Durable laboratory rubber friction test countersurfaces that replicatethe roughness of asphalt pavements

A. Kuosmanen a, T. Pellinen a, L. Hartikainen b,n, F. Petry b, S. Westermann b

a Department of Civil and Environmental Engineering, School of Engineering, Aalto-University, PO Box 12100, FI-00076 Aalto, Finlandb Goodyear S.A., Avenue Gordon Smith, Colmar-Berg L-7750, Luxembourg

a r t i c l e i n f o

Article history:Received 11 June 2014Received in revised form25 September 2014Accepted 28 September 2014

Keywords:Rubber frictionEpoxy asphaltConcreteSliding wearSurface topography

a b s t r a c t

This article presents the development of a method for manufacturing durable countersurfaces for testingthe friction of rubber samples in a laboratory. The surface sample was designed to replicate the surfaceroughness characteristics of a predefined road surface. Four different types of samples were manufac-tured: two bitumen bound stone mastic asphalt samples, one concrete bound sample, and one epoxybound sample. In addition, the surfaces were sandblasted to remove binder from the top of the surface.The similarity in surface roughness between the predefined target road surface and the laboratorysurface samples was evaluated using optical 3D surface roughness measurements and surface roughnesspower spectrums. The durability of the surfaces was investigated by repeated dry rubber sliding frictiontests. It was found that out of the studied surface samples the best one to both replicate the surfacecharacteristics of the target surface and to produce a durable surface in terms of wear was the one usingepoxy resin as a binder. By replacing bitumen with epoxy, more friction tests could be run withoutsignificant changes in the surface topography or rubber friction. In addition a reasonable correlationbetween the surface roughness power spectrums and the friction results was found, which supports theuse of roughness power spectrums as a roughness metric for road surfaces in the context of rubberfriction.

& 2014 Elsevier B.V. All rights reserved.

1. Introduction

Understanding tire–pavement interaction and especially tirefriction is very important for tire manufacturers to develop tireswith continuously increasing performance and safety. An impor-tant research tool for reaching this goal is a laboratory rubberfriction tester that can compare the friction performance of variousrubber samples. However, performing meaningful friction testswith rubber samples even in laboratory conditions is challengingdue to the effect of the pavement surface roughness. On one hand,the surface roughness of the laboratory road sample should matchthat of the intended tire–road application. On the other hand, theroughness should also remain as constant as possible throughouttest programs to provide comparable results for the differentrubber samples and to reduce test variability. It is known thatrubber–road friction evolves during the lifetime of the pavementas a result of the aggregate polishing and wearing, binder removal

and the aging of the binder [1]. Therefore, developing a testsurface for laboratory tests that minimizes these changes duringthe targeted life span of the surface allows more repeatable andmeaningful results from laboratory rubber friction tests. Develop-ing a method for manufacturing such surfaces and their quantita-tive comparison is the main topic of this study. The study focuseson dry rubber friction and does not include the effects of studdedtires, water, or abrasives.

Different kinds of sandpapers and corundum surfaces providequite durable countersurfaces that are easy to source for labora-tory rubber friction testing. However the surface roughness ofthese surfaces is typically quite different from those of pavements,which represent the final application for tires. Another possibilityis to take core samples from existing road surfaces, but this is ofcourse destructive to the target surface and cannot guaranteerepeatable properties for the samples in the long run. Differentmethods for producing durable surfaces that resemble asphalt forlaboratory testing can be found in literature, such as the mosaic ofcoarse aggregates glued together with epoxy resin presented in [2]and the two-step casting method explained in [3]. In the castingmethod the macro-roughness of the pavement was replicated bycovering the surface with liquid silicone, hence creating a negativeprofile of the surface. This negative profile was then used to cast apositive profile in a more durable material, namely synthetic resin

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/wear

Wear

http://dx.doi.org/10.1016/j.wear.2014.09.0110043-1648/& 2014 Elsevier B.V. All rights reserved.

n Corresponding author. Tel.: þ352 8199 4699.E-mail addresses: antti.kuosmanen@aalto.fi (A. Kuosmanen),

terhi.pellinen@aalto.fi (T. Pellinen),lassi_hartikainen@goodyear.com (L. Hartikainen),frank.petry@goodyear.com (F. Petry),stephan_westermann@goodyear.com (S. Westermann).

