do not necessarily reflect the views of UKDiss.com.
Development of Prediction Models for Skid Resistance of Asphalt Pavements and Seal Coat
Abstract
Maintaining adequate level of skid resistance is essential for road safety. As part of their regular maintenance program, the departments of transportation collect skid data on roads to ensure sufficient traction between road surface and vehicle tires. This study developed models for skid resistance of asphalt pavements and seal coat surfaces. These models predict skid number over time as a function of aggregate gradation, aggregate shape properties (texture and angularity) and its resistance to abrasion and polishing, and traffic level. The researcher examined 70 test sections in this study, half of these test sections were asphalt pavements and other half was seal coat surfaces. Skid data were collected using a skid trailer while surface friction characteristics of the test sections were evaluated using a dynamic friction tester and circular texture meter. In addition, the change in aggregate shape properties due to abrasion and polishing was studied using the MicroDeval test and Aggregate Image Measurement System. The results demonstrated that the developed models provide good correlation with the skid measurements in the field. This study produced a revised version of a utility called skid analysis of asphalt pavement (SAAP) that incorporate the new models for skid resistance of both asphalt pavements and seal coat. The SAAP can be used by the pavement engineers and contractors to predict skid resistance over time.
Keywords: Skid Resistance, Asphalt Mixtures, Seal Coat, Texture, Angularity, MicroDeval, AIMS, SAAP, Abrasion, Polishing
Introduction
Skid resistance is a key component in road safety. Proving an adequate level of skid resistance reduces vehicle sliding and stopping distance, especially in wet conditions (1). The friction between pavement surface and vehicle tires is related to the macrotexture and microtexture of pavement surface. The macrotexture of asphalt pavement is dependent on aggregate gradation, while the microtexture is dependent on aggregate shape characteristics. Aggregates with angular shape and rough texture provide higher level of skid resistance compared to aggregates with smooth surface. In addition, pavement surfaces with high macrotexture provide higher skid resistance compared to those with low macrotexture (2, 3, 4). Hall et al. (5) indicated that microtexture locates the magnitude of skid resistance, while macrotexture controls the slope of the skid resistance reduction as the speed increases. It is well established that the skid resistance deceases with time due to the polishing effect of traffic on pavement surface. The polishing action affects both microtexture and macrotexture of pavement surface (6, 7).
Skid resistance has two mechanisms; adhesion and hysteresis. These two mechanisms are highly affected by pavement macrotexture and microtexture (8). The adhesion is developed due to the direct contact between the tires and pavement surface especially in areas with high local pressure (9). Pavement microtexture is significant to the adhesion component that originated from molecular bonds between stone and rubber. In addition, pavement macrotexture contributes to the hysteresis component of the friction (10). Hysteresis is developed due to energy dissipation caused by the deformation of the tire rubber around bulges and depressions in the pavement surface (9). Adhesion and microtexture affect the skid resistance at all speeds, and they have prevalent influence at speeds below 30 mph. Hysteresis and macrotexture have little significance at low speeds; however, macrotexture is an essential factor for safety in wet conditions as speed increases (11).
The seal coat or chip seal is widely used as preventive maintenance treatment and considered relatively inexpensive pavement surface treatment. It can be used effectively on roads with both high and low traffic levels (12) (TxDOT 2003). Similar to hot mix asphalt (HMA) surfaces, the macrotexture and microtexture of seal coat surface have significant contributions to the skid resistance. The macrotexture of pavement surface is affected by the aggregate size and its embedment into the binder. Immoderate embedment may reduce the skid resistance of seal coat (13, 14). In addition, aggregate polishing due to traffic reduces the skid resistance, and the rate of skid reduction depends on the aggregate shape characteristics (2, 15). The seal coat surface treatments (TxDOT Grade 3 and TxDOT Grade 4) provided higher skid resistance compared to asphalt concretesurfaced pavements (TxDOT Type C), but skid resistance of the surface treatments may decrease significantly once its macrotexture decreases (2).
There are several attempts for developing prediction models for friction and skid resistance of asphalt pavements. Masad et al. (16) developed a new method to evaluate the change in the asphalt pavement skid resistance depending on aggregate texture, properties of mixtures, and environmental conditions. This method relies on the use MicroDeval test and Aggregate Image Measurement System (AIMS) to evaluate the resistance of aggregate to polishing and abrasion. In the field, they examined nine pavement sections and they prepared laboratory test slabs using different aggregate types and different mixture types. The results showed that aggregate type has significant effect on skid resistance, while the mix type was not statistically significant.
