Exploring Aggregate Morphological Characteristics under Laboratory Polishing for Enhanced Pavement Skid Resistance

Abstract

As the use of recycled asphalt pavement (RAP) in pavement construction grows for sustainable development, it becomes essential to investigate potential frictional deterioration over time. This study evaluated the friction properties of recovered RAP material aggregates compared with raw aggregates across various polishing cycles. The micro-Deval test was employed to simulate aggregate loss of texture, while morphological and friction properties were measured using an aggregate imaging measurement system (AIMS-II), along with a British pendulum tester (BPT) and dynamic friction tester (DFT). Additionally, Fourier transform infrared spectroscopy (FTIR) was employed to assess its potential in determining the origin and composition of RAP material aggregates. A simple method was used to fabricate custom aggregate rings, allowing for accurate testing in the DFT setup. The aggregate testing results revealed notable variations across the measurement techniques. AIMS-II analysis showed that traprock (maroon-colored) exhibited the highest surface texture, while DFT and BPT results indicated that certain limestones outperformed traprock in friction properties. Additionally, the testing results demonstrated that the RAP materials were comparable to, or even outperformed, certain limestone sources. However, because of potential variability within RAP stockpiles, careful quantification is necessary to assess their suitability. FTIR analysis demonstrated its ability to distinguish between carbonate-rich and silica-rich aggregates; however, further research is needed to build a library of aggregate sources. Finally, a machine learning algorithm identified the loss of aggregate DFT₂₀ values as the most significant aggregate property representing friction loss in asphalt mixtures.

Keywords

aggregate morphology skid resistance recycled asphalt pavement material aggregate rings aggregate microtexture

Introduction

Pavement skid resistance is a critical safety feature of road surfaces that affects the ability of a vehicle’s tires to maintain traction, especially under wet conditions. Long-term pavement frictional performance relies on factors such as asphalt mixture aggregate properties, traffic load and volume, road geometry, seasonal variation, and rainfall ( 1 ). The pavement surface properties, particularly surface texture characteristics, directly affect the pavement’s skid resistance and ability to reduce noise levels. The pavement surface texture, including microtexture and macrotexture, is primarily influenced by aggregate characteristics and mix gradation, providing the necessary friction at different speeds and conditions. Microtexture provides friction at the tire-rubber level, while macrotexture ensures water drainage and maintains friction at higher speeds ( 1 ). Asphalt pavement surfaces are subject to continuous traffic-induced polishing action, requiring the retention of adequate surface friction with continuous exposure to vehicular traffic to meet the safety standards of highways. Consequently, state DOTs have initiate programs to enhance skid resistance through material selection, surface treatments, and maintenance practices such as grooving or applying high-friction surface layers.

Morphological properties of aggregates, including texture, shape, angularity, and size, significantly influence asphalt pavement skid resistance. Explicitly, aggregate shape characteristics significantly influence the interlocking level of the aggregate particles and their interaction with asphalt binder during the compaction process ( 2 , 3 ). Thus, it is essential to accurately quantify the aggregate shape characteristics to understand their influence on pavement surface properties and select the best aggregate source for adequate long-term pavement friction performance. Two primary methodologies are commonly employed to quantify aggregate shape characteristics: traditional manual techniques and digital image processing (DIP) methods ( 4 – 9 ). Notably, the traditional methods for aggregate shape characteristics are notably labor-intensive and prone to influence from subjective judgments. Thus, many research studies have employed DIP methods to explore the aggregate shape characteristics. The aggregate image measurement system (AIMS) is widely utilized to monitor the change of aggregate shape characteristics with polishing ( 6 , 10 – 12 ). Many research studies have utilized the AIMS technique to evaluate the relationship between aggregate texture and pavement skid resistance and propose an empirical model for predicting the friction loss in asphalt mixtures based on mix gradation and the aggregate shape characteristics measured by AIMS at various micro-Deval test (MDT) times ( 11 , 13 , 14 ).

Aggregate friction properties are commonly measured by the British pendulum tester (BPT), dynamic friction tester (DFT), and Wehner/Schulze (W/S) device ( 15 – 22 ). These techniques utilize a rubber pad of a specific shape to assess the microtexture properties of pavement surfaces. BPT measures the surface friction at a low speed of 10 km/h using a 1.25 in. rubber slider width. In contrast, DFT has three rubber pads, each 0.6 in. wide, attached to a rotating disk. DFT quantifies the friction coefficient at varying speeds, ranging from 10 to 90 km/h. Specifically, the DFT value at 20 km/h (referred to as DFT₂₀) is indicative of the surface microtexture properties ( 23 , 24 ). Similarly, the W/S device has three rubber pads to measure the dynamic friction coefficient at a speed of 60 km/h (denoted as µ₆₀).

Forster explored the aggregate surface texture through image analysis techniques and found a correlation with aggregate friction, measured by BPT ( 25 ). However, other studies reported that BPT results relied on several factors, including contact path, slider load, coupon curvature, aggregate size, and particle arrangement, other than the aggregate surface texture ( 26 – 29 ). Additionally, a few studies found that the difference between BPT results typically fell within a small range of about four polished stone values, challenging the differentiation of higher-performing aggregates ( 30 – 32 ). Consequently, alternative testing methods, such as DFT and W/S devices, were developed to measure the friction properties of aggregates and asphalt mixtures to overcome the BPT limitations. Several studies adopted DFT to quantify the friction properties of raw and recovered recycled asphalt pavement (RAP) material aggregates ( 15 – 18 ). Saghafi et al. studied the impact of several factors—DFT rubber pad life, polishing devices, and aggregate arrangements—on the reproducibility of DFT₂₀ ( 17 ). It was reported that aggregate ring-shapes with 3%–16% gaps demonstrated comparable polishing and friction results (DFT₂₀) ( 17 ). It was concluded that the ranking of blended aggregates polished for 105 min in the MDT, as determined by DFT results, corresponded with the DFT of asphalt slabs ( 18 ).

