1/3 octave bands.

5 Project members are from BaneNor (Norway), DB (Germany), FS (Italy), Infrabel (Belgium), IP (Portugal), Network Rail (the UK), ÖBB-Infrastruktur (Austria), ProRail (the Netherlands), SBB (Switzerland), SNCF (France), SZCZ (Czech Republic), Trafikverket (Sweden). Please consult the project website: https://uic.org/projects/article/lownoisepad

- This leads to values that are well correlated with the to be expected noise emission differences in dB(A). A first proposal is given for a TDR single value is given in Equation 1.

𝑓4=1250

൫ 𝑇𝐷𝑅 ( 𝑓 ) −𝑇𝐷𝑅𝑟𝑒𝑓 ( 𝑓 ) +𝐴𝑤 ( 𝑓 )൯

𝑇𝐷𝑅 𝑠𝑖𝑛𝑔𝑙𝑒 𝑣𝑎𝑙𝑢𝑒= 10 ∗𝑙𝑜𝑔10 ቌ෎ 10

10

ቍ /4 (1)

𝑓1=630

where TDR(f) gives the dB/m per 1/3 octave band, TDRref(f) gives limit value as defined in the ISO 3095, Aw(f), the A-weighting per 1/3 octave band. A first validation was done by using available historical data from INFRABEL, ÖBB and DB. A first order fit between the proposed TDR Single value and Global noise reduction for several hundreds of train pass-by leads to the following table. Table 1: TDR single value versus Noise reduction for 2 sets of available data.

TDR single value (dB/m) Noise Reduction on LAeq,tp (dB) -2.5 <0 0 0 5 1.2 10 2.2 15 3.5 20 4.4

This proposal will be validated and finetuned once more measurement data from the 12 participating countries become available, later in the project. 2.3. WP3. Selection of rail pads to be tested

Based on the information collected from the Track database defined in WP2, a selection of rail pad(s) to be tested is made. There are two possibilities here: (i) existing rail pads will be tested and compared, or (ii) an optimization of an existing rail pad is requested by the participating I.M.

According to the reference book Railway Noise and Vibration: Mechanisms, Modelling and Means of Control by David Thompson (§6.6.5) [1], one can assess with the final term in Equation 2, or - 10*Log 10 (TDR2/TDR1) corrections on global train pass-by noise emission based on TDR differences.

However extra corrections are to be made, since from around 2 kHz and higher, the wheel contribution starts to become more important.

(1)

Today, within the LOWNOISEPAD project, historical measurement data is available: TDR2, TDR1 and the corresponding LAeq,tp@TDR1 and TDR2, specifically for typical “PANDROL” and “VOSSLOH” tracks. So we can use the 1/3 octave band data to assess TDR influence on LAeq,tp.

To avoid the use of TWINS modelling for every type of train, in order to do a detailed wheel/rail source separation and prediction of the new LAeq,tp, it is proposed to determine and use two extra corrections factors. These can be extracted out of an existing database of Train pass-by LAeq,tp 1/3 octave data measured on minimum two different pads, for different types of rolling stock and at different speeds, as explained in the flow-charts below (Figure1.):

Figure 1: Definition of ILPB(f): Insertion Loss for the same train running on 2 different railpads

with TDR1 and TDR2

Figure 2: Definition ILPBc and correction factor TDRw(f)

- A first correction factor named ILPBc is based on the actual measured direct noise Insertion loss ILPB in 1/3 octave bands. The correction factor varies with the excitation of the system (by rough or smooth wheels). It is the difference between the real measured Insertion loss and the pure theoretical formula -10*Log 10 (TDR2/TDR1) (Figure 2). - A second correction factor TDRw, 1/3 octave bands weighting function is in some cases needed to focus the TDR influence on a known dedicated frequency range. This factor is tuned on the available pass-by measurements and needs to be determined also for different speeds ranges and rolling stock. So, in Equation 3, all parameters in 1/3 octave bands between 100 Hz and 5 kHz can then be used:

Caniefanea Be NIEPBlesianson) + Sleepers (f) §. Rails (f) -10logyo(TOR2/TOR1)

LAeq,tp@TDR2= LAeq,tp @TDR1 – 10log10 (TDR2/TDR1) + ILPBc + TDRw (2)

where LAeq,tp @TDR1 is the pass-by noise at a given TDR, ILPBc and TDRw the above defined correction factors.

