NATIONAL COOPERATIVE HIGHWAY RESEARCH PROGRAM

NCHRP Report XXX

For Project 9-45

Test Methods and Specification Criteria for

Mineral Filler Used in HMA

Revised Draft Final Report 12/28/2010

From

University of Wisconsin-Madison

ii

Contents 1 Introduction and Research Approach ....................................................................... 1

1.1 Statement of Problem ........................................................................................ 1 1.2 Objectives ......................................................................................................... 1 1.3 Descriptions of Project Tasks ............................................................................ 1 1.4 Objectives and Experimental Plans of the Project Phases .................................. 2

1.4.1 Phase I ....................................................................................................... 3 1.4.2 Phase II ...................................................................................................... 4

2 Findings ................................................................................................................ 12 2.1 Types of Fillers Tested in this Study ............................................................... 12 2.2 Summary of Findings from the State Agencies’ Survey ................................... 13

2.2.1 Statistics of the Results from Questionnaire ............................................. 14 2.2.2 Summary of Survey Results ..................................................................... 16

2.3 Fillers Characteristics ...................................................................................... 17 2.3.1 Filler Geometry Tests and Results ............................................................ 19 2.3.2 Filler Composition Tests and Results ....................................................... 27

2.4 Summary of Fillers’ Properties ........................................................................ 33 2.5 Effect of Fillers on Mastic and Mixture Performance ...................................... 35

2.5.1 Effect of Fillers on Mastic and Mixture Workability Indicators ................ 39 2.5.2 Effect of Fillers on Mastic and Mixture Rutting Resistance ...................... 45 2.5.3 Effect of Fillers on Mastic and Mixture Fatigue Resistance ...................... 52 2.5.4 Effects of Fillers on Mastic and Mixture Low Temperature Performance . 60 2.5.5 Effects of Fillers on Mastic and Mixture Moisture Damage ...................... 70

3 Interpretation, Appraisal, and application .............................................................. 76 3.1 Quantifying of Filler Influence on Mixture Performance Indicators ................. 76

3.1.1 Strategy of Analysis ................................................................................. 76 3.1.2 Workability .............................................................................................. 78 3.1.3 Permanent Deformation ........................................................................... 81 3.1.4 Fatigue ..................................................................................................... 86 3.1.5 Low Temperature Cracking Resistance .................................................... 88 3.1.6 Moisture Damage Resistance ................................................................... 91

3.2 Assessment of Filler Test Methods .................................................................. 93 3.2.1 Fillers Selected for Sensitivity Analysis ................................................... 93 3.2.2 Results from the Sensitive Testing ........................................................... 96 3.2.3 Qualitative Assessment of Filler Testing Procedures ................................ 99

4 Conclusions and Suggested Future Research ....................................................... 101 4.1 General Conclusions ..................................................................................... 101 4.2 Specific Conclusions by Performance Indicators ........................................... 102 4.3 Recommend specification criteria for mineral fillers that optimize HMA performance. ........................................................................................................... 104 4.4 Suggested Future Research............................................................................ 105

5 References ........................................................................................................... 106

1

1 INTRODUCTION AND RESEARCH APPROACH This section of the report presents the statement of the problem, objectives, and descriptions of the phases and tasks of the research project.

1.1 Statement of Problem This project was initiated to determine the effect of mineral fillers on asphalt mixture performance with respect to common pavement distresses. To quantify the effect of fillers on mixture performance, critical filler properties which have significant influence on the mixture behavior must be identified. Based on these significant filler properties, recommendations of standard testing procedures to measure these critical filler properties and a means to incorporate these properties to enhance mix design shall be made.

1.2 Objectives The objectives of this study were to: 1. Identify and/or develop test methods for mineral fillers that characterize their mechanical and chemical effects on the performance of mastics (combinations of asphalt binder and mineral filler) and hot-mix asphalt (HMA) 2. Recommend specification criteria for mineral fillers that optimize HMA performance In this project, mineral filler is defined as any material meeting the criteria of AASHTO M 17 (ASTM D 242). The research investigated the behavior of mineral matter meeting these criteria as (1) a filler, an extender, or both, and (2) an inert or a reactive component in mastic and HMA. The test methods and related criteria are applicable to HMA, (including dense-graded HMA, stone matrix asphalt [SMA], and open-graded friction course [OGFC]), and suitable for use in both mix design and quality control.

1.3 Descriptions of Project Tasks The following seven tasks were undertaken to accomplish the project objectives:

Task 1 . Conduct a thorough review of the worldwide literature to identify (1) current practices on the specification and use of mineral filler; (2) the nature and characterization of the mechanical and chemical properties of mineral fillers as they relate to the performance of mastics and HMA; (3) test methods currently used to specify mineral fillers and measure their effect on the performance of mastics and HMA; and (4) effects of mineral filler on HMA pavement performance (including rutting, fatigue cracking (bottom-up and top-down), low-temperature cracking, moisture damage, and aging) and constructability.

2

Task 2 . Prepare an updated, detailed work plan, budget, and schedule for a laboratory evaluation program to: (1) Select one or more test methods from among the promising candidates identified in Task 1 based upon their demonstrated ability to (a) characterize the mechanical and chemical properties of mineral fillers and (b) measure the effect of mineral filler on HMA pavement performance and constructability; (2) Evaluate the selected test methods with a broad range of mineral fillers, including those commonly used across the United States, to determine the applicability and reliability of the methods; and (3) Validate and quantify the effectiveness of the selected test methods with respect to their potential for (a) predicting or measuring HMA pavement performance and constructability and (b) their use in quality control. Task 3. Submit an interim report of the findings of Tasks 1 and 2, including an annotated bibliography of the literature review. Task 4. Execution of the work plan approved in Task 3 and, based on the results of the laboratory evaluation, recommend one or more test methods with a demonstrated ability to (a) characterize the mechanical and chemical properties of mineral fillers and (b) measure the effect of mineral filler on HMA pavement performance and constructability. Task 5. Development of preliminary specification criteria for the test methods recommended in Task 4. Task 6. Assess the sensitivity of the recommended test methods and specification criteria for application in normal construction operations, including quality control. Task 7. Final report preparation that documents results, summarizes findings, draws conclusions, including (1) recommended test methods in draft AASHTO format and (2) recommended specification criteria and associated limits for each criterion. In the report, discuss benefits of and obstacles to implementing the recommended test methods and specification criteria for state departments of transportation and the HMA industry.

1.4 Objectives and Experimental Plans of the Project Phases The project was divided into two phases. For each project phase, a set of objectives was defined and a work plan was developed.

3

1.4.1 Phase I Two main objectives were defined for Phase I of the study:

1. Summarize the current knowledge for characterizing mineral fillers and

measuring their effects on mastics and HMA, and 2. Develop an experimental plan to include a select set of fillers to be tested

including test methods for fillers, mastics, and HMA that capture important effects of fillers on HMA performance.

To best accomplish these objectives the research team conducted the following main activities:

• Search of 10 well recognized literature data bases using 28 combinations of selected keywords.

• Review of abstracts of articles identified by the search and selection of a subset of relevant articles. More than 600 abstracts were identified out of which 228 articles were selected for detailed review and summarization.

• Survey of State Highway Agencies by e-mail using a questionnaire covering the most critical aspects of use and specifications of fillers.

• Two meetings with experts in Europe. Additionally, a detailed review of the European standards (Norms) related to aggregates and fillers was conducted.

The results of the survey and literature review were summarized in a letter report submitted to the NCHRP project panel. The main findings led to selection of a set of physical tests to characterize geometry of fillers, and chemical tests to characterize filler composition. These tests, listed in Table 1-1, were included in the experimental testing plan of the project. The review of European Norms and the interviews of a number of experts indicate a well developed framework in the norms for characterizing the essential properties of fillers. The test procedures are found to be similar to those used in the American Standards and studies but are unique in the sense that they require measuring specific properties of mastics, such as stiffening effects and moisture susceptibility. The European Norms also include criteria limits used to classify fillers into grades based on their stiffening effects and their content of calcium compounds. These findings were useful in the development of filler testing systems and acceptance criteria in this project.

4

Table 1-1 Tests Selected to Measure Critical Characteristics of Fillers

The activities of this phase also resulted in identification of the test methods and equipment that can be used for effective measurement of the effects of fillers on mastic and HMA performance in the laboratory. The selected tests are discussed in the following sections.

1.4.2 Phase II The results of Phase I were used to develop a testing plan for Task 4 of Phase II. The plan was designed as a three-part study as shown in Figure 1-1. In the first part, a large number of natural and imported fillers were collected based on mineralogy and common use. Table 1-1 lists the types of fillers selected for the study. Sixteen types of fillers were collected. For each type, two suppliers/sources were contacted. Thus, the total number of fillers considered in this project was 32 fillers. These fillers were tested to measure eight filler characteristics using the filler tests listed in Table 1-2. In part two of the testing plan, the focus was on testing mastics produced by mixing a subset of 16 fillers with selected asphalt binders. The 16 fillers selected represent a full factorial design of 24 with a “high” and a ”low” value for each of the four main filler characteristics, two chemical and two physical. Based on recommendation of the project panel, a 17th filler was added for inclusion in the mastic testing matrix, due to its high use in the field. These fillers were mixed with four binders to produce a total of 68 mastics. The 68 mastics were tested in accordance with the tests listed in Table 1-4.

1. Physical Property Test Protocol Remarks • Fractional voids Rigden Voids

EN 1097-4 No development is necessary.

• Size Distribution Laser Diffraction ASTM D4464

Type of surfactant and time of agitation were studied.

• Specific Gravity Helium Pycnometer No development is necessary. • Absorption Bitumen Number

(EN13179-2) & Laser Diffraction with Time

Not a standard method and some development is needed.

• Shape and Texture Microscopy using AIMS or UIAIA software

Not a standard method and some development is needed.

2. Chemical Property Test Protocol Remarks • Calcium Compounds X-ray Fluorescence ,

EN 196-21, EN 495-2 ASTM D3042 MN/DOT

Initial testing is needed to determine the appropriate test method.

• Water Solubility EN 1744-1:1998 No development is necessary. • Methylene Blue /

Plasticity Index AASHTO TP57 AASHTO T90

Limited development for TP 57 is needed.

• Organic Content- Loss on Ignition

EN 1744-1:1998 C17, AASHTO 267-86

Need to determine testing temperature.

5

The third part of the testing plan included testing HMA samples. The experimental design for HMA testing was based on mixing a selected set of 16 mastics with two aggregate gradations. The HMA tests are listed in Table 1-5. Based on the analysis of the mastics and HMA results, the filler properties that show a significant effect on HMA performance were identified and selected for further study in Tasks 5 and 6. The flow chart shown in Figure 1-1 describes the overall approach.

Figure 1-1 Laboratory evaluation plan for Task 4

1.4.2.1 Task 4- Part I – Collecting and testing of fillers Since natural fillers were used in the study, it was not possible to pre-define the characteristics of the fillers to be tested. The mineralogy and the source were used to target fillers that are expected to encompass a wide range in compositional and geometric characteristics. Table 1-2 includes the combinations desired for mineralogy and important physical properties or nature of the source. The list is based on the literature review, the results of surveys from SHA ’s, the research team experience, and contacts with aggregate suppliers. For each filler type, a minimum of two sources were targeted so that the filler library would include at least 32 different types of fillers. The testing of the 32 fillers included filler characteristics that were identified in Task-1: four physical and four chemical tests, as listed in Table 1-1.

1. Collected 32 fillers from 16 sources (Table 4).

2. Test the 32 fillers using filler tests listed in Table 1. Select 16 fillers to cover High and Low values for each of 8 filler characteristics.

3. Mix the 16 selected fillers with four binders and test the resulting 64 mastics using mastic tests listed in Table 2. Analyze results and select 16 mastics.

4. Mix 16 mastics with two aggregate gradations and test resulting HMA using HMA tests listed in Table 2.

5. Based on analysis of mastic and HMA testing, identify filler characteristics that have significant effect on HMA workability and performance.

