MULTIVARIATE ANALYSIS ON RUTTING OF COSTA RICAN ASPHALT CONCRETE MIXES

Transcription

1 MULTIVARIATE ANALYSIS ON RUTTING OF COSTA RICAN ASPHALT CONCRETE MIXES Pedro Castro-Fernandez, Ph.D. Candidate, Professor, Civil Engineering Department, University of Costa Rica Dr. Peter E. Sebaaly, Ph.D., Professor, Civil Engineering Department, University of Nevada - Reno Fabricio Leiva-Villacorta, PE Elizabeth Carmona-Rojas, PE José Pablo Aguiar-Moya, PE Jaime Allen-Monge, PE Gustavo Badilla-Vargas, PE Luis Cordero-Jimenez, PE Wendy, PE Civil Engineering Department, University of Costa Rica Escuela de Ingeniería Civil, Ciudad Universitaria Rodrigo Facio, Universidad de Costa Rica, San Pedro de Montes de Oca, San José, Costa Rica. Tel Fax pcastrof@racsa.co.cr Word count: 7641 This paper is submitted in response to the Call for Papers on Latin America sponsored by the International Activities Committee A0010 TRB 2006 Annual Meeting CD-ROM Paper revised from original submittal.

2 Castro-Fernandez, Sebaaly, Leiva-Villacorta, Carmona-Rojas, Aguiar-Moya, Allen-Monge, Badilla-Vargas, Cordero, Jimenez, 2 ABSTRACT Despite of having improved quality control / quality assurance techniques in recent years, rutting is a very common distress in Costa Rica s flexible pavements. Newer specifications and better materials, capable of matching current transit loads and environmental conditions, are required. This paper addresses the issue of rutting by considering alternatives to customary dense graded asphalt concrete, such as different types of gradations and additives. A full factorial experiment was addressed to identify key mix design considerations in regard to plastic deformation improvement. Three representative aggregate sources have been considered, together with four different binders (two neat binders and two SBS modified binders) and six different gradations (three dense graded and three gap graded). Testing was performed at both optimum asphalt content and maximum allowable asphalt content at a production facility (0.5 % over optimum). It was found that SBS modified binders outrank SMA gradations and neat binders by far. Aggregate source, gradation and asphalt content were found of very little effect, in comparison to the addition of SBS. When neat binders were used, nevertheless, aggregate source, gradation and type of asphalt binder became more important. As second stage of research, the best performing mixes were selected in order to conduct additional testing. From these test results it was concluded that the benefit of SBS modified binders is quite significant. It was also found that among dense gradations, current Costa Rican gradations showed the most rutting potential. Multivariate statistical techniques were applied to handle results, and to support final conclusions. 1. INTRODUCTION Costa Rican asphalt concrete mixes have not performed very well for a long time; moisture damage and rutting usually take place very shortly after construction. Despite having developed new quality control / quality assurance specifications, Costa Rica is still struggling to find better procedures and materials, which will ensure better performance at the short term. There is a need of new specifications that will not only allow, but also enforce, better technologies and materials. Searching for better performing hot mixed asphalt concrete, some research has been done, targeting for longer lasting mixes; this papers presents part of its results. The main focus of this paper is to identify specific actions to improve rut resistance, in regard to laboratory performance. Not only standard Costa Rican dense graded curves are considered, but also other dense gradations found to perform better in earlier research, as well as some gap graded mixes (stone matrix asphalt). Dense graded mixes with SBS modified binders are considered too. TRB 2006 Annual Meeting CD-ROM Paper revised from original submittal.

