National Academies Press: OpenBook

Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction (2017)

Chapter: Chapter 3 - Findings and Applications

« Previous: Chapter 2 - Research Approach
Page 24
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 24
Page 25
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 25
Page 26
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 26
Page 27
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 27
Page 28
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 28
Page 29
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 29
Page 30
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 30
Page 31
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 31
Page 32
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 32
Page 33
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 33
Page 34
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 34
Page 35
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 35
Page 36
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 36
Page 37
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 37
Page 38
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 38
Page 39
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 39
Page 40
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 40
Page 41
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 41
Page 42
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 42
Page 43
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 43
Page 44
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 44
Page 45
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 45
Page 46
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 46
Page 47
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 47
Page 48
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 48
Page 49
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 49
Page 50
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 50
Page 51
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 51
Page 52
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 52
Page 53
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 53
Page 54
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 54
Page 55
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 55
Page 56
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 56
Page 57
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 57
Page 58
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 58
Page 59
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 59
Page 60
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 60
Page 61
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 61
Page 62
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 62
Page 63
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 63
Page 64
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 64
Page 65
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 65
Page 66
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 66
Page 67
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 67
Page 68
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 68
Page 69
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 69
Page 70
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 70
Page 71
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 71
Page 72
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 72
Page 73
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 73
Page 74
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 74
Page 75
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 75
Page 76
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 76
Page 77
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 77
Page 78
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 78
Page 79
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 79
Page 80
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 80
Page 81
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 81
Page 82
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 82
Page 83
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 83
Page 84
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 84
Page 85
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 85
Page 86
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 86
Page 87
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 87
Page 88
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 88
Page 89
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 89
Page 90
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 90
Page 91
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 91
Page 92
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 92
Page 93
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 93
Page 94
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 94
Page 95
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 95
Page 96
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 96
Page 97
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2017. Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction. Washington, DC: The National Academies Press. doi: 10.17226/24959.
×
Page 97

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

24 Findings Sensitivity Study Temperature–frequency sweep tests were first performed to determine the modulus values of the materials and subsequently the aging ratio (AR). As represented in Equation 11, AR is the mathematical ratio of |G*| or dynamic modulus, |E*|, after aging to the value before aging. It is calculated at multiple temperatures (Tj) and frequencies (wi) to gain a com- plete picture of the impacts of oxidation. For comparison across different study materials, a singular frequency was required and 10 rad/s was chosen for this purpose. The ARs were used to assess the sensitivity of both the mastics and FAM to their corresponding binder oxidation levels. * , * , * , * , (11)AR G T G T mastic and binder E T E T FAM i j after aging i j before aging i j after aging i j before aging ( ) ( ) ( ) ( ) = ω ω     ω ω          The aging parameter (AP), described as the change in the parameter of interest with time, was also used to represent the oxidative aging of the mastic and FAM samples. Equation 12 shows an example of the AP calculated based on G*c from the unaged case, e.g., at t = 0, G*c0, and crossover modulus at different oxidation times, G*ct. The kt represents the aging rate of the material. 1 * 1 * (12) 0 AP kt G Gct c = = − The approach adopted here to evaluate sensitivity is based on the concept of the crossover modulus and the principles of second-order rate kinetics of binder oxidation. Because a direct analysis of the AR and binder AP did not provide a clear understanding of the quantification of the differences, a more involved methodology was followed. In the first application of this approach, the binder AP accuracy required to match varying levels of accuracy (1% to 20%) for the binder, mastic, and FAM ARs was estimated. In the second application, the errors in the AR when the AP was matched at 1% to 20% were evaluated. The primary conclusions drawn from these two analyses are summarized in Table 5 and Figure 11, which indicate the accuracy levels that are required for the binder AP to match the binder, mastic, and FAM AR at 1%, 10%, and 20% levels of accuracy. It should be noted that, based on recent advances reported in studies that compare the mechanical properties of FAM C H A P T E R 3 Findings and Applications

Findings and Applications 25 Error in AR Range of Required AP Accuracy Binder Mastic FAM 1% 1.1% – 1.9% 1.1% – 1.9% 1.5% – 3.6% 10% 11.0% – 18.9% 11.1% – 19.5% 14.8% – 35.6% 20% 22.0% – 37.7% 22.2% – 39.1% 29.6% – 71.2% Error in AR Range of Required AP Accuracy FAM 1.5% – 3.6% 14.8% – 35.6% 1% 10% 20% Binder 0.5% – 0.9% 6.3% – 9.1% 12.3% – 18.3% Mastic 1.1% – 1.9% 11.1% – 19.5% 22.2% – 39% 29.6% – 71.2% Table 5. Summary of sensitivity study findings. 0 25 50 75 Pe rc en t A cc ur ac y in B in de r A P Pe rc en t A cc ur ac y in B in de r A P Binder 0 25 50 75 Percent Error in AR Percent Error in AR Percent Error in AR (b) (c) (a) Mastic 0 25 50 75 Pe rc en t A cc ur ac y in B in de r A P FAM Line of Equality Line of Equality Line of Equality Change in Binder AR is 1.1 to 1.9 times less than the change in AP Change in FAM AR is 1.5 to 3.6 times less than the change in AP Change in Mastic AR is 1.1 to 1.9 times less than the change in AP 0 5 10 15 20 250 5 10 15 20 25 0 5 10 15 20 25 Figure 11. Summary of results from the sensitivity study.

26 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction and asphalt concrete, the sensitivity level of a full asphalt concrete mixture would be expected to be lower than for FAM, but by only a relatively small amount. The findings presented in this section are based on the AP-based sensitivity assessment; however, the same conclusions are reached when the assessment is based on G*. The significance of these findings is two-fold. First, they demonstrate that the properties of asphalt concrete are not proportionally sensitive to changes in the asphalt binder modulus. That is, increases in the binder modulus by a given factor do not necessarily result in equivalent factor changes for the asphalt concrete modulus. Second, as demonstrated most clearly in Figure 11, these findings show that for a laboratory aging procedure to replicate the impacts of binder oxi- dation on the modulus of asphalt concrete with a given level of accuracy (say 10%), the desired binder oxidation can be replicated at less accuracy by a factor of 1.5 to 3.6 (15%–36%). Con- versely, if a laboratory aging procedure is found to match the in-service level of binder oxidation with a certain percentage of error (say 10%), then the expected percentage of error in the resulting modulus value of an asphalt mixture that is tested after being subjected to that laboratory process would be 1.5 to 3.6 times higher (6.7%–2.8%). For the assessment of fatigue sensitivity, the parameter used to quantify the sensitivity was the strain ratio (SR). This SR parameter is based on the ratio of the strain levels (SL) that are required to achieve given number of failure cycles after and before aging as shown in the Equation 13. (13)SR SL SL after aging before aging = Assessments were made using a polynomial relationship between the AP and SR parameters. The general trend observed in fatigue sensitivity analysis is that if a laboratory aging procedure is found to match the in-service level of binder oxidation with a certain percentage of error (say 10%), then the expected percentage of error in the resulting fatigue properties (the SR in this case) of an asphalt mixture that is tested after being subjected to that laboratory process will be lower than or equal to 10%. Both sensitivity parameters, i.e., the AR and SR, and their relationship to the AP, show that the sensitivity of the mechanical properties to the oxidation in asphalt binders decreases when going from mastic to FAM. Presumably, the sensitivity would decrease even further in the asphalt mixture. Although general conclusions can be drawn for fatigue behavior using the relationship between the SR and AP, the actual quantification of the accuracy of the SR with respect to the error in the AP of the binder cannot be generalized for different binders and material types. A detailed discussion of the sensitivity study is included in Appendix C. Selection of the Chemical and Rheological Aging Index Properties Six material sources, detailed in Table 6, were used to select the chemical and rheological AIPs to track the oxidation levels in this project. The materials evaluated encompass a wide range of binder types. All of the mixtures were subjected to laboratory aging. For three of the mixtures, both original component materials and field cores from in-service pavements were available and evaluated. In the case of the WesTrack section, field cores were available at three different times after placement whereas only one level of aging was available for the FHWA ALF-styrene butadiene styrene (SBS) and control field cores. To evaluate the changes in the chemical and rheological AIPs in terms of oxidation, laboratory loose mixture aging was conducted at multiple temperatures ranging from 70°C to 95°C with durations ranging from 8 days to 35 days. Samples were collected at different time intervals for binder extraction and recovery and subsequent AIP testing.

Findings and Applications 27 Results Selection of Chemical AIP. The chemical AIPs were evaluated based on their correlation to the aging duration. Asphalt materials exhibit relatively similar kinetics, with an initial fast reaction period, also known as the spurt, followed by a slower reaction period that has an approximately constant rate (Glaser et al. 2013a, Han 2011, Jin et al. 2011, Petersen 1998, Petersen et al. 1996). The focus herein is on long-term aging. Therefore, the evaluation of chemical AIPs was conducted within the constant rate period because it corresponds to higher age levels. Figure 12 shows the comparisons between the three chemical AIPs evaluated (i.e., carbonyl area, C + S area, and C + S peaks) with respect to aging duration. The rate of oxidation is temperature-dependent. Therefore, the chemical AIPs measured from long-term aging at different temperatures for a given mixture are plotted separately. All of the data included in the analysis of the chemical AIPs correspond to aging trials where extracted and recovered binder analysis was conducted for at least three aging durations. As demonstrated in Figure 12, generally all three chemical AIPs strongly correlate with the aging duration. However, it can be noted that the effect of the sulfoxide functional group on the chemical AIP-based aging rates is of great importance. Although the C + S area and C + S peaks show very similar aging rates and rankings, the carbonyl area versus aging duration shows a different ranking of the materials. That is, in Figure 12(a), the carbonyl area is greater for the AAG-1 mixture than for the AAD-1 mixture, although their slopes are similar. However, in Figure 12(b), the C + S area is greater for the AAD-1 mixture than for the AAG-1 mixture, and the slope of the C + S area is higher for the AAD-1 mixture than for the AAG-1 mixture. The literature indicates that sulfoxides have a significant effect on rheology, and thus, this observa- tion suggests that it is important to consider the sulfoxide functional group when tracking oxi- dative aging. The C + S peaks versus aging duration graph exhibits the overall highest R2 values. Furthermore, the C + S peaks can be calculated using the direct output of the ATR–FTIR data, whereas the calculation of the C + S area requires numerical integration under the infrared (IR) spectrum. Therefore, the C + S peaks is an easy parameter to calculate and is not sensitive to the method chosen for the calculation. Thus, the C + S peaks was selected as the most promising chemical AIP identified and was used in the subsequent analysis of the rheological AIPs. Selection of Rheological AIP. The correlation between rheological AIPs and the C + S peaks for each mixture was used to evaluate the rheological AIPs. Data that correspond to different aging temperatures ranging from 70°C to 95°C were included in the evaluation. Past studies have demonstrated that the relationship between chemistry and rheology is not affected by aging temperature if the temperature is below 100°C (Petersen 2009, Elwardany et al. 2017a, Yousefi Rad et al. 2017). Therefore, the data for multiple aging temperatures are included for each evaluated mixture. Mixture AggregateType/Source Binder PG/Modification Field Material Availability Additive NC Granite/NC 64-22/None N/A N/A FHWA ALF- Control Granite/VA 70-22/None Available Hydrated Lime FHWA ALF- SBS-LG Granite/VA 70-22/SBS Available Hydrated Lime SHRP AAD The same source as FHWA ALF 58-28/None N/A Hydrated Lime SHRP AAG The same source as FHWA ALF 58-10/None N/A Hydrated Lime WesTrack Granite/NV 64-22/None Available Hydrated Lime Note: NC = North Carolina, VA = Virginia, NV = Nevada. Table 6. Asphalt mixtures used for AIP evaluation.

28 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction Log G* at 64°C and 10 rad/s was selected as the rheological AIP to evaluate oxidation levels within the project. It was found that oxidative age hardening affects G* most significantly at high temperatures and/or low frequencies. Therefore, the AIPs that were evaluated at low reduced frequencies were found to be the most effective. Figure 13 presents the correlation between log G* at 64°C and 10 rad/s and the C + S peaks for all the mixtures evaluated. Each data point corresponds to a different aging duration and/or aging temperature. A log scale of the rheo- logical parameters was used to evaluate the data in a linear trend for easier interpretation of the test results. The results include those of the binder extracted and recovered from loose mixture aging that was conducted at various temperatures and depths. The results demon- strate that the G* at 64°C and 10 rad/s is highly correlated to the chemical changes induced by oxidation for all of the mixtures evaluated. The G-R parameter also demonstrated a high correlation with the C + S peaks. Detailed data are included in Appendix D. In addition, it was found that the G-R is highly correlated with log G* at 64°C and 10 rad/s. The G-R parameter was evaluated at a very low frequency (0.005 rad/s) at which the direct measurement of the G* and phase angle is not possible. Therefore, the deter- mination of the G-R parameter required frequency sweep testing at multiple temperatures and the fitting of a master curve and time-temperature shift models to the rheological data because R² = 1 R² = 0.8867 R² = 0.9367 R² = 0.1406 R² = 0.9704 R² = 0.989 R² = 0.9645 2.0 3.0 4.0 5.0 Ca rb on yl A re a (A U) Oven Aging Duration (Day) NC - 95°C FHWA ALF-SBS - 85°C FHWA ALF-SBS - 95°C SHRP AAD-1 - 70°C SHRP AAD-1 - 95°C SHRP AAG-1 - 95°C WesTrack - 95°C R² = 0.9895 R² = 0.9438 R² = 0.973 R² = 0.4373 R² = 0.976 R² = 0.9965 R² = 0.938 5.0 6.0 7.0 8.0 9.0 Ca rb on yl + S ul fo xi de A re a (A U) Oven Aging Duration (Day) NC - 95°C FHWA ALF-SBS - 85°C FHWA ALF-SBS - 95°C SHRP AAD-1 - 70°C SHRP AAD-1 - 95°C SHRP AAG-1 - 95°C WesTrack - 95°C R² = 0.9931 R² = 0.9892 R² = 0.9549 R² = 0.8378 R² = 0.9786 R² = 0.9971 R² = 0.9614 0.07 0.09 0.11 0.13 0.15 Ca rb on yl + S ul fo xi de P ea k (A U) Oven Aging Duration (Day) (a) (b) (c) NC - 95°C FHWA ALF-SBS - 85°C FHWA ALF-SBS - 95°C SHRP AAD-1 - 70°C SHRP AAD-1 - 95°C SHRP AAG-1 - 95°C WesTrack - 95°C 100 5 15 20 25 30 35 100 5 15 20 25 30 35 100 5 15 20 25 30 35 Figure 12. Sensitivity of different chemical AIPs to aging duration: (a) carbonyl area, (b) C + S area, and (c) C + S peaks.

Findings and Applications 29 testing at 0.005 rad/s was not possible. In contrast, the G* at 64°C and 10 rad/s is a single point measurement that can be acquired directly from testing at a single temperature and frequency. The problems associated with the ZSV and crossover modulus are related to the interpolation and extrapolation of the data. In addition, the crossover modulus is less sensitive to the chemical changes caused by oxidation compared to the other rheological parameters evaluated. Therefore, the G* at 64°C and 10 rad/s constitutes the simplest and most efficient and effective rheological AIP evaluated. The complete results and discussion that led to the selection of the chemical and rheological AIPs are provided in Appendices D and E. Selection of Long-Term Aging Method To evaluate candidate aging procedures, preliminary aging trials were conducted using a typi- cal North Carolina mix with 9.5 mm NMAS and PG 64-22 binder, hereinafter referred to as the NC mix. The integrity of the specimens following aging, the rate of oxidation quantified using the AIPs of the extracted binder, versatility, and the cost of the various procedures were compared in order to select the most promising aging procedure. The selected procedure was then applied to a FHWA ALF-SBS–modified mixture, which is known to be highly susceptible to hardening with oxidation and difficult to compact, in order to verify that the procedure would not degrade the specimen integrity. Aging trials using the SBS mix included laboratory aging to match the oxida- tion level of the surface of an 8-year-old field core obtained from McLean, Virginia (the location of the FHWA ALF). Based on the literature review, two candidate aging methods were identified: oven aging and pressurized aging. For the preliminary evaluation of these two aging procedures, a standard binder PAV was utilized for the pressurized aging trials. Oven and pressurized aging methods lo g G *, at 6 4° C, 1 0 ra d/ s (kP a) R² = 0.978 R² = 0.9783 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Carbonyl + Sulfoxide Peak (AU) SHRP AAD-1 SHRP AAG-1 0.07 0.09 0.11 0.13 R² = 0.9879 R² = 0.9789 0.5 1.0 1.5 2.0 2.5 3.0 lo g G *, at 6 4° C, 1 0 ra d/ s (kP a) Carbonyl + Sulfoxide Peak (AU) NC WesTrack Fine 0.07 0.09 0.11 0.13 lo g G *, at 6 4° C, 1 0 ra d/ s (kP a) R² = 0.9907 R² = 0.9152 0.5 1.0 1.5 2.0 2.5 3.0 Carbonyl + Sulfoxide Peak (AU) FHWA ALF-Control FHWA ALF-SBS 0.07 0.090.08 0.110.1 0.12 Figure 13. Correlation between G* at 64°C and 10 rad/s with C + S peaks for six mixtures.

30 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction were applied to both loose mix and compacted specimens and the potential integrity problems associated with the two specimen types were evaluated. Study of Aging of Compacted Specimens Trials for aging the compacted specimens were conducted using both the oven and PAV. The current standard procedure for asphalt mixture laboratory aging, AASHTO R 30, consists of aging compacted specimens in the oven at 85°C for 5 days. However, previous research identified two specimen integrity problems using this procedure: 1. Distortion: Changes in the air void content and geometry due to slumping under self- weight have been reported when using AASHTO R 30 (Reed 2010). To overcome this issue, the NCHRP Project 9-23 protocol recommends wrapping specimens in metal wire mesh secured with three clamps to prevent the samples from geometry distortion (Houston et al. 2005). However, this approach has been reported only to reduce, but not eliminate, specimen distortion during aging (Reed 2010). 2. Oxidation gradient: NCHRP Project 9-23 demonstrated that the long-term oven aging of com- pacted specimens leads to both radial and vertical oxidation gradients in mixture specimens, which is a concern for its use in performance testing because properties differ throughout a specimen (Houston et al. 2005). To overcome these specimen integrity problems, two potential remedial approaches were tried in this study: 1. The application of pressure to increase the diffusion of the oxygen and, hence, potentially reduce the oxidation gradients. 2. The use of small specimens, 38 mm in diameter and 100 mm in height, to reduce the diffusion path distances and reduce the slump under self-weight. To evaluate compacted specimen aging as rigorously as possible, various procedures were tried, including pressurized and oven aging of both the large and small specimens. All specimens were fabricated using the NC mix according to the procedures explained in Chapter 2. Three criteria were used to evaluate the integrity of the specimens that were subjected to compacted specimen aging: 1. Initial integrity check: Specimen integrity was evaluated initially by visual inspection, dimension measurements, and air void content comparisons before and after aging to determine if the specimen had been damaged during the laboratory aging process. If an aging procedure was found to have disturbed the specimen integrity, it was eliminated from further consideration. 2. Performance testing: Performance test data were analyzed to detect any integrity issues that were not related to measurable geometry changes (e.g., microcracking). Dynamic modulus tests (AASHTO TP 79-12) and cyclic direct tension fatigue tests (AASHTO TP 107-14) were conducted, and the results were compared to the short-term aged mixture properties to deter- mine if the specimen integrity had been affected during aging (e.g., if the dynamic modulus value had decreased upon applying the aging procedure). 3. Oxidation gradient: Oxidation gradients in the aged specimens were evaluated through FTIR and DSR testing of binders extracted and recovered from locations in an aged specimen that varied in terms of distance from the specimen periphery. Differences in the rheology and chemical compositions, along with the distance from the specimen periphery, were used to detect the presence (if any) of an aging gradient. Study of Aging of Loose Mixtures The aging of loose mixtures provided another potential solution to overcome the issues associated with the AASHTO R 30 procedure. Geometry distortion was not a concern because

