Modified Asphalt Binders in Mixtures - Topical Report: Permanent Deformation Using A Mixture With Diabase Aggregate

1. Background

Pavement and laboratory tests performed on five surface course mixtures during the Federal Highway Administration's (FHWA) 1993 to 2001 Superpave Validation Study provided a good correlation between the rut depth in the asphalt pavement layer at 58°C and several laboratory mixture properties, including: (1) dynamic shear modulus, G*, at 40°C, (2) dissipated energy in the form of G*/sind at 40°C, (3) cumulative permanent shear strain at 40°C , (4) rut depths from the French Pavement Rutting Tester (French PRT) at 60°C, and (5) the creep slopes from the Hamburg Wheel-Tracking Device (Hamburg WTD) at 50°C.⁽¹⁾ The aggregate gradation and mixture volumetric properties were the same in all five mixtures; only the performance grade (PG) and type of asphalt binder (polymer modified vs. unmodified) were varied. The five asphalt binders were AC-5, AC-10, AC-20, Novophalt^TM, and Styrelf^TMI-D, having PG's of 58-34, 58-28, 64-22, 76-22, and 82-22, respectively.

G* and G*/sind at 40°C were measured using Frequency Sweep at Constant Height (FSCH). Cumulative permanent shear strain at 40°C was measured using Repeated Shear at Constant Height (RSCH). These tests were performed in accordance with American Association of State Highway and Transportation Officials (AASHTO) TP7-94, "Method for Determining the Permanent Deformation and Fatigue Cracking Characteristics of Hot-Mix Asphalt (HMA) Using the Simple Shear Test (SST) Device."⁽²⁾ The SST subjects a cylindrical specimen to simple shear. The acronym SST originally stood for Simple Shear Test, but it now also stands for Superpave Shear Tester. Shear properties were also measured at 58°C, but some of the data were highly variable and could not be used. The FHWA's Accelerated Loading Facility (ALF) was used to test the pavements for rutting. One relationship found during the study was:

where:

RD = Rut depth in the asphalt pavement layer at 58°C, mm.
CPSS = Cumulative permanent shear strain at 5,000 cycles and 40°C mm/m.

The relationship between pavement rut depth and each laboratory mixture test was determined so that mixture properties provided by other asphalt binders could be used to predict their relative ALF pavement rutting performances. Testing additional asphalt binders, mainly polymer-modified asphalt binders, is the subject of the study documented in this report. However, two changes to the mixture were made:

• The original aggregate blend consisted of 61-percent No. 68 diabase, 30-percent No. 10 diabase, 8-percent natural sand, and 1-percent hydrated lime. The amount of No. 68 aggregate was deficient so a new stockpile of aggregate was obtained. This was the only aggregate that needed to be replenished. The percent flat and elongated particles using a 3-to-1 length-to-thickness ratio were 16 percent for the new stockpile vs. 22 percent for the original stockpile. This was the only difference in the properties of the aggregates that was found. The percent flat and elongated particles for both stockpiles using a 5-to-1 length-to-thickness ratio were below 1.0 percent. Both stockpiles had average L.A. Abrasions, specific gravities, and percent water absorptions that were within the precision of the test.
  
  
• The hydrated lime in the new mixture was replaced with diabase fines because it was thought that the hydrated lime might interact with some of the polymer-modified asphalt binders to be tested in the new study. The effect of the asphalt binders on moisture susceptibility was also to be determined using no antistripping additive.

The asphalt binder content, volumetric properties, maximum specific gravity, percent natural sand, and the aggregate gradation, including the gradation of the material passing the 75-mm sieve, were not changed.

