E ﬀ ect of Mixture Design Parameters of Stone Mastic Asphalt Pavement on Its Skid Resistance

Featured Application: This paper seeks to analyze the e ﬀ ect of gradation and asphalt content on the skid resistance of stone mastic asphalt pavement. Abstract: Although it is well known that the stone mastic asphalt (SMA) pavement has good skid resistance, the skid resistance is not satisfactory and its durability is poor when the mixture design is unreasonable. In order to obtain excellent skid resistance for SMA pavement, this paper seeks to analyze the inﬂuence of mixture design parameters on the skid resistance of SMA pavement. The mixtures were designed with an orthogonal experiment. There were ﬁve factors, namely the percentage of aggregates passing the maximum size (P NMSA ), the percentage of aggregates passing the sieve size which is only one smaller than the maximum size (P NMSA-1 ), the percentage of aggregates passing the control sieve size (P CS ), the percentage of aggregates passing the sieve size which is only one smaller than the control sieve size (P CS-1 ), and asphalt content (AC), and each factor had four levels. The skid-resistance index (SI) obtained by the 3D measurement was used to evaluate skid resistance. The results show that the three parameters (P NMSA , P CS , and AC) are the key parameters to improve skid resistance. Among them, P NMSA has the greatest impact on the skid resistance, AC is the second, and the impact of P CS on skid resistance is the smallest. Moreover, the design parameters with best skid resistance are proposed.


Introduction
Skid resistance refers to the friction force that prevents the tire from slipping along the pavement surface [1]. It had been shown that the traffic accident risk increased significantly when the skid resistance was below a certain threshold value [2]. Therefore, skid resistance of the pavement is an important design parameter affecting driving safety.
Skid resistance/friction force consists of adhesion and hysteresis, mainly depending on the surface texture of the pavement [3][4][5][6]. The surface texture is classified according to wavelengths, including microtexture (0-0.5 mm), macrotexture (0.5-50 mm), megatexture (50-500 mm), and unevenness (500 mm to 50 m) [7]. The skid resistance of the pavement mainly depends on its microtexture and macrotexture. It is generally considered that the microtexture characterized by the friction coefficient plays a major role in skid resistance at low speeds; that the macrotexture characterized by mean texture depth (MTD), mean profile depth (MPD) or sensor measured texture depth (SMTD) has a great influence on the skid resistance at high speeds [8].
There are still no effective methods to accurately characterize microtexture until now. A number of studies used the British pendulum tester (BPN) to characterize microtexture and evaluate skid resistance at low speeds [9,10]. Liu et al. found that the friction coefficient measured by BPN was significantly affected by macrotexture. Hence, the general view that skid resistance of the pavement mainly depends on its microtexture is not always right [11]. Additionally, a dynamic friction tester (DFT) and a grip tester were also used to evaluate skid resistance at low speeds [12,13].
As mentioned previously, a number of parameters were applied to characterize the macrotexture of the pavement surface. These parameters could be obtained by different methods. Flintsch et al. used the circular track meter (CTM), laser inertial road profiler, and a sand patch to measure the pavement texture, and used MPD and MTD to characterize the macrotexture of stone mastic asphalt (SMA) and open-graded friction course (OGFC). Then, the correlation of parameters obtained by different test methods was established [14]. The stationary laser profilometer (SLP) was applied to obtain the texture elevation of the pavement surface and the texture spectral characteristics by digital signal processing. Simultaneously, the formulas of texture indicators with different wavelength ranges were proposed [15]. The texture elevation of the pavement surface could be also obtained by laser texture scanners (LTS), and geometrical indicators were used to characterize macrotexture, including MPD, average roughness, leveling depth, mean depth, surface roughness depth, and peak to valley height [16]. However, CTM, SLP, and LTS are quite expensive.
In light of expensive devices measuring pavement texture, three dimensional (3D) measurements by stereophotogrammetry or microscopy were considered to characterize macrotexture of pavement surface [17][18][19][20]. Compared with using two-dimensional (2D) line profiles, measurements in 3D can accurately characterize the roughness of the pavement surface. Firstly, a large amount of data from multiple images can be collected; secondly, 3D models of the pavement surface were established by some software, such as Digital Surf MountainsMap, 3D flow Zephyr, etc; finally, roughness parameters (arithmetic mean skewness, kurtosis, height, etc) were calculated. These parameters could be used to evaluate microtexture, macrotexture, and megatexture.
Asphalt pavement consists of asphalt and aggregates with different sizes [21][22][23][24][25][26][27][28]. Its texture is affected by many factors, such as aggregates type, aggregates sizes, mixture type, etc. Previous studies focused on how to characterize pavement surface texture and the influence of different factors (coarse aggregates type, mixture type, etc) on skid resistance. However, few studies were carried out to analyze the influence of mixture design parameters on skid resistance of pavement with constant coarse aggregates and mixture type. For a certain project, material source and mixture type are constant. Therefore, the skid resistance of pavement mainly depends on design parameters, such as gradation, asphalt content, etc. Although it is well known that the SMA pavement has good skid resistance, the skid resistance is not satisfactory, or its durability is poor when the mixture design is unreasonable [20].
This paper seeks to analyze the influence of mixture design parameters on the skid resistance of SMA pavement. The mixtures were designed with an orthogonal experiment. There were five factors, namely the percentage of aggregates passing the maximum size (P NMSA ), the percentage of aggregates passing the sieve size which is only one smaller than the maximum size (P NMSA-1 ), the percentage of aggregates passing the control sieve size (P CS ), the percentage of aggregates passing the sieve size which is only one smaller than the control sieve size (P CS-1 ), and asphalt content (AC), and each factor had four levels. The skid-resistance index (SI) obtained by the 3D measurement was used to evaluate skid resistance.

