NORDTEX FINAL REPORT ROAD SURFACE TEXTURE FOR LOW NOISE AND LOW ROLLING RESISTANCE

Transcription

1 NORDTEX FINAL REPORT ROAD SURFACE TEXTURE FOR LOW NOISE AND LOW ROLLING RESISTANCE VEJDIREKTORATET, RAPPORT 506, 2013

2 ROAD SURFACE TEXTURE FOR LOW NOISE AND LOW ROLLING RESISTANCE NORDTEX FINAL REPORT AUTHORS: Jørgen Kragh, Lykke Møller Iversen and Ulf Sandberg PHOTOS: Road Directorate p.31 Ulf Sandberg DATE: December 2013 ISBN (NET): LAYOUT: Road Directorate COPYRIGHT: Vejdirektoratet,

3 CONTENTS PREFACE 4 FORORD 5 SUMMARY 6 RESUMÉ 7 1. BACKGROUND AND AIM 8 2. METHOD APPLIED OVERVIEW 9 3. LIMITATIONS FIELD CALIBRATION NOISE AND TEXTURE DATA RELATION BETWEEN TEXTURE AND NOISE Relation between CPX noise levels and road surface texture levels Inter-correlations Correlations based on NordTyre data Estimated road noisiness parameters HOW CPX NOISE LEVELS REPRESENT TRAFFIC NOISE LEVELS ROLLING RESISTANCE Background Measurements in Denmark Measurements in Sweden and a synthesis of rolling resistance measurements CHANGING TEXTURE TO REDUCE NOISE ON SURFACES IN SERVICE HOW TO OPTIMIZE SURFACE TEXTURE FOR LOW NOISE Essential principles Limitations and sacrifices General advice on noise reducing road surfaces An example of a favourable surface identified in NordTex CHALLENGES IN SURFACE PROFILE MEASUREMENT Problems observed by ISO working group WG DRD investigation VTI problems WG 39 work on solutions and latest status DISCUSSION General Differences in texture spectra measured by the parties Ways to obtain texture for low tyre/road noise emission Small aggregate Steep grading curve Grinding the surface Rolling resistance and noise CONCLUSIONS ACKNOWLEDGEMENTS REFERENCES 45 APPENDIX 1 PAVEMENTS AND SELECTED MEASUREMENT RESULTS 46 APPENDIX 2 TABLES OF CORRELATION 53 APPENDIX 3 DIFFERENCES IN TEXTURE SPECTRA MEASURED BY THE PARTIES 56 APPENDIX 4 ABBREVIATIONS USED IN THE REPORT 59 3

4 PREFACE Nordic road administrations face challenges in their effort to reduce population exposure to noise from road traffic. In Denmark, for example, Government strategy emphasizes the potential associated with the reduction of tyre/road noise. The Norwegian Government target has been defined as 10 % reduction of the number of annoyed persons before 2020 as compared to the situation in The most efficient way to reduce the general traffic noise is to reduce the source of noise emission. At present the most important part of this noise emission is tyre/road noise. Noise reducing road surfaces are mainly experimental although Danish national and municipal road administrations in recent years have applied thin asphalt layers with small aggregate, and experiments are going on to optimise noise reduction as well as structural and acoustical durability. To support such work the connection between surface texture and traffic pass-by noise levels should be clarified. This has been studied in the Nordic countries, in particular by the Swedish institute VTI, while little expertise existed in Denmark and Norway prior to the project reported on here. In 2006, exchange of information on road surface texture and tyre/road noise began between Nordic parties, attempting at creating a common knowledge base. In 2009 the NordTex project was initiated in order to establish new Nordic competence. This report documents the results of the project. The following persons have contributed to the project. The Project Steering Group Norway: Jostein Aksnes (chair), Norwegian Public Roads Administration (Statens Vegvesen, Vegdirektoratet) Denmark: Jakob Fryd, Danish Road Directorate (Vejdirektoratet) Sweden: Pereric Westergren, Jesper Elsander, Swedish Transport Administration (Trafikverket) The Project Advisory Luc Goubert, Belgian Road Research Centre, BRRC, Belgium Ragnar Evensen, ViaNova Plan og Trafikk AS, Norway The Project Group Denmark: Jørgen Kragh (project leader), Bent Andersen, Lykke Møller Iversen, Erik Nielsen, Jens Oddershede, Danish Road Directorate (Vejdirektoratet) Norway: Doreen Siebert, Odd Durban Hansen, Norwegian Public Roads Administration (Statens Vegvesen, Vegdirektoratet), Truls Berge, Svein Å. Storeheier, SINTEF Sweden: Ulf Sandberg, Mikael Ögren, Anders Genell, Swedish National Road and Transport Research Institute (VTI) 4

5 FORORD De nordiske vejforvaltninger står over for store udfordringer i arbejdet med at nedbringe den støjbelastning, som vejtrafikken påfører befolkningerne. I den danske regerings strategi for begrænsning af trafikstøj påpeges potentialet for at nedbringe støjen fra bildækkenes kontakt med vejbelægningen. Den norske regerings erklærede mål er en reduktion af antallet af generede personer i forhold til situationen i 1999 på 10 % før Den mest effektive måde at reducere trafikstøj på er at formindske støjen ved kilderne, og som situationen er i dag stammer størstedelen af den udsendte støj fra kontakten mellem dæk og vejbelægning. Støjreducerende vejbelægninger har hidtil overvejende været anvendt på forsøgsbasis, selv om det danske Vejdirektorat og danske kommunale vejadministrationer i de seneste år i stigende omfang har udlagt tynde asfaltslidlag med lille maksimal stenstørrelse: Der pågår løbende en indsats for at optimere støjreduktionen samtidigt med at slidlagene har en acceptabel holdbarhed, både strukturelt og akustisk. Denne indsats kan fremmes hvis man får større indsigt i sammenhængen mellem vejbelægningens overfladetekstur og støjniveauet fra forbikørende biler. Det har der i mange år været arbej det med i de nordiske lande, navnlig ved det svenske VTI, mens der var meget begrænset ekspertise til rådighed i Danmark og Norge forud for NordTex-projektet. I 2006 begyndte en proces med at udveksle viden mellem nordiske interessenter om vejbelægningers overfladetekstur og dæk/vejbanestøjen, og i 2009 blev NordTex projektet påbegyndt med henblik på at etablere ny nordisk kompetence. Nærværende rapport dokumenterer resultaterne af projektet. I projektets forløb har følgende personer medvirket. Styregruppe Norge: Jostein Aksnes (formand), Statens Vegvesen, Vegdirektoratet Danmark: Jakob Fryd, Vejdirektoratet Sverige: Pereric Westergren, Jesper Elsander, Trafikverket Rådgivningsgruppe Luc Goubert, Belgian Road Research Centre, BRRC, Belgien Ragnar Evensen, ViaNova Plan og Trafikk AS, Norge Projektgruppe Danmark: Jørgen Kragh (projektleder), Bent Andersen, Lykke Møller Iversen, Erik Nielsen, Jens Oddershede, Vejdirektoratet Norge: Doreen Siebert, Odd Durban Hansen, Statens Vegvesen, Vegdirektoratet, Truls Berge, Svein Å. Storeheier, SINTEF Sverige: Ulf Sandberg, Mikael Ögren, Anders Genell, Statens väg- och transportforskningsinstitut (VTI) 5

6 SUMMARY The present report documents results of a joint effort by research institutes in Denmark, Norway and Sweden. The NordTex project has looked at relations between road surface texture and tyre/road noise, by studying results of earlier research and by conducting new experiments. The results are valid for passenger car tyres running on dense (non-porous) pavement. Noise levels were measured by means of CPX trailers, and road surface profiles were measured with laser based equipment. The relations between these parameters were analysed. On some of the pavements, rolling resistance was measured using a special trailer. The following conclusions are drawn: 1. The general trends mentioned in 6) below confirms results found in earlier projects. This is remarkable, because tyre design and normal mix recipes have developed essentially in the meantime 2. Low texture levels at large texture wavelengths are associated with low tyre/road noise levels. This was the most important parameter for the pavements studied 3. High texture levels at small wavelengths are also associated with low noise levels, although in the present data this relation is less clear than the relation at larger wavelengths 4. The trends in 2) and 3) are valid for the overall group of data, but results from some pavements did not follow these trends. The reasons are not understood 5. CPX trailer noise levels measured with standard reference test tyres are strongly correlated with statistical pass-by vehicle noise levels from light vehicles, and pavement texture optimisation may be based on CPX trailer measurements 6. To obtain a surface texture for low noise levels, texture levels shall be low in the megatexture range, i.e. the third-octave bands with centre wavelengths 63 mm 500 mm. This may for example be obtained by a) applying aggregate with a small nominal maximum size and with a shape facilitating the aggregates to be seated closely together after compacting b) grinding off peaks of the surface of a pavement having large aggregates. These are more resistant to wear from studded tyres than is the case with smaller aggregates 7. Smaller maximum aggregates imply less durability, in particular less wear resistance to studded tyres. Texture change due to such wear may be reduced by having a large portion of high quality aggregate larger than 2 mm in the asphalt mix 8. Polymer or rubber modified bitumen may contribute in obtaining asphalt mortar more resistant to wear from studded tyres 9. Competence concerning road surface texture measurement and its interaction with tyre/road noise emission has been enhanced significantly among project partners during the course of the project 10. The Danish Road Directorate has decided to discontinue its default use of so-called enveloping when analysing pavement surface profiles. Such enveloping is an attempt to interpret the surface texture in a way corresponding to the impact it makes on a tyre. The procedure has not been developed sufficiently to be useful in practice 11. Lower noise levels also means lower rolling resistance as long as dense road surfaces are considered, and therefore efforts to reduce traffic noise do not counteract efforts to reduce fuel consumption or CO 2 emission 12. The so-called G factor did not provide any new correlation between surface texture profiles and tyre/ road noise, which is not already provided by other parameters 13. The parameter denoted Skewness is a parameter sensitive to whether the texture is positive or negative. This parameter should be considered in future research 14. The method proposed in ISO for predicting the change in tyre/road and vehicle noise levels as a consequence of changes in the surface profile gave results which agreed well with noise level differences measured in NordTex. This indicates that the project results are relevant and useful. 6

7 RESUMÉ Rapporten dokumenterer resultater af et samarbejde mellem forskere i Danmark, Norge og Sverige. NordTexprojektet har undersøgt sammenhængen mellem teksturen af vejes overflade og støjen fra kontakten mellem dæk og vejbelægning, ud fra både tidligere projekter og nye målinger og analyser. Resultaterne gælder alene for personbiler, der kører på tætte (ikke-porøse) vejbelægninger. Støjniveauet blev målt ved hjælp af CPX-trailere, og belægningens overfladeprofil blev målt med laser-baseret udstyr. Sammenhængen mellem disse parametre blev analyseret. På nogle belægninger blev rullemodstanden målt ved hjælp af en speciel trailer. Konklusionerne er: 1. De overordnede tendenser omtalt i punkt 6) herunder bekræftede resultater af tidligere projekter. Dette er bemærkelsesværdigt, fordi design af bildæk og asfaltrecepter har ændret sig væsentligt i den mellemliggende tid 2. Lave niveauer af teksturen ved store bølgelængder indebærer lave niveauer af støjen fra dæk/ vejbelægning. Denne parameter var den vigtigste på de undersøgte belægninger 3. Høje niveauer af teksturen ved små bølgelængder giver også lave støjniveauer, men i de foreliggende data er denne sammenhæng mindre for de længere bølger 4. Tendenserne under 2) og 3) gælder for det samlede datamateriale, men resultaterne fra nogle af belægningerne følger ikke de generelle tendenser, uvist af hvilke årsager 5. Støjniveauer målt med CPX trailer udstyret med standard referencedæk er stærkt korreleret med støjniveauerne (SPB) fra lette køretøjers forbikørsel, og optimering af belægningers tekstur kan baseres målinger med CPX-trailer 6. For at få en overfladestruktur der giver lavt støjniveau, skal teksturniveauet være lavt i området for megatekstur, dvs. i tredjedeloktavbåndene med center-bølgelængder 63 mm mm. Dette kan for eksempel opnås ved a) at anvende aggregater med en lille nominel maksimal størrelse og med en form, der gør at stenene lejres tæt sammen ved komprimeringen b) at slibe toppene af stenene i overfladen af en belægning stor maksimal kornstørrelse, som er mere modstandsdygtigt over for pigdæk-slitage end slidlag med mindre sten 7. En mindre maksimal kornstørrelse giver mindre holdbarhed, især mindre slidstyrke over for pigdæk. Teksturens modstandsdygtighed over for slid kan øges ved at tilsætte en stor andel stenmateriale af høj kvalitet med kornstørrelser over 2 mm 8. Polymer- eller gummimodificeret bitumen kan bidrage til at asfaltmørtlen bliver mere slidstærk over for pigdæk 9. Kompetencerne inden for måling af overfladetekstur er i løbet af projektet øget betydeligt hos projektets partnere. Det samme gælder indsigten i sammenhængene mellem teksturen og den udsendte støj fra kontakten mellem dæk og vejbelægning 10. Vejdirektoratet vil indtil videre undlade enveloping ved analyse af overfladeprofiler. Hensigten med enveloping-proceduren er at bedømmelsen af teksturen skal svare til den måde den påvirker et bildæk. Proceduren er endnu ikke udviklet til praktisk brug 11. Lavere støjniveau indebærer lavere rullemodstand på tætte vejbelægninger, og derfor modvirker indsatsen for at nedbringe støjen ikke bestræbelser på at reducere brændstofforbruget 12. Den såkaldte G-faktor giver ikke ny information om sammenhængen mellem overfladens tekstur og dæk/vejstøjen, som ikke allerede er indeholdt i de øvrige målte parametre 13. Parameteren Skewness er følsom for om teksturen er positiv eller negativ. Denne parameter bør indgå i fremtidige undersøgelser af sammenhængen mellem tekstur og støj 14. Metoden i ISO til at beregne ændringen i niveauet af dæk/vejbane- og forbikørselsstøj som følge af ændringer i overfladeteksturen gav resultater, som passer fint med niveauforskellene målt i NordTex. Det indikerer, at projektets resultater er relevante og anvendelige. 7

