Investigation of Using Crushed Glass Waste as Filler Replacement in Hot Asphalt Mixtures

Abstract

Due to the depletion of natural aggregates and high maintenance cost of highway systems, developing sustainable asphalt concrete (AC) mixes that use waste materials instead of virgin raw materials is necessary. A large amount of waste glass material is globally generated per year that could be beneficial to sustain the asphalt industry. In this context, the present paper evaluates the properties and performance of AC mixtures that utilize crushed waste glass as a replacement material of filler aggregates. Three AC mixes with percentages of filler replacement in the range from 0%, 25%, to 50% were fabricated. Complex modulus testing was performed to evaluate the dynamic modulus |E*| and phase angle δ over a range of temperatures and loading frequencies. In addition, the flow number (FN) test was conducted to assess the rutting potential of the mixtures. The results showed that the mix containing 25% of crushed glass is likely to better resist fatigue cracking; however, the inclusion of glass in the AC reduced the rutting resistance compared to conventional hot mix asphalt (HMA). Finally, the results of the flow number test and the simple performance indicators were compared and used to rank the mechanical performance of the various mixtures.

1. Literature Review

Asphalt concrete (AC) has two primary components, namely, aggregates and asphalt binder. Aggregates on average range from 94 to 96% of the mixture which implies that alternative aggregates can help to create sustainable mixtures for road construction. Nowadays, numerous recycled and waste materials, such as reclaimed asphalt pavement, RAP, crumb rubber, plastics, and waste glass are being used and/or evaluated as a suitable replacement of virgin aggregates in the construction of pavement layers such as asphalt and base layers [1,2,3,4]. Recyclability and waste utilization not only decrease the harmful effects of waste disposal, but also reduce the depletion of natural resources, resulting in cost savings and economic benefits [5]. In addition, the use of such materials might lead to further enhancement of the performance of asphalt paving materials, thus representing a value-added use for solid waste. For these reasons, many research projects have been dedicated to searching for and assessing the utilization of alternative recycled materials in bituminous mixtures [3,4,6].

When exploring recycling based AC, besides material type, aggregate gradation, aging conditions, frequency of loading and temperature, additional factors need to be investigated. In the case of using waste glass (WG) materials, the effect on asphalt mixture properties (including dynamic modulus |E*| and phase angle Φ) and impact on performance need to be assessed. The resistance of AC to permanent deformation and fatigue cracking are two major performance measures that are typically considered along with durability [7]. Among the various alternatives, permanent deformation potential of AC can be evaluated using compressive uniaxial tests such as flow number (FN) [8], whereas fatigue cracking potential can be evaluated using indirect tensile testing [7,9].

Over the last decade, researchers opted to use crushed waste glass (CWG) in construction applications in order to divert WG from landfills and to reduce the environmental effect in the construction sector. CWG has been studied for use as an aggregate in asphalt subbase and unbound base layers, as well as concrete [10]. Ali and Arulrajah [11] and Chen et al. [12] examined the use of waste glass in the base and sub-base courses of pavements indicating that sufficient shear strength, bearing capacity, and crushing resistance are achieved. Similarly, Saberian et al. indicated that crushed glass had beneficial effects on unconfined compression strength, California bearing ratio, and resilient modulus of the base and subbase layers when combined with crumb rubber [13]. This was attributed to the larger angularity of glass particles having a higher internal angle of friction that resulted in stronger interlocking between aggregate particles.

With regards to the asphalt layer, good performance was reported in asphalt pavements with glass contents of 10% to 15% in surface course mixtures [14]. In this context, Bachand et al. [15] conducted a complex modulus test to compare the stiffness of HMA with 10% glass and different binder contents to that of a traditional HMA. It was concluded that small variation between all HMA were observed. Furthermore, a maximum aggregate size of 4.75 mm was recommended to control durability impact [16]. Arabani assessed the stiffness modulus of specimens with varying glass (0% to 20%) and hydrated lime levels at 5 °C, 25 °C, and 40 °C [17]. The results showed that the stiffness modulus of the specimens with glass and hydrated lime had an increasing value due to the higher cohesiveness between aggregates and binder produced by the hydrated lime’s anti-stripping properties. However, the stiffness modulus dropped as the glass content approached a predetermined threshold of 15%, defined as the ideal content. Likewise, Shafabakhsh and Sajed found an optimal glass content of 15% above which the stiffness modulus decreased [18]. The decrease in stiffness modulus was attributed to an abundance of glass particles, which can slip on together because of their smooth texture or break into smaller particles under heavy loads at the surface layer of the pavement structure. Moreover, Arabani and Kambooza [19] found that the dynamic modulus of traditional HMA was reduced by 70% compared to 50 % for mixes containing WG materials when the testing temperature was raised to 40 °C. This indicated that HMA containing glass had a lower susceptibility to temperature changes that might be the result of the heat transfer ability of glass in comparison with the stone aggregates. On the other hand, Airey et al. [20] used the indirect tensile strength method to examine specimens that contained 50% glass at 20 °C and found that the stiffness modulus is barely affected when glass is used in place of conventional aggregate. Similar conclusions were reached by other researchers, indicating as well that large proportions of recycled glass in asphalt mixtures may result in undesirable Marshall testing properties, such as a reduction in the mixture’s strength, density, voids filled with binder, and air void content [21].

