Fiber Additives in Asphalt Mixtures (2015)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

7 CHAPTER TWO LITERATURE REVIEW: USE OF FIBER ADDITIVES IN ASPHALT MIXTURES be used. Mineral fibers (also called mineral wool or rock wool) are manufactured by melting minerals then physically forming fibers by spinning [similar to making cotton candy (Science Channel n.d.)] or extruding. Minerals used to create mineral fibers include slag or a mixture of slag and rock (U.S. EPA 1995; Brown et al. 1996), basalt (Morova 2013), brucite (Guan et al. 2014), steel (Garcia et al. 2009, 2012 a and b, 2013 a and b; Serin et al. 2012), and carbon (Clevin 2000; Liu and Shaopeng 2011; Khattak et al. 2012, 2013; Yao et al. 2013). Carbon fibers and steel fibers (or steel wool) have been used in some fairly exotic ways to produce electrically conductive asphalt that can be used for deicing (Garcia et al. 2009, 2012 a and b, 2013 a and b) or to heal microcracks (Gallego et al. 2012; Liu et al. 2012; Garcia et al. 2012a, 2013a; Dai et al. 2013). Steel fibers have been used for research purposes, but because they corroded upon exposure to water, they were not effective in the long term (Freeman et al. 1989; Putnam 2011). Asbestos fibers were the first type of fiber used in hot mix asphalt; they were used from the 1920s (Serfass and Samanos 1996) until the 1960s when environmental and health issues curtailed the use of asbestos (Busching et al. 1970). Synthetic polymer fibers: The most commonly used polymer fibers are polyester, polypropylene, aramid, and combinations of polymers. Other fibers include nylon, poly para-phenyleneterephthalamide, and other less commonly used materials. Different polymers have different melt points, which need to be considered when adding to hot mix asphalt. Production of synthetic fibers typically involves drawing a polymer melt through small holes. Fibers can be bundled together into yarn (although yarn is not typically used today in asphalt concrete) (Busching et al. 1970). Reportedly, aramid fibers contract at high temperatures, which helps resist pavement deformation (Kaloush et al. 2010). Other plant-based fibers: These have been used in more limited areas. They may be derived from woody fibers (such as jute, flax, straw, and hemp), leaves (such as sisal), and seeds; or they may be fruit fibers, such as coir, cotton, coconut, or palm (Cleven 2000; Oda et al. 2012; Das and Banerjee 2013; Qiang et al. 2013; Abiola et al. 2014; Do Vale et al. 2014; Muniandy et al. 2014). Glass fibers: These have not been reported often in the literature but appear to have desirable properties, including This chapter summarizes the findings of the literature review on the use of fibers in asphalt mixtures. The findings are grouped by fiber materials, summarizing the types of fibers used and their typical properties; methods of testing fibers and the significance of those tests; mix design, production, and construction of fiber-reinforced mixtures; laboratory and field performance of fiber-reinforced dense-graded mixtures and open- or gap-graded mixtures; and the cost- effectiveness of the use of fibers. FIBER MATERIALS AND MIXTURES It is generally understood that asphalt is strong in compression and weak in tension (Busching et al. 1970). Adding fibers with high tensile strength can help increase the tensile strength of a mixture. In theory, stresses can be transferred to the strong fibers, reducing the stresses on the relatively weak asphalt mix. To effectively transfer stresses, there must be good adhesion between the fiber and the asphalt binder; a greater surface area on the fibers can aid this adhesion. In addition, the fiber needs to be uniformly dispersed in the mixture to avoid stress concentrations (Busching et al. 1970). If the primary reason for adding fibers is to reduce binder draindown, high strength is not required; fibers that will absorb or retain binder are used in these applications. The properties that affect fiber performance, how those properties are measured, and the types of fibers used are described in this section. Types of Fibers Many types and forms of fibers have been used in asphalt mixtures, either experimentally or routinely. Cellulose, mineral, and polymer fibers are the most common. The most commonly used types of fibers and their reported benefits and disadvantages are summarized in Table 1. Cellulose: Cellulose fibers are plant-based fibers obtained most commonly from woody plants, although some are obtained from recycled newspaper. These fibers tend to be branching with fairly high absorption; it is this nature that helps cellulose fibers hold on to high binder contents in mixtures. Cellulose fibers can be provided in loose form or in pellets. Mineral: Either naturally occurring fibers, such as asbestos (chrystolite), or manufactured mineral fibers can

8 high tensile modulus (~60 GPa), low elongation (3%–4%), high elastic recovery (100%), and high softening point (815°C). They are, however, brittle and must be handled carefully during construction (Abtahi et al. 2013). Waste or recycled fibers: The increasing importance of sustainability in construction has led to increased interest in reusing materials that would otherwise be disposed of, including waste fibers from a variety of sources. Putnam, for example, has explored the possibility of reusing waste carpet fibers and tire fibers from the auto manufacturing industry, with favorable results in terms of increased mixture toughness, permanent deformation, and moisture resistance (Putnam and Amirkhanian 2004). Chowdhury et al. (2005) investigated the use of fibers from recycled tires and found that they performed well, especially in reducing draindown. The advantages of natural fibers include low cost, acceptable strength and mechanical properties, and sustainability. One disadvantage is their tendency to absorb moisture, which can cause them to swell (Table 1) and can interfere with bonding of hydrophobic asphalt with the moisture-laden fiber. Natural fibers can also degrade at high temperatures or moisture conditions. Compatibility of the fiber and the asphalt can be improved with various surface treatments. Overall, however, it appears that some natural fibers, such as jute and sisal, can be used to replace synthetic fibers in asphalt mixes (Abiola et al. 2014). Fibers come in various forms in addition to loose and pelletized cellulose fibers. Fibers can be short and randomly oriented, long and unidirectional, tufts, or woven (Abiola et al. 2014). (Woven fabrics are not the focus of this synthesis.) TABLE 1 REPORTED BENEFITS AND DISADVANTAGES OF COMMON FIBER TYPES Fiber Type Reported Advantages Reported Disadvantages Cellulose • Stabilizes binder in open- and gap-graded stone matrix asphalt (SMA) mixtures. • Absorbs binder, allowing high binder content for more dura- ble mixture. • Relatively inexpensive. • May be made from a variety of plant materials. • Widely available. • May be from recycled materials such as newsprint. • High binder absorption increases binder cost. • Not strong in tensile mode. Mineral • Stabilizes binder in open- and gap-graded SMA mixtures. • Not as absorptive as cellulose. • Electrically conductive fibers have been used for inductive heating for deicing purposes or to promote healing of cracks. • Some may corrode or degrade because of moisture conditions. • May create harsh mixes that are hard to compact and may be aggressive, causing tire damage if used in surfaces. Polyester • Resists cracking, rutting, and potholes. • Increases mix strength and stability. • Higher melting point than polypropylene. • High tensile strength. • Higher specific gravity means fewer fibers per unit weight added. • Cost-effectiveness not proven/varies. Polypropylene • Reduces rutting, cracking, and shoving. • Derived from petroleum, so compatible with asphalt. • Strongly bonds with asphalt. • Disperses easily in asphalt. • Resistant to acids and salts. • Low specific gravity means more fibers per unit weight added. • Lower melting point than some other fiber materials requires control of production temperatures. • Begins to shorten at 300°F. • Cost-effectiveness not proven/varies. Aramid • Resists cracking, rutting, and potholes. • Increases mix strength and stability. • High tensile strength. • May contract at higher temperature, which can help resist rutting. • Cost-effectiveness not proven/varies. Aramid and polyolefin • Controls rutting, cracking, and shoving. • Combines benefits of aramid and polyolefin (polypropylene) fiber types. • Cost-effectiveness not proven/varies. Fiberglass • High tensile strength. • Low elongation. • High elastic recovery. • High softening point. • Brittle. • Fibers may break where they cross each other. • May break during mixing and compaction. • Cost-effectiveness not proven/varies. Source: Literature review.