8 1. BACKGROUND AND AIM Understanding the relation between pavement surface texture and the generation of tyre/road noise, together with the effects of pavement acoustic absorption, if any, is a key to optimizing the noise reducing properties of road surfacings. The very ambitious aim of the project was to establish principles for the design of road surfacing mixes yielding lower tyre/road noise levels and to perform initial test of such mixes in full scale demonstration projects. This topic is important to all Nordic countries while the impact of studded tyres is of little interest to most European countries. According to the original project description the project should 1) Establish a base of data on texture and noise levels; 2) Analyse data and identify texture characteristics associated with low tyre/road noise levels; 3) Derive promising mix designs, and 4) Build test sections and test these initially and after their first winter in service. An important aspect pointed out in the process of establishing the project was the need for building Nordic expertise concerning road surface texture measurement and texture influence on tyre/ road noise. Prior to the project start, the Swedish Transport Administration (Trafikverket) funded a study resulting in a state-of-the art report on road surface texture influence on tyre/road noise [1]. During the initial project stages it became clear that measuring pavement surface properties was even more difficult than anticipated, and the project layout had to be revised several times so points 1) - 4) mentioned above could only be fulfilled to a limited extent. These somewhat modified aims are dealt with in the following sections of the report. The intention has been for the present final report to provide a summary of results with reference to details documented in project milestone reports. 8

9 2. METHOD APPLIED OVERVIEW The first stage of the project was a comparison of 1. noise levels measured by means of three different CPX noise measurement trailers [2], and 2. road surface profiles measured with different equipment [3] To be able to compare texture measurement results it was necessary to define a measurement method. In a second project stage CPX noise levels and road surface texture should be measured on existing roads and an initial analysis should be made of the relations between surface texture and noise. Originally it was the intention to utilise results of this analysis to construct new test sections of road with surface textures expected to provide low tyre/road noise levels and then repeat the procedure from project stage two to confirm that low noise levels had in fact been obtained. In reality it turned out to be necessary to select among sections of road which by experience were assumed to be associated with low tyre/road noise levels. The results from this phase of the project were reported in [4] - [7]. The noise and texture measurements should be supplemented by measurements of tyre rolling resistance to avoid sub-optimization based on noise alone. Due to challenges met in measuring road surface profiles only a few rolling resistance measurement were carried out. These results are documented in [8]. Microphone arrangement on DRD open trailer decibella 9

10 3. LIMITATIONS The project has been limited to nonporous pavement, and it deals only with surface properties at a specific point in time. Long-term testing has been outside the scope of the present project. Due to a limited budget, the project had to be based on data from a limited number of pavements and to take advantage of road sections built and data recorded in a variety of other projects. Finally, an important limitation is that much of the tyre/road noise data collected have been measured using a CPX reference tyre which is intended to represent light vehicle noise. Thus the data are focused on light vehicle traffic, while they do not represent heavy vehicle traffic. Texture influence on noise from light and heavy vehicle tyres is known to be different. 10

11 4. FIELD CALIBRATION Figure 1 shows, as an example, the mean profile depths, MPD, measured in 2009 by DRD, STF and VTI respectively, in Task 2, Calibration, see [3] and [9]. There is a reasonably good agreement, actually better than in most other comparisons made recently, as presented to ISO and CEN working groups. 2,5 2,0 1,5 1,0 MPD [mm] DRD SVV-STF VTI Figure 2 shows examples of surface profile texture spectra measured by DRD and SVV-STF, respectively. At wavelengths smaller than 10 mm or so SVV-STF spectral levels are higher than DRD spectral levels. When these results were first presented the reason for the deviation was thought to be insufficient spike removal from the SVV-STF profile data and perhaps remaining influence from electrical noise, even though SVV- STF had attempted to correct for this. Later the deviations were found to be caused mostly by DRD applying enveloping in its analyses while SVV- STF did not do so; see also Section 11. 0,5 0,0 HRA 11/16 AC 11d AC110 AC 60 SMA 6+ PA8/PA/16 Figure 1. MPD measured in the right wheel track on six 100 m long calibration sections on M10, Kongelundsvej and Vigerslevvej [3], [8], but adjusted in 2013 by DRD Note: The results in Figure 1 deviate slightly from data in [3], because 1) DRD has corrected errors in the order of mm made in its original analyses, based on reanalyses carried out in 2013; 2) The VTI result from the two-layer porous asphalt PA 8/PA 16 has been adjusted in May 2013 by DRD, because the original VTI data by mistake were from a 160 m long section of Vigerslevvej preceding the 100 m long calibration section, and not from the calibration section itself. On the first 160 m VTI measured MPD = 1.13 mm; DRD measured MPD = 1.17 mm; on the road section m DRD measured MPD = 1.51 mm. Measuring with SVV-STF trailer with enclosure 11

12 Figure 3 and Figure 4 show examples of the results of comparing the A-weighted CPX trailer noise levels [2]. Figure 3 shows the CPX noise levels measured with the same trailer equipped with four different reference tyres, SRTT. The results found with each of the four tyres running on seven different pavements are shown as a function of the average result of measurements with the four tyres. An old SRTT belonging to TUG and having a hard tread, probably as a result of ageing, yielded the highest noise levels. The remainder of tyres were in newer condition and yielded results within ± 0.5 or 0.6 db on all pavements, which is considered a normal variation. Figure 2: Texture spectra, examples from [3], [9] but showing final data for the full length of each road section Figure 4 shows CPX noise levels measured on the same seven pavements with the three different trailers with their new SRTTs. The range is larger than in Figure 3 and it was concluded in [2] that the reason why the TUG trailer yielded 1 db lower noise levels than the DRI 1) trailer, is a combination of 1) the individual TUG SRT tyre being a little less noisy than the individual DRI tyre and 2) that the TUG tyre on the average rolled slightly nearer to the less worn lane centre than the DRI tyre on the DRI trailer. This is a consequence of the two trailers having different test tyre setups: the DRI trailer has a test tyre in each wheel track while the TUG trailer has a single tyre in the middle of the trailer and attempted to drive with this tyre in the right wheel track. 1) DRI has changed its name to DRD during the NordTex project period 12

13 Figure 3. CPX noise levels measured with four different tyres SRTT on the same trailer, from [2] Figure 4. CPX noise levels measured with three different trailers with their new SRTTs, from [2] 13

14 5. NOISE AND TEXTURE DATA The amount of available data used for analysing the relationship between road surface texture and tyre/road noise is summarized in Table 1 to Table 3. These tables refer to tables in Appendix 1 which contain selected data for each pavement. The Danish road sections mentioned in Table 1 and Table 2 were built in the European project SILVIA (Kongelundsvej) and in a national Danish project (M10). On each site there was a reference section with AC 11d, and a number of noise reducing thin asphalt layers. The Danish road sections on Kastrupvej mentioned in Table 2 and the road sections mentioned in Table 3 do also comprise a reference section and some noise reducing thin asphalt layers on each site. The sections on Kastrupvej and on M68 were built as a part of the European project SILENCE, while the sections on M64 were built as a part of the Danish-Dutch cooperation in the Dutch IPG project. The road sections at Igelsø were built in a Danish project to demonstrate a variety of noise reducing thin asphalt layers. The Norwegian test sections were built in the Norwegian project Environmentally friendly road surfaces (Miljøvennlige vegdekker). 14

15 Table 1. Overview of CPX measurements made in 2009 with SRTT at 50 km/h; the total number of pavements was N = 34, 18 of which in the Copenhagen area and 16 Norwegian pavements; in some cases this set of data has been denoted Data set #1 or <ds1_50> Sub set No. Site Pavement Measured by N [-] Measured Listed in ID Table No. 1 Kongelundsvej DRI01-DRI06 DRD Table 10 2 M10 DRI07-DRI13 DRD Table 10 3 Norwegian pavements STF02-STF19 STF Table 11 Reference tyre: SRTT Speed [km/h]: 50 Table 2. Overview of CPX measurements made in with CPX reference tyre A at 50 km/h; the total number of pavements was N = 55, N 1 = 18 in Denmark and N 2 = 37 pavements in Norway; in some cases this set of data has been denoted Data set #2 or <ds2_50> Sub set No. 1 Site Pavement Measured by N [-] Measured Listed in ID Table No. Kongelundsvej, M45 Kastrupvej DRI01 DRI02, DRI03 DRI06, DRI14 DRI15, DRI20 DRI29 DRD Table 14 2 Norwegian pavements STF Table 15 Reference tyre: A Speed [km]: 50 Table 3. Overview of CPX measurements made in with SRTT at 80 km/h; the total number of pavements was N = 46 pavements, 41 in Denmark and 5 in Norway; in some cases this set of data has been denoted Data set #1 or <ds1_80> Sub set No. Site Pavement Measured by N [-] Measured Listed in ID Table No. 1 Igelsø DRI30-DRI41 DRD Table 12 2 M64 DRI42-DRI55 DRD Table 12 3 M68 DRI56-DRI64 DRD Table 12 4 M10 DRI07-DRI013 DRD Table 12 5 Norwegian pavements STF Table 13 Reference tyre: SRTT Speed [km/h]: 80 15

16 6. RELATION BETWEEN TEXTURE AND NOISE 6.1 RELATION BETWEEN CPX NOISE LEVELS AND ROAD SUR- FACE TEXTURE LEVELS Figure 5 is a reproduction of a diagram from [4], showing the determination coefficient R 2 from linear regression analyses of the overall A-weighted CPX noise level on the third-octave band texture wavelength. The determination coefficient R 2 expresses the percentage of the variation in the measured noise levels which is explained by the variation in surface texture levels at that wavelength. The figure is based on a set of data from N = 29 roads with CPX noise levels measured using SRTT at 50 km/h: 13 DRD CPX and texture results from Kongelundsvej and M10, see Table 10 in Appendix 1, and data on noise levels and texture measured by SVV-STF on 16 roads in Norway, see Table 11. The highest percentage of explained variance R 2 = 72 % was found at 40 mm wavelength. These results are contaminated at small wavelengths, because DRD applied enveloping in the Danish results while STF did not do so in the Norwegian data. The enveloping caused lower texture levels than measured by SVV-STF and this has resulted in a lower percentage of explained variance. Figure 6 in the same way shows R 2 determined from Data set #2, see Table 2, for N =55 road sections with dense asphalt surfaces measured at 50 km/h with two old reference tyres A having slightly different age. The highest, although not high, determination coefficient R 2 is seen at 160 mm wavelength and there is virtually no correlation at small wavelengths. As Based on Data set #1 SRT tyres N = 29 roads M10, Kongelundsvej N 1 = 13 Norwegian roads N 2 = 16 Figure 5. Percentage of explained variance in overall A-weighted CPX noise level as a function of the texture wavelength; Data set #1, N = 29: L AcpxSRTT at 50 km/h [4] Figure 6. Percentage of explained variance in overall A-weighted CPX noise level as a function of the texture wavelength; based on Data set #2, N = 55: L AcpxTyreA at 50 km/h [4] Based on Data set #2 SRT tyres N = 55 roads N 1 =18 roads in DK N 2 = 37 roads in NO 16