Wear 321 (2014) 38–45

loaded with corundum particles. These methods can producedurable surfaces, but do not replicate the roughness of realpavements on all length scales. As in principle rubber frictiondepends on roughness at all length scales [4–6], a method forreplicating the full roughness spectrum needs to be developed.

To reach the goals of both surface roughness characteristics anddurability, countersurface samples were prepared using threealternative binders: bitumen, epoxy and cement. Epoxy andcement were chosen as binder candidates due to their well-known ability to bind aggregates together. In addition, a secondbitumen sample was built using polished aggregates. The mixtureswere compacted to small slabs and the applicability of eachsurface type was assessed using 3D surface topography scansand repeated laboratory rubber friction measurements.

2. Materials and methods

2.1. Experimental procedure

The experimental procedure for the four test samples isillustrated in Fig. 1. The purpose was to attain similar surfaceroughness as in a target stone mastic asphalt SMA8 road surface.

Thus, aggregate sizes up to 8 mm were used in all of the foursurface types. Additionally, the target was to achieve a durablesurface to allow more repeatable results from rubber friction tests.Since the hardness of the aggregates is one of the main factors ofroad surface wear resistance, the aim was to select the hardestavailable aggregates. Furthermore, because the aging of bitumen isone of the principal factors for asphalt pavement deterioration [7],two alternative binders, namely epoxy and cement, were used inaddition to bitumen. In order to avoid changes in the friction levelsdue to binder removal, as shown in [2], the binder was preemp-tively removed from the fresh sample surface by sandblasting.After the binder removal was completed, the initial friction androughness levels of the test sample surfaces were measured. Then,the surfaces were polished by sliding a rubber sample againstthem 100 times before measuring the friction and surface rough-ness again. These polishing sessions were repeated 1–10 times,depending on the durability of the samples.

2.2. Aggregate selection and job mix formula

Granite from Koskenkylä quarry, located in southern Finland, iswidely used on high-volume roads in Finland due to its highresistance to wear by studded tires. This type of granite typically

Fig. 1. Schematic diagram of the experimental procedure.

A. Kuosmanen et al. / Wear 321 (2014) 38–45 39

consists of quartz (30%), plagioclase (20%), orthoclase (25%), Biotite(10%), and Amphibole (10%) [8]. In order to build high durabilitylaboratory samples, Koskenkylä aggregates were used in the sur-face samples. The durability of the aggregates was measured usingthe Nordic abrasion test, described in the standard SFS-EN 1097-9.This ball mill type test is primarily used for determining theresistance to wear by the abrasion from studded tires. While thesurfaces were not exposed to wear by studs in this study, theNordic abrasion value was assumed to be a good indicator ofaggregate durability. The Nordic abrasion test result was 5.7% forthe chosen aggregate. The average values for different minerals aretypically in the range of 5–20% [9], with lower values indicatingbetter wear resistance.

One of the samples was also fabricated using a stone fraction of5/8 mm polished in a Nordic abrasion test machine. The purposeof this sample was to see if this pre-polishing would help tostabilize the surface micro-roughness and friction results at anearlier stage. However, as explained in the results of this article,the durability of this sample was not high enough to properlyassess the effect.

The target job mix formula for the SMA (stone mastic asphalt)samples was determined according to the Finnish Asphalt Speci-fications [10]. Additionally, limestone filler (9%) and 0.03% ofcellulose fiber were added to the bitumen and epoxy mixes asrequired by the Finnish Asphalt Specifications. The concrete mixwas designed based on a few experimental trials. Fig. 2 shows thegradation curves for the final surface samples. A small differencebetween the bituminous sample with unpolished aggregates andthe epoxy and concrete samples can be seen. This was due to theaggregates coming from different batches of the same material.Nevertheless, the Nordic abrasion test results were the same forboth batches. The gradation curves of the epoxy and the concretemixes were designed to represent the asphalt sample with

polished aggregates. The cement used in the concrete samplewas not included in the calculation of the gradation curve.

2.3. Sample preparation

The stone mastic asphalt samples were manufactured in thelaboratory according to standard SFS-EN 12697-35 and compactedusing a pneumatic smooth steel roller, according to the standardSFS-EN 12697-33. The physical properties of the SMA, epoxy, andconcrete samples are summarized in Table 1. The size of the testslabs was 12�12 cm2.

