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.
6 This chapter summarizes the findings of the literature review regarding lift thickness, mix type and size, and pavement performance. It opens with a discussion of the basics of asphalt mix compaction and the factors that influence the level of compaction. This is followed by a brief review of the evolution of the Superpave mix design system. With that background, the topic turns to the relationships among lift thickness, mix size, and air voids (or mat den- sity). The effects of different types of mixtures (fine graded, coarse graded, stone matrix asphalt [SMA], and open graded) and sizes of mixtures (nominal maximum aggregate size, or NMAS) are explored to the extent presented in the literature. The fourth section presents a review of the reported performance of different NMAS mixtures when compacted in different lift thicknesses. The influence of other features of the mixtures, most notably the type of gradation (fine versus coarse), on pavement performance is also explored. Finally, the cost implications of increasing lift thicknesses, the performance benefits, and possible ways to counter cost increases are recapped. A few notes on terminology are offered. The prevalent practice for specifying the density of an asphalt mixture in the United States is in terms of the percent of the mixture maximum theoreti- cal specific gravity, Gmm. Gmm represents the specific gravity of a voidless mix, that is, if all of the air voids could be squeezed out. The difference between 100% and Gmm equals the air voids in a compacted mixture. Thus, density and air voids can be considered interchangeable; 91% density (Gmm) equals 9% air voids. Also, the NMAS is defined in Superpave as âone size larger than the first sieve that retains more than 10 percent aggregate,â and the maximum aggregate size (MAS) is defined as âone size larger than the nominal maximum aggregate sizeâ (AASHTO M 323-17) in a standardized nest of sieves. Most departments of transportation (DOTs) now use the NMAS to designate the size of Superpave mixtures. Thus, a 12.5-mm mixture has a NMAS of 12.5 mm and MAS of 19.0 mm. Since it is now standard to refer to a mix based on its NMAS, the terms âmix sizeâ and âaggregate sizeâ are synonymous. 2.1 Basics of Compaction Compaction is the process of densifying an asphalt mixture by reducing the air void content (Leiva and West, 2008). Adequate compaction of asphalt mixtures during construction is widely recognized as one of the main factors affecting the ultimate performance of a pavement because it influences smoothness, strength, and skid resistanceâthe main factors of interest to drivers and engineers (Hughes, 1984). Achieving an acceptable level of compaction prevents densification by traffic leading to per- manent deformation (rutting), strengthens the pavement, helps to âwaterproofâ the pavement, and reduces the air void content to slow binder oxidation which leads to raveling and crack- ing (Brown, 1982). Overcompaction, however, can result in bleeding, shoving, and rutting C H A P T E R 2 Literature Review: Effects of Lift Thickness on Pavement Performance
Literature Review: Effects of Lift Thickness on Pavement Performance 7 (Williams et al., 2015). Hughes (1984) indicated that optimum in-place density corresponds to 3% to 5% air voids; below that the mix can be unstable and above there can be durability issues. Brown (1982) agreed, noting that rolling to densities higher than 95% to 97% can cause hairline, longitudinal cracking when the mix tries to move laterally. Compaction is provided first by the paver screed, then by compaction rollers. Geller (1984a) explained that compaction occurs through âthe application of pressure over [the] contact area.â As the mix becomes more compacted or cools, it begins to resist further compaction and the contact area decreases. This results in an increase in the contact pressure. Once the resistance to compaction is equal to or greater than the contact pressure of the roller, no more densification will occur. There are three main types of rollers: static steel wheel, pneumatic, and vibratory steel wheel rollers. Static rollers with two or three wheels were the most common before the 1950s; pneumatics gained popularity in the 1950s and 1960s; and vibratories were developed in the 1960s and have become very popular (Geller, 1984a). Geller (1984b) noted that there were three changes in the 1960s that put more pressure on roller operators: the move from method specs to end result specs, statistical sampling for quality control, and the development of nuclear gauges. The use of rubber-tired rollers is one way to reduce permeability by âsealingâ the surface (Hughes and Maupin, 1986; Retzer, 2008). With vibratory rollers, the speed must be matched to the frequency of vibration (Nittinger, 1977). If the speed is too great relative to the frequency, there will be too few impacts applied in the longitudinal direction, resulting in ripples in the surface. Thicker lifts require more force (amplitude) to compact, but too much force on a thin lift can result in the roller rebounding and bouncing as shock waves reflect off the underlying layer. A slow roller speed generally results in higher density but at the cost of reduced productivity and difficulties in keeping up with the paver. Choubane et al. (2006) cautioned that, to achieve adequate density in coarse mixes, high numbers of vibratory roller passes are sometimes employed, which can result in aggregate break- down. Brown (1982) commented that Corps of Engineers experience shows that vibratory steel wheel rollers can obtain good densities, but over-rolling can reduce the density. In an article discussing factors that may contribute to the tenderness observed in some Superpave mixtures, Prather (2000) outlined five factors affecting asphalt mixture compaction: physical properties of the aggregates and binders, lift thickness, weather conditions during com- paction, support conditions under the lift being compacted, and rollers and their operation. Leiva and West (2008) observed that the most important factors are temperatures, aggregate properties and gradation, binder type and content, compaction operations, and lift thickness. Decker (2006) noted that the effects of lift thickness, mix properties, and environmental condi- tions are even more critical when attempting to compact mixes at cold temperatures (base or air temperature <50Â°F [10Â°C]). Gudimettla et al. (2003) demonstrated in the laboratory that work- ability increases as mix temperature increases. Brown (1982) opined that achieving adequate density depends on a mix design that allows the mix to be âreadily compacted.â He noted that too high a fines content (percent passing the 75 Âµm [#200] sieve > 6% to 7%) stiffens a mix and makes it harder to compact, but an increase of 0.2% to 0.3% binder makes a mix much easier to compact. Leiva and West (2008) introduced the concept of accumulated compaction pressure (ACP) as an indicator of mix compactibility. ACP is a function of roller type, number of passes, and compaction pressure, which equals the gross weight of a static roller divided by the contact area of the drums (total width of the drums times the contact arc). The gross weight of vibratory roll- ers is increased by adding the centrifugal force from the vibration. The tire pressure of a pneu- matic roller is the compaction pressure. In analyzing data from the National Center for Asphalt
