Long-Term Field Performance of Warm Mix Asphalt Technologies (2017)

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.

21 Introduction This chapter compares the results of in-the-field transverse cracking for the HMA and WMA pavements based on the 2012–2013 (first-round) and 2014–2015 (second-round) dis- tress survey results. Identification of significant determinants for transverse cracking using two statistical based approaches is also discussed. Figure 3.1 shows a typical transverse crack in an asphalt pavement overlay. For most cases in this study, the trans- verse cracks are surface-initiated cracks, and they could be a combination of thermal cracking and reflective cracking, as described in Chapter 2. Comparisons of Transverse Cracking in the Field Comparisons Based on First-Round Distress Survey Results: 2012–2013 This section presents a summary of the first-round dis- tress survey results of the HMA versus WMA comparisons in terms of transverse cracking. Figure 3.2 (a) shows that out of the 28 projects evaluated, 14 in-service projects exhibited transverse cracking in the first-round distress survey. The transverse cracking comparisons include 35 HMA-WMA pairs. Those projects that did not exhibit transverse cracks are also included. Note that the four HMA-WMA pairs in the TN SR 46 project were not included in the transverse crack- ing comparisons because the overlay thickness was found to be 0.75 inches. While functional, this 0.75-inch overlay was considered to be one that can crack easily. In Figure 3.2 (b), H > W, H = W, and H < W denote that the HMA pavement has, respectively, longer, comparable, or shorter transverse crack lengths than the WMA pave- ment within a project. For those projects that did not exhibit transverse cracking, the HMA-WMA pairs are treated as “comparable.” Overall, among the 35 HMA-WMA pairs, the HMA and WMA pavements exhibited comparable transverse cracking resistance. Of the 6 pairs that were not comparable, the HMA pavements typically had longer transverse cracks than the WMA pavements (5 pairs) [see Figure 3.2 (b)]. These five pairs include the following projects: MO Hall St. (HMA > Sasobit, HMA > Evotherm); MN TH 169 (HMA > Evo- therm); and CO I-70 (HMA > Evotherm, HMA > Advera). However, in the second-round distress survey (after 2 years), these pairs showed comparable transverse cracking between HMA and WMA pavements. This finding suggests that some WMA pavements age and deteriorate faster than HMA pave- ments in terms of resistance to transverse cracking. A similar hypothesis was also proposed in NCHRP Report 763 (Martin et al. 2014): the initial stiffness of the WMA is less than the stiffness of conventional HMA and this gap can be reduced with increased time in the field. Figure 3.3 plots the transverse crack length versus pavement age for the first-round distress survey, with WMA and HMA projects separately identified. As shown, transverse cracking is mostly seen in pavements that are 4 or more years old. Younger pavements show fewer or shorter transverse cracks. Two proj- ects, the MN TH 169 project (subjected to low temperature in winter) and LA 116 project, show earlier transverse cracking. Comparison of the performance differences between WMA and HMA pavements can be found in Figure 3.2(b). The breakdown of the performance comparisons based on the different WMA technologies is shown in Figure 3.4. In general, there was no distinct difference between the WMA technology pavements and the HMA pavements. Because of the small number of projects in each of the five categories, the results cannot be statistically analyzed. Also, factors other than the HMA and WMA properties, such as pre-overlay condi- tions and pavement structure variation in the existing asphalt pavements, may confound the results. Figure 3.5 presents a summary of the comparison of trans- verse cracking among the WMA technologies. The WMA C H A P T E R 3 Transverse Cracking

transverse crack Figure 3.1. Typical transverse cracking in overlay. (b) Summary of Transverse Cracking Comparison 0 5 10 15 20 25 30 5 29 1 N o. o f P ai rs H>W H=W H<W (a) Projects Having Transverse Cracking Figure 3.2. Comparison of the length of transverse cracking for HMA-WMA pairs based on the first-round distress survey results. 0 20 40 60 80 100 120 140 160 180 Tr an sv er se C ra ck L en gt h, ft /2 00 ft Pavement Age, years MN LA 116 HMA WMA 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 Figure 3.3. Transverse crack length versus pavement age for the first-round distress survey.