Findings and Applications 31 the loose mix specimens are compacted following aging. In addition, the aging gradient was not a problem, because the loose mix is aged as a single layer of coated aggregate particles and, thus, oxygen and heat could circulate easily throughout the mix. Also, the increased surface area of the binder film that is exposed to oxygen was expected to accelerate aging in loose mix- tures compared to compacted specimens. However, the compaction of aged loose mix for per- formance testing remained a potential specimen integrity concern because aged binder is very stiff and, thus, was expected to be less compactable than unaged material. A past study on the loose mix aging of an asphalt rubber friction course (ARFC) reported that significantly more effort is required to compact long-term aged loose mix than short-term aged loose mix (Reed 2010). However, it is important to note that an ARFC represents an extreme case with rubber- modified asphalt and relatively thick asphalt film. It also has been found that the increased force/effort required to reach target air void contents when compacting aged loose mixtures may cause degradation of the aggregate structure and alter the mixture properties (Gatchalian 2006). Also, the compactability of aged loose mix can potentially be improved by increasing the compaction temperature or by adding a compaction aid such as zeolite. For this study, preliminary loose mix aging trials consisted of both oven and PAV aging of the NC mix at 85°C. Loose mix aging trials in the oven and PAV were conducted using the proce- dures described in Chapter 2. Two criteria were used to evaluate the integrity of the specimens compacted following loose mix aging: 1. Initial integrity check: The number of gyrations required to meet the target air void content was compared with that required for the short-term aged mix in order to assess compactabil- ity. Air void content measurements were used to verify that the desired compaction level was met. In addition, digital imaging processing software was used to analyze the internal coarse aggregate structure of the compacted short-term and long-term aged loose mixes to determine if the aggregate structure had been degraded by compacting the mixes following long-term aging. 2. Performance testing: Performance tests, including dynamic modulus tests (AASHTO PP 342) and cyclic fatigue tests (AASHTO TP 107-14), were conducted using specimens that were compacted following loose mix aging to assess any further potential integrity problems (e.g., dynamic modulus decrease upon aging). Evaluation of the Aging Procedures To select the most promising aging procedure, compacted and loose mix aging procedures were compared based on the following criteria: 1. Specimen integrity: Specimen integrity, as related to compacted and loose mix aging, as previ- ously discussed, is important for reliable performance evaluation. 2. Efficiency: The relative rate of oxidation achieved in each procedure, as evaluated through comparisons of chemical and rheological AIPs, can be quantified using ATR FTIR and DSR temperature–frequency sweep testing, respectively. 3. Practicality and versatility: The relative cost and availability of the required equipment were considered in selecting the most promising aging procedure. Furthermore, the versatility of the specimen geometries that could be produced for performance testing was an important consideration. It was necessary to verify the selected procedure for an additional mixture, particularly to evaluate potential concerns regarding specimen integrity. For this purpose, the FHWA ALF- SBS–modified mixture was selected as it is known to be both difficult to compact and highly susceptible to hardening with oxidation. Asphalt binder was extracted and recovered from a field core obtained after 8 years in service at the FHWA ALF in McLean, Virginia, for com- parison to the loose mix aging trial results. Aging for 21 days at 95°C was found to match the

32 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction level of oxidation of the binder extracted from the surface of the field core. This condition is considered to represent an extreme level of oxidation, as oxidation levels are drastically reduced with the depth of the pavement. To evaluate the integrity of the aged loose mix, the aged mix was compacted with no adjustment to the short-term aging compaction temperature and then subjected to dynamic modulus and cyclic fatigue performance testing. The number of gyrations required for compaction and the performance test results were compared to the results of the short-term aged specimens to assess the integrity of the specimens that were compacted after loose mix aging. Initial Integrity Check Results Compacted Specimen Aging. The initial integrity checks of the compacted specimens included visual inspection, air void content, and dimensions of the specimens. In cases where air void content and dimensional integrity problems were encountered, remedial strategies were developed to eliminate them. These strategies are detailed in Table 7. The comparison of the specimen air void contents before and after aging indicated that very minor air void changes occurred during the aging of both the small and large specimens with and without pressure. Note that the initial integrity checks for the small specimens aged in the oven at 85°C for 8 days indicated that no wire mesh was needed to avoid geometry distortion due to the relatively low weight of the small specimens. Loose Mix Aging. The number of gyrations required to reach the target air void content and analysis of the coarse aggregate structure were used as initial integrity checks for loose mix aging because the primary integrity concern was the ability to compact aged material for performance testing. Two compaction temperatures were tried: 144°C and 157°C. The results showed no significant difference in the compaction effort required for the short-term and long-term aged materials. The results thus indicate that it is possible to compact aged loose mix with no adjust- ment to the compaction temperature. Performance Testing Following the initial integrity checks, the specimen integrity of the aged mixtures was assessed using performance testing. Only a limited number of aging conditions were selected for perfor- mance testing due to the constraints of time and resources. For specimens that were aged using compacted specimen aging, the strategies detailed in Table 7 were utilized to minimize integrity- related problems. Dynamic Modulus Testing Results. Figure 14 presents the dynamic modulus master curves; the results shown correspond to the average values of two replicates. No meaningful difference is Material State Temperature Pressure Recommendation to Avoid Integrity Problems Compacted Large Specimen 85°C - Wire mesh support should be used for specimens. Compacted Large Specimen 85°C 0.30 MPa Specimen should be placed in a hammock-like support on its side. Compacted Small Specimen 85°C - Specimen should be placed in the oven on its side on a flat surface. Compacted Small Specimen 85°C 0.30 MPa Controlled air pressure application should be considered and specimen placed on its side on a flat surface. Note: “-” equals no pressure. Table 7. Summary of findings from Level 1 integrity check of compacted specimen long-term aging.

Findings and Applications 33 evident between the dynamic modulus values of the short-term aged and PAV-aged compacted specimens, indicating that either (a) no significant aging occurred or (b) the application of pressure damaged the specimens. The specimens aged using other methods show a significant increase in their dynamic modulus values. Loose mix aging appears to lead to slightly higher dynamic modulus values than compacted specimen aging in the oven for the same duration, thereby indicating that aged loose mix can be compacted for performance testing. In addition, these results suggest that higher levels of oxidation are achieved using loose mix aging com- pared to compacted specimen aging, given the same temperature and duration of conditioning in an oven. The compaction temperature utilized for the aged loose mixes had little effect on the dynamic modulus values. Cyclic Fatigue Performance Testing Results. Figure 15 presents the damage characteristic curves obtained from analysis of the cyclic direct tension test results for both the short-term and long-term aged materials. Typically, damage characteristic curves are used to describe the 1E+02 1E+03 1E+04 1E+05 |E* | (M Pa ) Reduced Frequency (Hz) S-NC L-O-85-8D-NC Compacted at 144°C L-O-85-8D-NC Compacted at 157°C C-O-85-8D-NC C-P3-85-1D-NC 1E+02 5E+03 1E+04 2E+04 2E+04 3E+04 |E* | (M Pa ) Reduced Frequency (Hz) (a) (b) S-NC L-O-85-8D-NC Compacted at 144°C L-O-85-8D-NC Compacted at 157°C C-O-85-8D-NC C-P3-85-1D-NC 1E+021E-08 1E-06 1E-04 1E-02 1E+00 1E+021E-08 1E-06 1E-04 1E-02 1E+00 Figure 14. Dynamic modulus results: (a) log-log scale and (b) semi-log scale. 0.0 0.2 0.4 0.6 0.8 1.0 C S Short-term Aged - Compacted at 144°C Oven, Loose Mix, 85°C, 8 days - Compacted at 144°C Oven, Loose Mix, 85°C, 8 days - Compacted at 157°C PAV, Compacted Spec., 85°C, 1 day, 300 kPa Oven, Compacted Spec., 85°C, 8 days Failure Point 1.0E+060.0E+00 2.0E+05 4.0E+05 6.0E+05 8.0E+05 Figure 15. Comparison of damage characteristic curves for the NC mixes subjected to different aging conditions (C = pseudo stiffness and S = damage parameter).

34 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction relationship between material integrity and damage and are path-independent (i.e., independent of loading and thermal history). The curves presented in Figure 15 represent the averages obtained from three different cross-head displacement amplitudes used in the tests. The damage character- istic curves of the stiffer materials usually are higher than the damage characteristic curves of the softer mixtures, as demonstrated also in Hou et al. (2010). Because aging is expected to increase the stiffness of asphalt mixtures, it was expected that the damage characteristic curves of the long- term aged mixtures would be higher than those of the short-term aged mixtures, unless an integ- rity problem existed in the specimen. The results demonstrate that all the oven-aged specimens (both aged loose mix and aged compacted specimens) have higher damage characteristic curves than the short-term aged specimens, indicating no integrity issues. In other words, the compac- tion of aged loose mix does not appear to lead to integrity problems. Furthermore, the results for the long-term aged loose mixture trials show similar damage characteristic curves regard- less of the compaction temperature, indicating that specimen integrity can be achieved without elevating the compaction temperature. In addition, the results indicate that the oven-aged loose mix specimens have slightly higher damage characteristic curves than the oven-aged compacted specimens, which is consistent with the dynamic modulus test results. The research team compared two sets of compacted specimens that were subjected to oven aging: aged 178-mm tall specimens, whereby the specimen ends were cut following aging to reduce the height to 130 mm for testing, and 130-mm tall specimens aged in the oven after cut- ting the ends of the specimens. The damage characteristic curves were similar for the two sets of samples. However, the specimens that were aged at a height of 130 mm (i.e., ends cut before aging) demonstrated a high propensity for end failure, indicating a higher level of oxidation at the ends of the specimen and, hence, a testing concern. The damage characteristic curves of the PAV-aged compacted specimens lie below those of the short-term aged specimen, indicating that the specimens were damaged by the application and/or release of pressure during aging. Thus, the application of pressure when aging compacted specimens should be avoided. A new fatigue failure criterion, termed DR (Wang and Kim 2017), was also used to evaluate mixtures with different aging levels. The DR criterion uses the average reduction in pseudo stiff- ness (i.e., C) up to failure. The DR value is calculated as the summation of (1 – C), illustrated in Figure 16, divided by the fatigue life (number of cycles to failure) for individual test replicates. DR is a material constant that is independent of mode of loading, temperature, and stress/strain amplitude. Note that the calculation of DR is in arithmetic scale rather than in log-log scale. Con- sequently, DR results are not as affected by test variability as another pseudo energy-based failure criterion, termed GR, that is calculated in log-log scale (Sabouri and Kim 2014). Number of Cycles (N) Ps eu do S tif fn es s (C ) Figure 16. Illustration of summation of (1–C).

Findings and Applications 35 Figure 17 depicts the DR energy-based failure criterion plots. Average DR values for different aging conditions are presented in Table 8. For severely aged materials, it is anticipated that the DR failure criterion lines will fall below those of short-term aged materials due to the loss of fatigue resistance associated with embrittlement imposed by oxidation. Any integrity problem in the specimens is also expected to result in lower DR values. Thus, the failure criterion results for the PAV-aged compacted specimens indicate a potential integrity problem, as the dynamic moduli values for the PAV-aged compacted specimens were lower than those from the short- term aged mixtures. All the other failure criterion lines fall close to that of the short-term aged materials, which indicates no integrity problems. Evaluation of Aging Gradient in Compacted Specimen Aging The chemical and rheological data for extracted and recovered binder that were obtained from the different aging trials were compared using AIPs. Based on the sensitivity study results (see Table 5 and Figure 11), a 2% change in the C + S absorbance and a 15% change in the binder complex modulus both correspond to an approximately 10% change in the dynamic modulus values of the mixture samples. A 10% change in the dynamic modulus value is considered a reasonable threshold of significance, and hence, a 2% change in the C + S absorbance and 15% change in the G* value at 64°C and 10 rad/s were used as thresholds for detecting significant differences when interpreting the AIP results. In compacted specimen aging, the oxygen diffusion from the periphery to the specimen center is impeded by the binder film and aggregate, thus leading to the high possibility of an oxidation gradient within specimens aged in a compacted state. The extraction and recovery of binder from different distances from the periphery of compacted specimens following long-term aging 0E+00 1E+04 2E+04 3E+04 4E+04 5E+04 6E+04 7E+04 0E+00 2E+04 4E+04 6E+04 8E+04 1E+05 Cu m ul at iv e (1- C) Nf (Cycle) Short-Term Aged, Compacted at 144°C Oven-Aged Loose Mix, 85°C, 8 Days, Compacted at 144°C Oven-Aged Loose Mix, 85°C, 8 Days, Compacted at 157°C Oven-Aged Compacted Spec., 85°C, 8 Days PAV-Aged Compacted Spec., 300 kPa, 85°C, 1 Day Figure 17. DR failure criterion lines of NC mix subjected to different aging conditions (Nf = number of load applications to failure). Mix ID Average DR Short-Term Aged, Compacted at 144°C 0.574 Oven-Aged Loose Mix, 85°C, 8 Days, Compacted at 144°C 0.541 Oven-Aged Loose Mix, 85°C, 8 Days, Compacted at 157°C 0.535 Oven-Aged Compacted Spec., 85°C, 8 Days 0.494 PAV-Aged Compacted Spec., 300 kPa, 85°C, 1 Day 0.454 Table 8. Average DR values for NC mix subjected to different aging conditions.

36 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction allows for the assessment of aging gradients. In this study, this assessment was accomplished by first coring and cutting the specimens into slices from which the binder was recovered. For the small specimens, the outer segments were obtained by slicing the specimens vertically, as depicted in Figure 18. The large specimens were cored using a 38-mm core bit and 75-mm core bit to obtain radial slices, as shown in Figure 19. Binder extraction and recovery was carried out only for the outer layer, middle layer, and core of the oven-aged large specimens. Figure 20 presents the results of the AIP tests for the various aging trials. The data shown in Figure 20 were further processed to calculate percent changes in C + S absorbance and G* between different aging conditions. These data are summarized in Table 9. Cells highlighted in green indicate conditions that show the percent change in G* values is less than 15%, meaning that the conditions used in calculating the percent change do not result in significant differences in material properties. The 15% criterion used in the comparison resulted from the sensitivity study, where 15% change in binder G* resulted in approximately 10% change in mixture |E*|. The results first demonstrate that the level of oxidation in the short-term aged mixture surpasses the level of oxidation in the RTFO-aged binder. The results shown in Table 9 indi- cate that the C + S absorbance and G* values of the long-term aged compacted specimens are Figure 18. Schematic of components of small specimen (38 mm ë 100 mm) used to evaluate oxidation gradients. Figure 19. Schematic of components of large specimen (100 mm ë 150 mm) used to evaluate oxidation gradients.

Findings and Applications 37 0.0635 0.0918 0.0746 0.0885 0.0905 0.0871 0.0868 0.0823 0.0801 0.0850 0.0926 0.0972 4.45 4.93 4.59 4.75 4.84 4.80 4.84 4.63 4.71 4.82 5.00 4.94 3.0 3.5 4.0 4.5 5.0 5.5 6.0 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 Ca rb on yl + S ul fo xi de P ea k (A U) Carbonyl+Sulfoxide Peak log G* lo g G * a t 6 4° C, 1 0 Hz (P a) PAV-Aged Small Specimen 85°C, 3 days Oven-Aged Small Specimen 85°C, 8 days Oven-Aged Large Specimen 85°C, 8 days PAV and Oven- Aged Loose Mix 85°C STAAged Binders 0.0635 0.0918 0.0746 0.0885 0.0905 0.0871 0.0868 0.0823 0.0801 0.0850 0.0926 0.0972 3.79 4.34 3.96 4.15 4.24 4.19 4.23 4.00 4.09 4.21 4.43 4.34 3.0 3.5 4.0 4.5 5.0 5.5 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 Ca rb on yl + S ul fo xi de P ea k (A U) Carbonyl+Sulfoxide Peak log G* lo g G *, 64 °C , 1 0 ra d/ s (P a) PAV-Aged Small Specimen Oven-Aged Small Specimen Oven-Aged Large Specimen PAV and Oven- Aged Loose MixSTAAged Binders Figure 20. Comparison between C + S absorbance peaks and log G* at 64°C and 10 rad/s for extracted and recovered binders from different compacted specimen aging processes, loose mix aging trials, and aged binders.

38 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction significantly higher than those of the short-term aged loose mixtures based on the previously defined thresholds. However, the AIP values of all the long-term aged compacted specimens are lower than those for the RTFO+PAV-aged binder. The AIP results of binders that were extracted and recovered from the large compacted speci- mens aged for 8 days at 85°C indicate greater levels of oxidation at the outer portion of the specimen than the core; this indicates the presence of an oxidation gradient. However, the AIP values corresponding to the outer segments and the cores of the long-term aged small specimens are similar, thus confirming that shorter diffusion paths can mitigate oxidation gradient con- cerns. The AIP values of binder extracted from the small specimens that were aged in the oven are relatively similar to the outer portion of large specimens aged under the same conditions. Based on the AIP results, the small specimens aged in the PAV experienced a higher level of oxidation than the oven-aged specimens, indicating that pressure expedites the oxidation of mixtures. However, for the small specimens aged in the PAV (3 days at 85°C and 300 kPa), a significant aging gradient is observed based on the G* results, thereby indicating that pressure does not alleviate the effects of an oxidation gradient. The long-term aging of the loose mixture in an oven yields AIP values that exceeded those of the compacted specimens aged for the same duration. Thus, aging loose mix appears to expedite oxidation significantly compared to aging compacted specimens. The AIP values of the long- term aged loose mixtures also exceeded the value of the RTFO+PAV binder. The long-term aging of the loose mix in the PAV for 2 days at 85°C and 2.1 MPa led to a level of oxidation relatively similar to that of the loose mix oven aging for 8 days at the same temperature. These results suggest that the addition of pressure can expedite aging almost four times faster than a conventional aging oven. However, it is important to note that the standard binder PAV does not allow enough space for aging sufficient quantities of material for performance testing in a single trial. Thus, if PAV aging of loose mix were to be adopted, a new and larger PAV would have to be developed. Summary of Compacted Specimen and Loose Mix Aging Compacted Specimen Aging. Although no integrity issues in terms of changes in air void content or specimen dimensions were encountered with the AASHTO R 30 protocol, the aging of large (100-mm diameter) compacted specimens with wire mesh support in an oven at 85°C led to the development of an aging gradient within the specimens. This lack of uniform properties throughout the specimen is of concern for performance testing and was observed directly as a % Change Between RTFO & PAV Between STA & RTFO+PAV Binder for Aged Binder, Outer Layer for Compacted Specimens, and Loose Mix Between Outer Layer & Core Between Outer Layer & Mid Layer Aged Binder PAV-Aged Small Specimen Oven-Aged Small Specimen Oven-Aged Large Specimen Oven-Aged Loose Mix PAV-Aged Loose Mix Note: STA = short-term aged. G* 139.9 90.5 86.2 77.8 G* 18.7 8.8 38.3 195.1 139.9 G* 24.1 C + S 5.8 C + S 2.2 0.3 3.2 C + S 23.1 21.3 16.4 13.9 24.1 30.3 G* 254.8 C + S 44.6 Table 9. Percentage change in chemical and rheological AIPs according to the different aging methods.

Findings and Applications 39 high rate of failure at the end locations where oxidation was most significant in the cyclic direct tension fatigue tests. However, the high rate of end failure could be overcome by aging specimens 178 mm in height and then trimming the ends to produce a 130-mm tall specimen for testing. The aging gradient observed in the large compacted specimens subjected to oven aging was eliminated by using small specimens (38-mm diameter with 100-mm height) due to the shorter diffusion paths of the smaller specimens. The application of pressure in compacted specimen aging was found to expedite aging. How- ever, oxidation gradients were observed in the pressure-aged specimens. In addition, although no changes in air void contents or specimen dimensions were induced by pressurized aging, the performance test results indicate that the application and/or release of pressure can damage specimens. Therefore, the results indicate that the most promising method for aging compacted specimens is to age small specimens in an oven without pressure. Loose Mix Aging. The primary concern associated with loose mixture aging is the ability to compact the material after long-term aging. However, in this study, the compaction of the NC mix after 8 days of oven conditioning at 85°C was possible with no adjustment to the compac- tion temperature. A similar number of compaction gyrations was required for both the short- and long-term aged loose mixes. The image analysis of the aggregate structure also indicated comparable compaction of both the short- and long-term aged loose materials. Furthermore, the performance test results indicate a significant increase in the dynamic modulus values of the long-term aged material compared to those of the short-term aged material, and the fatigue performance test results indicate no integrity concerns. In addition, loose mix aging exposes a large surface area of the binder to oxygen, and thus, a faster rate of oxidation was observed in the loose mix oven aging compared to the compacted specimen aging based on the measured asphalt binder chemical and rheological aging index values. The application of pressure also was found to expedite the oxidation of the loose mixes. However, only 500 g of loose mix could be aged at one time in the binder PAV. One Superpave gyratory-compacted specimen requires the prepara- tion of 7000 g to 8000 g of loose mix. Thus, the binder PAV would need to be run approximately 15 times to generate enough loose mix to prepare a compacted specimen for performance test- ing, which is inefficient and therefore was deemed impractical. Again, selection of pressure aging of loose mix would necessitate development of a new, larger PAV. Selection of Aging Procedure Based on the findings for the compacted and loose mix aging trials with and without pressure, the research team identified oven aging of loose mix as the optimal aging procedure. To avoid specimen integrity issues, the only option for the oven aging of compacted specimens is to age small specimens in an oven. This small specimen geometry allows for only dynamic modulus and direct tension testing. Other tests (e.g., permanent deformation tests) would not be possible under this scenario. Loose mix aging is more versatile than compacted specimen aging in that any specimen geometry (e.g., slabs or beams) can be produced using aged loose mix. In addition, loose mix aging leads to faster oxidation than compacted specimen aging and therefore offers efficiency gains. As discussed, the PAV aging of loose mix would require the development of a new, larger, mixture-specific PAV, which would be costly. Thus, given that aging loose mix can be accomplished relatively quickly in an oven using multiple pans, this method is considered a more practical approach at present. In this research, a conventional oven with inner chamber dimensions of 36 in. × 24 in. × 19 in. (W × H × D) was used. Using only six shelves, 18 pans (13 in. × 18 in. × 1 in.) can fit inside the oven. Loose mix spread in four pans is sufficient for the preparation of one Superpave gyratory-compacted specimen that is 150 mm in diameter and 178 mm in height. Also, the preliminary results indicate that the compaction of aged loose mix can be accomplished using the same temperature that is required to compact short-term aged material.

