National Academies Press: OpenBook

Volumetric Requirements for Superpave Mix Design (2006)

Chapter: Chapter 2 - Findings

« Previous: Chapter 1 - Introduction and Research Approach
Page 8
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 8
Page 9
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 9
Page 10
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 10
Page 11
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 11
Page 12
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 12
Page 13
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 13
Page 14
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 14
Page 15
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 15
Page 16
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 16
Page 17
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 17
Page 18
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 18
Page 19
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 19
Page 20
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 20
Page 21
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 21
Page 22
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 22
Page 23
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 23
Page 24
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 24
Page 25
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 25
Page 26
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 26
Page 27
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 27
Page 28
Suggested Citation:"Chapter 2 - Findings." National Academies of Sciences, Engineering, and Medicine. 2006. Volumetric Requirements for Superpave Mix Design. Washington, DC: The National Academies Press. doi: 10.17226/13999.
×
Page 28

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.

8The sections below present the specific findings of NCHRP Projects 9-25 and 9-31. This chapter is divided into eight sec- tions: Literature Review and Survey of Practice, Laboratory Testing, Analysis of Other Data Sets, Rut Resistance, Fatigue Resistance, Permeability and Age Hardening, Apparent Film Thickness and HMA Performance, and Summary. These dis- cussions describe the most effective relationships between performance-related properties and volumetrics as identified and/or developed during this research and graphically illus- trate what these models predict in terms of property changes as a function of VMA, design air voids, and related composi- tional factors. The practical implications of the findings pre- sented here are discussed in Chapter 3. Literature Review and Survey of Practice A variety of research papers and engineering reports were reviewed during the first few months of NCHRP Project 9-25 and presented in the Interim Report for that project; an updated version of the Literature Review was included in the NCHRP Project 9-31 Interim Report. One of the most important issues in HMA mix design is how to define “optimum” asphalt content. In the current Superpave system, this is defined as the binder content that produces 4% air voids at the given compaction level. In order to evaluate the effectiveness of this practice, a range of Super- pave mix designs, Marshall mix designs, and stone matrix asphalt (SMA) mix designs were reviewed, and the optimum binder content—defined in this case as the point at which minimum VMA is obtained—was determined. This is a more fundamental definition of optimum asphalt content than that which is currently used in the Superpave system. It was found that for these data, the optimum binder content based on min- imum VMA occurred at an average air void content of 3.4%, but could also be defined as occurring at an average of 75.3% VFA. In fact, the optimum asphalt content based on minimum VMA for these mixtures appeared to cover a range of air void contents—from about 2% to 5%. A dramatic increase in rut- ting potential has been associated with in-place air void con- tents of around 2% or less (8). Thus, very low design air void contents (say less than about 3%) should be avoided in wear- ing and intermediate course mixtures because low design air void contents should be expected to promote low in-place air voids and increase the possibility of constructing a pavement with poor rut resistance. Therefore, it appears reasonable to use a range for design air voids of from 3% to 5%. However, as discussed later in this report, engineers and technicians should be aware that changing the value for design air void content will significantly affect HMA performance. A variety of models were identified in the literature review for predicting performance-related properties from HMA composition and other properties. The Hirsch model for predicting HMA modulus, as developed during the early phases of NCHRP Project 9-25, was found to be more suit- able for relating modulus to volumetric composition than other existing models—Bonnaure’s equation and Witczak’s equation (7, 9, 10). Two models for predicting rut resistance were identified, both developed by Witczak and associates (11, 12). In both cases, the model predicted the results of a laboratory test for evaluating rut resistance and not field rutting. The models were similar, and both found that rut resistance increased with decreasing binder volume and air voids and increasing binder viscosity. A serious shortcom- ing of both models was the use of binder apparent viscosity values at 21.1 °C, rather than Superpave binder properties. A more useful model for predicting rut resistance was devel- oped during NCHRP Projects 9-25 and 9-31: it predicts that rut resistance increases with decreasing VMA relative to aggregate fineness and increasing binder viscosity (or com- plex modulus). Existing models for predicting the fatigue resistance of HMA have been empirically derived from laboratory flexural fatigue tests. Typically, such fatigue equations relate applied C H A P T E R 2 Findings

stress or strain, initial complex modulus, and either VBE or VFA to cycles of failure. In all cases, better fatigue resistance is predicted as a mixture becomes increasingly rich in asphalt binder, either as indicated by VBE or VFA (13–15). The empirical nature of these relationships and their relatively poor accuracy when applied to fatigue of actual pavements are serious shortcomings that lead to the development of a practical continuum damage approach to characterizing fatigue phenomena in HMA as part of NCHRP Projects 9-25 and 9-31 (16). A number of researchers in the past attempted to relate mixture volumetrics (most often VMA and or asphalt binder film thickness) to durability. This work mostly involved conjecture without substantial supporting data and so was inconclusive. The concept of binder film thick- ness remains controversial. In general, there is agreement in this early research that a certain amount of asphalt binder is needed in a mixture to ensure adequate durability and that the optimal binder content will depend to some extent on the properties of the aggregate used, including NMAS and specific surface (17–19). Most researchers have found a decrease in permeability and age hardening with decreased air void content, although such relationships usually exhibit a large amount of variability (19–21). Recent research on the permeability of Superpave mixtures in Florida has demon- strated that unlike the relatively fine, dense-graded HMA used in the past, coarse-graded Superpave mixtures can exhibit relatively high levels of permeability unless thor- oughly compacted (3). The substantial data set published by the Florida researchers has been analyzed to generate a use- ful equation for estimating mixture permeability from air void content and aggregate fineness, which is discussed later in this report. Only one method for predicting age hardening was located in the literature—Mirza and Witczak’s global aging system (22). This model predicts age hardening of asphalt binder in pavements based upon mean annual air temperature (MAAT), binder viscosity, depth in the pavement, and air void content. This model has several shortcomings, the most important being a reliance on binder apparent viscosity val- ues estimated from obsolete empirical measures of binder consistency and the prediction of age hardening only in terms of a change in apparent viscosity, rather than in terms of changes in the overall flow characteristics of the binder. A modification of the global aging system was developed that addresses some of these problems while maintaining consis- tency with the original model. This model is used later in this report to estimate the effect of changes in mixture composi- tion on typical age hardening of asphalt mixtures and binders. Although not specifically listed as one of the project objec- tives, analysis of laboratory data generated during NCHRP Projects 9-25 and 9-31 and of field data generated in a variety of other projects indicated that any modeling of the relation- ship between volumetric composition and performance must account for relative compaction—the air void content of either the laboratory specimen or in-place pavement relative to the air void content as designed. It is essential that this information be included in this report so that researchers attempting to validate the results of this research will under- stand the importance of accounting for the effects of relative compaction. However, to place the findings of NCHRP Proj- ects 9-25 and 9-31 in proper perspective, research found in the literature concerning the effect of in-place air voids on performance must also be discussed. Two significant such studies are NCHRP Project 20-50(14) and the research of Linden et al. (23, 24). In NCHRP Project 20-50(14), Seeds et al. analyzed data from the Long-Term Pavement Performance (LTPP) program. They found that the data could not be used to develop performance models relating in-place air voids to either fatigue or permanent deformation (23). In 1988, Lin- den et al. published results of a study conducted by Washing- ton State evaluating the relationship between pavement performance and in-place air voids (24). They reported that as a “rule of thumb,” every 1% increase in in-place voids results in about a 10% reduction in performance. This figure was a very rough, typical value, based on the results of several studies: (1) three analytical studies relating fatigue life to in- place voids; (2) a survey involving 28 state highway agencies; (3) an unpublished study of flexible pavements in Washing- ton State; and (4) observed fatigue cracking in three pave- ments placed in Washington with high air void contents. The analytical studies cited by Linden and Mahoney reported a 10% to 30% reduction in fatigue life for every 1% increase in in-place voids (25–27). No analytical studies on the effect of in-place air voids on rut resistance were cited in this study. The results of these studies should be considered inconclusive— NCHRP Project 20-50(14) was unable to develop any useful, reliable relationships between in-place air voids and per- formance; the study by Linden et al. was limited in scope, and even the authors admitted their conclusions represented only “a rule of thumb.” As part of NCHRP Projects 9-25 and 9-31, in late 2001 and early 2002, a survey was conducted of the manner in which state highway agencies are implementing Superpave specifi- cations for volumetric composition. Many states have slightly modified the requirements for Superpave mixture composi- tion as given in AASHTO M323 and R35. Most commonly, the design air voids content is expanded to a range of 3% to 5% and a maximum VMA is established at 1.5% to 2% above the established minimum values. A number of states have also slightly increased minimum VMA values, providing for somewhat richer mixtures than produced by the current version of Superpave. 9

Laboratory Testing As part of NCHRP Projects 9-25 and 9-31, a wide range of laboratory tests was performed on a variety of HMA. The lab- oratory tests were designed to provide information concern- ing the rut resistance, fatigue resistance, permeability, and resistance to age hardening of the mixtures studied. The most important of the procedures performed as part of this research included the following tests: • RSCH testing using the Superpave shear tester, at 58°C and 64°C (AASHTO T320-03); • Uniaxial fatigue testing at 10 Hz and 4°C and 20°C, which includes an initial measurement of the complex modulus |E*| (16); and • Mixture permeability, using the Florida permeability test (Florida Test Method FM 5-565). The mixtures tested were made using eight different aggre- gates and gradations: • A 9.5-mm limestone from Virginia, coarse and fine gradations; • A 19-mm gravel from Pennsylvania, coarse and dense gradations; • A 19-mm limestone from Kentucky, coarse and dense gradations; and • A 12.5-mm granite from California, dense and fine gradations. All of these mixtures were combined with a performance grade (PG) 64-22 binder. Some were also combined with a PG 58-28 binder and/or an air blown PG 76-16 binder. In most cases, the design gyration level was 100, but for some mixtures, Ndesign was 75. All of the California granite mixtures were designed using 125 gyrations. Early in the project, an attempt was made to design and evaluate some mixtures at 50 gyrations, but these were typically very weak and difficult to test; further testing of these mixtures was therefore aban- doned. All mixtures were made using three binder contents: the optimum binder content, optimum −1%, and optimum +1% (by total mix weight). The materials used represented a range of aggregate types, gradations, binder grades, and mix- ture compositions. Analysis of Other Data Sets As discussed later in this short report, some of the find- ings made during the research appeared very promising, but somewhat controversial. Therefore, in several cases ver- ification of the findings was attempted using data sets from other research projects. In evaluating the rut resistance of the NCHRP Projects 9-25 and 9-31 mixtures, the concept of resistivity was developed and appeared to relate very well to the results of the RSCH test. To further verify these results, field data from three sources was compiled and analyzed: 1. The MnRoad test track, as discussed in numerous reports by the Minnesota DOT and the University of Minnesota (5); 2. The NCAT test track, as documented by Brown et al. (6); and 3. The WesTrack project, as documented in FHWA’s Perfor- mance of Coarse-Graded Mixes at WesTrack—Premature Rutting (2). The approach in analyzing the uniaxial fatigue results involved a further development and simplification of con- tinuum damage theory. Because of the variability in fatigue data, the relatively small amount of testing performed as part of NCHRP Projects 9-25 and 9-31 and because of the novelty of the approach, further verification of the results was desired. This verification was performed by applying the same analytical approach to fatigue data gathered dur- ing the Strategic Highway Research Program (SHRP), as summarized in SHRP Report SHRP-A-404: Fatigue Response of Asphalt-Aggregate Mixes (15). Although these data were gathered using flexural fatigue tests, continuum damage theory predicts that the damage in the extreme fiber at the test conclusion for this procedure should be constant and can thus be related to the results of uniaxial tests. The permeability tests performed during this research were very limited. This occurred for two reasons—(1) the air void content of the mixtures was relatively low, resulting in low permeability values and (2) permeability tests performed in the laboratory will usually show lower values than those determined from field cores. Therefore, most of the speci- mens tested during NCHRP Projects 9-25 and 9-31 showed very low or zero permeability. To better understand the rela- tionship between mixture composition and permeability, data from the Florida permeability study was included in this analysis (3). Because understanding the extent and scope of the data used in developing the performance models developed dur- ing this research is essential to interpreting the findings pre- sented in this chapter, the external data sets summarized above are discussed in more detail in the sections below. This unfortunately increases the length and complexity of this report, but makes clear the fact that the findings are based on a much more robust set of data than that which was collected during testing performed under NCHRP Proj- ects 9-25 and 9-31. 10

