Negative texture, positive for the environment: effects of horizontal grinding of asphalt pavements

A pavement surface having deﬂections from a plane mostly directed downwards in valleys is said to have a “negative texture”, in contrast to a “positive texture” dominated by peaks. Negative textures are typical of porous asphalt pavements, but another way to achieve this feature is to grind oﬀ the peaks of the surface. This paper explores the eﬀects of grinding oﬀ texture peaks in the horizontal plane on a number of Swedish asphalt pavements in order to reduce noise and rolling resistance. Noise measurements were made to evaluate the ground-oﬀ surfaces versus the original surfaces, and, in most cases, also rolling resistance, texture and friction were also evaluated. It was found that grinding led to a more negative texture, tyre/road noise reductions up to 3 dB and tyre/road rolling resistance reductions up to 15%. It is concluded that horizontal grinding provides a maintenance operation with a signiﬁcant potential for reduction of noise and rolling resistance, without sacriﬁcing friction, though with limited longevity.


Introduction: tyre/road noise-reducing options
Traffic is one of the main noise sources in urban areas, and exposure to increased noise levels has an impact on human health as it is positively correlated to an increased risk of ischaemic heart diseases, high blood pressure, and cognitive impairment in children and sleep disturbance, among other health problems (World Health Organization, 2011).To reduce tyre/road noise, three possible approaches are feasible (Sandberg & Ejsmont, 2002): (i) optimise the pavement surface texture, (ii) provide high porosity in the top layer(s) of the pavement, and (iii) have a top layer with low stiffness.The first approach aims at reducing tyre deflections, thereby reducing tyre vibrations that emit noise while providing improved air drainage in the surface, which reduces the "air pumping" noise mechanism (Sandberg & Ejsmont, 2002).In the second approach, increased porosity in the top layer(s) will partly absorb the sound waves, while simultaneously improving air drainage which reduces the "air pumping" mechanism.In both cases, also, the amplification of the noise emission near the tyre/road contact edges by the so-called "acoustic horn" effect is reduced.Finally, the usage of a low stiffness modulus in the top layer is an approach that leads to reduced tyre deflection in the contact patch and thus reduced vibration.The approaches may also be combined to maximise the effects, which has actually been tried by applying a so-called poroelastic road surface (Goubert & Sandberg, 2016).
The first approach has the advantage of also reducing the rolling resistance of tyres rolling over the pavement (Sandberg, Bergiers, Ejsmont, Goubert, & Zöller, 2012), thereby reducing the energy consumption of the vehicles, which in turn reduces the emission of CO 2 .
In order to select which approach is best fitted to a given road section, the trade-offs implied by each alternative should be considered.For instance, a pavement texture optimisation should not compromise safety aspects, which means that friction levels should also be observed.
In this paper, the first approach is explored by means of horizontal grinding of the pavement surface, a technique that may also be viewed as "shaving off" the peaks in the texture.This technology was applied to a number of Swedish asphalt pavements of various designs and conditions.Unless otherwise stated, the work presented here was carried out by the Swedish National Road and Transport Research Institute (VTI) as part of a national Swedish project designated ViaFutura.Additionally, this paper briefly mentions a few grinding experiments that had already been performed.

Purpose of this paper
The objective of this paper is to describe the effects of the horizontal grinding technique applied to pavement surfaces, focusing on the environmental effects but also making sure that no safety problems are created.This was tested and demonstrated with a number of field experiments reported here.Additionally, potential problems and sacrifices occasioned by the technology are discussed.