23 Note: The number in parentheses indicates the number of HMA-WMA pairs in each crack length comparison. (d) HMA versus Water-Based Foaming H=W (9) (e) HMA versus Water-Containing Foaming H>W (1) H=W (5) H>W (1) H=W (14) (c) HMA versus Foaming H>W (3) H=W (8) H<W (1) (b) HMA versus Chemical H>W (1) H=W (7) (a) HMA versus Sasobit Figure 3.4. Breakdown comparison of the length of transverse cracking between HMA and specific WMA technologies based on the first-round distress survey results. (c) Comparison Between Organic and Foaming N o. o f P ai rs 0 5 0 0 2 4 6 8 10 Organic>Foaming Organic=Foaming Organic<Foaming (a) Comparison Between Chemical and Organic 0 5 0 0 2 4 6 8 10 N o. o f P ai rs Chemical>Organic Chemical=Organic Chemical<Organic N o. o f P ai rs 0 4 1 0 2 4 6 8 10 Chemical>Foaming Chemical=Foaming Chemical<Foaming (b) Comparison Between Chemical and Foaming Figure 3.5. Transverse cracking comparisons among WMA additives in terms of crack length based on the first-round distress survey results.

24 technologies are divided into three categories, namely chemi- cal (including Evotherm and LEA), organic (including Saso- bit), and foaming (including foaming, DBG, AquaBlack, Advera, Gencor and Aspha-min). The number of WMA- WMA pairs is not necessarily equal to the number of HMA- WMA pairs because more than one WMA technology is used in some projects, which generates more than one WMA-WMA pair, while only one WMA technology is used in other proj- ects, which results in no WMA-WMA pair. The comparison is based on a statistical t-test with a significance level of 0.05. Based on the limited results from the first-round distress survey, Figures 3.5 (a) through (c) show that the pavements using the three WMA types result in similar field performance in terms of transverse cracking. It is noted that all the projects had a pavement age of 7 years or younger when the first-round distress survey was conducted. Comparisons Based on Second-Round Distress Survey Results: 2014–2015 This section presents a summary of the second-round distress survey results of the HMA versus WMA comparisons in terms of transverse cracking. As shown in Figure 3.6 (a), 22 projects exhibited transverse cracking in the second-round distress sur- vey compared with 14 projects in the first-round distress survey. Eight more projects (mostly in the wet climatic zones) devel- oped transverse cracking during the interval between surveys, including the new MT I-15, TN SR 125, IA US 34, and LA US 61 projects. To compare the HMA versus WMA performance based on the second-round distress survey results, 39 HMA-WMA pairs were included in the analysis. For those projects that did not exhibit any transverse cracks, the HMA-WMA pairs are treated as “comparable.” Among the 39 HMA-WMA pairs, most of the HMA and WMA pavement performance was comparable, as illustrated in Figure 3.6 (b). Of the 8 pairs that were not comparable, the HMA pavements typically had longer transverse cracks than the WMA pavements (6 pairs) [see Figure 3.6 (b)]. These six pairs include the following proj- ects: OH SR 541 (HMA > Sasobit, HMA > Evotherm, HMA > Aspha-min); SC US 178 (HMA < Evotherm); LA 3121 (HMA < Evotherm); CO I-70 (HMA > Evotherm, HMA > Advera); and NE US 14 (HMA > Evotherm). Figure 3.7 plots the transverse crack length versus pavement age for the second-round distress survey. In the second-round distress survey, three new projects (MT I-15, TN SR 125, and IA US 34 projects) exhibited early transverse cracking (whether they were HMA or WMA) when the pavement was less than 4 years old. Most of the other projects exhibited trans- verse cracking when the pavement was in the field for 4 years, with an increase in cracking in the 8th year. In terms of the different WMA technologies, as shown in Figure 3.8, the WMA pavements show transverse cracking resistance that is comparable with that of the HMA pavements. Again, after breaking down the results into the different WMA technologies, the number of projects for comparison is too small for statistical analysis. The comparison among the WMA technologies is further examined based on the second-round transverse cracking survey results. Figures 3.9 (a) and (c) show that, out of seven WMA-WMA pairs, three organic modified WMA pavements (CO I-70, TN SR 46, and OH SR 541) have significantly longer transverse crack lengths than their companion WMA pave- ments (chemical and foaming), while Figure 3.9 (b) shows that chemical and foaming WMA pavements have similar resistance to transverse cracking. This finding is different from the first- round results that showed all WMA behaving similarly with respect to transverse cracking resistance. Given that these three pavements were all 7 years old or older when the second-round distress survey was conducted, it is possible that aging is more (a) Projects Having Transverse Cracking 0 5 10 15 20 25 30 35 6 31 2 N o. o f P ai rs H>W H=W H<W (b) Summary of Transverse Cracking Comparison Figure 3.6. Comparison of the length of transverse cracking for HMA-WMA pairs based on the second-round distress survey results.