40 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction Evaluation of the Selected Aging Procedure To verify the findings of the loose mix aging trials with the NC mix, the FHWA ALF SBS- modified mixture was subjected to loose mix oven aging trials at 70°C, 85°C, and 95°C. Small samples were removed periodically from the oven and subjected to extraction and recovery to determine the binder AIPs. In order to compare the level of oxidation achieved in the laboratory to that found under field conditions, the AIPs determined for binder extracted from the aged loose mix were compared with the AIPs determined for the asphalt binder extracted from the top lift of a field core extracted after 8 years in service. The results were used to approximate the oven condi- tioning time needed for the loose mix to reach the oxidation level of the field core at various depths. Two aging procedures were used to evaluate the integrity of the specimens that were compacted after loose mix aging: 8 days of conditioning at 85°C (consistent with the NC mix aging trials) and 21 days of conditioning at 95°C, which was found to correspond to the same oxidation level as the surface of the 8-year-old field core. The conditioning temperature of 95°C was selected because it can expedite aging faster than 85°C but is not expected to lead to volatilization or degradation of the polymers (Petersen 2009). Note that the asphalt binder oxidation level at the surface of the field core is thought to represent an “extreme” level of aging, as the surface of the field core was found to be severely oxidized compared to samples extracted from deeper within the pavement. Following compaction, the aged loose mix was evaluated using the same protocol utilized for the NC mix, which included an assessment of compactability and performance testing. Results. The 8-year-old field core obtained from the FHWA ALF was cut (sawn) to obtain 3 half-inch-thick slices. The asphalt binder was extracted and recovered from the field core slices, and then the AIPs were determined. The results were used to evaluate the oxidation level of the field core in terms of depth for comparison to the oxidation levels of the laboratory-aged samples. A significant oxidation gradient was found within the field core, with the surface of the field core being more severely oxidized than from deeper within the pavement. Figure 21 provides a graphic representation of the estimated oxidation levels for loose mix aging at 95°C that were needed to match the level of oxidation of the field core at different depths. Even at 95°C, 21 days of oven conditioning were needed to match the level of oxidation of the surface of the field core. However, it is worth noting that the smallest sample utilized in the performance tests of asphalt concrete is 38 mm in diameter or thickness. If a 38-mm specimen is obtained from the -45 -40 -35 -30 -25 -20 -15 -10 -5 0 0.112 D ep th (m m) Carbonyl + Sulfoxides Absorbance Peaks (AU) 21 Days of Aging at 95°C 15 Days of Aging at 95°C 13 Days of Aging at 95°C Tack Coat Layer First Slice of Top Layer Second Slice of Top Layer Third Slice of Top Layer 0.104 0.106 0.108 0.11 Figure 21. Comparison between field core aging gradient with respect to depth and long-term aging of loose mix in oven at 95°C.

Findings and Applications 41 top surface of a field core, the material in the top 19 mm of the specimen has aged more than the material at the 19 mm depth. Also, the material in the bottom 19 mm of the specimen has aged less than the material at the 19 mm depth. Therefore, it is reasonable to assume that the representative aging level for a depth of 38 mm would be close to the oxidation level at a depth of 19 mm. This explanation would indicate that 15 days are required for aging loose mix to match the age level of the top 38 mm of an 8-year-old field core extracted in Virginia. Additionally, for structural modeling and analysis purposes (e.g., using the FlexPAVE™ program), typically the averaged properties of each layer are used as inputs. Therefore, for this study, laboratory-prepared samples were aged in order to meet the average level of aging of each layer (in this case, 15 days of aging at 95°C). Based on these results, the oxidation level of the surface is considered to be a severe condition, because oxidation greatly dissipates with depth. Thus, loose mix aged at 95°C for 21 days is considered an “extreme” condition for evaluating the compactability and integrity of aged loose mix. Figure 22 shows comparisons of the degree of aging, using the C + S absorbance peaks and G* AIPs, among the aged loose mixes used for the compacted specimen integrity assessment, the short-term aged mix, and binders aged in the RTFO and PAV. The results demonstrate that 21 days of loose mix conditioning at 95°C greatly exceeds the oxidation level of the binder PAV, which is comparable to 8 days of loose mix aging at 85°C. The compactability of the long-term aged FHWA ALF-SBS–modified mix was evaluated as an initial specimen integrity check by comparing the number of gyrations needed to reach the target air void content with the number needed for short-term aged materials. The number of gyrations needed to reach the target air void contents was similar for both the short-term aged loose mixture and the two levels of long-term aged loose mixture (8 days at 85°C and 21 days at 95°C) with no adjustment of the compaction temperature. Thus, compacting the long-term aged loose mix was not problematic. Performance testing was utilized as an additional means to evaluate the integrity of the FHWA ALF-SBS–modified specimens that were compacted following long-term loose mix aging. 0.0577 0.0875 0.0705 0.0923 0.1188 0.1111 3.85 4.37 3.80 4.28 5.18 4.98 0.0 2.0 4.0 6.0 8.0 0.00 0.04 0.08 0.12 0.16 lo g G* a t 6 4° C, 1 0 ra d/ s (P a ) Carbonyl + Sulfoxide Peak log G* Ca rb o n yl + S ul fo x id e P ea k (A U ) Oven Aged Loose MixShort-termAged Mix Aged Binder Field Core Top 6 mm Figure 22. Comparisons between C + S absorbance peaks and log G* at 64°C and 10 rad/s for extracted and recovered binders from different loose mix aging trials, field core top layer surface slices, and aged binders (SBS-modified mixture).

42 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction Comparisons between the dynamic modulus and cyclic fatigue damage characteristic curves of the short-term aged material and the long-term aged material were used to evaluate the specimen integrity of the compacted long-term aged material based on expected trends of increased aging. Figure 23 presents the dynamic modulus master curves. The master curves represent the aver- aged values of two replicates. The results indicate that the oven-aged loose mix specimens have higher dynamic modulus values than the short-term aged specimens. Furthermore, the results suggest that the specimens that were compacted after 21 days of oven aging at 95°C have signifi- cantly higher dynamic modulus values than the specimens compacted after 8 days of oven aging at 85°C, as was expected based on the AIPs. If severe integrity problems had been present in the long-term aged specimens, then the dynamic modulus values would not be significantly higher than those of the short-term aged specimens, regardless of the aging level (because damage in the specimen reduces the dynamic modulus value), and thus, no integrity problems were detected in the dynamic modulus test results. Figure 24 presents the cyclic fatigue characteristic curves for the short- and long-term aged materials. These curves define how damage grows in a material and represent the averaged 2E+02 2E+03 2E+04 2E+05 |E* | (M Pa ) Reduced Frequency (Hz) Short-term Aged Oven-Aged Loose Mix, 85°C, 8 days Oven-Aged Loose Mix, 95°C, 21 days 1E+02 1E+04 2E+04 3E+04 |E* | (M Pa ) Reduced Frequency (Hz) (a) (b) Short-term Aged Oven-Aged Loose Mix, 85°C, 8 days Oven-Aged Loose Mix, 95°C, 21 days 1E-08 1E+011E-021E-05 1E-08 1E+011E-021E-05 Figure 23. Dynamic modulus test results: (a) log-log scale and (b) semi-log scale. 0.0 0.2 0.4 0.6 0.8 1.0 C S Short-term Aged Oven, Loose Mix, 85°C, 8 days Oven, Loose Mix, 95°C, 21 days Failure Point 0.0E+00 2.0E+061.5E+061.0E+065.0E+05 Figure 24. Comparison of damage characteristic curves from FHWA ALF SBS mixtures subjected to different aging conditions.

Findings and Applications 43 results of three tests that were conducted using various cross-head displacement amplitudes. The long-term aged specimens have higher damage characteristic curves than the short-term aged specimen, which follows expected trends. Thus, the performance test results indicate that no integrity issues are associated with the compaction of the aged loose mix. Figure 25 shows the DR energy-based failure criterion results for the SBS-modified mix. The failure criterion line for the specimens that were compacted following long-term aging for 21 days at 95°C falls significantly lower than the failure criterion lines for the short-term aged specimens and the specimens aged for 8 days at 85°C. This outcome suggests that the severe level of aging led to embrittlement and consequently degraded the resistance to fatigue and, hence, suggests no integrity problems. These results indicate that the compaction of long-term aged loose mix is possible with no adjustment to the compaction temperature, based on both the number of compaction gyrations required to reach the target air void contents and the performance test results. The long-term aged loose mix that was aged at 95°C for 21 days had an oxidation level that was equivalent to that of the surface of an 8-year-old field core obtained from the FHWA ALF in McLean, Virginia. This level is assumed to represent an extreme oxidation level that nonetheless allows for com- paction of the mix. Additional details on the selection of the aging procedure are included in Appendix F and elsewhere (Elwardany et al. 2017b). Selection of Laboratory Aging Temperature Petersen (2009) suggested that aging temperatures that exceed 100°C can induce chemical changes to the oxidation reaction in asphalt materials. To evaluate the performance impli- cations of long-term aging temperatures below and above 100°C, comparative tests between loose mixtures aged at 95°C and 135°C were conducted to evaluate the implications of loose mix aging at 135°C in terms of asphalt mixture performance using three mixtures, all prepared with the same FHWA ALF aggregate, but with different binders: ALF-SBS, SHRP AAD, and SHRP AAG. The SHRP AAD and SHRP AAG binders were selected due to their known dif- ferences in chemistry. The SHRP AAD binder has a high sulfur content (6.9%) and is highly structured (i.e., its components exhibit a high degree of incompatibility). Thus, the SHRP AAD binder was expected to be especially susceptible to changes in oxidation kinetics and mechanics at 135°C. The SHRP AAG binder has a low sulfur content (1.3%) and is less structured (more compatible) than SHRP AAD. Thus, the SHRP AAG binder was expected to be less susceptible to changes in oxidation kinetics and mechanisms when the temperature for loose mixture aging was 0E+00 1E+04 2E+04 3E+04 4E+04 5E+04 0E+00 2E+04 4E+04 6E+04 8E+04 C u m u la tiv e (1 -C ) Nf (Cycle) Short-Term Aged Oven-Aged Loose Mix, 85°C, 8 days Oven-Aged Loose Mix, 95°C, 21 days Figure 25. Comparison of DR failure criterion lines for SBS-modified mixtures.

44 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction increased from 95°C to 135°C. The FHWA ALF-SBS–modified mixture was selected in order to include a common asphalt-modified asphalt binder in the study and because field core data were available for it. Figure 26 presents a summary of the experimental plan that was implemented for each mixture to evaluate the implications of loose mixture aging at 95°C and 135°C. Note that the rheological AIP used in the selection of the laboratory aging temperature is G* at 64°C and 10 Hz, which differs from the selected rheological AIP of G* at 64°C and 10 rad/s. However, the use of G* at 64°C and 10 Hz versus G* at 64°C and 10 rad/s is not anticipated to affect the findings related to the selection of the laboratory aging temperature. First, two batches of loose mix were short-term aged at 135°C for 4 hours and then subjected to long-term aging at 95°C and 135°C using the proposed aging method. Small samples of the loose mixture were taken from the pans at periodic intervals for binder extraction and recovery. The changes in the asphalt binder oxidation level versus the aging duration were assessed for the two aging temperatures by means of chemical and rheological AIPs. Figure 27 shows the relationship between the G* value at 64°C and 10 Hz frequency and the C + S absorbance peaks for the binder samples extracted and recovered from the SHRP AAD mix aged at 95°C and 135°C. Figure 27 indicates that the binders aged at the two dif- ferent temperatures have different C + S absorbance peaks for the same G* value, indicating that a change in the oxidation reaction mechanism occurred when the aging temperature was increased from 95°C to 135°C. An analogous trend was observed for the FHWA ALF- SBS mixture and, to a lesser extent, the SHRP AAG mixture. The G* value was selected as the aging index value that matched the degree of aging between the two aging temperatures because binder rheology is speculated to be related to asphalt mixture performance more directly than binder chemistry. Also, the selection of G* as the primary aging index allowed the effects of different C + S absorbance peaks on mixture performance to be evaluated, Performance Testing (Dynamic Modulus and Cyclic Fatigue) Oven Aging of Loose Mix at 95 °C for Different Durations Oven Aging of Loose Mix at 135 °C for Different Durations FTIR-ATR TestingFTIR-ATR Testing Binder Extraction and Recovery Binder Extraction and Recovery DSR TestingDSR Testing Determining the Required Aging Durations at 95 °C and 135 °C to Match the Specified G* Values at 64 °C and 10 Hz Oven Aging of Loose Mix at 95 °C to Match Target G* Performance Testing (Dynamic Modulus and Cyclic Fatigue) Oven Aging of Loose Mix at 135 °C to Match Target G* Comparing the Test Results Figure 26. Summary of experimental plan.

Findings and Applications 45 because the same G* value for mixtures aged at two different temperatures will result in two different C + S absorbance peaks. The next step in the experimental plan was to determine the aging durations for the samples tested at 95°C and 135°C that would yield the same G* value. For the 95°C aging, 21 days of con- ditioning was determined for the FHWA ALF-SBS–modified mixture. Note that 21 days of con- ditioning at 95°C led to an equivalent oxidation level for the top 6 mm of an 8-year-old field core obtained from McLean, Virginia. The G* value for 21 days of conditioning at 95°C was deter- mined to be approximately four times the G* value for PAV-aged asphalt binder. Field cores that corresponded to the SHRP AAD and SHRP AAG mixtures were not available. Therefore, based on the finding for the FHWA ALF-SBS–modified mixture, four times the G* value of the PAV- aged binder was used as the target G* value that would reflect reasonable field aging levels of the pavement surface after a prolonged in-service period. Thus, the required aging durations at 95°C and 135°C to achieve G* values equal to four times that of PAV-aged binder were determined and used to evaluate the effects of aging at 135°C for both the SHRP AAD and SHRP AAG mixtures. Figure 28, Figure 29, and Figure 30 show the procedures that were used to match the aging levels between the loose mixtures aged at 95°C and at 135°C for the FHWA ALF-SBS, SHRP AAD, 0 0.5 1 1.5 2 2.5 3 3.5 4 lo g G * a t 6 4° C, 1 0 Hz (kP a) Short-Term Aged Loose Mix, 95°C, UC Loose Mix, 135°C, UC 0.120.07 0.08 0.09 0.1 0.11 Carbonyl + Sulfoxide Absorbance Peaks (AU) Figure 27. SHRP AAD loose mix prepared for long-term aging. G* = 57.068e0.0034 (H) G* = 43.1e0.0431 (H) 1E+01 1E+02 1E+03 1E+04 0 200 400 600 800 G *, 64 °C , 1 0 Hz (k Pa ) Aging Duration (Hours) Short-term Aged Loose Mix, 95°C, UC Loose Mix, 135°C, UC 52 H 21 D Target G* Figure 28. Determination of FHWA ALF-SBS aging durations.

46 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction and SHRP AAG materials, respectively. To match the aging levels between the loose mixtures aged at 95°C and 135°C, first the relationship between the log G* and aging duration was obtained for each aging temperature. Then, in the case of the FHWA ALF-SBS material, the G* value that corresponded to 95°C before compaction was determined. Based on the relationship between log G* and the time that corresponded to the loose mix aged at 135°C, the required duration of aging at 135°C to match this G* value was then determined via interpolation. For the SHRP AAD and SHRP AAG materials, the aging durations required to match the predetermined G* values, which corresponded to four times the value of G* for the binder after PAV aging (the target G*), at both 95°C and 135°C were determined. Note that the results presented in Fig- ure 29 and Figure 30 demonstrate that the SHRP AAD mix required shorter aging times at both 95°C and 135°C to achieve the requisite G* values than the SHRP AAG mix. This finding matches expectations in terms of the binders’ microstructures. That is, the SHRP AAD mix is incompatible and therefore was expected to have a high level of hardening susceptibility with the oxidation level compared to the SHRP AAG mix, which is compatible and therefore less structured. Comparisons of the performance of the mixtures prepared by aging at 95°C and 135°C (with equivalent binder G* values) were used to assess the performance implications of the aging temperatures. Note that to build the dynamic modulus master curves, frequency sweep tests G* = 52.818e0.0065 (H) G* = 65.4010e0.0700 (H) 0 200 400 600 16.8 H 8.9 D 1E+01 1E+02 1E+03 1E+04 1E+05 G *, 64 °C , 1 0 Hz (k Pa ) Aging Duration (Hours) Short-term Aged Loose Mix, 95°C, UC Loose Mix, 135°C, UC Target G* Figure 29. Determination of SHRP AAG aging durations. G* = 13.176e0.0055 (H) G* = 13.9933e0.0644 (H) 0 200 400 600 37.6 H 19 D 1E+01 1E+02 1E+03 1E+04 1E+05 G *, 64 °C , 1 0 Hz (k Pa ) Aging Duration (Hours) Short-term Aged Loose Mix, 95°C, UC Loose Mix, 135°C, UC Target G* Figure 30. Determination of SHRP AAD aging durations.

Findings and Applications 47 were conducted at multiple temperatures. In addition to material testing, the effect of the aging temperature on fatigue performance was evaluated at the pavement level. The pavement perfor- mance was predicted using FlexPAVE™. FHWA ALF-SBS. Figure 31 presents the mixture performance test results for the FHWA ALF-SBS mixture. Figure 31 (a) presents the FHWA ALF-SBS mixture dynamic modulus master curves that correspond to the specimens fabricated after aging at 95°C for 21 days and 135°C for 52 hours. The dynamic modulus test results for the specimens fabricated after short-term aging only also are provided for reference. The results indicate a significant increase in dynamic mod- ulus values with long-term aging. The results also indicate very little difference in the dynamic modulus master curves that correspond to mixtures aged at 95°C for 21 days and 135°C for 52 hours, suggesting that the chemical changes, represented by the C + S absorbance peaks, induced by aging at 135°C do not have a significant effect on the mixture dynamic modulus of the FHWA ALF-SBS mixture. Differences in the time–temperature shift factors between the short-term and long-term aged materials led to the differences in the reduced frequency range observed in the corresponding master curves. It can be seen that the data obtained from testing long-term aged specimens cover a larger reduced frequency domain than the short-term aged data, whereas the specimens that correspond to both conditions were tested over the same range of temperatures. Direct tension cyclic tests were performed on the FHWA ALF-SBS mixture specimens that are aged at 95°C for 21 days and at 135°C for 52 hours to assess the implications of aging temperature with regard to fatigue cracking performance. The fatigue tests were conducted using three cross- head displacement levels (low, intermediate, and high) that were selected based on the displace- ment of the test machine’s actuator, thus resulting in different on-specimen strain levels. Figure 31 Short-term Aged Oven, Loose Mix, 95°C, 21 days Oven, Loose Mix, 135°C, 52 hours 1E+2 1E+3 1E+4 1E+5 1E-8 1E-4 1E+0 1E+4 |E* | (M Pa ) Reduced Frequency (Hz) 0.0 0.2 0.4 0.6 0.8 1.0 0E+0 3E+5 6E+5 9E+5 1E+6 2E+6 C S 0E+00 1E+04 2E+04 3E+04 4E+04 5E+04 0E+00 2E+04 4E+04 6E+04 8E+04 Cu m ul at iv e (1- C) Nf (Cycle) (a) (b) (c) Figure 31. FHWA ALF-SBS mixture performance test results: (a) dynamic modulus curves, (b) C versus S curves, and (c) DR failure criterion lines.

