Improved Conditioning and Testing Procedures for HMA Moisture Susceptibility (2007)

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.

62.1 Phase I Experimental Program and Findings Details of the Phase I experimental program and its find- ings were documented in the Interim Report submitted on January 5, 2004. A summary of Phase I work is presented here. Table 1 presents the experimental matrix for the Phase I study. In this design, three aggregate sources with different levels of moisture sensitivity based on past research were tested using the three NCHRP Project 9-19 simple perfor- mance tests, ASTM D4867, and the HWTD. For the simple performance tests, specimens were tested unconditioned and after the ECS moisture procedure. The ASTM D4867 testing also included tests on unconditioned and conditioned speci- mens. The HWTD testing used submerged specimens only. The aggregates included in the Phase I experiment were selected based on their performance in previous studies. The Texas limestone aggregate had historically exhibited good resistance to moisture damage. Mixtures produced with the sandstone aggregate typically had AASHTO T283 tensile strength ratios in the range of 0.83 to 0.86, which were lower than the typical results obtained for the limestone aggregate. Mixtures produced with the Virginia granite had poor resistance to moisture damage without the aid of an anti- stripping agent. The same asphalt source (PG 70-22) was used to produce the mixtures for the Phase I experiment. The binder source used in the moisture-sensitive Virginia granite mixture was selected for the Phase I evaluation. The mixtures for the Phase I evaluation were laboratory reproductions of actual project mixtures made from the three aggregate sources, with the exception of using the same binder for all mixtures. Table 2 provides a summary of the mixtures used in the Phase I experiment. Table 3 presents a summary of the interpretation of the findings from the Phase I experiment. This table includes the ratings from past studies upon which the selection of the three aggregates was based, moisture sensitivity results and ratings from the five tests conducted in Phase I, and the cri- teria used in the ratings. The primary conclusion from Phase I of NCHRP Project 9-34 was that the dynamic modulus test was the most suited of the three simple performance tests for possible use with the ECS in an improved moisture sensitivity test. The ECS/ Dynamic Modulus test was able to differentiate between mixes made with stripping resistant aggregates (such as Texas limestone) and those with aggregates prone to stripping (such as Virginia granite). The Phase I testing showed that the dynamic modulus decreases significantly when a moisture- sensitive material is conditioned using the ECS conditioning procedure. For the three materials tested in Phase I, the ratio of the conditioned to unconditioned dynamic modulus cor- rectly identified the known moisture-insensitive and the known moisture-sensitive aggregates. The ratio of the condi- tioned to the unconditioned dynamic modulus ranked a sus- pected marginal aggregate as insensitive; however, based on ASTM D4867 and the HWTD, it appears that mixtures made with this aggregate and the specific binder used in Phase I are, in fact, moisture insensitive. The ECS/dynamic modulus test ranked the three Phase I mixtures the same way as ASTM D4867 ranked them. In contrast to the dynamic modulus test, when the ECS was combined with the two flow tests, the results were generally irrational. Strains and strain rates were consistently lower in ECS specimens than strains and strain rates in unconditioned specimens. The repeated loading in the ECS conditioning pro- cedure hardens the conditioned specimens, making it impos- sible to use the flow time test on unconditioned specimens and specimens conditioned with the ECS. The flow number test could possibly be used if the ECS conditioning procedure were modified to perform the flow number test during the conditioning process. However, this was not the case with this research since flow number tests were conducted after the ECS conditioning procedure. Although it appears feasible to C H A P T E R 2 Experimental Program

7conduct the flow number test during the ECS conditioning process, the flow number test has been shown to have a high level of variability in other studies. This high level of variabil- ity results in the test having poor sensitivity to changes caused by moisture conditioning. With four specimens per test and the current level of variability reported for the flow num- ber test, the flow number for conditioned specimens must decrease by approximately 40 percent for the conditioned and unconditioned tests to be considered significantly different. The high level of variability in the flow number test is a result of the flat slope of the permanent deformation curve for a large number of cycles prior to flow. This flat slope makes it very difficult to detect the exact point at which flow occurs in most mixtures. It is unlikely that the further development Aggregate Source NCHRP Project 9-19 Simple Performance Tests Type Location Resistance to Moisture Damage Flow Time Flow Number Dynamic Modulus ASTM D4867 HWTD Limestone Hunter, TX Good X X X X X Sandstone Sawyer, OK Marginal X X X X X Granite Richmond, VA Poor X X X X X Material Type Source Resistance to Moisture Damage Nominal Maximum Aggregate Size (mm) Design Asphalt Content Limestone Hunter, TX Good 12.5 5.4% Sandstone Sawyer, OK Moderate 19.0 5.3% Granite Richmond, VA Poor 19.0 4.5% Resistance to Moisture Damage Test Criteria Texas Limestone Oklahoma Sandstone Virginia Granite1 Expected Performance from Previous Studies Good Moderate Poor Minimum TSR= 80%2 Pass Pass Fail ASTM D4867 TSR 87% 89% 66% Stripping Inflection Point (SIP) Good Good Good SIP > 20,000 passes > 20,000 passes >16,000 passes TxDOT Maximum Impression Depth Pass Pass Pass HWTD Impression Depth at 20,000 passes 4.9 mm 2.2 mm 6.5 mm ECS/Flow Time Significant Difference in Flow Time No Flow No Flow Pass ECS/Flow Number Significant Difference in Flow Number No Flow No Flow No Flow Significant Difference in Dynamic Modulus Pass Pass Fail ECS/Dynamic Modulus Modulus Ratio 0.92 0.95 0.66 1All three mixes were prepared using the binder for the Virginia granite mix, and, therefore, the mixes for Texas limestone and Oklahoma sandstone were not prepared using specific binders from the sources identified in the mix design. 2TSR = Tensile strength ratio. Table 1. Phase I test matrix. Table 2. Summary of aggregate sources used in the Phase I experiment. Table 3. Interpretation of Phase I findings.

8planned for this test will decrease the variability to a level where it can be acceptable in a moisture sensitivity test. 2.2 Phase IA Experiment Design 2.2.1 Testing Program The research conducted during Phase IA included ECS/ dynamic modulus testing, ASTM D4867 tests, and HWTD tests. A total of eight different mixes was selected. Two broad categories of mixes were included in the experiment: (1) those reported to have poor field performance with respect to mois- ture damage and (2) those reported to have good field per- formance with respect to moisture damage. Selection of ma- terials was skewed toward poorly performing mixes since it is crucial to properly identify these mixes before construction. Thus, the work plan included five mixes reported to perform poorly and three mixes reported to perform well. For each mix, ASTM D4867 and HWTD tests required six and four test specimens, respectively. ASTM D4867 tests were conducted at Advanced Asphalt Technologies, LLC (AAT). The HWTD specimens were tested at PaveTex Engineering and Testing (PaveTex). For the ECS/dynamic modulus test- ing, six replicate specimens for each mix were considered. The ECS/dynamic modulus testing was conducted at the materials laboratories of the Northeast Center of Excellence for Pavement Technologies (NECEPT) at Pennsylvania State University (PSU), as well as the materials laboratories of the University of Texas at El Paso (UTEP). Therefore, there were a total of 12 ECS/dynamic modulus specimens for each mix—six for each laboratory. Table 4 presents the Phase IA testing matrix and location of tests. 2.2.2 Selection of Proper Sample Size for ECS/Dynamic Modulus Testing An important consideration for Phase IA of this study was the selection of the sample size for testing conducted with the ECS/dynamic modulus test. Reasonable estimates of the required sample size were determined considering confidence intervals for the mean and standard deviation, as discussed below. In general, the authors consider n specimens to be fabri- cated for each mix and each specimen to be tested for dynamic modulus before and after moisture conditioning with the ECS/dynamic modulus system. The property of interest is the percentage of the dynamic modulus retained after condition- ing (i.e., the ratio of modulus after conditioning to modulus before conditioning), designated as Retained %. Thus Retained % = [(Dyn. Modulus After ECS) / (Dyn. Modulus Before ECS)] × 100% The values of the Retained % for the n specimens will pro- vide information regarding the moisture susceptibility for the material. After consideration of possible outliers, the average of the remaining values will be the estimate of the ability of the mixtures to resist moisture-induced damage. In Phase I of this study, it was found that the estimated standard deviation for a single measured Retained % was 5.7 percent. With this as the assumed true standard deviation for a single measured Retained %, the standard deviation for the average of n such measurements would be . The values for this standard deviation of the average are given in Table 5 for values of n from 1 to 10. It is clear that additional samples beyond approximately six samples will not provide sufficient benefit to justify the added cost with respect to the estimation of the Retained %. It should be noted that the standard deviation of the aver- age of the measurements for each cell presented in Table 5 is the quantity that determines the power of the test and the precision of the confidence intervals. The accuracy of the estimation of the standard deviation of a single measured Retained % is the second important ques- tion when a sample size is being considered. This estimation is important as one is considering the properties of the testing process and the measurements in a given cell for the combi- nation of test device and test procedures of interest. It would be desirable to have a testing process for which the standard deviation is known to be small. For each cell of interest, a good estimate of the standard deviation of the measured values in the cell will be required. A useful approach is to consider the length of the confidence interval for the standard deviation that will result as a function of the number of samples. The 90-percent confidence interval for the true standard deviation 5 7. % / n Test Location Number of Superpave Gyratory Compacter (SGC) Specimens Per Mix Number of Conditioned Specimens HWTD PaveTex 4 4 ASTM D4867 AAT 6 3 ECS/dynamic modulus PSU and UTEP 12 (6 for each lab) 12 (6 for each lab) Table 4. Phase IA testing matrix and location of tests.

