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.
74 Experiment 1: Effect of Increased P-200 Content Introduction As stated earlier in the Work Plan section, the original plan was to modify the Georgia design by adding baghouse fines (BHF) to the mixture and therefore increase the P-200 material. The expectation was that the binder and increased dust would create a mastic and provide a more durable mixture. The Georgia mix design, already having a design air void content of 15.4%, had little room for additional filler if the samples for this part of the study adhered to the expected minimum air void content of 15.0%. The initial addition of 3.0% and 6.0% BHF showed a slight improvement to performance but reduced air void content below the minimum of 15.0%. The 6.0% addition of BHF appeared to provide an insignificant amount of benefit to the mixture; therefore, after deciding to swap to the South Carolina mix design, the decision was made to cut back the addition of BHF to the rate of 2.0% and 4.0%. The Georgia and South Carolina designs show marked differences in field performance but have the same aggregate type, gradation, optimum AC, binder type, and approximately the same fiber content. This led to a question as to what properties led to the difference in field performance and what performance test could be used to distinguish these properties. It was decided to test both the Georgia and the South Carolina mix designs for this evaluation. The Georgia design, while already at the expected mini- mum allowable air void content, was used as a comparison to see how the added BHF would affect both a good and poorly performing mix design. The mix design components for this part of the testing can be found in Table 41. The original Georgia and South Carolina design data are included in this section for comparison purposes. They have been labeled Georgia Control and South Carolina Control. The percentages of the stockpiles were altered to attempt to keep the mixture gradation blend equal to the original NCAT verification design. As can be seen in Table 42 and Table 43, the plus #4 sieves stayed relatively consistent, but an increase in the P-200 material caused the other fine sieves to shift by approximately the same percentage. Results and Discussion Effect on Air Voids It was anticipated that the added BHF would decrease the air voids in the mix. It was also expected that the film thickness would decrease; however, the projection was that the mixture performance would improve enough to mitigate any loss in film thickness or air void content. This may help determine if film thickness and air void content are critical design components, and, if so, at what threshold level. A summary of the mixture properties can be found in Table 44. The VCA ratio is based on the breakpoint sieve method. C H A P T E R 6 Part 2
Part 2 75 Mixture Type Georgia Good South Carolina Poor Mixture Designation Control +2%BHF +4%BHF Control +2%BHF +4%BHF Aggregate Type Granite Asphalt Type PG 76-22 Binder Modifier 2.5% SBS Polymer Anti-strip 0.5% by weight of binder Fiber, % 0.4 0.3 AC, % 6.0 Total P-200, % 2.0 3.9 6.0 1.7 3.7 5.6 Table 41. Experiment 1 mix design components. Percent Passing Sieve Georgia Mix Design Alterations Georgia Gradation Limits JMF NCAT +2%BHF +3%BHF +4%BHF +6%BHF 25.0 mm, 1" 100 100 100 100 100 100 19.0 mm, 3/4" 100 100 100 100 100 100 100 12.5 mm, 1/2" 92 96 96 96 96 96 85 - 100 9.5 mm, 3/8" 66 66 66 68 66 66 55 - 75 4.75 mm, #4 25 21 22 21 21 23 15 - 25 2.36 mm, #8 8 8 10 7 8 11 5 - 10 1.18 mm, #16 5 6 8 7 8 10 0.600 mm, #30 4 5 7 6 7 10 0.300 mm, #50 3 4 6 6 7 9 0.150 mm, #100 2 3 5 6 7 9 0.075 mm, #200 1.5 2.0 3.9 4.9 6.0 8.2 2 - 4 Table 42. Experiment 1 Georgia mix design alterations. Percent Passing Sieve South Carolina Mix Design Alterations South Carolina Gradation Limits JMF NCAT +2%BHF +4%BHF 25.0 mm, 1" 100 100 100 100 19.0 mm, 3/4" 100 100 100 100 100 12.5 mm, 1/2" 95 95 95 95 89 - 100 9.5 mm, 3/8" 70 74 75 75 63 - 75 4.75 mm, #4 21 21 22 22 15 - 25 2.36 mm, #8 8 8 9 10 5 - 10 1.18 mm, #16 5 6 8 0.600 mm, #30 5 3 5 7 0.300 mm, #50 3 4 6 0.150 mm, #100 5 2 4 6 0.075 mm, #200 2.2 1.7 3.7 5.6 0 - 4 Table 43. Experiment 1 South Carolina mix design alterations.
76 Performance-Based Mix Design of Porous Friction Courses The CoreLok air voids of the mixtures (Figure 60) did decrease to a point with added BHF. The Georgia design, which was already near the minimum threshold, did not show any addi- tional decrease in air voids between the 2.0% and 4.0% added BHF. The South Carolina design showed an incremental decrease in the air voids with the increased dust content. This consistent decrease is most likely due to the amount of extra room available in the design, as indicated from the higher VMA values in Table 44. Figure 60 shows, however, that even with an additional 4.0% BHF, the resulting air voids were well above the 15% minimum used in this study. Mix ID Total AC (%) Total P-200 (%) Average Air Voids (%) Avg. VMA, (%) Average VCAMIX/VCADRC Film Thickness (microns) GA Control 5.0 2.0 17.5 26.3 0.82 21.9 GA Control 6.0 2.0 15.4 26.6 0.82 27.1 GA Control 7.0 2.0 12.5 25.7 0.80 32.4 GA +2BHF 5.0 3.9 15.3 24.1 0.80 13.9 GA +2BHF 6.0 3.9 12.8 23.9 0.80 17.3 GA +2BHF 7.0 3.9 10.4 23.7 0.79 20.7 GA +3BHF 6.0 5.0 14.6 25.1 0.79 15.4 GA +4BHF 5.0 6.0 16.1 25.1 0.80 11.3 GA +4BHF 6.0 6.0 13.1 24.4 0.79 13.9 GA +4BHF 7.0 6.0 11.0 24.5 0.79 16.7 GA +6BHF 6.0 8.2 11.3 22.3 0.79 10.1 SC Control 5.0 1.7 23.6 31.9 0.92 28.0 SC Control 6.0 1.7 22.2 32.3 0.93 34.7 SC Control 7.0 1.7 20.6 32.6 0.94 41.2 SC +2BHF 5.0 3.7 22.3 30.4 0.91 15.9 SC +2BHF 6.0 3.7 20.7 30.8 0.92 19.8 SC +2BHF 7.0 3.7 18.3 30.4 0.91 23.8 SC +4BHF 5.0 5.6 21.3 29.8 0.90 12.0 SC +4BHF 6.0 5.6 19.3 29.8 0.91 14.8 SC +4BHF 7.0 5.6 17.6 30.1 0.91 17.8 Table 44. Summary of mixture properties with increased P-200. 17.5 15.3 16.1 23.6 22.3 21.3 15.7 12.8 13.1 22.2 20.7 19.3 12.5 10.4 11.0 20.6 18.3 17.6 0.0 5.0 10.0 15.0 20.0 25.0 Georgia Control Georgia +2BHF Georgia +4BHF South Carolina Control South Carolina +2BHF South Carolina +4BHF Av er ag e Co re lo k Ai r V oi ds (% ) 5.0% AC 6.0% AC 7.0% AC Figure 60. Experiment 1 CoreLok air voids summary.