17 already mentioned these results are contaminated at small wavelengths by DRD having applied enveloping while STF did not do so, and this probably has contributed to the lack of corralation at small wavelengths. Figure 7 is taken from [5] and shows the relation between the CPX noise levels from SRTT and the road surface texture levels at 80 mm wavelength for Data set #1, i.e. SRTT at 50 km/h on 34 roads (29 roads from Figure 5 plus 5 SVV-STF results from Kongelundsvej). As already mentioned, Figure 5 shows the highest R 2 at 40 mm wavelength while Figure 6 shows the highest R 2 at 160 mm. Nevertheless, in Figure 7 L tx80 has been chosen as an independent variable on the X-axis. This choice has also been made in various literature references, e.g. the octave band L tx80 in [10]. Based on Data set#1; N = 34 1= Kongelundsvej; N = 11 2 = M10; N = 7 3 = Norwegian roads; N = 16 The memo [5] contains several graphs of the type shown in Figure 7a and Figure 7b. Examples are shown in Figure 8 and Figure 9. The trend is for overall A-weighted CPX noise levels to be positively correlated with the surface texture third-octave band level at 80 mm wavelength. The same applies to some extent to the low frequency third octave band noise levels at 315 Hz but not to high frequency noise at 2000 Hz. Based on Data set#1; N = 34 1= Kongelundsvej; N = 11 2 = M10; N = 7 3 = Norwegian roads; N = 16 Figure 7b shows the same data as Figure 7a, but regression lines are shown for each of three sub sets of data. For the sub set shown in blue the range of texture levels is very limited and the texture/noise level relation is not significant. With the same texture level at 80 mm wavelength the pavements on M10 (green line) tended to yield about 3 db higher noise levels than the two other groups of pavement. This is looked into in some detail in Section Figure 7. Overall A-weighted CPX noise level from SRTT at 50 km/h as a function of the one-thirdoctave band texture level at 80 mm. N = 34 pavements in Data set #1. Fig. 7a (top): All data, L AcpxSRTT = L tx80 ; R 2 = Fig. 7b (bottom): Linear regression lines per sub set of data; taken from [5] 17

18 The data in Figure 9 illustrate the fact that there is a trend for the Norwegian pavements to have larger maximum aggregate size than the Danish wearing courses, and the higher noise levels on the Norwegian roads may be associated with this and with the use of studded tyres. The same trend was seen in data collected at 80 km/h [5]. Figure 10 shows 2-D correlation contours, i.e. the correlation 2) coefficient R from linear regression analyses for combinations of texture wavelength and acoustic third-octave frequency bands. The figure is a copy of Figures in [5] for the Norwegian pavements in Data set #2 ( Old reference tyre A at 50 km/h on N = 37 pavements) in the top part of the figure, and for the Danish pavements in Data set#1 (SRTT at 80 km/h on N = 41 pavements) in the bottom part of the figure. Table 17 and Table 18 in Appendix 1 give the values of R 2. Each table is based on data recorded with the same equipment, so the results are not contaminated by DRD having applied enveloping in the Danish results while STF did not do so in the Norwegian data. Figure 10 and Table 17 Table 18 show a positive correlation between the noise levels below 800 or at 1000 Hz and the surface texture levels at wavelengths larger than mm, and a negative correlation between high frequency noise levels and texture levels at small wavelengths. This has also been seen in earlier investigations. For example, the NordTex results in Figure 10 are in qualitative agreement with those in Figure 11 from [10]. Figure 11 was published in 1980, 2) Note, the figure shows the correlation coefficient R, not the determination coefficient R 2 as was the case in Figure 5 and Figure 6 Based on Data set#1; N = 41 1= Igelsø; N = 12 2 = M64; N = 13 3 = M68; N = 9 4 = M10; N = 7 Figure 8. CPX noise level from SRTT, L AcpxSRTT, at 80 km/h as a function of the one-third-octave band texture level at 80 mm, L tx80, for N = 41 pavements in Data set#1. The linear regression line is given by L AcpxSRTT = L tx80. R 2 = 0.462; taken from [5] Based on Data set #2; N = 55 1 = Kongelundsvej, M45, and Kastrupvej; N = 18 2 = Norwegian roads; N = 37 Figure 9. CPX noise level, L AcpxA, from old reference tyre A at 50 km/h as a function of the onethird-octave band texture level at 80 mm, L tx80, for N = 55 pavements in Data set #2. The linear regression line is given by L AcpxA = L tx80, R 2 = [5] 18

19 Third-octave band centre frequency [Hz] Third-octave centre wavelength [mm] Figure D correlation contours. The colours indicate the value of the correlation coefficient R. Figure 10a (top): Old reference tyre A at 50 km/h on N = 37 Norwegian roads, Data set #2; Figure 10b (bottom): SRTT at 80 km/h on 41 Danish roads, Data set #1. From [5] based on a Belgian and a Swedish study [11] attempting to determine noise-texture correlations as a function of the acoustic frequency and the texture wavelength. These Belgian- Swedish results were later confirmed by similar studies [10]. In Figure 11 top left, the maximum value of R = 0.95 at (315 Hz, 80 mm) and the minimum R = at (5000 Hz, 2 mm). Correspondingly: Top right: max 0.92 at (400 Hz, 80 mm); min at (3150 Hz, 2.5 mm); Bottom left: max 0.94 at (630 Hz, 80 mm); min at (4000 Hz, 2.5 mm); Bottom right: max 0.85 at (315 Hz, 63 mm), min at (2500 Hz, 2.5 mm). 19

20 Figure 11. Diagrams showing contours of correlation coefficients R from linear regression analysis of third-octave bands of acoustic frequency on thirdoctave band texture wavelength, originally from [10] but quoted here from [1]. All were radial tyres. Tyre P: PIARC reference tyre 165R15; Tyre S: Firestone 165R15; Tyre X: Michelin summer 175R14; Tyre W: winter tyre 165R15 What is unique about the NordTex study versus the Belgian-Swedish studies made before 1980 is that the NordTex study used the CPX method while the Belgian Swedish studies used the coast-by method, and that test tyres were different. Also, road surfaces were different, because in the late 1970 s there were few thin asphalt layers and no SMA surfaces. Despite these differences, the results are amazingly similar, and essentially the NordTex study confirms the results from the old Belgian-Swedish studies, although for modern tyres and roads. Thus, the noise/texture relations seem to be robust. However, these relations cannot be assumed to be valid for tyre/road noise from heavy vehicles. 6.2 INTER-CORRELATIONS Figure 12 [6] and Table 4 illustrate the relations between the overall A- weighted CPX noise levels and various pavement parameters: surface texture level L tx80 and L tx05 at third-octave band centre wavelength 80 mm and 5 mm, respectively, nominal maximum aggregate size, NMAS, and mean profile depth, MPD. L tx80 seems to be better correlated with the CPX noise levels than the other pavement parameters are, and L tx80 is highly correlated with MPD. These results are contaminated at small wavelengths because DRD applied enveloping in the Danish results while SVV-STF did not do so in the Norwegian data, and this probably contributes to the poor correlation between L tx05 and the CPX noise levels. 20

21 1 = Kongelundsvej; N = 11 2 = M10, N = 7 3 = Norwegian roads; N = 16 Figure 12. Relations between overall A-weighted CPX noise level and various pavement parameters, and interrelations between these. Based on measurements made by SVV-STF and DRD with SRTT at 50 km/h, <ds1_50>, see Table 10 and Table 11 Table 4. Values of R 2, summary from Figure 12, <ds1_50>; N = 34. Colours illustrate the degree of correlation L AcpxSRTT L tx80 NMAS MPD L tx05 L AcpxSRTT L tx NMAS MPD L tx Figure 13 illustrates the relations between the overall A-weighted CPX noise levels and the pavement parameters used in Figure 12 plus a so-called profile shape characteristic G (a German concept denoted Gestaltfaktor ). To calculate G, the cumulative probability distribution of the surface profile ordinates is generated. G is the percentage of this distribution which is higher than the average between the highest and the lowest individual points on the profile. The evaluation is made for every 100 mm of the profile length. A G value below 50 % indicates a peaky profile, while a G value larger than 50 % indicates a plateaulike profile with slits in it. Figure 13 is based on N = 21 pavements with available data on G [7]. The relations in the figure are all based on profiles measured by SVV-STF and therefore uncontaminated by DRD having applied enveloping while STF did not do so. 21

22 3 = Norwegian roads; N = 16 4 = Kongelundsvej, N = 5 Figure 13. Relations between overall A-weighted CPX noise level and various pavement parameters, and interrelations between these. Based on Data set #1; N=21; 3) Norwegian roads, red N = 16, Table 11; 4) Kongelundsvej, green N = 5, Table 10, with AC 8d excluded by mistake Table 5. Correlation coefficients R (Pearson correlation) summarised from Figure 13. The number of measurement results N = 21. Green = high positive; red = high negative correlation L CPXA L tx80 NMAS MPD G L tx5 Pearson Correlation L CPXA L tx NMAS MPD G L tx Table 5 summarises the correlation coefficients R based on the data illustrated in Figure 13. The highest correlation is between L tx80 and MPD. Best correlated with the CPX noise levels is the nominal maximum aggregate size NMAS, followed by L tx80. The G value is highly correlated with L tx05, but does not seem to provide new correlation with the noise levels that other parameters do not provide. MPD is sensitive to asymmetry of the profile and seems to represent the noise-texture relation better than G does. 22

23 6.3 CORRELATIONS BASED ON NORDTYRE DATA A multiple correlation was made for the NordTex project based on the data in Table 19 of the NordTyre final report [12]. The result is shown in Table 6. The study (of noise from different tyres versus texture parameters) contained the following parameters for 31 different pavements, including an ISO test track: L A MPD NMAS Age SRF L Me Average A-weighted noise level measured with the CPX method, using a set of 24 different car tyres, in db Mean Profile Depth, according to ISO , in mm Nominal Maximum Aggregate Size, in mm Surface age in years at time of noise measurement Type of surface (SMA or DAC) Megatexture level, according to ISO , in db rel m. L ME is believed to be closely related to L tx80 used in Table 5 The analysis confirmed the strong correlation between MPD and L ME. NMAS was found to provide by far the best correlation with L A, as was the case in Table 5. Age and megatexture were next. MPD correlation with the noise level was not significant. Note: The significance test was run as two-tailed, although it would have been better with one-tailed (provided all relations would be only one-directional). All Sig values should be divided by 2 to get the one-tailed values. Table 6. Results of a correlation analysis based on data from the NordTyre project. Sig. is the level of significance in two-sided test. Note that N = 30 for parameter SRF (Surface type), but 31 in all other cases. The only porous pavement (PA 6) was omitted from that part of the analysis, because it does not belong to any of the types SMA or DAC. Note also that Pearson correlation is R and not R 2 L A MPD NMAS AGE SRF L ME L A Sig. (2-tailed) Pearson Correlation **.448 * * N Pearson Correlation *.791 ** MPD Sig. (2-tailed) N Pearson Correlation.670 ** * NMAS Sig. (2-tailed) N Pearson Correlation.448 * Age Sig. (2-tailed) N Pearson Correlation * ** SRF Sig. (2-tailed) N L ME Sig. (2-tailed) Pearson Correlation.446 *.791 **.374 * ** 1 N ** Correlation is significant at the 0.01 level; * Correlation is significant at the 0.05 level (2-tailed) 23

24 6.4 ESTIMATED ROAD NOISI- NESS PARAMETERS Figure 14 from [6] shows the measured CPX noise levels as a function of ENDt (the Expected pass-by Noise level Difference due to texture differences) calculated using a method described in ISO [13] for Data set #1, with 41 CPX noise levels measured with SRTT at 80 km/h and texture data from the Danish roads listed in Table 12 in Appendix 1. There is fair correlation, with 20 % larger range in measured CPX noise levels than in the calculated ENDt. A logarithmic fit would provide a slightly higher determination coefficient R 2 = Figure 15 from [6] shows the Estimated Road Noisiness Level ERNL from [10], as a function of the ENDt. ERNL is the pass-by noise level from a passenger car estimated, by means of Eq. 1, from the octave band road surface texture levels L tx80 and L tx05 at 80 mm and 5 mm texture wavelength, respectively, with constants a = 0.5; b = 0.25 and c = 60. There is fine correlation and an almost 1:1 relation between ERNL and ENDt in these data. Figure 14. CPX noise level (minus an arbitrary 90 db) as a function of ENDt ERNL = a L tx80 - b L tx05 + c Eq. (1) Figure 15. Estimated road noisiness ERNL as a function of ENDt 24