The same job mix formula was used for the epoxy asphalt as forthe SMA sample with polished aggregates. The mixing wasperformed in three phases. First, the epoxy resin was mixed withthe hardener. Then, half of the epoxy–hardener mix was mixedtogether with the coarse aggregates (5/8 mm). Finally, when theepoxy had adhered evenly to the coarse aggregates, the fineaggregates were poured into the mixing bowl with the rest ofthe epoxy–hardener mix. The mixing was continued until themass was evenly blended and no segregated spots were observed.The mass was spread to a mold and vibrated on a vibration tablefor 2 min. After this, the surface was compacted with a roller. Aftercompacting, the slab was cured for 24 h at the temperature of22 1C. In addition, the slab was post-cured at a temperature of50 1C for 8 h to enhance its mechanical properties. After the curingthe slab was flipped over and the bottom side of the cured slabwas used for friction tests, ensuring a flat surface. The top surfaceof the SMA sample, on the other hand, was flat enough aftercompaction and did not need to be flipped over.

For the concrete samples, the purpose was to maximize theamount of cement in order to achieve high durability. Sampleswere manufactured in a laboratory with a mixer and a vibrationtable. The solids including aggregates and cement were mixed for

0.063 0.125 0.25 0.5 1 2 4 5.6 8 11.20

10

20

30

40

50

60

70

80

90

100

Sieve size [mm]

Pas

sing

[%]

Stone mastic asphalt, polished aggregatesStone mastic asphalt, unpolished aggregatesEpoxy and concrete surfacesLower and upper limit

Fig. 2. Gradation curves for the aggregate mixes. All mixes are contained within the lower and upper limits of the Finnish Asphalt Specifications, as shownwith dashed lines.

Table 1Material properties of the surface samples.

Designation Stone mastic asphalt samples Epoxy samples Concrete samples

Aggregate type Granite Granite GraniteBulk specific gravity of aggregates [kg/dm3] 2.67 2.67 2.67Nordic abrasion value [%] 5.7 5.7 5.7Binder type Bitumen 50/70 Epoxy Rapid cement (CEM I 52.5R)Dynamic viscosity [Pa s] 185 (at 60 1C) 1.529 (at 20 1C before curing) –

Binder specific gravity [kg/dm3] 1.03 1.101 3.1 (cement)Binder content [%] 7 7 0.42 (water–cement ratio)Plasticizer [% of cement weight] – – 1.5

A. Kuosmanen et al. / Wear 321 (2014) 38–4540

60 s in a mixing machine. Next, water was poured together withthe plasticizer and the mixing continued until a fluid mass wasattained. Then, the mixed mass was poured into a mold in threeparts. After each pouring, the mass was vibrated for 20 s on thevibration table. Finally, the concrete slab was leveled with a screedand covered for 24 h with a plastic coat to avoid the evaporationof water. Similar to the epoxy sample, the slab was finally flip-ped over and the bottom side of the cured slab was used forfriction tests.

2.4. Binder wearing process

To remove the binder from the surfaces, the bitumen boundsamples were first sandblasted with water using sand withaggregate sizes from 0.6 mm to 1.1 mm and water with an airpressure of 8 bar. The amount of sand needed was approximately22 kg/m2. The slabs were first blasted vertically from a distance of20 cm and then from four different directions with an angle of 451.The total duration of blasting was approximately 26 min/m2. Itwas observed, however, that the removal of bitumen was notcomplete after sandblasting with water, as seen in the middlesection of Fig. 3. In many areas, smaller aggregates were stillcovered by bitumen. Because of that, it was decided to sandblastthe slabs also without water in order to remove the rest of thebitumen. Fine sand with aggregate sizes from 0.1 to 0.6 mm wasblasted onto the surface from five directions with an air pressureof 6 bar. The amount of sand and the duration of blasting wereapproximately the same as with the water sandblasting. The finalresult is illustrated in the bottom section of Fig. 3.

To remove the cement from the top of the concrete surface,sandblasting without water was used. The technique was almostthe same as for the SMA sandblasting. However, hardened cementis much harder than bitumen and therefore the duration of thesandblasting was longer – approximately 165 min/m2. For theepoxy surface, the methods and specifications were almost iden-tical to those used for the concrete surface. The duration of thesandblasting needed for removing the epoxy from the top of thesurface was about 110 min/m2, which was slightly less than withthe concrete surfaces but much more than with the bitumensurfaces.