25 0 50 100 150 200 250 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 Tr an sv er se C ra ck L en gt h, ft /2 00 ft Pavement Age, years TN SR 125 IA US 34 MT I-15 HMA WMA Figure 3.7. Transverse crack length versus pavement age for the second-round distress survey. Figure 3.8. Breakdown comparison of the length of transverse cracking between HMA and specific WMA technologies based on the second-round distress survey results. Note: The number in parentheses indicates the number of HMA-WMA pairs in the transverse crack length comparison. H>W (2) H=W (14) (c) HMA versus Foaming H>W (3) H=W (8) H<W (2) (b) HMA versus Chemical H>W (1) H=W (9) (a) HMA versus Sasobit H=W (11) (d) HMA versus Water-Based Foaming H>W (2) H=W (3) (e) HMA versus Water-Containing Foaming

26 critical to the organic modified WMA pavement than to other WMA technologies in the long term. Further research is needed to evaluate the effect of long-term aging on WMA additives. Paired Ranking Analysis for Transverse Cracking The material properties obtained from the first-round field cores for both the HMA and WMA pavements were used to perform a paired ranking analysis and determine the signifi- cant determinants for transverse cracking. These properties include the following: 1. Mixture properties: dynamic modulus, creep compliance, and fracture properties at the intermediate temperature (68°F) and low temperature (14°F) including IDT strength, fracture work density, vertical failure deformation, and horizontal failure strain. 2. Binder properties: PG, G*sind, bending beam rheometer (BBR) m-value and stiffness, and fracture properties at temperatures of 41°F and 68°F including shear strength, fracture energy, and failure strain. First-Round Distress Survey Results Figure 3.10 presents a comparative summary of the numbers of HMA-WMA pairs that show a consistent trend between the material properties and transverse cracking. This comparison includes 14 in-service projects with 21 HMA-WMA pairs that exhibited transverse cracking. Detailed statistics of the paired comparison between HMA and WMA is presented in Appen- dix B. It was found that the ranking of the fracture work den- sity values obtained from IDT testing at the low temperature (14°F) provides the highest number of correlations (15 out of 21 pairs) with the ranking of transverse cracking. The rela- tionship between fracture work density and transverse crack- ing is negative, indicating that a higher fracture work density value correlates with less transverse cracking. The dynamic modulus values at 14°F and 10 Hz (15 out of 21 pairs) and BBR creep stiffness values (14 out of 21 pairs) also show good correlations but a positive relationship, indicating that a high modulus value leads to more transverse cracking. Second-Round Distress Survey Results Based on the second-round distress survey results, Fig- ure 3.11 presents a summary of the number of HMA-WMA pairs that show a consistent trend between the material prop- erties and transverse cracking. These comparisons are based on 15 in-service projects and 4 new projects, with 35 pairs that exhibited transverse cracking. Figure 3.11 shows that the ranking of the fracture work density values obtained from IDT testing at the low temperature (14°F) (25 out of 35 pairs) and the dynamic modulus values obtained at 14°F and 10 Hz (a) Comparison Between Chemical and Organic 0 4 3 0 2 4 6 8 10 N o. o f P ai rs Chemical>Organic Chemical=Organic Chemical<Organic (b) Comparison Between Chemical and Foaming 1 5 1 0 2 4 6 8 10 N o. o f P ai rs Chemical>Foaming Chemical=Foaming Chemical<Foaming (c) Comparison Between Organic and Foaming 3 4 0 0 2 4 6 8 10 N o. o f P ai rs Organic>Foaming Organic=Foaming Organic<Foaming Figure 3.9. Transverse cracking comparisons among WMA additives in terms of crack length based on the second-round distress survey results.