25 7. HOW CPX NOISE LEVELS REPRESENT TRAFFIC NOISE LEVELS Pass-by noise levels from passenger cars from Table 12 in Appendix 1 are shown in Figure 16 as a function of the CPX noise levels measured with SRT tyres on the same pavement. Figure 17 shows the same relation based on a larger set of data collected by DRD on 45 different pavements and repeated on these pavements plus another five new. Figure 17 includes the data from Figure 16. At 80 km/h the average difference between the CPX and the pass-by noise levels is 21.5 db and the correlation is high. 1 = Igelsø, N = 4 2 = M64 dense, N = 9 3 = M68 dense, N = 7 4 = M64 and M68 thin porous, N = 5 The noise levels measured with the CPX method using SRT tyres are well correlated with the noise levels measured with the SPB method for light vehicles. The residual standard deviation is approximately 1 db, which is roughly in line with earlier studies. Thus texture optimization based on the CPX method corresponds reasonably well with an optimization based on the SPB method. Figure 16. SPB noise levels from passenger cars at 80 km/h on 25 different pavements vs. CPX SRTT 80 km/h Figure 17. Passenger car pass-by noise levels as a function of the CPX noise level measured with SRTT. N = 95 on 45 and 50 pavements in Denmark 2010 and 2011, respectively 25

26 8. ROLLING RESISTANCE 8.1 BACKGROUND Projects on tyre/road rolling resistance, such as MIRIAM, COOEE, ROSANNE, and national Swedish activities during the last 10 years have demonstrated that new surface policies must consider tyre/road rolling resistance. Consequently, exploring low-noise surfaces by texture optimization must consider the effect that noise-optimization has on rolling resistance. Fortunately, low noise and low rolling resistance, when achieved by improved surface texture, do seem to go almost hand-in-hand. In the NordTex project, rolling resistance was measured in Denmark and Sweden by the Technical University of Gdansk (TUG) with a special trailer [8]. Measurements of rolling resistance were also intended in Norway but abandoned due to budget restraints. The trailer was run at 50 km/h and 80 km/h, sometimes also at 110 km/h. Three test tyres were used: SRTT (16 ), Avon AV4 and Michelin Primacy; see [8]. The first two of these tyres have been selected as reference tyres for noise measurements with the CPX method, and have also been used as temporary references for measuring rolling resistance. The SRTT and the Michelin Primacy are car tyres while the Avon AV4 is a tyre for light trucks, used as a proxy for heavy truck tyres. This tyre is known to work well for that purpose in noise measurements. The response of car tyres to various road surface textures differs from that of truck tyres. Within the NordTex project or in close connection with it 3), 14 road surfaces were measured in Denmark and 40 in Sweden. 8.2 MEASUREMENTS IN DENMARK As an example, Figure 18 shows the rolling resistance coefficients (RRC) measured at 110 km/h with three different tyres on the test sections of road on M10, while Figure 19 shows the 3) The Swedish Transport Administration funded rolling resistance measurements in Sweden to supplement the NordTex project RRCs measured at both 80 km/h and 110 km/h as a function of the mean profile depth MPD [8]. There is less variation between pavements than between tyres and there is a trend for higher rolling resistance with increasing MPD; see also Table 7. The large variation between tyres is due to the fact that the Avon AV4 light truck tyre has much higher RRC than the car tyres; AV4 tyre could not replace a car tyre. More comprehensive data, although not from the NordTex project, appear in a report from the MIRIAM project [14]. Most of those data have been collected at a later time, when methods and equipment had been improved and they represent a wider range of textures and more road surfaces 4). 4) These data contain information suggesting that something may be wrong with the data for SRTT in Figure 19, leading to too little sensitivity in RRC to variation in MPD. Most probably the RRC for HRA is too low 26

27 Figure 18. Rolling resistance coefficients measured with three different tyres on seven sections of M10, from [8] Figure 19. Rolling resistance coefficients measured with three different tyres on seven sections of M10 as a function of the surfaces mean profile depth; example from [8]. Data for UTLAC 8 are so close to data for AC 8o that the latter are difficult to see in the diagram 27

28 Table 7. Results of linear regression analyses of rolling resistance coefficients on MPD as measured on the seven sections of M10; based on [8] but with adjusted MPD-values; see the Note in Section 4. s R is the standard deviation of the residuals Tyre SRTT Avon AV4 Michelin Primacy HP Speed Slope Intercept R 2 s R [km/h] [-/mm] [-] [-] [-] In the MIRIAM report [14], the final relation between the rolling resistance coefficient RRC and MPD was suggested for light vehicles to be: RRC = Constant MPD + X IRI (Eq. 2) where MPD is Mean Profile Depth in mm, measured according to ISO IRI is the International Roughness Index X is a constant yet to be determined (but quite low for cars) Constant is a value unique to a certain tyre and several other circumstances; usually An important thing to notice in Eq. 2 is the slope of the relation with MPD, namely , which is larger than the value for SRTT in Table MEASUREMENTS IN SWEDEN AND A SYNTHESIS OF ROLLING RESISTANCE MEASUREMENTS This section presents a compilation made for the NordTex project of the results of rolling resistance measurements made by TUG at the request of VTI on 40 surfaces in Sweden, with three tyres, at two speeds. To reduce data, all measurement results from a certain road have been arithmetically averaged; i.e. for the two speeds and for the three tyres. This also reduces random errors. Speed is unimportant as speed influence is small and correlation between results measured at the two speeds is almost perfect. However, tyres may have slightly different sensitivity to texture, i.e. the RRC-MPD slope may differ, but it has been noticed that the ranking of surfaces is not affected by this. Thus, the grand averages given here represent an average for traffic having two thirds of light vehicles and one third of trucks and buses, assuming that tyre Avon AV4 represents the latter, which has been indicated in German results 5 ). The results are shown in Figure 20: RRC as a function of MPD. Each measurement series is distinguishable 5) A reference has not yet been approved for publication by the German sponsor by symbols and colours. The reason for distinguishing between series is that there are uncontrolled variations between these (and probably also within some series, although much less). Such variations are variations in temperature, and probably problems with stable calibration. For example, the two data sets yielding the highest rolling resistance coefficients were measured at 5-10 o C, while other data sets appearing in the lower part of the diagram were measured at o C. There could also be differences between data sets due to tyre changes between measurement series. No temperature correction has been made, since correction factors are unknown, while it is known that temperatures in south Scandinavian summer and autumn do affect rolling resistance as much as the road surface texture does. In Figure 20, the Danish measurements mentioned in Section 8.2, supplemented by results from Kongelundsvej have been added together with results of TUG measurements for the Round Robin Test (RRT) of MIRI- AM. Only the TUG data from MIRIAM were included, because other measur- 28

29 ing devices did not supply comparable data for all the surfaces. The most important result in Figure 20 is the slope of RRC versus MPD; i.e. the texture influence on rolling resistance. Regression data for all data sets are listed in Table 8. The slopes appear consistent ( ) when uncertain data sets (shown in parentheses) are neglected, although also the latter show reasonably consistent slopes. A slope of would be a good average for the three tyres used. If one would distinguish between the two car tyres on one hand and the light truck tyre on the other hand (these results are not shown in the figure), the slope for the car tyres would be approximately and for the truck tyre approximately This implies that, as is the case for noise [10], truck tyres are less sensitive to texture than car tyres, which is logical given the differences in dimension. RRC for average of 3 test tyres DK 2009 SE RV SE St Aska 2010 SE Skåne 2009 SE Skåne SE Skåne 2010 TAL MIRIAM RRT SE 2011 SE MPD [mm] Figure 20. Rolling resistance coefficients (RRC) measured with three different tyres at two speeds (RRC averaged over these) on 40 Swedish road surfaces as a function of the surface mean profile depth (MPD). Also results from Danish measurements mentioned in Section 8.2 and from TUG measurements in a Round Robin Test carried out in the MIRIAM project have been included Table 8. Results of linear regression analyses of rolling resistance coefficients RRC on MPD for the data in Figure 20. Each observation is an average for three tyres and two speeds. SE = Sweden, DK = Denmark. Values in () are uncertain due to small MPD range or too few observations Data set Number of observations Regression equation: RRC = Slope [-/mm] DK MPD SE RV MPD SE St Aska MPD (0.0017) (N.A.) SE Skåne MPD SE Skåne MPD SE Skåne 2010 TAL MPD (0.0021) (0.98) MIRIAM RRT MPD SE MPD (0.0012) (0.44) SE MPD (0.0009) (0.43) R 2 [-] 29

30 9. CHANGING TEXTURE TO REDUCE NOISE ON SURFACES IN SERVICE In order to optimize noise properties of in-service road surfaces by texture changes, VTI has tested ways of grinding the surfaces. Cutting narrow and shallow grooves in the longitudinal direction on a cement concrete surface with exposed aggregate (EACC). This does not change the texture between the grooves, but might provide extra drainage and potentially reduce air pumping noise. In 2012 this was tested on motorway E4 north of Uppsala. No improvement was found, perhaps because the texture in contact with the tyres was not changed and the grooves were too shallow to provide any substantial air drainage. Longitudinal diamond grinding made by rotating diamond blades separated a few millimetres. This was tested on a 20 years old cement concrete motorway at Arlanda Airport, and yielded a fine texture with narrow ridges ( fins ) so thin they probably create short-wavelength deflection in the tyres. The grooves in between provide air and water drainage, thus reducing air pumping noise. The result was 4 db of noise reduction for the SRTT and 3 db for the Avon AV4 tyre compared to a non-ground SMA16 surface. Circular grinding of asphalt to create a negative surface texture. This test was made in 2011 in Huskvarna where two-layer porous asphalt had been in service for one year [15]. The pavement was ground in one wheel track, see Figure 21. Porous asphalt has a concave ( negative ) texture but the largest aggregate (11 mm) had rounded upward faces. By grinding the top 1-2 mm of the surface with a horizontally rotating disc, an extra negative profile was created. Such grinding is fast and should not be very expensive. The idea was to provide supplementary and unique information to the NordTex project. For reference tyre SRTT, noise was reduced by up to 3 db due to the grinding and for the Avon AV4 tyre, noise was reduced by 0.5 db. Rolling resistance was reduced by 3-7 %. One year later the effect of grinding had vanished because intensive traffic (including studs) had worn off the flat faces of the chippings. A new Swedish project to study the circular grinding technique on road surfaces, including SMA surfaces, and its effect on noise, rolling resistance and emission of particulates has just started. A similar experiment has been initiated in Victoria, Australia. This suggests that it is possible to optimize pavement surface texture not only by selecting a proper pavement mix and paving procedure, but also by modifying the surface texture after the pavement has been in service for a while. 30