9 The individual types of fibers can have various structures and cross-sections. Chen and Xu used scanning electron microscopy to investigate the structure of some fibers, including asbestos, lignin (cellulosic), polyacrylonitrile, and polyester. They found that the synthetic fibers had “antenna features” at their ends that helped anchor them in the binder phase, creating a stronger network within the binder. The asbestos fibers had a smooth texture and a thin diameter, yielding a large surface area. The cellulosic fiber had a rough texture, and the diameter varied along the length of individual fibers. All the fibers were found, in this laboratory study, to increase binder stiffness, rutting resistance, and flow resistance. The asbestos and cellulosic fibers absorbed more binder (Chen and Xu 2010). Surface coatings are frequently applied to fibers during production for various reasons. Some reduce static, while others help with the manufacturing and packaging processes. Excess static can make handling and mixing fibers difficult. Some coatings, such as stain-resistant coatings used for carpet fibers, may not be compatible with asphalt binder. In a study by Putnam (2011), one fiber source was used with different binder grades and sources to explore issues related to the compatibility of the binder and the fiber coating. In another part of the study, one binder was used with polyester fibers with different coatings. The research revealed that different binder (crude) sources had different compatibilities with fiber finishes. Chromatography testing suggested that the fiber coating could affect the absorption of different fractions of the binder, particularly lighter fractions. The study also found that different finishes can affect binder properties, particularly tensile properties. However, the amount of finish applied to a fiber did not cause any significant differences (Putnam 2011). Methods of Testing Fibers and Fiber Mixtures The methods of testing fibers have largely come from the textile industry and may vary for different types of fibers because of their historical use. Properties of interest include the dimensions of the fibers, ash content, shot content, and properties related to compatibility with asphalt. For use on open- and gap-graded mixtures, the effects of the fibers on binder draindown are of primary interest; several variations of a draindown test are in common use. Mixtures are also frequently tested for abrasion resistance in the Cantabro test. The physical dimensions of the fibers are important because they can affect how well the fibers can disperse and interact with the other components of the mixture. For example, the lengths of the fibers can be modified to relate to the maximum aggregate size in the mixture; smaller aggregate sizes may use shorter fibers. Long fibers may be difficult to mix uniformly into the mixture in the lab or plant because they can get tangled and clump together (Tapkin et al. 2009, 2010; Do Vale et al. 2014). Sieve analysis is also sometimes used to characterize fiber size. This analysis can be performed using wire mesh screens or a device called an Alpine Air Jet Sieve, which fluffs the fiber and sieves it using a vacuum. The denier of a fiber is a measure of its fineness; it is an artifact of the textile industry. The denier is the weight in grams of 9,000 m of fiber. Fine silk fibers have a denier of 1.0. The denier relates to the surface area of fiber, which in turn is related to the potential asphalt demand and holding power in a mixture. The ash content is a measure of the organic content of plant-based fibers and is, therefore, a method of ascertaining that the fiber is organic. A small amount of fiber is heated to burn off the organic material. The remaining ash is the nonvolatile portion. ASTM D128 includes an ash determination procedure that is sometimes cited in organic fiber specifications. The shot content is a parameter frequently specified for mineral fibers. When mineral fibers are produced, there may be small globules of mineral material called shot. These globules do not contribute fibrous material, so the amount is typically limited. To determine the shot content, the fibers are placed in a nest of sieves and shaken in a shaker machine or by hand to separate the fibers (which are retained on the sieves) from the shot (which passes through). Shot content limits in specifications also report the applicable sieve size used to separate the fibrous and shot material. ASTM C1335 is one method used to measure the shot content. The compatibility and binding of organic fibers with asphalt binder are usually controlled by properties of the fiber, including pH, oil absorption, and moisture content. These properties are less often specified for synthetic fibers, which are sometimes petroleum based and therefore assumed to be compatible with asphalt. [However, some coatings that can be applied to fibers may not be compatible with the binder (Putnam 2011).] The pH is typically measured by soaking fiber in distilled water, then measuring the pH of the water with a pH meter. The oil absorption is an indication of the compatibility of the fiber with asphalt and is determined, usually, by suspending a measured amount of fiber in mineral spirits for a short time (typically 5 minutes), removing the fiber from the mineral spirits and shaking it to remove excess mineral spirits on the surface, then measuring the change in mass of the fiber. Oil absorption is reported in terms of how many times the mass of fiber is absorbed; for example, it may be specified that the fiber absorb five times its own mass. Gap- and open-graded mixes (SMA and porous) are generally tested to determine the percentage binder draindown. There are various methods for measuring this property. Stuart and Malmquist (1994) compared three test methods. The same compaction temperature was used in all three cases; this was 170°C, which is a common German plant discharge temperature.

10 • The German draindown test involves placing about 1 kg of mixture in a glass beaker after mixing, then covering the beaker with foil and holding it in an oven at the compaction temperature for 60 ± 1 min. Then the beaker is turned over, allowing the mix to fall into a tared bowl. The difference in the mass before and after storage, expressed as a percentage of the original mass of the mixture, is the draindown loss. The maximum allowable loss using this method is 0.3%. • The FHWA draindown test also involves holding a sample of mix at the compaction temperature for 60 ± 1 min. In this method, however, about 1 kg of mix is placed in a 2.36-mm sieve set on top of a bowl. After storage at the compaction temperature, the difference in the mass of the bowl reflects the mass lost through draindown. This difference, expressed as a percentage of the original mass of the mixture, is the percentage loss. • A third draindown test was developed by FHWA for open-graded friction courses. As with the other tests, 1 kg of mix is held for an hour at the compaction temperature, but in this case the mix is spread in a Pyrex pie plate before putting it in the oven. The mix is allowed to cool after the storage period, then the pie plate is turned over or inspected from below to determine how much binder has accumulated on the bottom of the plate. Five standard photographs illustrate different degrees of draindown; a visual comparison to these photos is used to assess draindown tendencies of the subject mix. Figure 1 shows three images used by the South Carolina DOT to illustrate below optimum, optimum, and above optimum binder contents for open-graded mixtures from its draindown test (SCDOT SC-T-91). FIGURE 1 Example of images demonstrating binder contents from draindown test (Source: South Carolina DOT, Test Method SC-T-91). The Austroads draindown test method is similar to the German method in that 1 kg of mix is placed in a tared glass beaker, then held for 60 ± 1 min before the beaker is turned upside down, allowing the mix to fall out. In the Austroads method, the oven temperatures are specified according to the mix type (open-graded or SMA) and whether or not a modified binder is used; the specified temperatures range from 160°C for an unmodified OGFC to 185°C for a modified SMA. In addition, a supplementary procedure can be used with polymer-modified binders if the amount of binder remaining in the beaker is more than 0.3% of the original mass of the mix. This procedure involves using a solvent to wash the residue from the beaker through a tared 0.600- mm sieve to determine whether a significant amount of the fine aggregate particles was trapped in the modified binder adhering to the beaker (Austroads 2006). As part of an effort to refine the mix design process for OGFC mixtures, Watson et al. (2003) explored various test methods and sample preparation techniques using mineral fiber with three different binder grades, modified and unmodified. The mineral fiber was added at 0.4% based on the total mass of the mix. Watson et al. (2003) compared the use of two different-sized sieves (4.75 mm and 2.36 mm) for the draindown test based on AASHTO T 305-97, because they suspected that the 4.75-mm sieve was too large for some finer OGFCs, allowing stone loss. Another modification was weighing the basket after removing the mixture to measure any asphalt remaining in the basket. The draindown testing was performed on mixes with PG (performance grade) 64-22 with and without mineral fiber, polymer-modified PG 76-22 with and without fiber, and rubber-modified PG 76-34 with fiber. The presence of fibers in the mix was the most significant factor related to the amount of draindown. Fibers reduced draindown dramatically: mixes without fibers had draindown amounts as high as 3%; adding only 0.4% fiber could reduce the draindown to minimal amounts. Adding the mass of asphalt clinging to the basket did not significantly affect the draindown estimate for the mixes tested. Both sieve sizes were found to be acceptable, but the 2.36-mm sieve had a lower standard deviation in the results. AASHTO T 305, Determination of Draindown Characteristics in Uncompacted Asphalt Mixtures, is now the most widely accepted test method in the United States for determining draindown (survey results). Similar to the methods discussed earlier, the difference in mass before and after oven storage is used to determine draindown. In this method, however, the mix is held in a woven wire basket, allowing the binder to drain off through the wire. Figure 2 illustrates this test in progress. Another test that is commonly used, particularly with open-graded mixtures, is the Cantabro test. In this test, compacted mixture specimens are tumbled in a Los Angeles abrasion test drum, without the steel balls used in the LA abrasion test. The change in mass before and after testing is an indication of the durability of the mixture. Fibers have been reported to improve this durability in some cases (Lyons and Putnam 2013). Watson et al. (2004) confirmed the suitability of the Superpave gyratory compactor to prepare specimens for Cantabro testing in place of the Marshall compacted specimens originally used. Attempts to test fibers as part of the binder phase have been fraught with difficulty. The binders with fibers tend to