2.5. Friction tests

The friction on each surface sample was measured on two lanes20 mm wide and 85 mm long. The wear cycles using a 20�20 mm2 rectangular sliding rubber block made out of a rubbercompound of a modern passenger car summer tire were appliedon one of the two lanes and the other one was used as a reference.The test conditions applied and the rubber sample type used

during the wear cycles were identical to those used for the frictionmeasurements. The friction tests were performed using a high-precision linear laboratory friction tester. During the test therubber block is loaded onto the asphalt sample with a specifiedload, and then slid for a given distance at a given speed, while thelongitudinal and vertical forces as well as the rubber blocktemperature are being measured. The test parameters were setso that the nominal contact pressure was 0.3 MPa and the slidingspeed was 7 mm/s. The asphalt surface was dry during the tests.The test parameters were chosen to support fundamental researchon rubber friction, and not necessarily to replicate typical roadconditions. A schematic of the operating principle of the device isshown in Fig. 4.

The resulting longitudinal force was divided by the verticalforce to provide a coefficient of friction (μ). The average value ofthe μ was calculated for the whole sliding distance to provide afriction value representing the whole lane. These values werefurther normalized by dividing them with the average value fromall studied lanes. An example of a friction measurement result canbe seen in Fig. 5. The data on this graph is normalized by dividingeach point by the average μ of this test only.

2.6. Surface scans

The surface scans were performed with a MicroCADlite fringe-projection type optical 3D scanner manufactured by GFMesstech-nik GmbH. The field-of-view of the device was 30�40 mm2 andspatial resolution 55 μm while the height resolution was 3 μm.The test lanes were captured by performing four scans on each.

Fig. 3. Freshly prepared asphalt surface (top) was sandblasted with water (middle).The removal of binder was finished by sandblasting without water (bottom).

Fig. 4. Schematic of the operating principle of the laboratory friction tester.

Fig. 5. Example of a measured friction curve on the reference lane of the bitumensample with non-polished stones. The curve is normalized by dividing each pointby the average value of the curve.

A. Kuosmanen et al. / Wear 321 (2014) 38–45 41

Due to the relatively small size of the area-of-interest, the areacould be covered completely down to the resolution of the device.

2.7. Surface analysis

In the context of modern rubber friction theories, the height-difference correlation function and the power spectral densitydenoted here as C(q) have been used as surface roughness metrics[4,11,12]. The benefit of these methods is that they characterizethe roughness of the surface on all measured length scales, none ofwhich can be a priori excluded from contributing to friction [4].The power spectral density of the surface roughness can bedefined as [4,11]

CðqÞ ¼ 1ð2πÞ2

Zd2x ⟨hðxÞhð0Þ⟩e� iq Ux

where x is a two-dimensional vector in the mean plane of thesurface and q is a two-dimensional wave vector, or spatialfrequency. The corresponding wavelength for each value of q canbe calculated as 2π/q. The C(q) function can also be described asthe Fourier transform of the autocorrelation function of the surfaceheight data. This spatial frequency dependent description of thesurface roughness is a central part of the calculation of kineticfriction by Persson in both the basic form of the theory [4] and itsextension taking into account the so-called flash temperatures [6].Because of the central role in this very promising rubber frictiontheory, the C(q) function was used in this study to characterize theasphalt surface roughness.

The surface analysis started by cropping out the part of thesurface where the rubber was sliding during the friction tests.Then, non-physical peaks and missing pixels were filled in byinterpolating between the surrounding pixels. After this the offsetand tilt were removed from the data and power spectral density,or C(q) curves were calculated for each measurement file. Finally,the C(q) curves from four different scans on each lane wereaveraged to provide an average roughness spectrum.

3. Results and analysis

The performance of each asphalt sample was evaluated usingfour criteria: initial state and evolution of surface topography,durability based on visual observations, and evolution of frictionresults. In addition, to gain a deeper understanding of the relationbetween the asphalt surface topography and rubber friction in thistest setup, the surface topography results were correlated againstthe friction results similar to the method presented in [13].