Rut Resistance A very good relationship has been found between mixture rut resistance and mixture resistivity. Resistivity indicates the resistance to binder flow exhibited by a particular aggregate structure. It is analogous to electrical resistivity, defined as “a materials opposition to the flow of electrical current.” Resis- tivity can also be thought of as the inverse of the coefficient of permeability for a granular material. It increases with increas- ing binder viscosity, increasing aggregate specific surface and decreasing VMA. Because HMA is almost always designed at close to 4% air voids, VMA will normally be proportional to VBE so that resistivity is closely but indirectly related to apparent film thickness; mixtures with thin binder film thick- ness will generally exhibit high resistivity values. Resistivity can be calculated using the following formula: (1) where P = resistivity, s/nm; |η*| = binder viscosity at the temperature of interest, Pa-s; Sa = aggregate specific surface, m2/kg; Gsb = aggregate bulk specific gravity; and VMA = voids in the mineral aggregate, volume %. It should be emphasized that Equation 1 is not an empiri- cal relationship developed during NCHRP Projects 9-25 and 9-31 for characterizing the rut resistance of HMA. Instead, it is the inverse of an existing equation for estimating the per- meability of a granular material (28). Therefore, the choice of variables, the values of the exponents, and the value of the constant 4.9 are not of the authors’ choosing—they result directly from Winterkorn’s formula for permeability and reflect on a fundamental level the factors governing fluid flow through porous media. Because of the extreme influence of temperature on the flow properties of asphalt, care must be taken in selecting the temperature at which viscosity is determined when calculat- ing resistivity for HMA. For laboratory tests, the viscosity should be determined at the same temperature at which the HMA is being characterized. For field rutting, the situation is more complicated. The value used should be some tempera- ture estimated to be characteristic of the overall potential for permanent deformation in the given climate. For example, within the current Superpave system, the critical temperature for rutting used in selecting PG binders is the yearly, 7-day- average, maximum pavement temperature, measured 50 mm below the pavement surface. This is the temperature used in this research in calculating resistivity for field projects. How- ever, it should be kept in mind that the manner of calculating the critical rutting temperature will likely evolve as further P = η* . S G VMA a sb 2 2 34 9 research is done on performance modeling of HMA pave- ments. Furthermore, some researchers and engineers may prefer other approaches to estimating characteristic temper- atures for rutting in HMA pavements. An important practical question in applying Equation 1 is how to estimate the specific surface of the aggregate. In analyz- ing a wide range of aggregate gradation data, it was found that for routine purposes the aggregate specific surface can be accu- rately estimated by summing the percent passing the 75-, 150-, and 300-μm sieves and dividing the total by 5. The sum of the percent passing the 75-, 150-, and 300-μm sieves is called the fineness modulus,300-μm basis, abbreviated as FM300.The rela- tionship between this parameter and aggregate specific surface was evaluated using data from eight projects, as shown in Fig- ure 1; the data used in this analysis was as reported for NCHRP Project 9-9, the NCAT Test Track, Pooled Fund Study 176, the Florida permeability study, MnRoad, FHWA’s Accelerated Loading Facility (ALF) rutting study, WesTrack, and data from NCHRP Projects 9-25 and 9-31 (2, 3, 5, 6, 29–31). This was the best and one of the simplest methods found for relating aggre- gate gradation to aggregate specific surface and is well suited for routine use in mix design work and HMA specifications. Spe- cific surface can be estimated reasonably accurately simply by dividing FM300 by 5.The r2 for this relationship is 90%,while the 95% prediction limit for new observations (including in Figure 1) is about ± 0.8. The current method of controlling aggregate specific sur- face involves establishing limits on the amount of aggregate passing the 75-μm sieve and controlling the dust-to-binder ratio. To compare this approach with FM300, Figure 2 shows the same data set used in Figure 1, but in this case the horizontal axis is the percent finer than 75 μm. The r2 value in this case is only 76%, and the 95% prediction limit increases to ±1.3. Clearly there is a relationship between the specific surface of a given aggregate and its mineral filler content, but this rela- tionship is only moderately strong. FM300 appears to be a sig- nificantly more accurate approach and is also more flexible in 11 0 2 4 6 8 10 12 0 20 40 60 FM300 A gg . Sp ec . Su rf. , m 2 /k g NCHRP 9-9 NCAT Track P.F. 176 FL/Perm. MN/Road 9-25/31 ALF WesTrack Fit 95 % P.L. 95 % P. L. S a = 0.203FM 300 Figure 1. Estimated Aggregate Specific Surface as a Function of FM300  P75  P150  P300 (r2  90%; Plot Includes 95% Prediction Limits for New Observations).

that it will allow producers with materials deficient in mineral filler to provide additional surface area by increasing the amount of material in the 75- to 300-μm size range. Some engineers may object to the use of FM300 because it appears possible to meet requirements stated in this manner using aggregate with little or no mineral filler; however, it should be remembered that current Superpave requirements have clear minimum and maximum values on the amount of material finer than 75 μm. Figure 3 is a plot showing maximum permanent shear strain (MPSS) determined using the RSCH test as a function of Ndesign × resistivity. Multiplying resistivity by Ndesign is neces- sary to account for differences in compaction energy, which can increase resistance to permanent deformation independ- ent of mixture composition. RSCH tests were performed at 54 °C and 60 °C, and the HMA tested incorporated a range of asphalt binders; the viscosity values used in calculating resis- tivity were determined for each binder at the temperature for the corresponding RSCH test. The specimens tested represent a wide range of mix composition and Ndesign levels. Addition- ally, air void contents were varied for these mixtures by alter- ing the asphalt content ± 0.5% from the design value. The rela- tionship in Figure 3 is quite good although it appears that the MPSS values at a given level of Ndesign × resistivity are somewhat higher for mixtures containing limestone aggregate compared with the gravel and granite aggregates. One possible explana- tion for this is that the limestone aggregate, being relatively soft, breaks down during the RSCH test more than do the harder aggregates. As discussed below, limited calibration of the resistivity equation suggests that this is not a serious prob- lem in applying this approach to field-rutting data. The resistivity approach to estimating mixture rut resist- ance was verified by using field-rut data from the MnRoad, NCAT, and WesTrack project (2, 5, 6). It must be emphasized that this calibration was performed using a substantial set of existing field data and not the limited laboratory data col- lected during NCHRP Projects 9-25 and 9-31. The data used in calibrating the rutting model is summarized in Tables 1 and 2. In calculating resistivity, the 7-day average high pave- ment temperature at a depth of 50 mm was used. Further- more, the amount of binder age hardening was estimated using a modification of Mirza and Witczak’s global aging model (discussed below). A statistical analysis of this data resulted in the following semi-empirical equation: (2) where RR = Rutting rate, mm rutting/m thickness/ESALs1/3 (equivalent single axle loads); P = Resistivity, in s/nm; Neq = Ndesign or number of blows with Marshall com- paction hammer; and RD = Relative field density = (100% − in-place voids)/ (100% − design voids). The relationship between this function and the observed rutting rate is shown in Figure 4. The r2 value for this model was 89%, which is very good considering that this model includes data from three widely different climates and uses only laboratory mix data and in-place air voids to predict the rutting rate. The 90% prediction limits shown in Figure 4 cor- respond closely to plus or minus a factor of 2.0 in the esti- mated rut depth. Thus, if the estimated rut depth found with Equation 2 were 8 mm, the 90% prediction limits would be 4 to 16 mm. The 90% confidence level was chosen because for rutting, only the upper confidence level is of practical inter- est, so this corresponds to a 95% one-sided prediction limit for design purposes. Although a factor of 2 might seem large for a confidence limit, this is equivalent to a factor of safety of 2, which is common in much practical engineering work. It should be emphasized that Ndesign in Equation 2 refers to the number of gyrations (or Marshall blows) required to RR P N RDeq= − − −224 1 08 0 650 18 6. . . 12 0 2 4 6 8 10 12 0 5 10 15 P75, % A gg . Sp e c. Su rf. , m 2 /k g NCHRP 9-9 NCAT Track P.F. 176 FL/Perm. MN/Road 9-25/31 ALF WesTrack Fit 95 % P.L. 95 % P.L. S a = 2.05 + 0.623P 75 Figure 2. Estimated Aggregate Specific Surface as a Function of Material Finer than 75 m (r2  76%; Plot Includes 95% Prediction Limits for New Observations). R2 = 81% R2 = 88% 0 2 4 6 8 10 0 5000 10000 15000 N x Resistivity, s/nm R SC H M PS S, % KY Limestone PA Gravel PA Limestone CA Granite Limestone Fit Gravel/Granite Fit Figure 3. RSCH Permanent Shear Strain as a Function of Gyrations  Resistivity.