When more measurement data becomes available in the project, the use and need of TDRw will investigated more in detail. Today we see that in some cases it is needed to cope with large variations in measured TDR values measured according to the CEN 15461[2]. Also, TDR values calculated according to CEN/TR 16891:2016 [3] will be used in the calculation and formulas above, since the latter method is less dependent on the performer of the impact tests according to the CEN 15461. 2.4. WP4. Test site selection + rail pad installation

A test location is to be selected. There will be only one component to be varied between the adjacent test sections: the rail pad.

The LOWNOISEPAD project intends to assess the acoustical influence/performance of the rail pad in the standard track in use at all twelve I.M. Since many parameters influence the typical noise emission generated by a train pass-by, all the following parameters are to be kept constant within the test sections: rail roughness, speed, ballast cross section, absorption reflection at microphone positions.

‘Aea,,@ TORI (f) — _ ILPBc(f) petal ea Rails (f) -og,s(TOR2/TOR) Pa ToRW(f) | Hew)

The comparison process is to be made on (at least) 2 adjacent 100m sites on the same track, named the reference section and the test section. It is proposed to evaluate for each section the LpAeq,Tp per 1/3 octave bands measured in a position as shown below, using the exact pass-by time Tp. The selection and the length of measurement sections: the sum of the reference and the test section depend on both practical and acoustical considerations but can’t be smaller than 160m (2x 80m). 2.5. WP5. Measurements

In WP5 the goal is to perform, at the by the I.M. selected sites, measurements on a track in service to compare the noise emission difference, in other words, the noise emission Insertion Loss generated by comparing the rail pad at the test section with the reference rail pad at the reference section. At each section, simultaneously noise emissions (ISO 3095) and vertical rail vibrations are to be measured.

Normalization and references -ISO 3095: 2013 Acoustics - Railway applications - Measurement of noise emitted by rail bound

vehicles. -EN 15461:2008+A1-2010: Characterization of the dynamic properties of track sections for pass

by noise measurements -EN 15610:2019: Railway applications - Acoustics - Rail and wheel roughness measurement

related to noise generation . Equipment and minimum instrumentation . It is proposed to apply the ISO 3095 requirements for instrumentation and also to take into account the following recommendations in order to avoid introducing deviations at sensor and instrumentation level. Use exactly the same types and sensitivity of accelerometers and microphones at both sections, also, the sensor fixation on rails and wind protection on the microphone should not be different. 2.6. WP6. Data processing and dissemination

The same processing will be applied for all I.M. – test sections. The raw data from both accelerometers on the rail and microphones will be first red with a file translator, and then “batch” processed in a similar way for all sites at the 12 I.M. The amount of measurement data will be huge. If e.g., 100 train pass-by are delivered per I.M., a total of about 1200 train pass-by will be processed. Processing protocol: An example (Figure 3) is used to show how the analysis can be performed based on previous analysis of similar measurement set-ups.

Figure 3: Example of data for processing: left side, raw data as sampled, right side, scaled and

synchronized data

The Figure 3 shows from upper to lower position respectively, rail acceleration at test section, sound pressure at test section, rail acceleration at ref. section, sound pressure at ref. section. Based on the rail vibration, the first and last axle is identified. From these, the pass-by time tp can be calculated (considering specific train geometry). Then the time delay dt between the first axle at the ref. section and the first axle at the test section in relation to the distance between these two leads to the estimation the train speed. 3. RESULTS

3.1 LOWNOISEPAD measurement data

For the 12 I.M., WP4 (test site selection, railpad installation) and WP5 (measurement preparation) are ongoing. Therefore, no measurements have been performed and processed within the LOWNOISEPAD project. By the end of September 2022, all measurement data will be transferred to LOWNOISEPAD for processing. Available historical data is used to show further how results can be presented.