8 Impact of Asphalt Thickness on Pavement Quality Technology (NCAT) Test Track, they found that it took more energy (ACP) to compact coarse mixes than fine mixes and to compact thinner lifts (t/NMAS of about 2 to 4 with an average of 3 compared to 3 to 7 with an average of 4.3). Analysis of variance (ANOVA) showed that t/NMAS was the most significant factor affecting ACP followed by temperature. Lift thickness, or the ratio of lift thickness to NMAS (t/NMAS), is widely recognized as a major contributor to the ability to achieve density. In general, thicker lifts are easier to compact; however, if lift thickness is too great, it will not be possible to compact the entire lift (Buchanan et al., 2004). The bottom of the lift may be outside the zone of influence of the roller and will be undercompacted. Thicker lifts provide more room for aggregate particles to be reoriented and retain heat longer, both of which aid compaction (Hughes, 1978). Through field and laboratory measure- ments, Chang et al. (2009) showed that thinner lifts cool faster than thicker lifts up to a point; increasing the lift thickness from 10 to 15 cm had somewhat less effect on time to cool to 80Â°C, the supposed cessation time, than increasing from 5 to 10 cm. Brown et al. (2005) noted that a 25-mm lift thickness placed at the NCAT Test Track cooled twice as fast as a 37.5-mm lift, lead- ing the authors to conclude that lifts should be at least 38 mm. Very little research has been aimed at modeling the compaction process. One of those research efforts developed a constitutive model describing the permanent change in volume of an asphalt mix during compaction. The model was first developed to describe the compaction of mix inside a Superpave gyratory compactor (SGC) on a macroscopic level using the finite element method (Masad et al., 2009). The asphalt mixture is modeled as a nonlinear viscoelastic material whose modulus and viscosity increase as the mixture is densified. Initially in the compaction process, the mix is highly compressible and large deformations occur within the mix. As the density increases, the deformations decrease as the aggregates move from a loose to a dense state (Masad et al., 2016a). The initial work modeled the compaction of five different Texas mixes in the gyratory. The gyratory compaction curves were used to estimate the model parameters. The finite element simulations reflect the compaction observed in the SGC and were able to capture the effects of changing the angle of gyration (Masad et al., 2009). The model was later used to simulate field compaction on four projects in Texas with different mix types, aggregates, binder grades, volu- metric properties, and pavement structures. In this case, the model was able to capture the effects of changing the amplitude and frequency of vibration, weight of roller, and differences in com- paction of confined versus unconfined edges of the mat (Masad et al., 2016b). The model has been incorporated into CAPA 3D (computer-aided pavement analysis) and can be used to predict field compactibility using laboratory measurements (Masad et al., 2009, 2016b). 2.2 Evolution of Superpave In the early 1980s, asphalt pavements nationwide were frequently failing to achieve the desired level of service and service life. Rutting was particularly common, but there were also problems with excessive fatigue and low-temperature cracking. There were other issues on the nationâs highway network that eventually led to the initiation of intensive, focused research efforts in six different areas (asphalt, concrete, bridges, maintenance cost-effectiveness, snow and ice control, and long-term pavement performance [LTPP]). The overall research pro- gram was called the Strategic Highway Research Program, SHRP (pronounced âsharpâ). The total budget for SHRP was $150 million over 5 years with $50 million dedicated to the asphalt research program, which suggests the magnitude of the problems with asphalt pavements at the time. More details on SHRP, and particularly the asphalt research program, are available in an NCHRP report (McDaniel et al., 2011), which is the source of the information presented in this section.
Literature Review: Effects of Lift Thickness on Pavement Performance 9 The main product of the asphalt research program under SHRP was a new mix design system dubbed Superpave, for SUperior PERforming asphalt PAVEments. Superpave included â¢ A new binder specification and tests, now known as performance grade (PG) binders and specified in AASHTO M 320; â¢ A new method for compacting specimens in the laboratory, the SGC; â¢ A set of aggregate property requirements; â¢ A volumetric method for designing asphalt mixtures, now standardized in AASHTO M 323 and R 35; and â¢ A series of proposed mixture performance tests intended to verify if a designed mix would perform as needed for the traffic and climate it would experience. (For a variety of reasons, these tests have been superseded by newer, more accurate tests.) Implementation of Superpave began in 1993â1994. By 1997, it was being used routinely in some states. Aside from the proposed performance tests, Superpave has now been adopted by almost every state. In fact, many states no longer use the term Superpave; they just call the mix- tures âasphalt mixturesâ or âhot mix asphalt.â As experience with the new system accumulated, the Superpave mix design system evolved. Superpave was largely geared toward stiffening and strengthening asphalt mixtures because rutting was such a prevalent issue at the time. This consideration led to stricter limits on aggre- gate properties, including angularity, texture, and gradations, than were typical with conven- tional Marshall designed mixtures. Most of the early mixtures were coarsely graded since these were commonly perceived to be more rut resistant. In addition, the early aggregate gradation requirements included control points, between which the gradation was required to pass, and a restricted zone (RZ), which the gradation was to avoid. Marshall mixes with a âsand humpâ in the intermediate size range of the gradation were frequently tender or unstable, so the RZ was defined to prevent the gradation from having this characteristic. The initial gradation control points and RZ are illustrated in Figure 1. A typical coarse graded, S-shaped 12.5-mm gradation is shown in Figure 2. The RZ became somewhat controversial fairly early in the implementation of Superpave in the mid to late 1990s. There were states, such as Georgia, that had successful mixes that passed through this zone. The RZ was more of a concern with coarse mixes where the gradation entered the zone from below. Fine mixes generally could pass through the zone from above without becoming tender. Because the RZ was not essential for all types of gradations, it was dropped from the Figure 1. Example early Superpave gradation chart, showing control points and restricted zone for a 12.5-mm mixture.