Mixtures with the five asphalt binders were tested by the SST using both the original and new aggregate blends. Table 1 and figure 1 show that the G*'s for each mixture at 40°C and the ALF-associated loading frequency of 2.0 Hz were not the same. A linear regression showed that the two sets of G*'s correlated to each other. The r² was 0.91. Frequencies ranging from 1.0 to 10.0 Hz provided similar differences in G*. Even though the data correlated to each other, the changes in G* meant that the relationships between ALF pavement rut depth and the various laboratory mixture properties, such as equation 1, were not valid. Because of this, it became more important to use a temperature closer to the PG's of the polymer-modified asphalt binders than to maintain a temperature of 40°C. The test temperature for the SST tests was increased to 50°C. This was thought to be the highest temperature that would provide repeatable data. Table 1 shows that there were discrepancies for the cumulative permanent shear strains where the strains at both 40 and 58°C were obtained in the Superpave Validation Study. The r² between the original strains at 40°C and the new strains at 50°C was 0.89.

The only method that could be used to relate the laboratory mixture properties to ALF pavement rutting performance was to develop new relationships using the new laboratory mixture properties. The applicability of this methodology is questionable. It assumes that the new set of mixture data provides the same pavement rutting performances as the original set of data. Because this should not be true, the predicted pavement rut depths should not be correct. Therefore, it must be assumed that the relative differences in the predicted rut depths are valid. If the new and original mixture properties provided by the five asphalt binders did not correlate to each other, this assumption would not be true.

Table 1. Comparison of SST properties provided by the original and new aggregate blends.

          ¹ND = No data.

Figure 1. G* of new aggregate blend vs. G* of original aggregate blend.

2. Objective

The objective of this study was to determine if the Superpave high-temperature rheological properties of polymer-modified asphalt binders correlate to asphalt mixture rutting resistance. The emphasis of this study was on evaluating the rutting resistances of mixtures containing polymer-modified asphalt binders with identical high-temperature PG's, but varied chemistries. This would indicate what types of modification provide properties that are, or are not, correctly captured by the current Superpave asphalt binder specification.

3. Materials

Eleven asphalt binders were obtained for this study. They consisted of eight polymer-modified asphalt binders: (1) styrene-butadiene-styrene [SBS] Linear, (2) SBS Linear Grafted, (3) SBS Radial Grafted, (4) ethylene vinyl acetate [EVA], (5) EVA Grafted, (6) Elvaloy, (7) ethylene styrene interpolymer [ESI], and (8) chemically modified crumb rubber asphalt [CMCRA]. As shown by this list, the asphalt binders include elastomeric and plastomeric modifiers, some with the same chemistry, but different geometry (linear vs. radial geometries, and grafted vs. ungrafted geometries). The term "grafted" includes any mode of chemically reacting a polymer with an asphalt binder, for example, vulcanization. There were three control asphalts: (1) air-blown, (2) unmodified PG 70-22, and (3) an unmodified PG 64-28. The target PG for the polymer-modified asphalt binders was PG 73-28. Descriptions and rheological properties of the asphalt binders are given in tables 2 and 3. Although identical PG's were desirable, the high-temperature PG's of the polymer-modified asphalt binders ranged from 71 to 77 after rolling thin-film oven (RTFO) aging. The unmodified PG 52-28 asphalt binder was not included in the study. The suppliers of the polymer-modified asphalt binders were allowed to modify this asphalt binder, the unmodified PG 64-28 asphalt binder, or a blend of both asphalts. The five asphalt binders used in the Superpave Validation Study were included, which meant that the total number of asphalt binders was 16.⁽¹⁾

Table 4 shows that the aggregate consisted of 91-percent crushed diabase and 8-percent quartzite natural sand. As previously indicated, the 1-percent hydrated lime was replaced with diabase dust. The aggregate gradation is shown in figure 2.

Mixture properties are given in table 5. The asphalt binder content was 4.85 percent by total mass of the mixture. Additional information on the asphalt binders, aggregates, and the mixtures are given elsewhere.^(1,3)

4. Tests

Asphalt binder properties were measured by a dynamic shear rheometer (DSR) after RTFO aging.⁽⁴⁾ Mixture rutting resistance was measured by: (1) G* and G*/sind using SST FSCH, (2) cumulative permanent shear strain after 5,000 cycles of repeated loading using SST RSCH, (3) rut depths from the French PRT at 6,000 wheel passes, and (4) creep slopes from the Hamburg WTD. The cumulative permanent shear strains from RSCH and the rut depths from the French PRT were considered the primary tests because they were specifically developed to measure rutting resistance. All mixtures were subjected to 2 h of short-term oven aging (STOA) at 135°C.^(1-2) All specimens were tested approximately 48 h after compaction.