Positive and negative textures
A pavement surface having its texture deflections from a smooth plane mostly directed downwards is considered to have a "negative texture", generally represented by its two-dimensional profile curve, which then also is called a "negative profile".Such a surface will predominantly have valleys projecting downwards from the profile centreline.The opposite is a texture having its deflections mostly directed upwards, which is considered to be a "positive texture", resulting in a "positive profile".In this case, the surface asperities predominantly are projecting upwards from the profile centreline, which generates a predominance of peaks in the contact area.
As discussed later, a negative texture results in more positive tyre/road noise and rolling resistance properties, since the local pressure distribution in the contact patch is more evenly distributed than with neutral or "positive textures"; the latter resulting in higher local pressure near the peaks.These features influence the vibration patterns in the tyre and thus noise emission and rolling resistance and, to some extent, also frictional properties.
This asymmetry in the profile can be quantified by various measures.One of them is "skewness", where a positive texture gives a positive skew ( > 0) and a negative texture gives a negative skew ( < 0).Skewness (r sk ) of a profile is defined in ISO 13473-2 (International Organization for Standardization, 2002).Equation (1) indicates how the skewness parameter is calculated, where RMS is the root mean squared deviation, L is the profile length and Z is the profile deviation from the centreline of the profile at the position x: To illustrate this concept, two artificial surface profiles were generated, one with positive skewness and the other with negative skewness.Other parameters, such as the RMS value, were kept constant.The two surface profiles are presented in Figure 1; note that the vertical scale is exaggerated.The skew of the upper profile is 1.28 and that of the lower profile is − 0.82.When rolling on a rough surface, the tyre tread rubber will not deform itself to perfectly envelop the surface.Most importantly, the rubber will not be able to establish contact with narrow valleys in the pavement texture or to follow steep gradients in the profile, due to the limited resilience of the rubber and the belt it is attached to.This has a direct implication for the noise generation mechanisms, as a different contact pattern will change the tyre's vibration excitation.One possibility to account for this effect is by using enveloping algorithms.Such algorithms estimate how the tyre tread surface deflects when contacting the pavement surface.
In Figure 1, the pavement profile curve is supplemented with such a curve, called the "enveloped profile".Different algorithms for calculating the enveloped profile can be found in the literature, as previously reported in the European project ROSANNE (Gottaut & Goubert, 2016) and further discussed in (Goubert & Sandberg, 2018).The enveloped profile in Figure 1 was generated by the authors using the procedure originally created by von Meier, as it was described by (Gottaut & Goubert, 2016).Asymmetry effects of pavement surface profiles and how existing and new measures respond to these are examined in some detail in other articles (Goubert & Sandberg, 2018;Sandberg, Goubert, & Vieira, 2018).Other authors have also considered the asymmetry effect on functional properties, including noise and friction (Izevbekhai & Voller, 2013;Liu & Shalaby, 2017;Mahboob Kanafi, Kuosmanen, Pellinen, & Tuononen, 2015).

Grinding technology and equipment
The grinding machines used in the Swedish experiments were produced by HTC Sweden AB but were of different sizes in each trial.Two such machines are shown in Figure 2.
Essentially, the machines have one, two or three grinding heads (circular rotating plates, 650-950 mm in diameter) on each of which there are 3-4 grinding discs of 170-270 mm diameter.See Figure 2 (a, d).Each grinding disc has 3-6 diamond tools mounted on the periphery.The latter grind the surface while the discs rotate, at the same time as the head rotates with its discs.The diamond tools are available with different coarseness, resulting in various levels of pavement surface microtexture.The entire machine moves forward at a speed determined by the operator.The lower the speed, the more solid matter is ground away.For example, in the experiment in 2017, using the small machine in Figure 2 (c, d), 100 m were ground at the rate of approximately 1 m per minute; in 2011 it was 2 m per minute.These machines are mainly used for grinding and polishing of industrial concrete floors.
This grinding technique is referred to as horizontal grinding as the grinding tools rotate in the horizontal plane.This shall not be confused with the much more common grinding technique, most often referred to as "diamond grinding", in which parallel "saw blades" rotate in the vertical plane (Gardziejczyk & Gierasimiuk, 2016;Li, Harris, & Wells, 2016;Sandberg & Mioduszewski, 2015).While both machine types use diamond tools, the horizontal type was used in this application to reduce macrotexture by "shaving off" the top of the texture, while the diamond grinding machines aim at reducing unevenness while simultaneously creating a longitudinal texture of narrow grooves.The latter is exclusively used on cement concrete pavements, while our horizontal grinding has been used only on asphalt pavements (but can equally well be used on concrete).
Each grinding operation was followed by careful vacuuming of the surface, followed by manual sweeping.Traffic blew away any remaining dust.

