Performance Testing for Superpave and Structural Validation

CHAPTER 7. CANDIDATE BINDER SPECIFICATION PARAMETER

INTRODUCTION

This chapter brings together full-scale ALF performance and laboratory mixture performance to assess the strengths and weaknesses of the various candidate binder parameters. The same composite statistical scoring technique is applied using diversified statistical measures, including slope regression significance (probability), Kendall’s tau measure of association and its associated significance (probability), and conventional correlation coefficient.

RUTTING/PERMANENT DEFORMATION

As previously discussed, the first measure of appropriateness for any binder or mixture laboratory test is whether the trends in the relationship are in the correct proportional or inverse direction. An extra step is needed to inspect the trends of the binder parameter with mixture performance and full-scale ALF performance because a composite score from the two will be utilized. It is possible, depending on the scatter in the data, that one of the two would be correct. Ideally, both should be correct. Table 100 lists the high-temperature binder rutting parameters and the corresponding laboratory mixture performance and ALF performance data that are compared for the 4-inch (100-mm) lanes. Table 101 lists the same for the 5.8-inch (150-mm) lanes. The slope of the linear regression was checked to screen relationships between binder parameters, mixture performance, and ALF performance to identify scenarios that should or should not receive continued analysis for the statistical composite score.

1 Pa = 0.000145 psi
1 mm = 0.039 inches

1 Pa = 0.000145 psi
1 mm = 0.039 inches

The relationships in table 100 to evaluate rutting in the 4-inch (100-mm) ALF lanes were all correct, and the relationships still held when the data points for lane 6 (terpolymer) and associated binder and mixture tests were removed. There were intermixed correct and incorrect trends for the rutting in the 5.8-inch (150-mm) ALF lanes. These results may seem surprising; however, the appearance of incorrect scenarios for the 5.8-inch (150-mm) ALF rutting does not signify poor or weak binder parameters. Rather, it is a direct reflection of the lack of variety in 5.8-inch (150‑mm) ALF rutting (see table 19 and figure 29 in chapter 3). This point puts the analysis at an impasse but is also an indicator of a very successful experimental design that targeted binders having equivalent high-temperature performance specifications but different intermediate-temperature performance specifications associated with fatigue cracking. This further suggests that the standard Superpave^® high-temperature specification is valid, at least for the materials in this research study, given that the materials were selected for the experimental design using the Superpave^® high-temperature rutting parameter. In other words, there could be an underlying relationship that identifies stronger or weaker parameters than Superpave^®, but the characteristics of the data simply cannot do so (see discussion in chapter 3 on capturing trends in light of scatter and number of data points).

The culmination of the numerical and statistical identification of the strongest and weakest binder parameters for rutting and permanent deformation are found in table 102 for all of the applicable 4-inch (100-mm) ALF lanes and in table 103 eliminating lane 6 (terpolymer). The entire composite score and each of the statistical components (regression slope significance, Kendall’s tau measure of association, significance of Kendall’s tau score, and correlation coefficient) used to compute the composite score are provided in the tables.

That the ranking of the strongest to weakest binder parameters did not essentially change whether the data points from lane 6 were included or excluded was somewhat unexpected. When the lane 6 terpolymer was removed, the stronger parameters became stronger and the weaker parameters became weaker. The rank order of the oscillatory-based nonrecovered stiffness and MSCR switched, but their scores were nearly identical in each comparison.

LSV and ZSV were the strongest statistical parameters associated with laboratory and full-scale rutting. The weakest was the standard Superpave^® parameter, which is counter to the alternative interpretation of ALF performance and experimental design because the binders were chosen based on the same Superpave^® high-temperature PG and exhibited statistically equivalent rutting. The next two strongest parameters quantify nonrecoverable deformation by different means; MSCR is a direct quantification while the oscillatory-based parameter is indirect but based on theoretical derivation and confirmed by comparison with direct MSCR. The variation of the standard Superpave^® parameter taken at a 0.25 radians/s frequency did better than the standard parameters taken at 10 radians/s, likely because of the intent to emphasize the softer portion of the binder response with polymer modification. MVR did better than the standard Superpave^® parameters but not as well as the modified, lower frequency Superpave^® parameter.