compact the specimen during quality-control (QC) testing, corresponding to Ndesign for the job mix formula (JMF). The air void content at Ndesign will, in this case, often deviate from 4.0%; however, changes in air void content and VMA are accounted for in Equation 2 in the resistivity term, which should also be calculated using QC data when possible. It should also be noted that using as-designed data when apply- ing Equation 2 to field data will often result in poor predic- tions of rutting rate because HMA mixes as-placed often vary substantially from their as-designed characteristics. If an esti- mate is needed of the effect of deviations during production from as-designed characteristics, rutting should be calculated 13 Section Mix Design Method Ndesign or Blows Aggregate Type Aggregate NMAS mm Aggregate Gradation Binder Grade Modifier Type NCAT Test Track Mixtures N1 Superpave 100 Slag/Limestone 12.5 ARZ PG 76-22 SBS N2 Superpave 100 Slag/Limestone 12.5 ARZ PG 76-22 SBS N3 Superpave 100 Slag/Limestone 12.5 ARZ PG 67-22 N/A N4 Superpave 100 Slag/Limestone 12.5 ARZ PG 67-22 N/A N5 Superpave 100 Slag/Limestone 12.5 BRZ PG 67-22 N/A N6 Superpave 100 Slag/Limestone 12.5 BRZ PG 67-22 N/A N7 Superpave 100 Slag/Limestone 12.5 BRZ PG 76-22 SBR N8 Superpave 100 Slag/Limestone 12.5 BRZ PG 76-22 SBR N9 Superpave 100 Slag/Limestone 12.5 BRZ PG 76-22 SBS N10 Superpave 100 Slag/Limestone 12.5 BRZ PG 76-22 SBS N11 Superpave 100 Granite 12.5 TRZ PG 76-22 SBS N12 SMA 50 Granite 12.5 SMA PG 76-22 SBS N13 SMA 50 Gravel 12.5 SMA PG 76-22 SBS S1 Superpave 100 Granite 12.5 BRZ PG 76-22 SBS S2 Superpave 100 Gravel 9.5 BRZ PG 76-22 SBS S3 Superpave 100 Limestone/Gravel 9.5 BRZ PG 76-22 SBS S4 Superpave 100 Limestone 12.5 ARZ PG 76-22 SBS S5 Superpave 100 Gravel 12.5 TRZ PG 76-22 SBS S6 Superpave 100 Limestone/RAP 12.5 ARZ PG 67-22 N/A S7 Superpave 100 Limestone/RAP 12.5 BRZ PG 67-22 N/A S8 Superpave 100 Marble/Schist 12.5 BRZ PG 67-22 N/A S9 Superpave 100 Granite 12.5 BRZ PG 76-22 SBS S10 Superpave 100 Granite 9.5 ARZ PG 67-22 N/A S11 Superpave 100 Marble/Schist 12.5 BRZ PG 76-22 SBS S13 Superpave 100 Granite 12.5 ARZ PG 76-22 SB MnRoad Mixtures 1 Marshall 75 Gravel/Granite 12.5 ARZ PG 58-28 N/A 2 Marshall 35 Gravel/Granite 12.5 ARZ PG 58-28 N/A 3 Marshall 50 Gravel/Granite 12.5 ARZ PG 58-28 N/A 4 Superpave 100 Gravel/Granite 12.5 ARZ PG 64-22 N/A 14 Marshall 75 Gravel/Granite 12.5 ARZ PG 58-28 N/A 15 Marshall 75 Gravel/Granite 12.5 ARZ PG 58-28 N/A 16 Superpave 100 Gravel/Granite 12.5 ARZ PG 64-22 N/A 17 Marshall 75 Gravel/Granite 12.5 ARZ PG 58-28 N/A 18 Marshall 50 Gravel/Granite 12.5 ARZ PG 58-28 N/A 19 Marshall 35 Gravel/Granite 12.5 ARZ PG 58-28 N/A 20 Marshall 35 Gravel/Granite 12.5 ARZ PG 58-28 N/A 21 Marshall 50 Gravel/Granite 12.5 ARZ PG 58-28 N/A 22 Marshall 75 Gravel/Granite 12.5 ARZ PG 58-28 N/A 23 Marshall 50 Gravel/Granite 12.5 ARZ PG 58-28 N/A WesTrack Mixtures 35 Superpave 96 Andesite 19.0 BRZ PG 64-22 N/A 38 Superpave 96 Andesite 19.0 BRZ PG 64-22 N/A 39 Superpave 96 Andesite 19.0 BRZ PG 64-22 N/A 54 Superpave 96 Andesite 19.0 BRZ PG 64-22 N/A Notes: SMA = stone matrix asphalt; RAP = recycled asphalt pavement; ARZ = above restricted zone; BRZ = below restricted zone; TRZ = through restricted zone; SBS = styrene-butadiene-styrene rubber; SBR = styrene-butadiene rubber; SB = styrene-butadiene Table 1. Properties of mixtures used in calibration of rutting model.

using both as-designed and as-produced data (using field air voids in each case to calculate relative density). The difference between these rutting rates will then provide an estimate of the effect on rutting rate of deviations from the mix design. A series of simple plots can be constructed using Equations 1 and 2 to illustrate the specific effect of changing VMA, design air voids, and aggregate fineness on rutting rate. These plots were constructed assuming typical values for Superpave mix- tures for |η*|, aggregate specific surface and Ndesign − 5,000 Pa-s, 4.8 m2/kg and 75 gyrations, respectively. Figure 5 shows esti- mated rutting rate (mm/m/ESALs1/3) as a function of design VMA and design air void content for a constant in-place air void content of 7%.As VMA increases, rut resistance decreases; the estimated rutting rate decreases by about 20% for each 1% decrease in VMA. Each 1% increase in design air voids decreases rutting rate by 18%. This might at first seem counter- intuitive, but by increasing the design air void level while main- taining the in-place air void content, the energy of compaction required to construct the pavement is increased significantly. Conversely, decreasing air voids under constant in-place air voids decreases the energy required for field compaction. Figure 6 shows the effect of in-place air voids on rutting rate at a constant design air void content of 4%. Each 1% decrease in in-place air voids decreases the rutting rate by about 18%. Note that the magnitude of the effect of changes in design air void content and in-place air void content appear to be nearly identical. In fact, if in-place air void content is allowed to vary with design air voids (i.e., in-place air voids of 8% for 5% design air voids, in-place air voids of 6% for design air voids of 3%), the factors nearly offset each other and there is little net change in rut resistance. As will be emphasized repeatedly in this report, in order to develop efficient HMA mix designs 14 Factors Types/Levels Sections/mixtures 43 U.S. climates Southeastern (Alabama), Northcentral (Minnesota), Intermountain (Nevada) Aggregate type Andesite, granite, gravel, limestone, marble, schist, slag Mix design methods Superpave—29 (96 and 100 gyration), Marshall—12 (35, 50 and 75 blows), SMA—2 (50 gyration) Aggregate NMAS 36 mixtures 12.5-mm, 3 mixtures 9.5-mm, 4 mixtures 19.0-mm Aggregate gradation 22 mixtures ARZ, 17 BRZ, 2 TRZ, 2 SMA PG grades PG 58-28, PG 64-22, PG 67-22, PG 76-22 Modified/unmodified binders 15 mixtures modified, 28 unmodified Modifier types SBS, SBR FM300 (QC) Min. 21.6, Max. 42.8, Avg. 29.4 VMA (QC) Min. 10.9 %, Max. 16.3 %, Avg. 14.6 % VTM (QC) Min. 1.9 %, Max. 7.4 %, Avg. 3.7 % VTM (In-Place) Min. 3.3 %, Max. 8.2 %, Avg. 6.2 % Notes: SMA = stone matrix asphalt; ARZ = above restricted zone; BRZ = below restricted zone; TRZ = through restricted zone; SBS = styrene-butadiene-styrene rubber; SBR = styrene-butadiene rubber Table 2. Summary of factors and levels included in calibration of rutting model. 0.0 0.5 1.0 1.5 2.0 2.5 0 2,000 4,000 6,000 8,000 P1.08N0.65RD18.6 R u t R at e, m m /m /E SA LS 1/ 3 NCAT MN/Road WesTrack Figure 4. Relationship Between Field Rutting Rate and Proposed Function of Resistivity, Ndesign and Air Voids; the Heavy Center Line Represents the Rutting Rate Predicted Using Equation 2 While the Thinner Lines Represent 90% Prediction Limits for New Observations. 0.0 0.5 1.0 1.5 12 13 14 15 16 17 18 Design VMA, Vol. % R u t R at e, m m /m /E SA Ls (1/ 3) 3% design / 7% in-place 4% design / 7% in-place 5% design / 7% in-place Figure 5. Effect of Design VMA and Air Voids on Rut Resistance of Superpave Mixtures at a Constant In-Place Air Void Content of 7%.

in the laboratory and then to effectively control these mixtures in the field, it is essential to understand how changes in design air void content effect performance. If in-place air voids are assumed to be independent of design air voids, increasing design air void content will improve performance because greater compaction energy is required to reach the target value for in-place voids. Under these conditions, decreasing design air void content will reduce performance because less com- paction energy is then required to reach the target in-place air void level. However, if in-place air voids more or less follow changes in design air voids, there will be little effect on per- formance as a result of changing design air voids. It should be noted (as discussed later in this chapter) that changes in in- place air void level also significantly affect permeability. Engi- neers contemplating changes in design air void content should carefully and realistically consider the ways in which such changes will affect pavement performance. Figure 7 is similar to the previous two plots and shows the effect of VMA and aggregate fineness, as indicated by FM300. This value was allowed to vary from 20% to 40%, which is a typical range for Superpave mixtures based upon quality con- trol gradation data from the NCAT Test Track, the MnRoad Project, and the WesTrack project. The aggregate fineness has a very large effect on rut resistance; changing the value of FM300 from 20 to 30 decreases the rutting rate by more than a factor of 2; increasing FM300 from 30 to 40 further decreases the rutting rate by a factor of about 1.9. It can be concluded that aggregate fineness, as indicated by FM300, should be care- fully controlled in order to better design Superpave mixtures for specific levels of rut resistance. Because rut resistance depends on both VMA and aggregate specific surface (as indi- cated in this case by FM300), these factors should ideally be controlled simultaneously. As discussed later in this report, control of aggregate specific surface also helps to limit mix- ture permeability. The main practical problem is how to establish such control without being unduly restrictive in the requirements for VMA and aggregate gradation. In order to put the previous analysis into perspective, Fig- ure 8 was constructed, which shows the relationship between rutting rate, asphalt binder grade, and Ndesign. Binder PG grade, like FM300, is a very important factor in determining mixture rutting rate; in this analysis, increasing the binder grade from a PG 58-28 to a PG 64-22 increases the rutting rate by a factor of 2.6. Increasing the binder grade from a PG 64-22 to a PG 70-22 increases the allowable traffic by a factor of 2.4. The effect of compaction is not nearly as large as that of binder grade. Increasing Ndesign from 50 to 75 reduces the estimated rutting rate by 23%. Increasing Ndesign again from 75 to 100 fur- ther decreases rutting rate by 17%, while again increasing Ndesign from 100 to 125 decreases rutting rate by 14%. In summary, the effects of changing various aspects of mix- ture composition on rutting resistance (mm/m/ESALs1/3) are as follows: • Decrease VMA 1%, increase design VTM 1%, or decrease field VTM 1%: decrease rutting rate by about 20%. • Increase in aggregate fineness by 10 as indicated by FM300: decrease rutting rate by a factor of about 2. 15 0.0 0.5 1.0 1.5 12 13 14 15 16 17 18 Design VMA, Vol. % R ut R at e, m m /m /E SA Ls (1/ 3) 4% design / 8% in-place 4% design / 7% in-place 4% design / 6% in-place 0.0 0.5 1.0 1.5 12 13 14 15 16 17 18 Design VMA, Vol. % R ut R at e, m m /m /E SA Ls (1/ 3) FM-300 = 20 FM-300 = 30 FM-300 = 40 Figure 7. Effect of Aggregate Fineness and Design VMA on Rut Resistance of Superpave Mixtures at a Constant In-Place Air Void Content of 7%. Figure 6. Effect of VMA and In-Place Air Voids on Rut Resistance of Superpave Mixtures at a Con- stant Design Air Void Content of 4%. 0.0 0.5 1.0 1.5 25 50 75 100 125 150 Ndesign R u t R at e, m m /m /E SA Ls (1/ 3) PG 58-28 PG 64-22 PG 70-22 Figure 8. Effect of Binder Grade and Ndesign on Rut Resistance of Superpave Mixtures (Design Air Voids  4%, In-Place Air Voids  7%).