27 Figure 3.10. Number of HMA-WMA pairs that have consistent rankings between material property ranking and transverse cracking ranking based on the first-round distress survey results. 0 3 6 9 12 15 18 21 N um be r o f p ai rs Positive Negative BBR stiffness (10.4°F) Binder shear strength (41°F) Binder fracture energy (41°F) Mix E* (14°F) Binder shear strength (68°F) Mix IDT strength (68°F) Mix work density (14°F) Binder failure strain (41°F) BBR m-value (10.4°F) Binder failure strain (68°F) Mix horizontal failure strain (68°F) Mix vertical failure deformation (68°F) Figure 3.11. Number of HMA-WMA pairs that have consistent rankings between material property ranking and transverse cracking ranking based on the second-round distress survey results. 0 5 10 15 20 25 30 35 N um be r o f p ai rs Positive Negative BBR stiffness (10.4°F) Binder shear strength (41°F) Binder fracture energy (41°F) Mix E* (14°F) Binder shear strength (41°F) Binder shear strength (68°F) Mix work density (14°F) Binder failure strain (41°F) BBR m-value (10.4°F) Binder failure strain (68°F) Mix horizontal failure strain (68°F) Mix vertical failure deformation (68°F)

28 (24 out of 35 pairs) continue to provide the highest number of correlations for the ranking of transverse cracking. These results are consistent with those of the first-round distress survey. Statistical Predictive Models for Transverse Cracking The prediction model for transverse cracking is highly complicated due to several factors, such as the difficulty in dif- ferentiating thermal cracking from reflective cracking in the field, the high variability of field conditions, and the poten- tial variability in crack initiation and cracking propagation mechanisms. As shown in Appendix E, the AASHTOWare Pavement ME Design analysis program can only estimate the general trend of field transverse cracking. The program’s pre- dictive quality is limited, especially for pavements with con- crete structures beneath them and/or under extreme climate conditions. As a result, a statistical-based approach is used in this study for field performance predictions. The PLS method was used to develop a predictive transverse cracking model by sorting out a wide spectrum of potential influencing factors and at the same time considering their col- linearity. These factors were determined based on literature findings and include the following: • Mixture properties: IDT strength at 14°F, creep compliance, m-value of creep compliance, fracture work density at 14°F, in-place air voids, and aggregate passing #4, #8, and #200 sieve size. • Binder properties: binder low temperature PG, binder BBR stiffness, m-value of binder, asphalt content, and effective binder content. • Pavement structure: HMA thickness and overlay thickness. • Climate, traffic, and pavement age parameters: low- temperature hours, average annual daily truck traffic (AADTT), and pavement age. Table 3.1 contains the range of each independent variable value measured during this project. To make the best use of the limited field performance data, the LOOCV method was applied to validate the prediction model. The LOOCV method uses every single data point once and only once. The PRESS method was used together with PLS and LOOCV to determine the optimum number of predictor variables. Appendix G provides details about the statistical approaches used in developing the performance predictive model. Transverse Crack Propagation Model The PLS method develops the crack propagation model using the six highest standardized coefficient predictor vari- ables, as shown in Equation (3.1). Details of the model develop- ment can be found in Zhang et al. (2015) and Shen et al. (2016) and in Appendix G. Y e (3.1)X X X X X X8.825 11.665 9.587 0.0033 0.267 1.047 0.00061 2 3 4 5 6= ( )− − + − − + where Y = transverse crack length, ft/200 ft; e = base of natural logarithm, approximately 2.718; X1 = fracture work density (14°F), MPa; X2 = creep compliance (32°F), 1/GPa; X3 = average yearly low-temperature hour, hr (from AASHTOWare Pavement ME Design); X4 = percentage passing #200 sieve, %; X5 = overlay thickness, inch; and X6 = AADTT. The effect of the individual predictor variables is in good agreement with findings in the literature: (a) mixtures with a higher fracture work density value and creep compliance value may be more fracture-resistant than mixtures with lower values (Hiltunen and Roque 1994, Wen 2013); (b) more hours at a low temperature (<15°F) result in higher transverse cracking potential (ARA 2003); (c) within a specific range, a higher percentage passing the #200 sieve introduces more fine aggregate particles, increases the demand of asphalt con- tent (higher surface area), and could result in a more flexible mixture with lower cracking potential (Schwartz et al. 2011); (d) a thicker overlay thickness can help to reduce thermal and reflective cracking by reducing the movement of existing cracks and reducing potential stress concentration (Hiltunen Predictor Variables Unit Range IDT strength, 14°F Mpa 2.17-5.56 Creep compliance (D1), -4°F 1/Gpa 0.04-0.16 Creep compliance (D2), 14°F 1/Gpa 0.05-0.18 Creep compliance (D3), 32°F 1/Gpa 0.07-0.71 m-value, creep compliance N/A 0.11-0.46 Fracture work density, 14°F Mpa 0.02-0.13 Air voids % 1.8-9.1 Binder low PG °C -6.5 to -28.4 Binder stiffness, BBR Pa 21.5-414.2 m-value, BBR, -6°C N/A 0.23-0.48 Asphalt content % 4.2-10.0 Effective binder content % 6.7-14.4 Percent passing #4 sieve % 45-82.2 Percent passing #8 sieve % 24.9-57.6 Percentage passing #200 sieve % 2.8-11.9 NMAS mm 9.5-12.5 Service life year 2-7 Low temperature hour hour 0.1-1269.8 HMA thickness inch 2.0-14.3 Overlay thickness inch 1.0-6.0 AADTT N/A 10-3380 Table 3.1. Predictor variables and their ranges considered in the transverse cracking models.