Several laboratory techniques are commonly used for polishing aggregates, simulating the wear on pavement surfaces caused by traffic. These methods aim to assess the durability and resistance of aggregates to polishing and degradation, providing insights into how pavement surfaces may evolve over time in friction and safety. Among these testing methods, MDT is commonly employed for aggregate durability evaluation ( 33 – 36 ). MDT simulates the polishing process by reflecting changes in aggregate morphology, including surface texture, shape, and angularity, in the presence of water. Micro-Deval loss (MDL) represents the weight loss resulting from polishing and degradation, with a lower MDL indicating high resistance to polishing ( 37 – 39 ). Notably, the MDT results had good reproducibility between laboratories ( 40 ). MDL serves as an indicator of aggregate wear in models predicting road surface polishing ( 41 ). Wu et al. conducted a comprehensive study to identify the criteria limits for durable asphalt pavement, considering several toughness and abrasion testing methods, and concluded that the 18% MDL at 105 min is an effective threshold for differentiating poor and good aggregates in aggregate abrasion and degradation resistance ( 39 ).

Study Goal and Objectives

This research study aimed to study the friction properties of RAP materials to assess any potential deterioration that occurred during the pavement’s construction and service life. Additionally, the most significant aggregate property of the friction loss of asphalt mixture was evaluated. The goal was achieved through:

Monitoring the change in the friction and morphological properties of recovered RAP material aggregate and raw aggregates at various polishing cycles.

Evaluating the effectiveness of utilizing Fourier transform infrared spectroscopy (FTIR) to distinguish between aggregate sources and identify the origin of RAP material aggregates.

Quantifying friction properties of asphalt mixtures, including stone matrix asphalt (SMA) and dense-graded mixtures.

A machine learning algorithm was performed to explore the correlation between the aggregate friction-related properties and asphalt mixtures, thus identifying the key aggregate properties affecting pavement surface friction.

Materials and Methods

Materials

This study evaluated eight raw aggregate sources—limestone, dolomite, and traprock, along with five RAP materials acquired from various quarries in Missouri, U.S.—as shown in Figure 1. The selection of aggregate sources was based on their frequent use in wearing course applications in Missouri. Traprock, a noncarbonate rock, is notably rich in quartz (SiO₂) and feldspar (Al₂O₃), whereas limestone and dolomite are carbonate rocks predominantly composed of calcite ( 42 , 43 ). Furthermore, two bonding agents, body filler and fiberglass resin (bondo), shown in Figure 2, were used to prepare the aggregate rings and coupons for aggregate friction testing. Four asphalt mixtures were fabricated using different limestone sources to investigate the influence of aggregate friction properties. All the SMA mixtures were designed to comply with the balanced mix design criteria ( 44 ).

Figure 1.

Locations of the acquired aggregate sources and recycled asphalt pavement materials.

Figure 2.

Bonding agents for the aggregate rings and coupons.

Aggregate and Mixture Sample Preparation

Sample Preparation for Aggregate Friction Properties Evaluation

Two sample geometries were prepared to evaluate the friction properties of aggregates: coupons for BPT and ring-shaped specimens for DFT. These aggregate samples underwent preparation for three distinct durations of MDT (0, 105, and 180 min), similar to AIMS analysis. The aggregate rings and coupons required careful manual arrangement of aggregate particles to achieve maximum packing density. This process ensured that the flat side of each particle was oriented downwards, covering the mold’s bottom surface thoroughly to maximize the contact area with the rubber sliders during testing ( 15 , 16 , 18 , 45 ). Also, a bonding agent was poured to hold the arranged aggregates and left for curing (for almost 2 h). Finally, the samples were demolded and tested using the appropriate testing methods, as presented in Figure 3.

Figure 3.

Aggregate coupons and ring-shaped specimens: (a) aggregate coupons for British pendulum testing and (b) aggregate ring-shaped specimens.

Preparation of Asphalt Mixture Slabs

The asphalt mixture slabs, intended for friction evaluation, were produced following AASHTO PP104-21 ( 46 ). Two square slabs were created and compacted for each mixture using a vibratory compactor, as demonstrated in Figure 4. The mix’s total weight was determined as the slab’s dimensions, the mixture’s maximum theoretical specific gravity (G_mm), and the targeted air void content (7% ± 1% V_a), all calculated using Equation 1. Additional details on the sample preparation process can be found elsewhere ( 45 ).

Mix Wt . = (lwt) (G_{mm} \cdot ρ_{w}) [\frac{(100 - V_{a})}{100}] .

(1)

where

l = the length of the slab (mm),

w = the width of the slab (mm),

t = the depth of the slab (mm), and

ρ _w = water density (in 0.001 g/mm³).

Figure 4.

Asphalt mixture slab preparation steps.

Testing Methods

Fourier Transform Infrared (FTIR) Spectroscopy

This study performed FTIR spectroscopy on aggregate powder. The FTIR spectra for each sample were recorded over a wavenumber range of 400–4,000 cm⁻¹ at room temperature, with a resolution of 4 cm⁻¹ and 16 scans per sample. After analyzing each sample, the diamond crystal was meticulously cleaned using dust-free tissue and ethanol to ensure accuracy and prevent contamination. The FTIR spectroscopy was conducted to evaluate its effectiveness in detecting the origin of the RAP material aggregates.

Acid Insoluble Residue (AIR)

The AIR test was conducted to assess the polishing resistance of aggregates by assessing the noncarbonate content in carbonate rock types, according to ASTM D3042 ( 47 ). Hydrochloric acid (HCl) was added to a container with 500 g of coarse aggregates (R#4) and agitated until effervescence ceased. This process was repeated with an additional 300 ml of HCl until no reaction occurred. The sample was then heated at 110°C for 1 h, repeating until reactions stopped. The solution was decanted, and diluted water was added to adjust the pH to over 5.5. Finally, the sample was washed over a #200 sieve (0.075 mm), and the oven-dried weight of the residue was recorded to compute the %AIR following ASTM D3042.