3.1. Surface topography matching

The main goals of this study were to provide a durable asphaltsurface for lab friction testing and to be able to replicate thesurface roughness power spectrum of a given asphalt road surface.The target road surface was characterized on-site using a similardevice as for the laboratory samples, but with a slightly higherspatial resolution. A summary of the surface topography results isgiven in Fig. 6. The C(q) functions are normalized by dividing eachvalue with the average root mean square (RMS) roughness of allthe analyzed surfaces. As can be seen in the figure, the bitumenand epoxy laboratory samples are able to replicate the roughnesspower spectrum of the target road surface to a good extent. Theconcrete sample, on the other hand, provides much lower overallroughness levels. This is likely caused by the high level of fineaggregates needed in the cement mix. These aggregates fill in thevalleys between the larger aggregates, effectively reducing theroughness on many length scales. Other, smaller, differences can

be seen between the curves. However, when considering typicalconfidence intervals for such curves as shown in Fig. 9, only theconcrete surface stands out as clearly different.

Fig. 6 also shows the self-affine behavior of the surface rough-ness above a certain roll-off frequency of around 1 rad/mm, whichcorresponds to a roughness wavelength of about 6 mm. This isclose to the size of the largest 8 mm aggregates used in the mix.Below this frequency, there is typically a plateau region in thecurve because the largest aggregate size has been reached and theroughness amplitude no more increases with increasing wave-lengths. In Fig. 6, a slight slope can be seen instead of a plateau.This is due to the small field-of-view used in the analysis. Theanalyzed image size was only 20 mmwide and 20 mm high, whichleft merely a few of the largest aggregates in each analyzed image.Above the roll-off frequency the roughness shows a self-affinescaling until the largest measured frequency, which is limited bythe lateral resolution (pixel size) of the measurement device. Inthis case the pixel size for the experimental asphalt scans was55 μm, and therefore based on the Nyquist–Shannon samplingtheorem the shortest measurable wavelength is 110 μm. In termsof spatial frequency the value is around 57 rad/mm, which is thehighest value seen in the plot for the experimental asphalts. Thetarget asphalt was scanned at a slightly higher spatial resolution of31 μm, which corresponds similarly to the maximum spatialfrequency of about 100 rad/mm seen in the plot.

3.2. Laboratory asphalt durability

3.2.1. Visual observationsThe wear lanes of the surface samples before and after the

wearing sessions are presented in Fig. 7. Both of the stone masticasphalt samples shown in Fig. 7a and b were highly deterioratedduring the tests. The sample with unpolished aggregates shown inFig. 7a was removed from the test after 50 rubber sweeps. Mostprobably the sandblasting deteriorated the surface by removingmastic from between the coarse aggregates and thus weakeningthe aggregate interlocking. The sample with polished aggregatesshown in Fig. 7b remained undamaged slightly longer, but ravelingwas also observed after 150 rubber sweeps. The difference in thedurability between these samples may have been caused by the

10−1 100 101 102 10310−8

10−7

10−6

10−5

10−4

10−3

10−2

10−1

100

Spatial frequency q [rad/mm]

C(q

)/RM

S [m

m3 ]

Bitumen Polished StonesBitumenConcreteEpoxyTarget asphalt

Fig. 6. Normalized roughness power spectral densities of the target road surfaceand the laboratory replicates. A good match is achieved with the epoxy andbitumen samples, while the concrete sample provides much lower roughnessamplitudes.

A. Kuosmanen et al. / Wear 321 (2014) 38–4542

smoother surface of the polished sample. Due to the smoother,more closed, surface of the slab with polished aggregates, lessmastic was blasted away from the valleys. Thus, the durability wasslightly higher compared with the other bitumen bound sample,but far below the target of the study.

Ten wearing cycles were performed for the concrete and epoxysamples, with a total of 1000 rubber sweeps applied. At the end ofthe wearing sessions, traces of oil from the rubber were clearlyvisible on the light concrete surface. The only visible damage onthe concrete surface was the broken aggregate near the left edge ofthe wear lane (Fig. 7c).

After the wearing session, the epoxy surface showed no notice-able signs of damage (Fig. 7d). Some traces of rubber and oil wereleft on the surface, but no stones were lost or broken. Strongadhesion between the epoxy and the aggregates prevented theloss of stones, and most probably no raveling would occur even ifthe number of wear cycles would be increased.