Past investigations on replacing filler with waste materials explored either partial or full replacements [22,23]. The aggregate portion which passes the No. 200 sieve (75 μm) is termed as filler, which influences the mechanical behavior and durability of asphalt mixtures. Mineral fillers were generally treated as being suspended in the asphalt binder without particle to particle contact and thus without contributing to binder stiffness [24,25]. Additionally, mineral fillers are part of the aggregate skeleton as they provide contact and/or friction between particles [26]. Thus, fillers play an important role in the performance of asphalt pavement mixtures. The filler present in the asphalt mix interacts with asphalt binder to form asphalt mastic. The filler activity in the mastic is due to the physical hardening and chemical interaction [27]. Based on this activity, fillers can be generally classified into two categories known as active fillers and passive (inert) fillers. The fillers which exhibit chemical activity in the mastic due to their alkaline nature and the acidic nature of the binder are termed as active fillers [27]. This chemical reaction is reported to improve the anti-aging potential, adhesion, and high temperature resistance in asphalt mastic and mixes [28]. The fillers such as hydrated lime, cement, steel slag, and so on fall under the category of active fillers. On the other hand, the inert or passive fillers exhibit little to no chemical activity in asphalt mastic but are usually responsible for causing stiffness or physical hardening in the asphalt mastic owing to their physical characteristics. Overall, the performance of asphalt mixes against distress such as permanent deformation, load and non-load dependent cracking, aging, and moisture sensitivity is largely dependent on the physical and chemical characteristics of fillers [29,30]. As a result, the type and quantity of fillers used in flexible pavements are critical to their cost-effectiveness and long-term performance. Cement, stone dust, and lime are the most common traditional fillers assessed in HMA [31,32]. Sebaaly et al. indicated that hydrated lime has the ability to improve the resistance of HMA mixtures to moisture damage, reduce oxidative aging, and enhance the resistance to fatigue and rutting, which led to observed improvements in the field performance of lime-treated HMA pavements [32]. Likewise, Choudhary [29] investigated the effects of different types of filler on the performance of asphalt concrete and found that mixes containing red mud and limestone dust worked well in terms of cracking and rutting deformation, while compositions containing carbide lime performed better in terms of moisture resistance. Moreover, Qassim et al. [33] and Murana et al. [34] investigated the performance of HMA containing Metakaolin as a partial or full replacement of fillers, and found that the addition of different proportions of Metakaolin has a significant influence on the mechanical behavior of the HMA. The Marshall’s stability, flow and, density grow as the Metakaolin content grows until 50%, whereas the indirect tensile strength increases continuously as the Metakaolin content increases up to 100% at 25, 40, and 60 °C.

Previous literature revealed a focus on one side of waste glass utilization as a replacement of fines aggregates in AC, generally limited to percentages between 10% to 15% [16,17,18], but a replacement of mineral fillers solely by waste materials showed improvement to the properties and performance of HMA [31,32,33]. Therefore, this study aims to investigate the utilization of recycled laminated waste glass as a replacement of mineral fillers in Superpave AC mixes from a performance perspective. In the process, the main characteristics of dynamic modulus, (i.e., stiffness), phase angle and the permanent deformation potential of asphalt concrete containing different glass powder percentages (0, 25, and 50%) are evaluated. Additionally, the flow number (FN), dynamic modulus (DM), and simple performance indicators results were compared and used to characterize the mechanical behavior of asphalt mixtures.

2. Materials and Experimental Work

2.1. Materials

The materials used in this research included limestone coarse and fine aggregates, two types of fillers, i.e., limestone (LP) and glass powder filler (GP); and a PG 64-16 performance grade binder. The aggregate characteristics were evaluated against the Superpave criteria as presented in

Table 1. Limestone aggregates characteristics vs. Superpave criteria.

Traffic, (Millions of Equivalent Single Axle Loads, Esals)	Property	ASTM	Superpave Criteria	Sample 1	Sample 2	Average	Status
10 to <30	Coarse Angularity	D5821-13	Min of 85%	98.46%	97.44%	97.95%	Passing
10 to <30	Fine Angularity	C1252	Min of 45 %	50	52	51	Passing
10 to <30	Flat & Elongated	D4791	Max of 10%	2%	2.40%	2.20%	Passing
10 to <30	Toughness	C131	Max of 40%	19.42%	20.66%	20.04%	Passing
10 to <30	Surface Texture (Particle Index)	D3398	rounded, smooth: 6 or 7 rough, angular:14–20 or more	13.9	14.15	14.025	Rough, angular

.