10 Impact of Asphalt Thickness on Pavement Quality national specifications in the early 2000s, as shown in Figure 3, though some states continued to specify it. The RZ had also been used to differentiate between fine and coarse mixes. The gradation curve for fine mixes passed above the RZ, and coarse mixes went below it. Without an RZ, some means of defining fine and coarse mixes was needed. Thus, the primary control sieve (PCS) was introduced. The PCS varies depending on the NMAS. Fine versus coarse is defined by where the maximum density line meets the 4.75-mm sieve for intermediate mixes, that is, 19.0 and 25.0 mm, and the 2.36-mm sieve for surface mixes, 9.5 and 12.5 mm (AASHTO M 323-17). The gradation of a fine graded mix passes above the PCS, and coarse mixes pass below it. Figure 3 shows the currently specified control points and PCS for a 12.5-mm mix. Some of the other commonly observed differences between Superpave and Marshall or Hveem designed mixes included lower binder contents, more use of modified binders, higher amounts of crushed coarse aggregates, and more angular sandsâall of which can make a mix more difficult to compact (Brown, 1997). Figure 2. Typical early Superpave coarse gradation for a 12.5-mm mixture. Figure 3. Current gradation requirements showing PCS replacing restricted zone and example of 12.5-mm fine gradation.
Literature Review: Effects of Lift Thickness on Pavement Performance 11 As shown later in this chapter, compaction difficulties were often encountered with coarse graded Superpave mixtures. Some states that had required coarse graded mixes, such as Florida (see Chapter 4), began to allow or even require fine graded mixes to be used instead. Other states have made various changes to their mixture specifications and design practices to try to make mixes easier to compact. Some of the changes include adding additional binder, changing design gradation levels, relaxing some aggregate shape and texture requirements, and more; some of these are discussed in the survey responses (see Chapter 3). Superpave is not static. As experience continues to grow, new materials or techniques are developed, and other advancements are made, Superpave will continue to evolve. 2.3 Relationships Between Lift Thickness, NMAS, and Air Voids While it has been known for decades that lift thickness and mix (or aggregate) size have a pro- found influence on the ability to densify an asphalt mix, the optimal levels of those factors have been the subject of debate over the years. As changes have been made in mix designs, materials, and compaction methods, the recommended values have evolved. For example, in 1982 the Corps of Engineers maintained that using a larger NMAS yields a more stable mix with better frictional properties at a lower binder content (Brown, 1982). More recent research has shown that workability, which reflects the ability to compact a mix, decreased as NMAS is increased (for a given aggregate type, gradation shape, and binder) (Gudimettla et al., 2003). In addition, the use of more angular aggregates can result in more rut-resistant mixes even if the NMAS is smaller (Cooley and Williams, 2009). Recommended lift thicknesses have also changed over time. In discussion of a paper given at an ASTM symposium (Brown et al., 1989), Dick Davis said the MAS should approach the lift thickness; he claimed that thousands of miles of pavement with the MAS exceeding two-thirds the lift thickness had performed well for decades. While this may seem a bit extreme, prior to Superpave a ratio of 2:1 was common (Brown et al., 1989; Hughes, 1989; Brown, 1997). At that time, the lift thickness ratio was typically related to the maximum aggregate sizeâthe sieve size through which 100% of the aggregate would pass. Superpave mix sizes are expressed rela- tive to the nominal maximum aggregate size, which is one sieve larger than the first sieve to retain more than 10% of the aggregate. The NMAS is one sieve smaller than the maximum, so the ratio of lift thickness to NMAS should be increased to about 3:1. In addition, the generally coarser aggregates and greater use of modified binders result in stiffer or harsher mixes that could be more easily compacted at a higher ratio. Therefore, a ratio of three times the NMAS is recommended (Brown, 1997, 1998; Brown et al., 2005). More recent experience and research suggests the t/NMAS ratio should be even higher. One of the most influential reports on the relationships among lift thickness, NMAS, and density was the research conducted under NCHRP Project 9-27 by NCAT and extensively reported in the literature (NCHRP Report 531, Brown et al., 2004). This study involved both laboratory and field phases. Attempts to relate t/NMAS to density in the laboratory by compacting different thicknesses in the gyratory compactor or using a vibratory compactor did not provide clear answers. This lack of correlation may have been because the stress states within specimens in laboratory compaction vary from those in the field. Therefore, the research team evaluated seven mixes from the NCAT Test Track that were placed at varied lift thicknesses from two to five times the NMAS. This part of the study did yield reasonable results. It was found that for fine graded mixes, the density at a t/NMAS ratio of 2 was 2.6% lower than at the optimum ratio (where the greatest density was achieved). For example, a
12 Impact of Asphalt Thickness on Pavement Quality fine graded 9.5-mm mix was compacted to 96.5% of laboratory density at the optimum t/NMAS ratio of 4.6:1 but the density was only 94.3% when the ratio was 2:1. For coarse gradations the impact was even greaterâa decrease of 4.8% (Brown et al., 2005). As the lift thickness increased, the deviation in the air void content from the lowest air void content decreased. These data showed diminishing returns as the lift thickness ratio increased above 3 to 4. Fine graded mixes generally showed less increase in density above a ratio of about 3, and coarse and SMA mixes generally showed substantial density increases up to a ratio of 4 or more. The lift thickness ratios recom- mended, then, were at least 3 for fine graded and at least 4 for coarse graded mixes. The lift thick- ness ratio could be up to 5 with no detrimental effects but caution should be exercised at higher lifts since achieving density, especially at the bottom of thick lifts, could become problematic. This experiment also demonstrated the fact that thinner lifts cool more quickly, which no doubt contributes to the lower densities achieved (Brown et al., 2004). In another part of the NCHRP Report 531 research, the researchers evaluated in-place air voids, lift thicknesses, and permeabilities at 20 construction sites incorporating different NMAS, gradations, t/NMAS ratios, binder grades, and Ndes levels; the sites all had approximately the same in-place density target. Trend lines indicated that as t/NMAS increased by 1, air voids decreased by 0.5% to 1.0%. The R-square values were very low because of a great deal of scatter in the data, but the p-values indicated a strong relationship between t/NMAS and in-place air voids (Brown et al., 2004). Cox et al. (2015a) collected data from 12 different Mississippi asphalt paving projects, includ- ing 31 different variables ranging from aggregate and mixture properties to field construction variables such as temperatures and rolling time or passes. Statistical