3.2.2. Friction evolutionThe evolution of the measured dry friction coefficient on the

concrete and epoxy lanes is shown in Fig. 8. The figure shows theaverage measured friction coefficient on the wear lane divided bythe average friction coefficient on the reference lane. This helpseliminate the effect of any environmental factors such as ambienttemperature or humidity. No clear trend can be established foreither of the surface types. It is therefore concluded that duringthe test period used, there was no significant change in the frictionmeasured on the concrete and epoxy surfaces.

3.2.3. Surface topography evolutionAnother way of assessing the durability of the asphalt surface in

laboratory friction testing was to quantify the changes in surfacetopography. A very slight increasing trend in the roughness at theshortest measured length scales was observed with increasing wearcycles, but this was not statistically significant. To illustrate the

stability of the surface topography, the C(q) functions of the epoxysurface at the initial and final stages are shown in Fig. 9. The C(q)functions are again normalized by dividing each value with theaverage root mean square (RMS) roughness of all the analyzedsurfaces. The graph shows that the roughness power spectrums areidentical down to the resolution of the measurement. The dashedlines show the 95% confidence intervals for each C(q) curve. Bothmain curves are easily contained within each other's confidenceintervals, so no statistically significant difference can be claimed.

Similar plots for the concrete surface are shown in Fig. 10.Again, no statistically significant difference can be observed in theroughness power spectrum throughout the study. A small devia-tion between the main lines can be seen at long wavelengths. Thisis due to a single stone cracking in the concrete wear lane near theend of the study.

Fig. 7. The surface samples before and after the wearing sessions: a) stone mastic asphalt with unpolished aggregates, b) stone mastic asphalt with polished aggregates,c) concrete surface and d) epoxy surface. Raveling can be clearly seen on both bitumen bound samples.

0 200 400 600 800 10000.9

0.92

0.94

0.96

0.98

1

1.02

1.04

1.06

1.08

1.1

Number of friction test sweeps [−]

µ wea

r lan

e / µ

frict

ion

lane

ConcreteEpoxy

Fig. 8. Average friction coefficient measured on the wear lane divided by the valuesmeasured on the reference lane. No clear trend can be established.

A. Kuosmanen et al. / Wear 321 (2014) 38–45 43

3.3. Correlation between surface roughness and friction

Both friction and surface characterization measurementsshowed that no significant changes took place in the surfaceroughness of the epoxy or concrete lanes during the test sessions.To verify that this conclusion was not simply a result of a lack ofaccuracy in the friction and surface roughness measurements, theaverage friction values from all the test lanes were correlatedagainst the corresponding surface roughness results. Only theinitial test results before the wear cycles from each lane wereused in the correlation to have all surface types equally repre-sented. The results are shown in Table 2. The roughness metric iscalculated using the C(q) values at a wavelength with the highestdirect correlation to the friction results, similar to what has beenreported in [13]. The surface and friction metrics were normalizedby dividing each value with the average of all values. The rubberblock temperature was continuously measured using a tempera-ture sensor installed inside the rubber block. The temperaturevalues were averaged over the duration of each test.

A multiple linear regression model to predict the friction valueswas generated using the surface roughness metric and the rubberblock temperature as inputs. The resulting model provided anR2 level of 0.73, indicating a clear correlation between the inputvariables and the friction levels. Both the temperature and thesurface roughness metric showed an increase in friction withincreasing values. The p-value of the surface roughness metricwas 0.03, indicating a significant effect from the surface roughnessat a confidence level of 0.05, while the p-value of 0.43 of thetemperature parameter did not. However including the tempera-ture in the model did increase the adjusted R2 of the fit, suggestingthat it does play a role despite the small temperature rangesobserved.

4. Discussion

The maximum number of wear test sweeps used in this studywas set to 1000 based on the target application. The number ofpolishing cycles used for example in studies with the Wehner–Schultze machine are typically of the order of 105–106 polishingcycles, such as in [2,14–16]. However, the polishing cycles in theWehner–Schultze studies typically refer to the number of passeswith a rolling rubber cone in the presence of water and abrasiveparticles. In the study reported here the metric for the wear cyclesis the number of dry rubber sliding friction test, which is expectedto be more severe, especially in terms of raveling. Due to thisdifference, the wear cycle numbers are not directly comparable. Itis of course likely that if the number of wear cycles is significantlyincreased, the epoxy and cement bound surfaces would eventuallyalso start to polish or deteriorate.