Part 2 77 Effect on Film Thickness The film thickness of the designs did significantly decrease with the added BHF (Figure 61). This was anticipated because the surface area of the P-200 is a significant part of the total surface area calculation. The original designs for both the Georgia and South Carolina mixtures had film thicknesses greater than the expected requirement of 24.0 microns. The added BHF dropped all of the modified designs below that point. Cantabro Results The Cantabro Test demonstrated that an increase in P-200 material could improve the durability of the mixture. A graphical depiction of the data can be seen in Figure 62, and a summary of all the data can be seen in Table 45. The added 3.0% and 6.0% BHF data are represented in the table but are absent from the figure due to only fabricating specimens at the optimum AC. As antici- pated, the Cantabro loss decreased with increased asphalt, but the Cantabro loss also decreased 21.9 13.9 11.3 28.0 15.9 12.0 27.1 17.3 13.9 34.7 19.8 14.8 32.4 20.7 16.7 41.2 23.8 17.8 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 Georgia Control Georgia +2BHF Georgia +4BHF South Carolina Control South Carolina +2BHF South Carolina +4BHF Fi lm T hi ck ne ss (m ic ro ns ) 5.0% AC 6.0% AC 7.0% AC Figure 61. Experiment 1 film thickness summary. Figure 62. Experiment 1 Cantabro loss summary. 25.4 17.5 17.6 57.3 34.2 26.5 19.3 13.0 9.3 37.9 18.9 16.9 12.8 8.2 7.1 26.8 9.0 9.2 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 Georgia Control Georgia +2BHF Georgia +4BHF South Carolina Control South Carolina +2BHF South Carolina +4BHF Av er ag e Ca nt ab ro L os s ( % ) 5.0% AC 6.0% AC 7.0% AC
78 Performance-Based Mix Design of Porous Friction Courses with the increased BHF. As seen in Table 45, the Georgia mixes with +2BHF and +4BHF at 5.0% AC had comparable stone loss to the mix at 6.0% AC but without added BHF. The South Carolina design showed significant improvement with the initial 2.0% additional BHF. As with the Georgia mix, Figure 62 shows that increasing the P-200 material by 2.0% pro- vides more durability to the South Carolina design than by increasing the asphalt binder content by 1.0%. The South Carolina Control at 7.0% AC has a percent loss of 26.8, while the South Carolina +2BHF design at the optimum AC of 6.0% shows a percent loss of 18.9. If the study shows that increased P-200 does more for durability than asphalt binder without sacrificing a significant amount of air voids and permeability, this could greatly benefit the industry. This could reduce the expensive cost of PFC mixtures and make them more economical. As shown in Figure 62, the addition of 2% BHF at 5% AC reduced Cantabro loss to values that were the equivalent of 6.0% AC without the added BHF (34.2 versus 37.9). The average Cantabro data was plotted to try and determine how to optimize the amount of P-200 material. The data for each AC are represented in separate graphs in Figure 63. The data for the Georgia design at optimum AC (Figure 63B) shows five data points because it includes the added 3.0% and 6.0% BHF. This graph, having more data points, may be more accurate than the other graphs with only 3 points. All of the data indicate that an optimum P-200 content may be somewhere between 4.5% and 5.5%. After a P-200 of 5.5, the effects of the increased dust have a negligible effect on the durability of the mix. An ANOVA with Tukey-Kramer groupings shows that the Georgia design showed improve- ment, as can be seen by the means, but the only statistical difference was between the Control and the added 6.0% BHF Table 46(A). The South Carolina mix with any percentage of added Mix ID Total AC (%) Total P-200 (%) Fiber (%) Average Air Voids (%) Cantabro Loss (%) Average St. Dev. COV (%) GA Control 5.0 2.0 0.4 17.5 25.4 1.9 7.6 GA Control 6.0 2.0 0.4 15.2 19.3 6.4 33.4 GA Control 7.0 2.0 0.4 12.5 12.8 3.2 24.7 GA +2BHF 5.0 3.9 0.4 15.3 17.5 3.3 19.1 GA +2BHF 6.0 3.9 0.4 13.5 13.0 3.0 23.2 GA +2BHF 7.0 3.9 0.4 10.4 8.2 1.7 20.1 GA +3BHF 6.0 5.0 0.4 14.4 10.3 3.5 34.2 GA +4BHF 5.0 6.0 0.4 16.1 17.6 1.4 7.8 GA +4BHF 6.0 6.0 0.4 13.1 9.3 1.3 13.5 GA +4BHF 7.0 6.0 0.4 11.0 7.1 0.7 9.3 GA +6BHF 6.0 8.2 0.4 11.4 10.1 1.0 9.9 SC Control 5.0 1.7 0.3 23.6 57.3 4.7 8.2 SC Control 6.0 1.7 0.3 22.2 37.9 11.9 31.3 SC Control 7.0 1.7 0.3 20.6 26.8 7.1 26.5 SC +2BHF 5.0 3.7 0.3 22.3 34.2 4.8 14.1 SC +2BHF 6.0 3.7 0.3 20.8 18.9 4.2 22.2 SC +2BHF 7.0 3.7 0.3 18.3 9.0 2.2 25.1 SC +4BHF 5.0 5.6 0.3 21.3 26.5 4.0 15.1 SC +4BHF 6.0 5.6 0.3 19.4 16.9 3.0 17.8 SC +4BHF 7.0 5.6 0.3 17.6 9.2 2.8 30.3 Table 45. Experiment 1 Cantabro results.
Part 2 79 BHF was statistically different from the South Carolina JMF Control Table 46(B). It is important to note that the Georgia Control passed the maximum 20.0% Cantabro loss criterion, while the South Carolina Control did not. The South Carolina Control improved by over 20% with the addition of 2.0% BHF, but it only improved an additional 2% when BHF was increased to 4.0%. This, along with the Georgia analysis, shows that increasing the P-200 material will help the durability of all the mixtures, but it may have a greater effect on mixtures with a larger initial Cantabro loss. (C) (B)(A) Figure 63. Effect of increased P-200 on Cantabro loss. (A) (B) Mix ID South Carolina N Mean Grouping South Carolina Control 3 39.0 A South Carolina +2.0BHF 3 18.9 B South Carolina +4.0BHF 3 16.9 B Mix ID Georgia N Mean Grouping Georgia Control 3 19.3 A Georgia +2.0BHF 3 13.0 A B Georgia +3.0BHF 3 10.3 A B Georgia +4.0BHF 3 10.1 A B Georgia +6.0BHF 3 9.3 B p = 0.039 R2 = 60% p < 0.001 R2 = 71% Table 46. ANOVA (` = 0.05) of Cantabro loss with increased P-200.
80 Performance-Based Mix Design of Porous Friction Courses Permeability As discussed in Part 1, there appears to be a direct correlation between air void content and permeability of the PFC specimens. The increased P-200 causes the air void content of the speci- mens to decrease; therefore, it was expected that the permeability of the specimens would also decrease. The Georgia 3.0% and 6.0% added BHF are included in both the summary (Table 47) and the graph (Figure 64). The South Carolina design showed little decrease in permeability with increased P-200. This is most likely due to the initial air void content of the mixture being relatively high. The Georgia design, which had an initial permeability value less than the rec- ommended 100 m/day, showed a decrease in permeability with increased P-200. The air void content of the mixtures directly reflects this as well. The only other criterion for permeability testing is the 35 m/day provided by Mississippi in its survey response. If that criterion is applied, only the Georgia design with an added 6.0% BHF fails. Since the permeability of the Georgia mix decreased with additional BHF when all the sam- ples were made at 6.0% AC, there was interest in determining how the added BHF affected Mix ID Total AC (%) Total P-200 (%) Fiber (%) Average Air Voids (%) Permeability (k) meter/day Average St. Dev. COV (%) Georgia Control 6.0 2.0 0.4 15.7 80 10.5 13.1 Georgia +2BHF 6.0 3.9 0.4 13.4 43 1.6 3.6 Georgia +3BHF 6.0 5.0 0.4 14.5 51 17.4 34.3 Georgia +4BHF 6.0 6.0 0.4 13.2 38 7.4 19.5 Georgia +6BHF 6.0 8.2 0.4 11.4 11 3.7 33.0 South Carolina Control 6.0 1.7 0.3 22.2 209 17.1 8.2 South Carolina +2BHF 6.0 3.7 0.3 20.6 222 17.3 7.8 South Carolina +4BHF 6.0 5.6 0.3 19.2 196 17.1 8.7 Table 47. Permeability summary with increased P-200. 80 43 51 38 11 209 222 196 0 50 10 150 200 250 300 Georgia Control Georgia +2BHF Georgia +3BHF Georgia +4BHF Georgia +6BHF South Carolina Control South Carolina +2BHF South Carolina +4BHF Av er ag e Pe rm ea bi lit y (k ), m et er /d ay Figure 64. Increased P-200 permeability results.
