Aggregate Properties and the Performance of Superpave-Designed Hot-Mix Asphalt (2005)

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.

67 CHAPTER 4 REVIEW OF PERFORMANCE DATA FROM FIELD TEST SECTIONS AND FULL-SCALE ACCELERATED TESTING Currently, a large percentage of the HMA produced in the United States is designed using the Superpave mix design system. The earliest Superpave projects were placed in 1992 (171). A number of experimental field sections have been built and documented by agencies. In 1999, NCAT evaluated the early performance of 44 Superpave sections placed between 1992 and 1998 (172). Some of these sections were revisited in 2001. Unfortunately, the consensus and source aggregate properties are not documented in these reports. There are a number of accelerated loading facilities in the United States. However, aggregate properties have not been experimental factors in the majority of testing completed to date. One exception is the Indiana DOT/Purdue Univer- sity APT Facility in West Lafayette, Indiana. Numerous research results from this facility were discussed previously (23, 61, 64–67). In addition, there are three test tracks that have been active since the completion of the Superpave mix design system: MnRoad, WesTrack, and the NCAT Test Track. 4.1 LTPP There are 773 SPS 1, 5, and 9 HMA test sections (there is often more than one test section per site) listed in LTPP DataPave 3.0 (173). A limited number of the SPS 1 and 5 sec- tions may have been designed using the Superpave method. The majority of the SPS 9 sections were designed using the Superpave method. Unfortunately, the only Superpave con- sensus aggregate property stored in DataPave is uncompacted voids in fine aggregate. A survey of DataPave 3.0 indicates there are records for 91 HMA layers (multiple layers per sec- tion) representing a range of uncompacted void results from 37.6% to 47.9%. Some sand equivalent test results are stored in DataPave 3.0, but only for seal coat treatments. A data extraction was performed to obtain uncompacted void content, rut depth, date of construction, date of last rut depth measurement, and annual ESALs. In total, uncom- pacted voids contents were available for 55 surface mixes. Rut-depth measurements were available for all of these sec- tions. There was no trend between total rut depth and uncom- pacted voids content. This was expected because of the range in age of the pavement sections and varying levels of traffic. Unfortunately, traffic data were only available in DataPave 3.0 for five adjacent sections in one state. There is a weak relationship (R2 = 0.46) between uncompacted voids content and rut depth divided by the square root of ESALs for the five sections. As shown in Figure 29, the relationship from the limited LTPP data does not match the relationship from the NCAT National Rutting Study (10). 4.2 MNROAD MnRoad’s mainline test road contains 16 HMA sections. Fourteen of these sections, constructed in 1992 and 1993, use the same three stockpiles: a coarse crushed gravel, fine gravel, and crushed granite (174). All fourteen original mainline HMA sections were constructed using the same gradation. The MnRoad mainline experimental variables were design com- paction effort, binder grade, and thickness. Two additional Superpave sections were constructed in 1997, one coarse and one through the restricted zone. Originally, 11 HMA sections were constructed on the low-volume test road using the same aggregate stockpiles and gradations as the mainline. Since aggregate type and gradation were not experimental factors in the MnRoad experiment, it is not possible to analyze the relationships between aggregate properties and performance. 4.3 WESTRACK Aggregate type was not an experimental factor for the 26 original WesTrack sections; a crushed gravel was used for all of the sections (175). Two gradations were used, coarse- graded and fine-graded, both 19.0-mm NMAS. The fines con- tent of the fine-graded mix was also varied. Due to premature failure, a number of the original sections were replaced: eight sections with a crushed andesite aggregate matching the orig- inal coarse gradation and two sections with two slightly dif- ferent dense-graded gradations, one with the original crushed gravel and one with the crushed andesite. However, design compaction effort (Hveem instead of gyratory) and binder grade were also modified for the dense-graded sections. One comparison that can be made is the effect of fractured face count between the two coarse aggregate sources used in the