In addition, the specific gravity tests for aggregates and mineral fillers were performed in accordance to the standards indicated in

Table 2. Specific Gravity of Aggregates and Glass filler.

Property	Value	Standard
Specific gravity of limestone Coarse aggregate	2.604	ASTM C 127
Specific gravity of limestone Fine aggregate	2.591	ASTM C 128
Specific gravity of limestone filler aggregate	2.563	ASTM C 854
Specific gravity glass cullet filler	2.331	ASTM C 854

.

2.2. Preparation of Asphalt Mixtures

In this study, three AC mixtures were assessed having an aggregate gradation with a nominal maximum aggregate size (NMAS) of 12.5 mm as illustrated in Figure 1.

The corresponding particle size distribution presented in

Table 3. Particle size distribution used for AC mix design.

Sieve Size (mm)	% Passing	Superpave Control Points
		Min	Max
19	100	100	100
12.5	91	90	100
9.5	70
4.75	50	44	74
2.36	34.4	28	58
1.18	22
0.6	15
0.3	12	5	21
0.15	5.26
0.075	3	2	10

shows that the selected gradation respects the Superpave control point’s criteria for a 12.5 mm designation [35].

The Labels “LP”,“GP25”,”GP50” are used in the upcoming sections to designate HMA mix with limestone filler, or 25% and 50% replacement of filler with glass powder, respectively.

Mix Design and Compaction

As part of the mix design, at each asphalt content, three specimens of 4700 g each, mixed at 160 °C and compacted to Ndes at 150 °C, and three loose samples of 2500 g were prepared and evaluated. The bulk specific gravity (Gmb) of the compacted specimens and the theoretical maximum specific gravity (Gmm) of the loose ones were measured using ASTM D2726 and ASTM D2041, respectively. Then, the optimum asphalt content at 4 % air voids was determined and the volumetric properties of the mix were checked according to the Superpave criteria for a 12.5 mm designation.

The Superpave gyratory compactor SGC was used to compact Gmb specimens with an effort corresponding to a case of medium to high traffic with 3 to 30 million equivalent single axle loads (ESALs). This is represented by the following compaction parameters: Initial number of gyrations (Nini) = 8, Design number of gyrations (Ndes) = 100, Maximum number of gyrations (Nmax) = 160. All the compacted specimens were compacted in a mold of 150.0 mm diameter with a height of 110.0–120.0 mm. It should be noted that to imitate the short-term aging of asphalt concrete during construction, the loose mixtures were conditioned for two hours at the compaction temperature. The guidelines from NCHRP Report 691 guided the selection of this aging protocol [36].

A summary of the mix design parameters for all mixtures is shown in

Table 4. Volumetric results of LP and GP mixes.

Mix	LP	GP25	GP50
OBC	4.63	4.1	4.15
Va %	4	4	4
VMA %	13.52	13.63	13.26
VFA%	68.05	71.2	69.8
% Gmm at Nini	85.77	85.56	85.14
% Gmm at Nmax	97.43	96.99	97.48
% Dust Proportion (DP)	0.7	0.71	0.7
Mixing Temperature °C	160	160	160
Compaction Temperature °C	150	150	150

.

The results of the mix design showed that the volumetric properties for LP and GP mixes are comparable, while the optimum binder content to achieve air voids volume Va of 4% varied between the three mixes. Mixes containing glass needed less binder than the control HMA.

For performance testing specimens, AC specimens were firstly compacted to a height of 175 mm and a diameter of 150 mm, then cored to 100 mm diameter and sawed to 150 mm height. Next, specimens were checked to verify that air voids (VA) are in the range of 7.0 ± 0.5%. In order to obtain the desired VA of 7.0 ± 0.5% in the cored and sawed specimens, trials on several weights of the loose mixtures were executed until the required weight of 6920 g was determined. Finally, for each AC mixture, 2 cylindrical specimens were prepared for (FN), |E*|, and phase angle δ testing.

2.3. Testing

2.3.1. Complex Modulus E* test

Asphalt concrete (AC) is a viscoelastic material. A material’s viscoelasticity is the property that shows both viscous and elastic qualities when deformed. To assess the reaction of asphalt mixes to in-service traffic and thermal loadings, the complex modulus test (E*) is conducted to characterize the stiffness measured in terms of the dynamic modulus |E*| and viscoelastic behavior exhibited by the phase angle (δ). The National Cooperative Highway Research Program NCHRP report 547 of the NCHRP Project 9-19 suggested |E*| as one of the most viable simple performance tests for rutting and fatigue cracking criteria [37]. Higher |E*| at small, mid, and large frequencies is associated with better rutting resistance, worst fatigue, and worst thermal cracking resistance [8].