Masad et al. (2) conducted a study that included measurements in the field and laboratory. In laboratory, several slabs with different asphalt mixtures and aggregate types were prepared and tested. A threewheel polisher was used to polish the test slabs, and the measurements of the friction and mean profile depth were collected using the dynamic friction tester (DFT) and Circular Texture Meter (CTMeter) after different polishing cycles. The results demonstrated high correlation between the aggregate properties and the mixture frictional characteristics. Based on laboratory stage, Masad et al. (2) developed a model to predict the initial, terminal, and rate of change in international friction index (IFI) as a function of aggregate characteristics obtained from AIMS system and aggregate gradation parameters. The data collected in the laboratory were compared to skid values measured in the field for the same asphalt mixtures. Masad et al. (2) proposed a system to predict the skid number of asphalt mixtures as a function of traffic level. Input parameters required for this model included aggregate texture measured using AIMS before and after polishing in MicroDeval, aggregate gradations, and traffic data.
Wu et al. (17) developed a new model to estimate the skid resistance based on 12 mixtures with various mix types and aggregate sources. The aggregates included sandstone and siliceous limestone and four mix types were evaluated. The model estimates the friction number at 60 km/hr. The researchers also demonstrated that aggregates with low skid resistance can be blended with good quality aggregates in order to achieve adequate skid resistance.
Kassem et al. (4) conducted a study to validate the IFI models developed by Masad et al. (2). Squaredshaped slabs were prepared in the laboratory using three different types of aggregates and different asphalt mixture designs were evaluated. Laboratory slabs were prepared in the laboratory and polished using a threewheel polisher. The frictional characteristics of the slabs were recorded with polishing. The results demonstrated that the coarse mixtures had better friction compared to fine mixtures. The results demonstrated high correlation between the measured and predicted IFI after considering aggregate texture and angularity indices in the developed model.
In order to improve the safety on highway pavements, the researchers proposed test methods and models to predict tirepavement friction and skid resistance as a function of aggregate characteristics, mix design and traffic level. These models need to be validated with additional data that cover a wide range of variables and parameters. In addition, these models should be extended or revised to predict the skid resistance of surface treatments such as seal coat. This study had two objectives:
 Investigate and examine surface and friction characteristics of test sections of asphalt mixtures and surfacetreated roads in Texas. The test sections covered a wide range of mixtures and aggregate types used in Texas.
 Validate and revise the skid prediction model for HMA; develop a prediction model for skid resistance of seal coat surfaces; and incorporate an improved method of traffic analysis, lane distribution of traffic data, and the effect of the percentage of truck traffic.
Selection of the Field Sections
The researchers measured the frictional characteristics and skid number on a number of HMA and seal coat test sections in Texas. The researchers identified and selected 35 test sections of HMA along with 35 test sections of seal coat. Four seal coat test sections were excluded due to excessive bleeding. During the selection of test sections, the research team made an effort to include surfaces with wide varieties of mixture gradations, aggregate sources, and climatic zones of Texas. Focus was given to identify test sections with higher traffic levels so that the team can observe higher polishing within relatively short time. Another important criterion of test sections selection was to find existing sections with history of skid measurement under TxDOT’s annual networklevel pavement evaluation program. TxDOT does not collect the networklevel skid data for all the roads every year. Typically, major highways (i.e., interstate highways) with higher traffic level are tested more frequently than other highways (i.e, farmtomarket roads). The annual skid testing frequency varies among different districts of TxDOT.
The test sections of asphalt mixtures included different mixture type (SMAC, SMAD, SMAF, CMHBF, Type C, Type D, TOM, PFC, CMHBC, and CAM), aggregate type (Limestone, Gravel, Granite, Sandstone, Dolomite, Rhyolite, Traprock, and Quartzite), year of construction (2004 to 2013), and were distributed across Texas (ATL, AUS, BMT, BRY, ODA, SAT, YKM, HOU, LRD, PHR, and LFK districts of TxDOT). Also, the test sections of seal coat included different grade type (Grade 3, Grade 4, and Grade 5), aggregates (Limestone, Gravel, Traprock, Sandstone, Dolomite, Rhyolite, LRA, and Lightweight), coating conditions (precoated and virgin), year of construction (2009 to 2013), and also were distributed across Texas (ATL, BMT, ODA, SAT, YKM, LRD, PHR, LFK, BRY). Details about the HMA and seal coat test sections are provided in a research report by Chowdhury et al. (18).