48 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction (b) presents the C versus S curves. The results show similar C versus S curves for the two long- term aging conditions, whereas the short-term aging C versus S curves are considerably lower, which is attributable to the short-term aged mixtures’ lower stiffness values (age level) compared to those of the long-term aged material. These results indicate no significant effect of aging temperature on the fatigue performance of the FHWA ALF-SBS mixture. In addition to the C versus S curves, the DR failure criterion was evaluated in this study. Fig- ure 31 (c) presents the failure criterion results for the FHWA ALF-SBS mixture, showing that the failure criterion lines for the long-term aged materials fall below the line for the short-term aged material, which is indicative of the lower fatigue resistance of long-term aged materials as a result of the brittleness caused by oxidative aging in the asphalt binder. However, the results indicate very similar failure criterion lines for the specimens prepared with loose mixture aged at 95°C for 21 days and at 135°C for 52 hours. These results suggest that the chemical changes induced by aging at 135°C do not signifi- cantly affect the performance of the FHWA ALF-SBS mixture. The FHWA ALF-SBS mixture contains SBS modification, which may mask the effects of microstructural changes induced by aging at 135°C. Hence, the evaluation of asphalt mixtures that contain unmodified asphalt binders (i.e., SHRP AAD and SHRP AAG) is also important. Table 10 presents the statistical t-test analysis outcomes for the dynamic modulus and fatigue test results that correspond to the FHWA ALF-SBS materials with different aging conditions. For each pair of compared samples, the data points that correspond to the selected reduced frequencies or S values were used for the analysis. The selection of the reduced frequencies and S values was based on the ranges of both the dynamic modulus master curves and C versus S curves. Very high reduced frequencies were avoided due the fact that the aging effect on the dynamic modulus is more pronounced at low frequencies or high temperatures. Also, very low S values were not considered because the properties of asphalt mixture specimens are more distinguishable at higher S values (or smaller C values) when different materials are compared with each other. The data points for two dynamic modulus test replicates and three fatigue test replicates were used for the statistical analysis. As mentioned earlier, a confidence level of 95% was employed to evaluate the difference between the pairs of aging treatments. Dynamic Modulus Sample Reduced Frequency 2.0E-06 1.0E-04 1.0E-03 1.0E-01 p-value Short-term aged Loose mix, 95°C, 21 days 0.0263 0.0014 0.0037 0.0201 Short-term aged Loose mix, 135°C, 52 hours 0.0061 0.0028 0.0051 0.0150 Loose mix, 95°C, 21 days Loose mix, 135°C, 52 hours 0.3529 0.1453 0.0591 0.3346 Cyclic Fatigue Sample S 1.0E+05 3.0E+05 5.0E+05 7.0E+05 p-value Short-term aged Loose mix, 95°C, 21 days 0.0005 0.0001 0.0006 0.0025 Short-term aged Loose mix, 135°C, 52 hours 0.0078 0.0004 0.0004 0.0029 Loose mix, 95°C, 21 days Loose mix, 135°C, 52 hours 0.3599 0.0918 0.1230 0.4560 Table 10. Statistical t-test analysis outcomes for dynamic modulus and cyclic fatigue test results for FHWA ALF-SBS materials.

Findings and Applications 49 The pairs with significant differences (p < 0.05) are shaded in Table 10. The statistical anal- ysis results suggest a significant difference between the dynamic modulus test results for the short-term aged and long-term aged materials. However, as inferred from the comparison of the master curves, no significant difference is seen between the ALF-SBS materials aged at 95°C and at 135°C. The p-values for the C versus S curves indicate a similar conclusion. The material properties shown in Figure 31 were input to FlexPAVE™ to predict the cracking performance of the study pavement. Considering the different positions of the failure criterion lines between the short-term aged and long-term aged conditions, very different field perfor- mance predictions were anticipated for these two conditions. A simple pavement structure was considered: a 10-cm asphalt concrete layer over a 20-cm aggregate base and 380-mm subgrade. For each binder type, based on the reported PG, different EICM data were selected for the performance predictions. EICM data for Washington, D.C., were used for the ALF SBS material, and climatic data for Ann Arbor, Michigan, and San Luis Obispo, California, were used for the SHRP AAD and SHRP AAG materials, respectively. The traffic input was 3,500 daily equivalent single-axle loads (ESALs). The analysis was performed for a 20-year service life. FlexPAVE™ was used to predict the distribution of damage within a cross-section of the asphalt pavement layer after 20 years of traffic loading. The wheel path was directly above the region of damage localization. To compare the results of the fatigue crack- ing predictions, the percentage of damage (referred to as “percent damage”) was computed as a function of time for each case. Percent damage is defined as the ratio of the sum of the damage factors within the reference cross-section area to the reference cross-section area itself, as shown in Equation 14. (Note: the percent damage area is defined schematically in Figure 33, also.) (14) 1 1 Percent Damage N N A A f i ii M ii M ∑ ∑=     ×= = where i = nodal point number in finite element mesh, M = total number of nodal points in finite element mesh, N/Nf = damage factor, N = number of load applications, Nf = number of load applications to failure, Ai = area represented by nodal point i in finite element mesh, and ΣAi = reference area. Figure 32 shows the distribution of damage within a cross-section of the asphalt concrete layer of the pavement for different aging treatments predicted using the FlexPAVE™ program. This figure suggests no significant differences between the distribution of damage predicted for asphalt loose mixture aged at 95°C compared to that aged at 135°C. However, as expected, the short-term aged material performed better than the long-term aged material. It is also noted that the long-term aging of the SBS mixture increases the top-down cracking propensity of the study pavement greatly. Although this observation is based on the unrealistic aging condition (i.e., constant aging through the thickness of asphalt layer), it signifies the importance of including accurate aging condition in LTPP prediction. FlexPAVE™ employs two overlapping triangles to form the reference cross-section area within which the damage evolution can be considered (Kim et al. 2017), as shown in Figure 33. The top inverted triangle has a 170-cm wide base that is located at the top of the surface layer and a vertex that is located at the bottom of the bottom asphalt layer. The 12-cm wide base of the

50 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction second triangle is located at the bottom of the bottom asphalt layer and its vertex is positioned at the surface layer. Figure 33 presents the final shape of these overlapping triangles, which defines the reference area for percent damage. FlexPAVE™ program output data were used to calculate the fatigue damage area as a function of traffic load repetition. Figure 34 presents the fatigue damage area versus the service life for the three aging treatments. A significant difference can be observed between the fatigue performance of the short-term and long-term aged materials, whereas the results for the two long-term aging treatments at 95°C and 135°C do not indicate any significant differences. These observations are in agreement with the fatigue performance data. SHRP AAD. Figure 35 presents the mixture performance test results for the SHRP AAD mix- ture. Figure 35 (a) presents the SHRP AAD mixture dynamic modulus master curves that cor- respond to the specimens fabricated after aging at 95°C and 135°C for 8.9 days and 16.8 hours, respectively. The dynamic modulus test results of the specimens fabricated after short-term aging only are provided also for reference. The results indicate a significant increase in the dynamic Short-Term Aging –0.1 –0.09 –0.08 Z (m ) –0.07 –0.06 –0.05 –0.04 –0.03 –0.02 –0.01 0 10 9 8 Z (cm ) 7 6 5 4 3 2 1 0 0 1.510.50 X (m) –0.5–1–1.51.510.50X (m) –0.5–1–1.5 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 –0.1 –0.09 –0.08 Z (m ) –0.07 –0.06 –0.05 –0.04 –0.03 –0.02 –0.01 0 0 1.510.50 X (m) –0.5–1–1.5 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Loose Mix, 95°C, 21 Days Loose Mix, 135°C, 52 Hr Figure 32. Damage contours for FHWA ALF-SBS mixture aged at different conditions. 170 cm 120 cm Top Layer Intermediate Layer Bottom Layer Figure 33. Area for percent damage definition.

Findings and Applications 51 modulus value with long-term aging. The results also indicate a significant difference in the dynamic modulus master curves that correspond to the mixtures aged at 95°C for 8.9 days and at 135°C for 16.8 hours, despite the mixtures’ equivalent binder rheology. A 10% threshold was used to determine the significance of difference between the dynamic modulus test results at the intermediate and high temperatures. The mixture aged at 135°C shows a reduction in modu- lus value compared to the material aged at 95°C. These results suggest that the chemical changes and/or other effects (e.g., absorption, drain-down) induced by aging at 135°C had a significant effect on the performance of the SHRP AAD mixture. Note that the SHRP AAD binder has high 0 10 20 30 40 50 60 70 0 50 100 150 200 250 Pe rc en t D am ag e Month Short-Term Loose Mix, 95°C, 21 Days Loose Mix, 135°C, 52 Hr Figure 34. Comparison of fatigue damage area versus service life for FHWA ALF-SBS mixture aged at different conditions. 1E+2 1E+3 1E+4 1E+5 1E-8 1E-6 1E-4 1E-2 1E+0 1E+2 |E* | (M Pa ) Reduced Frequency (Hz) 0.0E+0 4.0E+5 8.0E+5 1.2E+6 0E+00 5E+04 1E+05 2E+05 Short-term Aged Oven, Loose Mix, 95°C, 8.9 days Oven, Loose Mix, 135°C, 16.8 hours 0.0 0.2 0.4 0.6 0.8 1.0 C S 0E+00 1E+04 2E+04 3E+04 4E+04 5E+04 Cu m ul at iv e (1- C) Nf (Cycle) (a) (b) (c) Figure 35. SHRP AAD mixture performance test results: (a) dynamic modulus curves, (b) C versus S curves, and (c) DR failure criterion lines.

52 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction sulfur content and is highly structured, indicating high potential for changes between the oxidation products of laboratory aging at 135°C versus 95°C. Figure 35 (b) presents the damage characteristic curves for the SHRP AAD mixture. The results show that the damage characteristic curves for the two long-term aging conditions are both higher than the short-term aging damage characteristic curve, which is attributable to the short-term aged mixture’s lower stiffness value (age level) compared to that of the long-term aged materials. The same observation can be made between the two long-term aging tempera- tures. The results indicate that the damage characteristic curves for the mixture aged at 135°C are consistently lower than the curves for the mixture aged at 95°C, which is consistent with the dynamic modulus test results that indicate that the mixture aged at 135°C is less stiff than the mixture aged at 95°C. Moreover, the C value at failure, indicated by the end point of the C versus S curve, is considerably higher for the mixture aged at 135°C than at 95°C. The average C value at failure is 0.18 for the short-term aged material. The average C values at failure are 0.37 and 0.21 for the long-term aged material conditioned at 95°C and 135°C, respectively. The observed difference in the C values at failure for the long-term aged materials indicates that the mixture aged at 135°C is more brittle than the mixture aged at 95°C. Figure 35 (c) presents the DR failure criterion results for the SHRP AAD mixture and shows that the failure criterion line for the mixture that was long-term aged at 95°C is similar to that for the short-term aged mixture. However, the failure criterion line for the mixture that was long-term aged at 135°C is considerably lower than that for the short-term aged mixture, which is indicative of less fatigue resistance. These results suggest that long-term aging at 135°C leads to the degradation of fatigue resistance. The results presented suggest that long-term aging at 135°C should be avoided due to negative performance implications. In addition to the trends seen in the performance test results, visual observations of fractured specimens also indicate changes between the SHRP AAD mixtures aged at 95°C and 135°C. Table 11 presents the statistical analysis outcomes for the SHRP AAD mixture performance test results. The pairs with significant differences (p < 0.05) are shaded in the table. Both long-term aging conditions have a significant effect on the dynamic modulus Dynamic Modulus Sample Reduced Frequency 1.0E-05 1.0E-04 1.0E-03 1.0E-02 p-value Short-term aged Loose mix, 95°C, 8.9 days 0.0055 0.0002 0.0058 0.0063 Short-term aged Loose mix, 135°C, 16.8 hours 0.0264 0.0213 0.0188 0.0072 Loose mix, 95°C, 8.9 days Loose mix, 135°C, 16.8 hours 0.0537 0.0338 0.0218 0.0043 Cyclic Fatigue Sample S 1.0E+05 2.0E+05 2.5E+05 3.0E+05 p-value Short-term aged Loose mix, 95°C, 8.9 days 0.0337 0.0062 0.0002 0.0006 Short-term aged Loose mix, 135°C, 16.8 hours 0.0093 0.0036 0.0037 0.0044 Loose mix, 95°C, 8.9 days Loose mix, 135°C, 16.8 hours 0.4232 0.1304 0.0176 0.0134 Table 11. Statistical t-test analysis outcomes for dynamic modulus and cyclic fatigue test results for SHRP AAD materials.

Findings and Applications 53 values and fatigue performance for the SHRP AAD mixture. In spite of the matched binder rheology at the 95°C and 135°C aging temperatures, the significant difference is observed in the dynamic modulus and fatigue performance test data between the 95°C and 135°C aging treatment. Note that, due to the very brittle nature of the SHRP AAD materials aged at 135°C, the comparison is limited to low S values. As suggested by the t-test results, 95°C and 135°C have different effects on the SHRP AAD materials. Figure 36 presents the damage contours for the SHRP AAD mixture with different aging conditions. The short-term aged material shows better performance than the long-term aged asphalt materials. Notably, based on the damage contours, the asphalt mixture aged at 95°C shows better performance than the material aged at 135°C; the same conclusion was drawn from the material-level fatigue performance data as well. Figure 37 presents the fatigue damage area versus ESALs for the SHRP AAD mixture with dif- ferent aging conditions. A significant difference between the fatigue performance of the asphalt mixture aged at 95°C and at 135°C is evident. SHRP AAG. Figure 38 shows the mixture performance test results for the SHRP AAG mix- ture. Figure 38 (a) presents the SHRP AAG mixture dynamic modulus master curves that cor- respond to the specimens fabricated after aging at 95°C and 135°C for 19 days and 37.6 hours, respectively. The dynamic modulus test results of specimens fabricated after short-term aging Short-Term Aged 10 9 8 Z (cm ) 7 6 5 4 3 2 1 0 10 9 8 Z (cm ) 7 6 5 4 3 2 1 0 10 9 8 Z (cm ) 7 6 5 4 3 2 1 0 1.510.50 X (m) –0.5–1–1.51.510.50 X (m) –0.5–1–1.5 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 01.510.50 X (m) –0.5–1–1.5 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Loose Mix, 95°C, 8.9 Days Loose Mix, 135°C, 16.8 Hr Figure 36. Damage contours for SHRP AAD mix aged under different conditions. 0 10 20 30 40 50 60 70 0 50 100 150 200 250 Pe rc en t D am ag e Month Short-Term Loose Mix, 95°C, 8.9 Days Loose Mix, 135°C, 16.8 Hr Figure 37. Comparison of fatigue damage area versus service life for SHRP AAD mixture aged under different conditions.

54 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction only also are provided for reference. The results indicate a significant increase in the dynamic modulus value with long-term aging. The results also suggest similar dynamic modulus master curves that correspond to mixtures aged at 95°C for 19 days and 135°C for 37.6 hours, with slightly higher dynamic modulus values for the mixture aged at 135°C compared to 95°C. How- ever, the binder rheology test results shown in Figure 38 suggest that the binder contained within the mixture aged at 135°C is slightly stiffer than the binder contained within the mixture aged at 95°C. Thus, it is difficult to ascertain whether or not the results suggest a change in mixture performance as a result of laboratory aging at 135°C as opposed to 95°C. The results presented in Figure 38 suggest that little chemical change is induced by aging SHRP AAG at 135°C. Hence, any difference noted in performance that results from aging at 135°C as opposed to 95°C is speculated to reflect changes other than chemistry (e.g., absorption, drain-down). Figure 38 (b) presents the C versus S fatigue damage characteristic curves that correspond to the SHRP AAG mixture. The results show somewhat different trends in the C versus S curves for the two long-term aging conditions, with the short-term aging C versus S curve falling considerably lower than the long-term aged condition curves. Figure 38 (c) presents the failure criterion results for the SHRP AAG mixture, showing that the failure criterion line for the mixture long-term aged at 95°C is similar to that of the short-term aged mixture, but that the failure criterion line for the mixture long-term aged at 135°C falls somewhat lower, which is indica- tive of less fatigue resistance. Table 12 presents the statistical analysis results for each pair of aging treatments. The p-values indicate a significant difference between the performance of the short-term aged and long-term aged SHRP AAG materials. [The pairs with significant differences (p < 0.05) are shaded in the table.] However, although no significant difference can be observed between the dynamic mod- ulus values of the material aged at 95°C and at 135°C, a significant difference between the fatigue 1E+2 1E+3 1E+4 1E+5 1E-8 1E-6 1E-4 1E-2 1E+0 1E+2 |E* | (M Pa ) Reduced Frequency (Hz) 0.0E+0 4.0E+5 8.0E+5 1.2E+6 0E+00 5E+04 1E+05 Short-term Aged Oven, Loose Mix, 95°C, 19 days Oven, Loose Mix, 135°C, 37.6 hours 0.0 0.2 0.4 0.6 0.8 1.0 C S 0E+00 1E+04 2E+04 3E+04 4E+04 Cu m ul at iv e (1- C) Nf (Cycle) (a) (b) (c) Figure 38. SHRP AAG mixture performance test results: (a) dynamic modulus curves, (b) C versus S curves, and (c) DR failure criterion lines.

Findings and Applications 55 test results for these two temperatures is evident. This observation suggests that the 135°C aging temperature induces more brittle behavior than the 95°C aging temperature. The FlexPAVE™ program results also suggest less fatigue resistance of material aged at 135°C compared to 95°C. The damage contours presented in Figure 39 indicate that more severe fatigue cracking is associated with long-term aging at 135°C than at 95°C. Figure 40 also suggests a significant decrease in fatigue cracking resistance when the 135°C long-term aging treatment is applied. Proposed Long-Term Aging Procedure Based on the findings from this study, oven aging of loose asphalt mixtures at 95°C is pro- posed as the long-term aging procedure for fabrication of performance testing specimens. Aging asphalt mixtures in a loose mix state expedites oxidation compared to compacted speci- men aging under the same conditions. The performance test results indicate no problems with loose mixtures compacted after long-term aging. Loose mixture aging at 95°C provides shorter Dynamic Modulus Sample Reduced Frequency 2.0E-06 1.0E-04 1.0E-03 1.0E-01 p-value Short-term aged Loose mix, 95°C, 8.9 days 0.0102 0.0084 0.0044 0.0109 Short-term aged Loose mix, 135°C, 16.8 hours 0.0296 0.0120 0.0023 0.0049 Loose mix, 95°C, 8.9 days Loose mix, 135°C, 16.8 hours 0.0928 0.1169 0.3271 0.2768 Cyclic Fatigue Sample S 1.0E+05 2.0E+05 3.0E+05 5.0E+05 p-value Short-term aged Loose mix, 95°C, 8.9 days 0.0082 0.0020 0.0001 0.0314 Short-term aged Loose mix, 135°C, 16.8 hours 0.0062 0.0001 0.0004 0.0235 Loose mix, 95°C, 8.9 days Loose mix, 135°C, 16.8 hours 0.0161 0.0289 0.0415 0.0940 Table 12. Statistical t-test analysis outcomes for dynamic modulus and cyclic fatigue test results for SHRP AAG materials. Short-Term Aged 10 9 8 Z (cm ) 7 6 5 4 3 2 1 0 10 9 8 Z (cm ) 7 6 5 4 3 2 1 0 10 9 8 Z (cm ) 7 6 5 4 3 2 1 0 1.510.50 X (m) –0.5–1–1.51.510.50 X (m) –0.5–1–1.5 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 1.510.50 X (m) –0.5–1–1.5 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Loose Mix, 95°C, 19 Days Loose Mix, 135°C, 37.6 Hr Figure 39. Damage contours for SHRP AAG mixture aged at different conditions.

56 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction durations of laboratory aging to reach the oxidation levels in the field than lower temperatures, whereas aging temperatures higher than 100°C can result in changes in binder chemistry that do not occur in the field. Aging of WMA Mixtures Loose mixture aging in the oven at 95°C is the proposed long-term aging method for HMA mixtures. Although the compaction of long-term aged HMA loose mixtures typically is not prob- lematic, some WMA technologies (e.g., foam) may not maintain their compaction-aiding prop- erty if they are aged as loose mix. Consequently, a separate investigation of the feasibility of using compacted specimen aging was conducted to select the optimal laboratory aging procedure for WMA mixtures. Table 13 provides a summary of the Group B materials that were used in this study to evaluate WMA aging. This evaluation of WMA aging includes data obtained from two projects, Manitoba (MB) and NCAT. Both projects included a control HMA mixture. The results for HMA compacted specimen aging indicate that the oxidation gradient within the specimen is insignificant if the minimum sample dimension is equal to or less than 38 mm. Therefore, compacted specimen aging trials were conducted on thin disk specimens prepared from gyratory- compacted samples with a thickness of 25 mm. Compacted specimen aging was conducted at 85°C, which is the maximum temperature at which compacted specimen aging can be conducted without causing damage or distortion of the sample. In order to evaluate the efficiency of com- pacted specimen aging at 85°C compared to loose mixture aging at 95°C, both procedures also were conducted using the NCAT-HMA mixture. The results were compared to those for binder extracted and recovered from field cores obtained after 4 years in service to allow the determina- tion of the time required to match 4 years of field aging. Figure 41 (a) presents the G* values at 64°C and 10 rad/s of binder extracted and recovered from the MB-HMA and MB-Evotherm (WMA) field cores. Figure 41 (b) presents the field core 0 10 20 30 40 50 60 70 0 50 100 150 200 250 Pe rc en t D am ag e Month Short-Term Loose Mix, 95°C,19 Days Loose Mix, 135°C, 37.6 Hr Figure 40. Comparison of fatigue damage area versus service life for SHRP AAG mixture aged at different conditions. Site ID Location Binder / Modification Date Built Date Core Extracted MB Manitoba, Canada Control HMA, Foam WMA, Evotherm WMA 2010 2014 NCAT Alabama, United States Control HMA, Foamed WMA 2009 2013 Table 13. Group B materials used for WMA aging evaluation.