31 Figure 21. Non-ground surface (top) and ground surface (bottom). The coin diameter is 25 mm 31

32 10. HOW TO OPTIMIZE SURFACE TEXTURE FOR LOW NOISE 10.1 ESSENTIAL PRINCIPLES The previous sections highlight the following texture-related principles which are essential for creating a pavement with noise reducing properties (primarily for light vehicle traffic): For low noise levels at low and medium acoustic frequencies, texture at wavelengths larger than about 30 mm should be minimized. It follows that also L tx80 and L ME (megatexture) should be minimized For low noise levels at high acoustic frequencies, texture at wavelengths below about 10 mm should be maximized, meaning that L tx05 is maximized These two features mean that the texture of a surface shall not be minimized but be moved from the large to the small texture wavelengths as much as possible and practical. This may be achieved by the next feature: The nominal maximum aggregate size (NMAS) shall be as low as is practical and safe. Each mm of lower NMAS saves approximately 0.3 db of tyre/road noise The profile of the surface texture shall be as negative as possible; i.e. the top of the surface shall be as flat as possible and as much of its profile as possible shall be directed downwards. This is a typical feature of most opentextured and porous surfaces. Well rolled SMA also shows such features. The Mean Profile Depth (MPD) shall be low (lower than say 1 mm), but its relation to noise is not always clear. It should not be too low as this will cause air pumping and emphasize tonal effects in tyre/road noise LIMITATIONS AND SACRI- FICES For safety reasons, MPD must not be too low, as it may cause poor drainage and result in low wet friction and risks for hydroplaning at high speeds, but also at medium speeds for worn tyres. Since low NMAS usually is correlated with low MPD, at least for dense asphalt concrete and similar surfaces, it means that NMAS should not be too low. For low speed and medium speed roads, NMAS could probably be 4-6 mm, but for high-speed roads 8 mm would be a reasonable minimum value of NMAS. This normally requires SMA or another open-textured type. In order to obtain a safe MPD value, MPD should not be below 0.4 mm and it should be significantly higher at high speeds. The requirement to have at least 0.4 mm MPD is a recommendation given by VTI in discussions in Sweden. Another limitation is the requirements on wear resistance and structural durability. Usually, lower NMAS means increased risk of ravelling. In countries where studded tyres are used, wear is highly dependent on NMAS. Therefore, in Sweden and Finland 16 mm NMAS is normally used on the national highway system. In Norway NMAS of 16 mm is common on heavily trafficked roads and 11 mm on lightly trafficked roads. When NMAS is reduced to 11 mm or to 8 mm, wear increases. This causes not only a shorter lifetime, rutting and higher costs but also higher particle emission rates. Studded tyre wear also depends on the traffic speed. Thus, one may accept a NMAS of 11 mm or even 8 mm on some urban roads. To improve the wear resistance of an asphalt mix with reduced NMAS the Norwegian project Environmentally friendly road surfaces (Miljøvennlige vegdekker) showed that it is important to maximize the proportion of aggregate larger than 2 mm in the mix, and that the aggregate should be of very good quality. In addition the wear resistance can be improved by modifying the bitumen with polymer or rubber and by tightening control during the process of adding filler at the mixing plant. This will contribute to stronger and more elastic asphalt mortar GENERAL ADVICE ON NOISE REDUCING ROAD SURFACES If the effect of studded tyres can be neglected, such as in Denmark, 6 or 8 mm NMAS would be preferred, and textures should be relatively open, such as for SMA and some thin asphalt layers. An extra open version of SMA with 6 mm NMAS may be a good candidate, but there are some thin asphalt layers which may be at least as good. However, for high-speed roads, one 32

33 should prefer 8 mm NMAS in order to provide sufficient drainage. The present project has been limited to non-porous road surfaces, and porous surfaces are not mentioned here, although they may provide even better acoustic properties. Table 9. Qualitative description of surface texture levels on quiet and less quiet pavement Wavelength Pavement noise properties [mm] Quiet Medium Noisy Low Low High 5 High Low Low For regions where studded tyres are widely used, 8 mm NMAS seems to be useful on roads with posted speed limits up to 70 km/h (for example SMA 8 or similar thin asphalt layers), while at higher speeds 11 mm should be preferred (for example SMA 11 or similar thin asphalt layers). This definitely implies sacrifice in wear and lifetime compared to SMA 16. However, when noise needs to be reduced, this price needs to be paid; see also the comments on durable asphalt mortar in Section An alternative to replacing for example SMA 16 by SMA 11 is to grind the top of the surface as described in Section 9. In this way one may keep the 16 mm NMAS, and achieve the same noise properties as those of SMA 11 while maintaining the wear properties or perhaps obtaining even better properties than those of SMA 16. The cost is to grind the surface; in the worst case maybe each year, in favourable cases maybe every three years. The latter has to be verified by further tests and lifecycle costs need to be determined AN EXAMPLE OF A FAVOUR- ABLE SURFACE IDENTIFIED IN NORDTEX The track used by The Volvo Truck Corporation for noise testing, a so-called ISO track 6), has excellent 6) constructed in accordance with ISO 10844:1994. acoustic properties. CPX measurements in the NordTyre project [12] showed that this pavement was the quietest and other measurements showed that it was quieter than the successful double-layer porous asphalt on E4 through Huskvarna [16]. Figure 22 shows aggregate grading curves for the ISO test track in Hällered [17] and for a few other pavements. Figure 23 shows surface texture spectra measured by DRD 7) in 2012 on the ISO test track in Hällered and on the following other pavements in the NordTyre project [12]: 1) SMA 16 (ABS 16) at Hörby, the noisiest pavement represented in NordTyre 2) AC 8d (ABT 8) at Höor, noise-wise an average pavement according to multivariate analysis [12] 3) AC 6o (AB 6å) at Igelsø, the quietest of the real pavements represented in NordTyre The Volvo test track is interesting since it has a rather high MPD, which means that wet friction would not be a problem even on high-speed roads. It has been subject to weathering and test track traffic for many years and still has good properties. Its grading curve is almost identical to that of Swedish SMA 8 (ABS 8) according to specifications by Trafikverket. The grading 7) DRD used enveloping in its analyses, see Section 11.2 curve for the Norwegian SMA 16 in Figure 22 is similar to the medium curve for ABS 16 in the Swedish specification. The Igelsø AC 6o pavement is open-graded asphalt concrete with a gradation similar to that of regular SMA 6 pavement while other pavement parameters may deviate. The grading curve for SMA 16 shown in Figure 22 is for SMA 16 at Mastemyr while the texture spectrum for SMA 16 in Figure 23 is for SMA 16 at Hörby. The legend in Figure 23 also gives the average CPX noise levels measured at 80 km/h with the 31 different car tyres investigated in the NordTyre project [12] and the surface megatexture level L ME. The range of average tyre/road noise levels is 5 6 db. The quietest pavements have low texture levels at 250 mm 50 mm, and high texture levels at 5 mm wavelengths. The medium noisy pavement has low texture levels at 250 mm 50 mm, and low texture levels at 5 mm wavelengths. The noisiest pavement has high texture levels at 250 mm 50 mm, and low texture levels at 5 mm wavelengths. Table 9 is an attempt to summarize, in a qualitative way, the pavement surface texture properties related to tyre/ road noise emission. 33

34 100 Percent passing [%] ISO Hällered AC 6o Igelsø AC 8d Höör SMA 16 Mastemyr [mm] Figure 22. Aggregate grading curves for ISO test track at Hällered, for AC 6o at Igelsø; AC 8d at Höör and SMA16 at Mastemyr L tex [db re 1 µm] SMA 16: MPD = 0.99 mm; LME = 51.7; CPX = db AC 8d: MPD = 0.7 mm; LME = 44.2; CPX = 97.5 db AC 6o: MPD = 0.72 mm; LME = 45.2; CPX = 94.9 db ISO 10844: MPD = 0.86 mm; LME = 46.5; CPX = 94.7 db Centre wavelength [mm] Figure 23. Surface texture spectrum for ISO test track at Hällered and other spectra measured by DRD 1) in the NordTyre project [14]. The legend gives the mean profile depth MPD, the megatexture level LME and the average CPX noise level at 80 km/h from 31 different sets of passenger car tyres 1) See footnote 7) on p

35 11. CHALLENGES IN SURFACE PROFILE MEASUREMENT 11.1 PROBLEMS OBSERVED BY ISO WORKING GROUP WG 39 The first phases of the NordTex project met unexpected challenges as described in the following section. However, the following challenges had been recognized by the ISO working group ISO/TC 43/SC 1/WG 39 8), responsible for the ISO texture standards: Some measuring devices gave very different texture values from other devices Some devices gave unrealistically high values, as judged by experts in WG 39 Primarily modern equipment gave unrealistic values One organisation reported values which seemed unrealistic. This was later confirmed to be caused by the use of an inappropriate laser device As a result of this, a major manufacturer of laser sensors was called to serve as an expert to WG 39, and work was initiated to identify causes and suggest remedies. The causes were identified to be that users had stressed the performance of laser devices far beyond what was technically possible, as they required ever higher measuring speeds and ever higher measuring ranges, while at the same 8) The working group convenor is Ulf Sandberg time laser sensors on the market became more and more digitaloriented, which made proper regulation of laser output more difficult and electronic noise levels increasing. In many cases measured values were found to have little relation to the true values due to the previous problem, the occurrence of false spikes in recorded signals had increased. This problem had been neglected, and the removal of such false signals had not been sufficient some manufacturers/users had not adhered properly to the requirements specified in the standard for filtering the surface profiles before analysing the surface properties some specifications in the first version of the standard (ISO ) had been too liberal, leading to large deviations between results from devices having selected different options control methods for the performance of devices were not available The problems were serious to MPD calculation and even more serious to the calculation of spectra. Following this knowledge, which was gathered during the first years of the NordTex project, WG 39 worked on improving the standard to eliminate or reduce the problems. Some of these problems, although not the most serious ones, were suspected also to exist for some or all of the Scandinavian devices. The NordTex project required the utmost accuracy in texture measurements, since its purpose was to clarify fine texture features that might explain the noise reducing properties of pavements. Thus the NordTex project and the detected problems coincided in time in an unfortunate manner DRD INVESTIGATION Before SVV-STF performed its final analysis reported in [5] of the relations between texture and noise, DRD analysed its data in order to clarify if spikes could have made the results unreliable [18]. The conclusion was that the Danish texture data did not seem to suffer from such effects. This is probably due to the way data are collected. The measuring system consists of two lasers, one in each wheel track. The lasers measure at a frequency of 78 khz with a spot size smaller than 0.2 mm, and each reflected laser-pulse is categorized as valid or non-valid. The valid responses are averaged in a spatial resolution set to approximately 0.53 mm. At a driving speed of 50 km/h this corresponds to around three samples per spatial increment. Due to this procedure the system is not very sensitive to erroneous reflections and most positive or negative peaks are automatically removed. 35

36 Figure 24. DRD surface profile texture spectra resulting from analyses without (/-) and with (/+) enveloping 36

37 Texture spectra are determined as an energy average of spectra evaluated for consecutive lengths of m. Each spectrum is determined by FFT analysis using Hanning weighting and in the present project, enveloping of the profile was made by default according to the empirical method proposed by von Meier, van Blokland, and Descornet [19]. The intention was to apply a parameter d* = mm -1 in the enveloping, but subsequent investigations made during 2013 revealed that in fact the parameter had been set at d* = mm -1. This means that texture levels at small wavelengths have been reduced more than proposed in [19], because the surface texture will appear smoother, as the enveloping will smoothen out smaller valleys when d* is smaller. For very smooth or slowly fluctuating surface profiles this will not be a problem. DRD concluded in [18] that the DRD system delivered by Greenwood Engineering A/S, and used in the comparison measurements in 2009, gave the same texture spectrum, within the general measurement uncertainty, as determined using other DRD software on a file recorded by SVV-STF on two-layer porous asphalt on Vigerslevvej. Before analysing the SSV-STF file, spikes were removed using various procedures described in [18]. Figure 24 shows results of reanalyses made in 2013 by DRD of the profiles recorded during the calibration task mentioned in Section 4. The figure shows texture spectra from the full 400 m of hot rolled asphalt (HRA 16) and open graded asphalt concrete AC 11o on M10, determined without and with the enveloping procedure mentioned earlier in this section. Comparison with the spectra in Figure 2 reveals that most of the differences between DRD and SVV-STF results at wavelengths between 5 mm and 31.5 mm may be attributed to the enveloping applied by DRD. It must be concluded that the differences found between results from DRD and SVV-STF are mainly due to the fact that DRD applied enveloping while SVV-STF did not apply such a procedure. This unfortunate fact must be attributed to lack of experience in analysing road surface profiles VTI PROBLEMS VTI was the first to produce and use a mobile laser system for measuring road surface texture profiles. This system was produced by VTI and Selcom AB in cooperation and first implemented in It worked satisfactorily and was used widely until 2000, when the computer and its software for data collection had become outdated. A new computer and software were implemented during It was used in the first part of NordTex in These measurements showed that the system did not perform as well as it used to. Non-linearity was detected, the noise level was higher than before and drop-out rates higher than expected. Discussions with LMI Technologies Inc. who produces lasers, revealed that a laser of this age and history of use is not likely to work appropriately since the laser output decreases with its time in service. At the same time, the ISO WG 39 group decided it would be necessary to define a number of data-enhancing procedures; see Section Measurements of texture spectra are extra sensitive to the detected problems, and VTI decided not to measure texture spectra in NordTex. It was judged to be better not to supply texture spectra that most probably are not entirely correct, since the project was meant to explore detailed effects of texture, where the lowest possible uncertainty is needed. If erroneous data are used, results may be erroneous. However, MPD is not as sensitive to these problems, and the non-linearity could be handled by making sure that the distance between laser and road was nearly optimal, so it was judged 37