Frictional Characteristic Measurements
Measuring Microtexture and macrotexture
Field testing primarily included measurements of friction using the dynamic friction tester (DFT), mean profile depth (MPD) using the circular texture meter (CTMeter), and skid number using the TxDOT’s skid trailer. Figure 1a shows a layout of the test section used by the researchers when taking DFT and CTMeter measurements in the field (Figures 1b and 1c). The CTMeter device was used to measure the MPD, while the DFT was used to measure the coefficient of friction at different speeds (20, 40, 60, and 80 km/hr). During testing, the CTMeter and DFT devices were always positioned in the left wheel path of the outside lane. Six locations were tested in each section. Two locations were at the shoulder, and four locations were in the outer lane. Two DFT and six CTMeter readings were performed at each location. In some cases, where there was no shoulder, the researchers took CTMeter and DFT measurements between the wheel path to represent the initial skid values. The shoulder had higher friction value compared to wheel path as the later experienced frequent polishing under traffic. Measurements of macrotexture and friction were conducted on the outer lane as the skid number was measured by the skid trailer at the outside lane (in case of multiple lanes) on the left wheel path. Also, the outer lane experiences most polishing rates because most of the trucks and other vehicles use this lane.
Shoulder
Outer Lane
(a) Locations of DFT and CTMeter measurements 
(b) DFT testing 
(b) CTMeter testing 
Figure 1. Typical Field operations
The researchers calculated the international friction index (IFI) using the mean profile depth and friction obtained from the field according to Eq. 1.
IFI=0.081+0.732DFT20e40Sp
(1)
Sp=14.2+89.7MPD
(2)
Where:
MPD = mean profile depth measured using the CTMeter
DFT_{20} = coefficient of friction at 20 km/hr measured using DFT
Aggregate Characterization
The researchers obtained aggregate samples used in the construction of HMA and seal coat test sections. The researchers used the AIMS and MicroDeval devices to measure aggregate’s resistance to polishing and abrasion. The MicroDeval test is used to measure the resistance of aggregate to abrasion and it is conducted according to the American Association of State Highway and Transportation Officials (AASHTO) procedure (19). The AIMS was used to quantify the aggregate’s texture and angularity before and after polishing using the MicroDeval apparatus. Both of the texture and angularity of aggregates decrease with the time of polishing in the MicroDeval test. The loss of texture can be described using only three data points: texture measured before the MicroDeval test, after 105 min. and 180 min. of polishing in the MicroDeval test. Figure 2 shows an example of the loss of texture and angularity due to abrasion and polishing. Kassem et al. (4) provide more details about the MicroDeval and AIMS testing procedures.
Figure 2. Loss in Aggregate Texture and Angularity as a Result of MicroDeval Abrasion and Polishing of Virgin Aggregates
Mahmoud et al. (20) and Kassem et al. (4) suggested using Eqs.3 and 4 to describe change in aggregate texture and angularity as a function of polishing time in MicroDeval:
TX (t) =
aTX+ bTX* e(CTX*t)
(3)
GA (t) =
aGA+ bGA* e(CGA*t)
(4)
Where:
TX (t): change in texture as a function of time (min.)
aTX, bTX, CTX
: aggregate texture regression constants
GA (t): change in angularity as a function of time (min.)
aTX, bTX, CTX
: aggregate angularity regression constants
t: polishing time in MicroDeval
Aggregate Gradation Parameters
Masad et al. (2) have indicated that aggregate gradation is a fundamental factor that affects skid resistance. Masad et al. (2) and Kassem et al. (4) used the cumulative twoparameter Weibull distribution (Eq. 5) to describe the aggregate gradation. The Weibull distribution function was used to fit the aggregate size distribution and both scale (λ) and shape (κ) parameters were used to quantify the aggregate gradation.