analysis of the data revealed that t/NMAS had a greater effect on air voids than aggregate properties such as fine aggregate angularity, moisture content, methylene blue value, surface area of the fines, and others. Three different models to predict air voids using the significant variables all incorporated t/NMAS and suggested an optimal t/NMAS range of 4 to 6. In other, related work, Cox et al. (2015b) determined that longitudinal joints are even more difficult to compact in thin lifts (<38 mm) than in thick layers. In another study the authors (Williams et al., 2015) explored asphalt mix compactibility on a total of 22 field projects in Mississippi using 12.5- and 9.5-mm mixes. The t/NMAS ratios ranged between 2.5 and 7.1. They tested loose mix samples from the plants and roadways in addition to cores. They also recorded mat temperatures and nuclear densities during compaction. Regression models were used to relate air voids to other parameters. The final regression models related air voids to t/NMAS, initial surface temperature, difference between percent passing the No. 8 sieve and the maximum density line, accumulated com- paction pressure, aggregate surface areas, and total asphalt binder content. The results sug- gested that optimal t/NMAS ratios could be in the range of 4 to 6, rather than the commonly used 3 to 5. The study also confirmed that the ratio could be too high, resulting in lower than desired densities of about 6 or 7. Another laboratory study, this one in Wisconsin, explored the impacts of lift thickness (Bahia and Paye, 2001). When Wisconsin implemented the Superpave mix design method in 2000, they used high levels of angularity in the fine aggregates and generally coarse mixtures, especially on higher traffic volume roadways. These mixes often proved to be harder to compact. Reasons for this increased difficulty, according to contractors, included more manufactured sand, thin lifts, high fine aggregate angularity, and low binder contents. In a laboratory study, the researchers found a relationship between t/NMAS of mix in the gyratory compactor and density. Thickness was varied by changing the mass of mix in the mold. Smaller samples (<3,000 g) had higher air voids than larger samples. They also used the gyratory load plate assembly to measure the shear resistance of mix during gyratory compaction. After compaction, the mixes were extracted to determine if
Literature Review: Effects of Lift Thickness on Pavement Performance 13 aggregates were being crushed; no significant crushing was observed in any of the mixes tested. Laboratory results indicated lift thickness ratios should be between 4 and 6 times the NMAS. Buttlar et al. (2015) also found the use of fine graded mixes to be beneficial. Possible ben- efits of using fine graded versus coarse graded mixes included better compactibility, lower per- meability, less segregation, and improved smoothness. Coarse graded mixes were preferred in Illinois because they were viewed as more stable than fine graded mixes. With increased use of manufactured sand, however, this was not necessarily true. The authors pointed out that the FAAâs P-401 mixtures are fine graded mixes that can withstand heavy aircraft loadings. Illinoisâ existing gradation bands were coarser than those recommended under the Superpave mix design method. Illinois was still using the Superpave-defined restricted zone to prevent tender mixes. Illinoisâ thickness requirements met or exceeded the ratios recommended in NCHRP Report 531 for fine mixes (3:1), but not coarse (4:1). Buttlar et al. (2015) evaluated three fine graded 19.0-mm mixes and one coarse 19.0-mm control mix in the laboratory. The mixes were designed at a VMA (voids in mineral aggregate) of 13.4% Â± 0.1%, 4% air voids, and Ndes = 90. The fine mixes mainly differed in their gradations between the 0.600- and 9.5-mm sieves. The effective binder content was the same for all mixes (4.1%), but the asphalt absorption was somewhat higher for the fine mixes, so the total binder content was 0.1% to 0.3% higher. This small increase in total binder content shows that finer mixes are not necessarily significantly more expensive because of a higher binder content; asphalt absorption is a more important factor in determining cost. Results of the laboratory performance testing (Hamburg wheel tracking; disk-shaped compact tension, DC(T); dynamic modulus; and four-point beam fatigue) and accelerated loading (ATLAS) testing supported modifying the specifications for 19.0-mm HMA to reduce segregation potential and improve durability. The performance testing results consistently showed the fine mixes would be expected to perform as well as or better than the coarse mix. All of the planned accelerated load testing was not complete at the time of the report but results to date were very promising. A previ- ously constructed fine mix in one district was performing well (Buttlar et al., 2015). The specifications were revised to redefine the 19.0-mm gradations to be finer; the proposed changes were under review at the time of the report (Buttlar et al., 2015). Experience with fine graded mixes in one district showed that while bid prices were initially somewhat higher than for coarse graded mixes, with experience the bid prices came down and were comparable to that for coarse mixes. Contractors learned they were more likely to receive density bonuses with the finer mixes. 2.4 Performance Impacts As mentioned previously, in-place density is seen as the most critical factor affecting the perfor- mance of an asphalt pavement. Inadequate compaction can lead to rutting under traffic, as well as moisture damage and increased cracking. The latter two distresses can be related to the higher air void contents allowing the ingress of water and air into the pavement (Brown, 1982). This section summarizes the literature related to the air void content at which mixes become permeable and the mix parameters that affect the critical air void content. Then the literature related to forensic studies of other performance indicators, such as cracking, modulus, and fatigue life, is discussed. 2.4.1 Permeability It is well established that there is some link between lift thickness and the ability to compact an asphalt mix to an acceptable density. While there is an obvious link between air voids and permeabilityâwater and air move through the air voids, after allâthere is evidence that the