PART

I

PART 2

PART 3

6

Table 1-2 Types of Fillers Collected for Use in the Experimental Testing Plan Filler Number Mineralogy / Source Nature of Source

1. Limestone Hard 2. Limestone Soft 3. Dolomite Hard 4. Dolomite Soft 5. Granite Hard 6 Granite Soft 7. Basalt Hard 8. Gravel Sandstone 9. Gravel Quartzite

10. Gravel Chert 11. Bottom Ash 12. Fly Ash Type C 13. Fly Ash Type F 14. Fly Ash Non-Spec 15. Lime 16. Cement Type I and II

Based on the testing results of the 32 fillers, 16 fillers were selected for the mastic testing plan. The selection process was based on sorting fillers into groups of “High” and “Low” values for each of the four main characteristics. Table 1-3 lists the four main variables selected and shows the (24) design of experiment, in addition to the added filler. Thresholds for defining “High” and “Low” levels were based on the average and distribution of values for each variable. Because of this uncertainty, partial factorial experimental design was not considered.

Table 1-3 Experimental Matrix for Mastic Testing

Combination Number

Rigden V oids (2)

Size Distribution

(2)

Reactivity Calcium Comp.

(2)

Harmful fines Clay Content

(2) 1 H H H H 2 L H H H 3 H L H H 4 L L H H 5 H H L H 6 L H L H 7 H L L H 8 L L L H 9 H H H L

10 L H H L 11 H L H L 12 L L H L 13 H H L L 14 L H L L 15 H L L L 16 L L L L

7

Table 1-3 represents the initial target. However, during the course of the selection process, it was not possible to allocate each combination number to a filler. Therefore, the research team used engineering judgment to assign fillers to a combination number (1 through 16) provided that the filler would meet at least three of the four target (H,L) levels. The research team selected a total of 17 fillers instead of 16 to cover some of the gaps encountered with respect of the levels desired. The fillers selected and their specific properties are discussed in Chapter 2.

1.4.2.2 Task 4- PART II- Testing of Mastics Each of 17 selected fillers was mixed with four binders at one filler to binder (F/B) ratio, which was kept at the mass ratio of 1.0. The binders used included two PG 64 grades from two widely used sources, known to have different chemical compositions. One of the PG 64 grades was modified with Styrene Butadiene Styrene and with Poly-phosphoric Acid to a PG 70 grade. The following binders were used for the experimental plan:

a. PG 64-22 with high asphaltenes (from a light crude source-Flint Hills) b. PG 64-22 with low asphaltenes (from a heavy crude source-Valero) c. Binder (a) modified with PPA to a PG 70-22 d. Binder (a) modified with SBS to a PG 70-22

The 68 mastics (combinations of 17 fillers and four asphalts) were tested for mastic characteristic properties. Table 1-4 lists the mastic characteristics measured and the associated test methods. Mastic testing results were statistically analyzed to identify which of the filler variables have a statistically significant effect on the mastic characteristics, and the levels of filler characteristics that need to be included in the HMA testing. Also, the effect of binder properties was investigated such that the number of fillers and binder combinations that merit further evaluation were identified.

8

Table 1-4. Mastic Testing Program Conducted

Mastic Characteristic Response Variable Test Method Aging

1. Constructability Viscosity Rotational Viscosity (RV) Un-aged

2. Rutting Resistance

Accumulated Strain Dynamic Shear Rheometer (DSR)/ Multiple Stress Creep and Recovery (MSCR) 25 mm PP

Un-aged Non Recoverable

Compliance

3. Fatigue Resistance Fatigue Life Time Sweep

Un-aged and PAV only G*·sinδ DSR

4. Thermal Cracking Resistance S and m Bending Beam

Rheometer (BBR) PAV Aged only

5. Moisture Damage Resistance

Water Sensitivity EN 1744-4 Un-aged

Bond Strength PATTI / or DSR

1.4.2.3 Task 4-PART III- Mixture Testing Plan The analysis of mastic testing results conducted in Part II was used to select 16 mastics for further evaluation in HMA. The selection process focused on including mastics that show significantly different behavior in each of the five mastic characteristics. The mastics results for each characteristic were divided into four quartiles. Four mastics from each quartile were selected to ensure an acceptable level of redundancy in the mastic performance in each quartile. In addition, the research team made sure that the selected mastics were equally distributed between the four binders used. The mastics selected are discussed in Chapter 2. The experimental plan of HMA testing is shown in Table 1-5. Based on mixture and mastic test results, recommendations of tests to measure critical filler properties and a means to determine acceptable fillers to enhance mixture design were presented.

9

Table 1-5 Mixture Testing Program Conducted

Mixture Characteristic Measured Response Test Method

Constructability Gyrations to 92%Gmm

Superpave Gyratory

Compactor

Permanent Deformation Dynamic Modulus

(E*)

Asphalt Mixture Performance Test (AMPT)

Flow Number (FN) AMPT

Fatigue Resistance Cycles to 45% Drop in E*

Indirect Tension Test (IDT)

Thermal Cracking Fracture IDT

Moisture Damage Rut Depth Hamburg Wheel Test

In order to identify filler characteristics that have critical effects on HMA performance, an analysis approach was developed to first correlate mixture performance with mastic performance. If logical trends and correlations were found, mastic to filler correlations were explored to identify critical filler properties. The details of the approach are discussed in Chapter 3. The analysis allowed for identifying a set of filler properties that have significant effects on HMA performance, and for defining the trends in HMA properties as a function of these properties. Task 5 focused on establishing the criteria for acceptance of fillers.

1.4.2.4 Task 5- Develop preliminary specification criteria Based on the results of Task 4, it was clear that filler properties that have important effects on mixture and mastic performance indicators include the fractional voids as measured by the Rigden voids (RV), and to a less extent Calcium Oxide content (CaO) as measured by the X-ray fluorescence. In addition the specific gravity as measured by the Helium pycnometer is needed for the RV calculation procedure. This task focused on defining the specific trends between these filler properties and mastic or mixture properties. The strategy of analysis was based on correlating mixture properties with mastic properties first, and if clear trends could be defined, limits on mastic properties were defined based on the acceptance of mixture responses. Due to the high interactive effects between binders and fillers, no limits on filler properties could be defined, but models were developed to estimate mastic properties from filler and binder properties. Mixture responses including workability in terms of gyrations to a specific density, flow number at high pavement temperature, fatigue life as defined by number of cycles to 45% initial E*, moisture resistance as defined by the Tensile Strength Ratio and Hamburg wheel rut depth, and low temperature stiffness and strength. Although analysis of trends was conducted for all mixture properties, clear trends could be defined only for workability, rutting resistance, and low temperature strength of mixtures. Mixture

10

acceptance limits were either derived based on average performance of mixtures tested in this study, or established based on existing published literature. These limits were used to define mastic performance acceptance criteria.

1.4.2.5 Task 6- Assess the sensitivity of the recommended test methods and specification criteria for application in normal construction operations, including quality control

To assess the sensitivity of the recommended methods and specification criteria, the research team developed a plan to use filler testing equipment for each of the four properties at six laboratories. At each laboratory, two operators tested two replicates of six selected fillers. A document with details of testing protocol selected and method of reporting data was generated and sent with the fillers. To ensure consistency of equipment, some the devices were shipped to the various laboratories. The selection of the fillers for Task 6 was based on the distribution of the Rigden voids values of the fillers tested and the mineralogical types. The fillers selected included five natural fillers and one manufactured filler. The availability and the commonality of the fillers played a role in the final selection. Table 1-6 shows the list of fillers selected for Task 6, the mineralogical type, and the corresponding measured properties. The ranges of values for each selected filler property for all fillers tested is also included.

Table 1-6 List of Selected Fillers and the Corresponding Initial Values Measured in Task 4

Code Filler Type Rigden Voids (%)

Specific Gravity FM CaO (%)

LS2 Soft Limestone 35.40 2.62 3.68 46.30 DS2 Soft Dolomite 29.40 2.70 4.73 27.00 GH1 Hard Granite 42.60 2.66 4.06 3.50 GS2 Soft Granite 47.00 2.40 3.13 7.80 BH2 Hard Basalt 33.80 2.77 4.65 7.00 FS1 Steel Furnace Slag 40.00 2.80 4.87 50.30

Maximum values in the study 49.10 2.89 6.32 50.30 Minimum values in the study 26.20 2.14 2.98 0.95

Task 6 was conducted considering the following two points:

1. Filler tests could be included as part of aggregate evaluation in mixture design. It is expected that a test(s) for checking filler quality will be identified to evaluate the imported filler, if used, mixed with the dust collected from washing of the aggregates. The results of the tests could be included in the mix design report and be required during quality control. It is possible that bag house collectors could be monitored by contractors who could keep records on weekly or monthly basis, or as aggregate sources are changed.

11

2. To collect specific feedback on the test methods, a questionnaire was used to rate the ease of use of the equipment, practicality, and to determine if extensive training for operators is necessary in selected laboratories.

The laboratories included six labs that were willing to participate at no cost to project. Department of Transportation (DOT) labs and contractors were contacted to encourage participation. Additionally, operators from the research team participated. Participants received brief training on equipment by phone or in person and were supplied with split filler samples for testing. This final report represents the work completed for Task 7. The following sections include summary of the results and findings from the work conducted in Tasks 4, 5, and 6 of the project. The relationships between mixture performance properties and mastic / filler properties are described. Also, recommendations for test methods to measure important filler properties and proposed limits for specifications are included.

12

2 FINDINGS

2.1 Types of Fillers Tested in this Study Mineral fillers for asphalt applications are defined by the AASHTO M17 (ASTM D2424) as finely divided minerals such as rock dust, hydrated lime, hydraulic cement, fly ash loess, or other suitable mineral matter (1). In this study, for the purpose of selecting representative samples of materials used in the United States, fillers were broadly classified as natural and imported fillers. Natural fillers are defined as produced directly from the crushing of rocks used as source for aggregates. They also include fine particles collected in the dust collecting systems in the HMA production facilities. Imported fillers, also called manufactured fillers, include hydrated lime, cement, fly ashes, and other products that are not directly produced from crushing of rocks used as aggregates in HMA applications. Multiple aggregate and manufactured filler producers were contacted from a wide range of states to collect diverse sources and mineralogies. Producers who supplied the project with fillers were asked to specify the general mineralogy of the source. Hence, testing for filler mineralogy was not included as part of this study. Rather, the focus was characterizing the physical and chemical properties as detailed in Chapter 1. These characteristics were used to identify relationships with mastic and mixture performance properties in order to characterize the influence of the fillers on both asphalt mastics and mixtures. Collected fillers were sorted by mineralogy and hardness of source rocks. The Los Angeles (LA) abrasion of the fillers' rock source was used to define hardness. Hard fillers are derived from rocks with LA abrasion values equal to or less than 25.0%. Conversely, soft fillers are produced from rocks with LA abrasion values greater than 25.0%. Table 2-1 provides a summary of the 32 fillers included in this study and the state where the source quarry is located. The table also shows the identification code of the filler that will be used for fillers in this report.

13

Table 2-1 Summary of Fillers Collected for Testing in this Study.

No. Code Filler Type Filler Source (State)

1 LH1 Hard Limestone Michigan 2 LS1 Soft Limestone Iowa 3 LS2 Soft Limestone Nebraska 4 DH1 Hard Dolomite Wisconsin 5 DH2 Hard Dolomite Michigan 6 DS1 Soft Dolomite Wisconsin 7 DS2 Soft Dolomite Wisconsin 8 GH1 Hard Granite Wisconsin 9 GH2 Hard Granite California 10 GHB1 Hard Granite Wisconsin 11 GS1 Soft Granite California 12 GS2 Soft Granite Georgia 13 BH1 Hard Basalt Wisconsin 14 BH2 Hard Basalt Wisconsin 15 BV1 Vesicular Basalt Arizona 16 GRQ1 Siliceous Gravel Quartzite Ohio 17 GRQ2 Siliceous Gravel Quartzite Wisconsin 18 FAC1 Fly Ash Type C Minnesota 19 FAC2 Fly Ash Type C Nebraska 20 FAF1 Fly Ash Type F Ohio 21 FAN1 Fly Ash Non Spec Minnesota 22 FAN2 Fly Ash Non Spec Maryland 23 HL1 Hydrated Lime Wisconsin 24 HL2 Hydrated Lime Wisconsin 25 CM1 Cement Wisconsin 26 FS1 Steel Furnace Slag Indiana 27 FS2 Blast Furnace Slag Indiana 28 CBC1 Carbon Black Coarse Massachusetts 29 CBF1 Carbon Black Fine Massachusetts 30 CA1 Hard Caliches New Mexico 31 CA2 Soft Caliches New Mexico 32 AN1 Andesite Nevada

2.2 Summary of Findings from the State Agencies’ Survey The research team distributed questionnaires to collect information from state and provincial highway agencies. In total, 43 written responses were received, including 40 state highway agencies (including District of Columbia), the Saskatchewan province in Canada, FHW A, and Port Authorities of New York and New Jersey. For the 11 state highway agencies that did not respond, the research team reviewed the standard specifications for construction posted on the agencies’ web sites to answer as many questionnaire questions as possible.