First trials
The method of creating a more negatively skewed surface for asphalt pavements by grinding off the top of the asperities had been explored previously.An early pilot trial to create a "negative texture" for noise reduction was carried out in 2006 by Sandberg and Ejsmont (Sandberg, 2007) in cooperation with the company manufacturing the grinding machines, by grinding off the top 1 mm of a dense asphalt concrete pavement with maximum 11 mm aggregate (DAC11) on a 40 m long strip in an industry yard.Tests on a trafficked road would have required permission and closure of the road while grinding; thus, in this case, the industry yard pavement was accepted.
The effect on tyre/road noise was measured at 50 km/h, using the CPX (Close Proximity) method defined in ISO 11819-2 (International Organization for Standardization, 2017b).The result was a reduction of 1 dB (A-weighted overall CPX level) when using the SRTT reference tyre, but 0 dB for a tyre used as a proxy for truck tyres (these two tyres were later defined as P1 and H1 in ISO/TS 11819-3).This was discouraging, but it was recognised that the pavement was not suitable for this experiment since it was already very smooth-textured.
To apply the technique to a more suitable pavement, in 2011, Sandberg and Mioduszewski used horizontal grinding on a double layer porous asphalt concrete with maximum 11 mm aggregate (DPAC11), hoping to achieve a substantial noise reduction (Sandberg & Mioduszewski, 2012).This experiment was conducted before the ViaFutura project was started, but as the methodology and results are consistent with those in this project, this experiment is treated below together with the ViaFutura experiments.

International experience
In Australia, inspired by the Swedish results, in 2013 a project started on a highway south of Melbourne in which several types of pavements are tested annually for noise-influencing properties.A few of these were of the type "standard OGA" (Open-Graded Asphalt), having 10 mm maximum aggregate size and being 30 mm thick with a void content of 15%.One of these sections (#2) was treated by "shaving" the slow lane of the section, i.e. grinding off 1-2 mm of the top surface of the asphalt, using a 600 mm rotary diamond cutter (Buret, Mcintosh, & Simpson, 2014) normally applied for removing line markings.Tests were conducted during the first year following the installation of the surfaces and were part of a programme intended to take place over a total of five years.CPX tests (Buret et al., 2014;Buret, Mcintosh, & Simpson, 2017) have so far been conducted on eight occasions between March 2013 and March 2017.The results have been very positive, with the ground single-layer OGA being the quietest of the five tested pavements (including two non-treated double OGAs with 10 mm max.aggregate size).Noise reduction versus the non-treated OGA has been 2-3 dB(A) with no deterioration over the four years and the ground OGA is even better than the double-layer OGAs with no treatment (Buret et al., 2017).
Additionally, skid resistance measurements were carried out, using the VicRoads SCRIM device.These were performed three months after laying.The Sideways Force Coefficient was then found to be 0.85 for the untreated OGA and 0.95 for the treated (ground) OGA.The doublelayer OGAs had an SFC of 0.8 (Buret et al., 2014).In other words, the grinding was favourable for skid resistance.
Grinding operations of the kind studied here have been and are periodically done in the Netherlands, on a large scale and on 2.5 m width road lanes.According to personal communication with Mr Loes van Gompel, a representative of the grinding company, the purpose is to increase the skid resistance of the road surfaces, which can be done before rut depths reach 6 mm (L.van Gompel, personal communication, July 16, 2018).The grinding tools and the technology are the same as used in the Swedish project reported here, yet the application of the tools over a 2.5 m grinding width on a truck is a Dutch construction.Quantified results have not yet been reported.