Discussion of Implementability, Purchase Specification Applicability, and Other Caveats

The quantitative ranking of the strongest and weakest parameters is important, but not a complete deciding factor in and of itself. Specification tests should ideally be both discriminating but also practical for broader use by the asphalt binder supply industry, contractors, and owner agencies. It is challenging to score and quantitatively rank the implementability of the candidate specifications. Qualitative consideration of various caveats associated with each test is provided to help further narrow down recommended specifications.

ZSV and LSV were identified as the strongest parameters, and both can be conducted in DSR equipment already implemented by Superpave^®. ZSV can require a long time for each test, and LSV offers an improvement by speeding up the process. Both of these computed viscosities correctly reflected the beneficial contributions of polymer modification. However, these parameters are still a physical measure of viscosity in which apparent improvements can be achieved by means of stiffening from fillers or polyphosphoric acid, which do not impart comparable performance-improving characteristics of polymer modification. This research further confirms the MVR as a valid alternative to the Superpave^® high-temperature PG, but the development and application of the MVR was intended as a rapid verification of PG grade. This leaves the two parameters that measure nonrecoverable deformations, and both can be measured using DSR. The profession may be able to relate with oscillatory-based nonrecoverable stiffness more than MSCR because it is based on the same properties currently measured for PG grade: |G*| shear modulus and the phase angle δ. On the other hand, MSCR has advantages over the oscillatory-based nonrecoverable stiffness because MSCR provides an additional measure of the recoverable deformation by means of percent recovery, which AASHTO TP 70 integrates.⁽⁷³⁾

FATIGUE CRACKING

The various candidate intermediate-temperature binder fatigue parameters were also compared against both laboratory fatigue tests and full-scale ALF fatigue cracking. The comparisons of binder with ALF performance and binder with the strain-controlled axial cyclic fatigue test selected in the previous chapter were combined into a single composite score to identify stronger and weaker tests for discriminating fatigue cracking. Table 104 summarizes the checks that were conducted to make sure that both the axial fatigue test and the 4-inch (100-mm) ALF fatigue cracking had the same trend and correct direction, whether an inverse relationship or proportional relationship. All binder tests provided the correct trend except the binder stress sweep fatigue test, which had the opposite ranking. When the trends were checked using the 5.8-inch (150-mm) ALF lanes and associated laboratory mixture tests, only CTOD, failure strain in low-temperature DT test, large strain time sweep surrogate, and Superpave^® |G*|sinδ had correct trends. Binder yield energy was not present, probably due to data scatter and the number of data points. More binder tests exhibited correct trends when SBS 64-40 data, which challenged the laboratory fatigue characterization ranking, were removed.

Table 105 ranks the binder parameters from strongest to weakest based on the composite score corresponding to the fatigue cracking performance in the 4-inch (100-mm) ALF lanes. The individual components for each axial fatigue and ALF comparison used to calculate the score (regression slope significance, Kendall’s tau measure of association, significance of the Kendall’s tau score, and correlation coefficient) are provided as well. The ranking reveals that there are more discriminating parameters than the Superpave^® |G*|sinδ. CTOD has the strongest association with laboratory and full-scale ALF fatigue cracking followed by the binder yield energy. Both of these parameters mobilize the binder to very large strains and deformations, which research has identified as a needed mechanism to capture the beneficial effects from polymer modification. Number of cycles to failure from the time sweep cyclic fatigue test is the third strongest parameter and takes place at a smaller strain, but the approach illustrates that cyclic fatigue on binder and cyclic fatigue on mixture are equally valid. Brittle failure strain in DT at temperatures much lower than the intermediate fatigue region discriminates fatigue cracking better than the standard Superpave^® fatigue parameter for these particular mixes, which reinforces using deformations larger than are applied in Superpave^® |G*|sinδ. The weaker parameters identified were the creep slope m-value from BBR and EWF. BBR m-value was identified in the literature review as worthy of exploration but did not appear to provide any discrimination with the materials in this experiment, possibly due the small deformations and low-temperature region. The weaker EWF is a necessary step in the calculation of CTOD by means of the yield strength. This suggests the contributions of yield strength to EWF to compute CTOD is important.

*Incorrect trend direction

The results from the ranking analysis corresponding to the 5.8-inch (150-mm) ALF lanes with and without lane 9 (SBS 64-40) are shown in table 106 and table 107, respectively. Consistent with the previous ranking, CTOD and binder yield energy are present at the top, which further supports the discriminating ability of these tests.