14

2.2.1 Statistics of the Results from Questionnaire The goal of the survey was to collect feedback on the effect of fillers on HMA performance. The survey covered specifications for dense graded HMA, Stone Mastic Asphalt (SMA), and open-graded / porous HMA with modified and unmodified binders. In addition, the survey included a section related to Slurry Seal. The survey indicated that two types of fillers were recognized in this project:

(1) Naturally occurring in the aggregates (such as bag-house dust).

(2) Added fillers that contractors import from other sources (such as fly ash, hydrated lime, rock dust), also called “Imported Fillers”.

The following nine questions were listed in the questionnaire. The number of occurrences per answer relative to the total responses is shown in parentheses.

1. Does your agency have specifications / guidelines for natural mineral fillers?

Y es (4 of 54) No (50 of 54)

2. Does your agency/company have specifications / guidelines for imported mineral fillers

in any type of HMA? Y es (44 of 54) No (10 of 54)

If yes, what are the gradation limits you specify for imported fillers?

___ASTM D242 or AASHTO M17 (23 of 44), _Miscellaneous (21 of 44)_

3. Other than gradation, are there other characteristics of mineral filler that you believe have an important affect on HMA performance?

Y es (27 of 39) No (12 of 39)

(1) Plasticity (12 of 27) (2) Stripping of Mastics or HMA (5 of 27) (3) Loss on ignition (4 of 27) (4) Mastics Property (2 of 27) (5) Bar Linear Shrinkage (2 of 27)

(6) Clay content (2 of 27) (7) PH value (1 of 27) (8) Rigden void (1 of 27) 4. Does your agency/company use imported filler?

Y es (37 of 49) No (12 of 49)

(1) Lime (22 of 37) (2) Stone fines (18 of 37) (3) Fly ash (15 of 37) (4) Cement (13 of 37) (5) Silt (1 of 37) (6) V olcanic ash (1 of 37)

(7) Processed chat sludge (1 of 37) (8) Lime kiln dust (1 of 37) (9) Cement kiln dust (1 of 37) (10) Incinerator ash (1 of 37) (11) RAP (1 of 37)

(12) Slag (1 of 37)

15

5. What are the primary sources (aggregates) from which natural fillers (P200) are derived?

(1) Limestone (21 of 43) (2) Gravel (12 of 43) (3) Baghouse (7 of 43) (4) Granite (6 of 43) (5) Sandstone (5 of 43) (6) Dolomite (4 of 43) (7) Quartzite (4 of 43) (8) Quarry deposit (4 of 43) (9) Blast furnace slag (2 of 43) (10) Marble (1 of 43) (11) Reclaimed PCC (1 of 43) (12) Trap rock (1 of 43) (13) Gneiss (1 of 43) (14) Igneous (1 of 43) (15) Syenite (1 of 43)

6. Are there any distinct fillers that you suggest for consideration in this project?

Y es (8 of 42) No (34 of 42)

(1) Fly ash (5 of 8) (2) Lime (3 of 8) (3) Limestone (2 of 8) (4) Baghouse (2 of 8) (5) Quartzite (1 of 8) (6) RAP dust (1 of 8) (7) Cement (1 of 8) (8) Caliche –sodium nitrate (1 of 8)

7. Does your agency/company have defined procedures for mixing mineral fillers in HMA

during production in the field?

Y es ( 25 of 54) No (29 of 54) 8. Does/did your agency sponsor research projects related to the usage of mineral fillers?

Y es (4 of 39) No (35 of 39)

(1) Lime (1 of 4)

(2) Baghouse (1 of 4)

(3) Cement (1 of 4)

Experience Related to Slurry Seals

Because slurry seals include high content of fillers, and some issues have been identified with certain added (imported) fillers, please answer the following questions.

16

9. Does your agency specify types of fillers allowed for use in slurry seals?

Y es (12 of 43, 4 no issues) No (31 of 43)

If yes, have you experienced any issues with specific fillers in slurry seals, please provide information about sources and properties of these fillers.

(1) Cement (10 of 12) (2) Lime (5 of 12) (3) Limestone dust (2 of 12) (4) Fly ash (2 of 12) (5) Aluminum sulfate (1 of 12)

2.2.2 Summary of Survey Results The following summarize the survey findings.

(1) Specifications for natural mineral fillers. Most of the agencies (50 of 54) do not have specifications for natural mineral fillers (baghouse dusts). Baghouse dust is considered part of mineral aggregate. In fact, the agencies use the term “mineral fillers” specifically for imported mineral fillers.

(2) Specifications for imported mineral fillers. Most of the agencies (44 of 54) have specifications for manufactured mineral fillers. Twenty-three of the 44 agencies use ASTM D 242 or AASHTO M 17 as the specification for imported mineral fillers. For the agencies that do not use ASTM or AASHTO standards, the gradation requirements in their specifications are different from those in ASTM or AASHTO standards, specifically for the maximum size and percentage passing No. 200 sieve (0.075mm). The No. 30 (0.6mm) is most commonly used as the maximum size. The plasticity index limits in the specifications are also different from those in the ASTM or AASHTO. It is noted Georgia and Ohio DOTs use Superpave tests on mastics as specifications for mineral fillers.

(3) Other than gradation, are there other characteristics of mineral filler that you believe have an important effect on HMA performance?

There were 39 written responses to this question, of which 27 agencies believe other characteristics of mineral fillers have important effects on HMA performance. These characteristics include plasticity index, stripping properties, and carbon content of mineral fillers.

(4) Additional tests for mineral fillers. Only eight agencies suggested additional tests on mineral fillers. These tests include gradation requirements below No. 200 (0.075), chemical effects of fly ash and lime, and particle shape.

(5) Use of imported (manufactured) fillers

17

There are 37 agencies using imported mineral fillers. The four imported mineral fillers that are most commonly used are hydrated lime (22 of 37), stone fines (18 of 37), fly ash (15 of 37), and cement (13 of 37). These four mineral fillers have been included in the experimental plan of this project.

(6) Primary sources of natural mineral fillers The most common sources for imported (manufactured) mineral fillers are limestone (21 of 43), gravel (12 of 43), granite (7 of 43), sandstone (6 of 43), dolomite (5 of 43), and quartzite (4 of 43). Some agencies indicated the use of baghouse dusts (7 of 43), without specifying the source or type. This list of sources has been considered in the experimental plan.

(7) Distinct mineral fillers Only eight agencies recommended studying distinct fillers in this project. Fly ash (5 of 8) and lime (3 of 8) are the two fillers suggested by the agencies. These two fillers have been included in the experimental plan.

(8) Specifications for mixing mineral fillers in HMA Twenty-five of the 54 agencies have specifications for mixing mineral fillers in HMA. These specifications will be reviewed in detail in Phase II when the sampling location of mineral fillers in asphalt plants is recommended.

(9) Research Projects Only four agencies reported that they sponsored research projects related to mineral fillers influence on asphalt pavements. Nevada DOT studied the effect of lime in HMA; South Carolina DOT did a study on baghouse fines; Tennessee DOT sponsored research on cement as mineral fillers; and the fourth agency did not provide the research topic.

(10) Mineral fillers in slurry seal A total of 12 agencies specified the type of fillers used in slurry seal. The top two mineral fillers used by most agencies in slurry seal are cement (10 of 12) and lime (5 of 12). No major concerns were reported for using fillers in slurry seals.

2.3 Fillers Characteristics

The literature review conducted in Task 1 of the project indicated that the most commonly measured filler characteristics can be classified into two categories: physical / geometrical and chemical composition. For each category a number of properties have been found to be important with regard to influence on mastic and/or mixture performance.

18

1. Physical – Geometrical Characteristics a. Size and size distribution

b. Specific gravity c. Surface area

d. Fractional voids –Rigden voids e. Shape, angularity, and texture

2. Compositional/ Chemical Characteristics a. Plasticity index – AASHTO T90

b. Clay Content - Methylene Blue test c. pH value of diluted suspension

d. Water solubility/ susceptibility e. Elemental analysis

Based on critical review, and limited experimentation with equipment used to measure each of the properties listed, the research team selected a subset of tests for use in this study. The target measurable property and the corresponding testing protocol for the selected tests are shown in Table 2-1.

Table 2-1 Filler Characteristics Measured Geometrical Property Testing Protocol

• Fractional V oids Rigden V oids EN 1097-4

• Size Distribution Laser Diffraction ASTM D4464

• Bitumen Number Modified Penetrometer EN13179-2

• Specific Gravity Helium Pycnometer ASTM D5550 – 06

Compositional Property

• Calcium Compounds X-ray Fluorescence (Device Manual)

• Water Solubility Dissolution in Water EN 1744-1:1998

• Methylene Blue / Plasticity Index Modified Method AASHTO TP57, AASHTO T90

• Organic Content Loss on Ignition Test AASHTO 267-86

The following section describes the details of the selected filler tests and provides justification for the selection.

19

2.3.1 Filler Geometry Tests and Results The geometry of fillers can be defined by four characteristics: size, shape, angularity, and texture, similar to that of aggregates. It is well documented in literature that each of these characteristics can affect mastic and HMA performance (4, 7, 13, 18, 43-52). However, measuring each of these characteristics separately and defining its effects on HMA performance can be difficult. Therefore, in this study fractional voids, size distribution, and specific gravity were the geometric properties selected as best practical indicators of fillers’ geometry.

Fractional Voids’ Content: The capacity of filler to hold asphalt binder can be

determined by estimating the voids entrapped in a compacted sample of filler. The void content is estimated by compacting dry fillers in a specific size mold using a specified compaction effort. The concept is similar to the Fine Aggregate Angularity test currently utilized in the Superpave system to measure the angularity of fine aggregates (AASHTO 304). Measurement of fractional voids was introduced by Rigden in 1947 (2). Hence, most researchers refer to the fractional voids test as the “Rigden Voids” test (3,4,5). The current European Norms used to measure fractional voids (EN 1097-4) includes using the original Rigden apparatus developed in 1947 (6). Another version of the Rigden Voids was developed in the U.S in 1987 based on a number of studies and is used by at least one Highway Agency to qualify fillers for Stone Matrix Asphalt (4, 7). A number of studies concluded the four filler geometric characteristics (size, shape, angularity, and texture) affect Rigden voids and thus fractional voids can be considered as an effective indicator of the interactive effect of these characteristics on stiffening of binders (3, 8-19). In this study both versions of the Rigden voids test were evaluated and the original version as used in the EN 1097-4 standard was adopted due to larger sample used and availability of equipment. The Rigden voids (RV) results collected in this study are shown in Figure 2-1, sorted from lowest to highest value. The results show a wide range of values (26.2 % - 49.1%) for the tested fillers.

20

Average = 36.02%

20

25

30

35

40

45

50

55

FAC

1FA

C2

LS1

DH

2FA

N2

DS

2G

RQ

1C

M1

FAF1

LH1

BH

1B

H2

DS

1LS

2G

RQ

2B

V1

HL2

GH

B1

GH

2H

L1FS

1G

S1

CA

1FA

N1

AN

1G

H1

DH

1C

A2

GS

2FS

2

RV(

%)

Minimum = 26.20%Maximum = 49.10%

Figure 2-1 Distribution of Rigden Voids Values for the Study Fillers.

According to the literature, RV provides an indicator of the stiffening effect of the filler on the asphalt binder viscosity. It is shown that higher RV results in higher stiffening effect because more asphalt binder is required to fill the voids leaving less for separation between filler particles (2, 4, 12, 20, 23, 47, 53-56). It is also important to note that results shown in Figure 2-1 indicate the minimum and maximum values of RV correspond to manufactured (imported fillers) that are produced from industrial and/or production processes. The minimum values of the RV belong to fly ash type C (FAC1, and FAC2). The filler FS2, which a blast furnace slag, had the highest RV value.

21

The results also show that fillers selected cover a wide range of Rigden Voids (RV) that is substantial enough to cover the range of most fillers used in practice. Additionally, there appears to be no trend in RV values with source or mineralogy of natural fillers.