Part 2 81 permeability at other asphalt contents. An ANOVA of the results indicates a significant dif- ference (p-value = 0.000). Table 48 shows the mean permeability values and the Tukey-Kramer grouping from the ANOVA. Results show that the +2BHF and +4BHF at 5.0% AC provide similar permeability to the mix at 6.0% AC without added BHF. Just as with the Cantabro results shown previously, the mixture with +2BHF is as permeable and resistant to raveling as the mix with 1.0% more AC but without the added BHF. These results indicate that increased P-200 adds sig- nificant cohesive ability to PFC mixtures so that permeability and durability can be maintained at a significantly reduced AC. Hamburg Wheel-Tracking Test Results Table 49 summarizes the HWTT rut depths and RRIs determined according to Equation 6. The HWTT rut depth and RRI trends are similar to the Cantabro results. The South Carolina Control mixture failed around 2,540 passes (RRI = 1,270), but when the P-200 content was increased with 2.0% BHF, the performance improved and the mixture did not fail until 15,194 passes (RRI = 7,597). The 4.0% added BHF specimens showed improvement over both designs and reached 19,202 passes (RRI = 9,601) before reaching the failure criterion. This +4BHF design came close to passing the TxDOT criterion of 20,000 passes (RRI = 10,000). As can be seen in Figure 65 and Figure 66, the samples incurred primary (initial consolida- tion) and secondary (constant strain) deformation, but did not reach tertiary (shear defor- mation). Without tertiary deformation, there is no inflection point for these mixtures. The samples only exhibited densification, and no change due to shear was observed. The Georgia design showed marked improvement with the addition of 2.0% BHF, but no additional Mix ID Total AC (%) Total P-200 (%) Permeability (k) meter/day, (Mean) Tukey-Kramer Grouping Georgia Control 6.0 2.0 79.6 A Georgia +4BHF 5.0 6.0 75.3 A AGeorgia +2BHF 5.0 3.9 67.0 B Georgia +2BHF 6.0 3.9 42.8 B C Georgia +4BHF 6.0 6.0 37.8 C D Georgia +4BHF 7.0 6.0 14.2 D Georgia +2BHF 7.0 3.9 12.0 D Table 48. Permeability results at different AC and P-200. Mix ID Total AC (%) Total P-200 (%) Fiber (%) Average Air Voids (%) Greatest Rut Depth (mm) RRI Average (mm) St. Dev. St. Dev. (mm) COV (%) Average (mm) (mm) COV (%) GA Control 6 2 0.4 14.3 8.99 2.88 32 12,921 2,267 18 GA +2BHF 6 3.9 0.4 13.7 5.54 0.62 11.3 15,640 491 3 GA +4BHF 6 6 0.4 13.1 5.36 0.67 12.4 15,777 524 3 SC Control 6 1.7 0.3 21.7 15.85 1.84 11.6 1,270 500 39 SC +2BHF 6 3.7 0.3 20.8 15.14 4.64 30.6 7,597 4,016 53 SC +4BHF 6 5.6 0.3 20.4 12.81 1.99 15.5 9,601 1,883 20 Table 49. HWTT summary results for increased P-200.
82 Performance-Based Mix Design of Porous Friction Courses improvement was observed with the 4.0% BHF, specimens based on the HWTT rut depths and RRI values. Moisture Susceptibility Testing The testing for TSR in this experiment was critical. While increasing the P-200 can potentially create a mastic and consequently a stronger mixture, it also creates less free binder. Free binder is the extra binder remaining after the aggregate has been coated and binder absorption has taken place. The P-200 absorbs more of the free binder, leaving less to coat the coarse material. This can potentially lead to moisture susceptibility issues. The expectation was that the extra P-200 would increase the ITS of the mixtures and hopefully not decrease the TSR. If the ITS were increased significantly and the TSR did fall below the recommended criteria, it was thought that the anti-strip dosage could be increased or hydrated lime could be added to the mixture 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 0 5000 10000 15000 20000 A ve ra ge H am bu rg R ut D ep th (m m ) Hamburg Wheel Passes Georgia 2.0% Dust Georgia 3.9% Dust Georgia 6.0% Dust Max Allowable Rut Depth Figure 65. Georgia HWTT results with increased P-200. 0 2 4 6 8 10 12 14 16 18 0 5000 10000 15000 20000 Av er ag e H am bu rg R ut D ep th (m m ) Hamburg Wheel Passes South Carolina 1.7% Dust South Carolina 5.6% Dust South Carolina 3.7% Dust Max Allowable Rut Depth Figure 66. South Carolina HWTT results with increased P-200.
Part 2 83 to ensure passing results. The improvement in performance of the mixture should offset the additional cost of the change in dosage or type of anti-strip. The summary for this testing can be found in Table 50. The unconditioned tensile strength of each mixture with increased P-200 improved over the control for each design. As shown in Table 50, the conditioned strength of the mix improved over 40% when mix with 4% BHF as compared to the conditioned values of the control. The TSR value increased with the Georgia design but decreased with the South Carolina design. The South Carolina conditioned strengths did not increase with the same magnitude as the unconditioned strengths. The ANOVA was conducted on both the South Carolina and Georgia designs. It was initially performed separately, but the results for this testing were so significantly different that com- bining the results showed the same results. The results from the Tukey-Kramer grouping can be found in Table 51. Both the conditioned and unconditioned results showed good fit with a significant difference between the mixtures. The Georgia mix designs in regards to ITS with the additional 2.0 BHF showed some improvement over the control, but the means show that the 4.0 BHF showed no improvement over the 2.0 BHF. The South Carolina Control was signifi- cantly different from both the 2.0 and 4.0 BHF for the unconditioned strengths. The 2.0 BHF was not significantly different from the South Carolina Control in the conditioned strengths. As seen in Table 50, the South Carolina added BHF shows a decrease for the TSR results. The lower conditioned value for the 2.0 BHF caused the TSR value to be low and also caused it to be grouped with the Control. Overlay Test Results The OT results for the increased P-200 specimens showed a relative increase in performance when compared to the Control specimens. The summary of the results can be seen in Table 52, and a graphical depiction is shown in Figure 67. The South Carolina designs all terminated MIX ID TOTAL AC (%) TOTAL P-200,% AVERAGE AIR VOIDS (%) AVERAGE ITS (PSI) TSR Conditioned Unconditioned Conditioned Unconditioned GA Control 6.0 2.0 13.9 14.0 57.6 74.3 0.78 GA +2BHF 6.0 3.9 13.3 13.4 86.4 100.3 0.86 GA +4BHF 6.0 6.0 14.3 14.3 82.4 99.8 0.83 SC Control 6.0 1.7 21.2 21.2 36.8 45.2 0.81 SC +2BHF 6.0 3.7 20.6 20.5 38.8 62.0 0.63 SC +4BHF 6.0 5.6 19.0 18.9 54.4 77.3 0.70 Table 50. TSR summary for increased P-200. Mix ID Conditioned Unconditioned N Mean Grouping N Mean Grouping Georgia +2BHF 3 86.4 A 3 100.3 A Georgia +4BHF 3 82.4 A 3 99.8 A Georgia Control 3 57.6 B 3 74.3 B South Carolina +4BHF 3 54.4 B 3 77.3 B South Carolina +2BHF 3 38.8 C 3 62.0 B South Carolina Control 3 36.8 C 3 45.2 C p < 0.001 R2 = 98% p < 0.001 R2 = 94% Table 51. ANOVA for ITS of increased P-200 (` = 0.05).
