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.

7 Overview This chapter describes the research approach used in this project. The approach consists of three key components, as illustrated in Figure 2.1: (1) selection of the WMA field proj- ects based on predetermined selection criteria, (2) laboratory and field characterization of the WMA pavements and cor- responding control HMA pavements, and (3) data analysis to compare the long-term field performance of WMA and HMA pavements and to identify the significant determinants of WMA pavement performance using statistical methods and predictive models. Selection of WMA Field Projects Selection Criteria Selecting representative and statistically meaningful WMA field projects was of critical importance to this study. Specifi- cally, the following factors were considered for the selection of candidate WMA field projects: 1. Sample size. The sample of candidate WMA field projects had to be large enough to conduct meaningful statistical analysis or modeling, taking into account the many vari- ables that could affect field performance, such as WMA technology, climate, traffic, and pavement structure. The sample of projects had to include these variables in order to realize a thorough evaluation of the long-term performance of WMA and at the same time be practically reasonable. 2. History of performance. Field projects with pavements that showed distress or signs of potential distress were pre- ferred in the selection process. Without distress, it would be impossible to perform the statistical analysis and mod- eling that were needed to identify the significant perfor- mance determinants. 3. Existence of HMA control section. Because WMA was a substitute for HMA, it was essential to compare its perfor- mance with HMA under the same conditions. The exis- tence of a similar HMA control section could eliminate the effects of factors other than the asphaltic materials and, therefore, could be used to identify the engineering prop- erties of WMA that are significant performance determi- nants. The HMA and WMA pavements were either on the same traffic lane or parallel traffic lanes depending on the availability of HMA pavements. 4. Accelerated pavement testing. Compared with conven- tional pavements, the pavements used in accelerated pave- ment tests of WMA were loaded to cause accelerated deterioration within a relatively short period and to generate distresses that otherwise would need long-term highway traffic loading. They were included in the study; however, the limited aging of these mixtures was accounted for in the data analysis. 5. New pavement projects constructed in 2011 and 2012. The inclusion of the “new” pavement projects provided several advantages to this study: (a) their field performance could be monitored closely from the time of construction, (b) the field cores of the 2011 and 2012 pavements were especially useful for identifying the determinant proper- ties of pavement performance, and (c) because detailed background information about in-service pavements was not available in many cases, the selection of new projects could compensate for this limitation. 6. Other NCHRP studies. This study incorporated several field projects in common with other NCHRP studies. Close coordination with the researchers of other NCHRP proj- ects and sharing results, such as field measurements and laboratory test results, not only reduced expenses associ- ated with duplication, but also provided additional long- term performance data from these projects. The availability of HMA control sections was the most important criterion. All the WMA pavements included in this study had corresponding HMA control pavements that were C H A P T E R 2 Research Approach

8either at the same project location as the WMA pavements or were constructed nearby to provide similar traffic, materials, and structure conditions. Selected Field Projects Two types of field projects were incorporated into this study: five new projects that were constructed in 2011 and 2012, and 23 in-service field projects that were constructed before 2011. Figure 2.2 shows the distribution of all the selected projects. Each project contains at least one WMA technology type and a corresponding HMA control pave- ment, resulting in one or more HMA-WMA pairs for com- parative purposes. The five new WMA pavement projects are located in five states: Montana, Tennessee, Iowa, Texas, and Louisiana and are referred to as the MT I-15, TN SR 125, IA US 34, TX FM 973, and LA US 61 projects, respectively. Their key design and project information is presented in Table 2.1. The mix design information of the five projects is listed in Appendix A. Other construction information includes the following: 1. Pre-construction information: mixture design, WMA technology type, existing pavement structure, existing pavement conditions, target mixing and compaction tem- peratures, mile post or GPS information for the selected three 200-ft research test sections, and so forth. 2. Construction information: plant modification, weather, material information, aggregate moisture content, mix- ing and compaction temperatures, haul distance, in-place density, procurement of on-site gyratory-compacted samples, loose mixtures and raw materials, and any other important construction information that should be noted. 3. Post-construction information: quality control/quality assurance (QC/QA) data, procurement of cores, falling weight deflectometer testing, and location of field cores. The in-service projects had been in service for 4 to 10 years when the second-round distress survey was conducted in 2015. They included one HVS project and 22 other in-service projects. Historical construction data, such as construction year, climate, traffic, job mix formula, pre-construction infor- mation, project plans, QC/QA data, and previous pavement performance data were collected for all in-service projects to the extent possible. Tables 2.2 (a) through (d) present the essential design and project information for the 22 in-service projects and the 1 HVS project (CA 3a and 3b projects), as categorized by specific climatic zones (wet freeze, wet no- freeze, dry freeze and dry no-freeze). The mix design infor- mation of these projects is provided in Appendix A. Figures 2.3 (a) through (g) show the distribution of all the projects in terms of WMA type, climate zone, service year, pavement type, traffic level, anti-stripping type usage, and level of reclaimed asphalt pavement (RAP) used, respectively. As shown, these WMA projects covered three main WMA technology categories and four climate zones. They had been in service for at least 4 years and as many as 10 years when the second-round distress survey was conducted in 2014–2015. Two existing pavement structure types—flexible pavements and PCC/asphalt pavements with a cement-stabilized base— were included. One-third of the projects were interstate high- ways that were subjected to high volume traffic; two-thirds of the projects were state routes. An anti-stripping agent was used for one-half of the projects, and RAP was used for more than one-half of the projects. Conclusions & Proposed Actions Figure 2.1. Flow chart of the research approach. Figure 2.2. Locations of selected field projects.