of a single measured value of Retained % is a function of sam- ple size and standard deviation. The sample size function, F(n), is determined using chi square distribution with n − 1 degrees of freedom for the selected level of confidence. The values of F(n) are given in Table 6 for n between 2 and 10. It is clear from Table 6 that for the purpose of estimating the standard deviation of the measured values in a given cell, it would be good to have at least nine samples. However, con- sidering that the data from the two laboratories for a given cell can be pooled (resulting in 12 observations), the common standard deviation can be reasonably estimated considering six replicates per mix for each laboratory. 2.3 Materials Selection The most important criterion in selecting a specific mix for Phase IA was that the mix should have a known field performance record. The authors’ best estimate of field per- formance was qualitative rather than rigorously quantitative. The best approach was to consider the two general categories of mixes, those that performed poorly and those that per- formed well, rather than sequential ranking of all the mixes. The performance of a mix will vary when it is exposed to dif- ferent traffic, climatic, and construction conditions in the field. Therefore, comparing the performance of two mixes on tests in a controlled laboratory environment will not yield the most useful information on mix performance in the field. A mix might have failed severely in a short period of time in one project because it was exposed to a considerable amount of rain and heavy loads, drainage problems, and possibly had high air-void contents resulting from either poor mix design or poor compaction. A different mix might have exhibited marginal moisture damage just because it was well con- structed and was located in an environment with less rain, lighter traffic, and good drainage. For these reasons, it was difficult to identify mixes that might be referred to as “mar- ginal.” Quantitative ranking of different mixes in different environments is a difficult task unless accurate traffic and environment and construction data are available and consid- ered in conducting such a ranking. Qualitative data was the best type of information that could be obtained as the basis for the comparisons of mixtures in Phase IA. Selection of specific mixes for this project was based on the following factors: • Feedback from the project panel members, • Feedback from other experts regarding the history of mois- ture damage and field performance of candidate mixes, • Available creditable literature on previous research on moisture damage, and • Feedback from state personnel in the areas where the mixes/aggregates were commonly used. Five mixes reported to perform poorly and three mixes reported to perform well were identified for Phase IA of this research. Materials and pertinent mix designs for these mixes were received from materials producers. For each mix, the corresponding asphalt binder was also received from the per- tinent asphalt producer. The sections that follow discuss the materials used in this phase of the project. 2.3.1 Georgia Granite Georgia granite, widely available in Georgia, has very good frictional properties. However, this material has demonstrated poor resistance to moisture damage. Most of the problems noticed with this aggregate in Georgia date back to the 1980s; attempts to solve these problems led to most of the research regarding moisture damage in the state (23). Since detecting moisture damage problems with this material, Georgia DOT has been using 1 percent lime in mixes containing Georgia granite. Georgia granite was also used in one of the mixes placed in the Pavement Test Track of the National Center for Asphalt Technology (NCAT) and was included in 1995 research conducted by Tunnicliff and Root (24). Conversa- tions with past and present Georgia DOT personnel resulted in the selection of Georgia granite from a specific source for use in this study. 9 Value of n 1 2 3 4 5 6 7 8 9 10 Std. Dev. of Mean (%) 5.7 4.0 3.3 2.9 2.5 2.3 2.2 2.0 1.9 1.8 Table 5. Standard deviations of the average retained modulus for different sample sizes. Value of n 2 3 4 5 6 7 8 9 10 F(n) 15.29 3.83 2.3 1.72 1.42 1.23 1.09 0.99 0.92 Table 6. Sample size function F(n) for determination of 90-percent confidence interval for standard deviation.

2.3.2 Pennsylvania Dolomite Pennsylvania dolomite, a dolomitic limestone, has been used in many projects throughout the Commonwealth of Pennsylvania with no apparent signs of stripping. The good historical performance of this aggregate along with the fact that it has been used without any antistripping agent made it an excellent source to be included in the project. 2.3.3 Mississippi Chert Widely available in Mississippi, Mississippi chert has gener- ally shown poor moisture damage resistance in the field. The moisture susceptibility of this aggregate resulted in significant change to the mixes used in the state in the early 1990s. The im- provements included limiting natural sand, reducing the ratio of material passing the #200 sieve to the binder content, and using lime and liquid antistripping agents. The chert rock from one of the available sources has shown stripping problems with one source of binder and good resistance to moisture damage when the binder source was changed. The poor resistance to moisture damage has been observed with plant mixes even with lime added to the mix. The chert rock from one of the sources in Mississippi was also among the aggregates exhibit- ing poor performance in Tunnicliff and Root’s 1995 laboratory evaluation of antistripping additives in asphaltic concrete mix- tures (24). Chert rock from the same source has been used in one of the sections in the NCAT Pavement Test Track. 2.3.4 Kentucky Limestone This crushed limestone aggregate is from a quarry in Ken- tucky. The aggregate has shown stripping problems at the pavement surface in the field after 1 to 2 years of service. The material has also exhibited severe failure in the HWTD when used with the local natural sand and with a PG 64 binder. The performance, however, has been good with a polymer- modified PG 76 binder. The mix with a PG 64 binder was selected as one of the candidates for this research. 2.3.5 Arkansas Gravel Mixes prepared with gravel from one of the quarries in Arkansas have resulted in several pavement failures in Texas due to moisture damage. The mix design information for these mixes was available. The failed mixes had been prepared with a specific source of binder that is graded as a PG 64-22. The mix also failed when tested in the HWTD. The data from the completed ECS study at UTEP provided important infor- mation for this mix (19). In that research, the mix clearly demonstrated poor resistance to moisture damage. The per- formance of the mix had been reported to be excellent both in the field and in the HWTD when a PG 76-22 polymer- modified binder with 1 percent lime was used with this aggregate. The mix with the PG 64-22 binder that exhibited failure in the field (see above) was included in this study. 2.3.6 Oklahoma Sandstone The Oklahoma sandstone used in Phase IA of this research has been used with a PG 76-22 binder to make a mix that per- forms well. This mix was placed on IH-20 in the Atlanta District of Texas in 2001 and exhibited excellent performance based on measurements in 2003. This mix also demonstrated very little deformation in the HWTD (3 to 4 mm) when sub- jected to 20,000 passes at 50°C. The IH-20 study is a 5-year field/laboratory moisture damage research project that will be a valuable source of information for this mix. Therefore, this mix was selected as mix that performs well for this project. 2.3.7 Wisconsin Gravel Materials and mix design were received from a gravel source in Wisconsin that, historically, has shown excellent resistance to moisture damage. The gravel for the selected mix has been used with no lime or liquid antistripping agent. This mix utilizes a PG 58-28 binder and is the softest mix used in this research. 2.3.8 Wyoming Gravel Wyoming gravel, which has a high silica (SiO2) content, has been known as a highly moisture-sensitive aggregate if no hydrated lime is used. Various laboratory studies on this aggre- gate during SHRP and in the post-SHRP period have proven the moisture sensitivity of this aggregate with a range of differ- ent binders. As a result of this well-documented poor behavior, a specific mix design for this material, with known poor field performance, was requested and received from the Wyoming Department of Transportation (DOT) for use in this research. 2.4 Laboratory Testing Program The mixtures procured for Phase IA of the research were subjected to three types of tests: ASTM D4867, HWTD, and dynamic modulus before and after the ECS conditioning pro- cedure. Prior to preparation of the specimens for the tests, a procedure was followed to verify the mix designs. 2.4.1 Mix Design Information and Verification As discussed previously, eight mixtures were used in Phase IA of NCHRP Project 9-34. For each mixture, detailed 10