84 Performance-Based Mix Design of Porous Friction Courses prior to reaching the 93% load reduction and therefore had to be extrapolated. An outlier in both the Georgia Control and Georgia +2BHF designs was observed. These two outliers were removed from the data set prior to analysis. The Georgia +4BHF had no outliers but did show a large COV (49%). The peak load for each design decreased with increased P-200, while cycles to failure showed an increase with the increased P-200. The South Carolina +2BHF showed more extra polated improvement than the +4BHF design, but the Georgia design improved with each increase of the P-200. The ANOVA for the cycles to failure (Table 53) showed that the South Carolina +2BHF was significantly improved over the Control. The South Carolina +4BHF was not different from the Control or the +2BHF designs. The Georgia designs showed numerical improvement with increased P-200, but there was not a statistical difference between the designs when analyzing cycles to failure. In regards to peak load, an almost inverse relationship is observed (Table 54). The Georgia Control is statistically different from the Georgia +4BHF design, while the Georgia +2BHF design shows no difference. The South Carolina design showed some difference in peak load, but there was no significant improvement observed with the change in the P-200. Draindown Draindown was not evident for this experiment, nor was it expected. The only thing that changed in the designs for this part of the study was the addition of BHF, which would only decrease the amount of free binder. These results (Table 55) were as expected. Mix ID Replicates Average Air Voids (%) Average Peak Load (kN) Cycles to Failure Average St. Dev. COV (%) SC Control 6 19.2 1.798 1491 388 26.0 SC +2BHF 4 17.3 1.659 2335 480 20.6 SC +4BHF 4 22.5 1.638 1662 423 25.4 GA Control 3 12.8 2.621 583 166 28.4 GA +2BHF 3 11.0 2.456 682 177 25.9 GA +4BHF 4 13.4 1.998 941 464 49.3 Table 52. OT summary for increased P-200. 1491 2335 1662 583 682 941 0 500 1000 1500 2000 2500 3000 A ve ra ge O T C yc le s t o Fa ilu re SC Control SC +2BHF SC +4BHF GA Control GA +2BHF GA +4BHF Figure 67. OT results for increased P-200.
Part 2 85 I-FIT Results The I-FIT results for the mix designs with the increased P-200 can be found in Figure 68 through Figure 70. As previously analyzed, the peak load, Gf, and FI of each mix was calculated and summarized. The peak load (Figure 68) appears to increase with increased P-200. As seen in other testing, there appears to be negligible benefit between the Georgia +2BHF and +4BHF. The Gf of the mixtures (Figure 69) are relatively uniform regardless of mix design or addition of BHF. This may indicate that the mixture strength is increased with the additional P-200, but the amount of energy required to fracture the specimen is not affected. The FI (Figure 70) of the mixtures did not trend as expected. The FI is an indication of cracking potential; therefore, with increased P-200 it was expected that the FI would decrease due to the decrease in free binder. The South Carolina mix design did show a decrease in FI with increased P-200; however, the Georgia design showed a more parabolic trend. Mix ID Cycles to Failure N Mean Grouping SC +2BHF 4 2335 A SC +4BHF 4 1662 A B SC Control 6 1491 B C GA +4BHF 4 941 B C D GA +2BHF 3 682 C D GA Control 3 583 D p < 0.001 R2 = 75% Table 53. ANOVA Analysis for OT cycles to failure for increased P-200 specimens. Mix ID Peak Load N Mean Grouping GA Control 3 2.621 A GA +2BHF 3 2.456 A B GA +4BHF 4 1.998 B C SC Control 6 1.798 C SC +2BHF 4 1.659 C SC +4BHF 4 1.638 C p < 0.001 R2 = 73% Table 54. ANOVA for OT peak load for increased P-200 specimens. Mix ID Total AC (%) Fiber (%) Total P-200 (%) Draindown (%) Test Temp, °F 330 357 Georgia Control 6.0 0.4 2.0 0.0 0.0 Georgia +2BHF 6.0 0.4 3.9 0.0 0.0 Georgia +4BHF 6.0 0.4 6.0 0.0 0.0 South Carolina Control 6.0 0.3 1.7 0.0 0.0 South Carolina +2BHF 6.0 0.3 3.7 0.0 0.0 South Carolina +4BHF 6.0 0.3 5.6 0.0 0.0 Table 55. Draindown results for increased P-200.
86 Performance-Based Mix Design of Porous Friction Courses 1.694 1.973 1.892 1.263 1.472 1.592 0.000 0.500 1.000 1.500 2.000 2.500 Av er ag e Pe ak L oa d (k N ) Georgia SBS Control Georgia +2BHF Georgia +4BHF South Carolina SSS Control South Carolina +2BHF South Carolina +4BHF Figure 68. I-FIT peak load results for increased P-200. 1828 1853 2273 1924 1843 1969 0 500 1000 1500 2000 2500 3000 3500 Av er ag e Fr ac tu re E ne rg y (J /m 2 ) Georgia SBS Control Georgia +2BHF Georgia +4BHF South Carolina SSS Control South Carolina +2BHF South Carolina +4BHF Figure 69. I-FIT fracture energy results for increased P-200. 28.7 21.8 39.1 57.7 36.3 34.8 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 Av er ag e Fl ex ib ili ty In de x (F I) Georgia SBS Control Georgia +2BHF Georgia +4BHF South Carolina SSS Control South Carolina +2BHF South Carolina +4BHF Figure 70. I-FIT flexibility index results for increased P-200.
Part 2 87 The ANOVA (α = 0.05) showed that some of the mixtures were statistically differvent when comparing peak load and FI (Table 56). The peak load analysis shows that the increase in P-200 material did not statistically change the Georgia or the South Carolina results. The Georgia design with the additional 2.0% BHF was statistically different from both the South Carolina Control and 2.0% additional BHF designs. This indicates that even though adding the BHF to the South Carolina design did not change the peak load when compared to the Control, the addi- tion of the 4.0% BHF changed it enough to allow it to be grouped with Georgia +2BHF design. While the numerical difference in peak load for the South Carolina designs was not large enough to be significant, it did make a noticeable change when analyzed with a larger data set. This numerical increase in peak load is important to note if the peak load becomes part of the criteria for designing PFC. The Gf results of the mixtures were not statistically different. This can be seen in Table 57 with a p-value of 0.636 and a single Tukey-Kramer grouping. This analysis had a poor fit with an R2 of only 12.1%. The FI analysis, Table 58, showed that the South Carolina Control was statistically differ- ent from the Georgia Control and +2BHF designs. It was expected that the Georgia and South Carolina designs would be statistically different; however, the Georgia +4BHF had the second highest FI of the mixes. This indicates that it has the same resistance to cracking as the South Carolina Control. There is a large numerical difference between these designs, but they are not statistically different. As stated earlier, the Georgia data did not trend as expected. The no-notch I-FIT showed that with increasing P-200 there was a significant increase in the peak load required to fracture the specimens (Figure 71). The force required to break the speci- mens was increased with increased P-200 for both the Georgia and South Carolina designs. The Mix ID Peak Load N Mean Grouping Georgia +2BHF 5 1.973 A Georgia +4BHF 6 1.892 A B Georgia Control 5 1.694 A B C South Carolina +4BHF 6 1.592 A B C South Carolina +2BHF 5 1.472 B C South Carolina Control 4 1.263 C p = 0.002 R2 = 52% Table 56. ANOVA for I-FIT peak load with increased P-200. Mix ID Fracture Energy N Mean Grouping Georgia +4BHF 6 2273 A South Carolina +4BHF 6 1969 A South Carolina Control 4 1924 A Georgia +2BHF 5 1854 A South Carolina +2BHF 5 1843 A Georgia Control 5 1828 A p = 0.636 R2 = 12% Table 57. ANOVA for I-FIT fracture energy with increased P-200.