Micro-Deval Test (MDT)

MDT was performed to evaluate the abrasion resistance of aggregates following ASTM D6928 ( 37 ). The MDT aimed to evaluate the durability of aggregates and their ability to withstand polishing, abrasion, and grinding in the presence of water for 105, 180, and 240 min. Aggregate samples were coded as “BMD” for unpolished aggregates, and as “AMD105,”“AMD180,” and “AMD240” corresponding to polishing durations of 105, 180, and 240 min, respectively. MDT was run on four replicates for each aggregate type and run time, ensuring consistency and reliability in the obtained results. Furthermore, two gradings were used to study the impact of grading on the MDL values: grading A considered the aggregate size passing through a 1/2 in. sieve and retained on a #4 sieve (3/8 in./#4) while grading B represented the aggregate size passing through a 1/4 in. sieve and retained on a #4 sieve (1/4 in./#4).

Aggregate Image Measurement System (AIMS)

AIMS was utilized to quantify the change in aggregate morphology, surface texture, angularity, and sphericity before and after the MDT ( 48 ). Aggregate shape characteristics were quantified at three MDT run times: 0, 105, and 180 min ( 33 ). Figure 5 illustrates the limits of the AIMS indices, including texture (TX), gradient angularity (GA), and Form2D indices. Additionally, AIMS developed the coarse aggregate angularity texture (CAAT) index, which combines angularity and texture measurements to provide a comprehensive overview of aggregate micromorphology.

Figure 5.

The aggregate imaging measurement system aggregate morphology images.

British Pendulum Tester (BPT)

A BPT was employed to evaluate the friction properties of coarse aggregate coupons at a low speed of 10 km/h, according to ASTM E303-93 ( 49 ). BPT represents the loss of kinetic energy when a rubber slider is dragged across the test surface, reflecting the surface’s friction properties. BPT serves as a simple screening method to evaluate the microtexture properties of coarse aggregate coupons, utilizing rubber pads measuring 1.25 in.

Dynamic Friction Tester (DFT)

A DFT evaluated the friction characteristics of both slabs and aggregate rings under wet conditions following the ASTM E1911 ( 23 ). DFT assessed the surface friction across a speed range of 70 to 10 km/h, ensuring precision by conducting at least two repeated tests for each tested surface.

Three-Wheel Polishing Device (TWPD)

A TWPD was used to simulate the traffic-induced polishing action in the presence of water in the laboratory, according to AASHTO PP104 ( 46 ). The wheel assembly has three pneumatic tires rotating along an 11.2 in. diameter circular path compatible with both the DFT and CTMeter devices. The combined load exerted by the wheels was approximately 105 lb, with the tire pressure maintained at around 35 ± 5 pounds per square inch. Additionally, using DFT, the asphalt mixture friction properties were recorded at several TWPD cycles up to 100 K cycles.

Results and Discussion

The centrifuge extraction method was used to characterize the RAP material, as the ignition oven method would have degraded aggregate surface characteristics and induced micro-cracks. Visual observation revealed that the recovered RAP material aggregates consisted of two types of mineralogy—traprock and limestone—as supported by FTIR analysis (Figure 6). The aggregate morphology was quantified separately for each type, and a weighted average of the aggregate shape indices was subsequently computed.

Figure 6.

Fourier transform infrared spectroscopy analysis results: (a) raw aggregates and (b) recycled asphalt pavement (RAP) materials.

As illustrated Figure 6, the limestone and dolomite sources had almost similar peaks with different bands intensities and widths at around 1,450–1,400 (carbonate ν₃ asymmetric stretching), 890–870 (carbonate ν₂ out-of-plane bending), and 730–710 cm⁻¹ (carbonate ν₄ in-plane bending) wavenumbers, along with a secondary quartz Si–O mineral at 1,100 cm⁻¹, indicating a carbonate-rich composition. In contrast, the traprock source displayed distinct absorption bands near 1,000 cm⁻¹, corresponding to a strong broad Si–O asymmetric stretch, as well as bands around 800–780 cm⁻¹ and 560–520 cm⁻¹ that are consistent with silicate mineral vibrations, indicating a silica-rich mineral composition.

The FTIR analysis of RAP materials revealed the presence of two distinct aggregate sources within the RAP material. Specifically, RAP-1, RAP-2, and RAP-5 were predominantly carbonate-rich, RAP-4 was silica-rich, and RAP-3 contained a blend of traprock and limestone aggregates. The analysis revealed the possibility of utilizing FTIR to identify the origin of the RAP material aggregate once a comprehensive library is created. However, further analysis is needed to create a library for different aggregate sources and validate these findings.

Abrasion and Polishing Resistance of Aggregates

As expected, traprocks, typically rich in quartz (SiO₂) and feldspar (Al₂O₃), exhibited the highest resistance to polishing and degradation, indicated by their lowest MDT values, as shown in Figure 7. The MDL percentage of RAP materials was comparable to limestone sources, as the RAP material contained a fraction of traprock; this was supported by the AIR results (Figure 8). A variation in MDL values was observed for the same aggregate type (e.g., limestone), as shown in Figure 7. This variability can be attributed to the unique formations of rocks under different local conditions. For example, LS-2 had an MDL of 12.5%, whereas LS-3 had a higher MDL of 19%. This variation is attributed to the ratio of calcite (CaO) to dolomite (MgO) and the presence of minor constituents such as quartz and feldspar typically found in limestones. This finding highlighted the influence of aggregate mineralogy on abrasion resistance, regardless of rock type, emphasizing the need for a closer examination of the relationship between aggregate mineralogy and aggregate performance. It was found that the MDL percentage increased linearly with MDT run times, with the slope of this linear relationship defined as the abrasion rate, as presented in Figure 7.

Figure 7.

Micro-Deval test results of investigated aggregates.

Figure 8.

Acid insoluble residue (AIR) results.