A potential factor in the early deterioration of the bitumenbound samples is the sandblasting. The interlocking between thelarger stones on the surface is reduced with the removal ofbitumen and smaller sand particles. Epoxy creates stronger bondsbetween the stones and is therefore able to withstand the stressesexerted on it during the testing much better and perhaps retainmore sand particles during the sandblasting. If bitumen boundsurfaces would be pursued in the future for this application thesandblasting time could be reduced or instead of sandblasting thebitumen could be subjected to accelerated aging, perhaps by usingelevated temperatures, oxygen or UV light. One such agingmachine is presented, for example, in [17].

5. Conclusions

The aim of the study was to develop a road surface sample forlaboratory friction testing that on the one hand replicates the

10−1 100 101 10210−7

10−6

10−5

10−4

10−3

10−2

10−1

100

Spatial frequency q [rad/mm]

C(q

)/RM

S [m

m3 ]

Epoxy wear lane initial95% CI upper limit95% CI lower limitEpoxy wear lane final95% CI upper limit95% CI lower limit

Fig. 9. Normalized C(q) functions of the epoxy wear lane at the initial state and atthe end of the study. No significant change in the roughness power spectrum canbe observed.

10−1 100 101 10210−8

10−7

10−6

10−5

10−4

10−3

10−2

10−1

100

Spatial frequency q [rad/mm]

C(q

)/RM

S [m

m3 ]

Concrete wear lane initial95% CI upper limit95% CI lower limitConcrete wear lane final95% CI upper limit95% CI lower limit

Fig. 10. Normalized C(q) functions of the concrete wear lane at the initial state andat the end of the study. No significant change in the roughness power spectrum canbe observed.

Table 2Friction and surface metric results for the initial state of all studied test lanes.

Binder Lanetype

μnormalized

[dimensionless]Trubber[1C]

Normalizedsurfaceroughnessmetric[dimensionless]

Bitumen Reference 1.028 25.2 1.17Bitumen Wear 1.008 25.0 1.13Bitumen polishedstones

Reference 1.040 26.8 1.15

Bitumen polishedstones

Wear 1.031 26.8 1.07

Concrete Reference 0.931 25.5 0.47Concrete Wear 0.956 25.3 0.49Epoxy Reference 1.006 26.2 1.13Epoxy Wear 1.000 26.1 1.39

A. Kuosmanen et al. / Wear 321 (2014) 38–4544

surface topography of a given asphalt road surface, and on theother hand is durable enough to withstand repeated shear stressesexerted by laboratory rubber friction tests.

Out of three experimented binder materials, only the bitumenand epoxy samples matched the target surface in terms of rough-ness power spectrum to a reasonable extent while the concretesample had an overall too low roughness.

As for the durability, the samples bound by bitumen wereremoved from the friction tests at an early stage due to detach-ment of stones. The concrete and epoxy samples endured well,with no significant changes in friction or surface roughnessthroughout the test program.

The only sample that was both able to replicate the targetsurface roughness and endured the repeated friction tests was theepoxy sample.

When comparing the different test lanes with each other, areasonable correlation could be found between the surface rough-ness power spectrums and the dry laboratory rubber frictionresults.

Acknowledgments

The authors would like to thank Goodyear Tire & Rubber Companyfor the permission to publish these results, Project Manager BrunoMaitre for sponsoring the project, and Thomas Müsch and OlafTheissen for support with the laboratory friction tests.

This project was supported by the European Community underthe Marie Curie Industry–Academia Partnership and Pathways(PIAP-GA-2009-251606), as well as the Fonds National de laRecherche Luxembourg under AFR Grant number 1368189. Thesecontributions are greatly appreciated.

References

[1] M. Kane, D. Zhao, M.-T. Do, E. Chailleux, F. de Larrard, Exploring the ageing effect ofbinder on skid resistance evolution of asphalt pavement, Road Mater. PavementDes. 11 (2010) 543–557. http://dx.doi.org/10.1080/14680629.2010.9690346.