F (x, λ, κ) = 1
e(xλ)κ
(5)
Where:
x: aggregate size in millimeters
λ, κ: scale and shape parameters of Weibull distribution
Skid Number Measurements
Skid number data used in this study was obtained from two sources: TxDOT’s annual networklevel data collection, and project level measurements by the researchers. Each year TxDOT periodically measures the skid number for all its highways although at different intervals for different highways. The research team obtained the data form TxDOT’s Pavement Management Information System (PMIS) database. TxDOT PMIS database typically store the skid number data for each of the PMIS section which is typically 0.5mile long. The length of test sections included in this study varied between two miles to little over 15 miles. TxDOT measures skid number using a skid trailer with a smooth tire according to ASTM E 274, “Standard Test Method for Skid Resistance of Paved Surfaces Using a FullScale Tire”. The left tire is locked to measure the skid number at 50 mph (80 km/h). The skid number was measured at the outside lane (in case of multiple lanes) at the left wheel path.
Data Analysis and Results
Researchers, in this study, developed models to predict the skid number for both HMA and seal coat surfaces. Several parameters are taken into account in developing these models. These parameters include traffic data, aggregate texture and angularity, aggregate resistance to abrasion and polishing, and aggregate gradation. Statistical methods were used to develop prediction models for friction and skid resistance.
Analysis of Aggregate Gradation
The cumulative twoparameter Weibull distribution was used to describe the aggregate gradation as presented in Eq. 5. The MATLAB program was used to fit the Weibull function to aggregate size distribution. Figure 3 shows examples of Weibull functions for various HMA and sealcoat aggregate gradations. The xaxis represents the aggregate size in millimeters, and yaxis represents the cumulative percent passing of the aggregate. The scale (λ) and shape parameters (κ) were calculated by fitting the aggregate gradation to the cumulative twoparameter Weibull distribution. Higher values of λ were associated with coarser aggregate gradations. The research team also calculated these two parameters to describe the shape factor for all common mixture gradations used in Texas. Typically, the aggregate gradations running through the middle of the bands (allowed by TxDOT), for given mixture type, was used to calculate the default shape parameters.

(b) Examples of aggregate gradations of seal coat 
Figure 3. Weibull Distribution Function
Analysis of Aggregate Texture and Angularity
A total number of 56 different aggregate type/sources were examined this this study. The aggregate shape characteristics were measured using the AIMS device at three different levels of polishing using the MicroDeval device: before MicroDeval (BMD) or without any polishing, after 105 min. of polishing (AMD105), and after 180 min. (AMD180). However, the common practice at TxDOT is to measure the aggregate shape/texture characteristics before and after the microDeval abrasion test (0 and 105 min.). The researchers considered both procedures when developing analytical models to describe the change in angularity and texture of aggregates due to abrasion and polishing in the MicroDeval. One can see that the loss of texture and angularity is significant after 105 min. of polishing in the MicroDeval. After that polishing occurs at much slower rate. Equations 3 and 4 were used to describe the change in aggregate texture and angularity, respectively. Figures 4a and 4b show examples for the change in texture and angularity with polishing time. In addition, Figures 4a and 4b reports examples of the regression constants of Eq. 3 and 4, respectively.

(b) 
Figure 4. Regression Constants of Aggregate Texture
Determining the regression constants in Eqs. 3 and 4 requires three points of texture or angularity indices. The researchers used nonlinear regression analysis to predict the regression constants in Eqs. 3 and 4 using only two points of texture or angularity indices (BMD and AMD105) which is the standard practice at TxDOT. Sixteen aggregates sources were used in the regression analysis to develop equations to predict initial measurements, terminal measurements, and rate of change of texture and angularity. The SPSS software was used for the regression analysis. Equations 6 through 10 determine the regression parameters for texture loss for HMA surfaces based on two measurements before microDeval (BMD) and after 105 min. of polishing in MicroDeval (AMD).
 Texture coefficients:
aTX+bTX=0.999BMD+0.438
(
R2=1)
(6)
aTX=0.864AMD+14.985
(
R2=0.94)
(7)
cTX= 0.492+TL59.506(7.106 x ARI)
(
R2=0.60)
(8)
TL=
BMDAMDAMD
(9)
ARI=
AMD/BMD1(AMDBMD)2
(10)
where:
aTX+bTX
: Initial texture index
aTX
: Terminal texture index
cTX
: Rate of change in texture
BMD, AMD: texture index before and after 105 min. polishing in MicroDeval
TL, ARI: texture loss and aggregate roughness index respectively.