(E*) testing involves applying a repetitive sinusoidal dynamic compressive axial load (stress) to cylindrical (AC) samples. The stress amplitude is programmed so that it produces a recoverable strain not exceeding a maximum amplitude of 70 microstrains to maintain the material response in the linear viscoelastic range [38]. In this test, two replicates of each AC were examined at three different temperatures (5, 20, and 40 °C) and six different loading frequencies (20, 10, 5, 1, 0.5, and 0.1 Hz) in accordance with AAASHTO T 342-11. At each temperature/frequency combination, the average of the measured |E*| and δ of the two replicates were computed. Then, using the time temperature superposition (TTS) principle, the dynamic modulus and phase angle master curves were constructed at a reference temperature of 20 °C.

Dynamic Modulus Master Curve

The sigmoidal model of Equation (1) was used to create the dynamic modulus |E*| master curves as a function of the reduced frequency (${\mathrm{f}}_{\mathrm{r}}$), which in turn is a function of the actual frequency f and temperature T as seen in Equations (2) and (3) [39].

$$\mathrm{Log}\left(\left|\mathrm{E}*\right|\right)=\mathsf{\delta }+\frac{\mathsf{\alpha }}{1+{\mathrm{e}}^{\mathsf{\beta }+\mathsf{\gamma }\log \left({\mathrm{f}}_{\mathrm{r}}\right)}}$$

(1)

$${\log \mathrm{f}}_{\mathrm{r}}=\mathrm{logf}+\log ({\mathrm{a}}_{\mathrm{T}})$$

(2)

$$\mathrm{Log}\left({\mathrm{a}}_{\mathrm{T}}\right)={\mathrm{a}}_1{\mathrm{T}}^2+{\mathrm{a}}_2\mathrm{T}+{\mathrm{a}}_3$$

(3)

where δ, α, β, and γ are fitting parameters determined by numerical minimization of measured and fitted log |E*| values.$\delta$,$\mathsf{\alpha },\left(\mathsf{\beta }\&\mathsf{\gamma }\right)$ stand for the minimum value of |E*|, the vertical span of the modulus function, and the shape parameters, while ${\mathrm{a}}_1$, ${\mathrm{a}}_2$, and ${\mathrm{a}}_3$ stand for the polynomial model’s regression coefficients that were applied to fit the shift factor function log $({\mathrm{a}}_{\mathrm{T}}$) at any given temperature.

Figure 2 shows an example of developing the master curve for the “GP 25” by shifting measured |E*| values horizontally over the frequency axis. For analysis purposes, logarithmic and semi-logarithmic scales are usually used since in the first the variance between |E*| of AC mixes appears at smaller frequencies, while in the second, the variance appears more for higher values of frequencies.

Phase Angle Master Curves

The delay between the imposed peak stress and the peak strain response under cycling loading is defined as the phase angle δ [8]. Stress and strain are always in phase for completely elastic materials, but a phase difference is exhibited between stress and strain for viscous materials. The phase angle is utilized to determine the complex modulus components: storage and loss modulus; δ offers clear insight into the AC mixture’s degree of viscosity and/or elastic condition. Lower AC elasticity and more viscous behavior result from higher observed phase angles and vice versa. For an entirely elastic material, δ = 0°, whereas, for an entirely viscous material, δ = 90°.

To develop the master curve of δ, a method similar to that of |E*| master curve can be used. In this context, and since AC is a thermorheologically simple TRS material [40], values measured at a specific temperature can be adjusted horizontally to the reference temperature using the same shift factors found while constructing the |E*| master curve. Figure 3 illustrates a typical δ versus ${\mathrm{f}}_{\mathrm{r}}$ graph for the control “LP” mix at the reference temperature of 20 °C.

It is demonstrated that the shifted data of δ resembles a single smooth curve, proving the validity of using the shift factors from |E*| to form δ master curve. There are not many models in the literature that are designed to fit δ master curves. Of these, the approximation model developed by Booji and Thoone [41] and its modified version suggested by Yang and You [42] presented by Equations (4) and (5), respectively, are the most popular.

$$\mathsf{\delta }\left({\mathrm{f}}_{\mathrm{r}}\right)=-90\mathsf{\alpha }\mathsf{\gamma }\frac{{\mathrm{e}}^{\mathsf{\beta }+\mathsf{\gamma }\log \left({\mathrm{f}}_{\mathrm{r}}\right)}}{{(1+{\mathrm{e}}^{\mathsf{\beta }+\mathsf{\gamma }\log \left({\mathrm{f}}_{\mathrm{r}}\right)})}^2}$$

(4)

$$\mathsf{\delta }\left({\mathrm{f}}_{\mathrm{r}}\right)=-\mathrm{c}·90\mathsf{\alpha }\mathsf{\gamma }\frac{{\mathrm{e}}^{\mathsf{\beta }+\mathsf{\gamma }\log \left({\mathrm{f}}_{\mathrm{r}}\right)}}{(1+{\mathrm{e}}^{\mathsf{\beta }+\mathsf{\gamma }\log \left({\mathrm{f}}_{\mathrm{r}}\right)})²}$$

(5)

where $\mathsf{\delta }\left({\mathrm{f}}_{\mathrm{r}}\right)$ is the phase angle in degrees, ${\mathrm{f}}_{\mathrm{r}}$ is the reduced frequency in Hz, $\mathsf{\alpha },\mathsf{\beta },\mathsf{\gamma }$ are the regression parameters found in the sigmoidal function of |E*| expressed in Equation (1), and c is a fitting parameter found by numerical minimization between measured and predicted values of δ.