Moreover, traprock and five limestone sources, LS-1, LS-2, LS-3, LS-4, and LS-5, were considered to evaluate the influence of aggregate grading on the MDT results. A statistical two-way analysis of variance (ANOVA) was conducted to monitor the significance of aggregate grading and MDT run times on the MDL percent. The ANOVA hypotheses are the null hypothesis (H₀), which indicates all means are equal (μ₁ = μ₂), and the alternative hypothesis (H_a), which means at least one of the groups’ means is different from the others (μ₁ ≠ μ₂). Table 1 summarizes the ANOVA testing results of the significance of aggregate grading and MDT cycles. The ANOVA analysis revealed a significant impact of MDT cycles on MDT results, while the effect of aggregate grading on these results differed across various sources.

Table 1.

Analysis of Variance (ANOVA) Testing Results of Significance: Aggregate Grading and Micro-Deval Test (MDT) Run Times

ANOVA parameter	Traprock	Limestone sources
ANOVA parameter	TR-1	LS-1	LS-2	LS-3	LS-4	LS-5
MDT cycles
P-value	0.0409	0.0009	0.0071	0.0025	0.0072	0.0130
F	71.8	27.1	12.6	407.0	138.0	76.1
F crit	18.5	18.5	18.5	19.0	19.0	19.0
Grading
P-value	0.0136	0.0349	0.0712*	0.0768*	0.0364	0.0550*
F	23.4	1097.7	140.0	11.5	26.0	16.7
F crit	19.0	19.0	19.0	18.5	18.5	18.5

Insignificant change because of changing aggregate grading.

Analysis of Aggregate Morphology and Friction Properties

This section presents the influence of MDT on the aggregate friction-related properties, considering aggregate morphology and friction properties measured by DFT and BPT. As expected, the abrasive action of MDT tends to smooth out the aggregate surface irregularities. In general, it was observed that the friction-related properties decreased with MDT run times. Equation 2 expresses the change in the aggregate friction-related properties with MDT ( 33 ).

X (t) = a_{X} + b_{X} \times \exp (- c_{X} \times t) .

(2)

where

X(t) = the aggregate friction-related properties, aggregate morphology, DFT, and BPT values,

a _X, b_X, and c_X = the model constants, and

t = micro-Deval run time (in minutes).

Analysis of AIMS Results

The morphological characteristics of aggregates were quantified at various MDT run times (0, 105, and 180 min). Around 50 particles of each aggregate source were collected for further AIMS evaluation. The outlier analysis using interquartile range, represented by a box-and-whisker plot, was performed to ensure the reliability and accuracy of statistical analyses. Figure 9 presents an example of the performed outlier analysis. Outliers are depicted as individual points beyond the whiskers. It also appeared that processing the data (perhaps to remove outliers or adjust for errors) resulted in lower variability and fewer extreme values, as indicated by the smaller range and fewer outliers in the processed data than in the raw data.

Figure 9.

Example outlier analysis of LS-4 AIMS indices using box-and-whisker plots: (a) TX index and (b) GA index.

Furthermore, the average and standard deviation of AIMS indices were computed after removing the outliers, as shown in Figure 10. Generally, the aggregate morphology (i.e., TX, GA, CAAT, and Form2D) decreased because of interactions among aggregates or with the steel wall of the abrasion device, indicating that the aggregate edges and surface texture wore with polishing and became smoother. In contrast, no direct trend was observed between the AIMS sphericity index and MDT run time. To avoid misquantification of surface texture results by AIMS when testing aggregates of different surface colors, the recovered RAP material traprock was compared with the raw traprock, and the recovered RAP material limestone was compared with the raw limestone. TX indicated that the recovered RAP material aggregates exhibited superior or comparable texture to the raw aggregates, as the milling and crushing processes may fraction the particles and expose the internal surface texture. However, the GA of the recovered RAP material aggregates was lower than that of the raw aggregates. This discrepancy could be attributed to the wearing and rounding of aggregate edges during pavement construction, RAP stockpiling, or both, which would reduce particle angularity. Conversely, the asphalt mastic coating preserved the aggregate surface texture, which protected the texture during these processes.

Figure 10.

Example of average AIMS results across MDT run times: (a) texture index, (b) angularity index, (c) Form2D index, and (d) spherical index.

A Pearson’s correlation coefficient (PCC) was performed within the AIMS results to evaluate the linear correlation between two sets of AIMS indices, as illustrated in Figure 11. The correlation analysis indicated good linear correlations between GA and Form2D and sphericality and flat and elongated (F&E) indices and strong correlations between CAAT and TX and CAAT and GA parameters. MATLAB software was utilized to calculate the regression coefficients along with their 95% confidence interval, as demonstrated in Table 2.

Figure 11.

Correlation analysis results for AIMS indices: (a) PCC matrix for the AIMS indices, (b) AIMS GA versus Form2D, (c) AIMS F&E versus AIMS SP, (d) CAAT versus AIMS GA, and (e) CAAT versus TX.

Table 2.

The Correlations between Aggregate Morphology

Relationship	Models*	Goodness of fit	Equation ID
CAAT versus AIMS TX	$CAAT = a \times TX + b$ • a = 11.65 (11.26, 12.03) • b = 828.8 (715.4, 942.3)	SSE = 1.22E+07 R² = 0.99 RMSE = 181.7	(3)
CAAT versus AIMS GA	$CAAT = a \times TX + b$ • a = 2.725 (2.254, 3.196) • b = −2672 (−3820, −1523)	SSE = 2.7E+07 R² = 0.79 RMSE = 853.5	(4)
CAAT versus TX & GA indices	$CAAT = a \times TX + b \times GA$ • a = 9.959 (9.697, 10.22) • b = −0.53 (0.498, 0.562)	SSE = 2.66E+07 R² = 0.998 RMSE = 84.73	(5)
AIMS GA & Form2D	$GA = a \times Form 2 D + b$ • a = 599.1 (456.8, 741.3) • b = −1706 (−2680, −731.6)	SSE = 4.55E+06 R² = 0.663 RMSE = 350.6	(6)
F&E & AIMS SP	$F E = a \times SP + b$ • a = −6.931 (−8.319, −5.542) • b = 7.023 (6.082, 7.964)	SSE = 0.747 R² = 0.727 RMSE = 0.1421	(7)

Values in parentheses indicate the 95% confidence interval range for model coefficients.