38 that MPD could be measured appropriately, recognizing the basic problems in the existing standard. Therefore, VTI delivered only MPD results in the NordTex project WG 39 WORK ON SOLU- TIONS AND LATEST STATUS Texture measurement standards are developed by the working group ISO/ TC 43/SC 1/WG 39, which is under the noise committee (SC 1) since it was originally noise-related issues that needed standards on texture measurement. In , WG 39 detected serious problems with modern laser profilometer systems applied for macrotexture and megatexture measurement. Many systems seemed to deliver erroneous transients in their output signal, tending to dominate the measurement results on many surfaces. This also applies for the popular measure denoted MPD (Mean Profile Depth) which relies on correct peak detection, as specified in ISO ; but, moreover, high spatial frequency bands in spectral diagrams are highly contaminated by the occurrence of such transients. Some laser sensors were found to have high (electrical) noise level, in the worst cases even higher levels than the road surface texture levels. The main reason for the occurrence of transients was found to be that invalid readings, called drop-outs, were not correctly eliminated. In principle, such readings are impossible to avoid on many surfaces; for example because the laser spot in deep troughs or on steep slopes in the profile is hidden as seen from the sensor. On some black and shiny surfaces, the laser light power reflected back into the sensor might be too weak. In the past, laser systems have been able to cope with such problems, by due detection of such problems and interpolating the signal across problematic parts of the profile. Procedures for this are described in However, in recent decades some trends have been unfavourable to texture-measuring laser systems. One is that the laser systems have tended to go from pure analogue properties, meaning that the laser output has been rather easy to control in some feedback loop, to something more similar to digital properties, in which the laser output is possible to control over a much more limited range, and also needing faster control loops. This has been driven by a market dominated by digital applications of laser systems such as in DVD and CD systems. The market for old-fashioned easy-to-control lasers has become so small that in practice such components are no longer available. In addition, and equally unfortunate, users have not been aware of such laser profilometer system limitations, even though ISO and contain warnings. Thus they have ordered faster and faster systems from suppliers, as well as systems having a large vertical measuring range. Such performance makes it easier to measure with mobile systems in traffic, even at high speeds. But the users have failed to realize that such performance comes at the expense of higher dropout rates, less possibility to properly and with negligible delay identify dropouts, poorer resolution and more background (electrical) noise. All these things together have resulted, despite the technical development and efforts made by the manufacturing industry, in many modern systems which in practice are unsuitable for doing texture measurements. Nevertheless they are used for this, resulting in erroneous values and wrong decisions, obviously a very unsatisfactory situation. Solutions to the problems could be: Procedures to process measured profile data by eliminating effects of transients and electrical noise must be developed and included in updated standard(s). This has been denoted data-enhancing procedures by WG 39. WG 39 will submit a draft for a revised ISO with such procedures before the end of They are also needed for texture spectral measurements and will be included in such standards Laser profilometers for pavement texture measurements must go through test procedures to verify that their performance meets requirements in the standards. The opinion of WG 39 is that such procedures will not only warrant satisfactory systems being available in the future, but also that users are made aware of the sacrifices high requirements on speed and measuring range lead to. WG 39 has been mandated to develop such a procedure; initially in the form of a Publicly Available Specification (PAS). 38

39 12. DISCUSSION 12.1 GENERAL It must be emphasized that the project has been limited to passenger car tyre noise on dense asphalt pavement. The texture/noise relations on porous pavement and for heavy vehicle tyres are known to be different from the relations looked at here. The NordTex project unfortunately turned out to be slightly untimely. It began just after serious challenges in texture measurements had been identified. Most of these have now been overcome as a result of work done by the ISO working group. Fortunately, the most serious problems were not encountered with the equipment used in the NordTex project, so the measurement results should be reasonably good, and an important outcome of the project is that essentially enhanced levels of experience have been obtained among participants. For tyre/road noise measurements it was observed that CPX trailer noise measurements with standard reference test tyres (SRTT) are indeed representative of the noise emission from light road vehicles, with differences in reference tyre rubber hardness as a major reason for differences in noise levels measured with different trailers. This implies that pavement texture optimization for low passenger car tyre/ road noise emission may indeed be based on such measurements. When it comes to detail, several observations call for further investigation DIFFERENCES IN TEXTURE SPECTRA MEASURED BY THE PARTIES SVV-STF measured higher texture spectral levels at small wavelengths than DRD did on the same pavement. This was mentioned in Section Thus when data from both parties are pooled, one should expect low or no correlation between noise and texture in this range. This was in fact the case in some of the results mentioned in Section 6 on the relations between octave band texture and noise levels. Other analyses mentioned were based exclusively on sets of data collected using the same equipment, either that of DRD or that of SVV-STF. The main outcome of these latter analyses was a general confirmation of results found in earlier studies: To obtain low tyre/ road noise levels one shall aim at low texture levels at wavelengths around 80 mm and high texture levels at wavelengths around 5 mm. However, this does not mean that individual results follow the general trend, as illustrated in Appendix 3. The results shown in Figure 25 and Figure 26 do not seem to follow the overall trend while the reasons for this are unknown. In these cases, higher texture levels at small wavelengths were not associated with lower noise levels at high frequencies. These deviations from the general trend cannot be explained by the different ways of surface profile analysis applied by DRD and SVV-STF WAYS TO OBTAIN TEXTURE FOR LOW TYRE/ROAD NOISE EMISSION Results from individual pavements may deviate, but the general trend is that there was good (positive) correlation between noise levels in the low-frequency one third-octave bands ( Hz) and corresponding texture spectra at the larger texture wavelengths ( mm bands). On the other hand, at high frequencies noise levels correlated fairly well, but in negative direction, with small-wavelength macrotexture. These opposite correlations, at low and high frequencies, respectively, are part of the reason for poor correlation between the overall levels of noise and the overall levels of texture. These findings confirm results of earlier research, and therefore it is not really new knowledge. However, it is useful to have confirmation, that relations found decades ago are still valid, despite the fact that both tyre and road surface technology have 39

40 developed significantly. For example, SMA surfaces were rare 30 years ago, and the thin layer asphalt surfaces of today were almost unknown, while rough surface dressings of the earlier times have now become rare. When planning the NordTex project, the so-called Acoustic Optimisation Tool (AOT) was thought of as a potential means for developing promising new asphalt mix designs. During the project period, however, the AOT was applied in a parallel Danish project [20] with discouraging results, and the plans for using the AOT in the NordTex project were discontinued. To reduce noise, texture levels should be reduced at the texture wavelengths mm but increased at 2-6 mm wavelengths. In fact, texture levels should be reduced over the entire mega texture range (wavelengths mm). The strong correlation at the lower acoustic frequencies and weaker (but significant) correlation at high frequencies mean that the most important thing is to create a smooth surface at the larger texture wavelengths. This may be accomplished in at least the following ways: 1) the largest aggregate shall be placed as close as possible together, i.e. they should ideally be in contact with each other, to avoid large and deep valleys between the peaks in the texture (to avoid excessive deformation of the tyre rubber when it envelops the surface) 2) a steep grading curve will help, because a more uniform large aggregate, only separated a little by smaller grains, may be closer together (typical of thin asphalt layers) 3) ravelling (loss of stones) shall be avoided since this creates holes or large and deep valleys between the peaks in the texture 4) the surface could be smoothened by grinding off the top 1-3 mm of its peaks, provided the grinding is made with tools which create a proper microtexture and the resulting macrotexture is sufficient to provide proper water drainage 5) proper compaction of the surface after the screed of the paver is important, as the aggregate will then align itself in the most even way Small aggregate In practice optimum textures may be obtained by using small maximum aggregate sizes. In this way the texture may be relatively smooth at large texture wavelengths, while having high texture levels at the small texture wavelengths. In multiple correlation studies of noise versus texture and other road surface parameters, the nominal maximum aggregate size (NMAS) comes out as the most important. Ideally, the NMAS should be as small as possible, in practice as small as reasonably durable. In Denmark, where studded tyres are not permitted, a NMAS of 6 mm or 8 mm is favourable, whereas in Norway and Sweden, studded tyres would soon wear down 6 mm and 8 mm surfaces, so 11 mm would be a more economical option. However, at low speed (perhaps up to 70 km/h), 8 mm NMAS could be a solution in Norway and Sweden, although at the cost of reduced durability and increased emission of fine particles. A large portion of high quality aggregate > 2 mm contributes to stabilizing texture change due to wear from studded tyres, and the application of polymer or rubber modified bitumen assists in obtaining a wear resistant asphalt mortar (i.e. the mix of bitumen and fines). Pavements with harder mortar, however, endure less deformation and repeated loading without cracking Steep grading curve The grading of aggregates, other than their maximum size, may contribute to noise optimization. For small NMAS, SMA grading or even a more open grading may be favourable, as it results in higher texture levels at small texture wavelengths. A grading open enough to lead to porosity may further contribute by introducing sound absorption to assist in reducing the tyre/ road noise emission Grinding the surface A negative surface texture is known to be associated with low tyre/road noise emission. This requires a smooth top of the surface having valleys in it. Such a surface is characterized by a high value of the so-called G-factor, sensitive to asymmetry of the texture 40

41 profile. The present project found that the G-factor did not provide any useful correlation, which other parameters do not already provide. The MPD is also sensitive to asymmetry of the profile curve and seems to represent the noise-texture relation better than the G-factor. A standardised measure denoted skewness is claimed to be superior to the G-factor, but it was not applied in the NordTex project. Thus a lesson learnt could be to add skewness as a parameter to be measured in future investigations of the relation between road surface texture and tyre/ road noise. The effect of grinding the surface was tested on a small test section in a Swedish project carried out in parallel with NordTex. The gain in noise reduction was 1-2 db, and rolling resistance was also reduced, see Section Rolling resistance and noise The results collected in NordTex on rolling resistance supplemented other studies of rolling resistance versus texture. The main outcome was a refined sensitivity parameter (slope) in the texture/rolling resistance relation, see Section 8.3. A slope of seems to be a good average for the tyres used and a rather robust result. The slope for car tyres would be approximately and for the truck tyre approximately This trend is similar to the trend for noise levels: truck tyres are less sensitive than car tyres to differences in texture. The rolling resistance coefficient on the worst surface was approximately 50 % higher than that on the best surface. Within a regular range of common surfaces, the range is more likely to be in the order of 20 %. Noise and rolling resistance properties seem to go hand in hand ; i.e. lower noise also means lower rolling resistance. DRD open trailer decibella 41

42 13. CONCLUSIONS An important result of the NordTex project is the confirmation of general trends found in earlier projects, even though tyre design and normal mix recipes have developed essentially in the meantime. Also, the older investigations were based on pass-by noise levels rather than on CPX trailer noise levels. This implies a second important finding made during the project period, namely that the relation between results of CPX trailer noise measurements made with standard reference test tyres and the statistical pass-by vehicle noise levels from light vehicles is so strong that pavement texture optimisation may be based on CPX trailer measurements, which are cheaper to carry out than pass-by noise measurements, once a trailer system has been procured. This conclusion is valid for passenger car tyre noise on dense asphalt. Noise from heavy vehicle tyres or tyre/road noise on porous pavement may show different trends. The general trend is that low texture levels at large texture wavelengths are associated with low tyre/road noise levels. This turned out to be the most important parameter for the pavements studied. And high texture levels at small wavelengths are also associated with low noise levels, although this relation is less clear than the relation at larger wavelengths. These trends are valid for the overall group of data, but results from some pavements were seen to not follow these trends. The reasons for this are not understood. Thus, the main way to obtain surface textures for low noise levels was found to be to keep texture levels low in the megatexture range of wavelengths, which include third-octave bands with centre wavelengths 63 mm 500 mm. This may be implemented in various ways, such as by applying aggregate with a small nominal maximum size and with a shape facilitating the aggregates to be seated closely together after compacting. The same effect might be obtained by grinding-off the peaks of the surface of a pavement having large aggregates. The latter has the advantage of being more resistant to wear from studded tyres than a pavement having small maximum aggregate. The trade-off between grinding cost and improved pavement durability should be explored in future projects. A Swedish project with such aims has just started. Applying small maximum aggregates induces loss of durability, in particular where studded tyres are used. Adding a large portion of high quality aggregate > 2 mm could stabilize the texture change due to the wear from studded tyres. Likewise, the application of polymer or rubber modified bitumen could contribute to a more wear resistant asphalt mortar. A major purpose of the NordTex project was to establish Danish and Norwegian competence concerning the interaction between road surface texture and tyre/road noise emission. The project in this sense has indeed been successful in leading to enhanced levels of experience obtained 42