Pavement surface geometrical properties
To analyse how the texture and unevenness were affected by the grinding process, all the tested sections and their reference pavements (generally the same pavement without grinding) were measured by VTI using a laser RST vehicle (Arnberg et al., 1991) or similar equipment.The macrotexture, which has a texture wavelength in the range of 0.5-50 mm, was collected in accordance with ISO 13473-1 (International Organization for Standardization, 1997) and megatexture (wavelengths 50-500 mm) in accordance with ISO 13473-5 (International Organization for Standardization, 2009), while unevenness (wavelengths above 0.5 m) and rut depth were collected in accordance with CEN standards (European Committee for Standardization, 2008).Macrotexture was also characterised using skewness in accordance with ISO 13473-2 (International Organization for Standardization, 2002).

Tyre/road noise
The Close Proximity (CPX) method was employed in these experiments to analyse the noise resulting from the tyre/road interaction.The tests were carried out in accordance with ISO 11819-2 (International Organization for Standardization, 2017b) using the Tiresonic Mk4 test trailer.This trailer was designed and operated by the Gdańsk University of Technology (TUG) and is presented in Figure 3a.Two test speeds were used, namely 50 and 80 km/h, in accordance with ISO 11819-2 (International Organization for Standardization, 2017b).As reference tyres, both tyre P1, which acoustically represents light vehicles, and tyre H1, which represents heavy vehicle tyres, were used in accordance with ISO/TS 11819-3 (International Organization for Standardization, 2017a).
To allow for temperature normalisation, both air and road temperatures were measured simultaneously with the noise measurement.The normalisation for air temperature followed the procedure required in ISO/TS 13471-1 (International Organization for Standardization, 2017c) with a reference temperature of 20°C.The CPX levels were also normalised for tyre hardness and velocity variations according to ISO 11819-2 and ISO/TS 11819-3 (International Organization for Standardization, 2017b).

Rolling resistance
Measurements of the rolling resistance coefficient were made by a special trailer and crew from TUG: the so-called "R2 trailer" (see Figure 3d).The method is described in (Bergiers, 2016).As reference tyres, the P1 and H1 tyres defined in ISO/TS 11819-3 were used, plus a Michelin Primacy HP tyre (225/60R16), designated as tyre M1 in this text, was used as a second car tyre.

Skid resistance
Skid resistance was measured by the Saab Friction Tester.This is a test of longitudinal friction at optimum slip on a wetted surface (nominally 0.5 mm water depth) by a car equipped with a

Tested pavements
This paper reports trials made at four places, with the most essential descriptions summarised in Table 1.The type of pavement that was ground and its age are indicated, together with a rough estimation of how much of the peak texture was ground off.Note that 2011-GR1 was tested also in 2015, and 2013-GR1 and 2013-GR2 were tested also in 2014.
Each ground section was also associated with a reference section consisting of the same pavement which was not ground.These were located immediately before and after the ground section, and a minimum of 100 m of each part was measured, i.e. at least 200 m in total.An exception was 2013-GR3, which had a 100 m reference section in L1 in the opposite direction.
One illustration of a trial section is shown in Figure 4a.An illustration of a ground surface beside a non-ground surface is shown in Figure 4b.
Trial section 2017-GR1 is unique as it has two references.The reason is that when making the measurements on this site, it appeared that the reference section after the ground section (REF2) was quite different from the reference before the section (REF1) in terms of porosity.While REF1 had been totally clogged over its 6-year life, REF2 had some porosity left, which affected noise and permeability.Permeability measurements made with EN12697-40 (European Committee for Standardization (CEN), 2012) method gave an average outflow time of 272 s on REF1, but only 99 s on REF2 (the ground surface gave 552 s).This means that some pores were still somewhat open on REF2, making it quieter.

Texture changes
As an illustration of the effect of grinding, for 2017-GR1, the grinding process removed 3 kg/m 2 of material from the road surface.This corresponded to an average of 3 mm removal of the  2 for all the trial sections and their references.Additionally, skewness values of a few sections are shown.