Similarly Equations 11 through 15 determine the regression parameters for angularity change for HMA surfaces based on two measurements; before microDeval (BMD) and after 105 min. of polishing in the MicroDeval (AMD).
 Angularity coefficients:
aGA+bGA=0.999BMD+2.646
(
R2=1)
(11)
aGA=1.237AMD699.759
(
R2=0.95)
(12)
cGA= 1.891+TL111.658+(1.081 x ARI)
(
R2=0.61)
(13)
TL=
BMDAMDAMD
(14)
ARI=
AMD/BMD1(AMDBMD)2
(15)
where:
aGA+bGA
: Initial angularity index
aGA
: Terminal angularity index
cGA
: Rate of change in angularity
BMD, AMD: angularity index before and after 105 min. polishing in MicroDeval.
TL, ARI: Angularity loss and aggregate roughness index respectively.
Development of Predictive Model for IFI
Masad et al. (2) and Kassem et al. (4) developed IFI prediction models. The parameters for the IFI model developed by Masad et al. (2) (
amix
,
amix+bmix
, and
Cmix
) relied on factors that describe aggregate texture and its resistance to abrasion and polishing, aggregate gradation and number of polishing cycles in the laboratory. The parameters for the IFI model (
amix
,
amix+bmix
, and
Cmix
) developed by Kassem et al. (4), used the same factors, in addition to factors that describe the aggregate angularity. Kassem et al. (4) demonstrated the measured IFI and predicted IFI had better correlation when aggregate angularity is considered in addition to aggregate texture. In this study, the models proposed by Kassem et al. (4) were used and calibrated to fit the wide range of aggregates examined in this study. The model developed by Kassem et al. (4) was based on limited number of aggregate types (soft limestone, intermediate limestone, and sandstone). The study herein evaluated about 56 different aggregate types. The researchers used the SPSS software in the IFI model development. Similar to Kassem et al. (4), the IFI model include three analytical models for its parameters (
amix
,
amix+bmix
, and
Cmix
). The
amix
presents the terminal IFI, the
amix+bmix
presents the initial IFI, while
Cmix
presents the rate of change of the IFI. Equations 16 through 18 show the modified models.
amix= 49.3144+λ351.2890.00193(AMD)2
(16)
amix+bmix=0.33*ln1.43757*aTX+bTX+46.8933*λ+333.491*k2.42031*aGA+bGA+1.00801
(17)
Cmix=0.018+1.654CTX+1.346CGA
(18)
where:
amix
: terminal IFI
amix+bmix
: initial IFI
Cmix
: rate of change in IFI
λ, k: scale and shape parameters of Weibull distribution
AMD: the texture after 150 min. in MicroDeval
aTX, bTX
: regression constants for texture
aGA, bGA
: regression constants for angularity
CTX
: rate of change in texture
CGA
: rate of change in angularity
It should be noted that Eq. 19 is a function of the number of polishing cycles in laboratory (N). Since the IFI models were revised based on the traffic levels, a relationship developed by Masad et al. (2 2011) was used to convert the traffic level to corresponding number of polishing cycles (N). This relationship is presented in Eq. 20.
IFI (N) =
amix+ bmix* e(Cmix*N)
(19)
N = TMF x
101A+B x cmix+ Ccmix
(20)
where:
N: number of polishing cycles in thousands
A, B and C: regression coefficients (0.452, 58.95, 5.843 x
106)
, respectively.
cmix
: rate change in IFI
TMF: Traffic Multiplication Factor
The Traffic Multiplication Factor (TMF) is calculated using Eq. 21 and the adjusted traffic is calculated using Eq. 22.
TMF =
Days between construction and field testing x adjusted traffic1000
(21)
Adjusted traffic =
AADTx 100PTTx DLAADT100 + AADT x PTT x DLtruck 100
(22)
where:
AADT: average annual daily traffic for each section
DLAADT:
design lane factor of AADT
DLtruck
: design lane factor of trucks
PTT: Percent of truck traffic
Figures 5a and 5b show the correlation between the predicted and measured IFI for the HMA and seal coat test sections, respectively. The data points in Fig. 5 include the IFI measurements at the wheel path (WP) and at the shoulder or between the wheel path (BWP). Higher rsquared indicates higher correlations between the predicted and measured IFI.