11 crawl out of rolling thin film oven bottles during conditioning (Brown et al. 1996). In addition, in fabricating samples for bending beam rheometer or direct tension testing, it is difficult to get smooth samples with uniformly distributed fibers (Ayesha Shah, Purdue University communication, June 15, 2014). FIGURE 2 Draindown test in progress (Source: WSDOT). One recurring question from agencies is whether there is a test to detect the presence and uniformity of the distribution of fibers (Austroads 2007). Fibers may be visible in the mix, as illustrated in Figure 3, but this is not a quantifiable means of determining the presence of fibers. Comparisons of reported maximum specific gravity values reported in the literature for fiber-modified and control mixtures show that the values are not significantly different and cannot be used to detect the presence of fibers. Sometimes the presence of fibers can be detected using a solvent extraction and sieve analysis; fibers typically remain on the sieves. However, it can be difficult to determine whether the required amount of fibers has been added because of the small amounts of fiber per ton and the small size of the extraction sample. Also, this method does not assess the uniformity of distribution. Huang and White compared three fiber extraction techniques to assess the fiber content in Indiana mixtures with polypropylene fibers. The first technique used trichloroethylene to wash the asphalt binder and fiber from the mixtures, then the solvent was filtered to separate the fiber. Some fines were also removed in the process, so the material on the filter was ashed, which removed the fiber. Fiber content was determined on the basis of the difference in mass before and after ashing. The second procedure used water to float the fiber (and some fines) from the aggregate remaining after extraction using ASTM D2172. Ashing was again used to remove the fibers from the fines so that the weight of fibers could be determined. The final technique used sieve analysis (ASTM C136-84a) to remove fibers from the aggregate remaining after extraction using D2172. The fibers were caught on the sieves. As in the previous methods, ashing was used to separate the fiber and fine aggregates. Each method has its uses, depending on the kind of information needed. The first method would be an easy way to determine fiber content only. The second method also allows determination of binder content, while the third would be useful if aggregate gradation is needed (Huang and White 1996). FIGURE 3 Fibers visible in mixture clinging to shovel (Source: PennDOT). Research under way in Idaho will explore the use of X-ray computed tomography to compare the distribution of fibers in laboratory and field samples (see the case example in chapter four that describes ongoing research on fibers in dense-graded asphalt). MIX DESIGN WITH FIBERS Little appears in the literature regarding mix design procedures with fibers. When the topic is mentioned, it is usually as a brief introduction to sample preparation as part of a larger study. In general, mix design proceeds as usual with the addition of a draindown test (see state specifications and test methods in chapter three). Over time, the appropriate fiber content to use in a given mixture type may become almost standard for certain fiber types; for example, 0.3% by weight of mix is a very common addition rate for cellulose fibers in SMA. However, use of too high a fiber content may make compaction more

12 difficult, leading to higher air void contents in laboratory- or field-compacted mixtures (Serin et al. 2012; Crispino et al. 2013; Morova 2013). Chowdhury et al. (2005) found that using 1% of 6-mm-long recycled tire fibers stiffened SMA and porous mixes too much and made the mixes more susceptible to cracking. Binder contents are frequently higher in fiber mixes, especially with more absorbent cellulose or other plant- based fibers (Busching et al. 1970; Button and Hunter 1984; Toney 1987; Freeman 1989; Fortier and Vinson 1998; Cooley et al. 2003; Chen et al. 2009; Gibson et al. 2012). Absorption is a beneficial property to some extent because it allows for increased binder or mastic contents in the mixtures, which can aid durability, especially for gap- and open-graded mixes. Excessive absorption, however, may lead to expensive mixes. In addition to higher absorption, fibers may have high surface areas that need to be coated with asphalt. In a 2012 FHWA report, Gibson et al. summarized the results of a previous study comparing polymer-modified and a fiber-reinforced dense hot mix. The optimum asphalt contents for the various mixes are shown in Table 2. Work conducted by FHWA under NCHRP 90-07 demonstrated the increase in binder content associated with the use of fibers. Optimum binder contents were determined at 4% air voids at 75 gyrations in the Superpave gyratory compactor at a fixed compaction temperature of 140°C. Table 2 shows that the fiber mix had an optimum binder content 0.5% to 1.0% higher than the polymer-modified mixes and 0.8% higher than the unmodified control section (Gibson et al. 2012). TABLE 2 OPTIMUM BINDER CONTENTS FROM FHWA STUDY Modifier Optimum Binder Content (%) Terpolymer 4.4 Ethylene vinyl acetate (EVA) 4.4 SBS-LG 4.5 SBS radial grafted 4.6 Ethylene styrene interpolymer 4.6 Air-blown asphalt 4.8 Chemically modified crumb rubber 4.9 Unmodified PG 70-22 4.6 Unmodified PG 70-22 with 0.3% polyester fiber (by aggregate mass) 5.4 Source: Gibson et al. (2012). PRODUCTION, CONSTRUCTION, AND ACCEPTANCE OF FIBER MIXTURES Production of fiber mixes can be accomplished in different ways. Fibers can be added to the liquid binder in a process called wet-mixing or can be added to the aggregate in a dry-mixing procedure (Abiola et al. 2014). Fibers are often blown into the hot mix plant to help ensure uniform distribution (Figure 4) but are sometimes added to the plant in bags (Figure 5). Fibers can be added to drum plants through the reclaimed asphalt pavement collar, where they can be mixed with aggregate before the binder is added (Ryan Barborak, TxDOT communication, Sept. 2, 2014). In batch plants, fibers are added to the weigh hopper or pugmill (Shoenberger 1996; Watson et al. 1998). The addition rate must be coordinated with the plant production rate (metered) to ensure that consistent mix is produced (Schmiedlin 1998). FIGURE 4 One example of blowing fibers into drum plant: (a) fiber hopper, (b) blowing equipment, (c) feeding fibers through RAP collar (Source: PennDOT). (a) (b) (c) FIGURE 5 Adding fibers in premeasured bags at a batch plant (Source: R.S. McDaniel). Clumping of fibers in the mixture was reported in both the literature and the survey results. Figures 6 and 7 show examples of clumps of fibers found during and after mix compaction. Keeping the fibers dry is reportedly important to help prevent clumping and clogging of fiber injection equipment. In many cases, clumps can also be avoided by

13 increasing the mixing time (Watson et al. 1998; Cleven 2000), which is more easily accomplished in a batch plant. With some of the more brittle fiber types—such as steel wool and glass fibers—increasing the mixing time may cause the fibers to break, resulting in shorter fibers (Garcia et al. 2012a). In some cases, clumps of fibers have been observed when mix is discharged from the plant, but by the time mix is transferred into a silo, into haul trucks, and through a material transfer device, the clumps have dissipated (Nelson Gibson, FHWA communication, Aug. 25, 2014). FIGURE 6 Fiber clump (Source: PennDOT). FIGURE 7 Clump of fibers on surface of compacted mat (Source: PennDOT). Plant temperature also needs to be controlled to prevent thermal degradation of the fibers. Polypropylene fibers, for example, melt at a lower temperature than polyester fibers (163°C vs. 249°C), so more control of production temperatures is needed. This would be less of an issue with warm mix asphalt (WMA), but there are few examples of using fibers at WMA temperatures. Fiber mixes have also been reported to be “stickier” than some other mixes and may make hand work more difficult. For example, a dense mix with ¼-in.-long polyester fibers reportedly stuck to pneumatic tired rollers, hand tools, rakes, and other equipment (Toney 1987). On the positive side, some polymer fiber mixes have demonstrated a reduced tendency to segregate, presumably because of their stickiness and the fiber network (Jiang and McDaniel 1992). Few additional mixture tests were cited in the literature for routine production control or acceptance; conventional tests were generally found to be adequate. Of course, many additional types of tests have been performed for research purposes, as will be outlined in reference to specific projects later in this chapter. Busching et al. (1970) cautioned that testing fiber mixtures in compression only will not adequately characterize the mixture properties; some sort of cracking or flexural testing is also needed. PERFORMANCE OF FIBER MIXTURES Most of the literature reviewed for this synthesis was related to the performance of fiber mixes in the laboratory or the field. Individual projects compared different types of fibers in a variety of applications. Some of the most pertinent references are summarized here. (See the Bibliography for related literature not cited in this report.) Use of Fibers in Dense-Graded Mixtures In 1974, a badly cracked section of continuously reinforced concrete in Indiana was overlaid with approximately 2 in. of asphalt binder and 2 in. of surface. Six years later, the overlay was seriously deteriorated and needed to be overlaid again. At that time (October 1980), the Indiana Department of Highways (DOH, now Transportation) decided to place a test section of overlay containing polypropylene fibers to determine what benefits, if any, the fibers could provide. The 2-in. overlay was placed over the existing pavement (without milling). The fiber mix was placed in about 990 ft of the passing lane and an adjacent 550 ft of the travel lane. The remainder of the travel lane did not include fibers and served as the control. The fibers were added into the batch plant at a rate of 6 pounds per ton of mix (0.3%) (Galinsky 1984). After 2.5 years, the control section was exhibiting more than twice the amount of cracking observed in the fiber section and the cracks were of much greater severity [moderate to high severity, as defined in the Distress Identification Manual (FHWA 2003)]. Those cracks that did appear in the fiber section were less than 1/32 inch in width and had very little secondary cracking (braiding). Cracks in the control section were about ¼ in. wide, with some as wide as 3/8 in. or more. Significant secondary cracking was also observed. The control section was also “severely