14 Impact of Asphalt Thickness on Pavement Quality overall air void content is not sufficient to predict whether a pavement will exhibit excessive permeability. For example, Vardanega (2014) conducted an extensive review of literature and models related to asphalt mix permeability. The author concluded that although the t/NMAS ratio relates to the compactibility of an asphalt mix, it does not correlate well to permeability. Permeability is affected more by porosity and mix gradation. The size and interconnectedness of the air voids determine if a mat is permeable (Zube, 1962; Ford and McWilliams, 1988; Choubane et al., 1998; Jackson and Mahoney, 2014). Larger air voids have a greater chance of being interconnected within the mat, which is one reason coarser gradations are more likely to be permeable (Mogawer et al., 2002). Therefore, much of the research into performance issues relative to lift thickness evaluates the permeability of mixes in the laboratory or in the field. There is still value in attempting to relate permeability to density, since density is easily and routinely measured in the field. While it may not be possible to accurately predict permeability based solely on density (or air voids) in all cases, achieving an acceptable density for certain types of mixes can give some assurance the mix will be imperme- able. For example, Hughes and Maupin (1986) said one of the key factors to reduce moisture damage is achieving adequate compaction during construction; such damage would be nearly eliminated by compacting mixtures to 4% to 5% air voids at the time of construction. Some researchers have found good relationships between air voids and permeability when looking at particular types of mixes or limited data sets. For example, Jackson and Mahoney (2014) found that an exponential relationship defined that link for mixes used in Connecticut. Prowell (2001) found a strong relationship among in-place density, gradation, and permea- bility when doing a forensic evaluation of one pavement that was holding water. Nataatmadja (2010) related permeability to air voids using data from Mallick et al. (2001) for a 9.5-mm fine graded mix using five different types of models: power, exponential, quadratic, cubic, and hyperbolic functions. He showed the strongest correlation with a hyperbolic function (R2 = 0.9917). Conversely, in an analysis of 400 test sites (20 sections at each of 20 projects), Schmidt and Owasu-Abadio (2007) found no correlations between surface lift thickness and in-place density or thickness and permeability. However, they looked only at fine graded mixtures, which have been shown to be less sensitive to the effects of t/NMAS than coarse graded mixes. Many researchers have attempted to define the air void content (or density) at which mixtures become permeable. The values they report as thresholds vary, likely due to other factors besides air void contentâmost likely the interconnectedness of those air voids as noted earlier. Some of the thresholds reported include the following: â¢ Zube (1962) reported that field permeability increased markedly over 6% to 10% air voids. â¢ Brown (1982) stated that a Corps of Engineers study found that field permeability doubles when air voids increase by 1%. At 7% air voids or less, the mix is virtually impermeable. â¢ Mallick et al. (2001) found the field permeability of a 9.5-mm fine mix increased significantly above 8% air whereas a 9.5-mm coarse mix became permeable at 5% to 6%. Larger NMAS mixes became permeable at lower air voids; for coarse mixes those levels are 7% for 12.5 mm, 6% for 19 mm, and 5% for 25 mm. â¢ Allen et al. (2003), in Kentucky, found that below 92% of Gmm there was a dramatic increase in field permeability regardless of the size of the mixture. This was observed with both the NCAT field permeameter and the air induced permeameter (AIP) developed in this study. The AIP uses a vacuum to pull air through the pavement, which is indicative of the mat permeability. â¢ Comparison of laboratory-compacted samples versus field cores of four 9.5-mm and two 12.5-mm Superpave mixes from Pennsylvania showed that laboratory-compacted specimens exhibited lower falling-head permeability values than field cores at the same air void content
Literature Review: Effects of Lift Thickness on Pavement Performance 15 (or porosity). At air void contents greater than 7% to 8%, the permeability of 12.5-mm mixes increased âdrastically.â For 9.5-mm mixes, the mixes were impermeable below about 8% air voids (Solaimanian, 2010). The mix designs indicated these were all coarse graded mixtures (Raisi, 2009). â¢ In a study of 42 Superpave field projects, Hainan and Cooley (2003) investigated the effects of gradation (fine versus coarse), NMAS (9.5, 12.5, and 19.0 mm) and lift thickness (22.3â78.8 mm for t/NMAS ratios between 1.74 and 6.31) on permeability of cores. The results showed the permeability began to increase at greater than 6% air voids and increased sharply at air void contents above 8%. Furthermore, as lift thickness increased, permeability decreased. â¢ Cooley et al. (2001) evaluated mixes from 11 Superpave construction projects using a field permeability device. They found that, as NMAS increased, the air void content at which the pavement became permeable decreased. Threshold values determined in this field study were 7.7% air voids for coarse graded 9.5- and 12.5-mm mixes, 5.5% air for coarse 19.0-mm mixes, and 4.4% air for coarse 25.0-mm mixes. The in-place permeabilities at those air void contents were 100 Ã 10â5, 120 Ã 10â5, and 150 Ã 10â5 cm/s, respectively. â¢ Westerman (1999) observed that coarse graded Superpave mixes constructed in Arkansas through 1997 performed well until a very wet summer when permeability was observed in the field. Mixes with air voids greater than 6.5% were found to be permeable. In addition, a t/NMAS ratio of 4 was recommended. â¢ Mohammad et al. (2003) measured the laboratory permeabilities of Superpave cores from 17 projects in Louisiana. The 19-mm surface mixes were placed in 40- to 50-mm lifts and 25.0-mm intermediate courses placed in 50- to 100-mm lifts. Coarse graded mixes were per- meable (>125 Ã 10â4 mm/s) at air voids greater than 5.3%. None of the fine graded mixes, with air voids between 4.6 and 7.2, were permeable. The 19-mm mixes were slightly more perme- able than the 25-mm mixes, which was attributed to the greater lift thickness leading to fewer interconnected voids. Mixes placed in lift thicknesses greater than 60 mm were generally less likely to be permeable. In a field study of 42 construction projects, Hainin and Cooley (2003) found a relation- ship between permeability and in-place density based on NMAS and gradation. The projects sampled utilized five different types of mixes: fine graded 9.5-, 12.5-, and 19.0-mm mixes and 9.5- and 12.5-mm coarse graded mixes. The air void content at which mixtures became variable differed depending on whether the bulk specific gravity was determined using the CoreLok or the AASHTO T 166 water displacement method; in general, the critical air void content was higher when the CoreLok was used. As the NMAS increased, the air void content at which the mixes became excessively permeable decreased. Fine graded mixes were less permeable than coarse graded mixes at the same air void content. Brown et al. (2004) also found differences between the specific gravities, and therefore calculated air void contents, when measuring gravities using AASHTO T 166, the vacuum sealing (CoreLok) method, and dimensional analysis. They pointed out how critical the bulk and maximum specific gravities are in determining compliance with density specifications and pay factors. Another point of differences in measurements is a comparison of laboratory- and field- measured permeabilities. Many researchers have noted these differences and have observed that measuring permeability in the field can be difficult. One practical reason for this difficulty is obtaining a good seal of the permeameter to the pavement surface, especially with highly tex- tured surfaces (Russell et al., 2005). Another, more fundamental reason is that Darcyâs law, which is used to characterize permeability, assumes laminar, one-directional flow (Jackson and Mahoney, 2014). In a pavement, the flow of air or water is likely not laminar and the flow direc- tion can be horizontal as well as vertical (Jackson and Mahoney, 2014).