88 Performance-Based Mix Design of Porous Friction Courses trend of increased P-200 and peak load was linear for each of the designs. There were two outliers removed from the data (1 each from Georgia Control and Georgia +2BHF) prior to the analysis. The ANOVA (α = 0.05) in regards to peak load (Table 59) showed that the Georgia designs with added BHF exhibited statistical improvement over the Georgia Control. The South Carolina Control did not share a grouping with any of the designs and was therefore statistically different from all the other mixtures tested for this experiment. The South Carolina designs with added BHF also exhibited statistical improvement over the South Carolina Control. There was continu- ous numerical improvement for each design with added BHF; but the 2.0 and 4.0 BHF designs, for both Georgia and South Carolina, were not statistically different. Mix ID FI N Mean Grouping South Carolina Control 4 57.7 A Georgia +4BHF 6 39.1 A B South Carolina +2BHF 5 36.3 A B South Carolina +4BHF 6 34.8 A B Georgia Control 5 28.7 B Georgia +2BHF 5 21.8 B p = 0.007 R2 = 45% Table 58. ANOVA for I-FIT with increased P-200. 3.111 1.645 2.216 2.627 4.232 4.519 Av er ag e Pe ak L oa d (K N ) Georgia SBS Control Georgia +2BHF Georgia +4BHF South Carolina SBS Control South Carolina +2BHF South Carolina +4BHF Figure 71. No-notch I-FIT peak load results with increased P-200. Mix ID Peak Load N Mean Grouping Georgia +4BHF 6 4.519 A Georgia +2BHF 5 4.232 A Georgia Control 5 3.111 B South Carolina +4BHF 6 2.627 B C South Carolina +2BHF 6 2.216 C South Carolina Control 6 1.645 D p < 0.001 R2 = 95% Table 59. ANOVA for no-notch I-FIT peak load with increased P-200.
Part 2 89 The Gf of both the South Carolina and Georgia designs showed no statistical improvement by increasing the P-200 for each mixture (Figure 72). There was a continuous numerical improve- ment for each design, which may indicate that if the volumetric properties and permeability will allow more P-200, the mixture may be improved. The Georgia design could not be increased; however, the South Carolina design had room to increase the P-200 material even more than the additional 4.0% BHF. The ANOVA, Table 60, shows that the South Carolina poor mix had the lowest fracture energy results. While the additional BHF did improve the results for each mix, the difference was not significant enough to change the grouping. Experiment 2: Effect of Binder Modification Introduction This section discusses the differences in the Georgia mix design performance when binder modification and fiber content are altered. The expectation was that a change in the binder modification would provide insight into the resistance of PFC mixture to raveling and crack- ing. The Georgia design at the optimum AC was chosen for this part of the study. The fiber was omitted from this part of the study for dual purposes: (1) to determine if binder modification 3017 3762 4208 2572 2763 3386 0 1000 2000 3000 4000 5000 6000 Av er ag e Fr ac tu re E ne rg y (J /m 2) Georgia SBS Control Georgia +2BHF Georgia +4BHF South Carolina SBS Control South Carolina +2BHF South Carolina +4BHF Figure 72. No-notch I-FIT fracture energy results with increased P-200. Mix ID Fracture Energy N Mean Grouping Georgia +4BHF 6 4208 A Georgia +2BHF 5 3762 A B South Carolina +4BHF 6 3386 A B Georgia Control 5 3017 A B South Carolina +2BHF 6 2763 B South Carolina Control 6 2572 B p = 0.001 R2 = 49%. Table 60. ANOVA analysis for no-notch I-FIT fracture energy with increased P-200.
90 Performance-Based Mix Design of Porous Friction Courses alone would be sufficient to prevent draindown, and (2) to see if the fiber had an impact on the raveling or cracking potential of the mixture. The mix design components for this part of the study can be found in Table 61. The PG 76-22 (SBS) was the same binder used for Part 1, and the GTR binder was the same as the binder used for the Florida good mix in Part 1. The PG 82-22 (HiMA) was modified with a high dosage of SBS polymer that is approximately twice the typical dosage rate for PG 76-22. In order to ensure that the asphalt binder was the expected performance grade, samples of each binder were tested according to AASHTO M 320, Standard Specification for Performance-Graded Asphalt Binder. The results from this testing can be seen in Table 62. As can be seen in the table, all three asphalt binders graded as expected. Results and Discussion Volumetrics The Georgia Control design data (Table 63) are included in this section for comparison pur- poses. The mixture properties for this section are based on the Cantabro data. Cellulose fiber at a typical dosage rate of 0.4% was used in the control mix. The GTR design had a lower air void content compared to the other mixtures due to the increased asphalt volume from adding the GTR. The SBS design without fiber had a higher air void content than the Control design with fiber. The film thicknesses of the specimens was relatively uniform. The small deviations are likely due to the variation in mixture Gmb and Gmm. The GTR design had a higher film thickness due to the higher asphalt volume. A graphical representation of the effect of binder type on air voids and film thickness is shown in Figure 73 and Figure 74.The VCA ratio is based on the breakpoint sieve method. If the ratio is less than 1.0, it is assumed that stone-on- stone contact of the coarse particles exists. Table 61. Mix design components for Georgia with binder modifications. Design Component Mixture Type Georgia âGoodâ Aggregate Mineralogy Granite Asphalt Type PG 76-22 PG 67-22 PG 82-22 Binder Modifier 2.5% SBS 12% -#30 GTR 7.5% SBS Anti-strip 0.5% LOF 6500 by weight of binder Fiber, % 0.0 AC, % 6.0 P-200, % 2.0 2.0 2.0 Expected Binder Grade Blending Additive Continuous Grade PG Grade PG 76-22 Manufacturer SBS 77.1 - 25.4 76 - 22 PG 76-22 NCAT Lab GTR 81.0 - 23.8 76 - 22 PG 82-22 Manufacturer SBS 87.4 - 26.8 82 - 22 Table 62. Asphalt binder grade summary (AASHTO M320).
Part 2 91 Mix ID Total AC (%) Fiber (%) Avg. Air Voids (%) Average VMA (%) Average VCAMIX/VCADRC Film Thickness Georgia Control 6.0 0.4 15.4 26.6 0.82 27.1 Georgia PG 76-22 SBS 6.0 0.0 16.3 26.5 0.81 25.2 Georgia PG 76-22 GTR 6.7 0.0 13.9 25.9 0.80 28.2 Georgia PG 82-22 SBS 6.0 0.0 14.9 25.5 0.79 26.1 Table 63. Mixture properties summary with binder modification. 15.4 16.3 13.9 14.9 12.5 13.0 13.5 14.0 14.5 15.0 15.5 16.0 16.5 17.0 Co re Lo k Ai r V oi ds (% ) Georgia PG 82-22 SBS Georgia Control PG 76-22 Georgia PG 76-22 SBS Georgia PG 76-22 GTR Figure 73. CoreLok air voids with binder modification. 27.1 25.2 28.2 26.1 22.0 23.0 24.0 25.0 26.0 27.0 28.0 29.0 30.0 Av er ag e Fi lm T hi ck ne ss (m ic ro ns ) Georgia PG 82-22 SBS Georgia Control Georgia PG 76-22 SBS Georgia PG 76-22 GTR Figure 74. Film thickness with binder modification.