mix design information was requested from both the respon- sible highway agency and the hot-mix supplier. Production quality control data and acceptance data were also requested, but none of the agencies or suppliers provided these data. Table 7 summarizes the mix design data that were provided for each of the eight mixtures. Detailed mix design data were provided for seven of the eight mixtures used. For the Arkansas gravel, only the gradation, optimum binder content, and design compaction level were provided. The date that each mixture design was prepared is presented in Table 7 along with the date that samples were received for NCHRP Project 9-34. The time between the mixture design and the verification ranged from less than 1 month to 4.25 years. The findings from the mix design verification study are dis- cussed in Chapter 3. 2.4.2 Specimen Preparation All specimens for Phase IA of NCHRP Project 9-34 were fabricated by AAT using standard procedures. The sections that follow discuss procedures used in the specimen fabrica- tion process for: • Handling of binders and aggregates; • Laboratory mixing, aging, and compaction; • Fabrication of specimens for the ECS/dynamic modulus tests; and • Shipping of test specimens. 2.4.2.1 Binder Handling Samples of the binder used in each mixture were shipped to AAT by the respective material supplier in either 1-gal or 5-gal metal cans. Upon receipt at AAT, the binder samples were divided into quart containers by heating the original container in an oven set at 135°C, stirring with a mechanical stirrer, and pouring the binder into the individual quart con- tainers. A representative sample was obtained from one of the quart containers, and viscosities were determined at 135°C, 150°C, and 165°C—in accordance with AASHTO T316—to determine appropriate mixing and compaction tempera- tures. The quart containers were then used in the preparation of laboratory mixture batches. Quart containers were heated only once. Excess binder in the quart containers was dis- carded. The Phase IA testing program required approxi- mately 7.5 l (2 gal) of binder for each mixture. 2.4.2.2 Aggregate Handling Representative samples of the aggregates used in each mix- ture were shipped to AAT by the respective suppliers in sam- ple bags, plastic containers, and metal cans of varying sizes. The procedures described in the appendix of Mix Design Methods for Asphalt Concrete and Other Hot-Mix Types (25) were used to prepare the aggregate samples for laboratory batching. Coarse aggregate samples were separated into indi- vidual sizes, while individual samples of fine aggregate were mixed together to produce a homogeneous supply for subse- quent batching. Specimen batches were made assuming that the propor- tions given in the mixture design were based on washed gra- dations analyses. The research team attempted to verify that all of the mixture designs were based on washed gradation analyses; however, the team was unable to achieve this verifi- cation. The gradation of the blends was verified by perform- ing washed sieve analysis on one of the batches. The results of these sieve analyses are discussed in Chapter 3. 2.4.2.3 Mixing, Aging, and Compaction Four different size gyratory specimens were used in Phase IA. Specimens for mixture verification were 150 mm (5.90 in.) in diameter by 115 mm (4.53 in.) high. Specimens for the ASTM D4867 testing were 100 mm (3.94 in.) in diameter by 65 mm (2.5 in.) high. Specimens for the HWTD were 150 mm (5.90 in.) in diameter by 62 mm (2.44 in.) high. Speci- mens for the ECS/dynamic modulus tests before sawing and coring were 150 mm (5.90 in.) in diameter by 165 mm (6.5 in.) high. After sawing and coring, the ECS/dynamic modu- lus specimens were 100 mm (4.0 in.) in diameter by 150 mm (5.9 in.) high. All gyratory specimens were prepared to a tar- get air void content in accordance with AASHTO T312. The reported air void contents are those for the final test speci- men, which is the complete specimen for the verification, ASTM D4867 and HWTD tests, and the sawed and cored specimen for the ECS/dynamic modulus tests. An Interlaken compactor meeting the requirements of AASHTO T312 was used to prepare the HWTD and ECS/dynamic modulus test specimens. An Invelop Oy compactor meeting the require- ments of AASHTO T312 was used to prepare the ASTM D4867 specimens. Mixing and compaction temperatures for the binders were determined from viscosities measured at 135°C, 150°C, and 165°C—in accordance with ASTM D316. These viscosi- ties were converted to kinematic viscosities using the binder specific gravity measured at 25°C and the specific gravity tem- perature correction factors given in Annex A1 of AASHTO T201. Table 8 presents mixing and compaction tempera- tures for each mixture. Prior to compaction, materials for all specimens were short-term oven-aged in accordance with AASHTO R30 for 2 h at the compaction temperature. Again, the research team attempted to verify, but ultimately was unable to verify, that the volumetric properties of 11

Time between Design and Verification, yrs Sieve Size, mm 25 19 12.5 9.5 4.75 2.36 1.18 0.6 0.3 0.15 0.075 GA Granite 5/3/2000 8/5/2003 3.26 12.5 Coarse PG 67-22 Gyratory 75 100 100 99 83 50 36 27 20 15 9 5.1 4.7 4.0 2.592 2.477 2.795 1.0302 2.802 0.086 4.6 15.5 74.3 11.5 1.1 NA AR Gravel 10/3/2003 10/20/2003 0.05 19.0 Fine PG 64-22 Gyratory 100 100 100 88 80 60 48 36 27 16 7 4.3 5.0 4.0 NA1 NA NA 1.030 NA NA NA NA NA NA NA NA WI Gravel 7/8/2003 12/1/2003 0.40 12.5 Fine PG 58-28 Gyratory 100 100 100 97 89 70 53 38 27 14 7 4.8 5.4 4.0 2.504 2.404 2.684 1.031 2.726 0.597 4.8 15.3 73.8 11.3 1.0 87.0 MS Chert 4/14/2003 12/1/2003 0.63 12.5 Fine PG 67-22 Gyratory 68 100 100 95 89 65 47 35 27 14 7 5.3 5.4 4.0 2.371 2.276 2.513 1.038 2.559 0.735 4.7 14.3 72.1 10.3 1.1 NA KY Limestone 5/7/2003 5/24/2004 1.05 9.5 Coarse PG 64-22 Gyratory 100 100 100 100 95 64 42 29 21 11 6 4.5 5.4 4.0 2.458 2.360 2.640 1.030 2.669 0.427 5.0 15.4 74.1 11.4 0.9 81.0 OK Sandstone with Lime 3/20/2001 6/23/2004 3.26 12.5 Coarse PG 76-22 Gyratory 125 100 100 92 79 49 29 22 19 15 10 6.5 5.1 4.0 2.373 2.278 2.541 1.030 2.552 0.172 4.9 14.9 73.2 10.9 1.3 NA PA Dolomite 7/13/2000 10/15/2004 4.26 19.0 Coarse PG 64-22 Gyratory 75 100 100 82 66 40 26 16 12 9 6 4.0 4.6 3.8 2.597 2.498 2.772 1.030 2.803 0.406 4.2 14.0 72.9 10.2 0.9 90.0 WY Gravel 4/24/2002 12/21/2004 2.66 12.5 Fine PG 64-22 Marshall 75 100 100 95 77 53 39 25 20 13 10 6.4 5.0 5.0 2.452 2.329 2.577 1.030 2.644 1.015 4.0 14.1 64.6 9.1 1.6 16.9 Mixture Design Date Sample Receipt Date Property Asphalt Content, wt % Air Voids, vol % Maximum Specific Gravity of Mix Gradation, % passing AASHTO M323 Nominal Max Size, mm AASHTO M323 Gradation Classification Binder Grade Design Compaction Compaction Level, Ndesign/Blows Bulk Specific Gravity of Specimen Bulk Specific Gravity of Aggregate Bulk Specific Gravity of Binder Effective Specific Gravity of Aggregate Effective Binder Content, vol % Dust to Effective Binder Content, % Tensile Strength Ratio, % Absorbed Asphalt, wt % aggregate basis Effective Binder Content, wt % Voids in Mineral Aggregate, vol % Voids Filled with Asphalt, % 1NA = data not available. 2Binder specific gravity data in bold was assumed. Table 7. Summary of mix design data provided for the mixtures used in Phase IA.