3 Castro-Fernandez, Sebaaly, Leiva-Villacorta, Carmona-Rojas, Aguiar-Moya, Allen-Monge, Badilla-Vargas, Cordero, Jimenez, 3 The main goals of this paper are: To determine which gradation curves perform the best in regard to rutting. To assess the benefit of SMA gradations in regard to rutting, when neat binders are used. To determine whether mixes with SBS modified binders are less susceptible to rutting than mixes with neat binders. To determine how asphalt binder type effects rutting. Two asphalt binders are considered, standard AC30 (broadly used in Costa Rica) and PG70 (22). To determine how aggregate source effects rutting. To assess asphalt content effect on rutting. 2. BACK GROUND. Rutting continues to be a problem in the performance of hot mixed asphalt concrete. It is defined as the accumulation of small amounts of unrecoverable strain resulting from applied loads. Its sources are consolidation (reduction of air voids without lateral flow), and / or lateral movement (plastic flow due to shear failure). In recent years, the potential for rutting has increased, due to increased traffic volumes and higher inflation pressures (Ref. 1). Use of excessive asphalt cement is a common cause of lateral flow. Too much asphalt cement in the mix determines the loss of internal friction between aggregate particles, therefore the load is carried by the asphalt binder rather than the aggregate. Plastic flow is minimized by large size aggregate, angular and rough textured aggregates, and by adequate compaction. Some increased rutting resistance can be obtained by using stiffer asphalt binders (Ref. 2). Stone matrix asphalt mixes (SMA) are designed in a way that allows both stone-to-stone contact (therefore high strength is achieved), and a rich mortar binder (therefore moisture damage is minimized). Although new in United States of America, they are gaining popularity nation-wide (Ref. 3). National Center for Asphalt Technologies (NCAT) evaluated the performance of 85 SMA projects within USA; their main findings were: Ninety percent of SMA projects had rutting measurements of less than 4 mm. Twenty five percent showed no rutting at all. SMA is apparently more resistant to cracking than dense graded mixes. There was no evidence of raveling on SMA projects (Ref. 3). Elastometers or rubbers are polymers that resist deformation from applied stress by stretching and recovering their shape quickly when stress is removed. Styrene butadiene styrene (SBS) is an elastometer, and it allows asphalt concrete mixes to increase their tensile strength with elongation. Elastometers usually employed in asphalt binder modification normally allow for more flexible and resilent layers (Ref. 2). TRB 2006 Annual Meeting CD-ROM Paper revised from original submittal.

4 Castro-Fernandez, Sebaaly, Leiva-Villacorta, Carmona-Rojas, Aguiar-Moya, Allen-Monge, Badilla-Vargas, Cordero, Jimenez, 4 The asphalt pavement analyzer (APA) is a modification of the original Georgia loaded wheel tester. It consists of a loaded wheel over a pressurized linear hose which is tracked back and forth over a sample, to induce rutting. Laboratory rut depths measured by the APA have had good correlations with rut depths from other performance tests; although correlation to actual site rut depths is still not so good (Ref. 1). 3. EXPERIMENTAL DESIGN. This is a multi-factor random experiment. The following factors have been studied: Aggregate source: three levels, sedimentary crushed aggregate from the Eastern shore (Guapiles), and two sources of igneous crushed aggregate from the Central Region: a) lower absorption (Santa Ana), and b) higher absorption (Las Concavas). The Guapiles source is representative of about 60 % of current hot mixed asphalt concrete production in the country (most works done); the Santa Ana source is representative of about 20 % of current hot mixed asphalt concrete production in the country; and the Las Concavas source is representative of the kind of aggregate that used to be employed in the past (higher moisture damage susceptibility). Asphalt binder: two levels in regard to stone matrix asphalt (SMA), AC30 and PG 70 (22); and four levels in regard to dense graded mixes (DGM), neat AC30, neat PG 70 (22), SBS modified AC30 (2.5 %), and SBS modified PG70 (1.5 %). Gradation: three dense gradations; DGM1 comes from current Costa Rican standards (nominal maximum size of 12.5 mm is the most common); DGM2 (tighter granular structure, closer to maximum density line) and DGM3 (higher content of very coarse and very fine aggregate). DGM2 and DGM3 have been found to perform better than DGM1 in previous research. Three stone matrix asphalt gradations; SMA1 presents the highest content of aggregate finer than No. 4 but coarser than No. 200, and SMA3 presents the lowest content of aggregate finer than No. 4 but coarser than No. 200, with SMA2 in between. Notes: (1) Costa Rica still uses an asphalt binder classification system based on absolute viscosity at 60 C; although at certain times, special types of PG grades are supplied for particular projects. (2) PG grade in Costa Rica is reported as follows: PG xx (yy) refers to a binder whose maximum performance temperature is xx and its intermediate performance temperature is yy. No minimum performance temperature is tested, since the country doesn t have freezing temperatures. Maximum and intermediate performance temperatures are determined according to AASHTO MP1; aging on PAV is required to be at 100 C. TRB 2006 Annual Meeting CD-ROM Paper revised from original submittal.