Analysis of Mean Profile Depth (MPD)
The researchers also developed a predictive model for MPD as a function of aggregate gradation, polishing cycles (or traffic level). The purpose of this model was to predict MPD if such information is not available for a given asphalt mixture or seal coat. Nonlinear regression was conducted using the SPSS software and the model for HMA surfaces is presented in Eq. 23. The researchers also developed a predictive model for MPD as a function of seal coat size and polishing cycles presented in Eq. 24. Figures 6a and 6b show the correlation between the measured MPD and the predicted MPD for the HMA and seal coat test sections, respectively. Equations 23, 24 indicates that the MPD decreases with traffic and coarser mixture have higher MPD.
MPD =(λ/34.180) – (0.398/k) + (
k0.416
) – 0.003N (23)
MPD = (λ/5.403) + (3.491/k) + (
k0.104
) +
N0.47
– 2.594 (24)
where:
λ, k: Weibull distribution parameters for aggregate gradation.
N: number of polishing cycles in thousands
The researchers believe that the fair correlation between measured MPD and the predicted MPD for the seal coat test sections is affected by the construction of the seal coat (e.g., proper embedment of the rock into the asphalt binder or emulations, the use of the right emulations, uniform distribution of the rocks, etc.
R^{2} = 0.67

R^{2} = 0.68

Figure 5. Relationship between Predicted and Measured IFI for
R^{2} = 0.74
(a) HMA test sections 
R^{2} = 0.53
(b) seal coat test sections 
Figure 6. Relationship between measured and calculated MPD values
Skid Number Analysis
The researchers used the developed IFI models to predict skid number at 50 mph. The authors used the PIARC 1995 (21) standard equation to predict skid number at 50 mph given the IFI values.
SN(50) = 4.82 + 140.3 (IFI – 0.045)
e20Sp
(25)
where:
IFI: predicted international friction index
Sp:
speed constant parameter
The predicted SN (50) values calculated using Eq. 25 were compared to the SN measured in the field using a skid trailer at 50 mph. Figures 7a and 7b show the correlation between the measured and predicted SN for the HMA and seal coat test sections, respectively. Overall, a good correlation was found between the calculated and measured skid resistance. Although the rsquared has a value of 0.63 for HMA test sections and 0.59 for seal coat test sections, such correlation is considered good given the influence of other factors affecting skid resistance (e.g., geometry of roadway, climatic condition, construction quality, etc.). Construction quality can affect the surface characteristics in many ways such as segregation, bleeding of asphalt, and rough surface due to uneven paving. Asphalt bleeding can significantly reduce the skid number; however, it is associated with poor construction practice.
The researchers further investigated the effect of the traffic level on the skid resistance. Traffic level is categorized in four groups as presented in Table 1. The Traffic level is expressed as TMF as presented in Eq. 21. Figure 8 shows the range of skid number values at different traffic level. In general, the SN decreased with the increase in the traffic level. Higher traffic level causes more polishing to the surface of asphalt pavements and thus reduces SN.
Table 1. Traffic Groups Based on TMF
Level  Traffic Multiplication Factor 
Low  0 – 15,000 
Medium  15,000 – 40,000 
High  40,000 – 90,000 
Very High  >90,000 
R^{2} = 0.63
(a) HMA test sections 
R^{2} = 0.59
(b) seal coat test sections 
Figure 7. Relationship between the measured and predicted SN
Figure 8. Measured Skid Numbers in terms of Traffic Level
Skid Resistance Model Sensitivity Analysis
The researchers examined the sensitivity of the skid resistance models of the HMA and seal coat, to various factors including aggregate gradation, type, and traffic level.
Effect of Mixture Gradation
Four mixtures with different aggregate gradations were evaluated; TypeC dense graded mixture, TypeD dense graded mixture (finer than TypeC), porous friction course (PFC), and stone matrix asphalt (SMAC) or Grades 3, 4, and 5 for seal coat mixtures. The performance of these mixtures in terms of skid number was assessed using the developed model. All variables (e.g., traffic level, aggregate characteristics, etc.) were held constant and only the aggregate gradation was varied. Figure 9a shows that the mixtures with coarse aggregate gradations (such as PFC and SMAC) had higher skid numbers. Similarly, Fig. 9b shows that Grades 3 (coarser gradation) provides higher skid number compared to Grades 4 and 5. The coarse aggregate gradation provides higher macrotexture and thus yields higher SN.