• Increase of one level in high-temperature PG-grade: decrease rutting rate by a factor of about 2.5. • Increase Ndesign by one level: decrease rutting rate by about 15% to 25%. Several comments should be made concerning this analy- sis. First, although the model used in this analysis was based on a substantial data set, further refinement of the model using an even wider range of data is needed before it can be used with confidence for a wide range of conditions. Of par- ticular concern are the specific effects of mineral filler and polymer modification on rut resistance; the data set used in calibration of the resistivity model included mixtures made using a large number of modified binders, but these mixtures also tended to be those with the highest specific surface. Therefore, there is some confounding of these effects. As part of NCHRP Project 9-33, the rutting/resistivity model is being re-evaluated and refined; preliminary results indicate that the general form of the model is correct, as are most of the trends predicted by the model. Specific exponents in the final model will be somewhat different from those given in Equation 2. For example, initial analysis suggests that the exponent to Ndesign in Equation 2 should be −.949 rather than −0.595. Most importantly, it appears that, all else being equal, many mixtures made using polymer modified binders will exhibit substantially better rut resistance than predicted on the basis of resistivity alone (32). It must be emphasized that the various factors affecting rut resistance are additive and that although some may seem rel- atively insignificant, if these act together in the same way the results can be quite large. Engineers contemplating modifica- tion in current Superpave requirements (or specifications for other HMA types) must consider not only the effect of a par- ticular change in a given characteristic, but also the combined effects of all other such changes. Although aggregate angularity and gradation do not appear in either the resistivity equation or the related equa- tion for rutting rate, Equation 2 does include terms for both design compaction level and field compaction, which is accounted for through relative density—that is, field density/ air voids compared with design density/air voids. As aggregate quality decreases—that is, as an aggregate becomes less angu- lar and/or cubical and/or resistant to crushing—Ndesign (i.e., gyrations required to reach 4% air voids) will decrease, which will cause the rutting rate estimated using Equation 2 to increase. Thus, the proposed approach for accounting for the affect of mixture composition on rut resistance indirectly includes the effect of aggregate angularity and gradation through inclusion of terms for laboratory and field com- paction effort. Because the proposed relationship for rut resistance was based on mixtures that were mostly made with cubical, well-crushed aggregates with little or no natural sand, extreme caution should be used in applying this model to mixtures containing poor quality aggregates. Aggregate gra- dations included in the mixtures upon which Equation 2 were based included mostly coarse gradations, with significant numbers of fine and dense gradations, and a few gap-graded materials (SMA mixtures). However, no open-graded mixtures were included in these data. Therefore, Equation 2 should also not be applied to open-graded friction course mixtures until its accuracy for such materials has been verified. A second important limitation to the proposed model for rut resistance involves the behavior of mixtures at very low air void contents. It is well known that at in-place air void contents below about 2% to 3%, many HMA pavements will exhibit a sudden and dramatic decrease in rut resistance. This is attributable to excessive asphalt binder content, which prevents aggregate particles from developing the internal friction needed for good rut resistance. For this rea- son, it is generally accepted that air void contents below about 3% should be avoided when designing HMA mixes. This phenomenon is not directly addressed in the resistivity equation (Equation 1) or the associated equation for rutting rate (Equation 2). Therefore, the proposed approach to accounting for the effect of mixture composition on rut resistance should not be applied to mixtures with very low air void contents. Based upon the range of air void contents included in this analysis, the proposed equations should not be applied to mixtures designed at air void contents below about 3%, or to field produced mixtures with air void con- tents in QC testing below about 2.5%, or to pavements with in-place air void contents below about 4%. This qualifica- tion does not mean that the model is not accurate for these conditions—only that its accuracy has not been evaluated for such circumstances. Although these caveats to the proposed rutting model are substantial, in essence the proposed model should be valid for mixtures meeting or nearly meeting current require- ments for Superpave mixtures, heavy-duty Marshall mix designs, and SMA mixtures. As discussed in Chapter 3, it appears that the overall level of rut resistance in the vast majority of HMA designed using the Superpave system is adequate. However, some agencies have noted a decrease in fatigue resistance and an increase in permeability with the widespread adoption of Superpave mix design require- ments, and some have increased minimum VMA require- ments to improve fatigue resistance of these materials. The findings above suggest that aggregate specific surface should be increased along with VMA in order to maintain good rut resistance. As discussed below, this will have the added ben- efit of helping to limit HMA permeability. This and other ramifications of the findings presented above are discussed in greater detail in Chapter 3. 16

Fatigue Resistance Continuum Damage Approach to Fatigue Phenomena in HMA During NCHRP Projects 9-25 and 9-31, a practical approach was developed by applying continuum damage the- ory to characterizing and analyzing the fatigue response of HMA. The following discussion is a summary of this method of analysis and is a relatively minor extension of previous work on continuum damage theory done by other researchers (16, 33–37). The following equation for fatigue life was derived based upon continuum damage theory and an expo- nential damage rate: (3) where N = fatigue cycles; α = a material constant for viscoelastic material; f = loading frequency, Hz; C = damage ratio (damaged/undamaged modulus); C2 = continuum damage fatigue constant; 0 = applied strain amplitude (1⁄2 of peak-to-peak strain); and |E*|LVE = linear viscoelastic (LVE) complex modulus. To apply Equation 3 (and related functions) to fatigue phe- nomena in HMA, the values of α and C2 must be known along with the values for frequency and modulus. Analysis of uniaxial fatigue data gathered during NCHRP Projects 9-25 and 9-31 lead to the following empirical equation for esti- mating C2 as a function of |E*|,VFA, and the rheological index of the binder, R: (4) The rheological index R of the binder is a constant in the Christensen–Anderson and Christensen–Anderson– Marasteanu (CAM) models for complex modulus and phase angle of asphalt binders (38, 39). This constant is directly related to the dispersion of relaxation times for the binder, and so it increases as the width of the relaxation spectrum increases. In the literature of paving technology the width of the relaxation spectrum is often referred to as “rheologic type” and was traditionally characterized by empirical con- stants such as the penetration index (PI) and penetration- viscosity number (PVN) (40). The value of R relates well to these older indexes but is more rational and more exact in the way it characterizes the flow properties of asphalt binders (38, 40). Values for R typically range from about 1.2 to more than 3.0, with a typical value for an unaged binder being about 2.0 C E VFA R LVE2 0 629 0 576 1 5430 1= − − − −. * . . . N f C C E LVE = −( ) −( ) − + 2 1 2 1 0 2 2 α α α α αα  * (38–40). Oxidation, both during refining and during age hardening, can dramatically increase the value of R (38, 40). The primary objective of NCHRP Projects 9-25 and 9-31 was to develop relationships among mixture compositional char- acteristics and HMA performance. Although R is not a mix- ture compositional characteristic, it must be included in fatigue equations to ensure that the models are valid and accurate. It is not recommended that the R value be con- trolled as part of the mix design process; control of R must be done through the binder specification and should be the topic of other research projects. However, it must be emphasized that none of the fatigue models presented here involve inter- actions between R and other mixture characteristics. There- fore, even though variations in R are not directly addressed in this report, the effects of changes in mixture composition on fatigue resistance are still accurate and valid. Another important finding in analyzing the NCHRP Proj- ects 9-25 and 9-31 fatigue data was that the complex modu- lus in tension/compression, as determined at the start of the uniaxial fatigue tests, was significantly lower than that meas- ured in dynamic compression. Based upon comparing the measurements made during these fatigue tests and dynamic compression values predicted using the Hirsch equation, the following empirical equation was developed for estimating tension/compression modulus values (|E*|TC) from |E*| values determined for dynamic compression: |E*|TC = 0.000209|E*|1.57 (5) It was found that tension/compression modulus values com- puted using this equation agree well with flexural stiffness values predicted with Bonnaure’s equation and as measured during SHRP. Figure 9 is a comparison of tension/compression modulus values predicted using the Hirsch equation and Equa- tion 5 with measured flexural stiffness values (|S*|) measured during SHRP as part of the SHRP fatigue tests (15). 17 R2 = 81% 5.0 5.5 6.0 6.5 5.0 5.5 6.0 6.5 Log Flexural |E*|, Measured, psi Lo g T/ C |E* |, P re d. , ps i Figure 9. Predicted Complex Modulus in Tension/Compression Compared with Measured Flexural Complex Modulus for SHRP Mixtures (Bars  d2s Confidence Limits, Solid Line  Regres- sion Function, and Dashed Line  Equality).