5 Castro-Fernandez, Sebaaly, Leiva-Villacorta, Carmona-Rojas, Aguiar-Moya, Allen-Monge, Badilla-Vargas, Cordero, Jimenez, 5 Table 1 shows all treatment compositions, as combinations of aggregate source, asphalt binder and gradation. Rutting tests were conducted at both optimum asphalt content and critical asphalt content (optimum asphalt content plus 0.50 % by total weight of mix). 4. RAW MATERIALS All asphalt binders sampled and tested were produced by Costa Rican Asphalt Refinery (RECOPE), from the same oil crude (Boscan, Venezuela). Table 2 shows the properties of the asphalt binders. All asphalt binders show standard aging susceptibility, since all of them show an absolute viscosities ratio below 3.0 (RTFO to original at 60 C), and a RTFO mass loss below 1.0 %. All asphalt binders show standard thermal susceptibility, since all of them show a penetration index between 2.0 and +1.0, and a VTS index below 3.3. SBS modified AC30 is the least aging and thermal susceptible. According to AASHTO MP1 Performance grade specifications, the SBS modified binders are more rut resistant (both show a maximum performance temperature of 76 C), whereas both the AC30 and the SBS modified AC30 are more fatigue resistant (both show an intermediate performance temperature of 19 C). SBS polymer has an effect on binder maximum performance temperature; standard grades improve from 64 to 76 for AC30 and from 70 to 76 for PG70 (22). SBS has little effect on binder intermediate temperature; standard grades for both AC30 and PG70 (22) remain unchanged. Guapiles and Santa Ana aggregate samples were taken from stockpiles at actual asphalt mix plants; they represent what was being fed in them at the moment of sampling. Las Concavas aggregate samples were taken from stockpiles at the crusher plant; such aggregates are not currently been used in actual hot mixed asphalt concrete production. Figure 1 shows considered dense gradations, whereas Figure 2 shows stone matrix asphalt gradations. DGM gradations were selected according to what is used in Costa Rica (DGM1), and to what has been tested lately in order to improve performance (DGM2 and DGM3); in all cases Superpave gradation specifications were fulfilled. SMA gradations were selected according to earlier research, although the content of mineral dust had to be increased in order to fulfill air void requirements. SMA gradations had to be slightly changed for every aggregate source; proving the point of SMA gradations being highly aggregate dependant. In addition, SMA gradations fulfill NAPA gradation specifications (Ref. 3), except for the higher amounts of mineral dust used. Note: even though higher amounts of mineral dust could determine more rutting, it was required to go over the standard NAPA specifications on dust contents in order to achieve lower air voids. TRB 2006 Annual Meeting CD-ROM Paper revised from original submittal.

6 Castro-Fernandez, Sebaaly, Leiva-Villacorta, Carmona-Rojas, Aguiar-Moya, Allen-Monge, Badilla-Vargas, Cordero, Jimenez, 6 Table 3 shows aggregate properties. It can be seen how the Santa Ana aggregate is the heaviest (highest specific gravity), although the Guapiles source is the toughest (lowest L.A. abrasion). The Las Concavas aggregate is the most absorptive (dense gradations show water absorptions over 3.5 %). Fine aggregate angularity is poor in DGM1 (conventional in Costa Rica); although it normally improves in DGM2 and DGM3 (not conventional). Sand equivalent is low in Las Concavas aggregates, but well over 60 % in Guapiles and Santa Ana aggregates. All coarse aggregates are quite angular, since in all cases at least 95 % of the coarse aggregate shows two or more fractured faces. Notes: (1) Crushed limestone dust has been included in all SMA mixes, as a source of mineral filler. No mineral filler was added in DGM mixes. (2) Half percent of cellulose fiber by dry weight of aggregate was added in SMA mixes (pellets). It is supplied with an AC40 coating (28 % by weight). 5. SUPERPAVE VOLUMETRIC MIX DESIGN A Superpave volumetric mix design was performed for all treatments presented in Table 1, in order to select asphalt content levels to be used in rutting tests: a) optimum asphalt content (targeting for 4.0 % air voids at the design gyrations number), and b) highest allowable asphalt content in mix plant operations (0.50 % by total weight of mix over optimum). Table 4 shows volumetric mix design target properties, for both dense graded mixes and stone matrix asphalt mixes. Not all requirements were fulfilled at all times, specially in regard to stone matrix asphalt mixes, where it was hard to achieve very high contents of voids filled with asphalt. It has to be considered, nevertheless, that high contents of mineral dust could be considered as extenders, which would make it up for the reduced VMA on SMA mixes. Volumetric properties of mixes with optimum asphalt contents for all treatments are shown in Figures 3 and 4. Stone matrix asphalt mixes tend to allow for higher effective asphalt contents, specially when sedimentary aggregate from the Eastern shore was employed (Guapiles). Also stone matrix asphalt mixes show higher voids filled with asphalt (VFA). Very similar volumetrics were obtained for dense graded mixes with neat and modified binders. 6. FULL TREATMENT RUTTING ANALYSIS. At this stage all treatments were considered, using both optimum and highest allowable asphalt contents. Highest allowable asphalt content is considered as the critical effective asphalt content in regard to moisture damage (expecting increased damage when it is applied). TRB 2006 Annual Meeting CD-ROM Paper revised from original submittal.