Bitumen Number (BN): Although the test name includes the term “Bitumen” it does not involve testing with asphalt. The bitumen number provides another indicator of the geometric properties of fillers and is specified in the European Norms (EN13179-2). Measuring bitumen number is accomplished by measuring the consistency of a filler-water mixture to determine the amount of water required to allow a 5.0 to7.0mm of penetration of a metal cylinder in the mixture of water and filler after five seconds of mixing.

In this study, the bitumen number test was conducted according to European Norms EN13179-2. The distribution of the results is shown in Figure 2-2. The values are similar to those obtained in the RV values. The blast furnace slag (FS2 = 46) maintains its position as the extreme maximum. On the other hand, FAC1 and FAC2 did not result in the minimum values although both are below average. The lowest value for the BN resulted from soft lime stone (LS1 = 23).

22

Average = 32.17

15

20

25

30

35

40

45

50

LS1

DS

2FA

N2

LH1

DH

2FA

C1

GR

Q1

FAF1

DS

1B

H1

BV

1FA

C2

GS

1B

H2

AN

1LS

2C

A1

FS1

HL2

GR

Q2

GH

2G

HB

1G

H1

HL1

CA

2G

S2

FAN

1D

H1

FS2

Bitu

men

Num

ber

(mL)

Minimum = 23mLMaximum = 46mL

Figure 2-2 Distribution of Bitumen Number Results

Because the RV and BN tests are based on the same basic principal, one would expect similar results from the two tests. The plot in Figure 2-3 shows the correlation between the values of the BN and the RV values. A positive trend between RV and BN can be noted.

23

y = 0.89xR2 = 0.78

20.0

25.0

30.0

35.0

40.0

45.0

50.0

55.0

20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0

Rigden Voids (%)

Bitu

men

Num

ber (

mL)

Line of Equality

Figure 2-3 Bitumen Number and Rigden Void Relationship

The BN values are consistently lower than the RV values, which could be because the water used in the BN test lubricates filler particles allowing them to pack denser than in the dry compaction in the RV. However, the strong correlation between both parameters leads to the conclusion that only one needs to be used to obtain an indication of the voids. Because Rigden Voids is more user friendly and can be used with cementitious fillers that cannot be tested with the BN test, it was selected for this study.

Size Distribution: Particle size and size distribution of mineral fillers are

indicators of surface area. It has been reported that size distribution and surface area are not correlated to fractional voids. Thus, they need to be measured separately. Surface area may be measured directly, but this is complicated. Hence, surrogate procedures only were considered in this study. Size distribution can be measured using a number of techniques including sedimentation (hydrometer), light scattering, and sieving. In 1995, Harris and Stuart reviewed the merits of different techniques for measuring size distribution and concluded that laser diffraction (light scattering) is the best available method (20). They investigated the repeatability of the laser method, the optimum solution of the dispersing fluid, and the agitation time required for achieving better dispersion and consistent results. The laser diffraction method, with the consideration of the recommendations by Harris and Stuart, was used in this study for measuring filler size distribution. The device, is however, relatively expensive and initial estimates are in the range of US $30,000 to US $50,000.

24

In the laser diffraction test, a prepared sample of filler is dispersed in distilled water with a surfactant and then is circulated through the path of a laser light beam or some other suitable source of light. The particles pass through the light beam and scatter it. Photo-detector arrays collect the scattered light, which is converted into electrical signals to be analyzed using either Fraunhofer Diffraction, Mie scattering, or both. Scattering information is typically analyzed assuming particles are spherical. Calculated particle sizes are, therefore, presented as equivalent spherical diameters. The relative fineness of an aggregate can be determined using the fineness modulus (FM), which is calculated by dividing the sum of the percentages of P200 material coarser than 75, 50, 30, 20, 10, 5, 3, and one micron by 100. Finer fillers result in smaller fineness moduli. Fineness modulus was used as the indicator of filler particle size in this study. However, it is important to mention that other parameters can also be determined from laser diffraction. Parameters such as D10, D30, and D60 can also be determined from particle size analysis. These parameters are the particle sizes that correspond to 10, 30, and 60% of the material passing a specific sieve, respectively. The specific surface area (cm²/ml), commonly referred to as SA, can also be estimated from this analysis. The distribution of the FM values for the fillers tested is shown in Figure 2-4. It clearly shows these fillers have a wide range of particle sizes. The influence of the particle size distribution of filler on mastic and mixture properties is unclear based on the literature. The fly ash type N (FAN2, 2.98) from source two has the lowest FM value indicating that it is the finest among the fillers collected for this study. On the other hand, the Quartzite gravel (GRQ2, 6.32) is the coarsest based on its FM value. The use of the FM coupled with the RV values is believed to provide a complete picture of the fillers’ geometric properties as the RV is an indicator of surface area and size distribution while the FM is an indicator of shape and texture of the filler.

25

Average = 4.45

2.0

2.5

3.0

3.5

4.0

4.5

5.0

5.5

6.0

6.5

7.0

FAN

2G

S2

FAN

1FA

C1

LS2

GH

2G

S1

FAC

2G

H1

HL2

GH

B1

CM

1FA

F1LS

1FS

2G

RQ

1D

H2

BV

1B

H2

DS

2H

L1FS

1C

A1

DH

1C

A2

AN

1B

H1

LH1

DS

1G

RQ

2

FM

Minimum = 2.98Maximum =6.32

Figure 2-4 Distribution of Fineness Modulus

Specific Gravity: The specific gravity value is required for calculating the

fractional voids and is also needed for HMA design. Because of the problems encountered with the AASHTO T-84 procedure for fine aggregate specific gravity, some State Highway Agencies (SHAs) use other methods to measure this important property. The Pennsylvania DOT requires the use of the AASHTO T-133, which measures the density of Portland cement in kerosene. This method appears to be a logical substitute for AASHTO T-84. In addition, discussion with the manufacturers of the CoreLock device indicates the device could be useful for testing filler apparent specific gravity, but such measurements have not been attempted. In this study, it was determined that the Helium Pycnometer is the best alternative and was used according to the standard specification ASTM D5550 – 06 to measure specific gravity. The helium pycnometer method was selected for its simplicity, repeatability, and because the FHW A technical staff has experience with the method.

26

The results of the specific gravity testing of the fillers used in this study are plotted in Figure 2-5. Similar to the results of the RV, extreme values are measured for the manufactured fillers. The highest three specific gravities and lowest three specific gravities are measured for the manufactured fillers. Since the mastics tested in this project were produced at one mass ratio of filler to binder of 1:1, it is important to show how much the variation in specific gravity could change the volume fraction of filler in the mastics. The average volume fraction of the fillers, excluding the manufactured fillers, used in producing the mastics for this study is 28%. The volume fraction of the fillers tested range from 26.5% to 29.5%. Due to the small variation in volume fraction, it was concluded volume fraction is unlikely to significantly influence mastic properties. The issue is discussed in more detail in Chapter 2.5.

Average = 2.61

2.0

2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

2.9

3.0

FAF1

FAC

2FA

N1

GS

2FA

C1

CA

2G

H2

FAN

2G

RQ

2G

S1

DH

1C

A1

AN

1D

H2

LS2

GH

B1

LS1

LH1

GH

1G

RQ

1H

L1D

S2

DS

1B

H1

BH

2B

V1

HL2

FS1

CM

1FS

2

SG

Minimum = 2.14Maximum = 2.89

Figure 2-5 Distribution of Specific Gravity (SG) Results

Legend 1st Set of Letter(s): Filler Letter after 1st Set: Filler Type AN: Andesite FAF: Fly Ash Type F H: Hard B: Basalt FAN: Fly Ash Non Spec S: Soft BV: Vesicular Basalt FS: Furnace Slag CA: Caliches G: Granite Number: Source CB: Carbon Black GRQ: Gravel Quartzite 1: Source 1 CM: Cement HL: Hydrated Lime 2: Source 2 D: Dolomite L: Limestone FAC: Fly Ash Type C

27

2.3.2 Filler Composition Tests and Results It is recognized that stiffening effect of mineral fillers on binders is dependent on filler chemical composition (3, 9, 10, 12-15, 21-32, 38) . The chemical compounds that are most widely cited in the literature to have an influence include calcium compounds, organic compounds, water soluble compounds, and clay type materials. In this study, testing was conducted to measure each of these compounds using existing technology.

Calcium Compounds’ Content: Significant research has been conducted on the

potential reactivity of calcium compounds with asphalt. For example, in the European Norms, the content of hydrated lime and calcium carbonates are used as criteria to define the relative reactivity of mineral fillers. Also, some SHAs in the U.S. have adopted procedures for measuring acid insoluble residue and lime content of fillers (33). The methods are intended to characterize sources of limestone aggregates for calcium carbonate and hydrated lime quality. There are also standard protocols in the European Norms to measure calcium hydroxide (Ca(OH)2) and calcium carbonate (CaCO3) contents (34,35). A review of existing technology to measure calcium compounds indicates X-ray Fluorescence spectrometry is capable of identifying the presence and amounts of all elements. Therefore, X-ray Fluorescence was selected for use in this study. Since there is no standard procedure for conducting the testing the manufacturer manual of the Panalytical Epsilon 5 EDXRF device was adopted (36). The measurements for the fillers collected in this study demonstrated a very wide range of calcium oxide contents (CaO%), as evident by the results displayed in Figure 2-6. This wide range provides confidence that the results of this study are representative of the range of fillers currently used in HMA.

28

The slag fillers (FS1-50.3, and FS2–45.1) have exceptionally high calcium oxide contents, (CaO%) values. In addition, results agree with expectations as the limestone and caliches fillers have high CaO% contents, while the granites have low CaO% contents.

Average = 23.03

0

10

20

30

40

50

60G

RQ

2G

H2

GH

1G

S1

GH

B1

FAF1

AN

1B

H2

GS

2B

H1

BV

1C

M1

FAN

1FA

C2

FAN

2FA

C1

DH

1D

S1

DS

2H

L1D

H2

HL2

GR

Q1

CA

2LH

1C

A1

FS2

LS2

LS1

FS1

CaO

(%)

Minimum = 0.95%Maximum = 50.3%

Figure 2-6 Distribution of Calcium Oxide (CaO%)Results

Water Solubility: Water solubility, while not commonly measured in research or

used in the U.S. standards, is listed in the European Norms (35). The pH value of the filler solution has been used more widely as an indicator of reactivity. However, the most extensive study on pH value concluded that no relationship could be found between pH value and the laboratory performance of mastics or HMA (19). A number of agencies measure the value of pH for imported fillers, such as fly ash. However, since it is unclear from the literature if pH value affects the performance of HMA, testing for pH was not

29

conducted as part of this study. Instead, testing of water solubility of fillers was conducted in accordance with the European norm EN 1744-1:1998. In the water solubility test, pre-weighed dry filler is immersed twice in water. After 24 hours of conditioning, the water is filtered. The filler is dried to determine change in mass. Figure 2-7 depicts the results of water solubility testing. It is important to note that the majority of the fillers show very low values of percent loss in mass. However, few fillers show a gain in mass during the test. This unexpected result could be due to reaction of the filler particles with water resulting in chemical bonding of water molecules. Further evaluation of water solubility was aborted due to the results of this test in addition to lack of evidence in literature of its importance.

Average = 0.269

-15.0

-10.0

-5.0

0.0

5.0

10.0

FAC

2FA

N2

FAC

1FA

N1

HL2

GS

2FS

1G

H2

HL1

GS

1FS

2G

RQ

2LS

1FA

F1D

S1

DS

2C

A1

GR

Q1

LH1

CA

2G

HB

1D

H2

LS2

DH

1G

H1

AN

1B

V1

BH

1B

H2

Wat

er S

olub

ility

(%)

Minimum = -13.10Maximum = 8.45

Figure 2-7 Distribution of Water Solubility (loss due to dissolution) Results

30

Harmful Fines: Testing for the presence of harmful fines is included in the majority of aggregate specifications. The main materials considered harmful due to their chemical composition are clay and organic materials. Clay presence is a problem due its high moisture retention and surface charge. The most commonly used tests to detect active clay are the Methylene Blue (MB) test and the Plasticity Index (PI). The MB test is a standard test conducted by some SHAs. The plasticity index measurement is commonly used but is not considered to be as effective as the MB test in determining the content of active clay. Since many SHAs in the U.S. measure PI and MB, both were measured in this study. The repeatability of the MB test and how the intensity of the color measured are of concern. However, given the advancement in light spectroscopy, these concerns could be resolved using a color spectrometer. Additionally, the organic content of fillers was measured using the loss on ignition standard test (AASHTO 267-86) as will be discussed later.