3.2 LOWNOISEPAD SOFTWARE tool

A software tool is under development and will be fed by the project data and made available to the 12 I.M. at the end of the project in order to be able to visualize, compare, assess railpad changes from measurements performed in the 12 different countries. The software tool is fed by a database with the measurement data from the LOWNOISEPAD project. Today the software tool tested and fed only by some historical data for INFRABEL, ÖBB, and DB. Typical example is given in figure 4.

Figure 4: Calculation example on available data from the Belgian INFRABEL network. In red and blue measured data, in yellow right plot, a to be expected emission spectrum for the change in TDR

as shown in the left plot. In figure 5 3 examples show what can be an outcome of the software tool. Three “principal” screen shots with data from INFRABEL, ÖBB and DB are shown. The use of the software tool is easy:

- Select a country/I.M. and select there a train type and passage to be analysed: in the database there are pass-by data available from 2 different railpads, including the corresponding measured TDR, here according to the CEN 15461 - Select a rail pad, and thus a TDR, out of the database that you want to investigate - Select “calculation”, and instantly the to be expected pass-by level for the selected train/speed and railpad will be estimated. - The software tool shows the 1/3 octave results for pass-by, TDR, 10*log 10 (TDR2/TDR1), ILpb, ILPBC, ILPBC+TDRW, and also the to be expected noise reduction with reference to the softest railpad (TDR1) that was used during the measurement campaign.

Figure 5: 3 examples of the use of the software tools (data from INFRABEL, ÖBB, and DB)

4. CONCLUSIONS

At the end of the project, when all measurement data are introduced in the software tool, the project partners can evaluate the noise reduction on their specific type of track, rolling stock, speed, and temperature for different rail pads based on their typical TDR.

Within the project, data from at least 5, but probably more, optimized rail pads will become available. We see already a potential of noise reduction up to 3 dB and more, compared to very soft, non-optimized rail pads. There are already indications that optimized rail pads designed softer than very stiff EVA pads, to have better substructure protection, can result in lower emission than the EVA pads. This is to be confirmed in at least 2 countries. This will lead to recommendations for future rail pad design, properties, and geometry, and so clear quantification of low-cost noise reduction for future railway track fastener design. A software tool, validated by the available measurement data of the project can be developed to simulate the effect of railpad / TDR on Train noise. 5. ACKNOWLEDGEMENTS

We gratefully acknowledge the 12 I.M. for their driven participation: working on the database, installing different railpads, organizing and providing the measurements/measurement data. Also, thanks to UIC sustainability division for organizing/hosting the project, and to the UIC Track expert group, for their interest and support to the LOWNOISEPAD project. 6. REFERENCES

1. David Thompson, Railway Noise and Vibration: Mechanisms, Modelling and Means of Control ,

First edition, 2009 2. EN 15461:2008+A1-2010: Railway applications - Noise emission - Characterisation of the

dynamic properties of track sections for pass by noise measurements 3. CEN/TR 16891:2016: Railway applications - Acoustics - Measurement method for combined

roughness, track decay rates and transfer functions 4. Squicciarini G., Thompson D.J., Toward M.G.R. & Cottrell R.A. The effect of temperature on

railway rolling noise , Institute of Sound and Vibration Research, University of Southampton. 5. Oregui M., Núñez A., Dollevoet R. & Li Z. “Sensitivity analysis of railpad parameters on vertical

railway track dynamics”. ASCE Journal of Engineering Mechanics , Volume 143. 6. Auer F. Zur Verschleißreduktion von Gleisen in engen Bögen ; Wien/Graz 2010 27.10.2021 7. Verhelst W. & Heylen E. Development, validation and roll out of noise reducing rail pads on the

Belgian INFRABEL railway network , IWRN13