16 Impact of Asphalt Thickness on Pavement Quality Concerns about the coarse gradations of early Superpave mixes prompted the Colorado DOT to explore the possibility of using an air permeameter developed by Marquette University to test the in-place permeability of asphalt pavements to identify those that may develop performance problems related to high permeability (Retzer, 2008). The author cited McLaughlin and Goetz, who, in 1955, identified permeability as the most critical factor affecting durability. A total of 20 field projects incorporating different aggregate sizes, gradations, lift thicknesses, and com- paction equipment were evaluated. Since previous research indicated that permeability varies with NMAS and fine versus coarse gradation, those factors were included. The effects of rubber- tired versus steel-wheeled rollers were also studied. The results showed no significant differ- ence between fine and coarse gradations for their Â½-in. NMAS mixes. The Â¾-in. mixes were all on the fine side, so no comparison could be made at that size. Lift thicknesses between 2 and 3 in. did not have a great impact on permeability. Colorado DOT mixes appeared to be relatively impermeable below about 8% air voids. The use of rubber-tired rollers did reduce the perme- ability by about half (100 Ã 10â5 cm/s) below 8% air voids. No recommendations were made to change density requirements or to require permeability testing since the existing requirement of 92% to 96% Gmm seemed to yield pavements with acceptable permeability. The use of rubber- tired rollers, however, was encouraged. In summary, the literature shows different ranges of air void contents at which mixtures become permeable. In general, however, it can be said that coarse graded mixes tend to become permeable at lower air void contents than fine mixes. In addition, smaller NMAS mixes tend to be impermeable up to higher air void contents than larger NMAS mixes, as shown by Cooley et al. (2001), Mallick et al. (2001), Mallick and Cooley (2002), Mallick et al. (2002), Cooley (2003), and others. Mixes with air voids at 8%, a common field density requirement, are frequently observed as being permeable. NCHRP Report 531 recognized the variation in the points at which mixes become permeable and recommended air voids should be 6% to 7% or lower to avoid permeability problems (Brown et al., 2004). The ranges of air voids and their possible connectivity are illustrated in Figure 4 (Cooley et al., 2001). While the limits of the different phases may vary, this shows that mixes tend to be impermeable and voids are isolated at lower air voids; then, as air voids increase, connectivity and permeability increase. Eventually, at higher air voids, permeability increases dramatically as the voids become inter- connected, providing paths for water and air intrusion. 2.4.2 Forensic Studies Yet another source of variation in measured values that may contribute to the different per- meability thresholds reported is a difference between nuclear or nonnuclear gauge and core densities. This lack of agreement has been recognized for many years. For example, Kandhal and Koehler (1982) reported that in the mid-1970s, the Pennsylvania DOT observed premature loss of fines and coarse aggregates on a number of asphalt overlays where nuclear gauges were used to measure density on control strips to set roller patterns. This Pennsylvania study produced the widely referenced chart relating air void content versus extent of raveling (see Figure 5). Inves- tigation of these pavements attributed the premature distress to low densities. Previously, when cores were used to measure density, the requirements in Pennsylvania were â¥95% of Marshall density. When nuclear gauge use began in the 1970s, the specifications required 98% of the maximum control strip density. The surface mixtures exhibiting this distress were 100% passing the 12.5-mm (Â½-in.) sieve mixes with 80% to 100% passing the 9.5-mm (3â8-in.) sieve. They were placed in 25-mm (1.0-in.) lifts on most projects (six investigated) and in 38-mm (1.5-in.) lifts on two projects. One recommendation arising from this study was to increase the lift thickness to 38 mm (1.5 in.). Nuclear gauge and core densities did not agree, and the nuclear readings were more variable. However, in most cases, the nuclear gauge readings could be used if adjusted (calibrated) relative to cores.
Figure 4. Method for selecting critical in-place air voids and field permeability (Cooley et al., 2001). Figure 5. Air void content versus extent of raveling (Kandhal and Koehler, 1982). Reprinted, with permission, from Placement and Compaction of Asphalt Mixtures, ASTM STP 829, copyright ASTM International, 100 Barr Harbor Drive, West Conshohocken, PA 19428.
18 Impact of Asphalt Thickness on Pavement Quality Florida is another state where significant differences between nuclear and core densities were noticed (NAPA, 1999, Musselman et al., 1998). After detecting high air voids in mixes that the nuclear gauge had indicated were adequately compacted, the Florida DOT deter- mined that the nuclear gauge in backscatter mode was not accurate enough on coarse graded mixes with high air voids. They then changed their density acceptance process to use core densi- ties (Musselman et al., 1998). Many states now routinely correlate their nuclear density gauges with cores to avoid this lack of agreement (see survey responses, Chapter 3). Variation of the placed lift thickness from the designed thickness can also lead to performance issues. When investigating the cause of significant premature distress on an Indiana highway, Haddock and Prather (2004) discovered that the thickness of the 9.5-mm surface course, which was planned to be 32 mm (1.25 in.), was not achieved. The actual lift thicknesses ranged from about 15 to 31 mm (0.6 to 1.22 in.) with an average of 22.4 mm (0.88 in.). Cores also showed interconnected voids. The average air void content was 11.8%. Distresses observed 1 year after placement included longitudinal and fatigue cracking, weeping of water, and aggregate pop- outs. The aggregate bulk (Gsb) and effective (Gse) specific gravities, as well as the binder absorp- tion (Pba), were found to deviate from the mix design values; Gsb was lower and Gse and Pba were higher. The produced aggregate gradation was finer than the design target while the total binder content was equal to or slightly higher (0.2%) than the design. The aggregate properties and mixture volumetrics met the Indiana DOT specifications. The permeabilities calculated based on the measured air voids and lift thicknesses were high and highly variable. The authors con- cluded that the mixture was placed in too thin a lift; they recommended that the lift thickness should have been four times the NMAS. They also determined that the coarse aggregate content affects the void structure in the compacted mix and therefore affects permeability. In this case, the mix was too coarse to be placed in such a thin lift (despite the fact that the mix was finer than designed). Of course, air voids and lift thicknesses can affect more than the permeability and related distresses. For example, Zeinali et al. (2014) used flow number, flexural beam fatigue, and disk- shaped compact tension (DC[T]) to evaluate the performance of plant-produced mixtures from Kentucky relative to their density. The mixes were designed for fewer than 10 million equiva- lent single-axle loads (ESALs). The performance of the mixes, when compacted to 92% Gmm, was compared to the performance at the in-place density, which was between about 87% and 90%. The flow number was significantly affected by density. A decrease in the density by 3% to 5% from 92% decreased the flow number by 37% to 65%. As density increased, the fatigue life increased slightly (at 400 microstrains and 20Â°C), but the difference was not statistically signifi- cant. A reduction in the mixture density decreased fracture energy and resistance to thermal cracking significantly. These mixes were generally harder to compact than the Asphalt Institute laboratory standard mix; increasing the lift thickness was recommended to counter this. A perception that larger NMAS mixes are more rut resistant has led many states to use larger mixes; when used in lifts with lower t/NMAS ratios, this can lead to problems. However, smaller NMAS mixes could provide lower permeability and greater fatigue resistance. Therefore, some researchers have compared the performance of different NMASs and their effects on perfor- mance. As one example, Christensen et al. (2013) explored the possibility of using a wider range of mixture sizes in Wisconsin, thus giving contractors more flexibility. Prior to this project, the Wisconsin DOT required the use of 12.5-mm surface mixtures over 19.0-mm mixtures in lower lifts. The results of laboratory modulus testing and predictive modeling (using the Mechanisticâ Empirical Pavement Design Guide, among other tools) suggested the Wisconsin DOT could allow use of either 9.5- or 12.5-mm surface mixes, 12.5- or 19.0-mm intermediate and base courses, and 9.5-mm leveling courses. Mixes with an NMAS of 25 mm were predicted to possibly have high permeability and be prone to thermal cracking, so their use was not recommended. The t/NMAS should be 3 to 5 for fine mixtures and 4 to 5 for coarse mixes, according to this study.