Analysis of BPT and DFT Results

BPT and DFT were utilized to measure the friction properties of aggregate coupons and rings at various MDT run times (0, 105, and 180 min). Similar to the AIMS results, both BPT and DFT results decreased with MDT, as presented in Figure 12.

Figure 12.

Friction properties of the investigated aggregate sources at different polishing cycles: (a) BPT data and (b) DFT₂₀ values.

To study the correlation between DFT and BPT measurements, the DFT values at various speeds (60–10 km/h), were recorded and studied to explore the influence of aggregate microtexture on the DFT slip speed curve. Figure 13 shows an example of the DFT speed slip curve for two sources, TR-1 and LS-4, at different polishing conditions. The one-way ANOVA analysis revealed that the DFT values relied significantly on the DFT speeds; thus, the DFT values at 20, 40, and 60 km/h were considered to study the correlation between aggregate friction properties.

Figure 13.

Example DFT speed–slip curves under different MDT conditions for: (a) TR-1 and (b) LS-4.

Finally, the PCC matrix was performed to evaluate the correlation between the DFT values at different speeds and BPT data, as shown in Figure 14, a–d, which presents the linear regression that fitted the DFT values and BPN, with overall R² values of 0.875, 0.884, and 0.91, sequentially. Table 3 presents the coefficients of the linear models with 95% confidence intervals for each coefficient. Additionally, there was an increase in the R² value with increasing DFT speed, implying that the correlation between BPT measurements and higher-speed DFT measurements becomes more pronounced.

Figure 14.

Relationship between DFT and BPT results: (a) PCC matrix, (b) DFT₂₀ versus BPT, (c) DFT₄₀ versus BPT, and (d) DFT₆₀ versus BPT.

Table 3.

The Details of the BFT-BPN Models

Model 8	Model 9	Model 10
${DFT}_{20} = a_{20} \times BPN + b_{20}$	${DFT}_{40} = a_{40} \times BPN + b_{40}$	${DFT}_{60} = a_{60} \times BPN + b_{60}$
a₂₀ = 0.0198 (0.0171, 0.0226) b₂₀ = −0.315 (−0.40, −0.2296)	a₄₀ = 0.0197 (0.0171, 0.0224) b₄₀ = −0.349 (−0.431, −0.2679)	a₆₀ = 0.0197 (0.0174, 0.022) b₆₀ = −0.349 (−0.435, −0.292)
➢ SSE = 0.0401 ➢ R² = 0.875 ➢ RMSE = 0.036	➢ SSE = 0.03646 ➢ R² = 0.884 ➢ RMSE = 0.0343	➢ SSE = 0.0284 ➢ R² = 0.91 ➢ RMSE = 0.030

Note: DFT-BPN = dynamic friction tester measurement-British pendulum number.

Relationships between the Change of Friction-Related Properties and MDL

The correlation between the change of the aggregate morphology indices (%ΔX_i) and MDL was evaluated. The percent loss of aggregate friction-related properties and mass loss with MDT was calculated using Equation 11:

% Δ X_{i} = \frac{X_{i, BMD} - X_{i, AMD}}{X_{i, BMD}} \times 100

(11)

where

X_i, _BMD = the aggregate property or weight before MDT, and

X_i, _AMD = the aggregate property or weight after MDT.

Then, a PCC matrix was computed to evaluate the MDT influence on the loss of aggregate friction-related properties, as presented in Figure 15. The correlation analysis showed MDL had a strong positive correlation with the % loss of AIMS CAAT, TX, and GA and moderately with the % loss of DFT and BPT values. Figure 16 illustrates the linear regression that fitted the % loss of AIMS indices and MDL values.

Figure 15.

PCC matrix showing the influence of MDT on the loss of aggregate friction-related properties.

Figure 16.

Change in AIMS indices as a function of MDL, with confidence and prediction limits: (a) TX index and (b) CAAT index.

Machine Learning Analysis of Aggregate Properties on Asphalt Mixture Friction

To validate the most influential aggregate property in predicting friction loss in asphalt mixtures, the friction properties of four asphalt mixtures with different aggregate sources were evaluated at various TWPD polishing cycles using DFT.

Friction Properties of Asphalt Mixtures

DFT was also employed to measure the friction properties of asphalt mixture slabs at various TWPD cycles following AASHTO PP104-21 ( 46 , 50 ). While aggregate friction properties consistently decrease with polishing cycles, the asphalt mixture friction, represented by DFT₂₀, initially follows a different trend. During the first few thousand polishing cycles, the DFT₂₀ value increases during the first polishing cycles because of wearing the binder off, concealing the surface aggregates’ microtexture. Once this binder is worn off, exposing the aggregates texture, the friction values follow a similar trend to the aggregate friction properties as the aggregate microtexture starts to polish and lose texture, as shown in Figure 17 ( 51 ).

Figure 17.

Example of the friction properties (DFT₂₀ values) of asphalt mixtures.

While the DFT results for blended aggregates aligned with the friction results of the asphalt mixtures, indicating a consistent ranking across these two measures, the AIMS results did not align with the asphalt mixtures results. For instance, AIMS ranked LS-1 as having superior morphological properties, followed by LS-2 and then LS-3. However, when these aggregates are part of an asphalt mixture, the friction testing results rank LS-1 highest, with LS-3 coming in second and LS-2 in third place, highlighting that AIMS might mislead the surface texture results when testing aggregates of different surface colors, leading to erroneous aggregate rankings, specifically on texture ( 16 , 52 ).