Asphalt
The asphalt selected in this paper was Styrene-Butadiene-Styrene (SBS) modified asphalt. The modified asphalt was mixed with virgin asphalt (SK90#) and SBS (Beijing Yan-shan-chan-dao-gai 2#) at a ratio of 3:100 (mass ratio). Its properties were tested according to Chinese specification JTG E20-2011 [29], and the results are shown in Table 1. Moreover, the technical requirements must meet Chinese standard JTG F40-2004) [30]. Aggregates selected in this paper were gneiss. There were four stockpiles with different sizes, namely 10-15 mm, 5-10 mm, 3-5 mm, and 0-3 mm. The first three stockpiles were coarse aggregates and the latter one was fine aggregates. The mineral filler was ground limestone. Their properties were tested according to Chinese specification JTG E42-2005 [31]. Tables 2-4 show the physical properties of coarse aggregate, fine aggregate, and mineral filler, respectively.
For SMA pavement, the percent air void in coarse aggregates (VCA) must be verified. There is a controlled sieve size (P CS ) in this process. According to Chinese standard JTG F40-2004, P CS with both SMA-16 and SMA-13 are 4.75 mm, P CS with SMA-10 and SMA-5 are 2.36 mm and 1.18 mm respectively. P NMSA and P CS were key factors for SMA gradation design [32]. Moreover, surface texture is affected by AC [8]. Therefore, P NMSA , P NMSA-1 , P CS , P CS-1 , and AC were selected for mixture design factors. Each factor had four levels.
The mixture design is multi-factorial and multi-leveled. When there are more than three factors, there will be many experiments, thereby making our objectives hard to achieve. Thus, the orthogonal experimental design (OED) was chosen. OED selects some representative points from the comprehensive experiments based on the orthogonality. These representative points have the characteristics of "uniform and dispersion, neat and comparable". Moreover, OED is the main method of the factorial design, and is a highly efficient, fast, and economical experimental design method. In this paper, OED was applied to analyze the effect of mixture design parameters on the skid resistance of SMA pavement, and orthogonal design table L 16 (4 5 ) was selected. The value of four parameters (P NMSA , P NMSA-1 , P CS , and P CS-1 ) associated with gradations was determined according to gradations ranges in Chinese standard JTG F40-2004. The values of AC were selected according to OAC. Tables 5-8 show the factors and levels of different SMA types. Table 9 is the orthogonal experimental design table with five factors and four levels for each factor. It was assumed that there is no interaction between any two factors. SI was used to evaluate skid resistance. The test values of each factor at the constant level i were summed, then the mean values (k i ) and range (R) at corresponding levels could be calculated by the following formulas.
where k i represents mean values of SI for each factor at a certain level i (i = 1, 2, 3, 4). The skid resistance is better with higher k i . For each factor, k max is maximal among four levels, while k min is the minimum. R reflects the effect order of different factors. The factor has stronger impact on skid resistance with higher R.