In the modified MB test, a spectrophotometer is used to monitor change in color absorbance as a result of adding the Methylene blue solution to a sample of oven - dried mineral fillers mixed with sodium pyrophosphate. The procedure involves adding selected amounts of Methylene Blue solution until a target interval is defined along the Mythelene Bluse Value (MBV) scale at which a sharp increase in absorbance is identified, as depicted in the schematic below.

Methylene Blue Volume

Threshold Point

Target Interval

VL VH

The procedure is then repeated with finer steps to narrow the range of transition. In this study three steps were included starting with 5ml of Methylene Blue solution, followed by a second step with 1.5ml, and then a final step of 0.5ml. Thus, the procedure allowed for determination of the MBV with a precision of 0.5ml. The MBV is calculated as follows:

MVCVBM ×

=...

where, V = Methlyene blue solution volume needed (threshold)

31

C = Concentration of methylene blue M = Filler mass The results of the MB test are presented in Figure 2-8. The MB values observed in this study vary between zero and 31.6.

Average = 4.86

0

5

10

15

20

25

30

35

LS1

FAC

1FA

C2

FS1

FAN

1FA

N2

FS2

FAF1

LH1

GR

Q2

HL1

GS

2D

H2

HL2

DS1

DS2

GR

Q1

GH

1D

H1

GH

B1 LS2

BH1

CA1

CA2

BH2

BV

1G

S1

AN1

GH

2

MB

V

Minimum = 0.0Maximum = 31.56

Figure 2-8 Distribution of Methylene Value Test Results

The plasticity index test (AASHTO T90) is commonly used to measure the degree of plasticity of fines. Plasticity is the difference between the liquid limit and the plastic limit of the material passing No. 40 (425µm) sieve. In this study, results indicate most of the fillers collected are non-plastic. Only three fillers had a measurable plasticity value. These PI values for the three fillers are listed in the following table.

32

Table 2-2 Plastic Fillers Code PI BV1 3.00

GHB1 6.00 DS1 10.00

The organic content of fillers was measured using the Loss-On-Ignition (LOI) test (AASHTO T267). The test is widely used and no modification appears necessary. In the LOI test, a filler sample dried in a 110ºC oven for at least one hour. The dried sample is placed in a muffle furnace at 455 ± 10ºC to burn off the organic materials. The weight of the sample before and after ignition is used to determine the amount of material lost due to ignition. The results of the LOI test are included in Figure 2-9. The results show that the slag filler has the highest loss in mass of the fillers tested. As with most of the filler properties measured, extreme values occurred mostly in manufactured fillers.

Average = 0.71%

0.0

0.5

1.0

1.5

2.0

2.5

CM

1G

RQ

1LS

2G

S2FA

C1

LH1

FS2

LS1

DS1

GR

Q2

DS2

GH

1G

HB1

FAC

2H

L2G

H2

FAN

2BV

1G

S1AN

1FA

F1BH

1C

A1D

H2

BH2

CA2

FAN

1H

L1D

H1

FS1

LOI(%

)

Minimum = 0.00%Maximum = 2.12%

Figure 2-9 Distribution of Loss on Ignition Results

33

2.4 Summary of Fillers’ Properties The following summarizes the results of the fillers' testing. It is important to note that after the procurement of the carbon black fillers, the research team decided to abandon testing this filler due to difficulties, namely being extremely fine with low density which makes it easily airborne and difficult to handle. Also, Portland cement was not tested using MB and the BN due to its reactivity with water. A complete summary of filler testing results is included in Table 2-3.

Table 2-3 Summary of Filler Testing Results

No. Code Rigden Voids (%) FM CaO

(%) MBV SG LOI (%) PI Solubility

(%)

1 AN1 41.9 5.18 6.80 24.56 2.60 0.92 Non-Plastic 3.0 2 BH1 33.2 5.31 8.20 5.92 2.72 1.01 Non-Plastic 4.9 3 BH2 33.8 4.65 7.00 11.19 2.77 1.25 Non-Plastic 8.4 4 BV1 37.8 4.59 9.10 11.65 2.79 0.77 3.00 4.5 5 CA1 40.3 4.91 44.00 8.96 2.59 1.01 Non-Plastic 0.8 6 CA2 45.0 5.13 40.00 10.25 2.49 1.35 Non-Plastic 1.0 7 CBC1 NA NA NA NA NA NA NA NA 8 CBF1 NA NA NA NA NA NA NA NA 9 CM1 30.0 4.27 14.30 NA 2.87 0.00 Non-Plastic NA

10 DH1 42.8 5.07 26.00 2.79 2.59 2.01 Non-Plastic 1.5 11 DH2 27.0 4.54 32.01 0.99 2.61 1.01 Non-Plastic 1.4 12 DS1 34.7 6.02 26.00 1.17 2.71 0.11 10.00 0.7 13 DS2 29.4 4.73 27.00 1.82 2.70 0.35 Non-Plastic 0.7 14 FAC1 26.2 3.30 25.03 0.00 2.47 0.09 Non-Plastic -6.4 15 FAC2 26.2 3.88 23.13 0.00 2.37 0.50 Non-Plastic -13.1 16 FAF1 30.1 4.29 5.10 0.39 2.14 1.00 Non-Plastic 0.7 17 FAN1 40.5 3.22 21.80 0.01 2.38 1.66 Non-Plastic -1.3 18 FAN2 28.9 2.98 24.45 0.02 2.54 0.76 Non-Plastic -8.1 19 FS1 40.0 4.87 50.30 0.00 2.80 2.12 Non-Plastic 0.2 20 FS2 49.1 4.32 45.07 0.02 2.89 0.10 Non-Plastic 0.5 21 GH1 42.6 4.06 3.50 2.35 2.66 0.36 Non-Plastic 2.3 22 GH2 38.8 3.76 2.80 31.56 2.53 0.64 8.00 0.3 23 GHB1 38.3 4.26 4.70 2.81 2.62 0.40 6.00 1.3 24 GRQ1 29.5 4.39 32.70 2.06 2.66 0.03 Non-Plastic 0.9 25 GRQ2 36.5 6.32 0.95 0.66 2.55 0.21 Non-Plastic 0.5 26 GS1 40.2 3.85 4.60 14.47 2.58 0.85 Non-Plastic 0.4 27 GS2 47.0 3.13 7.80 0.88 2.40 0.05 Non-Plastic 0.1 28 HL1 38.8 4.74 27.60 0.80 2.70 1.91 Non-Plastic 0.3 29 HL2 38.1 4.15 32.20 1.05 2.79 0.54 Non-Plastic -1.0 30 LH1 32.2 5.63 43.14 0.62 2.65 0.10 Non-Plastic 1.0 31 LS1 26.2 4.30 49.20 0.00 2.64 0.11 Non-Plastic 0.6 32 LS2 35.4 3.68 46.30 3.87 2.62 0.04 Non-Plastic 1.5

The results shown were used to select fillers to blend with the asphalt binders for mastic testing. It was important to confirm that the filler properties used for the selection are

34

independent of each other. Table 2-4 shows the correlations matrix between all properties measured.

Table 2-4 Correlation Factors for the Filler Testing Results

Rigden Voids (%) FM CaO% MBV SG LOI(%) Solubility

(%) BN

Rigden Voids (%) 1.00

FM 0.06 1.00

CaO% -0.10 0.06 1.00

MBv 0.28 0.04 -0.40 1.00

SG 0.19 0.35 0.30 0.03 1.00

LOI (%) 0.26 0.11 0.02 0.12 -0.02 1.00 Solubility

(%) 0.35 0.44 -0.18 0.32 0.44 0.16 1.00

BN 0.89 -0.07 -0.06 0.08 0.02 0.33 0.15 1.00

As shown in Table 2-4, Rigden voids and Bitumen Number are highly correlated. Since solubility and loss on ignition (LOI) were found to have a very narrow distribution, they were not considered for selection. The specific gravity is needed for the Rigden voids and thus was considered as part of the RV test. From the analysis, selection of fillers to be used in mastic testing was based on only two primary geometric indicators (RV and FM) and two primary chemical indicators (CaO% and the MBV). Narrowing the primary properties to these four parameters provided a balanced selection of geometric and chemical indicators. As explained in Chapter 1, fillers used in the mastic testing plan were selected to cover two levels (High and Low) for each of the four primary properties. The fillers were distributed into 16 combination based on the primary filler properties according to a 24 factorial design. The definition of High and Low values was based on whether their property values are above or below the average values for each of the four primary properties. For example, fillers with high RV are those with values more than but not equal to the average RV value calculated for the sample set. Table 2-5 lists the 17 selected fillers. The table includes the measured filler properties and the combination level (H=high, L=low) that each filler meets. The requirements that are not met are shown in property columns with a (?) mark next to the level. In the column for selected filler code the number of experimental design levels (H, L) that the filler satisfies is shown in parentheses. For example for the Caliches (CA2) filler, it meets all four levels and thus the indictor (4) is shown next to it. It is important to note that due to the fact that not all fillers satisfied the high and low levels as required by the experimental design, the research team used engineering judgment in selecting fillers to cover the most commonly used.

35

Table 2-5 Selected Fillers to Proceed to Mastic Testing

Combi-nation

Number

RV Avg =

36.01%

FM Avg = 4.45

CaO Avg = 23.03

MBV Avg= 4.86

Code of Selected

Filler (Levels

satisfied )

Selected Filler Type

1 H (45.00)

H (5.13)

H (40.00)

H (10.25) CA2 (4) Soft Caliches

2 L (33.80)

H (4.65)

H ? (7.00)

H (11.19) BH2 (3) Hard Basalt

3 H (38.10)

L (4.15)

H (32.20)

H ? (1.05) HL2 (3) Hydrated Lime

4 L (35.4)

L (3.68)

H (46.30)

H ? (3.87) LS2 (3) Soft Limestone

5 H (41.90)

H (5.18)

L (6.80)

H (24.56) AN1 (4) Andesite

6 L (33.20)

H (5.31)

L (8.20)

H (5.92) BH1 (4) Hard Basalt

7 H (40.20)

L (3.85)

L (4.60)

H (14.47) GS1 (4) Soft Granite

8 L? (38.80)

L (3.76)

L (2.80)

H (31.56) GH2(3) Hard Granite

9 H (42.80)

H (5.07)

H (26.00)

L (2.79) DH1(4) Hard Dolomite

10 L (29.40)

H (4.73)

H (27.00)

L (1.82) DS2 (4) Soft Dolomite

11 H (49.10)

L (4.32)

H (45.07)

L (0.02) FS2 (4) Blast Furnace

Slag

12 L (26.20)

L (3.88)

H (23.13)

L (0.00) FAC2 (4) Fly Ash Type C

13 H (36.50)

H (6.32)

L (0.95)

L (0.66) GRQ2 (4) Siliceous Gravel

Quartzite

14 L (32.20)

H (5.63)

L ? (43.14)

L (0.62) LH1 (3) Hard Limestone

15 H (42.60)

L (4.06)

L (3.50)

L (2.35) GH1 (4) Hard Granite

16 L (30.10)

L (4.29)

L (5.10)

L (0.39) FAF1 (4) Fly Ash Type F

17 L? (47.0)

L (3.13)

L (7.80)

L (0.88) GS2 (3) Soft Granite

2.5 Effect of Fillers on Mastic and Mixture Performance The following section discusses the effect of fillers on mastic and mixture performance for each of the five performance criteria considered. For example the effects of fillers on mastic viscosity and mixture resistance to gyrations are discussed under workability. In addition to the workability section, there are four other sections discussing rutting, fatigue, low temperature cracking, and moisture damage resistance.

36

Upon completion of filler characterization, testing and evaluation proceeded with mastic and mixture testing using a sub-set of the fillers tested. The selected 17 fillers were blended with four binders as mentioned in Chapter 1 to create a total of 68 mastics. All mastics were produced at a 1:1 mixing ratio of filler to binder by mass in accordance with the Superpave specification, which recommends the mass ratio of filler to bitumen be in the range of 0.6 to 1.2. It is important to note that although the mastics were produced at a constant mass ratio of filler to binder, the variability in the volumetric fractions was limited to a narrow range of 26% to 32%, as shown in Figure 2-10. This is because the specific gravities of the natural fillers are very close as mentioned earlier in the filler test results section. Variability in Filler Volume Fraction

20%

22%

24%

26%

28%

30%

32%

34%

36%

38%

40%

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS2

HL2

Vol

ume

Frac

tion

(%)

Natural Manufactured

± One Standard Deviation

Average

Figure 2-10 Variability in Fillers Volumetric Fraction in Mastics

The data in Figure 2-10 shows the average volume fraction of filler to mastic approximately 28%. The data also shows for the natural fillers the range in volume fraction is within ±2 % of this average. Based on a detailed study by members of the research team in 2009 (39), it was determined initially that this variation in volume fraction is not important and cannot justify adjusting the mixing ratio by mass to achieve one volume faction of all fillers in mastics. In the 2009 study, the mastics were tested at multiple volumetric fractions ranging from 10% volume fraction to approximately 65%.