5. Evaluation Using G* and G*/sind

G* and the phase angle, , of each mixture were measured by FSCH at 7.0-percent air voids and 50°C. The total loading time was 0.1 s, which is 10.0 Hz. The data for the 16 mixtures at 10.0 Hz are given in table 6. The SST is shown in figures 3 and 4.

G*/sind is often a better indicator of rutting resistance for mixtures containing polymer-modified asphalt binders than G*. When G* is used to measure rutting resistance, it must be assumed that all mixtures have the same amount of recovered elastic strain after unloading. When this assumption is true, the change in G* is proportional to the change in the unrecovered, permanent strain. Thus, mixtures with lower permanent strains have higher G*'s. This assumption is not needed when evaluating G*/sind because it is inversely proportional to dissipated energy, or damage. At high temperatures, the susceptibility to rutting should decrease as G*/sind increases. Table 6 shows that G*/sind and G* ranked the mixtures similarly. A linear regression provided an r² of 0.98. The error of using G* to evaluate the mixtures instead of G*/sind is small for this set of data. The largest change in ranking is for the mixture with the PG 70-22 asphalt binder. This mixture has the second highest G* (78.9 MPa), but the fifth highest G*/sind (83.9 MPa).

An analysis of variance and Fisher's least squares difference (LSD) were used to rank the 11 mixtures at a 5-percent level of significance. The capital letters in table 7 are the statistical rankings. All mixtures with the same letter have averages that are not significantly different from one another. They are in the same group. All groups are designated by a single letter. However, the groups can overlap. An average with more than one letter indicates that it falls into more than one group. For example, if an average has the designation "A B", it falls into two groups, both A and B. The mixtures from the Superpave Validation Study were not included in the ranking because the main objective of this study was to evaluate the performances of mixtures with modified asphalt binders having similar PG's.

Table 7 shows that many of the 11 mixtures had significantly different G*/sind's, especially at 2.0 Hz (the use of this frequency is discussed below). The reason for this, as shown by table 8, is that the variability of G*/sind is low. Most coefficients of variation are less than 10 percent. The rankings in table 7 show that grafting did not significantly improve the properties of EVA, and its effect on SBS was marginal.

The correlation between the G*/sind's of the 11 asphalt binders and the 11 asphalt mixtures, using the standard frequencies of 10.0 Hz and 10.0 rad/s, was poor. The r² was 0.50. The r² using all 16 materials was 0.72. These relationships are shown in figures 5 and 6, respectively. A log-log transformation increased the r² for the 16 materials to 0.79. The relationship in figure 6 indicates that there is a trend of increasing mixture G*/sind with an increasing binder G*/sind. The regression line should go through the zero-zero origin, but it does not. This indicates that the relationship must be curvilinear.

There was no correlation between the G*/sind's of the mixtures at 10.0 Hz and continuous high-temperature PG. The r² was 0.14. The correlation using all 16 materials was poor. The r² was 0.59. These relationships are shown in figures 7 and 8, respectively. Figure 8 does show a trend of increasing mixture G*/sind with increasing high-temperature PG. Poorer correlations could be expected using the high-temperature PG's compared to the G*/sind's of the asphalt binders at 50°C because the mixtures were tested at 50°C.

The G*/sind's of the materials were also correlated using 2.0 Hz and 2.0 rad/s because these frequencies are associated with slow-moving traffic. The data are included in table 7. Figures 9 and 10 show that good correlations were obtained. The r² was 0.81 for the 11 materials and 0.85 for the 16 materials. As expected, the r²'s between the high-temperature PG's and the G*/sind's of the mixtures at 2.0 Hz and 50°C were not as high. The r² was 0.35 for the 11 materials and 0.67 for the 16 materials.

Previous | Table of Contents | Next