Effect of Traffic Level
Figures 9c and 9d demonstrate the effect of traffic levels or AADT on the skid resistance for HMA and seal coat test sections, respectively. The results showed that the skid number decreases with traffic level as expected; however, the SN had a steep slope or reduction at higher traffic levels. Pavement surface experiences most polishing at higher traffic levels which adversely affects the skid resistance.
Effect of Aggregate Type
Four different aggregate types (e.g. limestone, sandstone, dolomite, and different combinations) were examined. The traffic level and aggregate gradation were fixed. Figures 9e and 9f demonstrated that HMA or seal coat prepared using aggregates with rough texture such as sandstone provide higher skid number and low rate of skid reduction compared to mixtures with soft rock such as limestone. Thus, it is recommended to use aggregates with rough texture in the construction of HMA and seal coat especially those subjected to high traffic levels. Blending of aggregates with higher polishing resistance with local aggregate is recommended if the local aggregates have poor resistance to abrasion and polishing.
b)
a)

c)
d)

f)
e)

Figure 9. Sensitivity analysis for parameters affecting SN
Skid Analysis of Asphalt Pavement Computer Application
The researchers developed a computer application using Accessbased VBA language to execute the steps needed to calculate the skid resistance of asphalt pavements as well as the seal coat. This utility is called skid analysis of asphalt pavement (SAAP) and it is a revised version of the one developed by Masad et al. (2). The version include the modified model for the HMA and the new model for the seal coat. This utility can be used by pavement community and contractors to 1) ensure that a given mix design will provide the expected skid number over its service life, 2) optimize the mix design of HMA and aggregate blending to provide enhanced skid resistance based on the traffic level and available aggregates, 3) determine the proper seal coat grade and aggregate source to extend the skid service life, and 4) program and schedule preventive maintenance activities to ensure that pavements have adequate skid resistance. Figure 10 summarizes the steps of needed to predict skid number using the SAAP application as follows.
 Figure 10a shows the program interface. The interface provides a short introduction about the application and its purpose.
 The user needs to select surface type (e.g., HMA or seal coat) so the application uses the appropriate prediction model (Figure 10b).
 The user enters the aggregate gradation or select one of the standard seal coat of HMA mixture gradation used in the state of Texas (Figure 10c).
 The user selected the number of data points of before and after microDeval (e.g., two or three points) (Figure 10d)
 The user enters the aggregate texture and angularity values measured using AIMS before and after microDeval (Figures 10e and 10f).
 The SAAP estimates the initial MPD or gives the option to the user to enter it (Figure 10g).
 The user inputs the highway configuration and traffic data (Fig. 10h).
 The software can provide a prediction of skid resistance as a function of years in service (up to 15 years), or provide a classification of the pavement section based on its skid resistance after a specified number of years and corresponding threshold values as shown in Figs 10i to 10l.
Conclusions and Recommendations
This paper summarizes the research efforts by the authors to develop prediction models for the international friction index (IFI) and skid number (SN) for HMA pavements and seal coat. These models were developed based on parameters that describe the aggregate gradation, aggregate shape characteristics (texture and angularity) and its resistance to abrasion and polishing, in addition to traffic level. The results demonstrated good correlation between measured and predicted skid numbers in the field. The aggregate type and gradation as well as traffic level were found to have significant effect on skid resistance and rate of skid reduction. Coarse aggregate gradation provide higher macrotexture and high skid resistance compared to fine aggregate gradations. Higher traffic level caused steep reduction in skid number due to the significant surface polishing in a short time. Asphalt mixtures and seal coat prepared with aggregates with rough surface had higher skid resistance compared to asphalt mixtures with smooth aggregates. The results demonstrated that the AIMS and MicroDeval tests were found to be proper tools to evaluate the aggregate shape characteristics and its resistance to abrasion and polishing. The developed models can be incorporated in a Pavement Management System (PMS) at the network level to plan and program preventive maintenance activities to ensure that pavements have adequate skid resistance. These models can also be used during mix design procedure to optimize the aggregate selection and aggregate gradation to produce mixtures with adequate friction. The reader is referred to Chowdhury et al. (2017) for more information about this research study.