Pavement Surface Texture Characterization
The 3D evaluation parameters of rock surface roughness are various, including arithmetic mean deviation, maximum height, contour root mean square deviation, contour maximum peak height, contour maximum valley depth, contour support area, curved surface slope, etc., up to 17 evaluation parameters [33]. For the 3D geometric features characterization of a rough surface, it is often necessary to select several representative parameters for evaluation, which may result in a comparative one-sided evaluation, and not a comprehensive or accurate characterization of the rough surface.
Belem et al. proposed five typical 3D evaluation parameters to quantitatively characterize the rock joint features, including mean of elementary inclination angle over the whole surface (θ s ), root-mean-square of the joint surface gradient (Z 2s ), apparent anisotropy degree (K a ), surface roughness coefficient (R s ), and surface tortuosity coefficient (T s ) [34]. These parameters were not affected by the direction of the axis, which could accurately and stably reflect the geometric characteristics of the 3D rock joint surface. In addition, the sensitivity of the five parameters to the changes in surface geometry was different [35]. Therefore, the parameters had convenient applicability.
This paper selected the five parameters to characterize pavement texture. SI was used to evaluation skid resistance, and its calculation method was according to previous research [20]. Test specimens were slabs with a geometric size of 300 mm (length) × 300 mm (width) × 50 mm (height) produced according to Chinese specification JTG E20-2011. There were three specimens for SMA pavement with the same mixture type and gradation. Due to the large amount of data acquired by the images, it had a significant impact on the calculation speed and efficiency. There was a significant difference in the aggregate size, so the calculated area was selected according to the mixture types. The analysis area of SMA-16 and SMA-13 was about 100 mm (length) × 100 mm (width), while for SMA-10 and SMA-5, the analysis area was 55 mm(length) × 55 mm(width). Each specimen was tested three times in different places.

Establishing 3D Models of Pavement Texture
The pavement texture images were collected by a digital camera, and then the images were subjected to a series of processes, such as spraying a developer, filtering, etc. The processed photos were imported into a computer, and the 3D models of pavement texture were established. Figure 1 shows the 3D models of pavement texture.

184
SI was used to evaluate skid resistance. As mentioned before, ki represents mean values of SI 188

Calculation of SI
The five parameters (θ s , Z 2s , K a , R s , T s ) could be obtained based on the 3D models of pavement texture. The value of SI represents the skid resistance. The principal components analysis (PCA) was used to analyze the relationship between these five parameters and SI [20]. Then, SI could be calculated by the following formulas. The skid resistance is better with higher SI.

Results
SI was used to evaluate skid resistance. As mentioned before, k i represents mean values of SI for each factor at a certain level i (i = 1, 2, 3, 4), and R reflects the effect order of different factors. The above three parameters could be calculated by the corresponding formula. Tables 10-13 show the  calculation results.  Tables 10-13 show the skid resistance of SMA-16, SMA-13, SMA-10, and SMA-5 respectively. The following conclusions can be drawn from these tables.

i.
For different SMA types, the effect order of the three factors (P NMSA , P CS , and AC) is constant. Among them, P NMSA has the greatest impact on the skid resistance, AC is the second, and the impact of P CS on skid resistance is the smallest. Moreover, the three parameters have higher impact on skid resistance than the other two parameters in general. It is obvious that the three parameters are the key parameters to improve skid resistance. ii.
The effect order of P NMSA-1 and P CS-1 with different SMA types is various. For SMA-16, P NMSA-1 has a greater impact on skid resistance than P CS-1 , but the conclusion is the opposite for SMA-13. iii.
For P NMSA-1 , SMA pavements with different mixture types have the best skid resistance at the level (i = 1). For other design parameters, the corresponding levels (i) with different mixture types are significantly different when the skid resistance is best. iv.
Taking SMA-13 as an example, AC (5.6%) is lower than OAC (6.3%) when the skid resistance is best. It is possible that the surface texture of SMA pavement with lower AC is richer, resulting in better skid resistance. It is noticed that the pavement may have excellent skid resistance but poor pavement performance when AC is 5.6%. Therefore, the determination of the best design parameters should be considered both skid resistance and pavement performance. Note: Optimum group in Tables 10-13 refers the optimal design parameters when the skid resistance of SMA pavements for different mixture types is best.
The influence of each factor on the skid resistance of SMA pavement could be intuitively determined by the trend of k i with different levels. For each factor, the value increased with levels. Figure 2 shows the relationship between P NMAS and the skid resistance of SMA pavement. It can be seen that the skid resistance gradually reduces with the increasing of P NMAS . P NMAS refers to percentage of aggregates passing of the maximum size. Lower P NMAS means more aggregates remaining on the sieve. The macrotexture of asphalt pavement is mainly contributed by the coarse aggregate. The content of aggregates with the size larger than nominal maximum size leads to richer macrotexture, resulting in better skid resistance.