Noise
Results for the measurements with the CPX method are presented in Table 3 for each tyre, pavement and speed.The levels are normalised for air temperature (20°C), the nominal test speed (50 or 80 km/h) and tyre rubber hardness (66 Shore A) (International Organization for Standardization, 2017a).Noise reductions rather than the "absolute levels" are shown, as they are more accurate due to paired measurements made during the same runs.The noise reductions are shown for each one of the two reference tyres, P1 and H1, as well as the resulting CPX Index, calculated according to the ISO 11819-2 standard (International Organization for Standardization, 2017b).
The results are also presented as third-octave band frequency spectra (see Figures 6 and 7).

Rolling resistance
Results for the measurements of rolling resistance are presented in Table 4 for each tyre and pavement.Values for 50 and 80 km/h were averaged as they are highly correlated, and speed is not a major factor in rolling resistance (Sandberg et al., 2012).Reductions, rather than the absolute values, are shown, as they are more accurate due to paired measurements made during   the same runs.In this way, any potential drift with time or tyre temperature is eliminated, which otherwise may corrupt results.The results are also shown graphically in Figure 8.

Skid resistance
Results for the measurements of skid resistance are presented in Table 5 for each speed and pavement; the results are shown as the longitudinal wet skid resistance (SR) using the Saab Friction Tester (SFT).As this parameter was only of secondary interest, not all trial sections were tested.

Effects on pavement surface texture
The basic intention of the horizontal grinding is to create an "extra" negative profile.Considering the changes in texture values and studying close-up photos of the surface (only one example is shown here, another one in Sandberg & Mioduszewski, 2012), it appears that this was actually achieved.An excellent indication of this is given by the skewness values for test section 2017-GR1, for which the negative value was nearly doubled, despite the fact that the original pavement already had a negative skew.

Effects on noise
The resulting influences on noise levels varied with the tyre and the pavement type.Generally, the highest effects were achieved on the pavements with the highest macrotexture, such as the porous asphalt pavements and the SMA16.On SMA11, the texture was already low and noise reduction was negligible.The average spectral change for reference tyres for all tested pavements (i.e. both dense and porous) is presented in Figure 9 (P1) and Figure 10 (H1).Consistently, noise is reduced for the low-to-medium frequencies (250-1250 Hz for P1 and 250-1000 Hz for H1), which is attributed to the lower texture impact on the tyre, which is especially apparent when comparing the enveloped profile curves.On the other hand, the diagrams show a noise increase at high frequencies (1600-5000 Hz for P1 and 1250-5000 Hz for H1).Typical noise generation and amplification mechanisms associated with such high frequencies are air-pumping, stick-snap  and stick-slip (Sandberg & Ejsmont, 2002).That this has an effect on air pumping is logical, given that the texture is reduced and the drainage area between tyre rubber and pavement surface is reduced, something that will increase the air pressure gradients in the contact area.The effect seems to be present for both dense and porous pavements.Especially on porous surfaces, it may well be that some dust from the grinding got stuck in the deeper parts of the profile and thus reduced the drainage area, which was observed visually soon after the grinding, but it is not known how long the dust might have remained there.The authors have no explanation for the noise reduction at the highest frequencies (8000-10,000 Hz).Since this frequency range has no significant impact on the A-weighted overall levels   or on the perception of noise, it has not been subject to research.Although it may trigger research curiosity, it has little practical interest.The effects at low and high frequencies are opposite and it is therefore not obvious what the overall effect will be.However, from the A-weighted overall levels (Table 3) it appears that the noise increase at the high frequencies is not strong enough to balance out the noise decrease at the low-to-medium frequencies for tyre P1, but just barely and not always so for tyre H1.It is well known that truck tyres, like the H1 reference tyre, are less sensitive to texture, especially at the lower frequencies, so the result is logical, given existing knowledge (Sandberg & Ejsmont, 2002).
Nevertheless, when achieving 2-3 dB noise reduction for light vehicles, the authors consider this worthwhile, especially when it occurs "on top of" the already substantial noise reduction of porous asphalt (Sandberg & Mioduszewski, 2012).