Random Forest Analysis (RFA)

A parameter importance analysis was conducted using RFA to identify the aggregate friction-related property that most accurately depicts the friction loss in asphalt mixtures, as shown in Figure 18 ( 53 ). The analysis focused on specific aggregate properties, excluding others to avoid multi-collinearity because of strong correlations ( 54 ). The analysis revealed that the number of TWPD cycles, signifying traffic-induced polishing action, was the most important parameter. Among the aggregate friction-related properties, the DFT values at 105 min MDT ((DFT₂₀)_AMD105), followed by loss of sphericality and sphericality index at 105 min MDT were identified as significant predictors of friction loss.

Figure 18.

Feature importance analysis.

Conclusions

This study investigated the aggregate morphology and friction properties at various MDT run times. AIMS was utilized to measure the change in aggregate shape with MDT polishing condition. In contrast, BPT and DFT were employed to measure the aggregate friction properties. Additionally, two statistical analysis methods, ANOVA and PCC matrix, were employed to evaluate whether test results significantly differ under various conditions and to explore the correlation between measured aggregate friction-related properties. In conclusion, the following are the key findings of this study:

Statistical analysis indicated that the MDT cycles had a significant effect on the MDT results. However, the significance of aggregate grading differed among the various sources.

FTIR analysis can be used to identify the origin of the RAP material source. However, further analysis is needed to create a library for different aggregate sources and validate these findings.

The change of TX is the primary cause of mass loss, indicated by MDL, followed by loss of GA. However, the loss of sphericity had an ancillary influence.

The aggregate morphology and friction properties decreased significantly with MDT, indicating that a higher MDL indicates lower resistance to polishing and degradation and, thus, low long-term skid resistance.

MDT, while not a direct measure of aggregate friction properties, can be an indicative technique for assessing the potential loss of aggregate friction-related characteristics and morphology, offering insights into asphalt pavement’s long-term frictional performance.

While AIMS results indicated that traprock (maroon-colored) exhibited the highest surface texture, DFT and BPT results showed that certain limestones had superior friction properties to traprock. This highlights that AIMS measurements may be misleading when testing aggregates of different surface colors, potentially resulting in erroneous texture ranking.

There were linear correlations between BPT data and DFT values at 20, 40, and 60 km/h, with overall R² values of 0.875, 0.884, and 0.91, respectively.

As measured by DFT and BPT, the aggregate friction properties proved to be superior indicators of asphalt pavement friction loss than aggregate morphology parameters quantified by the AIMS technique.

Testing results revealed that the RAP materials were comparable to, or even outperformed, certain limestone sources; however, because of potential variability within RAP stockpiles, careful quantification is necessary to assess their suitability.

RFA identified that the loss of aggregate DFT₂₀ values as the most significant material property representing friction loss in asphalt mixtures, surpassing the aggregate ranking accuracy of AIMS indices.

Finally, this study focused on evaluating the influence of the polishing process on the aggregate morphology and friction properties. However, future research is needed to consider more aggregate sources to verify the aggregate friction-related property that most accurately depicts the friction loss in asphalt mixtures. This will aid in developing the ideal laboratory testing program for long-term friction properties in asphalt mixtures and selecting the optimal aggregate blend for the desired pavement friction performance.

Footnotes

Acknowledgements

The authors thank the Missouri Department of Transportation (MoDOT) for providing technical support and funding for this study. The authors also thank Jaxen McNeese for his assistance with sample preparation.

Author Contributions

The authors confirm contribution to the paper as follows: study conception and design: A. S. El-Ashwah, M. Abdelrahman; testing, analysis, interpretation of results, and draft manuscript preparation: A. S. El-Ashwah. All authors reviewed the results and approved the final version of the manuscript.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded by the Missouri Department of Transportation (MoDOT).

ORCID iDs

Ahmed S. El-Ashwah

Magdy Abdelrahman

References

Hall

J. W.

Smith

K. L.

Titus-Glover

Wambold

J. C.

Yager

T. J.

Rado

Guide for Pavement Friction. Final Report, NCHRP Project 1:43. Transportation Research Board, Washington, D.C., 2009.

Zofka

Marasteanu

Clyne

T. R.

Evaluation of Field Aging Effects on Asphalt Binder Properties. Road Materials and Pavement Design, Vol. 7, 2006, pp. 57–73. https://doi.org/10.1080/14680629.2006.9690058.

Abadie

Laboratory and Field Evaluation of Asphalt Pavement Surface Friction Resistance. Frontiers of Structural and Civil Engineering, Vol. 12, 2018, pp. 372–381. https://doi.org/10.1007/s11709-017-0463-1.

AASHTO Standard T326. Standard Method of Test for Uncompacted Void Content of Coarse Aggregate (As Influenced by Particle Shape, Surface Texture, and Grading). American Association of State Highway and Transportation Officials, Washington, D.C., 2022.

ASTM Standard D5821-13. Standard Test Method for Determining the Percentage of Fractured Particles in Coarse Aggregate. ASTM International, West Conshohocken, PA, 2017.

Masad

Luce

Mahmoud

Implementation of AIMS in Measuring Aggregate Resistance to Polishing, Abrasion and Breakage. Report No. FHWA/TX-06/5-1707-03-1, Texas Department of Transportation and the Federal Highway Administration, Austin, TX, 2006.

Al-Rousan

Masad

Tutumluer

Pan

Evaluation of Image Analysis Techniques for Quantifying Aggregate Shape Characteristics. Construction and Building Materials, Vol. 21, 2007, pp. 978–990. https://doi.org/10.1016/j.conbuildmat.2006.03.005.

Enad

Leslie

Eyad

Erdoğan

Garboczi

Comprehensive Evaluation of AIMS Texture, Angularity, and Dimension Measurements. Journal of Materials in Civil Engineering, Vol. 22, 2010, pp. 369–379. https://doi.org/10.1061/(ASCE)MT.1943-5533.0000033.