Literature Review: Effects of Lift Thickness on Pavement Performance 19 Lifts thinner than 1.5 in. should be avoided because of a tendency to cool too quickly. Analysis of the data in this project showed that mixes with larger NMASs were more prone to segregation. In addition, they cited Gudimettla et al. (2003), who found mixes with larger NMAS exhibited less workability than smaller NMAS mixes. In a comparison of fine and coarse graded mixtures using accelerated loading (heavy vehicle simulator, HVS), the testing showed that the fine graded mix performed as well as or slightly better than the coarse mix in terms of rutting. Laboratory testing of the rut resistance of plant- produced mixes in the Asphalt Pavement Analyzer (APA) correlated well with the HVS results. The mixtures were designed for 10â30 million ESALs using the same aggregates, unmodified binder, and effective binder content (Choubane et al., 2006). Lakkavalli et al. (2015) studied the relationship between lift thickness and pavement smooth- ness. The City of Calgary had implemented pavement smoothness specifications using the inter- national roughness index (IRI) measured with an inertial profiler. The researchers used those smoothness data to analyze the effects of lift thickness on smoothness by comparing before and after IRI values on a variety of projects. Four test sections were rehabilitated with mill and fill treat- ments consisting of 60 mm in one lift, 70 mm in one lift, 100 mm in two lifts, and 90 mm in two lifts. Two additional sections were reconstruction projects with two lifts of asphalt (250 mm total) over granular base on granular subbase. The results showed greater improvement and lower final IRIs in the two lift applications than in the single lifts. The thicker lifts were also more effective at improving the ride quality. The authors recommended that layers thicker than 60 mm be placed in two lifts to improve smoothness. Smith (2014) also looked into factors affecting the IRI of asphalt overlays. He analyzed IRI data from more than 1,500 miles of asphalt pavement overlays and found that average IRI decreased as overlay thickness increased and generally decreased as the number of lifts increased. When only a single lift was applied, the pre-overlay condition had a great influence on the final IRI. For three-lift overlays, the pre-overlay smoothness was not a determinant of final IRI. Decades earlier, Hughes (1978) also found that smoothness did increase as lift thickness increased on one construction project with three lift thicknesses ranging from 1.4 to 1.8 in. (35.6 to 45.7 mm). It was cautioned, however, that this may not be generally true as thick lifts may be rougher than two lifts of moderate thickness. Nehdi and Welker (2002) investigated the causes for premature failures of nine dense graded friction course surfaces. These mixes consisted of 100% crushed aggregates with relatively low binder contents and were intended to resist rutting caused by heavy truck traffic. The premature failures included raveling, potholes, and loss of fine aggregate, which occurred within 3 to 4 years after construction. Although the average values of binder content (5.15%), in-place air voids (4%), and percent compaction (96%) from cores looked acceptable on one job, it was found that variations in these values contributed to the observed distresses. It was also determined that the constructed lift thickness varied between about 30 and 65 mm; a lift thickness of 40 mm would relate to three times the NMAS. At the lower thicknesses, air voids increased and density decreased. In addition, the binder content after aging was reduced, indicating that excessive oxi- dation may have led to stripping of the binder. Segregation was also observed in the field. Rec- ommended remedies to the problems observed included using antistrip additives and material transfer vehicles, as well as better controlling and increasing lift thicknesses to prevent thin layers. 2.5 Costs and Benefits Little quantified (life cycle cost analysis, LCCA) information was found in the literature related to the cost implications of increasing lift thickness. However, several studies have explored the potential performance benefits of various mix types, sizes, and thicknesses, which imply cost
20 Impact of Asphalt Thickness on Pavement Quality benefits as well. Possible means of reducing costs have also been reported, though not in quantified dollars and cents. In addition to the benefits noted in the preceding section, studies demonstrating performance benefits and options are summarized in this section. One study that did include LCCA was conducted by NCAT to support the FHWA Enhanced Density Initiative (Tran et al., 2016). This literature review noted that past research showed that a 1% reduction in air voids increases fatigue life by 8.2% to 43.8%, rutting resistance by 7.3% to 66.3%, and service life by at least 10%. Using those improvements, LCCA of a simulated $1 million paving project with density at 93% compared to the same project compacted to 92% led to an estimated savings in net present value of $88,000 (8.8%). In addition, the review mentioned that a 2007 AASHTO Subcommittee on Materials survey indicated that more than one-third of the states had density requirements of less than 92%. These requirements were usually based on what was historically possible to achieve; with improvements in technology, design, and equipment, higher densities may be feasible. Technologies mentioned as possible means to increase densities included warm mix asphalt (WMA), high-frequency vibratory roll- ers, intelligent compaction, and joint construction techniques (sealants, joint heaters, attention to detail, etc.). The authors concluded that agency specifications can have a significant impact on the ability to achieve density. Simply requiring a higher density for 100% pay or a bonus can be effective, but appropriate lift thicknesses and project scopes are also needed. Sound bases are also necessary for compaction; if the underlying material is weak, the project scope should address that problem rather than just trying to overlay the weak material. Bell et al. (1984) explored the effects of density on various aspects of mix performance in a laboratory study. They compared mixtures compacted to 4% (considered excellent), 8% (good), and 12% (poor) air voids. The difference in stiffness (indicated by resilient modulus) from excel- lent to poor compaction led to an estimated âthreefold reduction in service life.â Fatigue life was 30 times greater at 4% than at 12% air voids. Blankenship and Anderson (2010) conducted a study for the Kentucky Transportation Cabinet to explore the relationship between pavement density and durability. Using beam fatigue, dynamic modulus, and flow number testing they concluded that increasing the density from 92% to 93.5% increased the fatigue life by 10% and increased flow number by 34%. Using flexural beam fatigue testing in controlled strain mode, Harvey and Tsai (1996) showed that for thick structures, reducing the âair void content from 8 to 5 percent increases fatigue life approximately 200 percent.