9 Project MT I-15 TN SR 125 IA US 34 TX FM 973 LA US 61 Construction Year 2011 2011 2011 2011 2012 Warm Mix Type Sasobit, Evotherm DAT, Foaming Evotherm 3G Sasobit, Evotherm 3G Evotherm 3G, Foaming Sasobit, Evotherm 3G Mixing Temp., °F H (315-320) W (283-300) H (320-350) W (290-320) H (330-340) W (265-280) H (310-340) W (249-290) H (325) W (295) Compaction Temp., °F H (290-300) W (269-285) H (311-335) W (275-290) H (253-265) W (220-239) H (268-291) W (221-268) H (277-316) W (245-250) Design Thickness, in. 2.5 1.25 1.5 2.0 2.0 Traffic 3 million ESALs (3,170 AADT, 26.3% truck) 0.39 million ESALs (3,470 AADT, 13% truck) 3 million ESALs (6,450 AADT, 10.9% truck) 3 million ESALs (11,300 AADT, 4.3% truck) 9 million ESALs (34,138 ADT, 14% truck) Aggregate Siliceous Gravel & Sand Limestone, Quartzite & Sand Gravel, Limestone & Dolomite Granite & Limestone NMAS, in. 3/4 1/2 1/2 3/4 1/2 Asphalt Binder PG 70-28 PG 70-22 PG 58-28 PG 70-22 PG 76-22 Anti-stripping Agent Hydrated Lime, 1.4% ARR-MAZ, 0.3% None None 0.6% Polymer- modified SBS Yes None N/A SBS Asphalt Content, % 4.6 6.0 5.44 5.2 4.7 Maximum Specific Gravity (Gmm) HMA (2.458) Sasobit (2.466) Evotherm (2.459) Foaming (2.453) HMA (2.352) Evotherm (2.355) HMA (2.423) Sasobit (2.428) Evotherm (2.429) HMA (2.406) Evotherm (2.405) Foaming (2.420) HMA (2.464) Sasobit (2.468) Evotherm (2.464) Sampling Date Sep. 15-17, 2011 Oct. 24-31, 2011 Sep. 6, 2011 Dec. 1, 2011 May 16-June 6, 2012 RAP or RAS None 10% RAP 17% RAP None 15% RAP Structure 2.5'' overlay + 7'' existing HMA + 16.2'' base (non- stabilized) + infinite subgrade 1.25'' overlay + 8'' bituminous base + 6'' min. aggregate base + infinite subgrade HMA & Sasobit: 1.5'' overlay + 5'' existing HMA + 7'' PCC + subgrade Evotherm: 1.5'' overlay + 3'' existing HMA + 9'' PCC + subgrade 2'' overlay + 8'' existing HMA + 10'' base + 141.1'' subgrade (lean clay) 2'' overlay + 8'' existing HMA + 8'' PCC + 6'' cement- treated soil subgrade Notes: 1. MT I-15: The existing HMA thicknesses vary: HMA (7”); Sasobit (12.5”); Evotherm (5”); Foaming (4.4”). 2. N/A: not available. 3. H: HMA, W: WMA. Table 2.1. Summary of new projects. Project MD 925 MO Hall St. MO Rte. CC MN TH 169 OH SR 541 PA SR 2012 VA I-66 IL 147 PA SR 2006 Construction Month/Year 9/2005 5/2006 2007 7/2010 9/2006 5-6/2009 7/2010 6/2010 5/2009 Warm Mix Type (content, %, by weight of binder) Sasobit Sasobit (1.5); Evotherm ET; Aspha- min (0.3) Evotherm DAT Evotherm 3G Sasobit (1.5); Evotherm ET (5.3 by mix); Aspha-min (0.3) LEA, Gencor (0.5) Astec DBG Astec DBG Sasobit, Advera Production Temp.,°F HMA (310- 350); Sasobit (270- 310) HMA (320); Sasobit (240); Evotherm (225); Aspha- min (275) HMA (320); Evotherm (280-290) HMA (300); Evotherm (265) HMA (320); Sasobit (260); Evotherm (235); Aspha-min (245) HMA (290- 310); Gencor (250- 265); LEA (240-260) HMA (317); Astec DBG (288) HMA (300); Astec DBG (270) HMA (310); Sasobit (265); Advera (260) Traffic (AADT) 10,480 21,000 8,618 12,600 650 254 57,000 775 523 Aggregate N/A Limestone Steel Slag Limestone N/A Limestone Limestone N/A Limestone Limestone (a) Wet Freeze Zone Table 2.2. Summary of in-service pavement projects. (continued on next page)

10 Project SC US 178 TN SR 46 TX FM 324 LA 116 LA 3191 LA 3121 Construction Month/Year 9/2007 10/2007 2-3/2008 3/2010 11/2008 3/2009 Warm Mix Type Evotherm DAT Sasobit, Evotherm DAT, Astec DBG, and Advera Sasobit, Evotherm DAT, Rediset, Advera Foam Astec Foam Evotherm 3G Production Temp., °F HMA (295); Evotherm (240) HMA Danley (320-350); HMA Franklin (320- 350); Sasobit (250); Evotherm DAT (240); Advera (250); DBG (260) H (330); W (240) N/A N/A N/A Traffic (AADT) 3,880 4,440 1,450 2,600 ADT 200 ADT 400 Aggregate N/A Limestone Limestone N/A N/A N/A NMAS, in. 3/8 1/2 3/8 1/2 1/2 1/2 Asphalt Binder 64-22 70-22 64-22 70-22 70-22 70-22 Anti- stripping Agent N/A Franklin (AD-Here 77- 00, 0.3%); Astec DBG (Pavegrip 650, 0.3%) 1% Lime N/A N/A N/A Asphalt Content, % H (5.3); W (5.4) 5.3 4.6 4.4 5.2 5.1 Gmm