14 deformed” with ruts as deep as 2¼ in. Rut depths in the fiber section were all less than 3/8 in. and averaged about 1/8 in. (Galinsky 1984). On the basis of the performance of the fiber mix, the Indiana DOH began using polypropylene fibers in more situations. In 1985, a new researcher visited the site of the experimental overlay to check on its condition after another year had passed. In May 1985, the control section had deteriorated to the point of functional failure. The rutting was severe, more than 2.5 in., and extreme plastic deformation (shoving) was seen. The extremely rough profile could have caused a loss of control of a vehicle, possibly throwing it into a bridge railing just south of the control section. The profile was so bad that truck drivers familiar with the roadway would merge into the passing lane to avoid the short control section (McDaniel 1985). The test section, on the other hand, had minimal rutting and no shoving. Cracks formed in the control section were numerous, wide and braided. Fewer, tighter cracks were observed in the fiber section, with very little braiding. Some cracks in the control section stopped at the joint and did not propagate into the test section, despite being quite severe. Because of the hazardous condition of the control section, a letter was sent to the district engineer and immediate action was taken to mill off and replace the control section (McDaniel 1985). Jiang and McDaniel (1992) evaluated the 8-year field performance of various thicknesses of asphalt overlays with and without cracking and seating the existing concrete and with and without polypropylene fibers in the intermediate and base layers of the overlays. The fibers were 10 mm long and added at 0.3% by weight of the mixture. They found that adding fibers to the base and intermediate layers of the conventionally overlaid section did not reduce cracking, because the cracking was mainly reflective cracking caused by substantial vertical and horizontal movements of the underlying concrete. However, the use of fibers did delay and reduce cracking on the cracked and seated sections. There was no apparent difference between the cracked and seated sections with fibers in the base versus the base and intermediate layers, suggesting that use of fibers in the base alone was effective at reducing cracking. Rutting was not severe on any of the sections but was lower on the sections with fibers. A later study in Indiana compared the field performance of seven asphalt additives or modifiers. In this case, polyester fibers were used in an asphalt overlay over jointed concrete pavement. The fibers were 6.35 mm long, added at a rate of 0.3% by weight of the mix, and dry mixed for 30 s, then wet mixed for 35 s in a batch plant. The other modifiers were various polymers, gelled asphalt, and crumb rubber. This study evaluated the ability of these modifiers to control both rutting and cracking. The styrene butadiene rubber (SBR), polymerized asphalt cement (PAC), and asphalt rubber mixtures were the most effective in terms of resisting cracking. Polyester fibers also performed well but had slightly more cracking than the top-tier performers. None of the mixes—not even the control section—demonstrated significant rutting under heavy interstate traffic. This finding suggested that additives were not necessary to achieve good performance; attention to detail and good construction practices could be sufficient (McDaniel 2001; McDaniel and Shah 2003). Maurer and Malasheskie (1989) compared the field performance of four fabric interlayers, one fiber-reinforced asphalt interlayer (stress-absorbing membrane interlayer, or SAMI), and a fiber-reinforced asphalt overlay to a control section with an unreinforced overlay. The objective was to identify treatments that could reduce reflective cracking in asphalt overlays. (For the purposes of this synthesis, this review will focus on the comparison of the fiber-reinforced SAMI and the fiber overlay to the control section.) The fiber- reinforced overlay consisted of a 38-mm wearing course with the addition of 0.3% (by weight of mix) polyester fiber. This treatment was included in the study because, if it performed as well as or better than the other treatments, it would be easier to implement and construct at a lower cost (Maurer and Malasheskie 1989). Test sections were constructed in 1984 on a principal arterial highway in Pennsylvania that exhibited a stable base in most areas and surface block cracking. The dense-graded mixes were produced in a batch plant. The underlying pavement was mostly portland cement concrete, except for a section in the center of the roadway that had previously held a trolley line and was paved with asphalt when the trolley was abandoned (Maurer and Malasheskie 1989). The fiber-reinforced SAMI was applied full-width as an alternative to placing paving fabrics. The membrane consisted of AC-20 with the addition of proprietary fibers at a rate of 6.0% by weight of asphalt along with an adhesion promoter (2.0% by weight of asphalt). The manufacturer used specialized equipment to heat, blend, and place the membrane. The paving contractor applied a layer of stone cover after membrane installation. Construction issues caused significant delays, especially on the first day of membrane placement. The issues included suspected contamination of the AC-20, steam from water in the adhesion promoter, and difficulty controlling the membrane application (Maurer and Malasheskie 1989). The fiber overlay consisted of a standard Pennsylvania Department of Transportation (PennDOT) surface mix with the addition of 0.35% fiber, equivalent to 3 kg of fiber per tonne of mix. The polyester fibers, in bags, were added to the batch plant at the beginning of the dry mix cycle. No construction difficulties or need for extra manpower

15 were encountered in placing this material (Maurer and Malasheskie 1989). After 44 months in service, the fiber-reinforced overlay was outperforming all the other treatments, with less reflective cracking (a reduction of more than 50% compared with the control). The fiberized membrane was the second most effective, with a reduction of 46.4% relative to the control. All the treatments provided some reduction in reflective cracking. Life cycle cost analysis, however, showed that the increased construction costs could not be justified at the time. In conclusion, the 1989 report recommended none of the treatments for implementation because of their lack of cost-effectiveness, but it did recommend that the test sections be revisited in 3 years to verify the life cycle (Maurer and Malasheskie 1989). Huang and White (1996) tested cores and slabs taken from polypropylene fiber-modified asphalt overlays in the state of Indiana. At the time of the research, the Indiana Department of Transportation (INDOT) used fibers extensively in overlays, a practice that had been shown to be effective at reducing reflective cracking over cracked and seated pavements and concrete pavements undersealed with asphalt. A total of 33 test sections were constructed on two high-traffic roadways in 1990. These test sections were cored, and slabs were cut for lab testing. In addition to control sections without fibers, the test sections included varying amounts of fibers in different pavement layers (Huang and White 1996). Testing included fatigue testing of beams cut from the pavement slabs and complex modulus testing on cores. In addition, bulk-specific gravities, maximum specific gravities, and sieve analyses were performed on samples from the pavements. The sieve analysis results (determined using the third testing method described earlier) showed that the actual fiber contents in the plant-produced mixes varied from the target values in most cases, probably owing to difficulties in feeding the fiber into the mixture at the discharge chute. The contents varied by 4% to 43% from the target; in most cases the fiber content was low, but in one case it was nearly 22% higher than designed (Huang and White 1996). The aggregate gradations and asphalt contents for all of the mixes, with and without fibers, were within specifications, but the field densities were low. The air void contents of the fiber mixes were higher than those of the controls, suggesting that the fibers made compaction more difficult (Huang and White 1996). Beam fatigue testing showed that the use of fiber extended the fatigue life of the overlays by as much as two times. Dynamic modulus testing results from one project indicated that the presence of fibers decreased the modulus but did not affect the phase angle (Huang and White 1996). A study by the Oregon DOT reported on 10-year performance of fiberized and polymer-modified test sections placed in 1985 after the application of more than 1.5 to 1.7 million equivalent single-axle loads. Additives were placed in the dense-graded top course (38 to 51 mm thick) over an unmodified base course (102 to 114 mm) over an existing pavement with severe alligator and thermal cracks. One section included polypropylene fibers and another included polyester fibers, both with AC-20 binder. There were two control sections and six sections with mixtures incorporating various anti-strip and polymer additives. Both fiber sections were comparable to the controls, with average rut depths of 13 to 16 mm. The polypropylene fibers performed much better than the control in terms of block cracking, while the polyester fibers performed better than the control. The polypropylene fiber had no block cracking, and the polyester had block cracking over less than 10% of the travel lane. The control section exhibited block cracking over between 30% and 50% of the travel lane. The fiber sections performed comparably to the controls in fatigue cracking (Edgar 1998). The Pennsylvania DOT was experiencing rutting problems on asphalt roadways in the 1980s, so it embarked on a field evaluation to explore various modification techniques that could improve the rutting performance. Anderson et al. (1999) reported on the 10-year performance of test sections constructed on I-80. Retained samples of the component materials were tested using the new Superpave protocols as well (Anderson et al. 1999). The field test sections included an unmodified control, four different polymer-modified binders [polyethylene, ethylene vinyl acetate (EVA), styrene butadiene styrene (SBS), and styrene butadiene (SB)], Gilsonite, and polyester fibers. The test sections consisted of an overlay over jointed concrete pavement. The modifiers were placed in the 63.5-mm-thick intermediate course and the 38-mm-thick surface course but not in the 75-mm-thick base course. The mixes were designed under the Marshall mix design procedure (Anderson et al. 1999). The fibers were added at 0.280% by weight of the mix into the mixing chamber of the batch plant during the dry mixing period. The dry mixing time was increased by 10–15 s to ensure that the fibers were uniformly distributed. No construction problems were noted (Anderson et al. 1999). The researchers attempted to add the fibers to the binder and test in the dynamic shear rheometer (DSR) and bending beam rheometer (BBR) but experienced great difficulty in preparing the specimens and ensuring that the fibers were well dispersed in them. This testing was abandoned and not reported (Anderson et al. 1999). Loose samples of some of the modified mixes, including the fiber mix, had been collected during construction. These