92 Performance-Based Mix Design of Porous Friction Courses Binder and Fiber Effect on Cantabro Results The Cantabro data showed that the designs without fiber performed better than the design with fiber. This seems to indicate that with the use of modified binders, fiber stabilizers may not be needed. The summary table (Table 64) and Figure 75 show that the SBS and GTR samples had similar percent loss, while the HiMA design had negligible loss. There were three specimens tested for each design so the sample size is adequate; however, there is a high COV of the SBS design so that any conclusions regarding Cantabro loss for that binder may be misleading. The almost neg- ligible loss for the HiMA design was expected due to the large amount of polymer in the binder. It was expected that the HiMA design would perform well for both durability and crack resistance. In the ANOVA, The difference between the control and the SBS mix is that the control mix contained 0.4% fiber stabilizer. These results indicate the fiber does not add additional strength or cohesion as measured by resistance to abrasion loss in the Cantabro Test. Table 65, shows that some of the mixtures were statistically different with a good model fit (R2 = 66.2%). The Georgia Control was statistically different from the HiMA design. All designs met the 20.0% maximum loss criterion. Even though the SBS and GTR designs are not statistically different from the Control, there is numerical improvement over the Control. The difference between the control and the SBS mix is that the control mix contained 0.4% fiber stabilizer. These results indicate the fiber does not add additional strength or cohesion as measured by resistance to abrasion loss in the Cantabro Test. Table 64. Summary of Cantabro loss with binder modifications. Mix ID Total AC (%) Fiber (%) Average Air Voids (%) Cantabro Loss (%) Average St. Dev. COV (%) Georgia Control 6.0 0.4 15.2 19.3 6.4 33.4 Georgia PG 76-22 SBS 6.0 0.0 16.1 12.3 6.2 50.3 Georgia PG 76-22 GTR 6.7 0.0 12.8 12.1 0.8 7.0 Georgia PG 82-22 SBS 6.0 0.0 14.4 4.7 1.1 23.2 19.3 12.3 12.1 4.7 0.0 5.0 10.0 15.0 20.0 25.0 30.0 Georgia Control Georgia PG 76-22 SBS Georgia PG 76-22 GTR Georgia PG 82-22 SBS Av er ag e Ca nt ab ro L os s ( % ) Figure 75. Cantabro loss for binder modification designs.
Part 2 93 Binder Effect on Permeability The permeability summary can be seen in Table 66, while a graphical depiction of the results is shown in Figure 76. The Control and SBS had similar permeability results, which may indicate that the fiber had no effect on the permeability of the specimens. The permeability of the GTR and HiMA designs considerably decreased over the Control. The GTR design had a higher vol- ume of asphalt due to the addition of GTR modifier, which may explain the decrease in perme- ability, but the only difference in the HiMA design was the increase in the polymer dosage. The HiMA has more than twice the amount of polymer as the SBS design. This increase in polymer did not decrease the air void content of the design but did significantly affect the permeability. As shown in Table 66, the COV for air voids with the different binders was less than 10%. Therefore, air void variation does not explain the reduced permeability for the GTR and HiMA mixes. COV is heavily influenced by the magnitude of the average; so in this case, all four mixes have similar variability as expressed in the standard deviation, but the two mixtures with a lower average k value appear to have much higher variability if looking only at the COV. Additional polymer and possible swelling of the GTR particles by absorbing oil from the binder may explain the low permeability for those two mixes. Hamburg Results by Binder Type The HWTT results (Figure 77) show that the SBS and GTR designs had a slightly greater rut depth than the Control design. The COV, seen in Table 67, for the SBS and Control designs are relatively high. This may account for the small difference in the measured rut depth. The HiMA design had the smallest rut depth and the lowest COV. The HiMA design has the stiffest asphalt binder so these results were expected. The Control performed better than the SBS and GTR design, but it is likely that the differences are within normal testing variability. Mix ID Cantabro Loss (%) N Mean Grouping Georgia Control 3 19.3 A Georgia PG 76-22 SBS 3 12.3 A B Georgia PG 76-22 GTR 3 12.1 A B Georgia PG 82-22 SBS 3 4.7 B p = 0.027 R2 = 66% Table 65. ANOVA for Cantabro loss with binder modifications. Mix ID AC (%) Fiber (%) Average Air Voids (%) Permeability (k) meters/day Avg. St. Dev. COV (%) Avg. St. Dev. COV (%) Georgia Control 6.0 0.4 15.7 0.5 2.9 80 10.5 13.1 Georgia PG 76-22 SBS 6.0 0.0 16.6 0.5 2.7 79 12.2 15.3 Georgia PG 76-22 GTR 6.7 0.0 13.3 1.2 8.6 33 14.1 42.5 Georgia PG 82-22 SBS 6.0 0.0 15.1 0.0 0.2 37 9.8 26.6 Table 66. Permeability summary with binder modification.
94 Performance-Based Mix Design of Porous Friction Courses 80 79 33 37 0 10 20 30 40 50 60 70 80 90 100 Georgia Control Georgia PG 76-22 SBS Georgia PG 76-22 GTR Georgia PG 82-22 SBS Av er ag e Pe rm ea bi lit y (k ) m et er /d ay Figure 76. Permeability results for binder modification. Figure 77. HWTT results for binder modification. Mix ID Total AC (%) Fiber (%) Average Air Voids (%) Greatest Rut Depth Recorded (mm) Average, mm St. Dev., mm COV (%) Georgia Control 6.0 0.4 14.3 8.99 2.88 32.0 Georgia PG 76-22 SBS 6.0 0.0 17.3 10.56 3.09 29.2 Georgia PG 76-22 GTR 6.7 0.0 13.5 10.84 1.64 15.1 Georgia PG 82-22 SBS 6.0 0.0 15.5 6.89 0.36 5.2 Table 67. HWTT summary with binder modification.
Part 2 95 Effect of Binder Type on Moisture Susceptibility The TSR summary results are presented in Table 68. The unconditioned strengths dropped slightly for all of the designs with binder modification, but according to the ANOVA in Table 69 there is not a statistical difference between the unconditioned strengths. There is a difference between the conditioned strengths. The Control design was statistically different from the SBS and HiMA designs. The SBS and HiMA designs showed a decrease in both the unconditioned and conditioned strengths when compared to the Control. This seems to indicate the fiber sta- bilizer may provide some cohesive strength or additional stability to the mix. The GTR and HiMA mixtures passed the ASTM requirement of 0.80, while all of the designs passed the AASHTO criterion of 0.70. The GTR design showed almost no moisture damage. This is important to note since the Florida GTR design from Part 1 also showed no sign of mois- ture damage. Therefore, the use of GTR in a PFC design may provide resistance to moisture damage. Additional laboratory testing with and without anti-strip agents is needed to verify if this is correct. Overlay Test Results The OT results for this part of the study can be seen in Table 70 and Figure 78. The SBS and HiMA designs did not reach the 93% loss reduction prior to reaching 1,000 cycles, so the data had to be extrapolated. The GTR design has a large COV (73.7%), while the other designs had reason- able COVs. The GTR design ranged from 112 to786 cycles to failure for this testing. There were no outliers observed in the GTR design, but there was a single outlier observed in the Control and SBS designs. Both of these outliers were removed prior to performing any analysis. Mix ID Total AC (%) Fiber (%) Average Air Voids (%) Average ITS (psi) TSR Conditioned Unconditioned Conditioned Unconditioned GA Control 6.0 0.4 13.9 14.0 57.6 74.3 0.78 GA PG 76-22 SBS 6.0 0.0 15.9 16.0 48.8 61.6 0.79 GA PG 76-22 GTR 6.7 0.0 14.6 14.8 66.8 70.2 0.95 GA PG 82-22 SBS 6.0 0.0 13.4 13.4 56.0 63.3 0.88 Table 68. TSR summary results with binder modification. Mix ID Conditioned ITS Unconditioned ITS N Mean Grouping N Mean Grouping Georgia Control 3 57.6 A 3 74.3 A Georgia PG 76-22 GTR 3 66.8 A B 3 70.2 A Georgia PG 82-22 SBS 3 56.0 B 3 60.3 A Georgia PG 76-22 SBS 3 48.8 B 3 61.6 A p = 0.002 R2 = 82% p = 0.346 R2 = 32% Table 69. ANOVA for ITS with binder modification.