Findings and Applications 57 data obtained from the NCAT-HMA and NCAT-Foam (WMA) sections. These results show that the HMA pavement section exhibits higher G* values than the WMA pavement section at all depths where measurements were taken. Figure 42 (a) shows the evolution of log G* at 64°C and 10 rad/s with laboratory aging dura- tions for the MB compacted specimen aging trials conducted at 85°C. The results show that the MB-HMA mixture has higher G* values than the MB-Foam and MB-Evotherm mixtures at a given laboratory aging duration, suggesting that the effects of reduced short-term aging from using WMA has long-term implications. However, the slopes of the WMA and HMA log G* versus aging duration curves are similar. The laboratory aging finding of elevated age levels in HMA compared to WMA matches the trends observed in the field core results presented in Figure 41. The MB-Evotherm and MB-Foam mixtures exhibit a similar G* evolution, indicating that the WMA technologies used did not significantly impact oxidative aging. Figure 42 (b) shows the evolution of log G* at 64°C and 10 rad/s with laboratory aging dura- tions for the NCAT compacted specimen aging trials conducted at 85°C. The NCAT-HMA loose mixture aging results for tests conducted at 95°C are included for comparison. The results of the NCAT mixtures match the findings of the MB mixtures; i.e., reduced short-term aging in WMA compared to HMA leads to lower G* values at a given aging duration, but the slopes of the 0 10 20 30 40 50 60 1 2 3 4 5 D ep th (m m) G*, 64°C, 10 rad/s (kPa) G*, 64°C, 10 rad/s (kPa) MIT-HMA MIT-Evotherm 0 10 20 30 40 0 10 20 30 40 50 60 D ep th (m m) NCAT-HMA NCAT-Foam (a) (b) Figure 41. G* values at 64°C and 10 rad/s versus depth for (a) MB field cores and (b) NCAT field cores (MIT = Manitoba Infrastructure and Transportation). y = 0.0249x + 0.1151 R² = 0.9825 y = 0.0181x + 0.1442 R² = 0.9995 y = 0.0217x + 0.0935 R² = 0.9935 0 0.2 0.4 0.6 0.8 0 5 10 15 20 25 30 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Duration (Days) MIT-HMA MIT-Evotherm MIT-Foam y = 0.0675x + 1.1788 R² = 0.98 y = 0.021x + 0.88 R² = 0.94 y = 0.0188x + 0.8397 R² = 0.95 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 0 5 10 15 20 25 30 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Duration (Days) (a) (b) Oven, Loose Mix, 95°C, NCAT-HMA Oven, Compacted, 85°C, NCAT-HMA Oven, Compacted, 85°C, NCAT-Foam Figure 42. Laboratory aging G* evolution results for (a) MB mixtures and (b) NCAT mixtures.

58 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction compacted WMA and compacted HMA log G* versus the laboratory aging duration curves are similar. The comparison of the loose and compacted specimen aging log G* versus aging dura- tion curves for the NCAT-HMA mixture shown in Figure 42 (b) indicates a significant increase in the efficiency of loose mixture aging at 95°C compared to compacted specimen aging at 85°C. The results presented in Figure 42 combined with the log G* results for binder extracted and recovered from field cores were used to determine the laboratory aging durations that are required to match 4 years of field aging at a depth of 19 mm. The MB project did not include any field sections that contained the Foam mixture, but the MB-Foam mixture was studied in the laboratory to investigate the effect of foaming on the aging of MB and NCAT base binders. Table 14 presents the required laboratory aging durations. The results show that the required laboratory aging durations for the WMA mixtures are slightly shorter than for the HMA for the same field aging. Furthermore, the large difference in laboratory aging durations that are required to match 4 years of aging in the NCAT versus MB sections demonstrates that climate has a significant effect on long-term aging. The required durations of compacted specimen laboratory aging at 85°C to match just 4 years of field aging are long (i.e., roughly 2 to 5 weeks), making compacted speci- men aging impractical. Compacted specimen aging at 85°C required 35.6 days to match 4 years of field aging for the NCAT-HMA section. In contrast, only 6.7 days of loose mixture aging at 95°C were required to match the same condition. Therefore, it is proposed that loose mixture aging at 95°C be used for laboratory long-term aging of both WMA and HMA mixtures. Future research is needed to investigate the relationship between laboratory aging durations and field aging for WMA materials and to resolve the compactability problems associated with WMA after long-term aging. It is expected that the compaction of long-term aged WMA loose mixture will require the use of HMA compaction temperatures. The compaction advantages of surfactant and wax additives may be lost after long-term aging. Foaming will cease prior to the completion of the long-term aging of foamed WMA loose mixtures. Climate-Based Determination of Predefined Aging Durations Kinetics Modeling of Loose Mix Aging Determining the laboratory aging durations representing a given time in service, climate, and depth within a pavement requires a solid understanding of binder oxidation kinetics and diffusion. Kinetics modeling enables the prediction of the rate of change in asphalt properties as a function of aging time and temperature. Therefore, kinetics models generally are calibrated using experimental measurements of AIPs at various aging durations under isothermal aging conducted at multiple temperatures. Diffusion models enable the prediction of oxygen partial pressure within binder films. In this project, a rheology-based kinetics model was developed based on existing chemistry-based kinetics models and AIP measurements obtained from the laboratory oven-aged loose mixtures and field cores. However, the time and resources available to this project did not allow the development of a fully validated diffusion model, which is a key suggestion for future research. Field Section Depth (mm) Required Compacted Specimen Aging Duration at 85°C (days) Required Loose Mixture Aging Duration at 95°C (days) NCAT-HMA 19 35.6 6.7 NCAT-Foam (WMA) 19 31.3 NA MB-HMA 19 16.5 NA MB-Evotherm (HMA) 19 16.1 NA Table 14. Laboratory aging durations required to match 4 years of in-service aging.

Findings and Applications 59 All of the models presented herein were obtained when investigating the AIPs of asphalt binders aged in thin films at elevated temperatures and/or under pressure. The oxidation of an asphalt will be kinetics-controlled if the binder film thickness is thin enough to allow oxygen to diffuse through the film faster than the rate of the reaction itself. It is critical that the film thickness is sufficiently thin to ensure a kinetics-controlled reaction when calibrating a kinetics model (Glaser et al. 2015). If the reaction is not kinetics-controlled, the diffusion effects will be significant and, therefore, must be considered within the model, which adds complexity. Glaser et al.’s (2013b) kinetics model framework discussed previously was applied to loose mixture aging to facilitate linking laboratory aging durations to field aging; log G* at 64°C and 10 rad/s frequency was selected as the AIP because mixture performance relates more directly to rheology than to chemistry, leading to Equation 15. log * log * 1 1 exp (15)G G M k k k t k Mto c f f c( )( )= + −   − − + where G* = long-term aged binder shear modulus at 64°C and 10 rad/s (kPa) and G*o = short-term aged binder shear modulus at 64°C and 10 rad/s (kPa). Table 15 provides the detailed properties of the 10 mixtures that were used to calibrate and validate the kinetics model presented in Equation 15. Because the binder source, aggregate source, and filler type may affect loose mixture aging rates, the experimental plan was designed to include a wide range of materials. Initially, an experiment was designed to verify that loose mixture oven aging leads to a kinetics- controlled reaction mechanism, which is a requirement for applying kinetics models without considering diffusion. Three loose mixtures were prepared using component materials from the WesTrack Fine sections with various binder contents (4.7%, 5.4%, and 6.1%). If loose mixture oven aging leads to a kinetics-controlled oxidation reaction, then the binders in these three mix- tures will oxidize at the same rate despite their various film thicknesses. However, if loose mixture oven aging leads to a diffusion-controlled oxidation reaction, then the rate of oxidation will be inversely proportional to the binder film thickness and thus to the binder content. After verifying that loose mixture aging results in a kinetics-controlled reaction, isothermal loose mixture aging was conducted using a broad range of mixtures at three temperatures: 95°C, 85°C, and 70°C. Asphalt binder was extracted and recovered from the loose mixtures at various aging durations to measure the rheological and chemical AIPs. The relationship between laboratory aging duration and log G* at 64°C and 10 rad/s was derived from loose mixture aging conducted at 95°C and used to deter- mine the parameter M in Equation 15 for all the study mixtures. Universal values of the reaction parameters included in kf and kc were obtained from the least mean square error optimization of Equation 15 to match the values of log G* at 64°C and 10 rad/s obtained from binders extracted and recovered from five calibration mixtures (FHWA ALF-Control, FHWA ALF-SBS, SHRP AAD, WesTrack Fine, and WesTrack Coarse) after various durations of laboratory aging at 95°C, 85°C, and 70°C. The remaining mixtures in Table 15 were used to validate that the kf and kc parameters are mixture-independent. Equation 15 was used to predict the values of log G* at 64°C and 10 rad/s cor- responding to the various durations of loose mixture aging at 85°C and 70°C. The predicted log G* values were compared against the measured data to evaluate the accuracy of the proposed model. To validate the kinetics model further, two loose mixtures, FHWA ALF-Control and WesTrack Fine with optimum binder content (5.4%), were subjected to non-isothermal aging. Then, the ability of the kinetics model to predict the oxidation of the loose mixtures under non-isothermal conditions was evaluated. Asphalt binder was extracted and recovered at several times during

60 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction the non-isothermal aging process and subjected to AIP testing. The measured AIPs were then compared against those predicted by the kinetics model. Evaluation of the Loose Mixture Aging Reaction Mechanism. Loose mixture oven aging was hypothesized to lead to a kinetics-controlled reaction mechanism because the film thick- ness of asphalt binders in loose mixtures is typically less than 15 microns (Radovskiy 2003). Glaser et al.’s (2015) evaluation of binder aging suggests that a film thickness of 15 microns will lead to a kinetics-controlled reaction. The hypothesis that loose mixture aging is a kinetics- controlled reaction is verified by the results presented in Figure 43, which shows the relationship between laboratory duration and both chemical and rheological AIPs for the three WesTrack Fine mixtures with various binder contents and, thus, film thicknesses. The results show equiva- lent AIPs at various aging durations irrespective of the mixture film thickness, thereby verifying the hypothesis that oven aging loose mixtures results in a kinetics-controlled oxidation reaction. Project Mix MIX ID BinderSource Binder Modifier / Anti- Stripping Agent Binder Grade Binder Content Aggregate Type Hydrated Lime (Content) FHWA ALF ALF Control ALF CTRL Citgo- Venezuelan Bachaquero None PG70-22 5.3% Limestone / Diabase (traprock) 1% ALF SBS ALF SBS Mid Continent SBS-LG PG 70-28 5.3% Limestone / Diabase (traprock) 1% SHRP Binder AAD-1 Binder + FHWA ALF Aggregate SHRP AAD California Coast None PG 58-28 5.3% Limestone / Diabase (traprock) 1% WesTrack Fine Section WesTrack - Fine West Coast None PG 64-22 4.7% Andesite / Granite / Sand 1.5%5.4% 6.1% Coarse Section WesTrack - Coarse Idaho Asphalt None PG 64-22 5.7% Crushed Andesite / Sand 1.5% NC DOT NCS9.5B NC Citgo- Wilmington, NC None PG64-22 6.6% Granite None LTPP South Dakota LSD N/A None 120-150 Pen 5.9% N/A None New Mexico LNM N/A None AC-20 7.6% N/A None Texas LTX N/A None AC-20 5.4% N/A None Wisconsin LWI N/A None N/A 5.9% N/A None Table 15. Mixtures used to calibrate and validate the kinetics model for loose mix oven aging.

Findings and Applications 61 These results indicate that kinetics models can be derived from loose mixture aging results with- out the need to consider diffusion. Kinetics Modeling of Loose Mixtures under Isothermal Aging. Table 16 presents the values of the kinetics model parameters. The universal values of the reaction parameters included in kf and kc (Eaf, Eac, Af, and Ac) were obtained from a least mean square error optimization of the data obtained at 95°C from five mixtures (FHWA ALF-Control, FHWA ALF-SBS, SHRP AAD, WesTrack Fine, and WesTrack Coarse), assuming M is equal to one. After optimizing the kf and kc parameters, the material-dependent M values were determined by conducting least mean square error optimizations using individual mixtures. It should be noted that M accounts for the mix-specific aging kinetics. The table shows that the material-dependent parameter, M, varies significantly (from 0.56 to 1.10) among the different materials evaluated. Higher M values mean higher oxidation susceptibility. Figure 44 presents comparisons between the measured and predicted log G* at 64°C and 10 rad/s values for the loose mixture aging temperatures of 70°C, 85°C, and 95°C for the mixtures C+S = 0.0013(D) + 0.0884 R² = 0.96 0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0 10 20 30 Ca rb on yl +S al fo xi de s A bs or ba nc e Pe ak s (A U) Aging Duration (Days) WT-Fine-%ac=6.1, Short-Term Aged WT-Fine-%ac=6.1, Oven, Loose Mix, 95°C WT-Fine-%ac=5.4, Short-Term Aged WT-Fine-%ac=5.4, Oven, Loose Mix, 95°C WT-Fine-%ac=4.7, Short-Term Aged WT-Fine-%ac=4.7, Oven, Loose Mix, 95°C log G* = 0.0552(D) + 1.6508 R² = 0.95 0 0.5 1 1.5 2 2.5 3 3.5 0 10 20 30 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) WT-Fine-%ac=6.1, Short-Term Aged WT-Fine-%ac=6.1, Oven, Loose Mix, 95°C WT-Fine-%ac=5.4, Short-Term Aged WT-Fine-%ac=5.4, Oven, Loose Mix, 95°C WT-Fine-%ac=4.7, Short-Term Aged WT-Fine-%ac=4.7, Oven, Loose Mix, 95°C Figure 43. Loose mix aging rates obtained from WesTrack Fine mix prepared with high (6.1%), optimum (5.4%), and low (4.7%) binder contents: (a) aging rates in terms of C + S absorbance peaks and (b) aging rates in terms of logarithm of binder shear modulus log G* at 64°C and 10 rad/s. Project Mix ID Af Eaf Ac Eac M FHWA ALF ALF-CTRL 1.25×1013 95.04 3.68×107 62.21 0.743 ALF-SBS 1.25×1013 95.04 3.68×107 62.21 0.623 SHRP Binder SHRP AAD 1.25×1013 95.04 3.68×107 62.21 1.104 WesTrack WT-Fine 1.25×1013 95.04 3.68×107 62.21 0.871 WT-Coarse 1.25×1013 95.04 3.68×107 62.21 0.725 Validation Set NCDOT NC 1.25×1013 95.04 3.68×107 62.21 0.937 LTPP LTPP-SD 1.25×1013 95.04 3.68×107 62.21 0.747 LTPP-NM 1.25×1013 95.04 3.68×107 62.21 0.546 LTPP-TX 1.25×1013 95.04 3.68×107 62.21 0.88 LTPP-WI 1.25×1013 95.04 3.68×107 62.21 1.016 Table 16. Kinetics model parameters using a single fitting parameter and universal reaction rates.

62 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction used to calibrate the kf and kc parameters. The results demonstrate good agreement between the measured and predicted values of log G*, thereby verifying the applicability of Equation 15 to the kinetics modeling of loose mixture aging. Figure 45 presents the comparisons between the measured and predicted log G* at 64°C and 10 rad/s values for the loose mixture aging temperatures of 70°C, 85°C, and 95°C for the validation mixtures. To apply the kinetics model to the validation mixtures, Eaf, Eac, Af, and Ac derived from the evaluation of the calibration mixtures were adopted. The material-dependent variable, M, was optimized against the AIP measurements obtained from the various laboratory aging durations at 95°C. The results demonstrate very good agreement between the measured and predicted values, thereby validating the use of Equation 15 coupled with the universal Eaf, Eac, Af, and Ac parameters. Kinetics Modeling of Non-Isothermal Aging. Field aging involves a variable temperature history. Therefore, the ability of the kinetics model to predict the oxidative aging of loose mix- tures under non-isothermal conditions was evaluated using two mixtures. Figure 46 (a) shows 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) lo g G *, 64 °C , 1 0 ra d/ s (kP a) lo g G *, 64 °C , 1 0 ra d/ s (kP a) lo g G *, 64 °C , 1 0 ra d/ s (kP a) lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 Aging Duration (Days) (d) WesTrack Fine (c) SHRP AAD(b) ALF SBS (a) ALF Control (e) WesTrack Coarse M = 0.623 M = 0.871 M = 0.725 M = 1.104 M = 0.743 Figure 44. Comparisons between measured and predicted log G* values for the calibration mixtures: (a) ALF-Control, (b) ALF-SBS, (c) SHRP AAD, (d) WesTrack Fine, and (e) WesTrack Coarse.

Findings and Applications 63 the non-isothermal aging history applied to the loose mixture specimens. The temperatures were varied between 40°C and 70°C over the course of 43 days. Equation 15, with the mixture- specific M value derived from laboratory aging at 95°C, was used to predict the AIP evolution under the non-isothermal conditions that were applied to the loose mixtures. The predicted AIP values were compared with those measured from the binders that were extracted and recovered at the times (in days) indicated in Figure 46 (a). Figure 46 (b) through Figure 46 (d) present comparisons between the measured and predicted AIP values. In general, the predicted and mea- sured values for log G* are in good agreement. These results suggest that the presented kinetics framework can be coupled with the field pavement temperature history and a diffusion model to predict the evolution of asphalt binder AIPs within a pavement over its service life. Kinetics Derivation of Laboratory Aging Duration Based on the successful prediction of non-isothermal aging, the kinetics modeling was extended to predict the required duration of loose mixture aging at 95°C in the laboratory to 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) M = 1.016 M = 0.747 M = 0.88 (a) NCS9.5B (c) LTPP-NM(b) LTPP-SD (d) LTPP-TX (e) LTPP-WI M = 0.546 M = 0.937 Figure 45. Comparisons between measured and predicted log G* values for the validation mixtures: (a) NCS9.5B, (b) LTPP-SD, (c) LTPP-NM, (d) LTPP-TX, and (e) LTPP-WI.

64 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction match field aging as a function of pavement temperature history and depth. It is important to note that loose mixture oven aging is kinetics-controlled whereas field aging is diffusion- controlled. Thus, to link laboratory aging to field aging in a rigorous way, the kinetics model would need to be coupled with a diffusion model. However, a suitable diffusion model is not available. Furthermore, current diffusion model frameworks require detailed mixture mor- phology information that may make the use of a diffusion model to determine the laboratory aging duration required to match the field aging impractical. Therefore, in lieu of a diffusion model, the kinetics modeling framework was simplified and calibrated against field data that corresponded to a wide range of climatic conditions and depths. Derivation of Laboratory Aging Duration Using the Kinetics Model. To determine the required laboratory aging duration at 95°C that is required to match a given field age level using kinetics modeling, Equation 15 must be written in terms of laboratory and field thermal histories, which are expressed by Equation 16 and Equation 17, respectively. Within these kinetics expres- sions, the temperature history and corresponding temperature dependence of the reaction rate are predicted by kf and kc, whose parameters are universal. In addition, Equation 16 and Equation 17 include the log G* values that correspond to short-term aging. To determine the required labora- tory aging duration at 95°C to match a given field temperature history, G*lab and G*field are equated, as shown in Equation 18. The time-dependent pavement temperature history is input into the field side of Equation 18 and 95°C is input as the laboratory aging temperature that allows the labora- tory aging time (tlab) to be solved. Equation 18 demonstrates that the material-dependent param- eter, M, appears on both the lab and field sides of Equation 18 and, therefore, cancels itself out. (a) (b) (c) (d) 30 40 50 60 70 80 0 5 10 15 20 25 30 35 40 45 Te m pe ra tu re (° C) Aging Time (Days) Loose Mix Sampling 2 D 6 D 6 D 13 D 5 D 10 D 1 D 0 5 10 15 20 25 30 35 40 16D 33D 43D G *, 64 °C , 1 0 ra d/ s Aging Duration Predicted Measured 2% Error 23% Error10% Error WesTrack, Fine 0 10 20 30 40 50 60 16D 33D 43D G *, 64 °C , 1 0 ra d/ s Aging Duration Predicted Measured 6% Error 2% Error8% Error ALF Control y = 1.0035x - 1.9924 R² = 0.99 0 10 20 30 40 50 60 0 20 40 60 M ea su re d lo g G *, 64 °C , 10 ra d/ s Predicted log G*, 64°C, 10 rad/s WesTrack, Fine ALF ControlLine of Equ ality Figure 46. Predictions of non-isothermal loose mixture oven aging: (a) non-isothermal laboratory aging history, (b) WesTrack Fine mix prediction, (c) ALF-Control mix prediction, and (d) overall prediction quality.