37

One of the main conclusions of the study was that at filler concentrations lower than 40% by volume, the influence of the filler is more dependent on the RV and the size distribution of the filler than the volumetric concentration. Furthermore, for natural fillers, volume fraction and Rigden voids are correlated (R2 = 0.57), as evident by Figure 2-11 below. However, no correlation between volume fraction and Rigden voids exists for manufactured fillers. Therefore, in specification development, (discussed in Chapter 3), manufactured fillers were not considered and Rigden voids was considered as a property for specification, but volume fraction was not.

All Fillersy = -0.0006x + 0.3011

R2 = 0.07

Without Manufatured FIllersy = 0.0011x + 0.2358

R2 = 0.57

0.20

0.22

0.24

0.26

0.28

0.30

0.32

0.34

0.0 10.0 20.0 30.0 40.0 50.0 60.0

Volu

me

Frac

tion

Rigden Voids (%)

FAC2

FS2

FAF1

HL2

Figure 2-11 Relationship between Volume Fraction and Rigden Voids

The asphalt mixtures were prepared using short-term aging. The fillers were introduced during the mixing processes at the same time the binder was added. The mix design called for 4% filler for the fine-graded mix and 5% for the coarse-graded mix by mass. The fines passing #200 (0.075mm) sieve were removed from the mix by dry sieving for the coarse aggregate and washing for the fine aggregate. This would allow for a controlled mix that isolates the effect of the added filler. Based on the outcome of the mastic modeling, when adding different fillers in the mix, the amount was adjusted to equate the volume of the original filler included in the mix design. This ensured that any change in the mix performance is due to the mineral filler type and not the change in mix components’ volumetrics.

38

0.45 Fine Gradation Chart

0

10

20

30

40

50

60

70

80

90

100

Seive Size (mm)

% P

assi

ng

Upper CPs

Lower CPs

Fine Gradation

Spec High

Spec Low

.075 .15 .3 .6 1.18 2.36 4.75 9.5 12.5 19 25 37.5

Figure 2-12 Aggregate Fine Gradation

0.45 Coarse Gradation Chart

0

10

20

30

40

50

60

70

80

90

100

Seive Size (mm)

% P

assi

ng

Upper CPs

Lower CPs

Coarse Gradation

Spec High

Spec Low

.075 .15 .3 .6 1.18 2.36 4.75 9.5 12.5 19 25 37.5

Figure 2-13 Aggregate Coarse Gradation

According to the asphalt institute, fine aggregate is the fraction that passes sieve number 8 or 2.36mm. Consequently, coarse aggregate represents the fraction retained on sieve number 8. The national asphalt pavement association (NAPA), on the other hand, defines fine aggregate gradation as that, when plotted on the 0.45 power gradation graph, falls mostly above the 0.45 power maximum density line. The term generally applies to dense graded aggregate (NAPA virtual Superpave laboratory). Consequently, the coarse gradation falls mostly below the maximum density line.

39

The fine aggregate gradation used in this study follows the defection of the NAPA. The coarse gradation falls below the maximum density line starting with the sieve #4 and smaller. The following subsections illustrate the effect of mineral fillers on mastic and mixture distress categorized by performance criteria.

2.5.1 Effect of Fillers on Mastic and Mixture Workability Indicators To evaluate the effect of filler on mastic workability, viscosity was measured at 135°C using a Brookfield Rotational Viscometer. Viscosity testing was conducted using a size 27 spindle. After 30min of temperature equilibrium, the spindle is rotated at 20rpm for 600sec. Viscosity measurements over the last 300sec of testing are averaged as the apparent viscosity. It should be mentioned that several experiments were conducted to evaluate the effect of shear rate on viscosity of mastics. Although the mastics were found to be shear dependent, the relative ranking of mastics did not change as a function of the shear rates covered in the testing. It was therefore appropriate to use the results at only one shear rate. The shear rate of 20rpm was used because it is currently the most commonly used shear rate in binder testing. The distribution of mastic viscosity values for the 68 mastics is displayed in Figure 2-14. In an attempt to simplify presentation of results, the natural and manufactured fillers are separated in all the plots for mastics and mixtures performance. Also, values for a given binder are connected by a line to show the relative variation.

40

0

2,000

4,000

6,000

8,000

10,000

12,000

14,000

16,000

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

Vis

cosi

ty (c

P)

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-14 Distribution of Mastic Viscosity at 135°C

It is clear that filler type plays a major role in mastic viscosity for any given binder. The mastics containing SBS exhibit consistently higher viscosities than mastics prepared with other binders, indicating SBS modification has the greatest effect on mastic viscosity. Additionally, the mastics containing SBS exhibited the highest variability in viscosity with filler type with a maximum mastic viscosity 3.5 times the minimum due to filler effect alone. This suggests the significant impact of filler type on binders containing SBS. While SBS modification has a notable effect on mastic viscosity, PPA modification and binder source did not lead to significant differences in mastic viscosity. Rather, the filler effect was greater than the binder effect. It appears filler GS2 interacts with all binders as the viscosity of mastics containing GS2 display higher viscosities than mastics containing the same binder but different fillers. Additionally, FS2, one of the manufactured fillers, when combined with the SBS modified binder produces a substantially higher viscosity than any other mastic evaluated. However, FS2 does not produce high viscosity when combined with the other binders which indicates the interaction of the SBS and FS2 is unique.

41

The distribution of relative viscosity is displayed in Figure 2-15. Relative values are computed by dividing mastic measurements by measurements on the binder in that mastic. Hence, relative viscosity is mastic viscosity divided by binder viscosity. It can be seen that there are similarities in the trends as a function of filler type. This indicates that the influence of filler type on relative viscosity is more important than binder type and binder modification. As noted previously, SBS combined with FS2 exhibits extreme behavior, which is reflected in an extremely high relative viscosity. The PPA modified mastics show consistently lower relative viscosities than mastics with other binders, indicating that fillers may negate the expected influence of the PPA modification. It is also noteworthy that for filler CA2, the ranking of relative viscosity with binder type is opposite of most other fillers. The relative viscosity values were used to rank the mastics in order to select a subset of 16 mastics to proceed with the mixture testing. It should be noted that some of the effects seen in Figure 2-15 could be related to the volume fraction variation. In fact there are some similarities in the trends seen in Figure 2-14 and the changes in fractional voids seen in Figure 2-10. For example filler BH2 and DS2 show some of the lowest relative viscosities and at the same time they show the lowest factional voids among the natural fillers. This observation will be used in modeling of the mastic viscosity as a function of filler properties.

0

2

4

6

8

10

12

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

Rel

ativ

e V

isco

sity

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-15 Distribution of Relative Viscosity of Mastics at 135°C

42

Mastics were divided into four quartiles based on relative viscosities to aid in selecting a subset for mixture workability testing. Table 2-6 shows the values for each quartile. A balance in the number of mixtures containing each binder was achieved such that for the total of 16 mixtures tested, there were four mixtures containing each binder.

Table 2-6 Summary of Quartiles for the Relative Viscosity Measurements Minimum Value 1.91 1st Quartile limit Value 3.1 Median limit Value 3.7 3rd Quartile limit Value 4.7 Maximum Value 10.4 Mean Value 4.1

Based on the quartile distribution, the mastics were divided into four categories: low reactivity (relative viscosity less than 3.1), medium-low reactivity (3.1 – 3.7), medium- high reactivity (3.7 – 4.7), and high reactivity (more than 4.7). Four mastics were selected for each category with the aim of maximizing the mineralogy type of fillers while taking into account the binder type and relative viscosity values. Table 2-7 shows the list of the 16 mastics selected. The table also includes the corresponding filler, mastic, and mixture testing results. The mixture testing was conducted using the standard procedure for the Superpave Gyratory Compactor and the workability indicator was defined as the number of gyrations to 92 % Gmm. All mixture testing was done at 135 C using 600 kPa pressure.

43

Table 2-7 Mastics Selection for Mixture Constructability Testing Based on Relative Viscosity.

Category Mastic FILLER TYPE RV FM CaO MBV Viscosity

(cP) Relative Viscosity

Fine Gradation

N92

Coarse Gradation

N92 L

ow

Rea

ctiv

ity

PFAC2 Fly Ash Type C 26.2 3.9 23.1 0.0 1889 1.9 13 28

FLH1 Hard Limestone 32.2 5.6 43.1 0.6 1675 2.8 18 40

VBH2 Hard Basalt 33.8 4.6 7.0 11.2 1806 2.9 12 33

PHL2 Hydrated lime 38.1 4.1 32.2 1.0 2941 3.0 14 28

Med

ium

Low

R

eact

ivity

SDS2 Soft Dolomite 29.4 4.7 27.0 1.8 4178 3.1 16 35

SCA2 Soft Caliches 45.0 5.1 40.0 10.2 4138 3.1 16 35

PGS1 Soft Granite 40.2 3.8 4.6 14.5 3294 3.3 16 29

PGRQ2 Siliceous Quartzite 36.5 6.3 1.0 0.7 3650 3.7 21 31

Med

ium

Hig

h R

eact

ivity

VGRQ2 Siliceous Quartzite 36.5 6.3 1.0 0.7 2660 4.3 14 33

FDH1 Hard Dolomite 42.8 5.1 26.0 2.8 2711 4.5 19 43

FGH2 Hard Granite 38.8 3.8 2.8 31.6 2781 4.6 19 43

VGS1 Soft Granite 40.2 3.8 4.6 14.5 2874 4.7 15 39

Hig

h

Rea

ctiv

ity

VCA2 Soft Caliches 45.0 5.1 40.0 10.2 3489 5.7 20 37

SAN1 Andesite 41.9 5.2 6.8 24.6 9298 7.0 18 37

FGS2 Soft Granite 47.0 3.1 7.8 0.9 4951 8.2 19 51

SFS2 Blast

Furnace Slag

49.1 4.3 45.1 0.0 13750 10.4 17 29

Selected Fillers Max 49.1 6.3 45.1 31.6

Selected Fillers Min 26.2 3.1 1.0 0.0

T otal Fillers Max 49.1 6.3 50.3 31.6

T otal Fillers Min 26.2 3.0 1.0 0.0

44

The results of the mixture workability, measured as the number of gyrations to 92% maximum density, are plotted in Figure 2-16. The figure is divided into four regions by binder type used.

05

10152025303540455055

FDH

1

FGH

2

FGS2

FLH

1

PFAC

2

PGRQ

2

PGS1

PHL2

SAN

1

SCA2

SDS2

SFS2

VBH

2

VCA2

VGRQ

2

VGS1

Num

ber

of G

yrat

ions

to

92%

Max

imum

Den

sity

Mastics

Coarse

Fine

Flint Hills + PPA PG 70-22

Flint Hills + SBS PG 70-22

Valero PG 64-22

Flint HillsPG 64-22

Legend: 1st Letter: Binder 2nd Letter: Filler 3rd/4th Letter: F: Flint Hills A: Andesite FS: Furnace Slag H: HardV: Valero B: Basalt G: Granite S: SoftP: PPA C: Caliches HL: Hydrated Lime 4th/5th Letter:S: SBS D: Dolomite 1: Source 1

FA: Fly Ash 2: Source 2

Figure 2-16 Distribution of the Number of Gyrations to 92% Maximum Density There is a clear distinction between coarse and fine mixtures. All of the coarse mixes are less workable than fine mixes based on the number of gyrations to 92% maximum density. For all mixtures containing a fine aggregate blend, there is little variability in the number of gyrations to 92% maximum density. The same holds for the coarse mixes with the exception of the Flint Hills binder which has a higher number of gyrations for all filler types than the other binders. The mixture results are interesting as the high range of mastic viscosities is not reflected directly in the mixture results. Based on these results, gradation is very important in determining mixture workability while asphalt binder and filler type is less significant. An ANOV A analysis was conducted using gradation and mastic as variables. Results are shown in the table below. As evident by the results, both gradation and mastic have low P-values, indicating they are both significant. However, as would be expected, gradation has an F-value approximately 93 times that for mastic, indicating gradation is of significantly higher importance for workability than mastic viscosity. For the purpose of this study and due to the variability in the natural materials incorporated in this study a statistically significant factor is decided based on a P-value below 10% combined with an R2 greater than 35%.