The Influence of NMAS on Skid Resistance
best. It is possible that the surface texture of SMA pavement with lower AC is richer, resulting 208 in better skid resistance. It is noticed that the pavement may have excellent skid resistance but 209 poor pavement performance when AC is 5.6%. Therefore, the determination of the best design 210 parameters should be considered both skid resistance and pavement performance.

211
The influence of each factor on the skid resistance of SMA pavement could be intuitively 212 determined by the trend of ki with different levels. For each factor, the value increased with levels.    Figure 3 shows the relationship between the percentage passing several control sieves and skid resistance of SMA pavement. With increasing P CS , the trend of k i with different mixture types stays the same, which first decreases and then increases. It means that the skid resistance of SMA pavement first decreases and then increases with increasing P CS . For P NMAS-1 and P CS-1 , k i with different mixture types and levels of factors is various. However, the value of k i does not change much, indicating that P NMAS-1 and P CS-1 have an insignificant effect on skid resistance of SMA pavement. resistance of SMA pavement. With increasing PCS, the trend of ki with different mixture types stays 223 the same, which first decreases and then increases. It means that the skid resistance of SMA

The Influence of AC on Skid Resistance
234 Figure 4 shows the relationship between AC and the skid resistance of SMA pavement.

235
It can be seen that the trend of ki is consistent for different mixture types in general with 236 increasing AC. ki gradually decreases as AC increases, indicating that the skid resistance decreases.

237
This is mainly due to the fact that when AC increases, the asphalt film is thicker, resulting in a 238 smaller macrotexture, which leads to weaker skid resistance. It is noticed that the trend of ki of   Figure 4 shows the relationship between AC and the skid resistance of SMA pavement. It can be seen that the trend of k i is consistent for different mixture types in general with increasing AC. k i gradually decreases as AC increases, indicating that the skid resistance decreases. This is mainly due to the fact that when AC increases, the asphalt film is thicker, resulting in a smaller macrotexture, which leads to weaker skid resistance. It is noticed that the trend of k i of SMA-5 is significantly different from the other SMA types. k i first increases and then decreases with increasing AC. When the aggregates sizes are small enough, the compaction work during the specimen process may lead to rearrangement of aggregates, resulting in an abnormality in the gradation. It may have an impact on skid resistance.

1.
The three parameters (P NMSA , P CS , and AC) are the key parameters to improve skid resistance. Among them, P NMSA may have the greatest impact on the skid resistance, AC is the second, and the impact of P CS on skid resistance is the smallest. Moreover, these three parameters have higher impact on skid resistance than the other two parameters (P NMSA-1 , P CS-1 ) in general.

2.
The skid resistance of SMA pavement decreases gradually with the increasing P NMSA and AC; the skid resistance of SMA pavement first decrease and then increase as P CS increases.

3.
For P NMAS-1 and P CS-1, the skid resistance of SMA pavement with different mixture types and levels of factors varies. However, there is no obvious difference between them, indicating that P NMAS-1 and P CS-1 have insignificant effect on skid resistance of SMA pavement. 4.
When the aggregates sizes are small enough, experimental conditions may have an impact on the skid resistance. 5.
The best design parameters are proposed, but the results only consider skid resistance. In order to determine optimal design parameters, both skid resistance and pavement performance must be considered.