Equations 3 and 4 potentially can be used to evaluate the effect of changes in mixture composition on fatigue resistance. However, because the fatigue testing upon which Equation 4 was based was somewhat limited and because of the newness of continuum damage theory, further verification of these results is desirable prior to application to HMA mix design and analysis. This verification was done by applying continuum damage theory to flexural fatigue data gathered during SHRP (15). These data included the results of 185 tests on mixtures made using eight different asphalt binders and three aggre- gates. Properties of this data set are summarized in Table 3. Application of continuum damage theory to flexural fatigue data in part is dependent upon the finding that at completion of a flexural fatigue test, when the flexural stiffness has decreased by 50%, the extreme fiber damage will be constant at about 86.4%, corresponding to a damage ratio of 0.136. In analyzing the SHRP fatigue data, the data was first ana- lyzed statistically using a model of the following form: (6) In analysis of the SHRP data, the value of the complex modulus in compression was estimated from the Hirsch model and then converted to a tension/compression value using Equation 5. This approach was taken, rather than using measured flexural modulus values, because in the mix design logF R+ ( ) + error log log * log logN A B E D E VFA TC = + + ( ) + ( )0 process only estimated modulus values are available. In Equa- tion 6, the coefficients to the log of the predictor variables correspond to exponents for these terms; according to con- tinuum damage theory, the exponent for the strain term is α/2. The model represented by Equation 6 was reasonably accurate, with an r2 value of 84% (adjusted for degrees of free- dom) and gave a value for D = α/2 = 1.74, slightly lower than the value of 2.00 that was used in analyzing the uniaxial fatigue data generated during NCHRP Projects 9-25 and 9-31. The next step in analyzing the SHRP data was to calcu- late the value of C2 for each test, using a rearranged form of Equation 3 and a value for the terminal damage ratio C of 0.136 and α = 1.74: (7) Uniaxial fatigue test data collected during NCHRP Projects 9-25 and 9-31 were re-analyzed using α = 1.74 rather than α = 2.00 (as was done in the initial analysis of these data). Then, C2 values and related information for both the SHRP data and for the NCHRP Projects 9-25 and 9-31 data were combined and analyzed statistically. The best model for esti- mating C2 for this combined data set was somewhat different from that given earlier as Equation 4: (8)C VBE N Ddesign design relativ2 0 612 0 380466= − − −. . e LVER E − − −8 88 1 22 0 780. . .* − = −( )⎡ ⎣ ⎢⎢⎢ ⎤ ⎦ ⎥⎥⎥ − + C f C N E LVE 2 0 2 2 1 1 2 1α α α α α α  * 18 Property Average Value Minimum Maximum Total number of tests 200 Number of uniaxial tests (n = 4 analyzed together) 43 Number of replicated flexural tests (n = 2, analyzed separately) 61 Number of non-replicated flexural tests 35 Mix design methods Superpave, Marshall Aggregate types Greywacke gravel, low absorption limestone, limestone (2 sources), granite, gravel Binder types SHRP core asphalts—eight binders of widely varying rheology and grade; one SHRP asphalt modified with three modifiers; NCHRP 9-25/31 binders: PG 58-28, PG 64-22, PG 76-22, all unmodified Estimated compaction (Ndesign), gyrations 76 29 125 Air void content, Vol. % 5.1 0.8 8.8 Voids in mineral aggregate, Vol. % 16.5 11.5 21.5 Effective asphalt binder content, Vol. % 11.3 6.1 16.4 Voids filled with asphalt binder, % 69.2 42.4 94.3 Test temperature, °C 19 4 25 Test frequency, Hz 10 Applied strain, × 106 (uniaxial tests) 50 to 100 at 4 °C and 100 to 200 at 20 °C Applied strain, × 106 (flexural tests) 339 200 1,200 Initial |E*| uniaxial at 20 °C, uniaxial tests, GPa 5.76 1.24 9.52 Initial flexural stiffness, GPa 4.53 1.02 11.35 Cycles to failure (50 % stiffness lost) 119,000 10,000 685,000 Table 3. Summary of properties of data used in developing fatigue model.

where VBEdesign = the effective asphalt binder content at the design compaction level, in volume %; Drelative = the bulk density relative to the design bulk density; R = the rheological index of the binder; and |E*| = in lb/in2. The inclusion of Ndesign in this model significantly improved its accuracy, but was complicated because of the different compaction methods used in the two data sets—the SHRP mix designs were prepared using Marshall compaction, while the NCHRP Projects 9-25 and 9-31 mix designs were pre- pared using gyratory compaction. Two different levels of design compaction were used in the SHRP mix designs: 50 blows and 75 blows (15). In the model represented by Equation 8, the number of gyrations for each compaction level for the SHRP mix designs were included as predictor variables; the analysis indicated that the equivalent number of gyrations for 50-blow Marshall compaction was 73 and for 75-blow Marshall compaction was 92. Equation 8 differs from Equation 4 in the use of effective binder content rather than VFA. In analyzing the combined set of fatigue data, it was found that VFA could be used as an effective predictor if it is adjusted to 4% air voids. However, this is a cumbersome cal- culation and in fact provides essentially the same information as VBE. VBE has the additional advantage that it is nearly independent of changes in design air void content for the range from 3% to 5%. The model represented by Equation 8 was very effective, with an r2 value of 89% (adjusted for degrees of freedom). The results are shown graphically in Fig- ure 10, which shows predicted and observed values for C2 for both data sets. As a final check on the accuracy of this analysis, the cycles to failure for the SHRP flexural fatigue data were predicted using Equations 3 and 8, which can be combined to form a single function for predicting fatigue life to a given damage ratio C: (9) where the damage ratio at the end of the test (when the flex- ural stiffness falls to 50% of its initial value) is 0.136 and the modulus values are as predicted using the Hirsch model corrected for tension/compression loading using Equa- tion 5. Figure 11 shows the cycles to failure predicted using Equation 9 and as measured during SHRP. Each point repre- sents the average of two tests and includes d2s confidence lim- its, which represent a 95% prediction limit for the difference between two independent observations. If most of these con- fidence limits include the line of equality, it indicates that the predictions agree well with the experimental value. In this case, 57 of 61 data points agree to within the d2s limits, indi- cating exceptionally good agreement between the predicted values and the observed fatigue limits. If two independent sets of fatigue measurements were compared, it would be expected that 58 of the 61 data points would agree to within the d2s limits, so the predicted values appear to be almost interchangeable with experimentally determined values. Effect of Mixture Composition on In Situ Fatigue Resistance Some discussion of the relationships among fatigue resist- ance and binder content, design compaction level, and field compaction is useful at this point to illustrate the practical implications of Equation 9. Many pavement engineers and technicians assume that lower values of Ndesign automatically result in higher binder contents, so lowering Ndesign will improve fatigue resistance because increased binder content will improve fatigue life. This is true if Ndesign is changed N fVBE N Ddesign design rela = × −1 05 10 6 1 677 1 041. . . tive LVE R C E 24 34 3 335 1 74 0 3 48 1 3 1 2 . . . . . * − −( ) ( ) 42 19 0.00010 0.00100 0.00010 0.00100 Observed C2 Pr e di c te d C 2 SHRP Flexural Data NCHRP 9-25/31 Uniaxial Data Equality Figure 10. Predicted and Observed Values for Contin- uum Damage Fatigue Constant C2 for SHRP Flexural Fatigue Data and NCHRP Uniaxial Fatigue Data. R2 = 91% 3.5 4.0 4.5 5.0 5.5 6.0 3.5 4.0 4.5 5.0 5.5 6.0 Log Measured Cycles to Failure Lo g Pr e di ct ed N f Figure 11. Predicted and Measured Log Cycles to Failure for SHRP Flexural Fatigue Data, with d2s Confidence Limits.

without modifying the given aggregate blend, but it is not true when a range of aggregates and mixtures are considered. Fur- thermore, there is no reason to believe that if Ndesign require- ments are changed, materials suppliers will not change aggregate gradations for their mixtures. In fact, in cases where asphalt binder is not paid as a separate item, there is strong economic incentive to modify aggregate gradations to obtain the minimum binder content that will not incur a penalty. Because changes in volumetric requirements cannot possibly include a requirement that Ndesign should be changed without any modification in aggregate gradation or sources, there is no basis for suggesting that implementing lower Ndesign values will result in increased binder contents and improved fatigue resistance. Decreasing Ndesign may improve the ease with which a mixture can be compacted in the field, but this will not nec- essarily mean that field compaction will be improved since there is an economic incentive to compact a pavement only to the highest air void content that will not incur a significant penalty. In summary, all else being equal, increasing Ndesign will improve fatigue resistance, and decreasing it will do the oppo- site. If an agency feels that higher binder contents and lower in-place air voids are needed to improve fatigue resistance, higher minimum binder contents (higher minimum VMA at a given design air void level) and improved field compaction requirements should be specified, perhaps in combination with lower Ndesign values if it is felt that this latter change will help materials suppliers and contractors deal with the first two changes. Fatigue resistance in situ involves more than the inherent fatigue resistance of the mixture because mixture stiffness will affect the magnitude of strains resulting from traffic loading, in addition to affecting the resulting rate of damage as pre- dicted by Equation 9 and similar functions. Therefore, in order to evaluate the overall relationships among volumetric composition and fatigue resistance, a simplified evaluation of field fatigue resistance was performed. The general approach involved using the Illipave algorithms, as described by Huang (41), to estimate tensile strains at the bottom of the bound materials in various pavement structures; Equation 9 was then used to determine the pavement fatigue life using a ter- minal damage ratio of 0.20. A variety of climates, pavement structures, and mixture compositions were considered: • Two climates: New York state and South Carolina; • Two pavement structures: 100-mm bound material over 150-mm granular subbase and 200-mm bound material over 300-mm granular subbase; • Three average subgrade stiffness conditions: weak, moder- ate, and stiff; • Four different times of year: mid-winter (January), early spring (March/April), spring thaw (April/May), and late summer (August); • Two binder grades: PG 64-22 and PG 76-22; • Design VMA ranging from 13 to 16; • Ndesign of 75; and • Design air voids of 3%, 4%, and 5%. It should be pointed out that the level of Ndesign was not var- ied since Equation 9 (and any similar fatigue equation derived from this analysis) predicts that fatigue life will increase with increasing values of Ndesign and will decrease with lower values of Ndesign, all else being equal. There is no mechanism for changing this relationship in the field. Mixture modulus val- ues were estimated using the Hirsch model. Binder R values were 1.70 for the PG 64-22 and 2.17 for the PG 76-22; these and other binder properties were taken from actual materials tested in Advanced Asphalt Technologies’ laboratory. As dis- cussed above, the statistical analyses showed no interaction between mix composition and binder R value. Therefore, using typical values for R should not affect the sensitivity of this analysis to changes in mix composition. Although an analysis of the affect of changes in R on fatigue resistance might be enlightening, it is clearly outside the scope of NCHRP Projects 9-25 and 9-31 and is not included in this report. The subgrade modulus values were allowed to vary according to the time of year, using the same values incorpo- rated into the 1991 edition of the Asphalt Institute’s Mix Design Methods for Asphalt Concrete and Other Hot-Mix Types (42) for the thickness design of flexible pavements as reported by Huang (41). The results of this analysis were then compiled in terms of relative fatigue life—in this case, fatigue life expressed as a fraction of that for a design VMA of 15%, design air voids of 4%. These relative fatigue life values were then summarized statistically using means and standard deviations. Plots were prepared showing average changes and d2s confi- dence limits in relative fatigue life with design VMA, design air voids, and in-place air voids. The results are shown in Figures 12 through 15. In Figure 12, the effect of changes in design air 20 0.0 0.5 1.0 1.5 12 13 14 15 16 17 18 Design VMA, Vol. % R e la tiv e N f 5% design / 7% in-place 4% design / 7% in-place 3% design / 7% in-place Figure 12. Effect of Design Air Voids and Design VMA on Relative In-Situ Fatigue Life, In-Place Air Voids Constant at 7% (Errors Bars  2s Confidence Limits).