Findings and Applications 65 Furthermore, if the short-term aging of the field and laboratory mixtures is assumed to be equal (i.e., G*o,lab = G*o,field), then Equation 18 reduces to Equation 19. Equation 19 shows that the labora- tory oven aging duration that is required to match field aging for a given mixture is independent of the G* of the short-term aged material and M. In other words, the laboratory aging duration that is required to match a given field condition is independent of the material-specific kinetics. This finding is significant because it indicates that mixture-specific kinetics model parameters are not required for the determination of laboratory aging durations. log * log * 1 1 exp (16),G G M k k k t k tlab o lab c f f c lab ( )( )= + −   − − +     log * log * 1 1 exp (17),fieldG G M k k k t k tfield o c f f c field ( )( )= + −   − − +     log * 1 1 exp log * 1 1 exp (18) ,field ,lab G M k k k t k t G M k k k t k t o c f f c field o c f f c lab ( )( ) ( )( ) + −     − − +     = + −     − − +     1 1 exp 1 1 exp (19) k k k t k t k k k t k t c f f c field c f f c lab ( )( ) ( )( )−   − − +     = −     − − +     Laboratory and Field Validation. Laboratory and field data were used to validate that the laboratory aging duration required to match a given field condition is independent of the mixture-specific kinetics. The laboratory non-isothermal aging experiment presented in Fig- ure 46 was used first for validation. Isothermal aging was conducted at 95°C and binder was extracted and recovered at different times to determine the G* value at 64°C and 10 rad/s for both the FHWA ALF-Control and WesTrack Fine mixtures. The results were used to determine the laboratory aging duration at 95°C that is required to match the G* value at 64°C and 10 rad/s of binder extracted and recovered during the non-isothermal aging experiments after 16, 34, and 44 days of aging. If the kinetics model derivation is valid, then the laboratory aging duration at 95°C that is required to match the age level in the non-isothermal experiment should be the same for the FHWA ALF-Control and WesTrack Fine mixtures, despite their differing oxidation susceptibilities, which are based on their material-dependent M values (FHWA ALF-Control, M = 0.743 versus WesTrack Fine, M = 0.871). Figure 47 presents comparisons between the 0 0.5 1 1.5 2 2.5 3 0 20 40 60 M ea su re d Ag in g D ur at io n at 9 5° C (D ay s) Non-Isothermal Oven Aging Duration (Days) ALF - Control WesTrack - Fine1995 y = 0.8665x + 0.2657 R² = 0.96 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 M ea su re d Ag in g D ur at io n at 9 5° C, A LF -C on tro l (D ay s) Measured Aging Duration at 95°C, WesTrack-Fine (Days) (a) (b) Figure 47. Laboratory non-isothermal aging validation: (a) measured aging duration for non-isothermal aging of ALF-Control and WesTrack Fine mixtures and (b) measured field aging for ALF-Control and WesTrack Fine mixtures.

66 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction laboratory aging durations at 95°C that are required to match the oxidation level at several times during the non-isothermal aging experiment. The results indicate that the aging duration at 95°C that is required to match the level of oxidation induced by the non-isothermal experi- ment is the same for the FHWA ALF-Control and WesTrack Fine mixtures, thereby validating the use of Equation 19 for equating laboratory aging to field aging. The FHWA ALF field data also were used to validate the kinetics derivation that was imple- mented to determine the laboratory aging durations. Asphalt binder was extracted and recovered from various depths of field cores extracted after 8 years of service from FHWA ALF-Control and SBS pavement sections in McLean, Virginia. The G* values at 64°C and 10 rad/s of the field-aged binders were measured. Isothermal laboratory loose mixture aging was conducted at 95°C and binder was extracted and recovered at various times to determine the G* value at 64°C and 10 rad/s for both the FHWA ALF-Control and SBS-modified mixtures. These data were then used to determine the duration of laboratory aging that is required to match the aging level of the field core. The measured laboratory aging durations at 95°C that are needed to match the field aging level at various depths were then determined. Figure 48 presents comparisons of the laboratory aging durations that are required to match the field core age level based on the G* values at 64°C and 10 rad/s values for the FHWA ALF-Control and SBS-modified mixtures. The FHWA ALF-Control (M = 0.743) and SBS-modified (M = 0.623) mixtures exhibited significantly different kinetics model M val- ues. However, the results demonstrate that the laboratory aging durations that are required to match the field age levels are similar for the two mixtures, despite their differing kinetics behavior. It should be noted that the data presented in Figure 47 were obtained solely from loose mixture aging, which is a kinetics-controlled reaction, whereas field oxidation is 0 10 20 30 40 50 60 0 5 10 15 20 25 Pa ve m en t D ep th (m m) Laboratory Aging Duration to Match Field Aging (Days) ALF Control ALF SBS (a) (b) y = 1.0277x + 0.8943 R² = 0.99 0 5 10 15 20 25 30 0 5 10 15 20 25 30 A LF C on tro l M ea su re d D ur at io n (D ay s) ALF SBS Measured Duration (Days) 6 mm depth 19 mm depth 54 mm depth (c) Figure 48. FHWA ALF field case study validation: (a) FHWA ALF location in McLean, VA (b) measured field aging at different depths, and (c) measured field aging for ALF-Control and ALF-SBS–modified mixtures.

Findings and Applications 67 diffusion-controlled. Figure 48 demonstrates that the duration of laboratory aging that is required to match field aging is the same for the FHWA ALF-Control and SBS-modified mixtures. However, these mixtures were constructed using the same aggregate structure, aggregate type, target air void content, and binder content. Thus, the diffusion proper- ties of the mixtures would be expected to be similar, and hence, this finding might not extend to all scenarios. It is not practical to consider diffusion in prescribing laboratory aging durations due to unknown morphological properties, such as the in situ density of asphalt layer. Development of Climatic Aging Index (CAI) The use of Equation 19 to determine laboratory aging durations as a function of pavement temperature history is computationally expensive. Furthermore, the kinetics model expressed in Equation 19 does not consider the diffusion that affects oxidation levels within pavements. Therefore, the application of Equation 19 to determine laboratory loose mixture aging dura- tions would likely be inaccurate without either the inclusion of a diffusion model or calibra- tion against the field data. To overcome these challenges, the kinetics model was simplified and calibrated against field data that correspond to a wide range of climatic conditions and depths in order to develop a CAI to determine laboratory aging durations using pavement temperature history. Equation 20 shows the kinetics model for the prediction of log G* as a function of pavement temperature and pressure history. log * log * 1 . . 1 exp (20),fieldG G M k P k P k P t k P tfield o c n f m f m c n field ( )( )= + −   − − +     The required inputs of this model are pavement temperature history, the mixture-specific kinetics parameter M, and the log G* that corresponds to the short-term aged condition. Although the effect of the oxidation spurt is clearly evident in laboratory aging that is con- ducted at an elevated temperature, as shown in Figure 44 and Figure 45, its effect on pave- ment temperature is relatively small because pavement temperatures do not go as high as 85° and 95°C as shown in Figure 44 and Figure 45. It is noted that the log G* and aging duration relationship is almost linear at 70°C even though that at 85°C and 95°C show the nonlinear relationship with the initial spurt. Therefore, when Equation 20 is used to predict log G* as a function of time using pavement temperature history data, the effect of the fast rate oxidation spurt is negligible, which allows Equation 20 to be simplified to the form given in Equation 21. log * log * (21)0,G G M k P tfield field c n= + × × × To further support the simplification of Equation 20 to Equation 21, kc and kf are shown as functions of temperature in Figure 49. It can be seen that, at temperatures below 15°C, kf is very small, which gives the exponential term in Equation 20 a value that is close to one. Consequently, at temperatures below 15°C, the red portion of Equation 20 tends towards zero. At temperatures between 15°C and 60°C, kc/kf is relatively close to one, which causes the red term in Equation 20 to be close to zero. Hence, at typical pavement temperatures, the red term in Equation 20 tends to be zero and Equation 20 can be simplified to Equation 21. Equation 19 demonstrates that the determination of laboratory aging durations at 95°C is independent of the mixture-specific parameters, M and G*o. Thus, with regard to the duration of laboratory aging, the pertinent parameters in Equation 21 are k (rate of reaction) and Pn

68 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction (the oxygen pressure term). The Pn parameter is affected by altitude and pavement depth. Because most of the pavement sections included in Group B that were used in the climatic index study are not in locations with high altitudes, the Pn parameter is affected largely by pavement depth. Therefore, the Pn parameter is replaced by the D term in Equation 22 to define the CAI, where D is an empirical depth correction factor to account for the differences in oxygen partial pressure, Pn, with pavement depth. The CAI in Equation 22 is now defined by the rate of reaction (k), reaction time (t), and pavement depth (represented by D). CAI (22)D k t= × × The slow reaction rate term, k in Equation 22, can be represented using Arrhenius equation parameters, as shown in Equation 23, where the units of time are days. In order to consider hourly temperature data, a summation of the CAI values is applied based on the hourly tem- perature history, as shown in Equation 24. CAI exp (23)D A E RT t a( )= × × − × CAI exp 24 (24) 1 t D A E RT oven a i i N∑= = × × − = where toven = required laboratory aging duration at 95°C to match field aging (day), CAI = climatic aging index, D = depth correction factor, A = frequency (pre-exponential) factor (unit-less), Ea = activation energy (kJ/mol), R = universal gas constant (kJ/mol•K), and T = hourly pavement temperature obtained from EICM at the depth of interest (Kelvin). The CAI is an empirical index that is employed to represent the extent of oxidation. To apply the CAI to determine the loose mixture aging duration at 95°C to match a given pavement tempera- ture history and depth of interest (toven), the laboratory aging duration needed to match the level of field aging was first determined experimentally using Group B materials. Loose mixture aging was conducted using the Group B materials at 95°C, and binders were extracted and recovered 1E-06 1E-05 1E-04 1E-03 1E-02 1E-01 1E+00 0 20 40 60 80 100 Fa st a nd S lo w R ea ct io n R at es (lo g G */d ay ) Temperature (°C) Fast Reaction Rate (kf) Slow (Constant) Reaction Rate (kc) (kc/kf) kf 1 0 Figure 49. Fast and slow reaction rates at different temperatures.

Findings and Applications 69 at periodic intervals and from corresponding field cores at various depths. The laboratory aging duration that is required to match the level of aging of the field cores was determined subsequently based on matching the G* values at 64°C and 10 rad/s. The pavement temperature history acquired from EICM data for the Group B sections was then input into Equation 24. For each section, CAI was calculated for pavement temperatures greater than 20°C. Determination of 20°C as the mini- mum temperature for aging in the field was made based on the results shown in Figure 50. This figure shows the prediction of log G* using the kinetics model for two binders with vastly different aging susceptibilities (SHRP AAD with M = 1.104 and ALF-SBS with M = 0.623). The predictions show that the two binders do not experience any aging when the temperature is 20°C for prolonged periods of time despite their different aging susceptibilities. Therefore, it is safe to assume that no aging occurs when pavement temperatures in the field fall below 20°C. The parameters Ea and A were regressed to provide the best fit between CAI values and measured laboratory aging durations without depth correction. Figure 51 (a) shows the relationship between the CAI values and measured laboratory dura- tions at 95°C (toven) without applying the depth correction factor D. The figure shows significant scatter in the data and an overall low R2 value. Figure 51 (b) presents the CAI calibration sepa- rately for three depths: the near-surface layer (6 mm from the surface), 20 mm from the surface, and deeper layers (below 20 mm). Figure 51 (b) also shows that separating the data according to depth improves the relationship between the CAI values and laboratory aging durations signifi- cantly, thereby highlighting the need for the depth correction. Thus, the Ea and A parameters were first calibrated using the data that correspond to a depth of 6 mm to provide a CAI value that is equivalent to the required laboratory aging duration at 95°C to match the pavement temperature history. Then, the D values were calibrated using data that correspond to depths of 20 mm and deeper. It is worth mentioning that the depth correction factor (D) is not only affected by diffusion mechanisms but also by photo-oxidation effects at the pavement surface. Figure 51 (c) shows the relationship between the laboratory aging durations and CAI values after applying the depth correction factors shown in Table 17. As shown in Figure 51 (c), the CAI values and measured durations that are needed to match field aging correlate linearly in a one-to-one relationship. Thus, the CAI values represent the required duration at 95°C that is needed to match the field aging for a given pavement temperature history and depth (toven), as shown in Equation 24. 0 5 10 15 20 25 0 100 200 300 400 lo gG * Duration (days) T=95°C T=70°C T=60°C T=50°C T=40°C T=30°C T=20°C ALF SBS (a) (b) 0 5 10 15 20 25 0 100 200 300 400 lo gG * Duration (days) T=95°C T=70°C T=60°C T=50°C T=40°C T=30°C T=20°C SHRP AAD Figure 50. Predictions of log G* under different isothermal loadings for prolonged periods of time for: (a) SHRP AAD binder and (b) ALF-SBS–modified binder.

70 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction M ea su re d Du ra tio n at 95 °C (D ay s) M ea su re d Du ra tio n at 95 °C (D ay s) y = 0.4944x R² = 0.28 0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35 M ea su re d Du ra tio n at 95 °C (D ay s) Climatic Aging Index, CAI WesTrack-Fine WesTrack-Coarse ALF-Control ALF-SBS LTPP-NM LTPP-SD LTPP-TX LTPP-WI y = 1x R² = 0.56 y = 0.4565x R² = 0.67 y = 0.2967x R² = 0.70 0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35 Climatic Aging Index, CAI Surface Layer (6 mm) 20 mm depth Deeper Layers (below 20 mm) y = 1x R² = 0.7142 0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35 Climatic Aging Index, CAI Surface Layer (6 mm) 20 mm depth Deeper Layers (below 20 mm) (a) (b) (c) Figure 51. CAI predictions: (a) overall CAI fitting without depth correction factor D, (b) CAI fitting based on layer depth without depth correction factor D, and (c) overall CAI fitting after applying depth correction factor (D).

Findings and Applications 71 It should be noted that climatic indices have been used in other applications to quantify the effect of temperature history on aging. The GAS model uses the MAAT to predict G* as a function of time and depth in Pavement ME Design. The cumulative degree days (CDD) concept, defined as the sum of the daily high temperatures above freezing, was proposed as a potential means to link laboratory aging durations to an equivalent pavement temperature history in NCHRP Project 09-49 and NCHRP Project 9-52 (Newcomb et al. 2015, Martin et al. 2014). The use of MAAT and CDD as aging indices was tried in this research, as well as prescribing laboratory aging durations as a function of the high temperature PG. The analysis of alternative aging indices to relate laboratory aging durations to field conditions are pre- sented in Appendix H. In summary, the following advantages of the CAI over the alternative aging indices were found: 1. The CAI is based on pavement temperature rather than air temperature. 2. The CAI can be applied at different depths and uses a depth correction factor (D) for variations in pavement temperature with depth. 3. The CAI considers the exponential relationship between aging rate and temperature. 4. The CAI captures hourly pavement temperatures. Laboratory Aging Duration Maps Using the coefficient values given in Table 17, Equation 24 was used to calculate the CAI val- ues for various locations in the United States using hourly pavement temperature history data obtained from the EICM to provide an overview of the proposed laboratory aging durations for various climates, field aging durations, and depths. Laboratory aging durations were calculated for three field ages: 4 years, 8 years, and 16 years. For each field age, the laboratory aging durations were determined at three depths (6 mm, 20 mm, and 50 mm) and rounded to the nearest day: 1. Determine the field aging duration to be simulated in the laboratory. 2. Determine the location of the pavement section of interest. 3. Obtain the air temperature history for the location of interest. 4. Run the EICM to obtain the hourly pavement temperature. Note that running the EICM requires additional inputs, such as the material and structural properties of the pavement section of interest. 5. Determine the depth in the pavement where aging is to be simulated; then, find the value of D according to Table 17. Note: The values of A and Ea also are given in Table 17. 6. Calculate the CAI value by adding together the CAI values calculated for each hour of pave- ment service life for temperatures greater than 20°C, in accordance with Equation 24. Note that for simplicity, the CAI value can be calculated using hourly data for one year and then multiplied by the number of years of interest. 7. For practicality, round toven to the nearest day because the CAI value represents the required oven aging duration at 95°C that is needed to match the field aging for a given pavement temperature history and depth (toven). Figure 52 shows the CAI-determined loose mixture aging durations at 95°C that are required to match 4, 8, and 16 years of field aging at a depth of 6 mm. Figure 51 demonstrates that MIX ID Depth Correction Factor (D) Arrhenius Equation, Pre-exponential Factor (A) Arrhenius Equation, Activation Energy Ea Surface Layer (6 mm) 1.0000 1.40962 13.3121 20-mm depth 0.4565 1.40962 13.3121 Deeper Layers (below 20 mm) 0.2967 1.40962 13.3121 Table 17. CAI fitting coefficients.

(a) (b) Figure 52. Required oven aging duration at 95°C to match level of field aging 6 mm below pavement surface for (a) 4 years of field aging and (b) 8 years of field aging. (continued)

Findings and Applications 73 (c) Figure 52. (Continued) Required oven aging duration at 95°C to match level of field aging 6 mm below pavement surface for (c) 16 years of field aging. climate has a significant effect on the required laboratory aging duration that is required to match a given field age. For example, as shown in Figure 52 (c), replicating 16 years of aging in Hawaii requires 40 days of oven conditioning, whereas only 4 to 5 days of oven conditioning are required to replicate the same duration of aging in Alaska. The evaluation of asphalt mixtures that have been conditioned to replicate aging at a depth of 6 mm may be useful for evaluating and predicting top-down fatigue and thermal cracking in pavements. Figure 53 shows the CAI-determined loose mixture aging durations at 95°C that are required to match 4, 8, and 16 years of field aging at a depth of 20 mm. A comparison between the labo- ratory aging durations presented in Figure 52 and Figure 53 demonstrates that significantly shorter laboratory aging durations are required to match the field aging at a depth of 20 mm compared to 6 mm, indicating that the temperature gradient and diffusion in pavements sig- nificantly affect oxidation levels. On average, the required laboratory aging durations that are needed to match field aging at a depth of 20 mm are 45% shorter than the durations required to match field aging at a depth of 6 mm. Also, significantly longer laboratory aging durations are required to match field age levels at a depth of 20 mm for warmer climates compared to colder climates. A depth of 20 mm represents a reasonable depth for the evaluation of surface layer asphalt mix- tures because it better reflects bulk behavior within a pavement structure than nearer the surface and avoids the effect of ultraviolet (UV) oxidation. Furthermore, the laboratory aging durations that are required to reflect field aging are much shorter and, therefore, more practical for the 20-mm depth

(a) (b) Figure 53. Required oven aging duration at 95°C to match level of field aging 20 mm below pavement surface for (a) 4 years of field aging and (b) 8 years of field aging. (continued)

Findings and Applications 75 (c) Figure 53. (Continued) Required oven aging duration at 95°C to match level of field aging 20 mm below pavement surface for (c) 16 years of field aging. compared to the 6-mm depth. Also, it should be noted that the recently developed small speci- men geometry for asphalt mixture performance testing consists of Ø38-mm diameter specimens. Therefore, if field cores are to be evaluated to complement laboratory investigations, the center of horizontal cores extracted from the pavement surface would be at a depth of approximately 20 mm. Figure 54 shows the CAI-determined loose mixture aging durations at 95°C that are required to match 4, 8, and 16 years of field aging at a depth of 50 mm. These results demonstrate that considerably shorter aging durations are required to simulate aging at a depth of 50 mm com- pared to depths of 20 mm and 6 mm, thus indicating the presence of a significant oxidation gra- dient with depth near the surface of the pavements. In Figure 54, the required aging duration of zero days in a few cold northern states means that long-term aging at 50 mm below the pavement surface in these cold regions are not significant enough to require long-term oven conditioning to mimic field conditions. Figure 55 and Figure 56 present the gradients in terms of age level with depth measured in field cores acquired from in-service pavements. Figure 55 shows that field aging is more or less constant below a depth of 50 mm. It is also noted that the measured log G* values below 50 mm depth are considerably larger than those from the laboratory short-term aged materials, which are shown in Figure 55 at the pavement surface. Different from the GAS model’s assumption that long-term aging below 50 mm depth is negligible, it can be concluded from this study that long-term aging does take place below 50 mm depth, but does not change appreciably. Consequently, the evaluation of asphalt mixtures that have been prepared to match field aging at a depth of 50 mm could be useful for evaluating intermediate and base asphalt layers that play a critical role in bottom-up cracking.

(a) (b) Figure 54. Required oven aging duration at 95°C to match level of field aging 50 mm below pavement surface for (a) 4 years of field aging and (b) 8 years of field aging. (continued)

Findings and Applications 77 (c) Figure 54. (Continued) Required oven aging duration at 95°C to match level of field aging 50 mm below pavement surface for (c) 16 years of field aging. 0 50 100 150 200 250 0.0 0.5 1.0 1.5 2.0 2.5 3.0 D ep th (m m) log G*, 64°C, 10 rad/s (kPa) WesTrack Fine - Short-Term Aged WesTrack Fine - 19 Years Old FHWA ALF Control - Short-Term Aged FHWA ALF Control - 8 Years Old Figure 55. Log G* values for measured field gradient throughout pavement depth compared to laboratory short-term aged materials for FHWA ALF-Control and WesTrack Fine section with optimum asphalt content (%ac) and high air void content (%Va).