As seen in Figure 4, the δ master curves fitted using the aforementioned models lacked precision. Therefore, another adjustment was proposed to the Yang and Lou model as expressed in Equation (6).

$$\begin{array}{c}\mathsf{\delta }\left(\mathrm{fr}\right)={\mathrm{k}}_1{[\log \left({\mathrm{f}}_{\mathrm{r}}\right)]}^3+{\mathrm{k}}_2{\left[\log \left({\mathrm{f}}_{\mathrm{r}}\right)\right]}^2+{\mathrm{k}}_3[\log \left({\mathrm{f}}_{\mathrm{r}}\right)]\\ -{\mathrm{k}}_4·90·\mathsf{\alpha }\mathsf{\gamma }\frac{{\mathrm{e}}^{\mathsf{\beta }+\mathsf{\gamma }\log \left({\mathrm{f}}_{\mathrm{r}}\right)}}{\left(1+{\mathrm{e}}^{\mathsf{\beta }+\mathsf{\gamma }\log \left({\mathrm{f}}_{\mathrm{r}}\right)}\right)²}\end{array}$$

(6)

where ${\mathrm{k}}_1,{\mathrm{k}}_2,{\mathrm{k}}_3,{\mathrm{k}}_4$ are constant regression parameters.

It can be shown that the proposed model predicted accurately δ master curves of both the standard HMA and the mix containing glass filler. Since good fits are obtained, the phase angle master curves of various mixes of the study fitted using Equation (6) were used to analyze the effect of replacing limestone filler by the glass powder on the phase angle of standard HMA.

2.3.2. Flow Number (FN) Test

To evaluate the resistance to permanent deformation of asphalt mixes, the NCHRP 9-19 Project developed the flow number test, a simplified performance test (dynamic creep and recovery test) [8]. In this test, a haversine compressive cyclic loading having a maximum load pulse of 0.1 second long followed by a 0.9 second rest period is applied on cylindrical AC specimens similar to those for |E*| testing. The accumulation of permanent strains is measured as a function of loading cycles. For the purposes of this investigation, the flow number testing was carried out in accordance with AASHTO T378-17 utilizing a maximal stress amplitude of 600 kPa per cycle at a temperature of 53 °C to replicate the conditions of the pavement surface in hot climate regions. The test was programmed to end after 10,000 loading cycles or when there was an accumulation of 7% permanent strain, whichever occurred first.

Figure 5 represents the typical plot of permanent strain accumulation under a specific combination of material, load repetitions, and environmental conditions. The volume of the mixture decreases and becomes denser in the primary zone, and the accumulated strain begins to increase fast. The secondary zone is a transition zone between the primary and tertiary zones, during which the initial rutting has finished and the strain slope has decreased. Finally, in the tertiary zone, shear deformation begins and rutting rises rapidly.

Model Used to Identify the Flow Number

The flow number, FN, is the quantity of load cycles required for AC to reach the flow point, or the beginning of the tertiary zone. The constant FN, stated in cycles, represents the location where the slope of the permanent strain curve is at its lowest value. Researchers at Arizona State University [43] suggested a reliable and logical method for determining FN in which the permanent axial strain is first fitted using the Franken model represented by Equation (7) below.

$${\mathsf{\epsilon }}_{\mathrm{p}}=\mathrm{A}{\mathrm{n}}^{\mathrm{B}}+\mathrm{C}({\mathrm{e}}^{\mathrm{Dn}}-1)$$

(7)

where ${\mathsf{\epsilon }}_{\mathrm{p}}$ is the permanent axial strain, n is the number of cycles, and A, B, C, and D are regression coefficients.

A, B, C, and D can be determined by a numerical minimization step between experimental and predicted values of strains per cycle. Once the Franken model is developed, the first and second derivative of Equation (7) are computed as shown in Equations (8) and (9), respectively, and FN is determined as the cycle where the second derivative changes from negative to positive sign.

$$\frac{{\mathrm{d}}_{{\mathsf{\epsilon }}_{\mathrm{p}}}}{{\mathrm{d}}_{\mathrm{n}}}=\mathrm{A}{\mathrm{Bn}}^{\left(\mathrm{B}-1\right)}+\mathrm{CD}{\mathrm{e}}^{\mathrm{Dn}}$$

(8)

$$\frac{\mathrm{d}{²}_{{\mathsf{\epsilon }}_{\mathrm{p}}}}{{\mathrm{d}}_{\mathrm{n}}²}=\mathrm{AB}(\mathrm{B}-1){\mathrm{n}}^{\left(\mathrm{B}-2\right)}+\mathrm{C}{\mathrm{D}}^2{\mathrm{e}}^{\mathrm{Dn}}$$

(9)

Figure 6 shows a curve fitting example for the measured permanent axial strain using the Franken model for one replicate of the “GP 50" mixture, and the flow number FN located at 572 cycles.