7 Castro-Fernandez, Sebaaly, Leiva-Villacorta, Carmona-Rojas, Aguiar-Moya, Allen-Monge, Badilla-Vargas, Cordero, Jimenez, 7 Rutting was assessed by means of the Georgia wheel track (Asphalt Pavement Analyzer). Rutting was measured from cylindrical samples compacted by a Superpave gyratory compactor at an air void content of 6.0 to 9.0 %. Rutting was measured after 100, 500, 1000 and 8000 cycles of loading with a pressure of 100 psi, at 60 C. The number of replicates at each experimental treatment was three. Note: in this research it was found later that air void content has little effect on rut depth as measured by APA Hierarchical cluster analysis. Hierarchical clustering is a multivariate statistical analysis technique that allows to identify significant different groups among observations that present different properties. Such technique considers canonical variables (combinations of variables) in order to explain why clusters are different. Therefore it allows to identify different sets of data, as well as to explain why such groups are different. It also gives some idea of how much correlated the measured properties are to each other. In this case, the difference between clusters is optimized when four of them are constructed. Table 5 not only describes which treatments get together to make every cluster, but also shows cluster properties (as averages and standard deviations). The following are the findings (aggregate dependant) from such analysis: Guapiles aggregate source: mixes with DGM1 and DGM2 gradations, and neat AC30 binder show high rutting. Mixes with DGM3 and neat AC30 binder, and mixes with neat PG70 (22) binder show low rutting. Mixes with SBS modified AC30 and SBS modified PG70 (22) show the lowest rutting. Santa Ana aggregate source: mixes with DGM1 and DGM2 gradations, and neat AC30 show highest rutting. Mixes with DGM1 and DGM2 with PG70 show high rutting. Mixes with DGM3 with both neat AC30 and neat PG 70 (22) show low rutting, with optimum asphalt content. Mixes with SBS modified AC30 and SBS modified PG70 (22) show the lowest rutting. Las Concavas aggregate source: same as Guapiles. According to available results, highest rutting means 2.8 mm after 1000 cycles of loading and 4.8 mm after 8000 cycles of loading; high rutting means 2.0 mm after 1000 cycles of loading and 3.6 mm after 8000 cycles of loading; low rutting means 1.3 mm after 1000 cycles of loading and 2.5 mm after 8000 cycles of loading; and lowest rutting means 0.8 mm after 1000 cycles of loading and 1.5 mm after 8000 cycles of loading. TRB 2006 Annual Meeting CD-ROM Paper revised from original submittal.

8 Castro-Fernandez, Sebaaly, Leiva-Villacorta, Carmona-Rojas, Aguiar-Moya, Allen-Monge, Badilla-Vargas, Cordero, Jimenez, 8 Therefore, rutting depends on asphalt binder stiffness, since neat AC30 gives more rutting than PG70 (22); as well as gradation, since DGM3 gradation gives less rutting; asphalt content, since asphalt contents over the optimum give higher rutting, depending on the aggregate source; and aggregate source. By using SBS modified binders, nevertheless, neat binder stiffness, gradation, asphalt content and aggregate source become unimportant. Best rutting resistance was achieved, at all cases, when SBS modified binders were employed. Table 6 allows to state the properties that better explain cluster differences. The two properties that better explain differences among clusters are APA deformation after 500 cycles and APA deformation after 1000 cycles, although APA deformation after 100 cycles and after 8000 cycles are also very significant. APA sample air voids are much less important, in regard to explaining variability among clusters, than measured APA deformations. Even though sample air voids is very important in regard to Canonical function 2, it can be noticed that Canonical function 2 is much less significant that Canonical function 1 (less percentage of variability explained by it), where sample air voids has very little loading. In Table 6 it can also be seen how air voids have little correlation with plastic deformation, based on the cluster analysis. This can be noticed in the fact of air voids not having high loadings in Canonical function 1, whereas rut depth at various numbers of cycles have high loadings Canonical correlation analysis. Canonical correlation is a multivariate statistical technique that allows to correlate sets of response and predictor variables. It lets to establish common effects among observations. It allows to notice which response and predictor variables are more significant in explaining differences among observations. Each canonical set consists of both response and predictor variables, which not only aid in explaining differences among observations, but also show a high correlation among them. The first canonical set explains the most variability among observations, the second canonical set explains less variability among observations than the first canonical but more than the third one, and so forth. This statistical technique is applied in order to see cause effect relationships that can aid in explaining differences among observations. Response variables are longer term rutting, as measured after 1000 cycles and after 8000 cycles. Predictor variables are shorter term rutting, as measured after 100 cycles and after 500 cycles, as well as sample air voids. Table 7 presents the first two canonical sets, which aid in explaining 100 % of common effects of variability among samples (considering each property has some correlation with other properties). There is a very strong relationship between short term APA deformations and long term APA deformations. APA sample air voids are not so important since its loading is always low (first and second Canonical sets). TRB 2006 Annual Meeting CD-ROM Paper revised from original submittal.