Effects on rolling resistance
The influence of the grinding on rolling resistance appeared to be more significant than that on noise since reductions around or above 10% will have significant effects (1-2%) on fuel or energy consumption of vehicles and consequently on CO 2 emissions.As illustrated by the results of the 2015 test, too little grinding may result in only negligible effects.The two car tyres P1 and M1 give higher reductions than the truck tyre H1 as truck tyres are known to be less sensitive to macrotexture (as for noise).

Effects on wet skid resistance
Many would expect that reducing macrotexture will reduce skid resistance.Our results did not indicate this.On the contrary, small increases in skid resistance were recorded.The reason is that the grinding tools leave a very rough microtexture which for normal speeds and water depths balances out the lower drainage of ground surfaces.It was obvious when touching the surface with fingers that the microtexture became rougher after grinding as the surfaces felt harsh and potentially abrasive.Figure 11 shows changes in the friction coefficient after the grinding process.Note that in all except one case the friction coefficient increased or remained the same.When comparing the friction measurements at 50 and 80 km/h, the coefficient dropped on average 0.02 (or 3%) for the reference surfaces whilst for the ground surface the observed drop was 0.05 (6%).This indicates that the frictional properties are more substantially affected at higher speeds, which is not surprising as the grinding process decreases macrotexture and, therefore, surface drainage capability.Nevertheless, one must make sure that the resulting macrotexture after the grinding is sufficient to provide acceptable wet skid resistance also at higher speeds and deeper water depths (Sandberg & Mioduszewski, 2015).On high-speed roads, this limits the pavements suitable for horizontal grinding to those which are either porous or have an MPD value above approximately 1.2 mm.

Overall considerations
A comparison of how the different grinding experiments resulted in different noise and rolling resistance reductions is presented in Figure 12.Unfortunately, several test sections lacked a few results, which led to only four points for each test tyre.Note that texture data for the 2013 GR2 section was not available, which explains that only the reference value is presented in that figure.It appears that for tyre P1 (cars), reductions in noise and rolling resistance go hand in hand and are substantial, while for tyre H1 (heavy vehicles), noise reductions were minimal, while rolling resistance is reasonably high.
The relevance of a high MPD value on the initial surface is also seen by inspecting the correlation matrix shown in Table 6.A higher initial MPD has a potential for producing more substantial decreases in noise and rolling resistance.According to the correlations matrix, the ground surface MPD had only a minor role in the results, yet a decreased MPD after grinding contributed to more substantial reductions in noise and rolling resistance.The most significant noise reductions were obtained on the 2017 GR1 test section that had a relatively high initial MPD.The most significant   rolling resistance reduction was obtained on the 2013 GR2 test section.Unfortunately, due to a mistake in the measurements detected too late, there was no texture measurement available after grinding to allow a better interpretation of such a significant decrease in rolling resistance.Moreover, a more comprehensive texture data analysis, including skewness, would be desirable.This was not possible as the raw data (profile) for the texture measurements was only available for one of the test sections.With a significance level of 5%, the observed correlations that were considered significant were between CPX50 change and MPD REF (significance of 3.5%), and between RR change and MPD REF (significance of 3.7%).
One minor disadvantage that has been noticed is that the ground surfaces tend to look a little glossy when viewed against the sunlight, since their flatness enhances specular reflection (see Figure 4).This might affect the drivers' perception of the road condition and thus might impact drivers' behaviour.
It is recommended, in order to ensure efficient grinding and for safety reasons, that horizontal grinding should not be carried out on pavement surfaces having an MPD value below about 1.2 mm (as grinding would not have a major effect then), and that grinding should not result in an MPD below about 0.6 mm (as below such textures, wet skid resistance and hydroplaning are substantially worsened).