Although the flow number is frequently used to predict rutting resistance [8,44], the FN index according to Zhang et al. [45] offers a more accurate assessment of the rutting performance of asphalt mixtures since it takes into account both the stresses and the quantity of load repetitions endured before reaching the tertiary flow. The FN index is expressed by Equation (10) below. For the purpose of performance analysis, higher values of FN indicate better rutting resistance whereas higher FN index values indicate a worse rutting resistance for the AC mixture.

$$\mathrm{FN}\mathrm{Index}=\frac{{\mathsf{\epsilon }}_{\mathrm{p}}}{\mathrm{FN}}$$

(10)

3. Results and Analysis

3.1. Master Curves Parameters

Table 5. Summary of master curve (MC) fitting parameters and statistics for various AC mixtures.

	|E*| MC Parameters					Shift factor Function Parameters				$\mathsf{\delta } \mathbf{M}\mathbf{C} \mathbf{Parameters}$
Mix Type	δ	α	β	γ	R²	a	b	c	R²	k1	k2	k3	k4	R²
LP	−0.295	4.746	−1.706	−0.426	0.997	0.001	−0.175	3.121	1.000	0.191	−0.831	0.851	1.135	0.994
GP 25	0.808	3.484	−1.422	−0.482	0.992	0.001	−0.207	3.593	1.000	0.028	−0.708	2.838	1.284	0.992
GP50	1.400	2.868	−1.318	−0.558	0.993	0.000	−0.136	2.703	1.000	−0.006	−0.034	1.223	1.110	0.996

shows a summary of the fitting parameters related to |E*|, shift factors, and δ for the study mixtures. As seen, excellent curve fitting is obtained for all mixes, according to the R² statistics (above 0.99), for experimental E*| and δ data by the constructed master curves utilizing the sigmoidal model given in Equation (1) and the proposed phase angle model of Equation (6).

3.2. Dynamic Modulus Results

Figure 7 and Figure 8 show the |E*| master curves for the HMA with conventional filler, “LP,” and those containing glass powder “GP” in the log-log and semi-log scale, respectively.

Generally, the dynamic modulus master curves reveal a reduction in the stiffness of the control HMA “LP” when limestone filler is replaced by the glass powder for both 25% and 50% percentages over the full range of frequencies considered. This implies that the use of glass powder at different contents as replacement of filler aggregates affects the properties of HMA. The reduction in |E*| seems to be smaller for a replacement percentage of 50% instead of 25%.

Due to the influence that |E*| has on asphalt performance [8], it is worthwhile to segment the full (|E|) master curve’s reduced frequency range into three zones where asphalt concrete exhibits various behaviors. Zone 1 includes the lower frequency range below 1 Hz and is characterized by soft asphalt mixture that is primarily rutted [46,47]. This typically reflects high temperatures and slower traffic loads. A higher (|E*|) is desired within this zone to minimize the rutting potential of the mixture. Zone 2, which covers the reduced frequency range between 1 and ${10}^3$ Hz, is characterized by an asphalt mixture that is moderately stiff and mostly impacted by fatigue cracking [46,47]. This normally reflects ordinary traffic speeds and moderate weather. Zone 3 (encompassing the range of frequencies greater than ${10}^3$ Hz) is characterized by an extremely stiff asphalt mixture that is primarily susceptible to thermal cracking and fatigue [46,47]. A smaller (|E*|) is desired within zones 2 and 3 to minimize the cracking susceptibility of the mixture.

Figure 9 illustrates the extent of this reduction in function of the reduced frequency which in turn is a representation of a wide range of combinations of loading conditions (temperature and frequency of loading). Overall, it can be seen that more reduction is occurring for increased values of reduced frequency, and this reduction is remarkably noticeable at higher frequency levels in Zone 3.

In Zone 1, the use of glass powder as 25% and 50% of filler material caused, respectively, a maximal reduction of 21% and 12% in the modulus of the control mix. This implies that mixes with 25 % glass filler may in general be more prone to rutting than mixes containing 50% of glass filler and conventional HMA which share nearly similar values of |E*| for ${\mathrm{f}}_{\mathrm{r}}$ values below 1 Hz. Over the range between 1 and ${10}^3$ Hz, the reduction in the stiffness increased to reach 23% and 18% for “GP25” and “GP50”. Similarly for zone 3, it can be observed that the stiffness reduction is further increasing for “GP25” and “GP 50” reaching a similar extent of 26% at a reduced frequency of ${10}^5$ Hz. Reduced stiffness in zones 2 and 3 indicates that the mixes having glass powder might have better resistance to fatigue and thermal cracking than the control mix “LP” assuming same pavement structure is being used.