H (2.460) W (2.463) HMA Danley (2.428); Sasobit (2.411); Evotherm (2.410); Astec DBG (2.444); Advera (2.422); HMA Franklin (2.425) HMA, Sasobit, Evotherm (2.508); Advera Rediset (2.498) H (2.525) W (2.541) H (2.453) W (2.486) H (2.507) W (2.490) Sampling Date 07/27/12 07/24/12 2/6-8/13 05/21/13 05/21/13 05/20/13 RAP N/A None None 15 15 15, 30 Structure 2" overlay + 5.7" HMA + 7.1" PCC + sand clay base 1.25" overlay + 4.26" HMA + 6" crushed stone 1.5" overlay + 5.7" HMA + 10" base 1.5" overlay + 5" HMA + 8.5" base 2" HMA + 6" HMA + 7" PCC 2" overlay + 12" cement- treated base (b) Wet No-Freeze Zone Project MD 925 MO Hall St. MO Rte. CC MN TH 169 OH SR 541 PA SR 2012 VA I-66 IL 147 PA SR 2006 NMAS, in. 3/8 1/2 1/2 3/4 3/8 3/8 1/2 1/2 3/8 Asphalt Binder 64-22 70-22 64-22 58-28 70-22 64-22 76-22 64-22 64-22 Anti- stripping Agent None ARR MAZ, 0.25% Pave Bond Lite, 0.25% None None None Pave Bond Lite, 0.5% None None Asphalt Content, % 5.0 5.3 5.4 4.2 6.1 5.9 H (5.0); W (5.4) 5.0 6.0 Gmm 2.519 2.451 2.469 2.549 2.429 2.476 H (2.62); W (2.605) H (2.450); W (2.47) H (2.467); Sasobit (2.462); Advera (2.469) Sampling Date 6/28/12 7/16- 18/12 7/17/12 8/28/12 6/18/12 6/26/12 6/26/12 7/19/12 6/21/12 RAP 15% 10% 20% N/A 15% None N/A 10% None Structure 2" + 5" HMA+ 8" Macadam stone 1.75" + 12" PCC + 0-3" base 3.75" + 7" PCC + 6" base 2" + 8" HMA+ 6" base 1.25" + 6.75" HMA + 9" granular base 1.5" + 5" HMA+ 4" aggregate base 1.5" + 5" HMA+ 9" + 10" base 1.5" + 9" HMA 1.5" + 5" HMA+ 4" stone base (a) Wet Freeze Zone Table 2.2. (Continued).

11 Project TX SH 251 TX SH 71 CA HVS 3a CA HVS 3b Construction Month/ Year 8/2008 6/2008 9/2009 9/2009 Warm Mix Type Astec DBG Evotherm DAT Gencor, Evotherm DAT, Cecabase Sasobit, Advera, Astec DBG, Rediset Production Temp., °F H (310)W (270) H (330) W (240) HMA (320), Gencor (284), Evotherm (248), Cecabase (266) HMA (335, 279), Sasobit (300,279), Advera (295,266), Astec DBG (295,257), Rediset (285,258) Traffic (AADT) 2,300 57,000 HMA (74,000), Gencor (159,000), Evotherm and Cecabase (160,000) HMA, Sasobit, Astec DBG and Rediset (160,000), Advera (50,000) Aggregate Limestone Limestone Granite Reed NMAS, in. 3/8 3/8 1/2 1/2 Asphalt Binder 70-22 76-22 64-16 64-16 Anti-stripping Agent 1% Akzo 0.8% Liquid None None Asphalt Content, % 5.1 4.8 7.0 8.3 Gmm H (2.45), W (2.4) 2.416 H (2.503) H (2.505) Sampling Date 2/5/13 2013 2012 2012 RAP, % None N/A Rubber (18% of binder) Rubber (18% of binder) Structure 2.0" overlay +4.3" HMA 2" overlay + HMA 2.5" gap-graded rubberized HMA + 2.5" HMA + 15.6" base 2.5" gap-graded rubberized HMA + 2.5" HMA + 15.6" base Notes: 1. N/A indicates that the information is not available and will be obtained. 2. H: HMA; W: WMA. 3. The traffic for the CA HVS project is based on equivalent standard axle load (ESAL). 4. The pavement structure for the VA I-66 project for WMA is different from that for HMA. WMA: 1.5" overlay + 3" HMA + 13" existing + 10" base. (d) Dry No-Freeze Zone Project WA I-90 WA SR 12 CO I-70 NE US 14 NV Construction Month/Year 6/2008 4/2010 7-8/2007 2008 8/2010 Warm Mix Type Sasobit Aqua Black Sasobit (1.5% by mass of binder); Evo therm DAT (0.5% of binder); Advera (0.3% of mixture) Advera, Evotherm DAT Ultrafoam Production Temp., °F HMA (330), Sasobit (276) HMA (325), Aqua Black (275) HMA (mixing 310, compaction 280); Sasobit (255, 235); Evotherm (250, 230); Advera (255, 235) H (330), W (275) H (330), W (275) Traffic (AADT) 13,000 6,550 30,000 2,140 5,000 Aggregate Basalt Basalt Crushed River Rock Limestone, Gravel N/A NMAS, in. 1/2 1/2 1/2 1/2 1/2 Asphalt Binder 76-28 64-28 58-28 64-28 64-28 Anti-stripping Agent None Superbond (0.25%) Hydrated Lime (1% by mass of aggregate blend) None Hydrated Lime, 1.5% Asphalt Content, % 5.5 5.2 6.3 5.0 4.6 Gmm 2.601 2.596 2.45 H-Adv (2.439), H-Evo (2.441) 2.451 Sampling Date 8/27/12 8/28/12 10/18/12 10/14/12 10/19/12 RAP, % 15-20 20 None <15 15 Structure 3" overlay + 11.28" HMA + 6.5" base (HMA)/5" base (Sasobit) 3" overlay + 7.8" HMA + 9" base 2.5" overlay + 10-11" HMA 3" overlay + 4" HL slurry stabilization + 1.5" existing asphalt + 4" bit sand base 6" HMA + 9" aggregate base (c) Dry-Freeze Zone Table 2.2. (Continued).