Combining vertical and horizontal grinding?
As already stated above, this type of grinding should not be mistaken for the pavement treatment widely known as diamond grinding.The latter also creates a highly negative texture, but often with side-effects, such as "fins" on the ridges of the grooves, that may more or less balance out the effect of the grooves.However, if the horizontal grinding is used as a post-treatment on a diamond ground pavement (grinding away "fins"), the authors think that very good effects are achievable with such a combined approach.It is recommended that this combination of grinding procedures be tested.

Durability and longevity
A crucial question is how durable are the grinding effects?The 2011 tests were repeated in 2012 (noise only, not included in the tables) and showed that the noise reduction effect was gone.Measurements in 2013-2014 also showed that the noise reducing effect was gone after one year.The rolling resistance reduction, however, was still there, albeit at a lower level.The reason for the loss of the noise reduction is thought to be that edges of the flat upper faces of the asperities gradually become rounded, thus increasing the enveloped profile amplitude, while simultaneously reducing the available drainage volume in the valleys of the profile.
With little doubt, the reason for this is that, in Sweden, studded tyres are used during the cold season and they are known to wear out pavement surfaces much faster than in countries where such tyres are prohibited.Therefore, in countries where studded tyres are not used, the authors are confident that the effects of grinding would last much longer.This is suggested, if not confirmed, by the very positive long-term results reported from Australia (Buret et al., 2017).It should be noted that the grinding procedure may not be very expensive if applied on a large scale and repeated at certain intervals, providing a reasonable cost-benefit ratio when including monetary evaluation for the society of the internal costs of the noise and rolling resistance effects.An LCA would be desirable but is not available at the moment.
When applied in Sweden or other countries or regions allowing studded tyres, it seems that horizontal grinding of asphalt pavements would be needed every second year.Whether this is economical or not is uncertain as this has not been evaluated.However, the lowering of rolling resistance, and therefore the reduced energy consumption and CO 2 emissions, would have a substantial effect in monetary terms, and may balance out the grinding costs, if grinding is carried out with large enough equipment, optimised for this purpose.
When applied in countries not allowing studded tyres, re-grinding would be needed at much longer intervals, which would be much more promising for an economically feasible maintenance procedure.

Conclusions
Horizontal grinding was applied on eight Swedish asphalt pavements to reduce noise and rolling resistance.The grinding operation created a more negative texture profile.This appeared to be acoustically favourable (up to 3 dB) but even more important, it reduced rolling resistance (up to 15%).This was achieved without sacrificing skid resistance for normal speeds and water depths.Unfortunately, in the Swedish tests, highly affected by the road wear of studded tyres in the cold season, the durability of the noise effect is low, and also the longevity of the rolling resistance reduction was poor.This was because asperities in the flat pavement surface created by the grinding were worn down by the studs to the more common rounded shape.
However, experience in Australia, where studded tyres are not allowed, suggests that the longevity of the noise reduction effect is good.Therefore, the authors believe that under normal pavement wear conditions, the horizontal grinding technology is feasible.Especially, it may be used for porous asphalt surfaces that have lost much of their initial noise reduction, where the grinding may restore noise reduction to a more favourable level.Furthermore, rough-textured pavements such as SMA16 may, after grinding, give similar, more favourable effects as the smoother SMA11 pavements.
As both noise and rolling resistance reductions are positively correlated to the initial MPD, it is recommended for efficiency reasons that horizontal grinding should not be carried out on pavement surfaces having an MPD value below about 1.2 mm.Furthermore, grinding should not result in an MPD below about 0.6 mm, as it would compromise safety.
The grinding speed is an important parameter as the ground-off material and the grinding depth are roughly proportional to the grinding speed.Only two different grinding speeds were tested here, 1 m per minute and 2 m per minute.Due to project constraints it was not possible to test different grinding speeds, and therefore it is difficult to indicate an optimal grinding speed based on the results presented here; however, this should be the subject of further studies.

Figure 1 .
Figure 1.Artificially generated textures with positive skew (a) and with negative skew (b) and their corresponding enveloped surface profiles.The red curve is the enveloped profile; see the text below.