mixture designs that were provided were based on specimens that were short-term oven-aged. 2.4.2.4 Sawing and Coring of ECS/Dynamic Modulus Test Specimens The ECS/dynamic modulus test specimens were manufac- tured by coring and sawing test specimens 100 mm (3.94 in.) in diameter by 150 mm (5.90 in.) in height from the middle of gyratory compacted specimens that were 150 mm (5.90 in.) in diameter by 165 mm (6.5 in.) in height. The procedure for preparing dynamic modulus specimens is described in AASHTO TP62. There are three reasons for using smaller test specimens obtained from larger gyratory specimens in the dynamic modulus test. The first is to obtain an appropriate aspect ratio for the test specimens. Research performed under NCHRP Project 9-19 found that a minimum height-to- diameter ratio of 1.5 was needed for the dynamic modulus test (26). The second reason is to eliminate areas of high air voids in the gyratory specimens. Gyratory compacted speci- mens typically have high air voids near the ends and the circumference of the specimen. The third reason is to obtain relatively smooth, parallel ends for testing. Dynamic modulus test specimens were prepared to the dimensional tolerances listed in Table 9. These tolerances are somewhat different from those specified in AASHTO TP62 and are the result of a study performed under NCHRP Project 9-29 to determine practical tolerances for dynamic modulus specimen prepara- tion (27). Several laboratories have adapted equipment for preparing dynamic modulus test specimens. The various approaches range from elaborate feed-control drills combined with sophisticated holders and double-bladed saws to standard drills and single-bladed saws with simple clamping arrange- ments. For this project, specimens meeting the tolerances listed in Table 9 were prepared using a portable core-drilling machine and a double-bladed saw. As shown in Figure 1, the portable core-drilling machine was mounted to a heavy stand on the laboratory floor to facilitate vertical drilling of the 13 Source Mixing Temperature (°C) Compaction Temperature (°C) GA Granite 163 152 AR Gravel 159 156 WI Gravel 150 135 MS Chert 158 148 KY Limestone 148 130 OK Sandstone with Lime 166 155 PA Dolomite 157 145 WY Gravel 165 145 Item Specification Remarks Average Diameter 100 mm to 104 mm (3.94 in to 4.09 in) See Note 1 Standard Deviation of Diameter 0.5 mm (0.02 in) See Note 1 Height 147.5 mm to 152.5 mm (5.81 in to 6.00 in) See Note 2 End Flatness 0.5 mm (0.02 in) See Note 3 End Perpendicularity 1.0 mm (0.40 in) See Note 4 1Measure the diameter at the center and third points of the test specimen along axes that are 90 deg apart. Record each of the six measurements to the nearest 0.1 mm. Calculate the average and the standard deviation of the six measurements. The standard deviation shall be less than 0.5 mm. The average diameter, reported to the nearest 0.1 mm, shall be used in all material property calculations. 2Measure the height of the test specimen in accordance with Section 6.1.2 of ASTM D3459. Record the average height. 3Using a straightedge and feeler gauges, measure the flatness of each specimen end. Place a straightedge across the diameter at three locations approximately 120 deg apart and measure the maximum departure of the specimen end from the straightedge using tapered-end feeler gauges. For each end, record the maximum departure along the three locations as the end flatness. 4Using a combination square and feeler gauges, measure the perpendicularity of each end. At two locations approximately 90 deg apart, place the blade of the combination square in contact with the specimen along the axis of the cylinder and place the head in contact with the highest point of the end of the cylinder. Measure the distance between the head of the square and the lowest point on the end of the cylinder using tapered-end feeler gauges. For each end, record the maximum measurement from the two locations as the end perpendicularity. Table 8. Mixing and compaction temperatures for the mixes used in the study. Table 9. NCHRP Project 9-29 specimen dimension tolerances (27). Figure 1. Portable core-drilling machine and stand.

specimen. The gyratory compacted specimen of 150 mm (5.90 in.) in diameter by 165 mm (6.50 in.) high was held in place under the drill by blocks of wood cut to provide a tight fit between the gyratory specimen and the stand. A sophisti- cated clamp for holding the gyratory specimen is not needed to obtain the tolerances on the specimen diameter listed in Table 9. Figure 2 shows the 100-mm (3.94-in.) diameter core and the waste portion of the gyratory specimen. Reasonably smooth, parallel ends for the test specimen were then provided by trimming the 100-mm-diameter (3.94-in.) core using the double-bladed saw shown in Figure 3. This step is more critical than the coring step and requires the 100-mm-diameter (3.94-in.) core to fit tightly in the saw clamp and sufficient waste material on each end to keep the saw blades from bending. All coring and sawing was done using water to cool the cut- ting tools. After all cutting was complete, the bulk specific gravity of the finished specimen was determined, in accor- dance with AASHTO T166, by first measuring the immersed mass, then the saturated surface dry mass, and finally the dry mass. A completed test specimen is shown in Figure 4. 2.4.3 Testing Sequence The mixtures were tested in the order given in Table 10. For each mixture, the verification was completed first. The order for the remaining tests depended on the workload in each laboratory. Generally, the ASTM D4867 specimens were fabricated and tested first, followed by the specimens for the HWTD testing, which were shipped to PaveTex. The 12 dynamic modulus test specimens were usually prepared last. These 12 specimens were ranked based on their air void contents, then separated into two groups to provide ap- proximately the same average and standard deviation of air void contents within each group. One group was shipped to PSU, and the other was shipped UTEP for ECS/dynamic modulus testing. Appendix B presents specimen identifica- tion numbers and air void contents for the specimens that were tested. All specimens that were shipped for either the HWTD or the ECS/dynamic modulus testing were packaged as a set of four specimens per box in cardboard boxes. Voids in the boxes were filled with packaging materials to minimize the potential for damage to the specimens. Upon receipt of the specimens at PaveTex, PSU, or UTEP, the boxes and specimens were inspected for damage during shipping. 2.4.4 ASTM D4867 This testing was performed at AAT in accordance with ASTM D4867. Six specimens, three conditioned and three unconditioned, were prepared and tested for each mixture. Briefly, the test includes compaction and preparation of at least six specimens. The compacted specimens should have 14 Figure 2. 100-mm-diameter core and waste ring. Figure 3. Double-bladed saw with 100-mm core. Figure 4. Final dynamic modulus test specimen.

air void contents between 6.0 and 8.0 percent. Half of the compacted specimens are conditioned through a freeze cycle (optional) followed by water bath. First, a vacuum level of approximately 525 mm of mercury (Hg) is applied to par- tially saturate specimens to a level between 55 and 80 percent. Vacuum-saturated samples are kept in a –18°C freezer for 16 h and then placed in a 60°C water bath for 24 h. After this period, the specimens are considered conditioned. The other three samples remain unconditioned. All of the samples are brought to a constant temperature, and the indirect tensile strength is measured on both dry (unconditioned) and con- ditioned specimens. The test method used in this research included the freeze cycle. Results from the ASTM D4867 tests are presented and analyzed in Chapter 3. 2.4.5 Hamburg Wheel Tracking Device (HWTD) The HWTD measures the combined effects of rutting and moisture damage by rolling a steel wheel across the surface of an asphalt concrete test specimen that is immersed in hot water (see Figure 5). Rutting, as a function of number of passes, is recorded and used to compute the following test parameters: • Maximum impression depth, • Creep slope, • Stripping slope, and • Stripping inflection point. Figure 6 presents a schematic of a typical rutting curve from the HWTD and the definition of the test parameters. The test was performed for all eight mixtures using cylin- drical specimens in accordance with TxDOT test method Tex-242-F. The testing was performed at 50°C for a total of 20,000 wheel passes. The mix is considered a failing mix if the measured rutting after a specified number of passes exceeds 12.5 mm. The number of passes at which the rutting pass/fail criterion is applied depends on the binder grade. For mixes with a high performance grade (PG) of 64, 70, and 76, the number of passes for rutting consideration is 5,000, 10,000, and 20,000, respectively. A summary of test method Tex- 242-F is provided in Table 11. The test was also conducted on specimens prepared in the form of slabs for two of the mixtures. This was done for a comparison of results from testing Superpave gyratory spec- imens with results from slabs. Results from the HWTD tests are presented and discussed in Chapter 3. 2.4.6 Dynamic Modulus and Environmental Conditioning System The environmental conditioning system that was devel- oped during SHRP was used for accelerated conditioning of the specimens. During development stages, this system was the subject of several research projects (15–17). The system was further researched and modified later (18, 20). This mod- ified version was used in the current research. The change in dynamic modulus as a result of water/load conditioning under the ECS is used as a measure of moisture damage in the test. 2.4.6.1 Water Conditioning The ECS water flow control device presented in Figure 7 was used to provide the accelerated water conditioning. This 15 Figure 5. Submerged specimens in the HWTD and the associated testing. Order Mixture 1 Georgia Granite 2 Arkansas Gravel 3 Wisconsin Gravel 4 Mississippi Chert 5 Kentucky Limestone 6 Oklahoma Sandstone with Lime 7 Pennsylvania Dolomite 8 Wyoming Gravel Table 10. Sequence of mixture testing.