12 Notes: “N/A” denotes that information is not available. DF: dry freeze; WF: wet freeze; DNF: dry no-freeze; WNF: wet no-freeze. (a) 12 17 18 0 5 10 15 20 N o. o f W M A S ec  on s Organic Chemical Foaming WMA Type (b) 6 10 4 8 0 2 4 6 8 10 12 N o. o f P ro je ct s DF WF DNF WNF Climate Zone (c) 11 10 7 0 2 4 6 8 10 12 N o. o f P ro je ct s 4 to 5 5 to 7 7 to 10 Pavement Aging Year (d) 20 8 0 5 10 15 20 25 N o. o f P ro je ct s Flexible PCC/Cement Stabilized Base Pavement Type (e) 18 10 0 5 10 15 20 N o. o f P ro je ct s <3 million >= 3million Traffic (ESAL) (f) 12 12 4 0 5 10 15 N o. o f P ro je ct s Yes No N/A Use of An-stripping (g) 14 10 4 0 5 10 15 N o. o f P ro je ct s Yes No N/A Use of RAP Figure 2.3. Distribution of selected WMA projects in terms of (a) WMA type, (b) climate zone, (c) service year, (d) pavement type, (e) traffic, (f) use of anti-stripping agent, and (g) use of RAP.

13 Field and Laboratory Characterization of WMA Projects Field Distress Surveys Two rounds of distress surveys were conducted in 2012– 2013 and 2014–2015, for all the in-service projects except for the CA HVS and TX SH 71 projects. The second-round distress survey was also conducted for the newly constructed projects except for the TX FM 973 project due to logistical reasons. Three 200-ft test sections of HMA and WMA pave- ments were selected to measure distresses. The first- and second-round distress surveys were conducted at the same 200-ft test sections. The distresses measured include rutting and cracking in accordance with the Long-Term Pavement Performance (LTPP) Distress Survey Manual (Miller and Bellinger 2003). In addition, thirteen 4-inch-diameter field cores and four 6-inch-diameter cores were acquired for each HMA and WMA pavement test section in the first-round fieldwork for the laboratory experiments. Cores were also extracted during the second-round fieldwork for the new projects. These cores were taken from the non-wheel path to limit the damage to the material. Additional cores were extracted at the fine tip of a wheel-path longitudinal crack or a transverse crack, if needed, to determine whether the crack initiated from the surface or the bottom of the pavement. Transverse Cracking Identification Transverse crack length and severity were manually mea- sured within the three 200-ft test sections of HMA and WMA pavements in both inside and outside wheel paths. In most cases in this study, transverse cracks in the existing asphalt layer overlapped the transverse cracks in the asphalt overlay and initiated from the top of the pavement, as shown in Fig- ure 2.4 (a). In the case of no existing cracks in the existing asphalt layer, as shown in Figure 2.4 (b), the transverse crack also initiated from the top. Both thermal cracking and reflec- tive cracking can occur in the form of transverse cracking in an asphalt overlay. The fact that transverse cracks in the HMA and WMA layers often overlapped with transverse cracks in the existing pavement indicates that the transverse cracks could be a combination of thermal cracking and reflective cracking. Wheel-Path Longitudinal Cracking Identification This study focused on fatigue-related wheel-path longi- tudinal cracking. Crack length and severity were measured within the three 200-ft test sections of HMA and WMA pave- ments in both inside and outside wheel paths. Figure 2.5 (a) shows an example of a longitudinal crack, and Figure 2.5 (b) shows that the crack initiated from the top and stopped within the asphalt layer. (b) (a) Underlying UnderlyingOverlay Overlay AdveraEvotherm Figure 2.4. Field cores taken at the tip of transverse cracks: (a) overlay cracking with transverse cracking in existing asphalt layer and (b) overlay cracking only without transverse cracking in existing asphalt layer.