9 Castro-Fernandez, Sebaaly, Leiva-Villacorta, Carmona-Rojas, Aguiar-Moya, Allen-Monge, Badilla-Vargas, Cordero, Jimenez, 9 7. BEST PERFORMING TREATMENT RUTTING ANALYSIS. SBS modified mixes with DGM3, using aggregates from Guapiles and Las Concavas performed the best in regard to the full factorial experiment. Santa Ana aggregate samples performed just as well when SBS modified binders were considered. SMA mixes perform as well as dense graded mixes with neat binders. Taking into account not only APA deformation measurements, but also overall volumetric properties, the treatments shown in Table 8 were selected for further research. SMA mixes with neat binders were selected based on their improved aggregate interlock (stone to stone contact). DGM3 mixes were selected based on their improved rut resistance. Table 9 shows best performing treatment composition. Aggregates from both Guapiles and Santa Ana were considered, together with both SMA3 and DGM3 gradations. SMA samples were mixed with both AC30 and PG70 (22); whereas DGM samples were mixed with both AC % SBS and PG 70(22) % SBS. At this stage, two asphalt contents were considered, both optimum and optimum plus 0.50 % by total weight of mix. In addition to APA rutting, best performing treatment samples were tested by: a) dry resilent modulus at 25 C, b) indirect tensile strength at 25 C, c) APA rutting performed on conditioned samples. All tests were performed at cylindrical samples compacted by means of a Superpave gyratory compactor, targeting for an air void content of 6.0 to 9.0 %. Conditioned samples tested by means of the APA were saturated at 55 to 80 %, followed by a 24 h storage period in a water bath at 60 C. Hierarchical clustering was employed to identify significant different groups among observations. Three different clusters could be identified among available data. Table 10 not only describes each cluster composition, but also shows cluster properties (as averages and standard deviations). The best performing mixes based on hierarchical cluster analysis are those with SBS modified binders, regardless of their aggregate source and asphalt content. Mixes with neat binders show, nevertheless, different performance based upon aggregate source: Guapiles aggregate source: when SBS modified binder was used instead of neat binder, APA deformation after 1000 cycles decreased 0.7 mm, APA deformation after 8000 cycles decreased 1.0 mm, dry resilent modulus at 25 C increased by a factor of 2.3, indirect tensile strength at 25 C increased by a factor of 2.2, APA deformation on conditioned samples after 1000 cycles decreased 1.1 mm, and APA deformation on conditioned samples after 8000 cycles decreased 1.4 mm. TRB 2006 Annual Meeting CD-ROM Paper revised from original submittal.