02 4 6 8 10 12 14 16 18 20 0 8,0006,0004,0002,000 10,000 14,00012,000 16,000 18,000 20,000 Number of Passes R ut D ep th , m m Creep Slope Stripping Slope Stripping Inflection Point Figure 6. Schematic of HWTD test results. Apparatus: HWTD The load applied by the wheel is 158 ± 5 lb. (705 ± 22 Newtons [N]) The wheel shall make approximately 50 passes across the test specimen per min. The maximum speed of the wheel must be approximately 1.1 ft/sec (0.305 m/s) and will be reached at the midpoint of the slab. Temperature Control System A water bath capable of controlling the test temperature within ± 4oF (2oC) over a range of 77 to 158oF (25 to 70oC). Rut Depth Measurement System A Linear Variable Differential Transducer (LVDT) device capable of measuring the rut depth induced by the steel wheel within 0.0004 in (0.01 mm), over a minimum range of 0.8 in (20 mm). The system shall be mounted to measure the rut depth at the midpoint of the wheel's path on the slab. Rut depth measurements must be taken at least every 100 passes of the wheel. Specimen: Specimen diameter shall be 6 in (150 mm) and specimen height should be 2.4 ± 0.1 in (62 ± 2 mm). Air void of test specimens must be 7 ± 1%. Procedure: Test requires two cylindrically molded specimens with the Superpave Gyratory Compactor. Place a specimen in the cutting template mold and use a masonry saw to cut it along the edge of the mold. The cut across the specimen should be approximately 5/8 in (16 mm) deep. Place the high-density polyethylene molds in the mounting tray and fit specimens into each one. Secure the molds in the mounting tray. Test temperature shall be 122 ± 2oF (50 ± 1oC) for all hot-mix asphalt specimens. Fill the water bath until the water temperature is at the desired test temperature. The temperature of the water can be monitored on the computer screen. Start the test after the test specimens have been in the water for 30 min at the desired test temperature. The testing device automatically stops the test when the device applies the desired number of passes or when the maximum allowable rut depth has been reached. Report: For each specimen, report the air void content, antistripping additive used, number of passes to failure, and rut depth at the end of the test. Table 11. Summary of test method Tex-242-F as used in this research.

device consists of a vacuum pump, a water source, valves, pressure gauges, and flow meters. The system was used to apply and monitor the flow of water. Constant flow of air or water can be achieved through suction applied by the vacuum pump. The water supply tank (not shown) was positioned approximately 1.5 m (5 ft) above the specimen to provide suf- ficient water head. The water was guided through spiraled pipes in a controlled hot water bath (not shown) before per- meating through the specimen to ensure proper temperature. The 60°C water was run through the encapsulated specimen from the top for 18 h. The water flow rate was approximately 8 cm3/min (0.5 in.3/min). 2.4.6.2 Load Conditioning A repeated haversine load was applied simultaneously with water conditioning (see Figure 8). Every pulse of the load had a duration of 0.1 sec followed by a rest period of 0.9 sec. This loading on the specimen continued for 18 h. The load level during conditioning was selected based on the temperature 17 Specimen Outlet GaugeSpecimen Inlet Gauge Gauge 1 Vent/Off Gauge 2 Vent/Off Fluid Selector Air/Water/Vacuum Air On/Off Water On/OffVacuum On/Off Vacuum Regulator Water Flowmeters Air Flowmeters Pressure Differential Gauge Figure 7. Water flow control device for the ECS. 0.00 0.05 0.10 0.15 TIME, SEC LO AD A XI A L ST RA IN TIME LAG, TI PERIOD, TP εOσO Figure 8. The specimen set-up for testing and the corresponding sinusoidal load.

of the site at which the mix was constructed and varied between 670 and 930 N (150 and 210 lb) depending on the site. An experiment was conducted to establish the load level, as discussed later in this chapter. Load and water condition- ing parameters are shown in Table 12. 2.4.6.3 Dynamic Modulus Testing The ECS/dynamic modulus tests were conducted at the laboratories of both PSU and UTEP. For each mix, a total of 12 replicate specimens were tested, with 6 at each laboratory. Dynamic modulus tests were conducted on the same speci- men three times. The first test was on the dry unconditioned specimen, followed by a second test after the specimen was exposed to vacuum partial saturation with distilled water for 30 min at 25°C. The last dynamic modulus test was con- ducted after the specimen was exposed to full load/water conditioning for 18 h at 60°C in the ECS. The dynamic modulus testing was conducted with a uni- axial sinusoidal load inducing approximately 100 μstrain in the specimen (see Figure 8). All dynamic modulus tests were conducted at 25°C. Selection of the 25°C test temperature was based on the findings of research under NCHRP Project 9-29, which concluded that dynamic modulus testing at moderate temperatures close to 25°C produced less variability in results than tests at extreme temperatures such as −10°C or 40°C, respectively. Specimen set-up and temperature control are also more easily managed at moderate temperatures. The loading frequencies for each specimen were 10, 5, 2, and 1 Hz, applied in decreasing order. Some of the earlier mixes, tested at PSU, were also the subject of a 25-Hz loading frequency. The 25-Hz frequency was dropped later for consistency with testing at UTEP. The dynamic modulus and phase angle are defined by Equations 1 and 2, respectively. (1) (2)φ = ⎛⎝⎜ ⎞ ⎠⎟ × T T i p 360 E * = σ ε 0 0 where |E*| = dynamic modulus, σ0 = amplitude of applied sinusoidal loading, ε0 = amplitude of resulting sinusoidal strain, φ = phase angle in degrees, Ti = time lag in seconds, and Tp = period of sinusoidal loading in seconds. At PSU laboratories, three Linear Variable Displacement Transducers (LVDTs) were used at 120° to capture deforma- tion of the specimen during both dynamic modulus testing and repeated loading of the conditioning phase (see Figure 9). At UTEP, two LVDTs were used to measure deformation during the dynamic modulus test, and no deformation data were captured during the conditioning phase. Dynamic mod- ulus testing parameters are presented in Table 13. 2.4.6.4 Modifications to the Dynamic Modulus Test System of Phase I Work in Phase II of NCHRP Project 9-29 indicated that the dynamic modulus test could be very repeatable, with the coef- ficient of variation for a single dynamic modulus test to be 13 percent. This value was based on pooling data on 576 dy- namic modulus measurements made in two laboratories using two test devices. This level of test variability compares well with data from NCHRP Project 9-19 that showed coefficients of vari- ation ranging from 13 to 16 percent (26, 28). For the six speci- mens planned for each cell of the ECS/dynamic modulus test- ing during Phase IA, the coefficient of variation of the average of the modulus measurements should be approximately percent, a reasonable value for this type of me- chanical test. The data on the dynamic modulus testing vari- ability were used as a guide in Phase IA for identifying outliers and assessing the quality of the collected data. In addition to es- tablishing benchmark variability data for the dynamic modulus test, NCHRP Project 9-29 identified the following factors that were carefully considered in the dynamic modulus testing of Phase IA: • Specimen temperature, • Gauge length for measuring strains, 13 6 5 3/ .= 18 Item Temperature Duration/ Frequencies Magnitude Loading Type Short-Term Conditioning (Static Vacuum Saturation) 25 C 30 min 625 mm Hg Temperature 60 C 18 h — — Load 60 C 18 h Site Specific Haversine (0.1 sec loading and 0.9 sec rest) Vacuum 18 h 100 mm Hg — Long-Term Conditioning (ECS) Water Flow 60 C 18 h 8 cm3/min — Table 12. Load and water conditioning parameters used in this research.