78 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction Aging Model to Predict Field Aging Throughout Pavement Depth Prediction of Field Aging Using Mix-Specific Kinetics Parameters The kinetics model developed herein could potentially be used as the basis for improving the prediction of changes in asphalt binder properties with oxidative aging within pavement performance prediction frameworks, including Pavement ME Design. To evaluate the ability of the kinetics model to predict the evolution of oxidative aging in pavements, Equation 25 was applied to predict the log G* at 64°C and 10 rad/s in the field using hourly pavement temperature history data at varying depths obtained from the EICM for the eight mixtures detailed in Table 18. log * log * 1 1 exp (25)_ ,G G M k k k t k tfield Predicted o field c f f c field ( )( )= + −   − − +     Figure 57 and Figure 58 present the results of the kinetics model predictions at various depths of the pavements. The predicted log G* values were compared to corresponding measurements 0 50 100 150 200 250 0.0 0.5 1.0 1.5 2.0 2.5 3.0 De pt h (m m) FHWA ALF Control - 8 Years Old FHWA ALF SBS - 8 Years Old LTPP South Dakota - 21 Years Old LTPP New Mexico - 18 Years Old LTPP Wisconsin - 17 Years Old WesTrack Fine - 19 Years Old log G*, 64°C, 10 rad/s (kPa) 0 20 40 60 80 100 0.5 1.0 1.5 2.0 2.5 3.0 3.5 D ep th (m m) log G*, 64°C, 10 rad/s (kPa) Low %ac and High %Va - 19 Years Old Low %ac and Medium %Va - 19 Years Old Optimum %ac and High %Va - 19 Years Old Optimum %ac and Medium %Va - 19 Years Old Optimum %ac and Low %Va - 19 Years Old High %ac and Medium %Va - 19 Years Old High %ac and Low %Va - 19 Years Old (a) (b) Figure 56. Measured log G* values as a function of pavement depth for (a) field cores obtained from various locations in the United States and (b) field cores obtained from 19-year-old WesTrack Fine sections constructed with different binder contents and air voids.

Findings and Applications 79 of log G* obtained for binder extracted and recovered from field cores. Note that, although the determination of the required laboratory aging duration that is needed to match field aging for a given mixture could be derived using the kinetics model without the mixture-specific parameters, i.e., G*o and M, the prediction of log G* at pavement depths requires these mixture-specific kinet- ics parameters. The mixture-specific parameter M is determined from the isothermal aging of loose mix at 95°C. It is assumed that the short-term aging in the field and laboratory is the same (i.e., G*o,lab = G*o,field). Because the kinetics model accounts for only the temperature dependence of the oxida- tion reaction, Equation 25 is not expected to yield accurate predictions of field aging. In the field, the partial pressure of the oxygen that is available to the binder will vary with depth, and thus, diffusion is expected to affect the field reaction rates significantly. In addition, UV aging is expected to contribute to oxidative aging near the pavement surface, which is not considered within the kinetics model that is calibrated using thermal oxidation laboratory aging. Therefore, the kinetics model was calibrated against field core measurements of log G* at a depth of 20 mm, which represents a reasonable depth for the evaluation of the bulk behavior of surface layers and avoids the effect of UV oxidation at the pavement surface. The calibration was accomplished using Equation 26. The C1 and C2 parameters were determined by linear regression between the kinetics model predictions and field core measurements of log G* at 20 mm depth. log * log * (26)_ 1 2 field_predictedG C C Gfield calibrated = + × where C1, C2 = calibration factors. Project Mix MIX ID Section Location Field Age Binder Source Modifier / Anti- Stripping Agent Binder Grade Binder Content FHWA ALF ALF- Control ALF- CTRL McLean, VA 8 Years Citgo- Venezuelan Bachaquero Anti- Stripping Agent (0.001) PG 70- 22 5.3% ALF-SBS ALF-SBS McLean, VA 8 Years Mid Continent None PG 70- 28 5.3% LTPP South Dakota LTPP- SD SD 21 Years N/A None 120-150 pen 5.9% New Mexico LTPP- NM NM 11, 18 Years N/A None AC-20 7.6% Texas LTPP-TX TX 18 Years (15 Years before microsurfacing)* N/A None AC-20 5.4% Wisconsin LTPP-WI WI 8, 17 Years N/A None N/A 5.9% WesTrack Fine Section WT- Fine Dayton, NV 0, 4, 19 Years West Coast Refinery None PG 64- 22 5.4% (Optimum) Coarse Section WT- Coarse Dayton, NV 2 Years Idaho Asphalt None PG 64- 22 5.7% (Optimum) *It was assumed that no aging occurred within the asphalt concrete layers after microsurfacing placement. Table 18. Mixtures used to calibrate field aging kinetics model predictions.

80 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction 0 20 40 60 80 100 120 140 160 180 200 0.0 0.5 1.0 1.5 2.0 2.5 3.0 D ep th (m m) log G*, 64°C, 10 rad/s (kPa) Measured for 8 Years Predicted for 8 Years 0 10 20 30 40 50 60 70 80 90 100 0.0 0.5 1.0 1.5 2.0 2.5 D ep th (m m) log G*, 64°C, 10 rad/s (kPa) Measured for 8 Years Predicted for 8 Years log G*, 64°C, 10 rad/s (kPa) 0 10 20 30 40 50 60 70 80 0.0 0.5 1.0 1.5 2.0 D ep th (m m) Measured for 18 Years Predicted for 18 Years (a) (b) (c) log G*, 64°C, 10 rad/s (kPa) 0 10 20 30 40 50 60 70 80 0.0 0.5 1.0 1.5 2.0 D ep th (m m) Measured for 21 Years Predicted for 21 Years (d) log G*, 64°C, 10 rad/s (kPa) 0 10 20 30 40 50 60 70 80 0.0 0.5 1.0 1.5 2.0 D ep th (m m) Measured for 17 Years Prediction for 17 Years (e) Figure 57. Predicted versus measured log G* values for (a) FHWA ALF-Control, (b) FHWA ALF-SBS, (c) LTPP South Dakota, (d) LTPP New Mexico, and (e) LTPP Wisconsin sections.

Findings and Applications 81 0 10 20 30 40 50 60 0.5 1.0 1.5 2.0 2.5 3.0 3.5 D ep th (m m) log G*, 64°C, 10 rad/s (kPa) (a) D ep th (m m) log G*, 64°C, 10 rad/s (kPa) 0 10 20 30 40 50 60 0.5 1.0 1.5 2.0 2.5 3.0 3.5 (b) D ep th (m m) log G*, 64°C, 10 rad/s (kPa) 0 10 20 30 40 50 60 0.5 1.0 1.5 2.0 2.5 3.0 3.5 (c) D ep th (m m) log G*, 64°C, 10 rad/s (kPa) 0 10 20 30 40 50 60 0.5 1.0 1.5 2.0 2.5 3.0 3.5 (d) D ep th (m m) log G*, 64°C, 10 rad/s (kPa) 0 10 20 30 40 50 60 0.5 1.0 1.5 2.0 2.5 3.0 3.5 (e) Figure 58. Predicted versus measured log G* values after 19 years of aging for WesTrack sections with fine gradation: (a) all sections, (b) extreme sections, (c) sections with optimum asphalt content, (d) sections with high asphalt content, and (e) sections with low asphalt content. Figure 59 presents the relationship between the kinetics model predictions of log G* at a depth of 20 mm and those measured from field cores. The results demonstrate a high correlation between the measured and predicted values of log G* (R2 = 0.75). The kinetics model tends to over predict field aging, which is expected due to diffusion. That is, the partial pressure of oxygen within the binder at a pavement depth of 20 mm will be lower than the oxygen partial pressure at the surface of loose mixture from which mixture-specific M and G*o values were determined. To enable accurate predictions of field aging, Equation 26 can be used with the experimentally determined coefficients given in Figure 59 (i.e., C1 = 0.4867 and C2 = 0.615).

82 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction Three WesTrack sections constructed with fine gradation were included in the calibration step shown in Figure 59. These sections were constructed with optimum binder content (5.4%) and various air void contents (i.e., high 12%, medium 8%, and low 4%). The calibration shown in Figure 59 includes three data points for 19-year-old field cores obtained from the three WesTrack Fine sections. Because the current kinetics prediction model does not account for the effect of mixture morphological properties on field aging, all three sections have the same predicted log G* value for 19 years of field aging (i.e., log G*_predicted = 2.325). However, the measured log G* values obtained from 19-year-old field cores show different values (i.e., log G*_measured values range between 1.765 and 2.077), indicating the need for the future consideration of the diffusion and mixture morphological effects on field aging. The importance of diffusion modeling is discussed further in the following paragraphs. The calibrated kinetics model, i.e., Equations 25 and 26, was used to predict log G* as a func- tion of depth using EICM hourly pavement temperature data and the log G* values at different pavement depths were compared to field core measurements. Although the calibrated kinetics model cannot account for diffusion differences with depth, the temperature gradients predicted from the EICM data were used within the kinetics model predictions. Figure 57 shows the com- parisons between the calibrated kinetics model predictions and the field core measurements for the FHWA ALF and LTPP projects. It is important to note that the binder was extracted and recovered only at a depth of 20 mm for the LTPP Texas field core, which is the reason those results are not presented. With the exception of the LTPP Wisconsin project, the calibrated kinetics model tends to under-predict aging near the pavement surface and over predict aging at depths below 20 mm, which matches intuition in the absence of a diffusion model for the depths below 20 mm and consideration of UV oxidation for the pavement near-surface. UV oxidation will lead to more aging than is taken into account by the thermal oxidation kinet- ics model herein. The partial pressure of oxygen within the binder will decrease as a function of pavement depth, thereby leading to reduced aging, which is not considered in the current model framework. Therefore, the results presented highlight the need for the development of a diffusion model to complement the kinetics model to improve the prediction of field aging in future work. The WesTrack project offers a unique opportunity to study the effects of asphalt mixture morphological properties on diffusion. The WesTrack Fine sections, detailed in Table 19, were used in a systematic study of the changes in binder and air void contents and were used also to conduct a preliminary evaluation of the effects of morphological properties on field aging. y = 0.615x + 0.4867 R² = 0.75 0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5 lo g G *_ m ea su re d (kP a) log G*_predicted (kPa) WT-Fine - Opt. %ac, Low %Va WT-Fine - Opt. %ac, Medium %Va WT-Fine - Opt. %ac, High %Va WT-Coarse LTPP-WI LTPP-NM LTPP-SD LTPP-TX ALF-CTRL ALF-SBS Figure 59. Calibration of kinetics prediction model of field aging by comparing predicted versus measured log G* values at a depth of 20 mm (%ac = asphalt content, %Va = air void content).

Findings and Applications 83 The calibrated kinetics model was used to predict log G* values after 19 years of field aging in the WesTrack sections and these values were compared against field core measurements. Air void content and film thickness do not affect loose mixture aging kinetics and, therefore, are not taken into account by the calibrated kinetics model. Consequently, the predictions of the log G* values were the same for all the WesTrack Fine sections. However, air void content and film thickness affect the diffusion of oxygen in the field, and therefore, the field core results differ among the different WesTrack sections. Figure 58 shows the comparisons between the calibrated kinetics model predictions and the field core measurements of log G* for 19-year-old WesTrack Fine sections. Figure 58 (a) includes all of the WesTrack sections and shows a significant spread in the data. The general trends in oxi- dation levels with changes in air void content and binder content suggest that mixture morphol- ogy affects field aging; however, the wide spread of the data also may reflect field production and placement variability. Figure 58 (b) shows the field core results of the most extreme cases (i.e., most severe aging versus least aging). The figure shows that the WesTrack section with the high- est air void content and lowest binder content experienced the most aging whereas the section with the lowest air void content and highest binder content experienced the least aging based on the log G* values, which matches expectations. A higher air void content allows oxygen to perco- late through the mix more easily than a lower air void content, and a lower film thickness creates a shorter diffusion path. Figure 58 (c), (d), and (e) present the results of the WesTrack field sec- tions with optimum binder content, high binder content, and low binder content, respectively. Generally, a higher air void content at a given asphalt content leads to higher field aging. These results highlight the need for the development of a diffusion model that considers the morphological properties of asphalt mixtures in order to predict field aging more accurately. In addition, the calibrated kinetics modeling framework presented here allows only for the predic- tion of the G* at 64°C and 10 rad/s. Therefore, future research is suggested to establish a means to predict the effects of aging over the range of temperatures and loading rates experienced by pavements and thus enable the prediction of the changes in asphalt mixture properties with time in pavement performance prediction frameworks. Determination of Loose Mix Aging Kinetics Parameter from USAT Binder Aging The mixture-specific parameter values within the kinetics model expressed in Equation 25 are determined using loose mixture aging. Aging loose mixture in the oven allows the physico- chemical effects of filler on asphalt binder oxidation rates to be captured, which is important when predicting field aging (Moraes and Bahia 2015, Wu et al. 2014, Petersen 2009, Little et al. 2006, Recasens et al. 2005, Huang et al. 2002, Jones 1997, Petersen et al. 1987). However, Location Field Age Mix Mix ID Binder Content Air Void Content Dayton, NV 19 Years WesTrack Fine (constructed in 1995) Low %aca, High %Vab 4.7% 12% Low %ac, Medium %Va 4.7% 8% Optimum %ac, High %Va 5.4% 12% Optimum %ac, Medium %Va 5.4% 8% Optimum %ac, Low %Va 5.4% 4% High %ac, Medium %Va 6.1% 8% High %ac, Low %Va 6.1% 4% a %ac stands for asphalt content. b %Va stands for air void content. Table 19. WesTrack Fine sections used to evaluate the effects of mixture morphological properties on field aging.

84 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction determining the kinetics model parameter values using loose mixture aging is cumbersome. The loose mixture must be prepared and conditioned in an oven at 95°C. The binder must be extracted and recovered at a minimum of three durations of laboratory conditioning. The extracted and recovered binder then must be subjected to DSR testing to determine the G* value at 64°C and 10 rad/s as a function of laboratory aging time. Therefore, an alternative means to calibrate the kinetics model using binder aging tests rather than mixture aging tests could improve the feasibility of implementation. Correspondingly, an experiment was conducted to relate the binder-derived aging rates to the corresponding loose mixture aging rates. The results of this effort allow two options for calibrating the kinetics model: 1. Loose mixture aging coupled with extraction and recovery (as described). 2. USAT binder aging coupled with an empirical model to predict loose mixture aging based on filler content and type. As previously discussed, the calibration of the kinetics model parameters requires the use of a kinetics-controlled oxidation experiment. Farrar et al. (2014) proposed the USAT for the efficient simulation of asphalt binder aging in the laboratory. The USAT uses thin binder films (0.3-mm thick) to induce a kinetics-controlled reaction (Farrar et al. 2014). Binder samples conditioned for different durations can be tested in the DSR to derive a kinetics model. There- fore, the USAT was selected for binder aging herein. To best mimic loose mixture aging, USAT short-term aging was conducted in an oven at 135°C for 4 hours followed by long-term aging in the same oven at 95°C. The evolution of log G* at 64°C for the USAT-aged binder was compared to that of binder extracted and recovered from equivalent loose mixture aging and was used to establish an empirical model to relate USAT binder aging to loose mixture aging. Figure 60 shows a summary of the experimental plan. Because the properties of the binder, filler, and aggregate may affect the relationship between the binder aging rates and the loose mixture aging rates, a broad range of materials, detailed in Table 20, was used to develop an empirical model relating the USAT binder aging rates to the loose mixture aging rates. A subset of the materials detailed in Table 20 was used to validate the model’s extendibility to other mixtures rather than to calibrate the empirical model. Binder USAT Aging Loose Mix Aging Oven Aging at 95°C for Various Durations Binder Extraction and Recovery DSR Testing Compare Age Hardening Rates Empirical Relationship between Loose Mix Aging Rates and Binder USAT Aging Rates DSR Testing Figure 60. Summary of experimental plan.

Findings and Applications 85 Figure 61 shows the log G* at 64°C and 10 rad/s results for both the USAT and loose mix aging trials for the five mixtures evaluated that contain hydrated lime. Figure 62 shows the results of the five different mixtures evaluated that do not contain hydrated lime. Hydrated lime is known to delay oxidation significantly (Petersen 2009, Little et al. 2006, Recasens et al. 2005, Huang et al. 2002, Jones 1997, Petersen et al. 1987). It is clear that the binder that was conditioned using the USAT aged significantly faster than when it was aged within a loose mixture, indicating that the filler inhib- its oxidation. Similarly, the binder that was conditioned to simulate short-term aging using USAT aged more quickly than binder extracted from short-term aged loose mixture. The inclusion of hydrated lime appears to lead to a greater difference in the log G* evolution between loose mixture aging and USAT binder aging, especially the slope of the secondary region that exhibits a linear relationship. Figure 63 presents the log G* evolution results for all ten mixtures evaluated using (a) loose mixture aging and (b) USAT binder aging. It can be seen that the relative log G* results among the different mixtures are remarkably similar for USAT and loose mixture aging, which indicates that the effect of the binder on the relative aging rate of different loose mixtures is more signifi- cant than the aggregate or filler type. An empirical model was developed to relate the binder and loose mixture aging rates as shown in Equations 27 through Equation 35 to obtain loose mixture aging rates using USAT testing with least squares regression. Equation 27 represents the shape of the log G* evolution, including the spurt and constant rate reaction periods. Separate model parameter values are included for the Project Mix MIX ID Binder Source Modifier / Anti- Stripping Agent Binder Grade Binder Content Aggregate Type Hydrated Lime Content %Pass Sieve #200 (P200) FHWA ALF ALF- CTRL Citgo- Venezuelan Bachaquero Anti- Stripping Agent (0.001) PG 70- 22 5.3% Limestone / Diabase (traprock) 1.0% 6.3% LTPP South Dakota LTPP-SD N/A None 120- 150 pen 5.9% N/A None 4.6% New Mexico LTPP- NM N/A None AC-20 7.6% N/A None 6.1% Texas LTPP-TX N/A None AC-20 5.4% N/A None 5.7% Wisconsin LTPP-WI N/A None N/A 5.9% N/A None 4.1% WesTrack Fine Section WesTrack Fine West Coast Refinery None PG 64- 22 5.4% Andesite / Granite / Sand 1.5% 5.4% Coarse Section WesTrack Coarse Idaho Asphalt None PG 64- 22 5.7% Crushed Andesite / Sand 1.5% 6.5% Validation Sections SHRP AAD SHRPAAD California Coast None PG 58- 28 5.3% Limestone / Diabase (traprock) 1.0% 6.3% AAG SHRPAAG California Valley None PG 58- 10 5.3% Limestone / Diabase (traprock) 1.0% 6.3% NC DOT NCS9.5B NC Citgo- Wilmington, NC None PG 64-22 6.6% Granite None 5.7% Table 20. Mixtures used to compare binder aging rates obtained from USAT binder aging and loose mix aging.

86 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction log G* = 0.0445 (D) + 1.7861 R² = 0.98 log G* = 0.0019 (D) + 0.696 R² = 0.99 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) (a) ALF Control log G* = 0.0161 (D) + 2.1611 R² = 0.64 log G* = 0.0285 (D) + 2.5199 R² = 1 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) (c) WesTrack Fine log G* = 0.0437 (D) + 1.4079 R² = 0.99 log G* = 0.0625 (D) + 1.6739 R² = 0.97 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) (b) WesTrack Coarse log G* = 0.0519 (D) + 1.4637 R² = 0.94 log G* = 0.0708 (D) + 1.7775 R² = 0.99 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) (d) SHRP AAD log G* = 0.0409 (D) + 0.567 R² = 1 log G* = 0.0436 (D) + 0.976 R² = 1 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) (e) SHRP AAG Figure 61. Measured USAT binder aging rates and loose mix aging rates at 95°C for mixtures with hydrated lime: (a) ALF-Control, (b) WesTrack Coarse, (c) WesTrack Fine, (d) SHRP AAD, and (e) SHRP AAG.

Findings and Applications 87 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) log G* = 0.0476 (D) + 0.2006 R² = 0.98 log G* = 0.0473 (D) + 0.6663 R² = 0.99 (a) LTPP-NM log G* = 0.058 (D) + 0.389 R² = 0.94 log G* = 0.0663 (D) + 0.9795 R² = 0.98 (b) LTPP-SD log G* = 0.0023 (D) + 1.1155 R² = 0.99 log G* = 0.0024 (D) + 1.6942 R² = 0.98 (c) LTPP-TX log G* = 0.0497 (D) + 1.1089 R² = 0.96 log G* = 0.0592 (D) + 1.466 R² = 0.99 (d) LTPP-WI log G* = 0.064 (D) + 1.434 R² = 0.98 log G* = 0.0719 (D) + 1.7854 R² = 0.99 (e) NCS9.5B Figure 62. Measured USAT binder aging rates and loose mix aging rates at 95°C for mixtures without hydrated lime: (a) LTPP New Mexico, (b) LTPP South Dakota, (c) LTPP Texas, (d) LTPP Wisconsin, and (e) NCS9.5B.