Statistical Analysis

A statistical analysis was conducted to test the significance of the observed differences between the average (|E*|) curves of LP and GP mixes over the analyzed range of frequencies. For this purpose, one-way analysis of variance (ANOVA) was performed at each reduced frequency level. The null and alternate hypotheses are as follows:

Null hypothesis ${\mathrm{H}}_0$: Means of |E*| for LP, GP25, and GP50 mixes are equal.

Alternative hypothesis ${\mathrm{H}}_{\mathrm{a}}$: at least one |E*| mean is different.

The analysis was conducted using a significance level α of 5%, thus, a p-value > 0.05 indicates that there is no enough evidence to conclude that the means are not equal at the specified reduced frequency. The ANOVA results presented in

Table 6. ANOVA test results.

fr	1 × 10−4	1 × 10−3	1 × 10−2	1 × 10−1	1 × 100	1 × 10	1 × 102	1 × 103	1 × 104
ANOVA	0.951	0.134	0.094	0.171	0.221	0.144	0.039	0.002	0.0015

indicated that below a reduced frequency of ${10}^2$ Hz, the difference between |E*|of the control “LP” and the mixes containing glass powder is not significant.

For ${\mathrm{f}}_{\mathrm{r}}$ levels where inequality between average |E*| of groups exists (p-value < 0.05), the Tukey HSD test was performed to find |E*| means that are significantly different from each other. The results in

Table 7. Tukey test results.

Mixes Compared	LP vs. G25	LP vs. G50	G25 vs. G50
p-Value at 102	0.032	0.041	0.78
p-Value at 103	0.003	0.005	0.19

confirm that a significant difference exists between |E*| of LP and GP mixes for ${\mathrm{f}}_{\mathrm{r}}$ above ${10}^2$ Hz.

3.3. Phase Angle Results

Figure 10 illustrates the master curves of δ for the different mixes considered. It can be seen that at the middle and high range of frequencies (above 1 Hz), “GP 25” exhibits more viscous behavior than “GP50” which is in turn less elastic (more viscous) than the control mix “LP”. This result is in agreement with the dynamic modulus analysis that indicated a stiffer asphalt for “LP” and “GP50” compared to “GP25” in Zones 2 and 3. In the lower range (below 1 Hz), the peak of the phase angle master curves occurred approximately at the same ${\mathrm{f}}_{\mathrm{r}}$ for “GP 50” and “ GP 25”indicating that the maximum viscous effect occurs in both mixtures at the same temperature or at the same loading time. Additionally, the peaks of δ curves for “GP25” and “GP50” occurred at a higher frequency compared to “LP,” and are higher respectively by 1.6%, and 5.6%. This implies that at peak Φ values the control mixture is less viscous than mixes using glass powder, which correlates with the highest |E*| (stiffness) values found for “LP” in the lower frequency range.

3.4. Flow Number Results

Table 8. Summary of flow number (FN) test results.

Mix	FN	$\mathbf{Permanent}\mathbf{Strain}{\mathsf{\epsilon }}_{\mathbf{p}, }\mathbf{Microns}$	FN Index
50%	572	16,014.8	28.0
25%	559	16,091.9	28.8
Control	1362	24,410.7	17.9

shows the results of FN testing for various mixes under consideration and Figure 11 presents a graphical plot comparing both FN and FN index data.

From the analysis of the results, both parameters indicate that mixes containing glass powder are more prone to rutting than standard HMAs. Furthermore, it was found that the replacement of 25% and 50% of filler materials reduced the FN of “LP” by 59% and 58% and increased the FN index by 60.6% and 56.2%. The results indicated that increased glass content provides better rutting performance in relation to the conventional HMA. However, as shown in Figure 11, despite the reduction in FN induced by the glass effect, both “GP25” and “GP50” mixes provided acceptable level of rutting performance with an FN above 190 cycles [48].

3.5. Assessment Using Simple Performance (SPT) Tests

Typically, at high and moderate temperatures, respectively, asphalt mixes are often prone to rutting and fatigue cracking. In this context, Witczak et al. proposed the rutting and fatigue stiffness factors [49] for the Simple Performance Test (SPT). The definition of SPT is "a test method(s) that accurately and reliably measures a mixture response characteristic or parameter that is highly correlated to the occurrence of pavement distress over a diverse range of traffic and climatic conditions" [49,50]. SPT can help to evaluate a mixture’s potential to rutting and fatigue distress under specific conditions, but it cannot anticipate the intensity or evolution of a distress over time. Pellinen and Witczak [50] assessed for a variety of (temperature and frequency) conditions the correlations between the measured rutting in the field and the stiffness factor parameter |E*|/sin δ. It was reported that the coefficient of determination ${\mathrm{R}}^2$ peaked at 54.4 °C and 5 Hz. At peak conditions, |E*|/sin δ had better statistical correlation to rutting than |E*|, which is in agreement with findings of other studies [51,52].