10 Castro-Fernandez, Sebaaly, Leiva-Villacorta, Carmona-Rojas, Aguiar-Moya, Allen-Monge, Badilla-Vargas, Cordero, Jimenez, 10 Santa Ana aggregate source: when SBS modified binder was used instead of neat binder, APA deformation after 1000 cycles decreased 1.4 mm, APA deformation after 8000 cycles decreased 2.3 mm, dry resilent modulus at 25 C increased by a factor of 2.5, indirect tensile strength at 25 C increased by a factor of 2.8, APA deformation on conditioned samples after 1000 cycles decreased 0.2 mm, and APA deformation on conditioned samples after 8000 cycles decreased 0.9 mm. Therefore mixes with SBS modified binders do show less rutting, as well as a higher resilent modulus at 25 C and a higher indirect tensile strength at 25 C. When considering DGM3 and SMA3 gradations, there was no effect of asphalt content on rutting. Table 11 shows the canonical functions that allow cluster separation. It can be seen that all properties are equally significant to split data in two clusters. There is a tendency, nevertheless, for mixes with SBS binders to show less rutting and higher stiffness (based on resilent modulus and indirect tensile strength). 8. CONCLUSIONS It was found that aggregate source, gradation and asphalt content are of very little importance, when SBS modified binders are applied. It should be considered, nevertheless, that SMA gradations were not used together with SBS modified binders; since it was an objective to determine whether SMA gradations would be enough by themselves, without considering the benefits of binder modification. This papers shows the benefit of SBS binder modification; from this point on, research is to be focused on whether SMA gradations plus SBS binders perform even better, or just as well as dense gradations plus SBS binders. When neat binders are used, aggregate source, gradation and asphalt content are very important in regard to rutting. It was found that DGM3 gradation performs better than DGM1 and DGM2 gradations, for the aggregate sources considered in this research project. Mixes with neat PG70 (22) performed better than mixes with neat AC30 in regard to rutting. There was no evidence of a difference in rutting for SMA mixes and DGM mixes. DGM3 outranks all SMA gradations. Despite of having low percentages of voids in mineral aggregate, Santa Ana dense graded mixes performed well in regard to rutting, when SBS modified binders were employed. Santa Ana mixes with neat binders did not perform as well as Guapiles and Las Concavas mixes with neat binders, partly due to their low percentages of voids in mineral aggregate. APA tests on conditioned samples didn t correlate well to APA tests on dry samples. Such test is a combination of both a plastic deformation test and a moisture damage test. APA tests on dry samples do somewhat correlate to resilent modulus at 25 C and indirect tensile strength at 25 C. TRB 2006 Annual Meeting CD-ROM Paper revised from original submittal.

11 Castro-Fernandez, Sebaaly, Leiva-Villacorta, Carmona-Rojas, Aguiar-Moya, Allen-Monge, Badilla-Vargas, Cordero, Jimenez, REFERENCES (1) National Cooperative Highway Research Program; Accelerated laboratory rutting tests: Evaluation of the Asphalt Pavement Analyzer; Transportation Research Board; Report 508; (2) Roberts, Freddy and others; Hot mix asphalt materials, mixture design, and construction; NAPA; (3) National Asphalt Pavement Association; Designing and constructing SMA mixtures: state of the practice; NAPA; (4) Fernandez, George C.J.; University of Nevada - Reno. Accessed TRB 2006 Annual Meeting CD-ROM Paper revised from original submittal.

12 Castro-Fernandez, Sebaaly, Leiva-Villacorta, Carmona-Rojas, Aguiar-Moya, Allen-Monge, Badilla-Vargas, Cordero, Jimenez, 12 LIST OF TABLES TABLE 1 Treatment composition. TABLE 2 Asphalt binder properties. TABLE 3 Aggregate properties. TABLE 4 Volumetric mix design target properties. TABLE 5 All treatment cluster separation and cluster properties. TABLE 6 Canonical functions in all treatment cluster analysis. TABLE 7 Canonical functions in all treatment cluster canonical correlation. TABLE 8 Properties of best performing treatments. Optimum asphalt content data. TABLE 9 Best performing treatment composition. TABLE 10 Best performing treatment cluster separation and cluster properties. TABLE 11 Canonical functions in best performing treatment cluster analysis. LIST OF FIGURES FIGURE 1 Dense mix gradations (DGM). FIGURE 2 Stone matrix asphalt mix gradations (SMA). FIGURE 3 Superpave optimum asphalt content volumetrics. By aggregate source. FIGURE 4 Superpave optimum asphalt content volumetrics. By mix type. TRB 2006 Annual Meeting CD-ROM Paper revised from original submittal.