68 LTPP Data y = -8E-06x + 0.0004 R2 = 0.4569 0 0.0001 0.0002 0.0003 0.0004 0.0005 0.0006 0.0007 35 37 39 41 43 45 47 Uncompacted Voids, % R ut D ep th , i n/ Sq rt. E SA Ls NCAT Rutting Study y=-0.0000791x + 0.0036647 R2 = 0.67 Figure 29. Uncompacted voids versus rut depth normalized by traffic (10). 0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 50 55 60 65 70 75 80 85 90 95 100 Coarse Aggregate Crushed Faces (%) Ru t D ep th /S q. R oo t E SA L Y = 0.001235 - 0.0000098(X) R2 = 0.42 Replacement Fine Plus Fine Coarse Figure 30. National Rutting Study and WesTrack coarse aggregate angularity relationship to performance (176). ◆ = National Rutting Study. ● = WesTrack. coarse-graded sections. Hand et al. (176) plotted coarse aggre- gate angularity (percent two crushed faces) versus rut depth divided by the square root of ESALs along with data from the NCAT National Rutting Study (17) as shown in Figure 30. The results from the WesTrack “fine” and “fine plus” mixes plot close to the regression line. The error on the “coarse” mixes is higher, but still comparable with the other sections. Prior to the implementation of the Superpave method, the use of such coarse mixes was uncommon. Therefore, it is unlikely that similar mixes were included in the National Rutting Study. The “replacement” sections appear to be outliers in the relationship. The replacement sections failed very quickly (at approximately 500,000 ESALs) after being placed in service. They were placed in service during hot weather. Data from the NCAT Test Track (177) show that rutting did not occur when the 7-day average air temperature dropped below 28°C. The average age of the sections evaluated as part of the National Rutting Study was 5.7 years (10). Therefore, on aver- age the sections in the National Rutting Study that were used to develop the relationship had been through five winters. It is unlikely that rutting would occur during these cooler periods, and yet ESALs would continue to accumulate. Accelerated loading was applied to the replacement sections during warm weather. Based on these considerations, the WesTrack results do not invalidate the results for coarse aggregate angularity from the National Rutting Study.

69 4.4 2000 NCAT TEST TRACK Gradation, aggregate type, and NMAS are experimental factors at the NCAT test track (178). Eight aggregate types— quartzite, granite, a limestone-slag blend, gravel, a limestone– recycled asphalt pavement (RAP) blend, a limestone-gravel- RAP blend, sandstone, and quartz gravel—are represented, providing a range of aggregate consensus and source proper- ties. Three NMASs were used: 9.5, 12.5, and 19.0 mm. Four major gradation shapes are included: above the restricted zone (fine), through the restricted zone, below the restricted zone, and SMA gap grading. Structural capacity was not a variable in the 2000 NCAT Test Track. The pavement sec- tion consisted of two 2-in. experimental lifts on 15 in. of HMA base. The 19 in. of HMA were placed on top of 5 in. of asphalt treated drainage layer, 6 in. of crushed stone, and 12 in. of A-2 improved subgrade. The maximum wire-line rut depth after 10 million ESALs was 7.27 mm (179). This level of rutting and the next highest rut depth occurred in two sections with unmodified binder that were placed at optimum +0.5% binder content. This illustrates that there were no true rutting “failures” at the 2000 NCAT Test Track; however, some observations can be drawn from the track performance in relation to aggregate properties. 4.4.1 Effect of Gradation When the Superpave method was first implemented, the restricted zone excluded many aggregate blends close to the maximum density line that had been used previously. Initially, it was felt that coarse-graded mixes would be more rut resis- tant than fine-graded mixes. However, the rapid failure of the coarse-graded mixtures in the WesTrack experiment created concern about coarse-graded Superpave mixes. Gradations passing above the restricted zone, through the restricted zone, and below the restricted zone were placed at the 2000 NCAT Test Track. Although each of the sponsoring agencies deter- mined the mixes to be placed on their sections, there are a number of cases in which the effect of gradation can be com- pared with the same aggregate source and binder grade. Figure 31 shows comparisons between coarse- and fine- graded mixes produced with PG 67-22 asphalt binder for three aggregate types. An analysis of variance was performed with rut depth as the response variable and gradation and aggregate type as factors. Wire-line rut depths taken at three random locations within each section were used as factors. As shown in Table 16, gradation is not a significant factor affecting rut depth. However, aggregate type and the inter- action between aggregate type and gradation are significant. Figure 31. Effect of gradation type on rut depth. Source Freedom F-statistic p-value Significant at 5% Aggregate Type 2 6.68 0.011 Yes Gradation 1 4.30 0.060 No Aggregate Type*Gradation 2 8.15 0.006 Yes Total 17 Error 12 Degrees of TABLE 16 Analysis of variance on effect of gradation on rut depth