88 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction mixtures with and without hydrated lime. The model also considers the amount of filler that is present in the mixture using the mass percentage of aggregate that passes the No. 200 sieve, as presented in Equations 28 through Equation 35. In order to account for the effect of filler on the reaction rate of loose mix aging during the slow (constant) reaction rate stage, Equation 28 is used to convert the slope of the USAT binder log G* curve to that of the corresponding loose mixture. Figure 64 shows the relationship between the slope correction factors (s) and percentage passing the No. 200 sieve (P200) and the regression equations used to develop Equation 28. Equation 29 is used to determine the intercept adjustment from USAT binder aging to loose mixture aging. The parameter F is another shape factor that corrects the first-degree term constant during the fast reaction state. Figure 65 summarizes the developed empirical model. log * , 8 8 8 8 , 8 (27) 2 2 G a D b D c D a b c d D e D ( ) ( ) ( ) ( ) ( ) = + + ≤ + + + − + >    3.51 0.32 , hydrated lime 1.60 0.1 . other fillers (28) 200 200 s P P ( ) ( ) = − −  (a) (b) 0 1 2 3 4 5 6 0 10 20 30 WesTrack Fine WesTrack Coarse ALF Control NCS9.5B SHRP AAD SHRP AAG LTPP Wisconsin LTPP New Mexico LTPP South Dakota LTPP Texas Loose Mix Aging at 95°C lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 10 20 30 0 1 2 3 4 5 6 WesTrack Fine WesTrack Coarse ALF Control NCS9.5B SHRP AAD SHRP AAG LTPP Wisconsin LTPP New Mexico LTPP South Dakota LTPP Texas USAT Aging at 95°C lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) Figure 63. Ranking comparisons: (a) loose mix aging rates and (b) USAT binder aging rates. y = -0.32x + 3.51 R² = 0.99 y = -0.10x + 1.61 R² = 1.00 0.8 1.2 1.6 2 3 4 5 6 7 Sl op e Co rr ec tio n % Passing No. 200 Sieve (P200) Hydrated Lime Other Mineral Fillers Figure 64. Relationship between slope correction factor and percentage passing No. 200 sieve.

Findings and Applications 89 034, hydrated lime 0.5, other fillers (29)I =  (30)a al u= (31)b b Fl u= − 8 8 8 8 (32) 2 F c b a d el u l l l( ) ( )( ) = + × + × − × + (33)c c Il u= − (34)d d Sl u= (35)e e Il u= − where D = aging duration (days), al , bl, cl, dl, el = loose mix dual-mechanism model parameters, au, bu, cu, du, eu = USAT dual-mechanism model parameters, S = slope correction factor to account for filler effects, I = intercept correction factor to account for filler effects, F = correction factor between fast and slow (constant) rates of reaction mechanism, and P200 = percentage passing No. 200 sieve. Figure 65. Summary of developed empirical model to relate binder and loose mix aging rates.

90 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction The proposed model was applied to the binder log G* results obtained from the USAT binder aging tests to predict the corresponding loose mixture aging evolution. Figure 66 and Figure 67 present comparisons between the measured and predicted loose mixture aging rates for the five mixtures that were used to calibrate the empirical model using mixtures with and without hydrated lime, respectively. The results demonstrate excellent agreement between the measured and pre- dicted log G* values. Figure 68 shows comparisons between the measured and predicted loose mixture aging rates for the independent validation mixtures. Note that two of the validation mixtures do not contain hydrated lime whereas the third does. Generally, the prediction accuracy is very good, thereby validating the use of the empirical equation to obtain the evolution of log G* for the loose mix using USAT binder test results. These results suggest that the calibrated mixture-specific param- eters of the kinetics model can be determined using USAT binder aging, thereby negating the need for loose mixture aging and corresponding extraction and recovery. Applications Integration of Pavement Aging Model in Mechanistic–Empirical Design In Pavement ME Design, the material properties, pavement structure, and traffic and climatic conditions are used as inputs to evaluate a trial pavement structural design. The trial design is then examined using the response models that are incorporated in the mechanistic–empirical 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) (c) WesTrack Fine (b) WesTrack Coarse (a) ALF Control Figure 66. Predicted age hardening rates for loose mix aging at 95°C from USAT binder aging for mixtures with hydrated lime: (a) ALF-Control, (b) WesTrack Coarse, and (c) WesTrack Fine.

Findings and Applications 91 design to determine stresses, strains, and deformations using layered elastic analysis. The response model output is used to predict the corresponding pavement distress development over the design period. The dynamic modulus of the asphalt mixtures in the individual layers within the pavement is a key property for predicting the pavement responses and correspond- ing distress development over the designated service life of the pavement. Based on the pre- dicted pavement performance, the trial pavement design is adjusted to obtain the thicknesses and material properties that are required to achieve adequate performance. The GAS model, developed by Mirza and Witczak (1995), is applied within Pavement ME Design to predict the changes in asphalt binder viscosity that are due to oxidative aging over the course of a pavement’s service life. The predicted asphalt binder viscosity values are then incorporated into the Witczak equation to predict the corresponding dynamic modulus (|E*|) evolution with oxidative aging. The GAS model assumes that aging is limited to the top 1.5 in. of the pavement. Therefore, within Pavement ME Design, the material properties below 1.5 in. are not considered. Consequently, bottom-up fatigue cracking predictions are not sensitive to pavement age. Based on the findings from this study, oxidative aging occurs throughout the entire thickness of the asphalt layers within a pavement. Hence, the current Pavement 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) (a) LTPP-NM (b) LTPP-SD (c) LTPP-TX (d) LTPP-WI Figure 67. Predicted age hardening rates for loose mix aging at 95°C from USAT binder aging for mixtures without hydrated lime: (a) LTPP New Mexico, (b) LTPP South Dakota, (c) LTPP Texas, and (d) LTPP Wisconsin.

92 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) (a) SHRP AAD 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) (b) SHRP AAG 0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 lo g G *, 64 °C , 1 0 ra d/ s (kP a) Aging Duration (Days) (c) NCS9.5B ME Design methodology needs to be modified and improved to consider the effects of aging on cracking performance. A detailed explanation of the GAS model is presented in Appendix B. The kinetics modeling framework developed herein could potentially be used to replace the GAS model in Pavement ME Design for the prediction of the binder G* evolution throughout a pavement’s service life. However, the predicted G* values must be input into a model to predict the corresponding dynamic modulus evolution. Therefore, in this study, the Witczak equation (Bari and Witczak 2006), Hirsch model (Christensen et al. 2003), and the North Carolina State University Artificial Neural Network (NCSU ANN) model (Sakhaei Far 2011) were used to evaluate the accuracy of these existing dynamic modulus predictive models when they are applied to highly aged materials. These models were evaluated using the laboratory- aged mixture dynamic modulus test results of the LTPP South Dakota asphalt mixture. Laboratory-fabricated loose mixture was short-term aged and then subjected to long-term aging at 95°C in the oven for 4, 8, and 16 days prior to producing specimens for dynamic modulus testing. The aged loose mixture samples were compacted in a Superpave gyratory compactor to achieve 4.5% air voids in the test specimens. Four small Ø38-mm diam- eter, 110-mm tall cores were extracted vertically from the inner 100-mm diameter of the gyratory-compacted specimens. The small specimens were subjected to dynamic modulus testing at 4°C, 20°C, and 40°C with loading frequencies ranging from 0.1 Hz to 25 Hz. Fig- ure 69 presents the resulting dynamic modulus master curves that demonstrate the rate of increase in the dynamic modulus values that is caused by oxidation decay with the duration of laboratory aging. Figure 68. Validation of predictive model: (a) SHRP AAD, (b) SHRP AAG, and (c) NCS9.5B mixtures.

Findings and Applications 93 Asphalt binder samples were extracted and recovered from the aged loose mixtures and sub- jected to DSR testing to determine the binder G* input for the dynamic modulus predictive models. Table 21 shows the G* results at 20°C and 10 Hz for the asphalt binders extracted and recovered from loose mixtures aged for 4, 8, and 16 days. The G* results demonstrate a con- tinuous increase in the asphalt binder G* value even after 8 days of aging. However, the asphalt mixture dynamic modulus test results show a comparably small change in dynamic modulus values between 8 and 16 days of aging. An intermediate temperature of 20°C and loading frequency of 10 Hz were selected for the dynamic modulus predictions. Figure 70 shows the comparisons between the model predic- tions and the measured dynamic modulus values at 20°C and 10 Hz frequency. The results demonstrate that the measured dynamic modulus values of the long-term aged material are significantly lower than the model predictions. The Witczak model over-predicted the dynamic modulus values most significantly. The Hirsch and ANN model predictions have a maximum error of approximately 50%. The preliminary evaluation of the three dynamic modulus prediction equations suggests that the Witczak model, currently used within Pavement ME Design, lacks the ability to predict dynamic modulus values of aged mixtures accurately. Therefore, the replacement of the Witczak equation in Pavement ME Design with a more accurate dynamic modulus prediction model merits consideration in future work. Alternative dynamic modulus predictive equations (i.e., NCSU ANN and Hirsch), although more accurate than the Witczak model, can still lead to significant error in the prediction of asphalt mixture dynamic modulus values, suggesting the need for recalibration using a broad set of materials with various age levels in future work. 1.0E+05 1.0E+06 1.0E+07 1.0E+08 1.0E-04 1.0E-02 1.0E+00 1.0E+02 1.0E+04 |E* | (M Pa ) Reduced Frequency (Hz) Short-Term Oven, 95°C, 4 Days Oven, 95°C, 8 Days Oven, 95°C, 16 Days Figure 69. Comparison of dynamic modulus master curves at different aging levels for LTPP South Dakota mix. Sample Aging Temperature Aging Duration G* at 20°C and 10 Hz Dynamic Modulus at 20°C and 10 Hz STA 135°C - 7,390 kPa 4,523 MPa 4 D 95°C 4 days 13,646 kPa 6,364 MPa 8 D 95°C 8 days 16,621 kPa 7,744 MPa 16 D 95°C 16 days 21,567 kPa 8,473 MPa Table 21. G* results for asphalt binders extracted and recovered from loose mix aged at different durations.

94 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction Conclusions The accurate characterization of asphalt mixture properties in terms of the service life of a pavement is becoming more important as more powerful pavement design and performance prediction methods are implemented. This project sought to develop a procedure to simulate the long-term aging of asphalt mixtures as a function of climate and depth for performance test- ing and prediction. The results of this project provide a basis for the future development of a methodology that integrates the effects of long-term aging in mechanistic–empirical pavement analysis programs such as Pavement ME Design. Sensitivity Study The goal of the sensitivity study was to estimate the sensitivity of the mechanical properties of asphalt concrete to asphalt binder oxidation. The sensitivity study provided thresholds by which to evaluate the significance of observed differences in asphalt binder AIPs in terms of asphalt mixture performance. The following conclusions pertain to the sensitivity study. • The properties of asphalt concrete are not proportionally sensitive to changes in the asphalt binder modulus. • To replicate the effects of binder oxidation on the modulus of asphalt concrete with a given level of accuracy (say 10%), the desired binder oxidation can be replicated at less accuracy by a factor of 1.5 to 3.6 (15% to 36 %). • If the binder AIPs are replicated with a certain percentage of error (say 10%), then the expected percentage of error in the resulting fatigue properties of an asphalt mixture that is tested will be lower than or equal to 10%. Selection of the Chemical and Rheological Aging Index Properties Candidate chemical and rheological AIPs were evaluated in order to select the AIPs needed to track the oxidation levels of field-aged and laboratory-aged materials when developing the 4.5E+03 6.4E+03 7.7E+03 8.5E+03 1.2E+04 1.4E+04 1.6E+04 1.9E+04 6.3E+03 7.6E+03 9.1E+03 1.2E+04 7.2E+03 8.7E+03 1.0E+04 1.2E+04 0.0E+00 6.0E+03 1.2E+04 1.8E+04 2.4E+04 Short-Term Aged Oven, 95°C, 4Days Oven, 95°C, 8Days Oven, 95°C, 16Days |E* | (M Pa ) |E*| _Measured |E*| _ Witczak |E*| _Hirsch |E*| _NCSU ANN Figure 70. Comparison between dynamic modulus values measured at 20°C and 10 Hz and predicted values based on extracted and recovered binder data measured at 20°C and 10 Hz.

Findings and Applications 95 long-term aging procedure and associated kinetics model. The following conclusions pertain to the selection of the chemical and rheological AIPs. • All three chemical AIPs evaluated [i.e., carbonyl plus sulfoxide (C + S) peaks, carbonyl area, and C + S area] showed good sensitivity to aging duration; however, the C + S peaks were the most reliable among the three chemical AIPs. • Including the effect of the sulfoxide functional group in the oxidative aging evaluation was crucial, as different aging rates were observed from carbonyl alone and from the C + S index values. • The dynamic shear modulus (G*) as a direct output of the DSR can be used to develop a relationship between the chemistry and rheology of asphalt binder. It is simple and more consistent with chemical indices than other rheology-based AIPs. Selection of the Long-Term Aging Method An experimental program was conducted to select the proposed long-term aging method. The following factors were evaluated to select the proposed laboratory aging procedure: (a) state of the material during aging (compacted specimen versus loose mix), (b) pressure level (oven aging versus pressurized aging), and (c) aging temperature (95°C versus 135°C). The following con- clusions pertain to the selection of the proposed long-term aging method. • The current standard procedure for the long-term aging of asphalt mixtures (AASHTO R 30), which consists of conditioning large 100-mm diameter compacted specimens in an oven at 85°C, leads to the development of an oxidation gradient from the specimen’s center to its periphery. The lack of uniform properties throughout the specimen is of concern for performance testing and was observed directly in cyclic direct tension fatigue tests through a high rate of end failure at specimen locations where oxidation was most significant. • The application of pressure in the compacted specimen aging process expedites oxidation. However, the performance test results indicate that the application and/or release of pressure damages specimens. • The aging gradient observed in large compacted specimens that were subjected to oven aging was eliminated by the use of small specimens (38-mm diameter with 100-mm height) due to their shortened lateral diffusion paths. Therefore, the oven aging of small compacted speci- mens is the most promising compacted specimen aging procedure, as no integrity issues were observed. • Aging asphalt mixtures in a loose mix state expedites oxidation compared to compacted speci- men aging under the same conditions. • The compactive effort required to compact long-term aged loose mixes is comparable to that required for short-term aged mixes, with no adjustment to the compaction temperature needed based on the results for two mixtures, PG 64-22 and PG 70-28 SBS-modified, the latter of which is known to be difficult to compact. • The performance test results indicate no problems with loose mixtures compacted after long-term aging. • Pressure expedites the aging of loose mix. However, the size of the standard binder PAV pro- hibits the generation of enough aged material for performance testing. If the pressure aging of loose mix were to be selected as the best aging procedure, then a larger PAV would need to be developed. • Loose mix aging in an oven is the most promising aging procedure to produce mixture speci- mens for performance testing in terms of efficiency and integrity, without the need to develop costly new equipment. In addition, any specimen geometry (e.g., beams) can be produced using aged loose mix, also making this procedure the most versatile option.

96 Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction • The results indicate that loose mix aging at 95°C is the optimal procedure for the long-term aging of asphalt concrete for performance testing. The literature indicates that the disruption of polar molecular associations and sulfoxide decomposition become critical at temperatures that exceed 100°C (Petersen 2009). In this study, 95°C was selected as the aging temperature instead of 100°C in order to avoid the aging temperature reaching close to 100°C due to pos- sible temperature fluctuations in the oven. Aging at lower temperatures precludes reaching field levels of oxidation within a reasonable time. • A significant change in the relationship between binder rheology and chemistry can occur for certain asphalts when the aging temperature is increased from 95°C to 135°C. This change implies corresponding changes in the kinetics and mechanisms of the oxidation reactions that are associated with an increase in temperature from 95°C to 135°C, which is consistent with findings described in the literature. • Asphalt mixture performance can be negatively impacted by long-term aging at 135°C. Despite having matching rheological characteristics, two of the three mixtures evaluated in this study exhibited decreases in both dynamic modulus values and fatigue resistance. These results suggest that aging at 135°C for performance assessment and prediction should be avoided. • When the loose mix laboratory aging temperature is at or below 95°C, the relationship between binder chemistry and rheology is unaffected by the aging temperature, based on all three mixtures evaluated, indicating that the aging temperature does not affect the oxidation reaction mechanism. • The rate of oxidation increases with an increase in temperature, and thus, the results suggest that the optimal loose mixture laboratory aging temperature is 95°C. • It is proposed that loose mixture aging at 95°C should be used for the laboratory long-term aging of both WMA and HMA mixtures. Future research is necessary to investigate the advan- tages and disadvantages of compacting WMA after long-term aging. Climate-Based Determination of Predefined Aging Durations After selecting the proposed long-term aging method, a means to determine the laboratory aging durations that are required to represent a given time, climate, and depth within a pave- ment was developed. Project-specific laboratory aging durations that are required to match the AIPs of field cores at varying depths were determined for a broad set of materials. The project-specific aging durations were used to calibrate a kinetics-derived CAI that can be used to determine the required laboratory aging duration to match the field aging at any location of interest. The following conclusions pertain to the selection of the proposed long-term aging method. • Loose mixture oven aging leads to a kinetics-controlled oxidation reaction, indicating that kinetics models can be derived and applied to loose mixture aging without the need to con- sider diffusion. • A kinetics model for loose mixture aging was developed and then validated using log G* at 64°C and 10 rad/s frequency as the AIP within Glaser et al.’s (2013a) kinetics model frame- work that was developed initially for binder aging. The kinetics model can be calibrated using AIP measurements obtained from isothermal aging at a single temperature. • The laboratory aging duration required to match a given field condition is independent of the material-specific kinetics. • The CAI, developed by simplifying the kinetics model and calibrating against field data, can prescribe laboratory aging durations to match a given field condition using hourly pavement temperature histories at depths of 6 mm, 20 mm, and 50 mm or below.

Findings and Applications 97 Aging Model to Predict Field Aging This study evaluated the ability of the kinetics model developed herein to predict the evolu- tion of oxidative aging in pavements. The prediction of the changes in asphalt binder properties with oxidative aging is important for the simulation of changes in asphalt mixture properties that are induced by aging within pavement performance prediction frameworks, including Pavement ME Design. The following conclusions pertain to the kinetics aging model. • The kinetics model was calibrated successfully (R2 = 0.75) against field core measurements of log G* at a depth of 20 mm, which represents a reasonable depth for the evaluation of the bulk behavior of surface layers and avoids the effect of UV oxidation at the pavement surface. • The calibrated kinetics model tends to under-predict aging near the pavement surface and over predict aging at depths below 20 mm, which matches intuition in the absence of a dif- fusion model and consideration of UV oxidation at the pavement near-surface. These results highlight the need for the development of a diffusion model that considers the morphological properties of asphalt mixtures in order to predict field aging more accurately. • An empirical model was developed and validated to relate the USAT asphalt binder aging rates and the loose mixture aging rates. The model allows USAT binder testing to be used to determine the mixture-specific coefficients that are included within the calibrated kinetics model, thereby negating the need for cumbersome loose mixture aging and corresponding extraction and recovery. Integration of the Pavement Aging Model in Mechanical–Empirical Design Within pavement performance prediction frameworks, including Pavement ME Design, the predicted changes in binder properties as a result of oxidation must be input into a model to predict the corresponding changes in asphalt mixture properties. Therefore, a preliminary investigation into the accuracy of existing dynamic modulus predictive models with regard to highly aged materials was conducted using the Witczak equation (Bari et al. 2006), Hirsch model (Christensen et al. 2003), and the NCSU ANN model (Sakhaei Far 2011). The following conclu- sions pertain to the evaluation of dynamic modulus predictive models. The three evaluated dynamic modulus prediction equations lack the ability to accurately pre- dict dynamic modulus values from binder properties. The Witczak model, used in Pavement ME Design, over-predicts the dynamic modulus values most significantly. The Hirsch and ANN model predictions have a maximum error of approximately 50%.

Next: Chapter 4 - Suggested Future Research »
Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

TRB's National Cooperative Highway Research Program (NCHRP) Research Report 871: Long-Term Aging of Asphalt Mixtures for Performance Testing and Prediction presents a proposed standard method for long-term laboratory aging of asphalt mixtures for performance testing. The method is intended for consideration as a replacement for the method in AASHTO R 30, “Mixture Conditioning of Hot Mix Asphalt (HMA),” which was the most commonly used method for aging asphalt materials for performance testing for input to prediction models for the past 25 years. The method improves on R 30 in that the laboratory aging time is specifically determined by the climate at the project location.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!