• Strain level used in the testing, • End friction reducer, and • Data analysis and quality of raw data collected. Modifications that were made to the dynamic modulus to reduce variability during Phase IA follow. Specimen Temperature. A very important consideration in dynamic modulus testing is control of the specimen tem- perature. The NCHRP Project 9-19 test protocols require controlling temperature to ±0.5°C and recommend using dummy specimens with embedded thermocouples to moni- tor specimen temperatures. In lieu of dummy specimens, the protocols recommend specific equilibrium times that appear to be too short based on work completed in an FHWA pooled-fund study (29). For ECS/dynamic modulus testing, the problem of temperature control is further complicated by the partial saturation and membrane used in the testing. For Phase IA, equilibrium times for the initial unconditioned dynamic modulus, the dynamic modulus after saturation, and the dynamic modulus after conditioning were deter- mined using specimens instrumented with thermocouples. These equilibrium times were then used in the subsequent testing. Later in this chapter, the experiment followed to ensure proper temperature equilibrium will be discussed. Gauge Length. In Phase I, strains were measured over a 100-mm gauge length as specified in the NCHRP Project 9-19 test protocols. For this gauge length on a 150-mm-high spec- imen, the strain measuring system is mounted 25 mm from the end of the specimen. One of the issues identified in NCHRP Project 9-29 is the parallelism and flatness of the sawed specimens used in the dynamic modulus testing. To reduce errors associated with end effects, the NCHRP Project 9-29 research recommended reducing the gauge length to 70 mm based on previous research conducted in NCHRP Project 9-19 (26). The reduction in gauge length increases the reproducibility of data from the two deformation sensors mounted on the specimen. Side-to-side differences in meas- ured strains were generally less than 20 percent in NCHRP Project 9-29 compared to 50 percent or more in NCHRP Project 9-19. For Phase IA, a 70-mm gauge length was used. Strain Level. The NCHRP Project 9-19 dynamic modulus test protocol specifies controlling the strain on the specimen between 50 and 150 μstrain. Due to possible nonlinear effects, particularly at high temperatures, as reported by Pellinen (28), these tolerances were reduced to 75 to 125 μstrain in NCHRP Project 9-29. Strain data for Phase I were generally collected below the NCHRP Project 9-19 minimum values, resulting in more variable data. For Phase IA, the NCHRP Project 9-29 limits of 75 to 125 μstrain were used. End Friction Reducer. An end friction reducer was not used in the Phase I testing. The double latex membrane rec- ommended by the NCHRP Project 9-19 test protocols was deemed impractical for use with the ECS conditioning pro- cedure. This system was also considered impractical by the NCHRP Project 9-29 researchers for use in production mix- ture design testing. Teflon sheets with a thickness of 0.28 mm were found to be an acceptable alternative in NCHRP Project 9-29 and were used as end friction reducers in the Phase IA testing. The sheets were perforated to allow permeation of water through the specimen. 2.4.6.5 Data Analysis and Quality of Raw Data The NCHRP Project 9-19 dynamic modulus test protocol permits various data analysis methods to be used to calculate the dynamic modulus from the measured stresses and strains. 19 Gage points: Transducers 100 mm 15 0 m m 70 m m 3 @ 120 120° Figure 9. Schematics showing configuration of LVDTs on the specimen. Parameter Value/Type Temperature 25 ± 0.5 o C Load Pattern Sinusoidal Frequencies 251 , 10, 5, 2, and 1 Hz Load Level Variable Displacement Measurement 3 LVDTs at 120 o Axial Direction Measurement Span in Axial Direction 70 mm Strain Level 100 ± 25 µstrain 125-Hz frequency applied to a limited number of specimens at PSU. Table 13. Description of parameters for dynamic modulus testing.

In NCHRP Project 9-29, such latitude for data analysis was considered unacceptable for mixture specification testing. Standard data collection and analysis algorithms were devel- oped in NCHRP Project 9-29 and implemented in the first article simple performance test devices evaluated in the proj- ect. The standard methods are based on regression analysis of sinusoidal data and, in addition to the modulus and phase angle data, produce data quality statistics that indicate to the testing technician the acceptability of the test data. The data quality statistics and the recommended criteria levels for good quality data (based on NCHRP Project 9-29) are sum- marized in Table 14. The standard errors of the load and deformations indicate how closely the applied loading and the measured deforma- tions reproduce sinusoidal forms. The dynamic modulus analysis is only applicable to sinusoidal loading. Poor loading waveforms, both in shape and frequency, and noisy defor- mations will increase these standard errors. Both are less than 10 percent. The deformation drift is a measure of the perma- nent strain that accumulates during the dynamic modulus test. It is expressed as a percentage of the measured strain amplitude and is limited to 4 times the measured strain amplitude, or roughly 400 μstrain. The deformation unifor- mity and phase uniformity measure how close responses from the individual sensors on the specimen are to each other. These are essentially the coefficient of variation for the deformation and the standard deviation of the phase angle and are less than 20 percent and 3 degrees, respectively. An Excel spreadsheet was developed in NCHRP Project 9-29 to process raw data and to compute the dynamic modulus, phase angle, and data quality statistics. This spreadsheet was modified to include deformation measurements from three LVDTs and was used by the technicians performing the tests in Phase IA to quickly assess the quality of the dynamic mod- ulus test data and repeat testing as needed. 2.4.6.6 Establishing Temperature Control Procedure at Various Stages During Phase IA, significant attention was paid to control- ling temperature during various stages of the test. This was specifically important because different temperatures were used during dynamic modulus testing and conditioning. To ensure that the specimen temperature was maintained at 60°C during conditioning and that dynamic modulus meas- urements were performed at 25°C, an experiment was con- ducted at UTEP that consisted of installing a thermocouple inside the specimen and subjecting the specimen to the tem- perature conditioning procedure. The only exception was that the dynamic modulus measurements were not per- formed. A datalogger was used to record the temperature of the specimen every 5 min. The specimen temperature moni- tored during various steps is presented in Figures 10 through 13. Figure 10 shows that approximately 1 h is required to reach equilibrium after placement of the specimen inside the chamber, which is maintained at 25°C. Figure 11 shows the equilibrium time required before the dynamic modulus measurement of the vacuum-saturated specimen is taken. The figure shows that the specimen ini- tially has a lower temperature and reaches equilibrium after approximately 1 h inside the chamber. Therefore, at least a 1-h waiting period is required to conduct vacuum-saturated dynamic modulus measurements after placement of the spec- imen inside the chamber maintained at 25°C. A typical result of the specimen temperature during con- ditioning is shown in Figure 12. The results summarized in Figure 12 suggest that the specimen reaches a temperature of 20 Statistic Criteria for Good Quality Data Standard Error of the Load < 10 percent Standard Error of the Deformations < 10 percent Deformation Drift < 400 percent Deformation Uniformity < 20 percent Phase Angle Uniformity < 3 degrees Table 14. NCHRP Project 9-29 recommended data quality statistics. 20.0 22.5 25.0 27.5 30.0 0 15 30 45 60 75 90 105 Time, minutes Te m pe ra tu re , ° C Chamber Specimen Chamber Specimen 20.0 22.5 25.0 27.5 30.0 0 15 30 45 60 75 Time, minutes Te m pe ra tu re , ° C Figure 10. Equilibrium time for unconditioned dynamic modulus measurements (Tests at UTEP). Figure 11. Establishing equilibrium time for vacuum-saturated dynamic modulus measurements (tests at UTEP).