16 retained samples were reheated and used to compact gyratory specimens for testing in the Superpave shear tester (SST) and the indirect tensile (IDT) tester. IDT testing indicated that the tensile strengths of all of the modified mixes were similar, with the two mixes with polymer-modified binders being slightly stronger than the Gilsonite and fiber mixes [3.5 vs. 3.4 MPa (510 vs. 490 psi), respectively]. The critical cracking temperatures for the fiber mix and the SB mix, however, were about 3°C cooler than for the Gilsonite and EVA-modified mixes (Anderson et al. 1999). SST testing showed that the fiber mix had the third highest modulus, after the Gilsonite and EVA mixes. The frequency sweep and phase angle data suggested that the fiber and SB mixes would perform best in terms of fatigue cracking. The field performance confirmed this, as the fiber and SB test sections exhibited less severe fatigue cracking than the Gilsonite and EVA (Anderson et al. 1999). Field performance showed that all the modified mixes performed well in terms of rutting; the control section also performed well, but had more rutting than the modified sections. The fiber mix did not show any secondary cracking around the sawn and sealed joints after 8 years. The Gilsonite section, in particular, showed excessive raveling and was replaced by PennDOT after 9 years. Top-down longitudinal cracking was observed in three sections but not in the fiber- reinforced section (Anderson et al. 1999). In a laboratory study at Clemson University, continuous polyester fibers from roofing manufacturing trim waste were shredded to two lengths [6.35 mm and 12.7 mm (1/4 in. and ½ in.)] in a paper shredder to see whether they could be used as a replacement for other fibers in dense asphalt mixtures. The indirect tensile strength and moisture sensitivity of mixtures with these fibers at different addition rates (0.35% and 0.50% by weight of the mix) were evaluated (Anurag et al. 2009). The optimum asphalt content of the fiber mixes was higher than that of the control mix because additional binder was needed to coat the high surface area of the fibers. The air voids and voids in mineral aggregate (VMA) of the fiber mixes were also higher than those of the control. Marshall stability and flow values increased with the addition of fibers. The tensile strength, tensile strength ratio, and toughness of the fiber mixes were higher than those of the control. The shorter fibers at the higher addition rate (6.35 mm at 0.50%) were found to perform best in this study (Anurag et al. 2009). Kaloush et al. (2010) compared the performance of a dense asphalt mixture containing 1 lb/ton of polypropylene and aramid fibers to a control mixture without fibers. The plant-produced mixes, from a paving project in Tempe, Arizona, were sampled for later testing at Arizona State University using current characterization tests. Overall, the fiber-reinforced mixture outperformed the control mixture, specifically: • The fiber mix exhibited higher peak stress and higher residual energy in triaxial shear testing compared with the control mix. This was attributed to the reinforcement effect of the fibers and greater resistance to shear failure and rutting. • The fiber mix demonstrated flow numbers 15 times greater than the control mixture in the repeated load permanent deformation test. The results suggested that the fiber mix could store more energy than the control, which again indicated greater resistance to permanent deformation. • In the dynamic modulus test, the fiber mix developed higher moduli than the control mix at all temperatures and frequencies, though the difference between the mixes was greater at high temperatures than at low temperatures. At high temperatures, the effects of the aggregate and fiber structure dominated over the binder properties, which were dominant at lower temperatures. Since the two mixes contained the same binder, differences were lessened at low temperatures. • The fatigue testing results were mixed depending on strain level and temperatures. At 40°C the fiber mix had a longer fatigue life, but at 70°C the fatigue lives of the two mixes were similar. At 100°C and high strain levels, the control mix had higher fatigue life; at lower strains, the fiber mix was superior. This behavior was explained by comparing the tensile strength of the fibers with the bonding strength between the fiber and the binder; at high temperatures, the strength of the bond determined the reinforcing effectiveness. • Indirect tensile testing at low temperatures (0°C, -10°C, and -20°C) indicated that the fiber mix would be more resistant to thermal cracking, with a strength 1.5 times greater than the control. In addition, the fiber mix demonstrated higher fracture energy, which relates to reduced thermal cracking. • Fracture mechanics analysis (C*-integral) indicated that the fiber mix would be much more resistant (40 times more) to crack propagation than the control. It was noted that the control samples tended to split open during testing while the fiber-reinforced samples did not (Kaloush et al. 2010). Overall, the lab characterization tests generally showed that the fiber mix would perform better than the control in resisting permanent deformation and thermal cracking, and would sometimes perform better in fatigue. The applicable test results were used as inputs into the Mechanistic- Empirical Pavement Design Guide (MEPDG) software to estimate the effects on pavement performance under different traffic conditions and in different pavement layer thicknesses. Using the MEPDG to simulate the effects of different layer thicknesses, the researchers estimated that the control mix

17 would need to be placed 2 in. thicker than the fiber mix (5.5 vs. 3.5 in. for the assumed conditions) to resist an equivalent amount of rutting (such as 0.4 in.) to develop. The fatigue analysis showed that the fiber mix would experience less fatigue cracking, though the difference in cracking varied depending on the layer thickness (Kaloush et al. 2010). After 2 years in the field, the control sections had “about three times the amount of low-severity cracking” as the fiber sections (Kaloush et al. 2010). A study by Xu et al. (2010) compared the laboratory performance of four fiber types (polyester, polyacrylonitrile, lignin, and asbestos) to a control with no fibers. The polyacrylonitrile, lignin, and asbestos fibers were added at 0.3% by mass of the mix; the polyester fiber was used at varying contents from 0.20% to 0.50% to evaluate the effect of addition rate on performance. Dense mixes were evaluated in terms of rutting, flexural strength and strain, fatigue, tensile strength, and resistance to freeze-thaw cycles. All four fibers reduced rutting in a one-fourth scale accelerated loading test, but the polymer fibers reduced it the most (19% and 32% at 2,500 cycles for the polyacrylonitrile and polyester, respectively, compared with 8.4% and 11.4% for the lignin and asbestos). Fibers also improved the flexural strength and ultimate strain at 0°C and -10°C. The lignin and asbestos fibers had slightly better performance than the synthetic fibers, probably because of their high surface areas and branched structure. Similarly, all the fibers increased the fatigue life of the mixture, with the polyacrylonitrile, polyester and asbestos performing better than the lignin (Xu et al. 2010). The polymer fibers also resulted in the greatest improvement in indirect tensile strength. After freezing and thawing, however, these fiber mixes were only slightly stronger than the control. The asbestos and lignin fiber mixes actually had strengths lower than the control after freezing and thawing. The relatively poor performance after freezing and thawing may be related to higher air void contents in these mixes (Xu et al. 2010). Bennert (2012), in a study for the New Jersey DOT, compared the performance of plant-produced mixtures with and without a combination of polyolefin and aramid fibers. Both mixtures were produced in the same batch plant on the same day using the same mix design; the 9.5-mm surface mix was designed for a traffic volume of 3 to 10 million equivalent single-axle load with 5.9% PG 64-22 binder. The mixtures were characterized using dynamic modulus (AASHTO TP 79), flow number (AASHTO TP 79), beam fatigue (AASHTO T 321), and cycles to failure in the overlay tester (TxDOT TEX-248F). Somewhat surprisingly, the results showed that the mixture without fibers had a much higher high temperature modulus than the mix with fibers—as much as two to three times higher. The mixture without fibers was only slightly stiffer than the fiber mix at low temperature. Analysis of the mixture phase angle from the same tests showed that the mix without fibers was more elastic than the fiber mix. Similarly, the flow number tests showed that the mixture without fibers had a greater resistance to rutting than the fiber mix, evidenced by exhibiting about five times the cycles to 5% strain (2,124 vs. 427). The beam fatigue test, which relates to the initiation of cracking, showed that the two mixtures had nearly identical resistance. The overlay test results, on the other hand, showed that the fiber mix had much greater resistance to crack propagation than the mix without fibers (174 cycles to failure vs. 6 cycles) (Bennert 2012). Since the mixture testing results were somewhat unexpected, the binders from the dynamic modulus samples were extracted, recovered, and tested in the dynamic shear rheometer (DSR) and multiple stress creep recovery (MSCR) tests to explore the possibility that differences in the binders were influencing the mixture results. The high performance grades of the binders (AASHTO M 320) were very comparable, with that from the no-fiber mix being slightly higher than that from the fiber mix (68.8°C vs. 67.4°C). The recoverable strain from the MSCR test (AASHTO TP 70) was almost four times greater for the no-fiber mix than for the fiber mix at 58°C and almost nine times greater at 64°C, despite the fact that both mixes reportedly contained the same PG 64-22 binder. This indication of much greater elasticity of the binder in the no-fiber mixture could help to explain some of the differences in mixture behavior; it led to the speculation that the binder used in the mix without fibers may have had some contamination with a polymer- modified binder. A small amount of polymer could increase the elasticity of the binder and mixture, leading to higher stiffness and improved rutting resistance (Bennert 2012). In another study, Bennert also compared the performance of a conventional 12.5-mm mixture produced for the New Jersey DOT with a similar mix containing aramid and polyolefin fibers, both using a PG 64-22 binder. Plant- produced mixes were sampled and tested in the laboratory (Bennert, n.d.). Dynamic modulus testing (AASHTO TP 79) of the plant-produced mixtures showed that the two mixes had similar stiffnesses at all frequencies. The fiber- reinforced mixture had slightly better resistance to rutting in the asphalt pavement analyzer (AASHTO T 340) than the control (2.70 vs. 3.14 mm). Likewise, results of testing in the repeated load flow number test (AASHTO TP 79) also showed slightly better performance for the fiber mix (959 vs. 747 cycles) (Bennert n.d.). On the other hand, overlay test results [according the NJDOT B-10 overlay test for determining crack resistance of hot mix asphalt (HMA)] showed that the mix without fibers had a statistically significant greater resistance to crack