14 Rut Depth Identification Rut depth was measured within the three 200-ft test sec- tions of the HMA and WMA pavements every 50 ft for both the inside and outside wheel paths. Figure 2.6 illustrates the rut depth measurement process performed in the field. Laboratory Testing Tables 2.3 and 2.4 provide a summary of all the labora- tory mixture and binder tests performed during this study. The material properties obtained from these laboratory tests were used in the determination of significant determinants for field performance. Axial Dynamic Modulus Test In this study, two types of dynamic modulus tests were con- ducted: (1) an axial dynamic modulus test using an Asphalt Mixture Performance Tester (AMPT) and (2) an indirect tensile dynamic modulus test using MTS. The axial dynamic modulus tests were conducted accord- ing to AASHTO T 342 Standard Method of Test for Determin- ing Dynamic Modulus of Hot Mix Asphalt. The test applies a uniaxial sinusoidal (i.e., haversine) compressive stress to an unconfined PMLC cylindrical specimen (5.9 inches in height and 3.9 inches in diameter) to achieve a target vertical strain level of approximately 100 microns in an unconfined test mode. Figure 2.7 (a) shows the set-up for the axial dynamic Wheel Path Longitudinal Crack (b) (a) Figure 2.5. Longitudinal cracking: (a) longitudinal crack found in the wheel-path and (b) field core taken from tip of a wheel-path longitudinal crack. Figure 2.6. Rut depth measurement device in the field (left) and taking measurements in the field (right).

15 modulus test. Specimens were tested at temperatures of -10°C, 4.4°C, 20°C, 37.8°C, and 54.4°C and at loading fre- quencies of 20 Hz, 10 Hz, 5 Hz, 1 Hz, 0.5 Hz, and 0.1 Hz at each temperature for the development of master curves and for use in the statistical analysis. Specimens were compacted to the same air void levels (7 ± 0.5%). Triplicates were tested for each specimen type. The dynamic modulus is mathematically defined as the maximum (i.e., peak) dynamic stress (s0) divided by the peak recoverable axial strain (e0): (2.1) 0 0 ∗ = σ ε E IDT Dynamic Modulus Test The IDT dynamic modulus tests were conducted using PMFC cores due to the geometry constraint for the axial test discussed previously. The testing method followed the pro- cedure implemented by Wen and Kim (2002). Figure 2.7 (b) shows the set-up for the IDT dynamic modulus test. The size Testing IDT DynamicModulus IDT Creep Compliance IDT Fracture at Intermediate Temp. IDT Fracture at Low Temp. Hamburg Wheel Tracking AMPT E* and Flow Number (for new projects only) Testing Conditions Temp. (°F): −4, 14, 32, 50, 68, 86 Frequency (Hz): 20, 10, 5, 1, 0.1, 0.01 Temp. (°F): − 4, 14, 32, 50, 68, 86 Duration: 100 seconds Temp. (°F): 68 Loading rate: 2 in./min Temp. (°F): 14 Loading rate: 0.1 in./min 122°F Temp. (°F): 40, 70, 100, 130 Frequency (Hz): 25, 20, 10, 5, 2, 1, 0.5, 0.2, 0.1 Flow number: pavement high temp. at 600 kPa Material Properties Dynamic modulus Creep compliance î IDT strength; î Fracture work density; î Vertical failure deformation; î Horizontal failure strain î IDT strength; î Fracture work density; î Vertical failure deformation; î Horizontal failure strain î Rut depth; î Stripping inflection point (SIP); î Cycles î Dynamic modulus; î Flow number References/ Standards Wen and Kim 2002 Wen and Kim 2002, AASHTO T 322 Wen 2012 Wen 2012 AASHTO T 324 AASHTO T 342 Notes: 1. IDT: indirect tensile; AMPT: Asphalt Mixture Performance Tester. 2. Flow number test temperature was determined based on the high pavement temperature of specific project locations obtained from LTPPBind Version 3.1 software at 50% reliability and 98% reliability. 3. AMPT dynamic modulus and flow number tests were conducted only for new projects using plant-mixed laboratory- compacted specimens. Table 2.3. Summary of laboratory testing—mixtures. Testing Performance Grade (PG) Multiple Stress Creep Recovery (MSCR) Monotonic at Intermediate Temp. Monotonic at Low Temp. Testing Conditions Different temperatures depending on the test (DSR, BBR) Temp.: Pavement high temperature Stress: 0.1, 3.2 kPa Temp.: 68°F Shear strain rate: 0.3 s-1 Temp.: 41°F Shear strain rate: 0.075 s-1 Material Properties PG high temp.; PG low temp.; BBR stiffness; m-value Nonrecoverable creep compliance (Jnr); Percentage of recovery (R) Maximum stress; Fracture energy; Failure strain Maximum stress; Fracture energy; Failure strain References/ Standards AASHTO MP 1/T 313/T 315 AASHTO T 350 Wen et al. 2010 Wen 2012 Notes: 1. DSR: dynamic shear rheometer, BBR: bending beam rheometer. 2. MSCR test temperature was determined based on the high pavement temperature of specific project locations obtained from LTPPBind Version 3.1 software at 50% reliability and 98% reliability. Table 2.4. Summary of laboratory testing—binders.