60°C between 2.5 and 3 h of conditioning and remains con- stant during the remaining conditioning period. The figure also suggests that the chamber temperature and water bath temperature must be set at a level higher than 60°C to ensure the specimen temperature is maintained at 60°C. To measure the conditioned dynamic modulus, the speci- men temperature needs to be reduced from 60°C to 25°C. To expedite the temperature drop, the chamber temperature is initially set at 15°C for 1.25 h and then raised to 25°C. The specimen and chamber temperature measurements are pre- sented in Figure 13. The results suggest that approximately 2.25 h of equilibrium time is needed after the temperature drop from 60°C is initiated. In a separate experiment at PSU, the evaluation of the tem- perature condition was conducted using nine thermocouples on the surface of the specimen and at the center of a dummy specimen, and it was found that the temperature is well con- trolled within the specified range. During each actual test, tem- perature was monitored using a thermocouple at the center of the dummy specimen. A temperature example for one of the specimens is shown in Figures 14 and 15. The figures indicate the time it takes for the temperature to rise from 25°C to 60°C 21 20 30 40 50 60 70 Water Bath Specimen Chamber 0.00 5.00 10.00 15.00 20.00 Time, hours Te m pe ra tu re , ° C Figure 12. Specimen temperature during conditioning (tests at UTEP). Chamber Specimen 10 20 30 40 50 60 0.00 0.50 1.00 2.50 3.00 1.50 2.00 Time, hours Te m pe ra tu re , ° C 20 25 30 35 40 45 50 55 60 65 5/ 16 /0 4 5: 00 5/ 16 /0 4 10 :0 0 5/ 16 /0 4 15 :0 0 5/ 16 /0 4 20 :0 0 5/ 17 /0 4 1: 00 5/ 17 /0 4 6: 00 5/ 17 /0 4 11 :0 0 5/ 17 /0 4 16 :0 0 5/ 17 /0 4 21 :0 0 Time Te m pe ra tu re (o C ) Specimen MD314.13 5/16/2004 Unconditioned specimen tested at 10:42 Vacuum-saturated specimen tested at 13:46 Water/load conditioning started at 14:01 and continued for 18 hours 5/17/2004 Water/load conditioning stopped at 8:01 Conditioned specimen was tested at 15:48 Figure 13. Equilibrium time for conditioned dynamic modulus measurements (tests at UTEP). Figure 14. Temperature measured at the center of a dummy specimen during an actual test (tests at PSU).

and the time it takes for the temperature to drop from 60°C to 25°C. The times needed to reach equilibrium at PSU were longer than the times at UTEP, perhaps due to differences in the equipment used. During the PSU experiments, it was observed that after completion of 18-h, 60°C conditioning, it took 6 to 7 h before the specimen temperature became suitable for testing dynamic modulus at 25°C. Times for different events during the test are shown in Figure 14. A close-up of the time it takes for the temperature to drop from 60°C to 25°C is shown in Figure 15. It should be mentioned again that the temperature of the cham- ber is first dropped from 60°C to 15°C and maintained at this level for about 1.5 h before setting it at 25°C. Also, in all experiments for establishing equilibrium time, a dummy speci- men was partially saturated along with actual test specimens. The significance of proper testing temperature control for dynamic modulus is evident from Figure 16. This simple experiment conducted at PSU indicates that, for a typical mixture, a deviation of 2°C from 25°C results in a modulus decrease or increase of 10 percent (for testing at 25 Hz) and 18 percent (for testing at 1 Hz). 2.4.6.7 Establishing Load Levels during Conditioning In the original ECS conditioning procedures developed under SHRP at Oregon State University and those developed at UTEP during the 1990s, the haversine load during condi- tioning was maintained at a constant peak of 200 lb (890 N). The selection of this conditioning load level is not well docu- mented in SHRP Report A-403 (30). The report states, “[t]his loading level was selected from others (not reported here) to be moderate enough to minimize permanent deformation.” The report presents data from a mixture composed of Mate- rials Reference Library (MRL) asphalt AAG-1 and MRL aggregate RB at air void levels of 5 and 8 percent. The total accumulated permanent axial strain for the 8-percent air void mixture was approximately 2.5 percent; for the 5-percent air void mixture, it was approximately 1.5 percent. Maintaining the same temperature (60°C) and the same load (890 N) during the conditioning procedure for all mixes has some deficiencies, as site conditions are not properly taken into account when mixes are used in different areas with different pavement temperatures. Since the temperature is maintained at a constant 60°C during the ECS procedure, the magnitude of load needs to be adjusted for different mixes to account for differences in the site temperature and binder grades. In Phase IA, a study was undertaken to address this issue and to establish the conditioning load levels for differ- ent pavement temperatures. Alternatives to Establishing the Load Level. There are three alternatives for the conditioning load—constant load, constant dynamic strain, and constant mechanical damage. These are discussed below: 22 Specimen MD314.13 5/16/2004 ECS Conditioning ends at 8:01 Dry specimen was tested at 10:42 Partially saturated specimen was tested at 13:46 Load/Water conditioning began at 14:01 and continued for 18 hours 5/17/2004 Conditioned Specimen was tested at 15:48 20 25 30 35 40 45 50 55 60 65 5/ 17 /0 4 14 :0 0 5/ 17 /0 4 13 :0 0 5/ 17 /0 4 12 :0 0 5/ 17 /0 4 11 :0 0 5/ 17 /0 4 10 :0 0 5/ 17 /0 4 9: 00 5/ 17 /0 4 8: 00 5/ 17 /0 4 7: 00 5/ 17 /0 4 6: 00 5/ 17 /0 4 15 :0 0 5/ 17 /0 4 16 :0 0 5/ 17 /0 4 17 :0 0 5/ 17 /0 4 18 :0 0 Time Te m pe ra tu re (o C ) Figure 15. Evaluating the time it takes for the temperature to decrease from 60°C to 25°C (tests at PSU).

• Constant Load. This is the approach used currently in AASHTO TP34 and the Texas modified ECS conditioning procedure. The conditioning load is 890 N (200 lb). This is the easiest approach to implement, but the strain in the specimen and the level of non-moisture-induced damage varies with mixture stiffness, particularly binder grade. Greater dynamic strains and levels of non-moisture- induced damage will occur in mixtures made with softer grades of binder. This approach probably biases the test re- sults in favor of stiffer binders. • Constant Dynamic Strain. The second approach is to ad- just the load level to obtain constant dynamic strains in the specimen. In this case, the load level would be adjusted, based on the grade of the binder, such that the dynamic strains induced in the specimens remain constant. Stiffer mixtures would be tested at higher loads to maintain con- stant dynamic strain. • Constant Mechanical Damage. The third approach is to adjust the load level to obtain a constant level of non- moisture-induced mechanical damage in the specimens at the end of conditioning. In this case, the load levels would be adjusted such that the permanent deformation in the test without moisture would be constant. Again, this would result in stiffer mixtures being tested at higher loads to maintain constant mechanical damage. Initial Assessment of Load Level. An initial assessment of the load level was based on the grade of binder used in the mixture. The rationale behind this effort was to maintain approximately the same level of non-moisture-induced mechanical damage in the specimens. For mixtures that are moisture sensitive, additional moisture-induced damage would occur. Since the ECS conditioning procedure is per- formed at 60°C regardless of the grade of binder used in the mixture, stiffer binders should be tested at higher load levels to maintain a constant amount of mechanical damage in the test. Research reported by Kaloush, (31), has shown that per- manent strains can be predicted from resilient strains using the following model: (3) where εp = accumulated permanent strain, εr = resilient strain, N = number of load cycles, and T = temperature in °F. Rearranging Equation 3 to solve for the permanent strains produces (4) For a particular temperature and number of load cycles, which is the case in the ECS conditioning procedure, Equa- tion 4 shows that the accumulated permanent strain is proportional to the resilient strain. Based on Equation 4, con- ditions resulting in the same resilient strain should produce the same accumulated permanent strain. The resilient strain is proportional to the modulus of the material, so when test- ing materials with different moduli, the load level should be adjusted to maintain a constant resilient strain level. ε εp r N T= [ − + +( )10 3 1555 0 3994 1 7340. . log( ) . log( ) ] log . . log . lo ε ε p r N ⎛ ⎝⎜ ⎞ ⎠⎟ = − + ( )+3 1555 0 3994 1 7340 g T( ) 23 0 1,000 2,000 3,000 4,000 5,000 6,000 0 5 10 15 20 25 30 Loading Frequency (Hz) 23°C 25°C 27°CM od ul us (M Pa ) Figure 16. Effect of deviation from 25°C temperature on the measured dynamic modulus at different frequencies (tests at PSU).