To determine the stiffness factor |E*|/sin δ, i.e. the SPT parameter for rutting evaluation, the |E*| and δ values at 54.4 °C and a frequency of 5 Hz were obtained from every mix type’s respective master curve. The computed stiffness factor values are presented and ranked in

Table 9. Ranking of FN and rutting stiffness factor evaluated at 5 Hz for the LP and GP mixtures.

54.4 °C	E*	δ	E*/sin δ	E*/ sin δ Rank	FN Rank
LP	369.64	32.04	696.72	1	1
GP25	181.89	26.97	401.04	3	3
GP50	309.06	34.01	552.48	2	2

from best (1) to worst (3) along with the matching FN value. Similar ranking was depicted by the flow number testing and the SPT interpretation. Additionally, the results demonstrate that when 25% and 50% of filler content is replaced by glass, the stiffness factor is reduced by 42 % and 20%, respectively.

As for fatigue assessment using SPT, Witczak et al. [49] found a reasonable and significant correlation (${\mathrm{R}}^2$ = 0.7) between the fatigue stiffness factor evaluated at 21.1 °C and the fatigue cracking distress. The fatigue stiffness factor, also known as the loss modulus, was recommended and has been used to describe asphalt mixtures’ susceptibility to fatigue cracking [37,53]. A lower loss modulus suggests improved fatigue resistance. To conduct the analysis, |E*| and δ values at 21.1 °C and a frequency of 10 Hz were obtained from the corresponding master curve. Then, the loss modulus was calculated and ranked from best (1) to worst (3) as shown in

Table 10. Loss modulus evaluated at 10 Hz.

21.1 °C	E*	δ	E*. sin δ	Rank
LP	8224.09	20.92	2936.97	3
GP25	6358.48	25.22	2709.59	1
GP50	7219.07	23.70	2901.88	2

.

In general, it can be observed that glass use as filler material enhanced the fatigue resistance of HMA. The mix having 25% of glass powder showed more improvement than that having 50%.

Overall, a good agreement can be noticed between the fatigue and rutting assessments using SPT parameters and the interpretation of the dynamic modulus. The control mix “LP” with the highest modulus in the lower frequency range offered the best rutting stiffness factor, and the mix containing 25% glass powder had the best fatigue resistance, with the lowest |E*| in the medium and upper frequency ranges.

4. Conclusions

To investigate the effect of replacing filler with waste glass materials on the properties and performance of HMA, two commonly used laboratory performance tests, DM and FN, were performed on three asphalt mixes incorporating different glass contents of 0%, 25% and 50%. The main conclusions of this study are listed below:

• The utilization of waste glass powder to replace 25% and 50% of filler minerals reduces the stiffness of the conventional HMA by 21% and 12% at low frequencies, and 23% and 18% at medium frequencies. Nevertheless, the reduction is mostly exhibited at large frequencies, reaching 26% for both mixes utilizing glass fillers.
• The difference between |E*|of the control (LP) mix and the mixes containing glass powder is statistically significant for ${\mathrm{f}}_{\mathrm{r}}$ values above ${10}^2$ Hz.
• The optimum binder content of glass mixes was relatively smaller than their standard HMA counterpart.
• Higher waste glass content of 50% as filler provides higher |E*| values compared to 25% along the analyzed range of frequencies.
• The glass powder incorporated as filler, especially for the content of 25%, improved the fatigue and thermal cracking compared to the control HMA. On the other hand, rutting resistance was reduced; however, it was acceptable with a flow number above 190 cycles.
• In the case of phase angle, the use of the glass powder increased the viscous behavior of the asphalt mixture at intermediate and higher frequencies, and provided results comparable to the control HMA in the lower range of frequencies.
• The model suggested to fit the phase angle master curve provided better accuracy than the theoretical models examined, and therefore can be calibrated for regional use to predict δ of conventional and modified asphalt mixtures containing waste glass filler materials.

The findings revealed by this study should serve as a basis for further investigation in areas that may include:

• Conducting further tests to evaluate the resistance of mixes containing laminated waste glass powder (fillers) to moisture damage and stripping. According to a research by Behbahani et al. [54], zycosoil, a nanoscale anti-stripping agent, can be used to enhance the mechanical characteristics and moisture sensitivity of a mixture including asphalt and glass. Thus, it is recommended for future studies to evaluate asphalt concrete mixtures that include nanomaterials as anti-stripping material along with the glass powder.
• Investigating the properties and performance of HMA incorporating 100% of laminated waste glass powder as a replacement of mineral fillers.
• Testing the effect of using glass powder with different percentages at the binder and mastic levels.