96 Performance-Based Mix Design of Porous Friction Courses An ANOVA was conducted on both the peak load and cycles to failure data. Both sets of analysis have goodness of fit values greater than 90%, which indicates that the models show a good fit with the data. The cycles to failure data (Table 71) shows that the Control and GTR designs are statistically different from the HiMA and SBS designs. This was expected since the HiMA and SBS designs did not fail prior to the 1,000 cycle cutoff. The HiMA and the SBS designs are modified only with polymer and perform well according to the OT. The SBS design without fiber performed 3.5 times better than the design with fiber (Control). This may indicate that the fiber stiffens the mastic and absorbs some of the available free binder, which reduces the cracking resistance of the mix in regard to OT performance. The peak load analysis (Table 72) showed that the HiMA design was statistically different from the other designs. The HiMA design had the lowest peak load recorded of the four designs but had the highest cycles to failure, while the GTR design had the highest peak load and the lowest recorded cycles to failure. This seems to suggest that the performance difference may be related to elasticity of the designs. The HiMA design has a significant amount of polymer, which Mix ID Replicates Average Air Voids (%) Average Peak Load (kN) Cycles to Failure Average St. Dev. COV (%) Georgia Control 3 12.8 2.621 583 166 28.4 Georgia PG 76-22 SBS 3 13.3 2.129 2137 104 4.9 Georgia PG 76-22 GTR 4 11.6 2.977 388 286 73.7 Georgia PG 82-22 SBS 4 12.3 1.243 2877 551 19.1 Table 70. OT summary for binder modification designs. 583 2137 388 2877 0 500 1000 1500 2000 2500 3000 3500 4000 Georgia Control Georgia PG76- 22 SBS Georgia PG76- 22 GTR Georgia PG82- 22 SBS Av er ag e O T Cy cl es to F ai lu re Figure 78. OT results for binder modification designs. Mix ID Cycles to Failure N Mean Grouping Georgia PG 82-22 SBS 4 2877 A Georgia PG 76-22 SBS 3 2137 A Georgia Control 3 583 B Georgia PG 76-22 GTR 4 388 B p < 0.001 R2 = 93% Table 71. ANOVA for OT cycles to failure results with binder modification.
Part 2 97 provides greater elastic recovery. The cracking resistance of the GTR as measured by cycles to failure is significantly less than that of the polymer. Effect of Binder Type on Draindown The draindown results for this part of the study (Table 73) were the most pertinent because the stabilizing agent (fiber) had been removed from these designs. The purpose was to see if binder modification would negate the need for fiber as a draindown solution while still provid- ing a well performing mixture. All of the samples were tested at 330°F and 357°F except for the HiMA design. The HiMA design required a mixing temperature of 340°F; therefore, the test temperatures for the HiMA design were 340°F and 367°F. The HiMA design tested at 367°F was the only sample to fail the 0.3% maximum draindown criterion. The samples tested at 357°F for the SBS and 340°F for the HiMA designs were close to failing. The GTR design showed almost no draindown. This may indicate that the use of GTR negates the need for fiber when incorporated in PFC mixtures. It should be noted that during fabrication of specimens for the SBS and HiMA designs, draindown was observed in the aging pans for the Cantabro and permeability specimens that were com- pacted at their recommended compaction temperatures (300°F and 320°F). The performance samples that were aged and compacted at 275°F showed minimal draindown in the aging pans for the SBS and HiMA designs. I-FIT Results The I-FIT results of notched samples with the binder modifications can be seen in the fol- lowing figures. There was one outlier removed from the GTR design set, while the Control set also had one outlier that was removed in the Part 1 analysis. In regards to peak load, the GTR design was statistically different from the SBS and HiMA designs but not from the Control (Fig- ure 79 and Table 74). The Gf and FI results for notched samples showed a numerical increase Mix ID Peak Load N Mean Grouping Georgia PG 76-22 GTR 4 2.977 A Georgia Control 3 2.621 A B Georgia PG 76-22 SBS 3 2.129 B Georgia PG 82-22 SBS 4 1.243 C p < 0.001 R2 = 93% Table 72. ANOVA for OT peak load results with binder modification. Mix ID Total AC (%) Fiber (%) Draindown (%) Test Temp, °F 330 357 340 367 Georgia Control 6.0 0.4 0.0 0.0 Georgia PG 76-22 SBS 6.0 0.0 0.0 0.3 Georgia PG 76-22 GTR 6.7 0.0 0.0 0.0 Georgia PG 82-22 SBS 6.0 0.0 0.3 1.0 Table 73. Draindown results for binder modification designs.
98 Performance-Based Mix Design of Porous Friction Courses with the binder modifications over the Control, but the designs were not statistically different except for the HiMA design (Figure 80 and Table 75). The HiMA design showed an obvious increase in FI over the other designs (Figure 81) and was statistically different from the other mix designs. The model had a goodness of fit of 57.43% (Table 76). The HiMA with the high polymer content provides more elasticity to the mixture, allowing some recovery to occur after the peak load and fracture has transpired. The no-notch I-FIT specimens showed an increase in the peak load for the SBS design when compared to the Control (Figure 82). There was a single outlier removed from the Control and GTR designs prior to performing the analysis. The HiMA design was statistically different from the other designs, with only an average peak load of 2.413 kN. The other average peak loads were greater than 3.00 kN (Table 77). These results show the same trend as the OT results in regards to the peak load. The Gf results show that the GTR and HiMA designs are statistically different. It was surpris- ing that Gf values for the no-notch samples modified with GTR were lower than corresponding mixes modified with polymer (Figure 83 and Table 78). The indication from these results is that the polymer-modified mixes are more flexible than the GTR mix. This is contrary to the field experience evidenced by the good performing mixes in Florida and New Jersey that used GTR, unless the field performance was affected more by other factors such as quality of construction than by type of binder modification. The Gf of the HiMA design is higher than all other designs. However, the reported Gf may be even lower than the actual value, because three of the HiMA samples hit the backstop on the I-FIT machine and never terminated. The specimens never reached the 0.1 kN cutoff; therefore, 1.694 1.603 2.046 1.426 0.000 0.500 1.000 1.500 2.000 2.500 Georgia Control Georgia PG 76- 22 SBS Georgia PG 76- 22 GTR Georgia PG 82- 22 SBS Av er ag e Pe ak L oa d (k N ) Figure 79. Notched I-FIT peak load results for binder modification designs. Mix ID Peak Load (kN) N Mean Grouping Georgia PG 76-22 GTR 5 2.046 A Georgia Control 5 1.694 A B Georgia PG 76-22 SBS 6 1.603 B Georgia PG 82-22 SBS 6 1.426 B p = 0.003 R2 = 52% Table 74. ANOVA for notched I-FIT peak load for binder modification designs.