43 by project participants. In this connection the Danish Road Directorate has decided to discontinue its default use of enveloping when analysing pavement surface profiles. It has been found, in NordTex and in other parallel projects, that low rolling resistance and low noise seem to walk hand in hand when dense road surfaces are considered, i.e. lower noise levels also means lower rolling resistance. Therefore efforts to reduce traffic noise do not counteract efforts to reduce fuel consumption or CO 2 emission. The parameter introduced in Germany to characterize pavement surface profiles, the so-called G-factor, did not provide any new correlation between surface texture and tyre/road noise, which other parameters did not already provide. However, a lesson learnt could be in future investigations to add a parameter denoted Skewness as a parameter to be measured. This parameter is sensitive to the direction of the profile, i.e. whether the texture is positive or negative, which may be useful information on the profile, other than the range and rate of variation in surface height. Analyses made in the NordTex project looked at the method for predicting changes in tyre/road and vehicle noise levels as a consequence of changes in the surface profile, which was proposed in the international standard ISO specifying test tracks for measuring noise from road vehicles and their tyres. This prediction was found to agree well with the noise level differences measured on the dense asphalt pavements dealt with in the NordTex project. This may be seen as an indication that the project results are relevant and useful. 43

44 14. ACKNOWLEDGEMENTS The project was funded by the Nord- FoU, a co-operation between the national Nordic road administrations to initialise, finance and run R&D projects. The Swedish Transport Administration funded measurements of texture and rolling resistance to supplement the base of data. Assistance in rolling resistance and CPX noise level measurements was provided by Jerzy Ejsmont, Piotr Mioduszewski and Grzegorz Ronowski, Technical University of Gdansk, whose contributions are gratefully acknowledged. 44

45 15. REFERENCES [1] U. Sandberg, Road surface texture influence on tyre/road noise, VTI Report VV dnr BY20A2008:12388, [2] J. Kragh, Road surface texture low noise and low rolling resistance. CPX trailer comparison Copenhagen 2009, Danish Road Institute Report , ISBN Electronic [3] S. Å. Storeheier, Road surface texture for low noise and low rolling resistance Task 2 Nordic road surface texture measurement method. Field calibration and results, SINTEF Memo , Project No. 90E321 [4] S. Å. Storeheier, Road surface texture for low noise and low rolling resistance. Task 5 Texture/Noise Relations, SINTEF project memo , Project No. 90E321 [5] S. Å. Storeheier, Road surface texture for low noise and low rolling resistance. Task 5 Texture/Noise Relations, SINTEF project memo Version 2, , Project No. 90E321 [6] S. Å. Storeheier, personal communication to J. Kragh 16 May 2013 [7] S. Å. Storeheier, personal communication to J. Kragh 29 May 2013 [8] J. Kragh, Rolling resistance Copenhagen 2009, Danish Road Institute Technical Note , ISBN Electronic [9] S. Å. Storeheier presentation at seminar June 2010, CPH [10] U. Sandberg, J. A. Ejsmont (2002), Tyre/road noise reference book, ISBN Number: , Kisa, Sweden, website: info [11] U. Sandberg; G. Descornet (1980), Road Surface Influence on Tire/Road Noise Part I, and G. Descornet; U. Sandberg (1980), Road Surface Influence on Tire/Road Noise Part II. Proc. of Inter-Noise 80, Miami, pp [12] J. Kragh, NordTyre Tyre labelling and Nordic traffic noise, Draft Final Report Analysis of data collected during the summer 2012, Danish Road Directorate, 2 April 2013 [13] Acoustics Specification of test tracks for measuring noise emitted by road vehicles and their tyres, ISO 10844:2011 (E) [14] U. Sandberg; J. A. Ejsmont; A. Bergiers; L. Goubert; F. Anfosso- Lédée; M. Zöller; R. Karlsson (2011), Road surface influence on tyre/road rolling resistance. Deliverable No. 4 of MIRIAM. Downloadable from the MIRIAM website ( Road-Surf-Infl_Report% pdf accessed 5 June 2013) [15] U. Sandberg; P. Mioduszewski (2012), Gaining extra noise reduction and lower rolling resistance by grinding a porous asphalt pavement, Proc. of Inter-Noise 2012, New York City [16] U. Sandberg; P. Mioduszewski (2012), The importance for noise reduction of the bottom layer in doublelayer porous asphalt, Proc. of Acoustics 2012 Hong Kong, May 2012 [17] E. Wennberg, personal communication to J. Kragh May 2013 [18] G. Pigasse, J. Kragh, J. Oddershede, B. Andersen, Memo on spike removal, Danish Road Directorate, Doc. 14/ , rev. 15-Jun-13 [19] A. von Meier, G.J. van Blokland and G. Descornet. The influence of texture and sound absorption on the noise of porous road surfaces. In PI- ARC 2nd International Symposium on Road Surface Characteristics, pages 7 19, 1992 [20] J. Oddershede, J. Kragh, H. Bendtsen, Road surface texture and tire/road noise A pilot comparison of prediction and measurement, Proceedings of Internoise 2012 New York, August

46 APPENDIX 1 PAVEMENTS AND SELECTED MEASUREMENT RESULTS Table 10. Pavement sections in Denmark (N = 18) and some measurement results from 2009, SRTT at 50 km/h, Data set #1 Meas. by ID # Pavement Site Ref speed SRTT L AcpxP1 MPD L ME [km/h] [db] [mm] [db re 10-6 m] DRD DRI01 AC 8d Kongelundsvej S DRI02 AC 6o Kongelundsvej N DRI03 AC 11dN Kongelundsvej N DRI04 AC 11dS Kongelundsvej S DRI05 SMA 6+ Kongelundsvej S DRI06 UTLAC 6-TB 6k Kongelundsvej N DRI07 AC 8o M10 N DRI08 AC 11d M10 N DRI09 AC 11o M10 S DRI10 HRA M10 N DRI11 SMA 6P M10 S DRI12 SMA 8 M10 N DRI13 UTLAC 8-TB 8k M10 S STF DRI02 AC 6o Kongelundsvej N DRI03 AC 11dN Kongelundsvej N DRI04 AC11 ds Kongelundsvej S DRI05 SMA 6+ Kongelundsvej S DRI06 UTLA6 6-TB6k Kongelundsvej N

47 Table 11. Pavement sections in Norway (N = 16) and some SVV-STF measurement results from 2009, Old SRTT at 50 km/h, Data set #1 SRTT ID # Pavement Site Ref speed L AcpxP1 MPD L ME [km/h] [db] [mm] [db re 10-6 m] STF02 SMA 11 E6 Horg F STF03 SMA 11 E6 Omkjv F1(08) STF04 SMA 11 E6 Omkjv F2(07) STF05 SMA 8 E6 Omkjv F2(07) STF06 AC 8d RV 715 Trolla STF07 SMA 8 RV 715 Trolla STF08 AC 11d RV 715 Trolla STF09 SMA 11 RV 715 Trolla STF10 AC 6d RV 715 Trolla STF11 SMA 6 RV 715 Trolla STF12 SMA 11 ref RV 715 Trolla STF13 T8s RV020Elverum STF16 SMA6 E18 Mastemyr STF17 SMA8 E18 Mastemyr STF18 SMA11 E18 Mastemyr STF19 SMA16 E18 Mastemyr

48 Table 12. Pavement sections in Denmark (N = 41) and DRD measurement results from 2009 (M10) and 2010 (other pavements), SRTT at 80 km/h, Data set #1 ID # Pavement Site Ref speed SRTT L AcpxP1 L vehp MPD L ME [km/h] [db] [db] [mm] [db re 10-6 m] DRI30 AC 8o Igelsø Igelsø S DRI31 SMA 8 Igelsø Igelsø S DRI32 SMA 6+8 Igelsø Igelsø S DRI33 SMA 6+11 Igelsø Igelsø S DRI34 AC 6o Igelsø Igelsø S DRI35 AC 11d Igelsø Igelsø S DRI36 AC 8o Igelsø Igelsø N DRI37 SMA 8 Igelsø Igelsø N DRI38 SMA 6+8 Igelsø Igelsø N DRI39 SMA 6+11 Igelsø Igelsø N DRI40 AC 6o Igelsø Igelsø N DRI41 AC 11d Igelsø Igelsø N DRI42 AC 6o - Microville Herning-I M64 S DRI43 AC 8o - Microville Herning-I M64 S DRI44 AC 11d Herning-I M64 S DRI45 PA 8 CL2 Herning-I M64 S DRI46 SMA 6 Herning-I M64 S DRI47 SMA 6+ Herning-I M64 S DRI48 SMA 8 Herning-I M64 S DRI50 TB6k Herning-I M64 N DRI51 TB8k Herning-I M64 N DRI52 SMA11 Herning-I M64 N DRI53 PA 8 CL1 Herning-I M64 N DRI54 PA 6 CL2 Herning-I M64 N DRI55 AC 11d Herning-I M64 N DRI56 SMA 4+8 Herning-II M68 Ø DRI57 SMA 6+8 Herning-II M68 Ø DRI58 SMA 6+11 Herning-II M68 Ø DRI59 SMA 6+ Herning-II M68 Ø DRI60 AC 6o Herning-II M68 V DRI61 AC 11d Herning-II M68 V DRI62 PA 6 Herning-II M68 V DRI63 PA 8 Herning-II M68 V DRI64 SMA 6+ Herning-II M68 V DRI07 AC 8o M10 M10 N DRI08 AC 11d M10 M10 N DRI09 AC 11o M10 M10 S DRI10 HRA M10 M10 N DRI11 SMA 6P M10 M10 S DRI12 SMA 8 M10 M10 N DRI13 UTLAC 8-TB 8k M10 M10 S

49 Table 13. Pavement sections in Norway (N = 5) and STF measurement results from , SRTT at 80 km/h, Data set #1 ID # Pavement Site Ref speed SRTT L AcpxP1 MPD L ME [km/h] [db] [mm] [db re 10-6 m] STF02 SMA11 Horg F1 E6 N STF16 SMA6 Mastemyr F2 E STF17 SMA8 Mastemyr F2 E STF18 SMA11 Mastemyr F2 E STF19 SMA16 Mastemyr F2 E Table 14. Pavement sections in Denmark, N = 18, and DRD measurement results from 2009, CPX reference tyre A, Data set #2 ID # Pavement Site Ref speed Tyre A L AcpxP1 MPD L ME [km/h] [db] [mm] [db re 10-6 m] DRI01 AC 8d Kongelundsvej S DRI02 AC 6o Kongelundsvej N DRI03 AC 11d Kongelundsvej N DRI04 AC 11d Kongelundsvej S DRI05 SMA 6+8 Kongelundsvej S DRI06 UTLAC 6 Kongelundsvej N DRI14 SMA6+8 M45 S DRI15 SMA6+8 M45 N DRI20 AC 11d Kastrupvej N DRI21 AC 11d Kastrupvej S DRI22 SMA 6 Kastrupvej S DRI23 SMA 6+8 Kastrupvej S DRI24 SMA 4+8 Kastrupvej N DRI25 SMA 4 Kastrupvej S DRI26 SMA 6+8 Kastrupvej N DRI27 AC 6o Kastrupvej S DRI28 SMA 6 Kastrupvej N DRI29 SMA 6+8 Kastrupvej N

50 Table 15. Pavement sections in Norway (N = 37) and STF measurement results from 2008, CPX reference tyre A at 50 km/h, Data set #2 ID # Pavement Site Ref speed Tyre A MPD L ME L AcpxP1 [km/h] [db] [mm] [db re 10-6 m] Rv715 Trheim Rv715 Trheim Rv715 Trheim Rv715 Trheim Rv715 Trheim Rv715 Trheim E6 Melhus E6 Melhus E18 Oslo E18 Oslo E18 Oslo E18 Oslo E16 Hønefoss E16 Hønefoss E16 Hønefoss E6 Stange E6 Stange Rv2 Kongsv Rv2 Kongsv Rv161 Oslo Rv161 Oslo E6 Stjørdal E6 Stjørdal E6 Trheim E6 Trheim Rv20 Elverum Rv62 Eidsvåg Rv118 Rygge Rv2 Kongsv., Rasta, Rv2 Kongsv., Rasta, Rv170 Bjørkelangen E6 Omkjvn. S Moholt E6 Omkjvn. N F3 Moholt E6 Omkjvn. Sgående F4 Tunga Rv715 ref Trolla E6 Stjørdal Ng tunnel Rv62 Hp4 Eidsvåg