29 and Roque 1994; Fromm and Phang 1972); and (e) higher traffic loads (AADTT) can lead to more crack propagation. Figure 3.12 (a) shows the relationship between field- measured and predicted transverse cracking according to Equation (3.1), with the HMA and WMA results separated. This model has a coefficient of determination (R2) value of 0.82, standard error of the estimate of 0.47, and a Mallow’s Cp of 7.0, all of which indicate good prediction quality. The regression model has the capacity to predict transverse crack- ing for both HMA and WMA pavements. Figure 3.12 (b) presents the validation of the crack propagation model using the LOOCV method. As shown, most of the validated data are located reasonably close to the line of equality. The R2 value of 0.6 and standard error of the estimate of 0.79 illustrate a relatively good validation result. Transverse Crack Initiation Model For the crack initiation model, a PLS procedure similar to the one used for developing the crack propagation model was applied in conjunction with the BL regression method to develop a probabilistic-based predictive model. Sixty-one responses (including 22 with field cracking and 39 without field cracking) were used for the model regression. The optimum number of variables was determined to be four according to the PRESS plot. The four predictor variables were then selected using PLS regression analysis based on the high- est standardized coefficients (absolute values). These predic- tor variables are IDT strength (+), low-temperature hour (+), percentage passing the #200 sieve (-), and pavement age (+). The signs in the parentheses indicate the trend between the predictor variables and the responses. Using the BL regression method, a probabilistic-based crack initiation model was developed, as shown in Equation (3.2), where P indicates the probability of the initiation of trans- verse cracking, and therefore, 1-P means the probability of no crack initiation. P y e ˆ 1 1 1 (3.2)i X X X X8.42 0.0019 0.284 1.52 0.8671 2 3 4 ( )= = + ( )− − + − + + where Pˆ = probability of initiation of transverse cracking, %; yi = ith pavement section; e = base of natural logarithm, approximately 2.718; X1 = average yearly low-temperature hour, hr (from AASHTOWare Pavement ME Design); X2 = percentage passing the #200 sieve, %; X3 = IDT strength (14°F), MPa; and X4 = pavement age, year. According to the model, within the value ranges used in this study, a high number of hours at low-temperature (<15°F), low percentage passing the #200 sieve, high IDT strength value, and long pavement age could be critical predictor variables for the initiation of transverse cracking. Specifically, overlay mix- tures with high IDT strength values could be related to high levels of stiffness and brittleness and therefore would be prone to transverse cracking. Using the crack initiation model, the probability of crack- ing for each project was estimated; Figure 3.13 (a) shows these results as the solid bars. Of the projects with transverse cracks shown in Figure 3.13 (a), 15 out of 22 pavement projects were predicted to have higher than 50 percent probability for crack initiation. In the without transverse crack group, shown in Figure 3.13 (b), 36 of 39 pavement projects were predicted to have less than 50 percent probability for crack initiation. These findings indicate that the prediction results match the field conditions fairly well. To validate the model, the LOOCV (a) 0 2 4 6 8 0 2 4 6 8 Pr ed ic te d Tr an sv er se C ra ck (L n) , ft /2 00 ft Field-Measured Transverse Crack (Ln), ft/200 ft HMA WMA Line of Equality R2 = 0.82 SEE=0.47 0 2 4 6 8 0 2 4 6 8 V al id at ed T ra ns ve rs e Cr ac k (L n) , ft /2 00 ft Field-Measured Transverse Crack (Ln), ft/200 ft Validated Line of Equality R2 = 0.6 SEE=0.79 (b) Figure 3.12. Crack propagation model development and validation: (a) relationship between predicted and field-measured transverse cracks and (b) validation of transverse crack propagation model (SEE = standard error of the estimate).