Part 2 99 1,828 2,034 2,178 2,475 0 500 1000 1500 2000 2500 3000 3500 Georgia Control Georgia PG 76- 22 SBS Georgia PG 76- 22 GTR Georgia PG 82- 22 SBS Av er ag e Fr ac tu re E ne rg y (J /m 2) Figure 80. Notched I-FIT fracture energy for binder modification designs. Mix ID Fracture Energy (J/m2) N Mean Grouping Georgia PG 82-22 SBS 6 2475 A Georgia PG 76-22 GTR 5 2179 A Georgia PG 76-22 SBS 6 2034 A Georgia Control 5 1828 A p = 0.253 R2 = 20% Table 75. ANOVA for notched I-FIT fracture energy with binder modifications. 28.7 38.7 31.1 86.0 0.00 20.00 40.00 60.00 80.00 100.00 120.00 140.00 Georgia Control Georgia PG 76- 22 SBS Georgia PG 76- 22 GTR Georgia PG 82- 22 SBS Av er ag e Fl ex ib ili ty In de x (F I) Figure 81. Notched I-FIT flexibility index for binder modification designs. Mix ID FI N Mean Grouping Georgia PG 82-22 SBS 6 86.0 A Georgia PG 76-22 SBS 6 38.7 B Georgia PG 76-22 GTR 5 31.1 B Georgia Control 5 28.7 B p < 0.001 R2 = 57% Table 76. ANOVA for notched I-FIT flexibility Index with binder modification.
100 Performance-Based Mix Design of Porous Friction Courses 3.111 3.302 3.110 2.413 0.000 0.500 1.000 1.500 2.000 2.500 3.000 3.500 4.000 Georgia SBS Control Georgia PG 76-22 SBS Georgia PG 76-22 GTR Georgia PG 82-22 SBS Av er ag e Pe ak L oa d (k N ) Figure 82. No-notch I-FIT peak load with binder modifications. Mix ID Peak Load (kN) N Mean Grouping Georgia PG 76-22 SBS 6 3.302 A Georgia Control 5 3.111 A Georgia PG 76-22 GTR 5 3.110 A Georgia PG 82-22 SBS 6 2.413 B p < 0.001 R2 = 68% Table 77. ANOVA for no-notch I-FIT peak load with binder modifications. 3,017 3,819 2,579 4,567 0 1000 2000 3000 4000 5000 6000 7000 Georgia SBS Control Georgia PG 76- 22 SBS Georgia PG 76- 22 GTR Georgia PG 82- 22 SBS Av er ag e Fr ac tu re E ne rg y (J /m 2) Figure 83. No-notch I-FIT fracture energy with binder modifications. Mix ID Fracture Energy (J/m2) N Mean Grouping Georgia PG 82-22 SBS 6 4567 A Georgia PG 76-22 SBS 6 3819 A B Georgia Control 5 3017 A B Georgia PG 76-22 GTR 5 2579 B p = 0.011 R2 = 45% Table 78. ANOVA for no-notch I-FIT fracture energy with binder modifications.
Part 2 101 the machine had to be manually stopped. Since the specimens did not reach the 0.1 kN cutoff, all of the area under the curve was not attainable. The data was trimmed at its lowest point after the peak load for the Gf calculations. Experiment 3: Evaluation of the Effect of Layer Thickness on Performance Introduction The determination of splitting tensile strength (or ITS) at varying lift thicknesses was per- formed to determine if the NMAS of the mixture had an effect on ITS. This testing was important, because PFC mixtures are often placed at thicknesses less than 1 inch. Therefore, the standard dimensions of TSR samples for dense-graded mixes may not apply to PFC mixes, or may not be representative of field conditions for PFC mixes. Comparison testing of a 4.75 mm dense- graded mixture along with 9.5 mm and 12.5 mm PFC mixtures was conducted. The 4.75 mm dense-graded mix was selected because it is also often placed less than 1 inch thick. If ITS of the 4.75 mm mix is not affected by variation in layer thickness, it would be reasonable to assume the same may be true for PFC mixes. The 4.75 mm samples were prepared from mix produced for Lee County Road 159 test sections near Auburn, Alabama, and the 9.5 mm PFC was prepared from mix samples for NCAT Test Track section E9-1a. The Georgia Control mix used in Part 1 was selected for the 12.5 mm PFC mix. The test thicknesses of 2.5 in., 0.75 in., and 2.5 x NMAS were chosen for the evaluation. At 2.5 times NMAS, thicknesses resulted in approximately 0.5 inch for the 4.75 mm mix, 1.0 inch for the 9.5 mm mix, and 1.25 in. for the 12.5 mm mix. Six-inch diameter samples were com- pacted to standard height with 50 gyrations using a gyratory compactor and sawed to the desired testing thicknesses. The 4.75 mm dense-graded mixture specimens were prepared for testing according to AASHTO T 283. According to AASHTO T 283, if the specimen thickness is less than 2.5 in., the specimen diameter must be 4.0 in. Therefore, in order to conform to the AASHTO T 283 requirement, a 4.0 inch diameter core was taken from the 6-inch lab- compacted specimens for thicknesses other than 2.5 in. The PFC specimens for this part of the study were conditioned in the same manner as the TSR samples from the previous sections. Results and Discussion The specimens were all saturated and subjected to one freezeâthaw cycle prior to testing. Three specimens were tested at each thickness for each of the different NMAS mixtures. A sum- mary of the results can be seen in Table 79. The dense-graded mixture had a larger ITS than both the PFC mixtures. This was expected due to the difference in air voids between the dense-graded mix and the PFC samples. An ANOVA (α = 0.05) was conducted on the ITS results from each mix design to determine if ITS for the varying lift thicknesses were statistically different. According to the analysis shown in Table 80, Table 81 and Table 82, the lift thickness had no effect on the ITS. The analysis for the 12.5 mm mixture had a p-value of 0.071. This is close to the distinguishing value of 0.05. According to the Statistical Sleuth (Ramsey and Schafer, 2002), a p-value of 0.05 to 0.10 is sug- gestive but inconclusive. Bennert and Cooley (2014) reported, however, that thinner layers resulted in increased tensile stresses at the surface and may increase the potential for top-down cracking.
102 Performance-Based Mix Design of Porous Friction Courses Mix ID NMAS (mm) AC (%) Sample Diameter, in. Sample Thickness, in. Sample Air Voids (%) Avg. ITS (psi) Std. Dev. ITS (psi) COV ITS LR 159 2.5xNMAS 4.75 6.2 4.0 0.45 4.8 175 2.2 1.2 LR 159 0.75inch 4.75 6.2 4.0 0.79 6.8 183 10.0 5.5 LR 159 2.5 inch 4.75 6.2 6.0 2.50 6.4 187 8.0 4.3 E9-1a 2.5xNMAS 9.5 6.0 4.0 0.93 13.6 95 3.2 3.3 E9-1a 0.75 inch 9.5 6.0 4.0 0.79 14.8 83 6.4 7.7 E9-1a 2.5 inch 9.5 6.0 6.0 2.51 16.9 89 13.5 15.3 GA 2.5xNMAS 12.5 6.0 4.0 1.22 11.4 95 7.9 8.3 GA 0.75 inch 12.5 6.0 4.0 0.73 10.1 83 11.4 13.7 GA 2.5 inch 12.5 6.0 6.0 2.53 12.9 75 3.3 4.3 Table 79. Summary results for ITS based on lift thickness. Source DF SS MS F p Mix ID 2 231.0 115.5 2.04 0.210 Error 6 338.9 56.5 Total 8 569.9 Std. Dev. = 7.515 R2 = 41% Note: DF = degrees of freedom, SS = sum of squares, MS = mean of squares, F = F-statistic, p = p-value. Table 80. ANOVA of ITS for the 4.75 mm dense-graded mixture. Source DF SS MS F p Mix ID 2 209.1 104.5 1.34 0.330 Error 6 468.4 78.1 Total 8 677.5 Std. Dev. = 8.835 R2 = 31% Table 81. ANOVA of ITS for the 9.5 mm PFC design. Source DF SS MS F p Mix ID 2 578.1 289.1 4.25 0.071 Error 6 408.1 68 Total 8 986.2 Std. Dev. = 8.247 R2 = 59% Table 82. ANOVA of ITS for the 12.5 mm Georgia PFC design.