[2] M.-T. Do, Z. Tang, M. Kane, F. de Larrard, Pavement polishing—development ofa dedicated laboratory test and its correlation with road results, Wear 263(2007) 36–42. http://dx.doi.org/10.1016/j.wear.2006.12.086.

[3] M.Bürckert, F.Gauterin, H.-J. Unrau, Untersuchung des Einflusses der Grobtexturauf Messergebnisse mit dem SKM‐Messverfahren, Berichte der Bundesanstalt fürStraßenwesen - Straßenbau Heft S 78, (2012) 28–29. ISBN 978-3-86918-269-8.

[4] B.N.J. Persson, Theory of rubber friction and contact mechanics, J. Chem. Phys.115 (2001) 3840. http://dx.doi.org/10.1063/1.1388626.

[5] B.N.J. Persson, O. Albohr, U. Tartaglino, A.I. Volokitin, E. Tosatti, On the natureof surface roughness with application to contact mechanics, sealing, rubberfriction and adhesion, J. Phys.: Condens. Matter. 17 (2005) R1–R62. http://dx.doi.org/10.1088/0953-8984/17/1/R01.

[6] B.N.J. Persson, Rubber friction: role of the flash temperature, J. Phys.: Condens.Matter 18 (2006) 7789–7823. http://dx.doi.org/10.1088/0953-8984/18/32/025.

[7] X. Lu, U. Isacsson, Effect of ageing on bitumen chemistry and rheology, Constr.Build. Mater. 16 (2002) 15–22. http://dx.doi.org/10.1016/S0950-0618(01)00033-2.

[8] A. Torppa, M. Räisänen, Kiviainesten ominaisuuksien vaikutus hiljaistenpäällysteiden kulumiseen, in: H. Tervahattu (Ed.), Reduction of rolling noise,Final report (VIEME), Ministry of Transport and Communication, Helsinki,2008, p. 61, ISBN 978-952-201-682-9.

[9] E. Erichsen, A. Ulvik, K. Wolden, P.-R. Neeb, Aggregates in Norway—propertiesdefining the quality of sand, gravel and hard rock for use as aggregate forbuilding purposes, Geol. Soc. – Geolo. Surv. Nor. Spec. Publ. 11 (2008) 37–46.

[10] The Finnish Asphalt Specifications (in Finnish), Päällystealan neuvottelukuntaPANK, Vantaa, 2011, ISBN 978-952-99985-1-7, 2011.

[11] P.R. Nayak, Random process model of rough surfaces, J. Lubr. Technol.93 (1971) 398–407. http://dx.doi.org/10.1115/1.3451608.

[12] M. Klüppel, G. Heinrich, Rubber friction on self-affine road tracks, RubberChem. Technol. 73 (2000) 578–606. http://dx.doi.org/10.5254/1.3547607.

[13] L. Hartikainen, F. Petry, S. Westermann, Frequency-wise correlation of thepower spectral density of asphalt surface roughness and tire wet friction,Wear 317 (2014) 111–119. http://dx.doi.org/10.1016/j.wear.2014.05.017.

[14] B. Allen, P. Phillips, D. Woodward, A. Woodside, Prediction of UK surfacing skidresistance using Wehner Schulze and PSV, in: Proceedings of the InternationalConference Managing Road and Runway Surfaces to Improve Safety, Chelten-ham, England, 2008.

[15] S. Friel, M. Kane, D. Woodward, Use of Wehner Schulze to predict skidresistance of irish surfacing materials, Airf. Highw. Pavement (2013) 817–828.http://dx.doi.org/10.1061/9780784413005.067.

[16] M. Kane, I. Artamendi, T. Scarpas, Long-term skid resistance of asphaltsurfacings: correlation between Wehner–Schulze friction values and themineralogical composition of the aggregates, Wear 303 (2013) 235–243.http://dx.doi.org/10.1016/j.wear.2013.03.022.

[17] M. Kane, D. Zhao, E. Chailleux, F. de Larrard, M.-T. Do, Development of anaccelerated pavement test reproducing the effect of natural ageing on skidresistance, Road Mater. Pavement Des. 14 (2013) 126–140. http://dx.doi.org/10.1080/14680629.2012.749804.

A. Kuosmanen et al. / Wear 321 (2014) 38–45 45