Bessa

I. S.

Branco

V. T. F. C.

Soares

J. B.

Neto

J. A. N.

Aggregate Shape Properties and Their Influence on the Behavior of Hot-Mix Asphalt. Journal of Materials in Civil Engineering, Vol. 27, 2015, p. 04014212. https://doi.org/10.1061/(ASCE)MT.1943-5533.0001181.

10.

Masad

Fletcher

Aggregate Imaging System (AIMS): Basics and Applications. Texas Transportation Institute, Texas A&M University System, College Station, TX, 2005

11.

Masad

Luce

Mahmoud

Chowdhury

Relationship of Aggregate Texture to Asphalt Pavement Skid Resistance Using Image Analysis of Aggregate Shape. Final Report, Highway IDEA Project 114: 2007, NCHRP Highway-IDEA Program Officer Transportation Research Board National Research Council, Washington, D.C., 2007.

12.

Kassem

Masad

Awed

Little

Laboratory Evaluation of Friction Loss and Compactability of Asphalt Mixtures. Report No. SWUTC/12/476660-00025-1, U.S. Department of Transportation, University Transportation Centers Program, College Station, TX, 2012.

13.

Chowdhury

Kassem

Aldagari

Masad

Validation of Asphalt Mixture Pavement Skid Prediction Model and Development of Skid Prediction Model for Surface Treatments. Report No. FHWA/TX-17/0-6746-01-1. Texas Department of Transportation and the Federal Highway Administration, College Station, TX, 2016.

14.

Aldagari

Al-Assi

Kassem

Chowdhury

Masad

Development of Predictive Models for Skid Resistance of Asphalt Pavements and Seal Coat. International Journal of Pavement Engineering, Vol. 23, 2022, pp. 695–707. https://doi.org/10.1080/10298436.2020.1766685.

15.

Izzo

Measuring Microtexture of Aggregates Using the Dynamic Friction Tester (DFT). Virtual Transportation Short Course, TX, 2020.

16.

Fernando

E. G.

Arambula

Crockford

Laboratory Evaluation of Aggregate Friction and Texture Properties. Transportation Research Record: Journal of the Transportation Research Board, 2021. 2675: 1396–1405.

17.

Saghafi

Abdallah

I. N.

Nazarian

Practical Specimen Preparation and Testing Protocol for Evaluation of Friction Performance of Asphalt Pavement Aggregates with Three-Wheel Polishing Device. Journal of Materials in Civil Engineering, Vol. 34, 2022, p. 04021397. https://doi.org/10.1061/(ASCE)MT.1943-5533.0004018

18.

Zhou

Scullion

Fernando

Souliman

M. I.

Quantifying the Impact of Reclaimed Asphalt Pavement on the Skid Resistance of Surface Mixtures. Transportation Research Record: Journal of the Transportation Research Board, 2023. 2677: 03611981231166387.

19.

Heitzman

Moore

Evaluation of Laboratory Friction Performance of Aggregate for High Friction Surface Treatments. NCAT Report No. 17-01. National Center for Asphalt Technology at Auburn University, Auburn, AL, 2015.

20.

Wang

Chen

Yin

Oeser

Steinauer

Influence of Different Polishing Conditions on the Skid Resistance Development of Asphalt Surface. Wear, Vol. 308, 2013, pp. 71–78. https://doi.org/10.1016/j.wear.2013.09.013.

21.

Wang

Chen

Xie

Stanjek

Oeser

Steinauer

A Study of the Laboratory Polishing Behavior of Granite as Road Surfacing Aggregate. Construction and Building Materials, Vol. 89, 2015, pp. 25–35. https://doi.org/10.1016/j.conbuildmat.2015.04.032.

22.

Wang

Liu

Kollmann

Oeser

Evaluation of the Polishing Resistance Characteristics of Fine and Coarse Aggregate for Asphalt Pavement Using Wehner/Schulze Test. Construction and Building Materials, Vol. 163, 2018, pp. 742–750. https://doi.org/10.1016/j.conbuildmat.2017.12.147.

23.

ASTM Standard E1911-19. Standard Test Method for Measuring Surface Frictional Properties Using the Dynamic Friction Tester. ASTM International, West Conshohocken, PA. 2019.

24.

El-Ashwah

A. S.

Broaddus

Abdelrahman

Predicting the Friction Coefficient of High-Friction Surface Treatment Application Aggregates Using the Aggregates’ Characteristics. Journal of Materials in Civil Engineering, Vol. 35, 2023, p. 04023089. https://doi.org/10.1061/(ASCE)MT.1943-5533.0004739.

25.

Forster

S. W.

Pavement Microtexture and Its Relation to Skid Resistance. Transportation Research Record: Journal of the Transportation Research Board, 1989. 1215: 151–164.

26.

Won

M. C.

C. N.

Evaluation of Laboratory Procedures for Aggregate Polish Test. Transportation Research Record: Journal of the Transportation Research Board, 1996. 1547: 23–28.

27.

Perry

M. J.

Woodside

A. R.

H. Woodward. Observations on Aspects of Skid-Resistance of Greywacke Aggregate. Quarterly Journal of Engineering Geology and Hydrogeology, Vol. 34, 2001, pp. 347–352.

28.

Fwa

T. F.

Choo

Y. S.

Yurong

Effect of Aggregate Spacing on Skid Resistance of Asphalt Pavement. Journal of Transportation Engineering, Vol. 129, 2003, pp. 420–426. https://doi.org/10.1061/(ASCE)0733-947X(2003)129:4(420).

29.

Liu

Fwa

Choo

Effect of Surface Macrotexture on Skid Resistance Measurements by the British Pendulum Test. Journal of Testing and Evaluation, Vol. 32, 2004, pp. 304–309.

30.

Kandhal

P. S.

Parker

Jr. Bishara

E. A.

Evaluation of Alabama Limestone Aggregates for Asphalt Wearing Courses. Transportation Research Record: Journal of the Transportation Research Board, 1993. 1418: 12–21.

31.

Mahmoud

E. M.

Development of Experimental Methods for the Evaluation of Aggregate Resistance to Polishing, Abrasion, and Breakage. Doctoral Dissertation, Texas A&M University, College Station, TX, 2007.