Figure 10. Steps Needed to Predict Skid Number SAAP
References
 Noyce, D. A., Bahia, H. U., Yambo, J. M., and Kim, G. (2005). Incorporating Road Safety into Pavement Management: Maximizing Asphalt Pavement Surface Friction for Road Safety Improvements. Draft Literature Review and State Surveys, Midwest Regional University Transportation Center (UMTRI), Madison, Wisconsin.
 Masad, E., Rezaei, A., Chowdhury, A., and Freeman, T. J. (2011). Field evaluation of asphalt mixture skid resistance and its relationship to aggregate characteristics (No. 056273). Texas Transportation Institute.
 Kassem, E., Masad, E., Awed, A., and Little, D. (2012). Laboratory Evaluation of Friction Loss and Compactability of Asphalt Mixtures (No. SWUTC/12/476660000251).
 Kassem, E., Awed, A., Masad, E., and Little, D. (2013). Development of Predictive Model for Skid Loss of Asphalt Pavements. Transportation Research Record: Journal of the Transportation Research Board, (2372).
 Hall, J. W., Smith K. L., TitusGlover, L. (2009). Guide for pavement friction. National cooperative highway research program, contractors final report for NCHRP project 0143.
 Flintsch, G. W., McGhee, K. K., and de Leon, E. (2005). Field validation of macrotexturebased hot mix asphalt segregation detection methods. Journal of the Association of Asphalt Paving Technologists.
 Forster, S. W. (1989). Pavement microtexture and its relation to skid resistance. Transportation Research Record, (1215).
 Anupam, K., Srirangam, S.K., Scarpas, A. and Kasbergen, C., (2013). Influence of Temperature on TirePavement FrictionII: Analysis. Transportation Research Record, Vol 4, pp. 114124
 Cairney, P. (1997). Skid Resistance and Crashes: a review of the literature (No. ARR 311).
 Ivey, D. L., D. L. Bullard, J. R. Lock, and L. I. Griffin III (1992). Texas Skid Initiated Accident Reduction Program: Final Report. Report 9101F, TTI: 21889/910, TX92/9101F, Texas Department of Transportation.
 Galambos, V., Hegmon, B., & Rice, K. (1977). Brinkman. Pavement Texture and Available Skid Resistance.
 Texas department of transportation, Standard Specifications for Construction and Maintainance of Highways, Streets, and Bridges. TxDOT 2003
 Krugler, P. E., T. J. Freeman, J. E. Wirth, J. P. Wikander, C. K. Estakhri, and A. J. Wimsatt (2012). Performance Comparison of Various Seal Coat Grades Used in Texas (No. FHWA/TX12/064961).
 Roque, R., D. Anderson, and M. Thompson (1991). Effect of Material, Design, and Construction Variables on SealCoat Performance. Transportation Research Record, No. 1300.
 Rezaei, A., E. Masad, and A. Chowdhury (2011). Development of a Model for Asphalt Pavement Skid Resistance Based on Aggregate Characteristics and Gradation. Journal of Transportation Engineering, Vol. 137, No. 12.
 Masad, E., Luce, A., Mahmoud, E., and Chowdhury, A. (2007). Relationship of Aggregate Texture to Asphalt Pavement Skid Resistance Using Image Analysis of Aggregate Shape. Final Report for Highway IDEA Project.
 Wu, Z., King, B., Abadie, C., and Zhang, Z. (2012). Development of design procedure to predict asphalt pavement skid resistance. Transportation Research Record: Journal of the Transportation Research Board, (2306).
 Chowdhury, A., Kassem, E., Aldagari, S., Masad, E. (2017) Validation of Asphalt Mixture Pavement Skid Prediction Model and Development of Skid Prediction Model for Surface Treatments. Technical Report. Report FHWA/TX17/06746011. Texas Department of Transportation.
 AASHTO (2002). AASHTO Designation TP 5800: Standard Test Method for Resistance of Coarse Aggregate to Degradation by Abrasion in the MicroDeval Apparatus. AASHTO Provisional Standards, American Association of State Highway and Transportation Officials, Washington, D.C.
 Mahmoud, E., and Masad, E. (2007). Experimental methods for the evaluation of aggregate resistance to polishing, abrasion, and breakage. Journal of Materials in Civil Engineering, 19(11).
 Wambold, J.C., Antle, C.E., Henry, J.J., and Rado, Z. (1995). International PIARC Experiment to Compare and Harmonize Texture and Skid Resistance Measurements: Final Report.