Resilient strains can be estimated by dividing the applied load by the mixture modulus. Mixture modulus values for different binder stiffnesses can be estimated using the Witczak predictive equation (32): (5) where E = dynamic modulus, 105 psi; η = bitumen viscosity, 106 Poise; f = frequency, Hz; Va = air void content, %; Vbeff = effective bitumen content, % by volume; ρ34 = cumulative % retained on 19-mm sieve; ρ38 = cumulative % retained on 9.5-mm sieve; ρ4 = cumulative % retained on 4.76-mm sieve; and ρ200 = % passing 0.075-mm sieve. For the AAG-1 mixture used in the original SHRP ECS research, the following parameters are used in the dynamic modulus predictive equation to obtain the modulus repre- sentative of the ECS conditioning procedure at 60°C: η = 0.003253, 106 Poise; f = 10 Hz; Va = 8%; Vbeff = 10.75% by volume; ρ34 = 5%; ρ38 = 32%; ρ4 = 52%; and ρ200 = 5.5%. This results in a modulus at 60°C of 47,000 psi. Moduli for the same mix can be estimated using Equation 5 and 60°C vis- cosity values representative of various binder grades. In NCHRP Project 1-37A, representative viscosity temperature parameters were developed to predict viscosity values for var- ious binder grades using the ASTM viscosity-temperature susceptibility relationship: log . . . ( )E = − + −1 249937 0 029232 0 001767200 200 2ρ ρ . . .− − −0 002841 0 058097 0 8022084ρ Va V V beff beff + ⎛⎝⎜ ⎞⎠⎟ + − + Va . . .3 871977 0 0021 0 0039584 38ρ ρ − + + − − 0 000017 0 005470 1 38 2 34 0 6033 3 0 . ( ) . ( . ρ ρ e ` . log( ) . log( ))313351 0 393532f − η (6) where η = viscosity, cP; TR = temperature, Rankine; A = regression intercept; and VTS = regression slope of viscosity temperature suscepti- bility. These binder grades are summarized in Table 15 along with the representative viscosity at 60°C for that grade, the modulus from Equation 5 assuming the volumetric parame- ters representative of the mix used in the original ECS research, and the estimated loading to produce equivalent resilient strains at 60°C. The estimated load is the ratio of the modulus for the given binder grade to the standard modulus of 47,000 psi multiplied by the standard ECS conditioning procedure load of 200 lb. Both the constant dynamic strain approach and the con- stant mechanical damage approach would yield the same load adjustment factors since the Kaloush permanent deformation equation reduces, for a specific temperature and number of load cycles, to the permanent strain being proportional to the resilient or dynamic strain. 2.4.6.8 Laboratory Study for Establishing the Conditioning Load Level Following the initial assessment, a laboratory study was undertaken to verify the load levels by conducting repeated load permanent deformation tests on a limited number of samples made with different grades of binder. This testing was used to confirm that similar resilient and permanent strains were obtained at 60°C when the load was varied with binder grade (binder grades were as given in Table 15). An ex- periment was designed to confirm these load levels using a single mixture with three different binders. Table 16 summarizes volumetric properties of the mixture selected for this study. It was a 12.5-mm mixture that used crushed limestone coarse aggregate and a blend of manufac- tured and natural sand. The natural sand was used at 14.0 percent of the total aggregate. This represented approxi- mately 41.0 percent of the fine aggregate fraction. Specimens for the load level study were prepared using three binders: a log log logη = +A VTS TR 24 Grade VTS A 60 Viscosity (cP) Modulus 10 Hz (psi) Estimated Load (lb) 52-34 3.602 10.707 105,210 31,371 133 58-28 3.701 11.010 227,151 42,031 179 64-22 3.680 10.980 521,740 57,804 246 70-22 3.426 10.299 1,123,552 77,534 330 76-22 3.208 9.715 2,285,502 101,500 432 82-22 3.019 9.209 4,403,607 129,670 552 Table 15. Estimated loads for ECS procedure by binder grade.

neat PG 58-28, a neat PG 64-22, and an elastomeric modified PG 76-22. The experimental design for the load level study consisted of performing permanent deformation tests without mois- ture at 60°C using three load levels for each binder grade. The load levels bracket the estimates presented in Table 15. Repli- cate specimens were tested for each binder/load level combi- nation. Table 17 summarizes the experimental design. The experiment required fabricating and testing 18 specimens. Table 18 summarizes the air void content of these specimens. Each specimen was tested under repeated haversine load at UTEP at 60°C without water conditioning. The load period for each cycle was 0.1 sec with the rest period being 0.9 sec. To simulate the effect of confinement during the ECS condi- tioning procedure, the specimens were encapsulated in a membrane and tested while a partial vacuum of 2.5 in. (8.5 kPa) of Hg was applied. The conditioning load pulses were applied for 3 h (9,500 cycles), and resilient strains and per- manent strains were recorded. The strain measurement set- up was modified to measure the strains during conditioning while the membrane was on the specimen. This was achieved by placing the targets on the specimens and affixing the LVDT mounting system outside the membrane. The mount- ing targets were modified to add a screw that passes through the membrane and is connected to the LVDT mounting sys- tem. The hole was plugged using Super Glue to make sure that the vacuum level was maintained. The results of this experimental program are presented in Chapter 3. 2.4.6.9 Resolving Testing System Differences between the Two Laboratories One concern that needed to be addressed was ensuring that the testing systems at the PSU and UTEP laboratories produced comparable results. This was accomplished by 25 Nominal Maximum Aggregate Size 12.5 mm Ndesign 75 Coarse Aggregate Angularity. One Face/ Two Face 100/100 Fine Aggregate Angularity 47.2 Flat & Elongated, % (Ratio 5:1) 3.0 Sand Equivalent, % 55 Binder Content, % 4.75% Gyratory Compaction, % Gmm Nini 86.4% Ndes 96.0% Nmax 97.3% Voids in Mineral Aggregate (VMA), % 14.6 Voids in Total Mixture (VTM), % 4.0 Voids Filled with Asphalt (VFA), % 72.6 Fines to Effective Binder Ratio (F/A) 1.2 Gradation, % passing sieve size, mm 37.5 100 25 100 19 100 12.5 97 9.5 75 4.75 39 2.36 30 1.18 24 0.6 18 0.3 11 0.15 7 0.075 5.6 Table 16. Volumetric properties of the mixture used in the load level study. Repeated Load Permanent Deformation Test on Dry Unconditioned Specimens at 60 CBinder Load Level (lb) Replicate 1 Replicate 2 130 X X 168 X X PG 58-28 180 X X 190 X X 210 X X PG 64-22 240 X X 240 X X 280 X X PG 76-22 410 X X PG 58-28 PG 64-22 PG 76-22 Specimen Air Voids Specimen Air Voids Specimen Air Voids MD 310.1 6.2 MD 309.1 6.3 MD 308.1 6.7 MD 310.2 6.4 MD 309.2 6.2 MD 308.2 6.6 MD 310.3 7.2 MD 309.3 6.7 MD 308.3 6.8 MD 310.4 6.5 MD 309.4 6.8 MD 308.4 6.4 MD 310.5 6.8 MD 309.5 6.5 MD 308.5 6.3 MD 310.6 6.3 MD 309.6 7.3 MD 308.6 6.4 Average 6.6 Average 6.6 Average 6.5 Standard Deviation 0.36 Standard Deviation 0.40 Standard Deviation 0.20 Table 17. Experimental design for the load level verification study. Table 18. Air void content of load level verification study specimens.

conducting dynamic modulus tests on a specimen made from Polyethylene Terephthalate (PET) at 25°C and at different frequencies. The tests were first conducted at PSU, and then the specimen was shipped to UTEP for testing. The synthetic specimen was again tested at PSU after completion of UTEP tests. The results are shown in Figure 17. This graph indicates that the modulus results from the tests at both laboratories are considerably close. The difference between the average modulus values from the two laboratories is about 4 percent of the overall average. 26 Dynamic Modulus vs. Frequency 3,200 3,250 3,300 3,350 3,400 3,450 3,500 3,550 3,600 0 2 4 6 8 10 12 Frequency (Hz) PSU 6/28/04 UTEP 7/8/04 PSU 7/16/04 PSU 7/19/04 UTEP 7/8/04 D yn am ic M od ul us (M Pa ) Figure 17. Comparison of results from dynamic modulus tests on a synthetic specimen from laboratories at PSU and UTEP.