voids and design VMA are shown at a constant in-place air void level of 7%. For every 1% increase in VMA, the fatigue life increases from about 13% to 21% (typically about 16%). For every 1% increase in design air void content, the fatigue life increases about 7% to 14% (typically about 10%). This later finding may at first seem counter-intuitive, but it must be remembered that the analysis summarized in Figure 12 was generated assuming constant in-place air voids—therefore, increasing design air voids mostly has the effect of increasing the compaction effort during construction. For comparison, Figure 13 represents an analysis in which in-place air voids are allowed to vary with design air voids—that is, in-place air voids were assumed to be 6%, 7%, and 8% for design air voids of 3%, 4%, and 5%, respectively. In this case, the advantage of using higher air voids disappears—and, in fact, it appears to become a disadvantage in that it significantly decreases fatigue life. However, this is only because as design air voids change at constant VMA, it directly affects VBE—as air voids increase at constant VMA, VBE decreases. To illustrate the good relationship between effective binder content and fatigue resistance, Figure 14 was con- structed. This plot is nearly identical to Figure 13, but the horizontal axis is VBE rather than VMA. There is an excel- lent, nearly linear relationship between relative fatigue life and effective binder content—even though this figure was generated using different pavement structures, climates, and times of the year. For every 1% increase in VBE, there is typ- ically a 13% to 15% increase in relative fatigue life. There- fore, to control the fatigue resistance of HMA, VBE should be specified. Alternately, asphalt binder content by weight can be specified as a function of aggregate specific gravity, but this is a somewhat more cumbersome approach. Fur- thermore, it is clear that if in-place air voids are allowed to vary with design air voids, there is little net affect on fatigue resistance. As discussed previously, the same situation exists for rut resistance—that is, changing in-place air voids simultaneously with design air voids has little net effect on rut resistance. Although there appears to be some advantage to linking design and in-place air voids, such an approach would be impractical since in most paving projects the in-place air voids cannot be predicted with any certainty. Paving engi- neers and technicians should nevertheless understand the relationship among design air voids, in-place air voids, and performance: • At a constant level of in-place air voids, increasing design air voids will improve performance because it forces more compaction energy to be used during construction. • At a constant level of design air voids, increasing field air voids will decrease performance because it will result in less com- paction energy being used during construction (it will also increase the permeability of the pavement, potentially decreasing resistance to age hardening and moisture damage). • If design air voids and in-place air voids vary in a similar way, there will be little effect on performance. 21 0.0 0.5 1.0 1.5 12 13 14 15 16 17 18 Design VMA, Vol. % R e la tiv e N f 5% design / 8% in-place 4% design / 7% in-place 3% design / 6% in-place 0.0 0.5 1.0 1.5 7 8 9 10 11 12 13 14 15 16 Design VBE, Vol. % R e la tiv e N f 5% design / 8% in-place 4% design / 7% in-place 3% design / 6% in-place 0.0 1.0 2.0 3.0 4 6 8 10 12 14 In-Place Voids, Vol. % R e la tiv e N f 5% design voids 4% design voids 3% design voids Figure 15. Effect of In-Place Air Voids and Design Air Voids on Relative In-Situ Fatigue Life (Errors Bars  2s Confidence Limits). Figure 14. Effect of Design Air Voids and Design VBE on Relative In-Situ Fatigue Life, In-Place Air Voids of 6%, 7%, and 8% for Design Air Voids of 3%, 4%, and 5%, Respectively (Errors Bars  2s Confidence Limits). Figure 13. Effect of Design Air Voids and Design VMA on Relative In-Situ Fatigue Life, In-Place Air Voids of 6%, 7%, and 8% for Design Air Voids of 3%, 4%, and 5%, Respectively (Errors Bars  2s Confidence Limits).

Because of the strong relationship between VBE and fatigue resistance, VBE should be kept constant if design air voids are varied—that is, VMA should be increased or decreased in the same way as design air voids. This approach is, in fact, not at all new as it is the precise methodology suggested for Mar- shall mix designs in the Asphalt Institute’s Superpave Mix Design (SP-2 Manual) (43) where minimum VMA values increase 1% for each 1% increase in design air void content. For example, the minimum VMA value for a 9.5-mm NMAS aggregate blend is 14% for 3% design air voids, 15% for 4% design air voids, and 16% for 5% design air voids. The effec- tive asphalt binder content in each case is 11%. The last plot in this series is Figure 15, which summarizes the effects of design air voids and in-situ air voids simulta- neously. For every 1% increase in in-place air voids, relative fatigue life decreases by a nearly constant amount of about 22%. This means that an increase in in-place air voids of 2% will decrease fatigue resistance by nearly 50%. However, as mentioned above, this probably understates the importance of in-place air voids to fatigue life because it neglects the effect of changes in air voids on permeability and age hard- ening. This finding can be compared with those of Linden et al. cited earlier (24). Linden et al. cited three analytical studies in which a 10% to 30% reduction in fatigue life was predicted for every 1% increase in in-place air voids (25–27). This is in good agreement with the findings of NCHRP Projects 9-25 and 9-31. However, the rule of thumb of a 10% overall reduction in performance for every 1% increase in in-place air void content by Linden et al. is some- what lower than the figure found in this analysis, but con- sidering the very approximate nature of the research of Linden et al., the results should not be considered to con- tradict the findings of NCHRP Projects 9-25 and 9-31. Although an in-depth study of the effect of in-place air voids on pavement performance is outside the scope of this research, successful implementation of the results of this research will depend in part on achieving proper field compaction of mixtures designed according to the recom- mendations put forth in this report. In summary, the analysis presented above indicates several important relationships exist between the fatigue resistance of HMA mixtures and volumetric composition: • At given values for Ndesign, design air voids, and in-place air voids, fatigue resistance increases with increasing VBE. • At a given design values for VBE, design air voids, and in- place air voids, fatigue resistance will increase with increas- ing values of Ndesign. • At given design values for VBE, air void content, and Ndesign, fatigue resistance will increase with decreasing in-place air void content. Permeability and Age Hardening Permeability Tests As discussed earlier in this report, the permeability tests performed as part of NCHRP Projects 9-25 and 9-31 were unfortunately of limited value. This was for two reasons: (1) the air void content of the specimens was relatively low (typically from about 3% to about 7%), which, even in field specimens, would result in very low permeability values; and (2) the permeability of laboratory specimens is often much lower than that of field cores. Therefore, the permeability of most of the specimens fabricated during this research was so low as to be impractical or even impossible to measure. This does however lead to an important finding: permeability testing of laboratory-fabricated specimens is usually not effective because the permeability will be much lower than that of field specimens and will tend to be quite variable. For purpose of mix design and mix design selection, it is proba- bly more practical to rely upon models for estimating per- meability rather than measuring permeability in the laboratory, which perhaps might show fairly low permeabil- ity values for mixtures that might exhibit unacceptably high permeability in the field. Because of the shortcomings of the permeability tests per- formed during NCHRP Projects 9-25 and 9-31, use has been made of the substantial permeability data set generated dur- ing the Florida study reported on by Choubane et al. (3). Properties of this data set are summarized in Table 4. This study involved permeability testing of a large number of field cores and a limited number of laboratory-fabricated speci- mens. It should be pointed out that these pavements were constructed relatively early during the implementation of Superpave and their composition does not reflect that of Superpave pavements currently being constructed in Florida, which in general now have higher VMA and binder contents. A number of approaches were used to analyze these data sta- tistically to develop an accurate and useful function for pre- dicting the permeability of HMA. It was determined that the most effective approach was to use a relatively simple model in which permeability is proportional to effective air void content (VTMEff), which in turn is a function of total air void content and aggregate specific surface: (10) where (11) (12)V Sa0 1 53 1 87= −. . . VTM VTM VEff = − 0 and k VTMEff= 108 22

The r2 for this model, adjusted for degrees of freedom, was 65%. Although this does not appear to be an extremely strong correlation, the high variability in the permeability measure- ments must be considered when evaluating this model. Choubane et al. did not evaluate the repeatability of their measurements (3). However, as part of NCHRP Projects 9-25 and 9-31, an estimate of the standard deviation of these meas- urements was made by grouping specimens from the same project and same material and having air void contents within 1% of each other. Variance values were then estimated for each of these groups. An overall average variance was then calculated, weighted according to the number of specimens in each group. Because very low permeability values (below 50 × 10−5 cm/s) showed much lower variability than the other measurements, these were eliminated from the calculation. The estimated variances for the remaining permeability measurements appeared to fall into a similar range: the pooled estimate of the standard deviation using this method of 150 × 10−5 cm/s and incorporating 80 different measure- ments. The large number of measurements incorporated into this estimate means that it should be quite reliable. Figure 16 is a plot of measured permeability versus effec- tive air void content for several sets of data. This plot includes data for the Florida field cores, Florida laboratory specimens, and NCHRP Projects 9-25 and 9-31 laboratory specimens. The plot includes the 95% prediction interval for new obser- vations for the Florida field cores. The width of this predic- tion interval—about 310 × 10−5 cm/s—is very nearly equal to twice the estimated standard deviation of 150 × 10−5 cm/s. It therefore appears most of the scatter observed in Figure 16 is probably the result of experimental error rather than of lack of fit in the model and that the proposed method of estimat- ing mixture permeability is significantly more accurate than indicated by the r2 value of 65%. One of the problems associated with permeability testing of HMA is that permeability values measured on field cores generally are much higher than comparable specimens prepared in the laboratory. In Figure 16, the permeability for the laboratory specimens is about 1/6 of the value for similar field cores. The very low permeability of laboratory prepared specimens suggests that testing such specimens is probably not useful since it will almost always show very low or zero permeability and when it does not, the results are likely to be highly variable. Instead, Equations 10 through 12 should be used to estimate in-place permeability, based upon aggregate specific surface and measured or anticipated in-place air void content. In specifying Superpave and other HMA types, rea- sonably low levels of permeability should be maintained to help prevent excessive age hardening and to reduce suscepti- bility to moisture damage. In order to achieve such control, aggregate specific surface and in-place air void content must be simultaneously controlled. As with other aspects of con- trolling Superpave volumetric composition, a critical issue becomes how specifically to exert such control. Florida researchers have suggested that Superpave surface-course mixtures should exhibit permeability values below 100 × 10−5 cm/s. For an in-place air void content of 7%, this corresponds to an FM300 value of 26%. However, minimum FM300 values 23 0 200 400 600 800 1000 1200 0.0 2.0 4.0 6.0 8.0 10.0 Effective Air Voids, Vol. % k x 10 5 , c m /s 19 mm Low Sa 19 mm High Sa 12.5 mm Low Sa 12.5 mm High Sa Fit 95 % PI Figure 16. Permeability of Specimens Tested During the Florida Study and During NCHRP Projects 9-25 and 9-31 as a Function of Effective Air Void Content (Solid Line  Regression Line for Predicted Perme- ability; Dashed Lines  95% Prediction Interval). Property Average Value Minimum Maximum Total number of tests 113 Total number of field projects 7 Total number of mixtures 13 Aggregate types Alabama limestone, Florida limestone, Georgia granite, RAP Aggregate NMAS and gradation 12.5-mm and 19-mm, all BRZ Binder grade, type PG 67-22, unmodified Estimated aggregate specific surface, m2/kg 4.47 3.57 5.34 Air void content, Vol. % 8.1 3.7 14.6 Voids in mineral aggregate, Vol. % 17.7 13.2 23.7 Effective asphalt binder content, Vol. % 9.6 8.5 10.7 Voids filled with asphalt binder, % 54.9 38.5 73.3 Permeability, × 10-5 cm/s 344 5 1014 Table 4. Summary of properties of Florida permeability study data set.