30 method was used. Each sample data item was validated once using LOOCV; Figure 3.13 (b) presents the results as the bars with grids. The cross-validated data have almost the same probability range distribution as the predicted data, indicat- ing the model’s good prediction power. At the same time, this indicates that when new validation data are introduced, the model should still work well. Potential Significant Determinants for Transverse Cracking Table 3.2 provides a summary of the identified potential significant determinants from two rounds of paired ranking analysis as well as the statistical models. In the table, “+” means a positive relationship, that is, a higher value will lead to more transverse cracking; “-” means a negative relationship, that is, a higher value will lead to less transverse cracking. Within each category, the properties are listed sequentially based on their relevant influence on transverse cracking. For example, in the ranking method, the property that shows the highest number of pairs in consistent ranking between the field and laboratory results is listed first; in the PLS model method, the property that has the highest standardized coefficient is listed first. As seen, mixture fracture work density at 14°F was identified as a potential significant determinant by both paired ranking and modeling methods. The test method to obtain the frac- ture work density is relatively simple and fast, with no need to mount linear variable differential transducers (LVDTs). It can be performed as a routine test during mix design or mix evalu- Figure 3.13. Validation of transverse crack initiation model: (a) with transverse cracks and (b) without transverse cracks. (a) (b) 0 10 20 30 0-25 25-50 50-75 75-100 3 4 4 11 5 4 2 11 N um be r o f P av em en t S ec ti on s Probability Range, % Predicted Validated 0 10 20 30 0-25 25-50 50-75 75-100 28 8 1 2 26 10 1 2 N um be r o f P av em en t S ec ti on s Probability Range, % Predicted Validated 1st Round Ranking Mix fracture work density, 14°F, (−) Mix E*, 14°F, (+) Mix IDT strength, 68°F, (+) Binder BBR stiffness, 14°F, (+) Mix vertical failure deformation, 68°F, (−) Mix horizontal failure strain, 68°F, (−) 2nd Round Ranking Mix fracture work density, 14°F, (−) Mix E*, 14°F, (+) Binder BBR m-value, 14°F, (−) Mix vertical failure deformation, 68°F, (−) Statistical PLS Model, Crack Propagation Creep compliance, 32°F, (−) Average yearly low-temperature hour, (+) AADTT, (+) Overlay thickness, inch, (−) Percentage passing the #200 sieve, %, (−) Mix fracture work density, 14°F, (−) Statistical PLS Model, Crack Initiation Pavement age, (+) IDT strength, 14°F, (+) Average yearly low-temperature hour, (+) Percentage passing the #200 sieve, (−) Table 3.2. Potential significant determinants for transverse cracking. ation. Other mixture properties such as dynamic modulus at 14°F, mixture IDT strength, mixture vertical failure deforma- tion, and mixture creep compliance could also be potential significant determinants for transverse cracking. Generally, mixture properties have a more direct relationship with the field performance and are preferred over binder properties.