Evaluation of Top-Ranked Binder Fatigue Cracking Parameters from Ontario Highway 655

A collaborative effort between the Ontario Ministry of Transport, Ontario Hot Mix Asphalt Producers, and Queens University built pavement test sections to understand the influence of asphalt binder specifications on low-temperature thermal cracking. These test sections were similar to the ALF full-scale accelerated pavement experimental design because the mix design and construction were the same, and the only variable was asphalt binder. An overview of the binders used and physical properties measured by Queens University and TFHRC is provided in table 108.⁽¹⁰⁴⁾ The binders were characterized for CTOD, binder yield energy, and Superpave^® |G*|sinδ. The materials and performance data from the Ontario experiment offer an opportunity to explore CTOD and binder yield energy because cracking other than classical low-temperature thermal cracking appeared.

1 Pa = 0.000145 psi
°F = 1.8(°C) + 32
1 mm = 0.039 inches

Detailed crack maps provided by the Ontario Ministry of Transportation were used to classify cracking into longitudinal, centerline, edge, alligator, and transverse after 5 years of service (2003–2008).⁽¹⁰⁵⁾ It has been reported that the southbound traffic contained trucks having heavier loads than the northbound traffic because southbound trucks are returning from logging activities. Significantly different amounts of cracking are found in the two directions. This performance suggests that the difference between cracking in the northbound and southbound lanes could be load-associated cracking rather than low-temperature thermal cracking. Some limited alligator fatigue cracking appeared, but some short transverse cracking limited to within the wheel paths could be the beginning of interconnected alligator fatigue cracking.⁽¹⁰⁴⁾ The total number of cracks, total length of cracks, and length of longitudinal cracks are provided in table 109 through and table 111, respectively. The tables do not contain any centerline cracking. The ranking changes slightly depending on the type of cracking.

1 m = 3.28 inches

1 m = 3.28 inches

Again, the previously described statistical scoring was used, except only binder properties were compared against the full-scale highway test section cracking. The rankings of three binder tests are provided in table 112 through table 114 for the comparisons with the total number of cracks, total lengths of cracks, and length of transverse cracks, respectively. Although the results are weaker than ALF, CTOD has the strongest association with the observed cracking while the binder yield energy and Superpave^® |G*|sinδ are weaker and sometimes have incorrect trends altogether. These results, combined with the ALF results, help further identify and confirm that CTOD is a discriminating parameter for fatigue cracking.

°F = 1.8(°C) + 32

°F = 1.8(°C) + 32

*Somewhat.
**Mostly.

°F = 1.8(°C) + 32

Discussion of Implementability, Purchase Specification Applicability, and Other Caveats

The strongest binder parameters identified in table 105 as being better that Superpave^® |G*|sinδ are the CTOD, binder yield energy, time sweeps, and low-temperature DT failure strain. The advantage of binder yield energy is that it can be measured in the DSR. However, University of Wisconsin researchers have recently postponed further development of this test in favor of alternative strain sweep characterization procedures that take advantage of VECD methodologies because of one particular shortcoming with the binder yield energy test: some modified binder exhibit two peaks (initial yield and ultimate yield) in the binder yield energy test, which presents a challenge as to when and where the strain energy is to be calculated (see figure 119). The VECD-based stress sweep binder test was outside the scope of this study at the time this report was written. Low-temperature failure strain was the next strongest parameter. Although low-temperature DT testing combined with BBR testing provides a more rigorous low-temperature PG grade than BBR alone, DT testing has already fallen out of favor by agencies. If this was to be reconsidered for fatigue resistance, the failure strain alone is likely vulnerable to not appropriately catching fatigue and cracking resistance because softer or lower-quality binders could exhibit higher strain tolerance. This was not the case for the legitimate, high-quality binders in this study. DT test failure strain would have to be accompanied with a strength measurement that the DT test can provide, but DENT testing at intermediate temperatures for CTOD is already a type of tension test where extension and strength are measured. Nonetheless, the largest hurdle for implementation of CTOD using DENT is the need for a new piece of test equipment if a laboratory does not already have one to measure force-ductility (AASHTO T 300).⁽¹⁰⁶⁾ Ruggedness evaluation of existing AASHTO T 300 equipment for measuring CTOD would be a necessary next step.