45

Table 2-8. Number of gyrations to 92% Maximum Density ANOVA: Gradation and Mastic Variable F-Value P-Value Gradation 466.64 0.000 Mastic 5.18 0.000

R2= 89.35% Further analysis was conducted to separate the effect of fillers and binders using variables of aggregate gradation, binder source, binder modification, and filler mineralogy. Results shown in Table 2-9 below confirm gradation is the most important variable. However, binder source and binder modification are also important as evident by low P-values. Filler source or type appears to have a statistically significant effect as well since the P-value is below 10%. The filler effect is however less important than the other factors as shown by the low F statistics value.

Table 2-9. Number of gyrations to 92% Maximum Density ANOVA: Gradation and Mastic Components

Variable (levels) F-Value P-Value Gradation (2) 454.34 0.000 Binder Source (2) 7.39 0.009 Binder Modification (3) 15.12 0.000 Filler mineralogy (9) 1.87 0.072

R2 = 89.06%

2.5.2 Effect of Fillers on Mastic and Mixture Rutting Resistance Mastic and binder rutting resistance was evaluated using the Multiple Stress Creep and Recovery (MSCR) test in accordance with ASTM D7405. Testing included three stresses (0.1kPa, 3.2kPa, and 10kPa) and two temperatures (58°C and 64°C). The non-recoverable creep compliance, (Jnr) and percent recovery, at a stress level of 3.2 kPa and 58°C, were chosen as the indicators of rutting resistance. These test conditions are chosen since changing the temperature to 64°C and the stress level to 10kPa led to the same ranking of mastics. A lower Jnr signifies superior rutting resistance. The distribution of mastic MSCR results is presented in Figure 2-17.

46

0.01

0.10

1.00

10.00

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

Jnr(

1/kP

a)

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-17 Distribution of Mastic Jnr at 58°C

As evident by the high variability in mastic Jnr values for each binder, the interaction between filler and binder has a significant effect on mastic Jnr. The mastics containing SBS exhibit superior rutting resistance based on Jnr values than all other mastics and also exhibit less variability in Jnr with filler type. On the contrary, PPA modification appears to have lower interaction with fillers as PPA mastics generally have similar Jnr values to that of mastics containing the neat Flint Hills binder. The mastics containing the other binders follow the same general trend for the majority of filler types, indicating the filler effect is more dominant than the binder. In addition to examining the distribution of mastic Jnr, it is important to look at the change in Jnr that occurs upon addition of fillers was evaluated. The relative Jnr, which is the mastic Jnr divided by the binder Jnr was used to accomplish this. The distribution of relative Jnr values is shown in Figure 2-18.

47

0.0

0.2

0.4

0.6

0.8

1.0

1.2

No

Fille

r

AN1

BH1

BH2

CA2

DH1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

Rela

tive

Jnr

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-18. Distribution of Relative Jnr at 58°C

All relative Jnr values are below the value of “one”, indicating addition of any of the fillers increases rutting resistance. This is intuitive as addition of fillers causes an increase in stiffness and hence an increase in rutting resistance. With the exception of the PPA modified mastics, all the binders exhibit similar relative Jnr values given the same filler type in the mastic. This indicates, for three of the binders, the filler effect is greater than binder effect with respect to changing Jnr. The results show that the PPA is the most sensitive binder to the filler used. This is not surprising as chemically modified binders are expected to be more chemically reactive to mineral components. The percent recovery was also examined. The distribution of mastic percent recovery at 3.2kPa and 58°C is presented in Figure 2-19.

48

0%

10%

20%

30%

40%

50%

60%

70%

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

% R

ecov

ery

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-19. Distribution of Mastic Percent Recovery at 58°C As evident by the results, mastics produced with SBS modified binder exhibit much higher percent recoveries than other mastics. Additionally, the SBS modified mastics show the highest variability with filler type. The variability ranges from 18% to 62%, indicating filler type has a significant effect on polymer modified mastics. Mastic relative percent recovery was also examined as presented in Figure 2-20 below.

49

0.0

1.0

2.0

3.0

4.0

5.0

6.0

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

Rela

tive

% R

ecov

ery

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-20. Distribution of Relative Percent Recovery at 58°C Results indicated for all binders except PPA modified binders, addition of fillers increases percent recovery. However, with the exception of GS2, all PPA mastics exhibit a relative percent recovery less than one, indicating a decrease in elasticity with addition of fillers. Additionally, the two neat binders display little variation in relative percent recovery with filler type. The mastics with SBS modified binder show the most sensitivity to filler type with relative values ranging from 1.6 to 5.8. Similar to the relative viscosity, the quartile method was used for the selection of mastics to be used as part of the asphalt mixtures in mixture rutting testing. Table 2-10 shows the quartile limit values for the Jnr results. The final mastic selection is shown in Table 2-11. This selection was conducted to maximize the number of fillers selected and to balance the number of mixtures containing each binder without violating the limits of each quartile.

50

Table 2-10 Summary of Quartiles for the Relative Jnr Measurements Jnr at 58 C Minimum Value 0.05 1st Quartile limit 0.17 Median limit 0.21 3rd Quartile limit 0.26 Maximum Value 0.63 Mean 0.23

Table 2-11 Mastics Selection for Mixture Rutting Testing Based on RELATIVE Jnr

Category Mastic FILLER TYPE RV FM CaO MBV Jnr Relative Jnr

Low

R

eact

ivity

PDS2 Soft Dolomite 29.4 4.7 27.0 1.8 2.05 0.63 PLS2 Soft Limestone 35.4 3.7 46.3 3.9 1.45 0.45

PFAF1 Fly Ash Type F 30.10 4.29 5.10 0.39 0.40 0.29 FFAC2 Fly Ash Type C 26.2 3.9 23.1 0.0 1.97 0.26

Med

ium

Lo

w

Rea

ctiv

ity VLH1 Hard Limestone 32.2 5.6 43.1 0.6 2.13 0.26

VHL2 Hydrated Lime 38.1 4.1 32.2 1.0 2.06 0.25 SGRQ2 Siliceous Quartzite 36.5 6.3 1.0 0.7 0.63 0.22 FFS2 Blast Furnace Slag 49.1 4.3 45.1 0.0 1.64 0.22

Med

ium

H

igh

Rea

ctiv

ity SBH2 Hard Basalt 33.8 4.6 7.0 11.2 0.58 0.21

PCA2 Soft Calichies 45.0 5.1 40.0 10.2 0.65 0.20 FGH2 Hard Granite 38.8 3.8 2.8 31.6 1.35 0.18 VGS1 Soft Granite 40.2 3.8 4.6 14.5 1.38 0.17

Hig

h

Rea

ctiv

ity VGH1 Hard Granite 42.6 4.1 3.5 2.4 1.32 0.16

FDH1 Hard Dolomite 42.8 5.1 26.0 2.8 1.10 0.15 SAN1 Andesite 41.9 5.2 6.8 24.6 0.30 0.11 SGS2 Soft Granite 47.0 3.1 7.8 0.9 0.15 0.05

Selected Fillers

Max 49.1 6.3 46.3 31.6

Selected Fillers

Min 26.2 3.1 1.0 0.0 Total Fillers Max 49.1 6.3 50.3 31.6 Total Fillers Min 26.2 3.0 1.0 0.0

Flow number testing was used to determine mixture rutting resistance according to procedures recommended in the NCHRP 9-19 project report with target air voids in the mixture equal to 7%. According to NCHRP 9-19, the deviator stress level used in the flow number test should be within the range of 70 to 210kPa for the unconfined tests. In this study, the flow number test was conducted using a uniaxial compression load without confinement at 58°C. A loading stress level of 200kPa was selected to attain tertiary flow in a reasonable number of cycles for various mixtures with different binders, gradations, and mineral fillers. All flow number tests were conducted 58°C. The mixtures were tested with 0.1sec of creep loading at a stress level of 200kPa and 0.9sec of recovery per cycle. Flow number which represents the number of cycles of repeated creep at which the mixture enters a tertiary creep flow region under a haversine repetitive

51

loading was used as the indicator of rutting resistance. Thus, a high flow number is desirable. The distribution of mixture flow number is presented in Figure 2-21.

0

500

1,000

1,500

2,000

2,500

3,000

FDH

1

FFAC

2

FFS2

FGH2

PCA2

PDS2

PFAF

1

PLS2

SAN

1

SBH2

SGRQ

2

SGS2

VGH1

VGS1

VHL2

VLH

1

Flow

Num

ber

Mastics

Coarse

Fine

Flint Hills + PPA PG 70-22

Flint Hills + SBS PG 70-22

Valero PG 64-22

Flint HillsPG 64-22

Legend: 1st Letter: Binder 2nd Letter: Filler 3rd/4th Letter: F: Flint Hills A: Andesite FS: Furnace Slag H: HardV: Valero B: Basalt G: Granite S: SoftP: PPA C: Caliches HL: Hydrated Lime 4th/5th Letter:S: SBS D: Dolomite 1: Source 1

FA: Fly Ash 2: Source 2 Figure 2-21. Distribution of Mixture Flow Number

The results demonstrate that the binder and gradation effects are most significant. Given the same mastic, coarse mixtures exhibit higher flow numbers than fine. The results shown above indicate the mixtures containing SBS modified binder exhibit superior rutting resistance which is consistent with mastic results. The SBS mixes with coarse gradation and AN1 and GRQ2 exhibit extreme behavior indicating a unique filler effect. Mastics with SBS and these two fillers do not have unusually low Jnr values, indicating the addition of the coarse aggregate leads to the extreme behavior. PPA modification also increases flow number somewhat but not to the extent SBS modification does. A statistical analysis was conducted to further support graphical observations. An ANOV A analysis was conducted to determine the importance of gradation and mastic on mixture flow number. The results shown in Table 2-12 reveal both gradation and mastic are very important as evident by their low P-values. The gradation F-value is higher than the mastic F-value, indicating aggregate gradation is more significant than mastic.

Table 2-12. Flow Number ANOVA: Gradation and Mastic

Variable F-Value P-Value Gradation 25.45 0.000

Mastic 13.58 0.000 Model R2 = 77.19%

Since mastic type is identified as statistically significant, another ANOV A analysis was conducted using aggregate gradation, and mastic components, namely, binder source,

52

binder modification and filler mineralogy as variables. Results indicated aggregate gradation and binder modification are most important. Filler mineralogy is also important as evident by the low P-value. Binder source, however, is insignificant. It should be noted that the model R2 value is about 74% which indicates there could be other variables not included in the model. Other possible variables could include filler geometry and/or filler binder interactions but they were not included in this first step ANOV A due to limitation of number of observations. It is clear though that filler mineralogy has an important effect as shown by the highest value of the F- statistics.

Table 2-13. Flow Number ANOVA: Gradation and Mastic Components Variable F-Value P-Value Gradation 21.9 0.000 Binder Source 0.09 0.765 Binder Modification 22.63 0.000 Filler Mineralogy 2.25 0.038

R2 = 73.5%

2.5.3 Effect of Fillers on Mastic and Mixture Fatigue Resistance Both the Superpave standard fatigue parameter G*·sinδ and fatigue life as determined from a time sweep were used to evaluate mastic and binder fatigue resistance. All fatigue tests were conducted at 25°C using the standard DSR testing procedure. The distribution of mastic G*·sinδ at 25°C is presented in Figure 2-22.

53

0

500

1,000

1,500

2,000

2,500

3,000

3,500

4,000

4,500

5,000

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

G*·

sinδ

(kPa

)

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-22. Distribution of Mastic G*·sinδ

As evident by the results, there is considerable variability in mastic G*·sinδ with both filler and binder type. However, there appears to be no consistent pattern in the effect of filler type on G*·sinδ as the values among different binders do not follow any pattern. Additionally, it is difficult to distinguish between mastics containing different binders. The mastic containing Flint Hills and GRQ2 has an exceptionally high G*·sinδ, indicating a unique filler-binder interaction. Additionally, Valero combined with CA2, FAF1, and FS2 also produces relatively high G*·sinδ values. To evaluate the impact of adding fillers to binders, relative G*·sinδ, (mastic G*·sinδ divided by binder G*·sinδ), was examined. The distribution of relative G*·sinδ is displayed below.