13 TABLE 1 Treatments composition. Aggregate source Guapiles (sedimentary from Eastern shore) Santa Ana (igneous from Central Valley) Las Concavas (igneous from Central Valley) Asphalt binder Gradation type Dense graded (DGM) Stone matrix asphalt (SMA) DGM1 DGM2 DGM3 SMA1 SMA2 SMA3 AC PG70 (22) AC30 + SBS PG70 (22) + SBS AC PG70 (22) AC30 + SBS PG70 (22) + SBS AC PG70 (22) AC30 + SBS PG70 (22) + SBS Note: the full factorial experiment also included two asphalt binder contents, optimum and optimum plus 0.50 % by total weight of mix (highest allowable plant asphalt content). 13

14 TABLE 2 Asphalt binder properties. Property type Physical properties Aging susceptibility Thermal susceptibility Performance grade Property Asphalt type AC30 PG70 (22) AC % SBS PG70 (22) % SBS Density at 25 C (g/cm3) Penetration at 25 C (1/100 mm) Absolute viscosity at 60 C (Poise) Absolute viscosities ratio at 60 C (RTFO / original) RTFO mass loss (%) Penetration index VTS index Original condition maximum performance temp ( C) RTFO condition maximum performance temp ( C) PAV condition intermediate performance temp ( C)

15 TABLE 3 Aggregate properties. Aggregate source Gradation Bulk specific gravity Guapiles (sedimentary from Eastern shore) Santa Ana (igneous from Central Valley) Las Concavas (igneous from Central Valley) Absorption (%) Fine aggregate Coarse aggregate (*) Uncompacted void content (%) (**) DGM DGM DGM SMA SMA SMA DGM DGM DGM SMA SMA SMA DGM DGM DGM SMA SMA SMA (*) At least 95 % of coarse aggregate has 2 or more fractured faces. Sand equivalent (%) L.A. abrasion (%) (**) Determined by testing each gradation as is, so that the effect of gradation on uncompacted air void content can be shown. (***) Ratio of 3:1. Flat & elongated particles (%) (***) 15

16 TABLE 4 Volumetric mix design target properties. Properties Dense graded mixes (*) Stone matrix asphalt mixes (**) Asphalt content (% TWM) N.A. Maximum 6.0 Air voids (%) 3.0 to 5.0 (as close as possible to 4.0) 3.0 to 5.0 (as close as possible to 4.0) Voids in mineral aggregate (% VMA) Higher than 14.0 Between 16.0 and 18.0 Voids filled with asphalt (% VFA) % G N ini Smaller than 89 Smaller than 89 % G N max Smaller than 98 Smaller than 98 (*) According to Superpave mix design specifications. (**) Additional requirements: a) voids in coarse aggregate smaller than calculated voids in rodded volume of coarse aggregate, b) drain down at optimum asphalt content not higher than 0.30 %. 16

17 TABLE 5 All treatment cluster separation and cluster properties. Cluster (*) Mix properties Air voids (%) Mean Stand. dev. APA deformation at 60 C after 100 cycles (mm) Mean Stand. dev. APA deformation at 60 C after 1000 cycles (mm) Mean Stand. dev. APA deformation at 60 C after 8000 cycles (mm) Mean Stand. dev. Description of treatments within cluster Santa Ana mixes with AC30 binder Guapiles and Las Concavas (except DGM3) mixes with AC30. Santa Ana mixes with PG70 binder Guapiles and Las Concavas with PG70. Guapiles and Las Concavas (DGM3) with AC30. Santa Ana DGM3 with both neat binders and optimum asphalt content Mixes with either AC30 + SBS or PG70 + SBS. (*) Hierarchical cluster separation by Ward s method. Cluster separation highly significant (confidence level over %). 17

18 TABLE 6 Canonical functions in all treatment cluster analysis (*). Canonical functions 1 Higher loadings (**) Variable Loading APA deformation after 100 cycles APA deformation after 500 cycles APA deformation after 1000 cycles APA deformation after 8000 cycles APA sample air voids (*) Cluster separation is highly significant. (**)Loading refers to the correlation between a common effect (canonical function) and an actual variable. Cumulative variance between clusters explained by considered variables, which is reproduced by canonical functions. As a percentage

19 TABLE 7 Canonical functions in all treatment canonical correlation. Canonical sets 1 2 Higher loadings (*) Predictor variables Response variables Variable Loading Variable Loading APA deformation after 100 cycles APA deformation after 500 cycles APA samples air voids APA deformation after 100 cycles 0.91 APA deformation after 1000 cycles 1.00 APA deformation after 8000 cycles 0.32 APA deformation after 1000 cycles APA deformation after 8000 cycles (*) Loading refers to the correlation between a common effect (canonical function) and an actual variable Variance among observations explained by response variables, which is reproduced by predictor canonical functions. As a percentage. (**) (**) Common effects on predictor variables correlated to common effects on response variables. Consider every set has a response canonical function, as well as a predictor canonical function