51 Table 16a. Dense-graded pavement sections tested in Sweden, N = 27, at a speed of 50 km/h. VTI measurement results , using CPX reference tyres P1 and H1. No temperature correction of noise levels made. Results of tests when texture was not measured have been excluded. ARG means asphalt rubber, gap-graded and TAL means thin asphalt layer ID # Pavement Pavement designation designation Site Meas. Pavem. Ref Tyre P1 Tyre H1 year age speed L AcpxP1 L AcpxH1 MPD L ME international Swedish [years] [km/h] [db] [db] [mm] E22-2 SMA16 ABS16 E22 Malmö-Lund ,8 92,6 1,20 NA E22-3 DAC11 ABT11 E22 Malmö-Lund ,8 90,1 0,41 NA AR-3 ARG11 ARG11 E6 Vellinge ,4 91,4 0,76 NA AR-5 ARG11 ARG11 E6 Vellinge ,0 90,7 0,75 NA AR-6 ARG16 ARG16 E6 Malmö ringroad ,6 0,99 NA RV34-1 SMA16 ABS16 RV34 Linköping-Kåparp ,7 93,7 1,09 NA RV34-2 DAC11 ABT11 RV34 Linköping-Kåparp ,5 90,5 0,61 NA RV34-1 SMA16 ABS16 RV34 Linköping-Kåparp ,5 91,8 0,86 NA RV29-3 TAL16 TSK16 Peab RV29 South of Tingsryd ,6 91,7 1,12 NA E22-TAL1 TAL16 TSK16 Sksk E22 Ronneby-Br_Hoby ,5 91,6 0,99 NA E22-TAL2 TAL16 TSK16 Sksk E22 Ronneby-Br_Hoby ,6 92,3 0,66 NA RV13-Ref SMA16 ABS16 E22 Southwest of Hörby ,0 92,2 0,89 NA RV13-1 DAC8 ABT8 RV13 West of Höör ,0 89,5 0,35 NA RV13-2 SMA8 ABS8 RV13 West of Höör ,1 88,7 0,79 NA RV13-3 DAC11 ABT11 RV13 West of Höör ,4 90,4 0,37 NA RV13-4 SMA11 ABS11 RV13 West of Höör ,8 90,2 0,76 NA RV13-Ref SMA16 ABS16 E22 Southwest of Hörby ,5 92,9 0,99 NA RV13-1 DAC8 ABT8 RV13 West of Höör ,2 90,4 0,56 NA RV13-2 SMA8 ABS8 RV13 West of Höör ,6 90,4 0,68 NA RV13-3 DAC11 ABT11 RV13 West of Höör ,4 91,1 0,45 NA RV13-4 SMA11 ABS11 RV13 West of Höör ,5 91,5 0,71 NA RV11-1a TAL11 TSK11 RV11 Tomelilla-Lunnarp ,8 92,1 0,63 NA RV11-1b TAL11 TSK11 RV11 Tomelilla-Lunnarp ,2 91,4 0,65 NA RV34-1 SMA16 ABS16 RV34 Linköping-Kåparp ,2 92,6 0,91 NA NA-1 SMA11 ABS11 Nya Allén, Göteborg ,7 91,1 1,09 NA NA-2 SMA8spec ABS8 special Nya Allén, Göteborg ,6 89,9 0,81 NA JF-1 SMA8slag ABS8 stslagg Skälbyvägen, Järfälla ,8 89,1 0,82 NA 51

52 Table 16b. Dense-graded pavement sections tested in Sweden, N = 24, at a speed of 80 km/h. VTI measurement results , using CPX reference tyres P1 and H1. No temperature correction of noise levels made. Results of tests at speed > 90 km/h or when texture was not measured have been excluded. In some cases measurements were made at 70 and 90 km/h. Interpolation has then been made to 80 km/h to ensure comparability. ARG means asphalt rubber, gap-graded and TAL means thin asphalt layer ID # Pavement Pavement designation designation Site Meas. Pavem. Ref Tyre P1 Tyre H1 year age speed L AcpxP1 L AcpxH1 MPD L ME international Swedish [years] [km/h] [db] [db] [mm] E22-2 SMA16 ABS16 E22 Malmö-Lund ,4 99,9 1,20 NA E22-3 DAC11 ABT11 E22 Malmö-Lund ,9 98,6 0,41 NA AR-3 ARG11 ARG11 E6 Vellinge ,3 99,3 0,76 NA AR-5 ARG11 ARG11 E6 Vellinge ,4 98,8 0,75 NA AR-6 ARG16 ARG16 E6 Malmö ringroad ,4 99,4 0,99 NA RV34-1 SMA16 ABS16 RV34 Linköping-Kåparp ,7 101,0 1,09 NA RV34-2 DAC11 ABT11 RV34 Linköping-Kåparp ,8 98,7 0,61 NA RV34-1 SMA16 ABS16 RV34 Linköping-Kåparp ,9 99,2 0,86 NA RV29-3 TAL16 TSK16 Peab RV29 South of Tingsryd ,8 98,7 1,12 NA E22-TAL1 TAL16 TSK16 Sksk E22 Ronneby-Br_Hoby ,7 98,8 0,99 NA E22-TAL2 TAL16 TSK16 Sksk E22 Ronneby-Br_Hoby ,6 99,4 0,66 NA RV13-Ref SMA16 ABS16 E22 Southwest of Hörby ,4 99,7 0,89 NA RV13-1 DAC8 ABT8 RV13 West of Höör ,0 98,1 0,35 NA RV13-2 SMA8 ABS8 RV13 West of Höör ,2 96,5 0,79 NA RV13-3 DAC11 ABT11 RV13 West of Höör ,5 99,2 0,37 NA RV13-4 SMA11 ABS11 RV13 West of Höör ,9 97,8 0,76 NA RV13-Ref SMA16 ABS16 E22 Southwest of Hörby ,6 100,2 0,99 NA RV13-1 DAC8 ABT8 RV13 West of Höör ,6 97,5 0,56 NA RV13-2 SMA8 ABS8 RV13 West of Höör ,0 97,0 0,68 NA RV13-3 DAC11 ABT11 RV13 West of Höör ,0 98,2 0,45 NA RV13-4 SMA11 ABS11 RV13 West of Höör ,5 98,5 0,71 NA RV11-1a TAL11 TSK11 RV11 Tomelilla-Lunnarp ,5 98,7 0,63 NA RV11-1b TAL11 TSK11 RV11 Tomelilla-Lunnarp ,9 97,9 0,65 NA RV34-1 SMA16 ABS16 RV34 Linköping-Kåparp ,0 99,5 0,91 NA 52

53 APPENDIX 2 TABLES OF CORRELATION 53

54 Table 17. Correlation coefficient R for combinations of texture wavelength and CPX noise level; old reference tyre A at 50 km/h on 37 Norwegian pavements, Data set #1, as in the top part of Figure 10 on p. 19 Frequency [Hz] Observations: Low frequencies: R 2 > 0.64: 200mm 80mm, Hz; max R 2 = v/ 200mm/800 Hz High frequencies: R 2 > 0.39: 8-6,3mm, Hz; max R 2 = Wavelength [mm] 54

55 Table 18. Correlation coefficient R for combinations of texture wavelength and CPX noise level; reference tyre SRTT at 80 km/h on 41 Danish pavements, ds1_80, as in the bottom part of Figure 10 on p. 19 Frequency [Hz] Observations: Low frequencies: R 2 > 0.59: 125mm 50mm, Hz; max R 2 = v/ 100mm/630 Hz High frequencies: R 2 > 0.30: 4-2mm, z; max R 2 = Wavelength [mm] 55

56 APPENDIX 3 DIFFERENCES IN TEXTURE SPECTRA MEASURED BY THE PARTIES SVV-STF measured higher texture spectral levels at small wavelengths, than DRD did on the same pavement. This was mentioned in Section Therefore, when data from both parties are pooled, one should expect low or no correlation between noise and texture in this range. This was in fact the case in some of the results in Section 6 on the relations between octave band texture and noise levels. Other analyses mentioned were based exclusively on sets of data collected using the same equipment, either that of DRD or that of SVV-STF. The main outcome of these latter analyses was a general confirmation of results found in earlier studies: To obtain low tyre/ road noise levels one shall aim at low texture levels at wavelengths around 80 mm and high texture levels at wavelengths around 5 mm. However, this does not mean that individual results follow the general trend, as illustrated in the following. The results in the upper part of Figure 7 on p. 17 have been repeated in the caption of Figure 25. Two groups of pavements having a texture level L txt80 of about 45 db re 10-6 mm (actually db) come out with db different noise levels: one group shown in green, measured by DRD on M10, the other shown in red, measured by SVV-STF on RV 715 at Trolla. With the same texture at 80 mm wavelength the noise level should be the same at low frequencies, according to observations mentioned in Section 6. Thus, to obtain higher overall A-weighted noise levels on M10 than on RV 715, with the same texture levels at 80 mm and therefore the same noise levels at low frequencies, one would expect higher noise level contributions from high frequencies on M10. Such higher noise levels at high frequencies should be associated with lower texture levels at small wavelengths. But the data in Figure 25 show that the texture levels measured by DRD on M10 were actually higher, not lower, than the texture levels measured by SVV-STF on RV 715, as displayed in the upper part of Figure 25. This is the case, even though DRD in its analyses applied enveloping and therefore reduced the texture levels at small wavelengths. The trailers used for noise measurement were compared on M10, see [2] and Figure 4, with the result that noise levels measured with the SVV-STF trailer were 0.3 db or so higher than those measured with the DRD trailer. Thus the results shown in Figure 25 do not seem to follow the overall trend, but the reasons for this are unknown. Figure 26 illustrates that also within the Norwegian set of data the general trend is not necessarily followed by individual results. Figure 26 repeats the Norwegian results from Figure 25, supplemented with results from SMA 11 on E18 at Mastemyr. On the latter pavement L CPXA was as high as on the Danish M10 pavements looked at in Figure 25 while the texture level at 80 mm, 42 db re 10-6 mm was slightly lower, see Figure 7. Low frequency noise levels on SMA 11 at Mastemyr were identical with the noise levels on the pavements on RV 715 at Trolla, while noise levels at 1 khz and higher frequencies were up to 5 db higher on SMA 11 at Mastemyr than on the pavements at Trolla. Also in this comparison, higher texture levels at small wavelengths were not associated with lower noise levels at high frequencies. 56

57 Figure 25. Texture (top) and CPX noise level spectra (bottom) measured by DRD on M10 and by SVV- STF on RV 715 at Trolla. The legend gives the pavement type; mean profile depth MPD; macrotexture level LME; and overall CPX noise level. The small diagram to the right identifies the data points 57

58 Figure 26. Texture (top) and CPX noise level spectra (bottom) measured by SVV-STF on RV 715 at Trolla and on E18 at Mastemyr. The legend gives the pavement type; mean profile depth MPD; macrotexture level LME; and overall CPX noise level 58

59 APPENDIX 4 ABBREVIATIONS USED IN THE REPORT Abbreviation ABS ABT AC d AC o AOT ARG BRRC CPX DAC DRD DRI ds1_50 EACC ENDt ERNL HRA G-factor IRI L A L CPXA L ME L txyy MPD NMAS PA RRC SMA SPB SRF SRTT STF SVV-STF TAL TB k TUG UTLAC VTI Explanation Swedish designation for SMA Swedish designation for AC d / DAC Dense graded asphalt concrete; same as DAC Open graded asphalt concrete Acoustic optimisation tool Asphalt rubber, gap-graded Belgian Road Research Centre Close proximity Dense asphalt concrete; same as AC d Danish Road Directorate Danish Road Institute Data set #1 measured at 50 km/h Exposed aggregate cement concrete Expected pass-by noise level difference due to texture differences Estimated road noisiness level Hot rolled asphalt Profile shape characteristic (German concept denoted Gestaltfaktor) International roughness index A-weighted noise level or LCPXA = A-weighted noise level measured with CPX trailer Megatexture level Third octave-band texture level at centre wavelength yy mm Mean profile depth Nominal maximum aggregate size Porous asphalt Rolling resistance coefficient Stone mastic asphalt Statistical pass-by Surface type Standard reference test tyre SINTEF Norwegian Public Roads Administration and SINTEF Thin asphalt layer Danish designation for UTLAC Technical University of Gdansk Ultra thin asphalt layer Swedish National Road and Transport Research Institute 59

60 The Danish Road Directorate s headquarter is situated in Copenhagen and local offices are situated in Aalborg, Herning, Skanderborg, Middelfart, Næstved, Fløng and Herlev. You will find more information on VEJDIREKTORATET Niels Juels Gade 13 Postboks Copenhagen K Tel.: [email protected] vejdirektoratet.dk