18 propagation than the fiber mix after both short-term (194 vs. 129 cycles) and long-term (179 vs. 118 cycles) aging. The short-term aging was plant aging during construction; the long-term aging was in the laboratory according to AASHTO R 30. The beam fatigue test yielded similar behavior. The reason the mix without fibers performed better in fatigue than the reinforced mix is unknown, but it is possible that the addition of fibers reduced the effective asphalt content of the mix (Bennert n.d.). The FHWA included a polyester-fiber-reinforced test section in a study comparing the performance of modified and unmodified binders using the accelerated loading facility (ALF) to test fatigue and rutting resistance. The ultimate goal was to explore candidate binder tests to improve the performance grade specifications, especially regarding testing modified binders. This ALF study tested PG 70-22 (unmodified control), Arizona wet process crumb rubber (CR-AZ), catalytically air-blown binder, styrene butadiene styrene with ~3% linearly grafted styrene SBS polymer by weight (SBS-LG), terminal blend CR (CR- TB) (5.5% rubber plus 1.8% SBS), terpolymer (Elvaloy with 0.4 % polyphosphoric acid), PG 70-22 with polyester fiber, and SBS 64-40 (approximately 3.5% SBS in a soft- base asphalt). Dense-graded Superpave 12.5-mm mixtures with these binders were produced in a drum plant. The optimum asphalt content for the control mix was 5.3%, so the binder content was fixed at 5.3% for all the mixes except the Arizona CR, which is a gap-graded mix and has a binder content of 7.1%. This mix also had different aggregates, which may have contributed to the much higher binder content. The following are additional details about the mixes, test sections, and their performance: • The Arizona CR was placed at 50 mm over 50 mm of PG 70-22 mix. The control and other modified mixes were placed at 100 mm. Thicker 150-mm sections were placed using the control, two SBS-modified, air-blown, and terpolymer mixes for comparison of the effects of lift thickness. • The maximum specific gravity of all the mixes was comparable (2.699 to 2.705) with the exception of the gap-graded mix, suggesting that the presence of a small amount of polyester fiber does not affect specific gravity. • Loading was applied with the ALF device at a temperature of 64°C. The fiber mix withstood 125,000 wheel passes and reached a maximum rut depth of 12.50 mm. Rutting on the other 100-mm sections exceeded 12.50 mm in less than 50,000 wheel passes (with the exception of the CR-TB, for which testing stopped after 50,000 passes at a rut depth of 9.06 mm). The differences were not statistically significant except for the CR-TB; that is, the fiber performed similarly to SBS-LG, CR-AZ, control, and air-blown binders. It performed better than terpolymer. • The temperature was increased to 74°C and a different location was loaded. In this case, the fiber mix performed better than all the sections except the linearly grafted SBS and the CR-TB. Rutting in the fiber section was similar at the two different temperatures. • In terms of fatigue cracking, the fiber section performed second best after the Arizona CR, followed by the terminal blend CR, as measured by percentage of area cracked and cumulative crack length. • The fiber section was “very resistant to fatigue cracking.” It performed second only to CR-AZ as measured by load passes to surface crack initial, load passes to 25 mm cumulative crack, and load passes to 25% cracked area. • Laboratory fatigue testing results did not match the field performance of the fiber mix. • The authors concluded that the fiber mix demonstrated “very good fatigue cracking resistance … in the dense- graded mixture reinforced with polyester fibers. The fatigue cracking of this section was measurably better than those of the polymer-modified section even though a less-resistant unmodified asphalt binder was used in the mix. This was the second most effective performer of its thickness group behind the composite, gap- graded crumb rubber asphalt pavement. The presence of fiber had no significant beneficial or negative impact on rutting performance” (pp. 229–230). A 2009 paper reporting on the same experiment indicated that when microcracks appeared on the surface of the ALF section with fibers, they did not coalesce into a larger crack as was seen in the other sections (Kutay et al. 2009). In a study of materials from Cyprus, polypropylene (PP) and glass fibers were used together in dense-graded hot mix asphalt. The authors reported that a previous study compared the performance of different types of fibers and concluded that glass and PP fibers outperformed polyester and nylon fibers. Therefore, they decided to try combining the two in this study. The concept was that the PP fiber would become “tacky around its melting point” to improve bonding and the glass fibers would provide a high stiffness. The PP fibers were blended into the liquid asphalt and the glass fibers were added to the aggregates; both fibers were 12 mm in length. The binder and aggregates were then mixed and specimens were prepared in the Superpave gyratory compactor (Abtahi et al. 2013). The content of PP fibers was 2%, 4%, and 6% by weight of the binder. These contents were combined with 0.00%, 0.05%, 0.1%, and 0.2% glass fibers by weight of the aggregate. For the PP in binder, softening point, penetration, and ductility tests were also performed. As the PP fiber content increased, the penetration and ductility decreased and the softening point increased (Abtahi et al. 2013). Marshall stability tests were performed on the mixtures with PP and glass fibers (as well as mix with no fibers and