32.

Masad

Rezaei

Chowdhury

Harris

Predicting Asphalt Mixture Skid Resistance Based On Aggregate Characteristics. Report No. FHWA/TX-09/0-5627-1. Texas Transportation Institute, The Texas A&M University System, College Station, TX, 2009.

33.

Mahmoud

Masad

Experimental Methods for the Evaluation of Aggregate Resistance to Polishing, Abrasion, and Breakage. Journal of Materials in Civil Engineering, Vol. 19, 2007, pp. 977–985. https://doi.org/10.1061/(ASCE)0899-1561(2007)19:11(977).

34.

Masad

Kassem

Little

Characterization of Asphalt Pavement Materials in the State of Qatar. Road Materials and Pavement Design, Vol. 12, 2011, pp. 739–765. https://doi.org/10.1080/14680629.2011.9713893.

35.

Allam

L. M.

Arya

Comparative Analysis of Idaho and Micro-Deval Aggregate Degradation Test Methods. Journal of Materials in Civil Engineering, Vol. 26, 2014, pp. 198–201. https://doi.org/10.1061/(ASCE)MT.1943-5533.0000771.

36.

Wang

Liu

Schulze

Oeser

Development of Morphological Properties of Road Surfacing Aggregates During the Polishing Process. International Journal of Pavement Engineering, Vol. 18, 2017, pp. 367–380. https://doi.org/10.1080/10298436.2015.1088153.

37.

ASTM Standard D6928-17. Standard Test Method for Resistance of Coarse Aggregate to Degradation by Abrasion in the Micro-Deval Apparatus. ASTM International, West Conshohocken, PA, 2017.

38.

ASTM Standard D7428-15. Standard Test Method for Resistance of Fine Aggregate to Degradation by Abrasion in the Micro-Deval Apparatus. ASTM International, West Conshohocken, PA, 2017.

39.

Parker

Kandhal

P. S.

Aggregate Toughness/Abrasion Resistance and Durability/Soundness Tests Related to Asphalt Concrete Performance in Pavements. Transportation Research Record: Journal of the Transportation Research Board, 1998. 1638: 85–93.

40.

Lang

A. P.

Range

P. H.

Fowler

D. W.

Allen

J. J.

Prediction of Coarse Aggregate Performance by Micro-Deval and Other Soundness, Strength, and Intrinsic Particle Property Tests. Transportation Research Record: Journal of the Transportation Research Board, 2007. 2026: 3–8.

41.

Kane

Artamendi

Scarpas

Long-Term Skid Resistance of Asphalt Surfacings: Correlation Between Wehner–Schulze Friction Values and the Mineralogical Composition of the Aggregates. Wear, Vol. 303, 2013, pp. 235–243. https://doi.org/10.1016/j.wear.2013.03.022.

42.

Geological Survey Program. Traprock–PUB2907. Missouri Department of Natural Resources, 2020. https://dnr.mo.gov/document-search/traprock-pub2907/pub2907

43.

Geological Survey Program. Limestone–PUB2902. Missouri Department of Natural Resources, 2020. https://dnr.mo.gov/document-search/limestone-pub2902/pub2902

44.

Buttlar

W. G.

Abdelrahman

Rath

Meister

Gettu

El-Ashwah

A. S.

Evaluation of Recycled Components in Stone Matrix Asphalt Mixes. Report No. cmr 23-016. Missouri Department of Transportation, Jefferson City, MO, 2023.

45.

ASTM Standard D3319-23. Standard Practice for the Accelerated Polishing of Aggregates Using the British Wheel. ASTM International, West Conshohocken, PA, 2023.

46.

AASHTO Standard PP104-21. Sample Preparation and Polishing of Asphalt Mixture Specimens for Dynamic Friction Testing. American Association of State Highway and Transportation Officials, Washington, D.C., 2021.

47.

ASTM Standard D3042. Standard Test Method for Insoluble Residue in Carbonate Aggregates. ASTM International, West Conshohocken, PA, 2017.

48.

Masad

Rezaei

Chowdhury

Field Evaluation of Asphalt Mixture Skid Resistance and Its Relationship to Aggregate Characteristics. Texas Transportation Institute, College Station, TX, 2011.

49.

ASTM Standard E303-22. Standard Test Method for Measuring Surface Frictional Properties Using the BRITISH Pendulum Tester. ASTM International, West Conshohocken, PA, 2022.

50.

El-Ashwah

A. S.

Abdelrahman

Comparative analysis of empirical decomposition algorithms in predicting tire-pavement friction from asphalt surface textures: A Hilbert Huang transform (HHT) analysis. Discover Civil Engineering, Vol. 1, 2024, p. 89. https://doi.org/10.1007/s44290-024-00076-w

51.

El-Ashwah

A. S.

Abdelrahman

Analysis of the Frictional Performance Development of Recycled Stone Matrix Asphalt (SMA) Mixtures. Innovative Infrastructure Solutions, Vol. 9, 2024, p. 394. https://doi.org/10.1007/s41062-024-01696-y.

52.

Azzam

Khasawneh

M. A.

Al-Omari

A. A.

Masad

Kassem

A Statistical Approach for Predicting Skid Resistance of Asphalt Pavements. International Journal of Pavement Research and Technology, Vol. 14, 2021, pp. 647–654. https://doi.org/10.1007/s42947-020-0250-6.

53.

Breiman

Random Forests. Machine Learning, Vol. 45, No. 1, 2001, pp. 5–32. https://doi.org/10.1023/A:1010933404324.

54.

El-Ashwah

A. S.

Abdelrahman

Relating Aggregate Friction Properties to Asphalt Pavement Friction Loss Through Laboratory Testing, Statistical Analysis, and Machine Learning Insights. International Journal of Pavement Engineering, Vol. 26, No. 1, 2025, p. 2456739. https://doi.org/10.1080/10298436.2025.2456739.