should vary both with air void content and with application— that is, mixtures in protected layers of the pavement can have higher permeability values. Specific guidelines for minimum FM300 values are given in Chapter 3. Age-Hardening Tests As part of NCHRP Projects 9-31 and 9-25, a variety of mix- tures were subjected to long-term oven conditioning and the extent of the resulting age hardening was measured using the field-shear test to measure the complex modulus before and after conditioning. It was expected that the results of this experiment could be related to the permeability of the mix- tures. The results in part did confirm a relationship between permeability and age hardening, in that the amount of age hardening clearly increased with increasing air voids. How- ever, equally clear was that the extent of age hardening also depended strongly on the specific aggregate and binder used in a mixture. The age-hardening data was analyzed using a multiple regression model with indicator variables to account for the effects of aggregate/binder combinations and using air void content as a covariate: (13) where AHRi = age-hardening ratio for ith observation; β0 = intercept (average response for aggregate/binder “0” [Virginia limestone and PG 64-22 binder]); β1 = average effect for aggregate/binder “1” (Virginia limestone and PG 58-28); Xi1 = indicator variable for aggregate/binder “1” and = 1 for aggregate/binder “1” and 0 otherwise; AHR X X X X X X i i i i i i i = + + + + + + β β β β β β β 0 1 1 2 2 3 3 4 4 5 5 6 6 + +β7VTMi i β2 = average effect for aggregate/binder “2” (Virginia limestone and PG 76-16); Xi2 = indicator variable for aggregate/binder “2” and = 1 for aggregate/binder “2” and 0 otherwise; β3 = average effect for aggregate/binder “3”(Pennsylva- nia gravel and PG 64-22); Xi3 = indicator variable for aggregate/binder “3” and = 1 for aggregate/binder “3” and 0 otherwise; β4 = average effect for aggregate binder “2” (Kentucky limestone and PG 64-22); Xi4 = indicator variable for aggregate/binder “4” and = 1 for aggregate/binder “4” and 0 otherwise; β5 = average effect for aggregate binder “5” (California granite and PG 64-22); Xi5 = indicator variable for aggregate/binder “5” and = 1 for aggregate/binder “5” and 0 otherwise; β6 = average effect for aggregate binder “6” (California granite and PG 58-28); Xi6 = indicator variable for aggregate/binder “6” and = 1 for aggregate/binder “6” and 0 otherwise; β7 = coefficient for effect of air void content (VTMi) on age-hardening ratio; and i = error term for ith observation. The r2 value for this model was 90.5%. All coefficients were highly significant, with the exception of the coefficient for the indicator variable for the Virginia limestone/PG 76-16 binder combination. Figure 17 shows the average effect of different aggregate/binder combinations on age hardening—that is, on the vertical axis are the values of the constant β0 and the coeffi- cients βi1 through βi5. The differences are significant and cannot be easily interpreted in terms of mineralogy or binder grade. Figure 18 shows the effect of air void content on age hardening, after removing the effect of different aggregate/binder combi- nations. As an example of this adjustment, consider mixtures 24 0.8 1.0 1.2 1.4 1.6 VA Li me sto ne /PG 64 VA Li me sto ne /PG 58 VA Lim es ton e/P G 7 6 PA Gr av el/P G 6 4 KY Li me sto ne /PG 64 CA G ra nite /PG 64 CA G ra nite /PG 58 A ge H ar de n in g R at io Figure 17. Average Age Hardening for Various Mixtures Subjected to Long-Term Oven Conditioning, as Calculated from Dynamic Modulus at 25 C and 5 Hz Using the Field Shear Test (Error Bars are for Bonfer- roni Joint 95% Confidence Intervals).

made using the California granite and the PG 64-22 binder. The average age-hardening ratio for all such mixtures was 1.28, while the average age-hardening ratio for the Virginia limestone/ PG 64-22 binder (the “0” aggregate) was 1.00 (see Figure 17). Therefore, the California granite/PG 64-22 mixtures had an average effect on age-hardening ratio of +0.28. To remove this effect from the data plotted in Figure 18, 0.28 was subtracted from the observed age-hardening ratios for all California granite/PG 64-22 mixtures. Figure 18 then shows the effect of air voids along with all errors. The effect of increased air void content on age hardening is significant (p<0.001), but the effect of different aggregate/binder combinations appears to be stronger than the effect of air void content. It can be concluded that control of air voids in HMA can only partially control the extent of age hardening in flexible pavements. This might mean, for example, that surface cracking in some mixtures might be the result of a particular combination of aggregate and asphalt binder being especially susceptible to age harden- ing and not necessarily the result of an inappropriate mix design or poor construction. Additional research is needed to better understand the relationship between aggregate mineral- ogy, asphalt-binder chemistry, and age hardening of HMA. In study of mixtures prone to surface cracking, evaluation of the age-hardening resistance of specific aggregate/binder combi- nations should be considered along with other tests. Some additional comments on Figures 17 and 18 are war- ranted. It appears from examining this plot that the amount of age hardening increases more rapidly at higher air void contents than at lower. However, the amount of variability makes such a hypothesis difficult to evaluate with certainty. Increased age hardening at air void contents above 4% is con- sistent with the concept of effective air voids—that is, that permeability of HMA is effectively zero below a certain air void content, which varies from mixture to mixture. An attempt was made to relate the age hardening of the different mixtures to the estimated zero permeability air voids content (related to aggregate fineness), but no such relationship was apparent. This should not be taken as definitive proof that such a relationship does not exist, only that it could not be statistically detected in this particular experiment. Effect of Mixture Composition on Age Hardening The only model identified in the literature review for esti- mating the effect of mixture volumetrics on age hardening is the global aging system developed by Mirza and Witczak (22). Unfortunately, this model makes use of traditional measurements such as penetration, softening-point temper- ature, and capillary viscosity, which are then converted to apparent viscosity values. This makes the model difficult to apply in a meaningful way to the Superpave system. Further- more, the age hardening is predicted only in terms of age- hardening ratios and not in terms of binder master curve parameters, which means that the global aging system is also difficult to apply in developing models and plots to illustrate the effect of changes in mixture composition on age harden- ing. For these reasons, a modification of Mirza and Witczak’s global aging system was developed, which provides results very similar with the original system but makes use of rational rheological measurements and binder master curve parameters. The modified global aging system was used to analyze several hypothetical situations to evaluate the effect of air voids and aggregate specific surface on age hardening. Age- hardening ratios were predicted at an age of 60 months for MAAT values of 7.2 °C, 15.6 °C, and 23.9 °C. In-place air void contents were assumed to be 5%, 7%, and 9%, while assumed values for FM300 were 20, 30, and 40. Age-hardening ratios were calculated for both mixture complex modulus (|E*|) at 10 Hz and binder steady-state viscosity at temper- atures of 0, 25, 40 and 60 °C. The analysis was performed for the PG 58-28 and PG 76-16 binders used in various other parts of NCHRP Projects 9-25 and 9-31. The age hardening for binder viscosity was estimated because high binder viscosities could contribute significantly to pave- ment distress by preventing healing of surface cracks dur- ing hot weather. Two examples of this analysis are shown in Figures 19 and 20. Figure 19 shows mixture age hardening at 25 °C and 10 Hz for the PG 58-28 binder for a MAAT of 15.6 °C. Figure 20 shows binder age hardening at 40 °C, also for a MAAT of 15.6 °C. Estimated age-hardening ratios, as should be expected, increase dramatically with increasing MAAT. The mixture age-hardening ratios were generally highest for the PG 58-28 binder at “test” temperatures of 25 °C and/or 40 °C; age-hardening ratios for binder viscos- ity decrease with increasing “test” temperature. Also, the age-hardening ratios for the PG 58-28 binder were always higher than for the PG 76-16 binder. Several important, 25 0.8 0.9 1.0 1.1 1.2 1.3 1.4 0.0 2.0 4.0 6.0 8.0 10.0 Air Voids, Vol. % A dju st ed A .H . R at io VA Limestone/PG 64 VA Limestone/PG 58 VA Limestone/PG 76 PA Gravel/PG64 KY Limestone/PG 64 CA Granite/PG 64 CA Granite/PG 58 Figure 18. Age-Hardening Ratio after Removing Aggregate/Binder Effect as a Function of Air Void Content.