19 with PP alone). As the PP content increased, the stability increased and flow decreased at each glass fiber addition rate. The addition of glass fibers at 0.05% and 0.1% increased the stability and decreased flow, but less improvement was seen when the glass fiber content increased to 0.2% (Abtahi et al. 2013). In terms of volumetric properties, increasing the fiber content increased the voids in the total mix and decreased the voids filled with asphalt, hence the VMA also increased. There was a decrease in unit weight of the mix with increasing fiber content. The optimum combination found in this study was 6% PP with 0.1% glass fibers. The stability for the combination was 25% higher than that for the unmodified control. The decrease in voids filled with asphalt was considered to be a benefit that would help the mix resist flushing in hot climates, where significant expansion of the binder would occur (Abtahi et al. 2013). Fibers in Stone Matrix Asphalt and Open-Graded Asphalt Mixes Stone matrix asphalt mixtures are gap-graded mixtures in which the voids in the mineral aggregate are mostly filled with asphalt mastic (binder, filler, and sometimes fibers). Open-graded mixtures, as their name implies, have open void space, which allows water to flow into and out of the mixture; therefore, these mixes are also called porous asphalt or permeable asphalt mixes. The main purpose of using fibers in these mixes is to control binder draindown; both will be discussed in this section. Stuart and Malmquist (1994) summarized the properties and purported benefits of using SMAs fairly early in the U.S. usage of this type of mix, after about 20 had been placed in the United States. On the basis of previous European experience with this type of mix in surface courses, it was expected that SMAs would perform better in terms of rutting under heavy traffic. These mixes are gap-graded with high coarse aggregate, binder, and mineral filler contents. Because of the lack of intermediate aggregates in the mixture, stabilizers are typically added to help retain the binder in the mixture; that is, to prevent draindown during production, transport, and laydown (Stuart and Malmquist 1994). Stuart and Malmquist reported on a study to evaluate the effects of different types of stabilizers in SMA, including loose cellulose fibers, pelletized cellulose fibers, loose rock wool fibers, and two polymers with AC-20 binder. Six stabilizers were evaluated in terms of their effects on mixture resistance to rutting, low temperature cracking, aging, and moisture damage, as well as draindown. Three of these stabilizers (one fiber and two polymers) were also used to construct an SMA surface on US-15 in Maryland. The two loose cellulose fibers evaluated were of domestic origin, while the pelletized cellulose and the rock wool were European. Because the SMA technology had been introduced to the United States from Europe, many of the early projects involved European stabilizers (Stuart and Malmquist 1994). The mixes were designed using the Marshall mix design with 50 blows per face. Then the mixes were evaluated in terms of draindown (by three methods: German, FHWA, and “pie plate”); resistance to rutting (three methods); resistance to low temperature cracking (two methods); resistance to aging (short- and long-term); and resistance to moisture damage (three methods) (Stuart and Malmquist 1994). The four fiber mixes evaluated in this study exhibited similar low amounts of draindown, but the polymer-modified mix had relatively high amounts of draindown and did not pass the German and open-graded friction course draindown tests. On the basis of the initial results, one loose cellulose and one loose rock wool fiber were dropped from further testing because they were expected to perform similarly to the remaining loose cellulose and pelletized cellulose fiber in the study (Stuart and Malmquist 1994). There were no significant differences in the resistance to rutting of the remaining two fiber and two polymer-modified mixes as measured by the Georgia loaded wheel test, the French pavement rutting test, and the gyratory testing machine. Similarly, there were no significant differences in the low temperature cracking resistance. The two polymer- modified mixes demonstrated less age hardening than the fiber mixes but were not effective at controlling draindown. It was noted that none of the actual mixes placed on US-15 exhibited any draindown during construction, despite the fact that the two polymers did drain down in the lab; the discrepancies between testing and performance were reported to be “difficult to explain.” After 18 months in the field, all the SMA sections were performing without noticeable distress (Stuart and Malmquist 1994). Brown et al. (1996) conducted a laboratory study of mortars for SMA mixtures using different fine aggregate types, two mineral fillers, modified and unmodified asphalts, and three types of fibers—cellulose, rock wool, and slag wool. The goals were to determine whether Superpave PG binder tests could be used to characterize SMA mortars and to determine how the components of the mortar affect performance. Fine mortar was defined as the binder plus stabilizer, mineral filler, and aggregate that passed through the 75-µm (#200) sieve and was considered for testing as a binder under the performance grade system. The total mortar was also tested in some cases; it included the fine mortar plus aggregates that passed through the 2.36-mm (#8) sieve. The fibers were added at 1.9% to 3.0% by weight of the mortar, which would be typical of the fiber content in the mortar fraction of SMAs (Brown et al. 1996).

20 Limited testing showed that fibers did not have a great effect on either the DSR or BBR results of the fine mortar. Cellulose stiffened the mortar slightly, but the rock and slag wools did not. There were difficulties in conducting the testing, such as mortars crawling out of the rolling thin film oven bottles, mortars not flowing in the pressure aging vessel pans, and complications in molding BBR and direct tension specimens (Brown et al. 1996). The results of testing both the total mortar and the fine mortar indicated that most of the stiffening came from the mineral filler; the fibers did little to stiffen the mortar at most temperatures. However, at high temperatures, such as those encountered during production and placement, the fibers did stiffen the mortar appreciably. This high temperature effect is credited with reducing the draindown during construction and may be the main reason to use fibers in SMA (Brown et al. 1996). In tests at the Nantes, France, test track, porous asphalts with mineral, glass, and cellulose fibers retained their high void content better than unmodified and polymer-modified overlays. In another experiment at Nantes, a very thin fiber- modified overlay showed excellent resistance to reflective cracking from an underlying fatigued pavement (Serfass and Samanos 1996). The authors identified the following benefits provided by the use of fibers: • Fixing the asphalt binder in the mix and preventing draindown; • Reinforcing the mastic (binder plus fibers); and • Reducing temperature susceptibility of the mastic because of the 3D network created. These benefits enable asphalt mixes to be designed that are rich in bitumen and therefore have increased durability, resistance to aging, resistance to fatigue and thermal cracking, and high stability (Serfass and Samanos 1996). Watson et al. (1998) summarized the Georgia DOT’s (GDOT’s) history of using open-graded friction courses. GDOT had used OGFCs for decades before banning them in 1982 after numerous problems with draindown, oxidation, raveling, and stripping of the pavement layer under the OGFC. Beginning in 1993, GDOT began using a modified 12.5-mm OGFC that included polymer-modified binder, fibers, and hydrated lime placed at 41 to 50 kg/m2 (75 to 90 lb/yd2). The use of polymer-modified binder and fibers reportedly allowed the buildup of thicker films coating the aggregates, which reduced weathering and early oxidation. Hydrated lime was added to both the OGFC and the underlying layers to prevent stripping. Mineral fibers were used at about 0.4% by weight of the mix to prevent binder draindown and increase mix strength. They also reportedly worked with the modified binder to increase film thickness; calculated film thicknesses in OGFCs with fibers were about 400% greater than those in conventional dense-graded mixes and about 30% to 40% greater than in previous OGFCs. Similar mixes are still routinely used in Georgia. A paper in 2000 by Cooley et al. compared the performance of cellulose with mineral fibers in OGFC. Cooley et al. credited a 1998 survey by Kandhal and Mallick as one impetus for their study; that survey reportedly revealed that many states specified mineral instead of cellulose fibers in OGFCs because of concerns that the cellulose would absorb water and cause moisture-related damage to the pavement. Cooley et al. conducted a field inspection of a 6-year- old Georgia DOT trial project that showed no significant performance differences between sections with cellulose and those with mineral fibers in terms of surface texture, rutting, cracking, and raveling. Cores from the field sections were tested for permeability; the cores with cellulose and cellulose with polymer-modified binder had the highest permeabilities, but the differences were not statistically significant. Differences in water absorption into Marshall compacted specimens did not appear to be significantly different for the cellulose and mineral fiber specimens, though the loose cellulose mixes did have the highest absorption. Mixes with loose cellulose, two cellulose pellets (pelletized with 34% and 20% asphalt), and mineral fibers did not perform differently in terms of tensile strength ratio (TSR). No visual stripping was observed in any of the mixes. Submerged asphalt pavement analyzer (APA) rut depths were low, but the loose cellulose mix did have lower rutting (5.2 mm at 8,000 cycles compared with 7.6 mm for the mineral fiber). The authors concluded that cellulose was as effective as mineral fibers and no moisture problems should be expected because of the use of cellulose. In another study, Watson (2003) inspected 13 SMA projects in five states after 5 to 10 years in service. On the basis of visual examination, he concluded that SMAs with fiber and unmodified binder performed as well as SMAs with polymer-modified binder. The types of fibers were not identified, but they probably included cellulose and possibly mineral fibers. Putnam and Amirkhanian (2004) compared the laboratory performance of cellulose, polyester (recycled raw materials), scrap tire, and waste carpet fibers (nylon) in an SMA with a PG 76-22 binder and granite aggregate. The waste carpet fibers were in the form of tufts of fibers and were added at 0.3% for each fiber type. An optimum asphalt content was determined for each mix. The synthetic fibers had lower optimum asphalt content than the cellulose because they were less absorptive. The mixes were evaluated in terms of draindown (AASHTO T 305), moisture sensitivity (ASTM