â For thin structures, the same reduction in air voids increased fatigue life by 100% to 150%. Similarly, the fatigue life also increased as asphalt content increased, with about a 10% increase in life for thick structures and 20% for thin structures for each 0.5% more asphalt. Caution is needed when increasing the asphalt content to reduce the risk of rutting, but achieving a lower air void content during construction improves both fatigue and rutting resis- tance. They evaluated one combination of granite aggregates with an AR-4000 asphalt cement at five asphalt contents (4.0%, 4.5%, 5.0%, 5.5%, and 6.0%), three air void contents (1% to 3%, 4% to 6%, and 7% to 9%) and two strain levels (300 and 150 microstrains). There are cases, such as maintenance or pavement preservation applications, where thin lifts are essential. How to design these types of mixes so that they can be constructed properly was explored by Scullion et al. (2014). This research for the Texas DOT led to three so-called per- formance mixes that were designed to be placed in thin lifts, 25 mm (1 in.) or less, to replace typical maintenance mixes placed in 50-mm (2-in.) lifts. The dense-graded mix is called crack attenuating mix (CAM); there is also a gap-graded fine SMA and a fine permeable friction course (PFC). The performance mixes must meet requirements in the Hamburg wheel tracking device to resist rutting and the overlay tester to resist cracking. The mixes are characterized by high binder contents (>5.5% for CAM and >6% for PFC and SMA), polymer-modified binder,
Literature Review: Effects of Lift Thickness on Pavement Performance 21 angular aggregates, and no use of reclaimed materials. In terms of material costs, they are quite expensive, but since they are placed at about half the lift thickness, they are considered competi- tive and may even be up to 30% less expensive in initial costs. With an NMAS size of 4.75 mm, placement at 25 mm is feasible, though it was noted that âsome trial and errorâ was encountered in determining roller patterns on some trial projects. Performance after 4 to 5 years is reportedly good. Predictions based on the performance tests suggest that the service lives of the perfor- mance mixes will be improved over the conventional maintenance mixes. 2.6 Summary of Findings from Literature This chapter provides background information on the basics of asphalt mix compaction and the evolution of Superpave mixtures. Then the literature relating to lift thickness, nominal maxi- mum aggregate size in the mixture, and density is reviewed. The impacts on pavement perfor- mance of failing to achieve adequate density are summarized. Last, the literature related to costs associated with the performance impacts of varying lift thicknesses and densities is reviewed. Optional ways to improve in-place density without increasing lift thicknesses are also presented. Adequate asphalt pavement density is widely recognized as one of the most important factorsâif not the single most important factorâaffecting pavement performance. Density is obtained by compacting the mixture under the paver screed and then by rollers. There are three main types of rollers that have differing operational characteristics and applications. The ability of the rollers to adequately compact a mixture depends on a wide range of factors including mix properties (particularly stiffness), weather conditions, base support conditions, rolling opera- tions, and mixture lift thickness. Many researchers and practitioners have recognized that the ratio of lift thickness to the mixtureâs nominal aggregate size has a great impact on the ability to adequately compact the mix. When the Superpave mix design system was implemented in the mid to late 1990s mixes were often coarser than the previously used mixes. Superpave was developed at a time when rutting was a severe problem, and coarse mixes were perceived to be more rut resistant. The coarse nature of these early Superpave mixes made them especially difficult to compact. One result of this lack of density was excessive permeability of the mat. Research into the problems and experi- ence with these mixes indicated that thicker lifts were needed with coarse graded mixes to enable compaction. Eventually refinements were made to Superpave and to many state specifications to allow the use of finer, less harsh mixes. Many researchers have explored the relationships among lift thickness, mixture size, and pavement density. The frequent goal of many of those research efforts has been to identify the optimal ratio of lift thickness to mixture size. Decades ago the mixture maximum aggregate size was the basis for the lift thickness ratio. More recently the mixture nominal maximum aggregate size has been used. Values for the optimal density and the recommended t/NMAS ratio reported in the literature vary depending on a number of factors. In general, however, there is consensus in more recent research that the density of the mat should be greater than about 92% of Gmm and densities of 93% to 94% would be preferred. Recommended lift thickness ratios are usually at least 3:1 for fine graded mixtures and 4:1 for coarse graded mixtures to facilitate compaction to the optimum level. Research leading to this conclusion has included laboratory, field, and accelerated testing. Several research efforts have shown that fine graded mixtures can be as rut resistant as coarse graded mixtures. While fewer studies have reported on the effects of using smaller NMAS mixes, the limited research identified showed that smaller NMAS mixes could perform well and could be cost-effective.
22 Impact of Asphalt Thickness on Pavement Quality Failure to achieve adequate density, which can be caused by insufficient lift thickness, can cause rutting under traffic, increased cracking, moisture damage, and more. Excessive perme- ability allows air and water to enter the pavement, accelerating binder oxidation (embrittlement) and moisture damage. Several researchers have explored acceptable permeabilities for various sizes and types of asphalt mixtures, and some have related permeability to lift thickness. Other researchers have established relationships between lift thickness and other parameters including smoothness. There is little published literature quantifying the effects of increasing lift thickness. A few studies have assessed the life cycle cost impacts of increasing pavement density, and the impacts were substantial. A 1% increase in density, which could be facilitated by placing in a thicker lift, could increase the fatigue life, rutting resistance, and overall service life dramatically.