Figure 2 .
Figure 2. Grinding machines used in this study.(a): model HTC 2500 iX, with one of its three grinding heads lifted to expose the four grinding discs.(b): machine HTC 2500 iX in operation.(c): machine HTC950RX in operation.(d): the 3 × 2 diamond cutter tool HTC950RX.

Figure 3 .
Figure 3.The measurement equipment used.(a): CPX measuring trailer by TUG.(b): reference tyre P1. (c): reference tyre H1 (P1 and H1 were used both in the noise and rolling resistance measurements).(d): rolling resistance trailer operated by TUG.(e): additional tyre used in rolling resistance measurements only (Michelin Primacy).smalltest tyre, at 50 and 80 km/h (European Committee for Standardization, 2011).It is one of the world's most commonly used friction testers on airfields, and also used on roads.

Figure 4 .
Figure 4. (a) Example of a test section immediately after grinding: 2017-GR1.(b) example of a test section (2017-GR1) not ground (left) and ground (right).The contrast in the photo is somewhat enhanced to highlight the differences.

Figure 5 .
Figure 5. Examples of pavement profiles in the REF1 section (a) and in the ground (GR1) section (b); black for the original profiles; red for the enveloped versions (see Figure 1 for explanations).

Figure 6 .
Figure 6.CPX spectra for the reference tyre P1, representing light vehicles, at speeds of 50 and 80 km/h, for ground (GR) and reference (REF) test sections.

Figure 7 .
Figure 7. CPX spectra for the reference tyre H1, representing heavy vehicles, at speeds of 50 and 80 km/h, for ground (GR) and reference (REF) test sections.

Figure 8 .
Figure 8. Reduction of the rolling resistance coefficient for each test section and for three tyres.

Figure 9 .
Figure 9. CPX spectral change for reference tyre P1, average of all test sections at (a) 50 km/h and (b) 80 km/h.

Figure 10 .
Figure 10.CPX spectral change for reference tyre H1, average of all test sections at (a) 50 km/h and (b) 80 km/h.

Figure 11 .
Figure 11.Friction coefficient before and after grinding, measured at the speeds of 50 and 80 km/h.Two of the test sections lack data.

Figure 12 .
Figure 12.Average noise and rolling resistance reduction, while simultaneously comparing MPD values (represented by the markers' diameters) for the reference and ground surfaces.The MPD values are expressed by the size of each circular marker in the figure.A smaller circle means a smaller MPD value.

Table 1 .
Summary description of the ground pavements: designations, locations, traffic volume (AADT), dimensions and ground-off texture.

Table 2 .
Macrotexture and megatexture values in mm, as well as skewness.
Notes: 2013-GR2: MPD value not available due to corrupt data.2017-GR1: this one has two ref sections; first fully clogged (cl), second with some porosity (po).Skewness available for 2017-GR1 only.

Table 3 .
Results of CPX measurements of noise properties; reductions of CPX levels in A-weighted dB (original minus ground surfaces).
Notes: 2017-GR1: this one has two ref sections; first fully clogged (cl), second with some porosity (po).2013-GR3 only tested at 50 km/h due to the speed limit.

Table 4 .
Results of measurements of the rolling resistance (RR) coefficient; presented as reductions in % from the original reference (not ground) to the ground surfaces.

Table 5 .
Friction measurements result showing the resulting friction coefficient (SR) for each surface and measurement speed [unitless].

Table 6 .
Correlation matrix, with correlation coefficients given as percentages.that only the elements above the main diagonal are shown here.CPX50 change and CPX80 change represent the change due to grinding in the measured CPX levels at 50 and 80 km/h respectively.RR change represents the rolling resistance change due to grinding.MPD REF and MPD GRD represent the MPD values for the reference and ground surfaces, respectively.MPD diff is the difference between MPD GRD and MPD REF.Values calculated for the average result of both reference tyres. Note