21 D4867 modified), and rut testing (APA). Draindown was determined at the optimum asphalt content and at higher contents to see how well the fibers could stabilize an excess amount of binder (Putnam and Amirkhanian 2004). The different types of fibers were equally able to prevent draindown at the optimum asphalt content. At higher binder contents, however, cellulose performed most effectively, followed by polyester, tire, and carpet fibers (the last two were comparable). The high stabilizing capacity of the cellulose was attributed in part to its higher absorption compared with the synthetic fibers (Putnam and Amirkhanian 2004). Although there were no significant differences in the wet or dry strengths of the fiber-reinforced mixes and the TSR values were all in excess of the minimum required, the cellulose fiber resulted in a lower mix toughness. The authors commented that the synthetic fibers would therefore be expected to bridge cracks better than the cellulose and might have a stronger bond with the asphalt binder (Putnam and Amirkhanian 2004). There were no significant differences in the ability of the fiber-reinforced mixes to resist rutting in the APA (Putnam and Amirkhanian 2004). Hassan et al. (2005), in a study for Oman, explored the effects of 6-mm-long cellulose fibers (0.4% by weight of mix), SBR-modified binder (4% SBR), and a combination of fibers and SBR compared with a control with no additives. The study found that the polymer was more effective at resistance to raveling in the short term, while both polymer and fibers improved the long-term resistance (resistance in an aged condition). Fibers reduced draindown more than polymer alone. Tayfur et al. (2007) compared the performance of unmodified and modified SMAs for their resistance to permanent deformation using indirect tensile strength, static and repeated creep, and wheel-tracking tests. The modifiers included granular amorphous polyalphaolefin, cellulose fibers, polyolefin, bituminous cellulose fiber, and styrene butadiene styrene. The researchers found that all the modified mixes had higher tensile strengths than the unmodified control, with the polyolefin and SBS having the highest tensile strengths. The SBS mixes had the greatest resistance to permanent deformation in the wheel-tracking test; the fiber mixes had some of the highest deformations in this test. The SBS mixes also had the highest resilient modulus among the modified mixes; the control had the highest resilient modulus at 5°C but not at 25°C or 40°C. Overall, the SBS mix performed most effectively; the fiber mixes did not perform particularly well (Tayfur et al. 2007). As the importance of sustainability has increased in roadway construction, it has also become an important consideration in airfield construction. Stempihar et al. (2012) conducted a lab and field study to explore the feasibility of using fiber-reinforced porous asphalt mixtures for airfield pavements. The addition of fibers was considered a potentially sustainable paving practice because they might improve the performance of the pavement. Airfield pavements in cool climates need to be able to withstand heavy loads from aircraft, extreme variations in temperatures, and snow plowing in winter; fibers could potentially help with all these issues. An increase in service life would also increase sustainability by reducing carbon emissions from maintenance and reconstruction, and from production of new paving materials. Use of recycled or waste fibers would also increase sustainability (Stempihar et al. 2012). This study compared the laboratory performance of fiber- reinforced asphalt concrete (FRAC) mixture samples from a paving project at the Jackson Hole Airport (JAC) in Jackson, Wyoming, with a control mixture without fibers. The control mix was reproduced in the laboratory using the same materials from a mix that was placed at the Sheridan County Airport (SHR) in Sheridan, Wyoming. The mixtures were evaluated in terms of dynamic modulus, fatigue, indirect tension, and Cantabro mass loss. A blend of polypropylene and aramid fibers was added to the batch plant at a rate of 1 lb/ton (0.5 kg/MT). The fibers were added to the hopper after the bag house so they would not be pulled into the bag house. Both mixtures used a PG 64-34 binder and similar binder contents (5.70% at JAC and 5.6% at SHR) and were open-graded mixes with a maximum aggregate size of 19 mm, conforming to the FAA P-402 porous friction course specification control points. The JAC mixture also included 0.75% hydrated lime (Stempihar et al. 2012). Confined dynamic modulus testing according to AASHTO TP 62-03 showed that the FRAC was significantly stiffer than the SHR mixture at higher temperatures, which should represent increased rutting resistance. There were no substantial differences in the dynamic moduli at lower temperatures (Stempihar et al. 2012). Beam fatigue testing (AASHTO T 321-03) showed that the fiber mix performed better in fatigue than the control mix at strain levels of 400 µm and 600 µm, but the performance was similar at 800 µm (Stempihar et al. 2012). Tensile strength testing was conducted on the mixtures at 0°C, 10°C, and 21.1°C according to AASHTO TP 9-02. The FRAC outperformed the control in terms of tensile strength, energy at fracture, and total energy. The authors noted that “although the specimen cracks, the fibers hold the specimen together, which requires more energy for the asphalt sample to fail” (Stempihar et al. 2012, p. 64). Finally, the Cantabro mass loss of the two mixtures was compared. In this test, specimens 100 mm (4 in.) in

22 diameter and 63.5 mm (2.5 in.) tall were tumbled in an LA abrasion drum (without the steel balls) for 300 revolutions. The difference in mass before and after testing is used to determine the percentage mass loss. The mixes performed similarly in this test, with mass losses of only 2.6% and 3.7% for the fiber and control mixes, respectively (Stempihar et al. 2012). The authors also examined the sustainability of fiber mixes through the estimated CO2 equivalent emissions. The emissions during construction of fiber-reinforced or -nonreinforced mixes would be similar, so any overall differences would arise from a difference in the service lives of the pavements or a difference in thickness. The use of fibers was estimated to result in an increase in the service life and could yield a 33% decrease in CO2 emissions, depending on the extent of the increase (Stempihar et al. 2012). A cost estimate was also developed: the cost of adding the fibers was estimated to be approximately 11% in this case. For this to be cost-effective, the equivalent uniform annual cost was determined for the JAC and SHR mixes. The typical service life of an open-graded mix was assumed to be about 8 years. If the use of fibers increased the service life from 8 to 8.9 years, the additional cost of the fibers could be justified (Stempihar et al. 2012). The authors concluded that the use of fibers in airfield pavements is feasible and offers the potential for increased service life under heavy loading and high tire pressures (Stempihar et al. 2012). Lyons and Putnam (2013) compared the laboratory performance of cellulose fibers, CR-modified asphalt, and SBS-modified asphalt in porous asphalt mixtures. They found that the addition of fibers and polymers led to reductions in the porosity and permeability of the porous mixtures. However, it also led to improvements in draindown, abrasion resistance (Cantabro), and indirect tensile strength. Cellulose and crumb rubber were most effective at reducing draindown compared with the unmodified control. Crumb rubber and a combination of cellulose fibers with SBS-modified binder were most effective at improving the abrasion resistance of the mixtures. Finally, cellulose did not have a significant effect on tensile strength, but SBS and crumb rubber did lead to increased strength (Lyons and Putnam 2013). Do Vale et al. (2014) studied the effects of using coconut fibers in SMA. The northeastern part of Brazil is a leading producer of coconuts. They found that the addition of cellulose and coconut fibers increased the TSR. But SMA mixes with coconut fibers did not perform as well in fatigue as mixes with cellulose or no fiber. This was possibly because the high absorption of the coconut fiber increased the stiffness of the mix. The researchers also noted that long coconut fibers were difficult to mix with the aggregate and could have lowered the strength of the mix by interfering with aggregate interlock. Work using shorter coconut fibers was planned. Discussion of Performance of Fiber-Reinforced Mixtures The literature survey on the performance of fiber-reinforced mixtures shows that the results are mixed. The use of fibers is not reported to cause any performance problems, provided the mix design, fiber dosage, and mix production are adequate. In some cases, fibers are reported to improve cracking or rutting resistance; in others they appear to have no effect. There are many possible explanations for these apparent discrepancies, including differences in the materials used in different studies, construction or laboratory mix preparation issues, and natural variability. Work by Cleven (2000) suggests an additional explanation. Cleven reported that the use of fibers may have a greater impact on the performance of marginal or low-quality mixtures. His findings showed that fibers did not affect low temperature cracking resistance until the binder began to fail. When cracking began to develop in the binder, the fibers were mobilized and helped reduce the cracking (Cleven 2000). This conclusion is supported by Kutay et al. (2009), who observed that when cracking initiated in the ALF fiber section, the fibers helped reduce the severity of the cracking. A paper by Gibson and Li (2015), also using the FHWA ALF, showed that fiber-reinforced mix performs better in fatigue than polymer-modified mix at high strain levels, but not at lower strain levels. These observations might also explain some of the seemingly disparate results reported in the literature. For example, the Indiana test section placed in 1980 exhibited much better rutting and cracking resistance than the control section without fibers (Galinsky 1984; McDaniel 1985). The severe rutting and cracking in the control section, however, showed that the control mixture was not of sufficient quality to withstand the interstate traffic loadings applied. Later, when fibers were added to a much more visible and closely controlled study in Indiana (McDaniel 2001; McDaniel and Shah 2003), all the mixes, including the unmodified control, performed very well for more than 10 years under interstate traffic loadings. Adding fibers or a variety of polymer binders to a high-performing mixture did not have as great an impact on performance. COSTS AND BENEFITS OF FIBER ADDITIVES IN ASPHALT MIXTURES Some of the cited studies demonstrate the benefits of using fibers, including these: • Reduced draindown in open- and gap-graded mixtures,

23 • Increased resistance to rutting and cracking, • Improved durability, and • Increased toughness and stability. However, documented benefit–cost ratios or cost- effectiveness studies are lacking in the literature. Only the study by Stempihar et al. (2012) included a cost estimate. As previously reported, the cost for the fiber mix in that study was about 11% higher than the cost for the control mix. This increased cost could be justified by an increase in the service life of 0.9 to 1.1 years.