16 of the specimens was 1.5 inches in thickness and 3.9 inches in diameter. The volumetric properties of the samples were recorded before conducting the tests. Triplicates were tested for each specimen type. The IDT dynamic modulus test applies a sinusoidal com- pressive stress to the diametric axis of an unconfined PMFC cylindrical test specimen to achieve target strain levels (40–60 horizontal microstrain and <100 vertical microstrain) in the linear viscoelastic region. The tests were conducted at tem- peratures of -4°F, 14°F, 32°F, 50°F, 68°F, and 86°F and at loading frequencies of 20 Hz, 10 Hz, 5 Hz, 1 Hz, 0.1 Hz, and 0.01 Hz at each temperature for the development of master curves. Equation (2.2) presents the mathematical relation- ship between load and deformation in the IDT-loading mode: 2 (2.2) 0 1 2 2 1 2 0 2 0 ∗ = pi β γ − β γ γ − βE P ad V U where P0 = peak-to-peak load, N; a = loading strip width, m; d = thickness of specimen, m; V0 = peak-to-peak vertical deformation, m; U0 = peak-to-peak horizontal deformation, m; and g1, g2, b1, and b2 = geometric constants. IDT Creep Compliance Test The IDT creep compliance tests were conducted in accor- dance with AASHTO T 322 Standard Method of Test for Deter- mining the Creep Compliance and Strength of Hot Mix Asphalt Using the Indirect Tensile Test Device. A constant load was applied for 100 seconds to measure the horizontal deforma- tion of the asphalt mixture at different temperatures (-4°F, 14°F, 32°F, 50°F, 68°F, and 86°F). The determination of creep compliance is shown in Equation (2.3). (2.3)1 2[ ]( ) ( )( ) = − β + βD t d P U t V t where D(t) = creep compliance, 1/Pa; P = applied load, N; d = thickness of specimen, mm; V(t) = vertical deformation, mm; U(t) = horizontal deformation, mm; and b1 and b2 = coefficients that can be found in the paper by Wen and Kim (2002). Flow Number Tests Flow number tests were conducted only for the new pave- ment projects using PMLC specimens to evaluate the rutting (a) Axial Dynamic Modulus Test Set-Up (b) IDT Dynamic Modulus Test Set-Up Figure 2.7. Dynamic modulus test set-ups: (a) axial and (b) IDT.

17 resistance of the asphalt mixtures. Samples from the TN SR 125 project were compacted in a mobile asphalt laboratory at the plant during construction without reheating. Because of limited resources on site, samples from other projects were compacted at the Washington State University (WSU) asphalt pavement laboratory with reheating. These tests were performed according to AASHTO TP 79 Standard Method of Test for Determining the Dynamic Modulus and Flow Number for Asphalt Mixtures Using the Asphalt Mixture Performance Tester (AMPT). The test temperature was determined based on a 7-day maximum pavement temperature 0.78 inches below the surface using LTPPBind Version 3.1 software at 50 percent reliability. IDT Fracture Tests IDT fracture tests at intermediate (20°C) and low (-10°C) temperatures were used to characterize cracking resistance of mixes in this study. Based on past research (Wen and Kim 2002, Wen 2013), the fracture property obtained from an IDT test at room temperature was considered a promising engi- neering property that correlated well with the field fatigue performance of WesTrack pavements. Other researchers also used fracture properties obtained from the IDT fracture test at low temperatures to characterize the thermal cracking resistance of mixes (Das et al. 2013, Solanki et al. 2015, and Wu et al. 2015). Both the fatigue and thermal cracking prop- erties of the asphalt mixtures can be evaluated by the IDT monotonic fracture tests using a similar laboratory set-up. IDT fracture tests can not only provide IDT strength data, which indicate the maximum load that asphalt mixtures can withstand, but also give other fracture properties that can be used to characterize cracking resistance. Fracture work den- sity is defined as the fracture work divided by the specimen volume; fracture work is the area beneath the load versus vertical displacement curve, as shown in Figure 2.8. Vertical failure deformation is the critical deformation at the peak load. Horizontal failure strain is the horizontal strain at the peak stress in the stress-strain curve. Higher vertical failure deformation and horizontal failure strain indicate the mixes’ better ability to stretch under extreme loading. In this project, IDT fracture tests were conducted on the samples 1.5 inches thick and 3.9 inches in diameter. Appen- dix D provides the suggested procedures for the mixture IDT fracture cracking testing in AASHTO standard format. Hamburg Wheel Tracking Tests The HWT test has been widely used to characterize the rutting and moisture susceptibility of asphalt mixtures, and it can also indicate rutting in the field (Yildirim and Kennedy 2001, Yildirim and Stokoe II 2006). In addition, the HWT tests are more suitable for field cores than the flow number tests for evaluating rutting. In this project, HWT tests were performed at the Louisi- ana Transportation Research Center asphalt laboratory to evaluate rutting resistance and moisture susceptibility of the asphalt mixtures. The tests were performed according to AASHTO T 324 Standard Method of Test for Hamburg Wheel-Track Testing of Compacted Hot Mix Asphalt. The tests were conducted at 122°F under wet conditions. For the PMLC specimens, the air void contents were controlled at 7.0 ±0.5 percent. The air void content of each specimen was recorded for the PMFC cores. HWT tests are typically stopped when a 0.5-inch rut depth is reached or at the end of 20,000 cycles, whichever comes first, making it difficult to compare the rutting resistance using either the number of cycles at the completion of the test or the rut depth at the end of the test. Therefore, a single rutting resistance index (RRI) was developed in this study to quantify the rutting resistance. The RRI considers both rut depth and number of cycles, as shown in Equation (2.4): RRI N 1 RD (2.4)( )= × − where RRI = rutting resistance index; N = number of cycles at completion of the test; and RD = rut depth at completion of the test, inch. The RD is typically less than 1 inch. A high RRI value indicates high rutting resistance. For instance, the following three scenarios demonstrate the use of the RRI: 1. Good rutting resistance: 0.1-inch rut depth at 20,000 cycles, RRI = 18,000. 2. Average rutting resistance: 0.5-inch rut depth at 20,000 cycles, RRI = 10,000. 3. Poor rutting resistance: 0.5-inch rut depth at 10,000 cycles, RRI = 5,000. Figure 2.8. Load versus vertical displacement curve.