practical findings can be made based upon the results of this analysis: • Mixture age hardening as indicated by complex modulus increases with increasing air voids and decreasing aggre- gate specific surface. This effect is not extremely large— typically, age-hardening ratios decrease 2% to 7% for each 1% increase in FM300 and increase 5% to 14% for each 1% increase in field air voids at a MAAT of 15.6 °C. However, the combined effect of high air voids and low aggregate specific surface can increase age hardening by 50% or more. The amount of age hardening that occurs in a mix- ture not only is dependent upon the air voids and aggregate fineness, but also is strongly dependent upon the specific binder used and the MAAT. • Age hardening as indicated by binder viscosity values can be extremely high—often greater than 100. The effect of increasing air voids by 2% is to increase age hardening by about a factor of 2 at a MAAT of 15.6 °C and by a factor of about 3 at a MAAT of 23.9 °C. The very high binder vis- cosities that can potentially exist in aged pavements could contribute significantly to surface cracking by preventing any healing from occurring at the pavement surface during hot summer weather. • In general, the effect of increasing air voids by 2% on age hardening is comparable with the effect of decreasing FM300 by 5%. More careful control of aggregate specific sur- face should help maintain good resistance to age harden- ing in HMA. Apparent Film Thickness and HMA Performance One of the objectives of this project was to evaluate the relationship between film thickness and HMA performance. Since the 1950s, some pavement engineers have proposed that film thickness is an important characteristic in determining the durability and fatigue resistance (44–48). Film thickness is generally estimated by dividing the effective volume of binder in a mix (in units of m3/kg aggregate) by the specific surface of the aggregate (m2/kg). In the late 1950s Campen and his associates proposed that HMA mixes should be designed with film thickness values between 6 and 8 μm (44, 45). Much later, Kandhal and Chakraborty suggested that film thickness values between 9 and 10 μm should be used with mixes designed according to the Superpave system in order to prevent premature aging (48). Despite the many proponents of film thickness, its use to design or specify HMA mixes remains controversial. The Superpave system does not include any requirements or guidelines for film thickness. Many pavement engineers object to the term “film thickness” on the grounds that individual asphalt films do not exist in an HMA mix and that instead asphalt is a continuous phase in what is in reality a particulate composite. Although this latter view is technically correct, the fact remains that film thickness values can be calculated for HMA mixes and these values relate two important character- istics of HMA mixes—asphalt binder content and aggregate specific surface. To address the objection that asphalt films do not really exist in asphalt mixes, the term “apparent film thick- ness” (AFT) is in general used throughout this report. In general, the research performed during this project does not support the direct use of AFT values in the design and specification of HMA mixes. At the same time, it should be pointed out that this research has demonstrated that AFT in many cases will indirectly relate to HMA performance. The strongest of such relationships is that between rut resistance and AFT. Resistivity is proportional to the square of aggregate specific surface and is inversely proportional to the cube of VMA, and because most Superpave mixes are designed at or very close to 4.0% air voids, there is a direct relationship between VMA and effective binder content. Therefore, there should be a very good relationship between AFT and resistiv- ity and between AFT and rut resistance. To evaluate this rela- 26 0.0 2.0 4.0 6.0 8.0 10 20 30 40 50 60 FM300 A gi ng R a tio (5 yr /In i.) 9% in-place voids 7% in-place voids 5% in-place voids Figure 19. Predicted Mixture Age-Hardening Ratio at 25 C and 10 Hz as a Function of In- Place Air Void Content and FM300 for a MAAT of 15.6 C. 1 10 100 1,000 10,000 10 20 30 40 50 60 FM300 A gi ng R a tio (5 yr /In i.) 9% in-place voids 7% in-place voids 5% in-place voids Figure 20. Predicted Binder Age-Hardening Ratio at 40 C as a Function of In-Place Air Void Content and FM300 for a MAAT of 15.6 C.

tionship, Figure 21 shows rutting rate as a function of AFT for data from the NCAT test track, MnRoad, and WesTrack (this plot can be compared with Figure 3). The relationship is only moderately strong, but rutting rate clearly increases with increasing AFT film thickness. This plot would suggest that HMA mixes with AFT values greater than 9 μm may be prone to excessive rutting. The relationships between AFT and other aspects of HMA performance are not as straightforward. Fatigue resistance increases with increasing effective binder content; therefore, if aggregate specific surface is kept constant, fatigue resistance will increase with increasing AFT. Permeability decreases with increasing specific surface and decreasing in-place voids. Therefore, at a constant value of in-place air voids, perme- ability will decrease with decreasing AFT. However, as asphalt binder content is reduced for a given HMA mix, AFT will decrease and the mix will become more difficult to compact, potentially leading to higher in-place voids and greater per- meability. Mixes with higher binder contents will be easier to compact and so may often exhibit lower in-place air voids and lower permeability. This phenomenon is possibly the source of the proposed relationship between AFT and durability. In summary, the results of this study suggest that AFT should relate to most aspects of HMA performance. All else being equal, rut resistance will in general increase with decreasing AFT. Other relationships between AFT and per- formance are indirect; therefore, the use of AFT for specifying and/or controlling HMA mixes is not recommended. Instead, such control should be exerted through the relationships pre- sented previously, linking various aspects of HMA composi- tion to rut resistance, fatigue resistance, and permeability. Summary During NCHRP Projects 9-25 and 9-31, laboratory tests were conducted to evaluate the effect of changes in VMA, air void content, VFA, aggregate specific surface, and related factors on various performance-related properties of HMA. These data, along with several data sets in the literature, were used to develop semi-empirical models for estimating rut resistance, fatigue resistance, and mixture permeability. Mirza and Witczak’s global aging system was modified to provide a more rational model for predicting age hardening, consistent with both the Christensen–Anderson model for binder mod- ulus and the newly developed Hirsch model for estimating the modulus of HMA. The following important findings were made based upon these tests and analyses: • It appears reasonable to allow design air voids for Superpave mixtures to vary within the range from about 3% to 5%. However, engineers and technicians that wish to deviate from the current design air void level of 4.0% should understand how such changes can affect HMA performance. • A variety of models for relating mixture volumetric com- position to performance were identified in the literature; however, these models are not well suited for evaluating the effect of mixture composition on performance for the Superpave system of mixture design and analysis. There- fore, models have been developed (or existing models refined) during NCHRP Projects 9-25 and 9-31 for esti- mating mixture performance on the basis of volumetric composition. • Many state highway agencies have modified the require- ments for VMA, air voids, and related factors for Super- pave mixtures. Three modifications are most common: (1) an expansion of the design air void content from 4% to a range of 3% to 5%; (2) establishing a maximum VMA value at 1.5% to 2.0% above the minimum value; and (3) a slight increase in the minimum VMA values, typically by about 0.5%. • Aggregate specific surface is very nearly proportional to the sum of the weight percent material passing the 75, 150, and 300 m sieves. This factor, called the fineness modulus, 300 μm basis (FM300), can be used to control aggregate specific surface in mixtures made using the Superpave system to ensure adequate mixture performance and good workability. • Rut resistance as indicated by laboratory tests and as measured in a wide range of field test tracks/test roads was predicted to within about a factor of 2 using a model incorporating mixture resistivity, design compaction, and relative field compaction. • The rutting/resistivity model suggests that each 1% decrease in VMA, 1% increase in design air voids, and/or 1% decrease in field air voids increases rut resistance by about 20%,as indicated by rutting rate in mm/m/ESALs1/3. • Increasing FM300 by 6% (at constant VMA) typically increases rut resistance by about a factor of 2.0 to 2.5. 27 R2 = 69% 0.000 0.050 0.100 0.150 0.200 0.250 4.0 6.0 8.0 10.0 12.0 14.0 Apparent Film Thickness, microns Ru tt in g Ra te , m m /m /E SA Ls 1/ 3 NCAT MN/Road WesTrack Fit Figure 21. Relationship Between Rutting Rate and Apparent Film Thickness.

• For the types of HMA used in NCHRP Projects 9-25 and 9-31—that is, mixtures made using good-quality, highly angular aggregates with little or no natural sand— increasing the high temperature binder grade one level will increase rut resistance by about a factor of 2.5, as indicated by rutting rate in mm/m/ESALs1/3. For HMA designed according to current Superpave requirements, binder grade appears to be the most important considera- tion in determining rut resistance of HMA; volumetrics are an important but secondary factor. It must be emphasized that replacing the good-quality aggregates normally used in Superpave mixes with poorly crushed gravel and/or large amounts of natural sand would almost certainly cause a substantial decrease in rut resistance and might also result in mixtures much more sensitive to changes in volu- metric composition. • Increase in Ndesign by one level decreased rut resistance by about 15% to 25%. • A practical approach to fatigue analysis of HMA based on continuum damage theory was developed during NCHRP Projects 9-25 and 9-31. This technique was ini- tially developed through analysis of laboratory test data collected during NCHRP Projects 9-25 and 9-31 and then verified and refined through successful application to flex- ural fatigue data gathered during SHRP at the University of California at Berkeley. • Fatigue resistance is affected by VBE, design compaction (Ndesign), and field compaction, expressed in terms of field density relative to laboratory/design density. Every 1% increase in VBE increases fatigue life by about 13% to 15%. Every 1% increase in field air void content (at a constant design air void content) decreases fatigue resistance by about 20%. • Permeability of HMA increases with increasing air voids and decreasing aggregate specific surface. Perme- ability can be effectively modeled using the concept of effective air voids—the total air void content minus the air void content at zero permeability. Furthermore, the zero air voids content increases with increasing aggre- gate fineness. • A simple, reasonably accurate equation has been devel- oped based upon permeability data gathered by Choubane et al. in a study on the permeability of Super- pave mixtures in Florida (3). According to this model, per- meability increases by about 100 × 10−5 cm/s for every 1% increase in air voids or 3% decrease in FM300 for air void contents above the zero-permeability limit. • The permeability of HMA specimens prepared in the lab- oratory tends to be significantly lower than permeability values measured on field cores of comparable mixtures. For this reason and because of the highly variable nature of permeability measurements, laboratory measurements of mixture permeability are not recommended for use in rou- tine mixture design. However, the effect of air void content and aggregate fineness on permeability should be consid- ered during the mix design process. • The age hardening of the HMA studied depended not only upon air void content, but also upon the specific combina- tion of aggregate and asphalt binder. Additional research is needed to better understand the effect of aggregate–asphalt binder combinations on mixture age hardening. • A modified version of the Mirza–Witczak global aging system was used to examine the effects of air voids, aggre- gate fineness, and other factors on mixture and binder age hardening. For a MAAT of 15.6 °C, the mixture age- hardening ratio decreased about 2% to 7% for every 1% increase in FM300. The age-hardening ratio increased about 5% to 14% for every 1% increase in in-place air voids. Although not extremely large effects, considered over the possible range for FM300 and field air voids, these factors can significantly affect mixture age hardening. • The modified global aging system predicted extreme amounts of age hardening as indicated by binder viscos- ity. These extreme age-hardening ratios are the result of changes in binder rheology that occur during the aging process and could significantly affect mixture performance because of the severe reduction in healing rates that might occur with such large increases in binder viscosity. Addi- tional research is needed to better understand the relation- ship among age hardening, binder viscosity, healing, and fatigue cracking in HMA pavements. • The various models developed during this study suggest that several indirect relationships exist between AFT and various aspects of HMA performance. The most signifi- cant of these is between AFT and rut resistance—as AFT increases, rut resistance decreases. Mixtures with AFT values above 9 μm may be prone to excessive rutting. However, because the relationships between AFT and per- formance are indirect, it is not recommended that AFT be used in specifying or controlling HMA mixtures. 28

Next: Chapter 3 - Interpretation, Appraisal, and Applications »
Volumetric Requirements for Superpave Mix Design Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

TRB's National Cooperative Highway Research Program (NCHRP) Report 567: Volumetric Requirements for Superpave Mix Design examines whether changes to the recommended Superpave mix design criteria for voids in mineral aggregate, voids filled with asphalt, and air voids content might further enhance the performance and durability of hot-mix asphalt.

  1. ×

    Welcome to OpenBook!

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

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

    No Thanks Take a Tour »
  2. ×

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

    « Back Next »
  3. ×

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

    « Back Next »
  4. ×

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

    « Back Next »
  5. ×

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

    « Back Next »
  6. ×

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

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

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

    « Back Next »
Stay Connected!