70 Brown et al. (179) performed a similar analysis for all of the granite sections on the east tangent of the track. Coarse- graded (below the restricted zone [BRZ]), fine-graded (above the restricted zone [ARZ]), and gradations passing through the restricted zone (TRZ) were included. Each gradation was placed with an unmodified PG 67-22 binder as well as with PG 76-22 binder produced with both SBS (styrene-butadiene- styrene polymer) and SBR (styrene-butadiene rubber) modi- fiers. The interaction between gradation and binder grade is shown in Figure 32. An ANOVA is shown in Table 17. In the case of the east curve data, gradation is a significant factor, with the coarse-graded mixes having the largest rut depth followed by the mixes passing through the restricted zone. The mixes with the fine gradations were the most rut resistant. Practically speaking, there was little difference between the rut depths, and all three gradations would per- form well. 4.4.2 Relationship Between Aggregate Properties and Performance The following aggregate tests were performed on the aggre- gate sources used in the 2000 NCAT Test Track: • Bulk specific gravity, • F&E, • Uncompacted voids in coarse aggregate, • Uncompacted voids in fine aggregate (FAA), • Sand equivalent, • Methylene blue, • LA abrasion, • Micro-deval, and • Sulfate soundness. Coarse aggregate angularity was not performed even though gravel sources were used at the track. The tests were per- formed on the aggregate stockpiles, and the blend properties were determined mathematically using a weighted average (percent of blend and percent coarse or fine aggregate were used as weighting factors). A stepwise regression was performed using the final-wire line rut depth as the response variable for all of the Superpave sections. When only the aggregate properties shown above were used as predictor variables, LA abrasion was the first predictor entered followed by percent passing the 0.075-mm (No. 200) sieve. However, the relationship was not signifi- cant for either variable. Other response variables that were Figure 32. Effect of gradation shape and binder type (east curve) (179). Source Degrees of Freedom Adjusted Mean Squares F-stat p-value Significant?1 Gradation (Grad) 2 14.1783 30.18 0.000 Yes Modifier 2 18.9431 40.33 0.000 Yes Grad*Modifier 4 3.082 6.56 0.000 Yes Error 72 0.4697 — — — 1 5% level of significance. TABLE 17 Results of analysis of variance on effect of modifier and gradation on rut depth (179)

71 related to durability are less likely to occur in this short time period. 4.5 SUMMARY OF DATA FROM IN-SERVICE PAVEMENTS AND ACCELERATED LOAD FACILITIES Numerous Superpave sections have now been in-service for 10 years or longer; however, the data from these sections, particularly aggregate properties, are not readily accessible. The best potential source of data, the LTPP Program, only stores limited Superpave aggregate consensus property data. Attempts to use this data were hindered by a lack of traffic data. The most significant accelerated testing related to aggre- gate properties has been conducted at the Indiana DOT/Purdue University APT Facility. This work was discussed in Chap- ter 2. Three test tracks have been operated since the imple- mentation of the Superpave method: MnRoad, WesTrack and the NCAT Test Track. Aggregate properties were not an experimental variable at MnRoad or WesTrack with the excep- tion of the coarse aggregate angularity of the replacement sections at WesTrack. Coarse aggregate angularity data from WesTrack was in general agreement with the relationship developed by Cross and Brown (17). There were a number of different aggregate types and gradations placed at the 2000 NCAT Test Track. All of the sections performed well. Strong correlations were not evident among rutting, VMA, construc- tion density or densification under traffic, and aggregate prop- erties. A weak relationship was obtained between rut depth and fine aggregate uncompacted voids for the sections con- structed using PG 76-22 binder. evaluated included VMA, initial construction density, and densification (change in field density) after 10 million ESALs. The uncompacted void content of the fine aggregate had a significant relationship with the as-constructed density. LA abrasion was the first variable entered for change in density, and uncompacted voids content of fine aggregate was the fourth variable entered. The relationships for both variables were significant. In order to further analyze the data, the data set was divided by binder grade with all of the PG 67-22 sections grouped together and all of the PG 76-22 sections grouped together. It was felt that the modified binders may be mask- ing the effect of the aggregate properties. There were only nine Superpave sections placed with PG 67-22. Therefore, a stepwise regression could not be performed with the previ- ous group of predictor variables. Instead, individual regres- sions were performed with aggregate properties such as fine aggregate uncompacted voids, coarse aggregate uncompacted voids, and F&E at the 31 ratio, which might be expected to be correlated with rutting. None of the aggregate properties pro- duced significant relationships. For the 19 sections with PG 76- 22 binder, an ANOVA performed as part of the regression analysis indicated a poor but significant relationship for the fine aggregate uncompacted void content (p = 0.011, R2 = 0.31). It should be noted that all of the sections in the 2000 NCAT Test Track performed well in rutting. It is difficult to develop relationships between a response variable, such as rut depth, and aggregate properties when all of the rut depths are small, even though traffic, climate, and pavement struc- ture were constant for all of the sections. The application of 10 million ESALs occurred in a 2-year period. Distresses