18 Binder Extraction and Recovery Asphalt binders were extracted and recovered from the field cores. The extraction process was performed based on AASHTO T 164 Standard Method of Test for Quantitative Extraction of Asphalt Binder from Hot Mix Asphalt and the recovery was conducted according to AASHTO R 59 Stan- dard Practice for Recovery of Asphalt Binder from Solution by Abson Method. The chemical used for extraction was a com- bination of 85 percent toluene and 15 percent ethanol by volume. Both WMA and HMA field cores were put into a conventional oven at 230°F until they were loose enough to separate. The separated mixtures were cooled down to room temperature before binder extraction. The minimum mass of the samples used for binder extraction was determined by nominal maximum aggregate size (NMAS). Usually, several extractions were needed until the extracted material was no darker than a light straw color. Binder Performance Grading Binder performance grading was performed according to AASHTO M320 Standard Specification for Performance- Graded Asphalt Binder. The continuous performance grade (PG) was calculated based on test results at both high and low temperatures. The recovered binder was treated as short-term aged binder, and the high temperature PG was determined in accordance with AASHTO T 315 Standard Method of Test for Determining the Rheological Properties of Asphalt Binder Using a Dynamic Shear Rheometer. The m-value and stiffness value used for determining the low temperature PG were deter- mined in accordance with AASHTO T 313 Standard Method of Test for Determining the Flexural Creep Stiffness of Asphalt Binder Using the Bending Beam Rheometer. Binder Multiple Stress Creep Recovery Tests The multiple stress creep recovery (MSCR) tests of rolling thin film oven-aged (RTFO-aged) original binder and recov- ered binder were performed in accordance with AASHTO T 350 Standard Method of Test for Multiple Stress Creep Recov- ery Test of Asphalt Binder Using a Dynamic Shear Rheometer. This method is designed to evaluate the ability of a modified binder to maintain its elastic response at two different stress levels while being subjected to 10 cycles of stress and recovery. The RTFO-aged binders were prepared using original binder collected at a field asphalt plant in accordance with AASHTO T 240 Standard Method of Test for Effect of Heat and Air on a Moving Film of Asphalt Binder as well as binders extracted and recovered from field cores. A 25-mm parallel plate geometry was used with a 1-mm gap setting. The samples were tested at creep stress levels of 100 Pa and 3,200 Pa. The creep por- tion of the test lasted for 1 second followed by a 9-second recovery. The first stress level was repeated for 10 cycles and then increased to the next level. All the binder tests were per- formed at a high temperature that was determined based on the environmental high pavement temperature of specific project locations as determined from LTPPBind Version 3.1 software at 98 percent reliability. Binder Fatigue and Thermal Cracking Tests Monotonic tests at 68°F and 41°F were conducted to char- acterize the damage properties of the asphalt binders. Appen- dix D provides details of the monotonic tests with a dynamic shear rheometer (DSR). Figure 2.9 presents a typical mono- tonic test result where the shear strength is the maximum shear stress the asphalt can withstand, and the failure strain is the shear strain at the peak stress. Shear failure strain has been reported to correlate with top-down fatigue cracking (Wen and Bhusal 2013). Fracture energy is defined as the area underneath the shear stress versus shear strain curve up to the peak stress. Fracture energy at intermediate temperatures has been reported to correlate with field bottom-up fatigue cracking of asphalt pavements (Johnson et al. 2009). Data Analysis WMA and HMA Field Performance Comparisons Field performance comparisons between each WMA pave- ment and its corresponding HMA control pavement were conducted in terms of transverse crack length, wheel-path longitudinal crack length, and rut depth. The crack lengths of the WMA and HMA pavements were compared based on statistical t-test results at a significance level of 0.05. The null hypothesis was that there was no significant difference in Figure 2.9. Binder shear stress versus strain curve obtained from monotonic test.