20 TABLE 8 Properties of best performing treatments. Optimum asphalt content data. Aggregate source Gradation Asphalt type Asphalt content (% by total weight of mix) Guapiles (sedimentary from Eastern shore) Santa Ana (igneous from Central Valley) VMA (%) VFA (%) APA deformation after 1000 cycles at 60 C (mm) SMA2 AC SMA3 PG70 (22) DGM3 AC30 + SBS DGM3 PG70 (22) + SBS SMA3 AC SMA3 PG70 (22) DGM3 AC30 + SBS DGM3 PG70 (22) + SBS APA deformation after 8000 cycles at 60 C (mm)

21 TABLA 9 Best performing treatments composition. Aggregate source Guapiles (sedimentary from Eastern shore) Santa Ana (igneous from Central Valley) Asphalt binder Gradation type Dense graded (DGM): DGM3 Optimum asphalt content Optimum asphalt content plus 0.50 % by total weigth of mix Stone matrix asphalt (SMA): SMA3 Optimum asphalt content 21 Optimum asphalt content plus 0.50 % by total weigth of mix AC (*) 4 (*) PG70 (22) AC30 + SBS PG70 (22) + SBS AC PG70 (22) AC30 + SBS PG70 (22) + SBS (*) Gradation SMA2 was selected instead of SMA3, due to mix workability.

22 TABLE 10 Best performing treatment cluster separation and cluster properties. Cluster (*) Mix properties APA plastic deformation after 1000 cycles at 60 C (mm) APA plastic deformation after 8000 cycles at 60 C (mm) Dry resilent modulus at 25 C (kpa) Description of treatments within cluster Mean Stand. Mean Stand. Mean Stand. dev. dev. dev Mixes with modified binders, both AC30 and PG70 binders Mixes with Santa Ana aggregates, with neat binders Mixes with Guapiles aggregates, with neat binders. (*) Hierarchical cluster separation by Ward s method. Cluster separation highly significant (confidence level over %). 22

23 TABLE 10 (cont d) Best performing treatment cluster separation and cluster properties. Cluster (*) Mix properties Indirect tensile strength at 25 C (MPa) Wet APA plastic deformation after 1000 cycles at 60 C (mm) Wet APA plastic deformation after 8000 cycles at 60 C (mm) Description of treatments within cluster Mean Stand. Mean Stand. Mean Stand. dev. dev. dev Mixes with modified binders, both AC30 and PG70 binders Mixes with Santa Ana aggregates, with neat binders Mixes with Guapiles aggregates, with neat binders. (*) Hierarchical cluster separation by Ward s method. Cluster separation highly significant (confidence level over %). 23

24 TABLE 11 Canonical functions in best performing treatment cluster analysis (*). Canonical functions 1 2 Higher loadings (**) Variable (*) Cluster separation is highly significant. Loading APA deformation after 100 cycles APA deformation after 500 cycles APA deformation after 1000 cycles APA deformation after 8000 cycles Dry resilent modulus at 25 C Wet resilent modulus at 25 C Dry Indirect tensile strength at 25 C Wet indirect tensile strength at 25 C Dry resilent modulus at 25 C Wet resilent modulus at 25 C Dry Indirect tensile strength at 25 C Wet indirect tensile strength at 25 C Wet APA deformation after 100 cycles Wet APA deformation after 500 cycles Wet APA deformation after 1000 cycles Wet APA deformation after 8000 cycles Cumulative variance between clusters explained by considered variables, which is reproduced by canonical functions. As a percentage % % (**)Loading refers to the correlation between a common effect (canonical function) and an actual variable. 24

25 % Passing No.200 No. 50 No. 30 No. 16 No. 8 No mm 12.5 mm 19.0 mm No. 100 Size (mm) 0.45 FIGURE 1 Dense mix gradations (DGM) DGM3 DGM1 DGM2 25

26 % Passing No.200 No. 50 No. 30 No. 16 No. 8 No mm 12.5 mm 19.0 mm No. 100 Size (mm) 0.45 FIGURE 2 Stone matrix asphalt gradations (SMA) SMA3 SMA1 SMA2 26

27 % V a VMA % V be FIGURE 3 Superpave optimum asphalt content volumetrics. By aggregate source. 60 VFA Guapiles SMA Santa Ana SMA Las Concavas SMA Guapiles DGM Santa Ana DGM Las Concavas DGM

28 a % V VMA % V be FIGURE 4 Superpave optimum asphalt content volumetrics. By mix type. 60 VFA SMA AC30 SMA PG70 DGM AC30 DGM AC30 SBS DGM PG70 DGM PG70 SBS 28