19 crack length between HMA and WMA. The rut depths were compared based on a practical criterion, a 1⁄16-inch threshold value, because the manual measurement of a rut depth had a precision of 1⁄16 inch. Determination of Significant Determinants Due to the confounding factors in the field and the dif- ficulty in clearly differentiating cracking mechanisms and types (surface-initiated or bottom-initiated), statistical based approaches were deemed more appropriate to identify the sig- nificant determinants of field performance. In this project, two statistical based approaches were used jointly: (1) using paired ranking and (2) using partial least square (PLS) statistical modeling. The material property selected by both methods is considered to be the most promising significant determinant. Other selected properties are suggested as potential significant determinants that can be evaluated further. The paired ranking method assumes the WMA and HMA can be analyzed in pairs because they share similar traffic level, pavement structure, climate, and construction condi- tions. The analysis follows three steps: 1. Compare a variety of material properties obtained from laboratory tests for each of the HMA-WMA pairs and ascertain a material property ranking (HMA > WMA, HMA = WMA, or HMA < WMA), based on statistical t-test results and a significance level of 0.05. 2. Compare the specific field performance result (e.g., the transverse crack length) for each of the HMA-WMA pairs and obtain the field performance ranking (HMA > WMA, HMA = WMA, or HMA < WMA), based on statistical t-test results and a significance level of 0.05. 3. Determine the number of HMA-WMA pairs that exhibit consistent trends/rankings between the material proper- ties and the field performance. The properties that provided the highest number of pairs (highest level of consistency with field performance) can be considered as potential significant determinants for field per- formance. The paired ranking method detects the material properties that can provide the most consistent trends with field performance for the 27 field projects (TX FM 973 was not included). Although practically powerful, one specific limitation is that this approach does not integrate the effect of other parameters such as pavement structure and climate; it exclusively compares the effects of material properties. The statistical PLS method can assess the critical influenc- ing factors on pavement performance by constructing a sta- tistical predictive model. The identified influencing factors can consist of both material parameters and non-material parameters (such as pavement structure and climate). In addition, compared with the conventional multiple linear regression method, the PLS method can solve collinearity issues (exclude highly correlated factors) and at the same time identify significant material property indicators by devel- oping a predictive model. The PLS method only selects the parameters that contribute strongly to the prediction model, and is suitable for a relatively small dataset. This makes it practically useful for pavement performance predictions because obtaining large amounts of field performance data and laboratory data is usually time-consuming and costly. The method of PLS modeling for pavement performance prediction is described in the next section. Performance Predictive Models Pavement ME Design Analysis and Performance Predictions The ability to predict field performance with respect to vari- ous distresses is a main concern with regard to asphalt pave- ments that include both HMA and WMA. The AASHTOWare Pavement ME Design program is widely used to predict the performance of pavements. However, the reliability of the pro- gram and its distress models, especially for WMA pavements, concerns users. The comprehensive database of material prop- erties and the use of long-term distress surveys in this study provide a valuable opportunity to evaluate the effectiveness of the pavement performance models within the AASHTOWare Pavement ME Design analysis. In this project, nationally calibrated factors in the Pavement ME Design program were used to predict the performance of asphalt pavements. Inputs for this analysis were categorized into two types: Level 1 material properties were used for over- lays and Level 2 and Level 3 material properties were used for layers underneath overlays. Results from this analysis were compared with the field distress survey results, and thus, the effectiveness of the long-term performance predictive models within the Pavement ME Design software was tested for HMA and WMA pavements. Statistical PLS Method and BL Regression Method In this project, the statistical partial least square (PLS) method (Yeniay and Göktas 2002) was used to develop field cracking and rutting predictive models for both WMA and HMA pavements. Multiple linear regression (MLR) was not chosen because it could result in an over-fitting model that is unable to predict new responses well (Hawkins 2004). Com- pared with MLR, the PLS method can better solve collinear- ity problems and ensure that only parameters that contribute greatly to the prediction model are selected. Here, collinearity

20 refers to a high level of correlation between two predictor variables that often exist in a complex field situation and must be avoided in the development of predictive models. Other advantages of the PLS method include simplicity, reliability, versatility, and suitability for relatively small datasets (Peng and Lai 2012). By using the PLS method in conjunction with leave-one-out-cross-validation (LOOCV) and the prediction sum of squares (PRESS), the model simultaneously decides the optimum variable number of independents, determines the coefficients of the model parameters, and validates the model’s capability to predict new performance. It is noted that the initiation and propagation of pavement cracking can vary (Roque et al. 2010, Zhang et al. 2015). A pave- ment might develop its first crack at an early stage, whereas the severity of the crack (length, width, and quantity of crack) might not increase significantly for a long time. For crack initiation, a probabilistic-based model is more suitable than a deterministic-based model (Finn et al. 1986). In this project, a binary logistic (BL) regression method was used together with the PLS regression method to develop probabilistic-based crack initiation models for transverse cracking and longitudinal cracking. Instead of providing a yes or no answer to the prediction of crack initiation, the probability for a pavement to crack under a particular condi- tion (material, environment, structure, etc.) is determined. Depending on the probability results, engineers and agencies can set up different threshold values as a limit of crack initia- tion to aid in decision-making.