54

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

Rel

ativ

e G

*·si

nδ

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-23. Distribution of Relative G*·sinδ

All mastics have relative G*·sinδ values above one, indicating addition of fillers increases G*·sinδ, which is intuitive due to the stiffening effect of fillers. There are fluctuations in relative G*·sinδ with filler type. For most filler types, the mastics containing modified binders (SBS and PPA) exhibit higher relative G*·sinδ values than the neat binders. The results show a distinct ranking for the results based on binder. However, the fluctuations in the relative G*·sinδ with filler type create a few cases of overlap in the ranking among binders. Fatigue life was also defined as the number of cycles corresponding to a 50% reduction in initial complex modulus in a strain-controlled time sweep test for the mastic. The SBS modified mastics were too stiff to test at the same strain level as the other mastics. Consequently, the SBS modified mastics were tested at a lower strain level than the others. Hence, only relative fatigue life (i.e. mastic fatigue life divided by binder fatigue life) values are presented in Figure 2-24.

55

0.0

0.5

1.0

1.5

2.0

2.5

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

Rela

tive

Fati

gue

Life

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-24. Distribution of Mastic Relative Fatigue Life

Results above indicate addition of filler to binder has potential to both increase and decrease fatigue life. The mastics with SBS have relative fatigue lives significantly higher than the other mastics with other binders. Also, with the exception of DH1, addition of all fillers increases the fatigue life of the SBS binder. The mastics with the PPA modified binder exhibit little change in relative fatigue life with filler type, indicating PPA modification minimizes the effect of different filler mineralogies on fatigue life. Also, for all mastics with PPA modified binder, the relative fatigue lives are well below one, indicating a decrease in fatigue resistance upon addition of filler to the PPA modified binder. This observation is opposite to the trends of the MSCR results at higher temperature. For mastics containing neat binders similar relative fatigue lives can be seen, implying binder source does not have a significant effect on mastic fatigue performance. There are however some exceptions such as filler GH1 and GRQ2 which show very different interaction with the two neat binders. The relative fatigue lives were used to determine the levels of the reactivity of the fillers and to select the mastics for mixture fatigue testing. Table 2-14 includes the levels determined for the relative fatigue life. As shown the values range from a minimum of

*SBS was tested at a lower strain than others due to high stiffness

56

0.09 to a maximum of 2.10. This wide range is mainly resulting from the significant change in fatigue life of the binder modified with PPA after mixing with fillers. It appears that PPA modification significantly influences the base binder as indicated by the very large increase in cycles to failure. However this increase is lost when the binder is mixed with the fillers. The maximum value of the relative fatigue life for the PPA mastics is 0.41, which is a 59% reduction in fatigue life.

Table 2-14 Quartile Distribution of the Relative Values for the Fatigue Life and G*sinδ

Fatigue Life Relative Value

G*SinD Relative Value

Mean 0.77 Mean 2.54 Minimum 0.09 Minimum 1.41 1st Quartile 0.29 1st Quartile 2.06 Median 0.74 Median 2.36 3rd Quartile 1.06 3rd Quartile 3.10

Maximum 2.10 Maximum 4.22 Table 2-15 includes the mastics selected for the mixture testing based on the limits of the quartiles shown in Table 2-14. The table also lists the filler mineralogy of the fillers included in the selection as well as the manufactured filler types.

Table 2-15 Selection of mastics Based on Fatigue Life

Category Mastic BINDER FILLER TYPE RV FM CaO MBV Fatigue

Life

Relative Fatigue

Life G*Sinδ Relative

G*Sinδ

Low

R

eact

ivity

VGH1 VAL GH1 42.6 4.06 3.5 2.4 4146 0.09 2436 1.64

VHL2 VAL HL2 38.1 4.15 32.2 1.0 9693 0.21 3036 2.04

PDS2 PPA DS2 29.4 4.73 27.0 1.8 42180 0.25 2227 2.29

PGH2 PPA GH2 38.8 3.76 2.8 31.6 46360 0.27 2761 2.84

Med

ium

Low

R

eact

ivity

PCA2 PPA CA2 45.0 5.13 40.0 10.2 48998 0.29 3083 3.2

PGS2 PPA GS2 47.0 3.13 7.8 0.9 69919 0.41 2933 3.01

VLS2 VAL LS2 35.4 3.68 46.3 3.9 25764 0.56 2988 2.01

FBH1 FH BH1 33.2 5.31 8.2 5.9 33793 0.69 2296 1.63

Med

ium

Hig

h R

eact

ivity

FFS2 FH FS2 49.1 4.32 45.1 0.0 36423 0.74 1981 1.41

FFAC2 FH FAC2 26.2 3.88 23.1 0.0 40526 0.82 2165 1.54

SDH1 SBS DH1 42.8 5.07 26.0 2.8 10448 0.85 3139 3.59

SGRQ2 SBS GRQ2 36.5 6.32 1.0 0.7 12627 1.02 2701 3.09

Hig

h

Rea

ctiv

ity VGS1 VAL GS1 40.2 3.85 4.6 14.5 49243 1.06 2689 1.81

FGH1 FH GH1 42.6 4.06 3.5 2.4 58140 1.18 3101 2.20

SGH1 SBS GH1 42.6 4.06 3.5 2.4 25428 2.06 2931 3.35

SBH2 SBS BH2 33.8 4.65 7.0 11.2 25929 2.10 2923 3.34

Selected Fillers Max 49.1 6.3 46.3 31.6

Selected Fillers Min 26.2 3.1 1.0 0.0

Total Fillers Max 49.1 6.3 50.3 31.6

Total Fillers Min 26.2 3.0 1.0 0.0

57

The fatigue testing of mixtures was conducted using indirect tensile loading. Testing consisted of stress controlled time sweeps according to a procedure proposed by Kim et al in 2004 (31). All tests were conducted at 25°C with an oscillatory load of 3.25kN and 10 Hz frequency. The fatigue life (NF45) was determined as the number of cycles corresponding to a 45% reduction in maximum E*. The maximum E* value measured during the time sweep were also reported. It should be noted that binders and were tested using strain controlled time sweeps while mixtures were tested using stress controlled time sweeps. The distribution of mixture fatigue life is shown in Figure 2-25.

0

5,000

10,000

15,000

20,000

25,000

FBH

1

FFAC

2

FFS2

FGH

1

PCA2

PDS2

PGH

2

PGS2

SBH

2

SDH

1

SGH

1

SGRQ

2

VGH

1

VGS1

VHL2

VLS2

Nf4

5

Mastics

CoarseFine

Flint Hills + PPA

PG 70-22

Flint Hills + SBS PG 70-22

Valero PG 64-22

Flint HillsPG 64-22

Legend: 1st Letter: Binder 2nd Letter: Filler 3rd/4th Letter: F: Flint Hills A: Andesite FS: Furnace Slag H: HardV: Valero B: Basalt G: Granite S: SoftP: PPA C: Caliches HL: Hydrated Lime 4th/5th Letter:S: SBS D: Dolomite 1: Source 1

FA: Fly Ash 2: Source 2 Figure 2-25. Distribution of Mixture Fatigue Life

Mixture fatigue results indicate there is some variability in fatigue life with filler type. However, it is clear that binder type and gradation have more significant effects on fatigue life. For the same mastic, all fine gradations exhibit longer fatigue lives than coarse mixes. Additionally, the mixtures with modified binders generally have higher fatigue lives than the neat binders. Based on graphical results, there does not appear to be significant differences in fatigue lives of mixtures containing the two neat binders, indicating binder source is insignificant. The coarse mixture with PPA and GH2 exhibits extreme behavior, with a fatigue life significantly higher than the others. This suggests a unique interaction between PPA and GH2 when the coarse gradation is used. It is speculated the reason this is not observed for the fine mix because the fine particles in the fine gradation mitigate the effects of the GH2. It is important to note that the fatigue life measured for the mastics paint a different picture for the PPA mastics. An ANOV A analysis was conducted to determine the relative importance of gradation and mastic on mixture fatigue life. Results displayed in Table 2-16 indicate both

58

aggregate gradation and mastic composition are significant, but aggregate gradation is far more important as evident by the high F-value compared to that of the mastic.

Table 2-16. Mixture Fatigue Life (NF45) ANOVA: Gradation and Mastic Variable F-Value P-Value Gradation 41.67 0.000 Mastic 5.54 0.000

R2 = 66.82%

Another ANOV A analysis was used to evaluate the relative importance of gradation, filler mineralogy, binder source, and binder modification. The results displayed in Table 2-17 confirm graphical findings that gradation and binder modification are the most important factors determining fatigue life. Filler mineralogy is still showing a statistically significant influence on the measured fatigue life as indicated by the P-value in the ANOV A results. It is clear, however, that the gradation has the most important influence and that the R2 of the model is relatively low ( 61.17%). Other factors such as filler geometric could potentially improve correlation if included.

Table 2-17. Fatigue Life ANOVA: Gradation and Mastic Components Variable F-Value P-Value Gradation 37.73 0.000 Binder Source 2.99 0.090 Binder Modification 9.99 0.000 Filler Mineralogy 2.05 0.067

R2 = 61.17% The maximum E* observed in the fatigue tests was also examined as maximum E* that can be incorporated into the MEPDG analysis when designing pavements. The distribution of maximum E* is presented in Figure 2-26. Maximum E* results indicate little difference due to aggregate gradation and binder type. There is considerable variation in maximum E* in the neat Flint Hills binder. Combination of Flint Hills and FS2 with the coarse gradation leads to an E* considerably lower than the fine mixture. An analysis of variance (ANOV A) was conducted to determine if gradation and mastic are important factors on maximum E*. Results are shown in Table 2-18.

59

0

500

1,000

1,500

2,000

2,500

3,000

3,500

4,000

FBH

1

FFA

C2

FFS2

FGH

1

PCA

2

PDS2

PGH

2

PGS2

SBH

2

SDH

1

SGH

1

SGRQ

2

VG

H1

VG

S1

VHL2

VLS2

Max

imum

E*

(MPa

)

Mastics

CoarseFine

Flint Hills + PPA

PG 70-22

Flint Hills + SBS PG 70-22

Valero PG 64-22

Flint HillsPG 64-22

Legend: 1st Letter: Binder 2nd Letter: Filler 3rd/4th Letter: F: Flint Hills A: Andesite FS: Furnace Slag H: HardV: Valero B: Basalt G: Granite S: SoftP: PPA C: Caliches HL: Hydrated Lime 4th/5th Letter:S: SBS D: Dolomite 1: Source 1

FA: Fly Ash 2: Source 2 Figure 2-26. Distribution of Mixture Maximum E* at 25°C and 10 Hz Frequency

Table 2-18. Maximum E* ANOVA: Gradation and Mastic

Variable F-Value P-Value Gradation 4.01 0.052 Mastic 1.61 0.104

R2 = 17.42% While the P-value for gradation is below 0.1, neither gradation or mastic is considered significant due to the low R2 value, which indicates gradation and mastic cannot account for the variability in maximum E*. An additional ANOV A was used to determine if filler mineralogy, binder source, or gradation are important. The extremely low R2 of 11.59 indicates the parameters are incapable of accounting for the variability in maximum E*. Thus, it is concluded no parameters are significant, as shown in Table 2-19. Based on statistically analyses, it seems unreasonable to use maximum E* as an indicator of analysis of filler effect on fatigue resistance.

Table 2-19. Maximum E* ANOVA: Gradation and Mastic Components Variable F-Value P-Value Gradation 3.30 0.075 Filler Mineralogy 1.54 0.178 Binder Source 6.83 0.012 Binder Modification

2.88 0.136

R2 = 11.59%

60

2.5.4 Effects of Fillers on Mastic and Mixture Low Temperature Performance

The binders and mastics were evaluated for their low temperature performance using the standard Superpave Bending Beam Rheometer (BBR) testing procedure at -12 ºC. Additionally, the Asphalt Binder Cracking Device (ABCD) was used to determine the cracking temperature. Results reported for the BBR test included the stiffness and m-value at 60sec. The distribution of mastic low temperature stiffness is shown in Figure 2-27.

0

100

200

300

400

500

600

700

800

900

1,000

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

Stiff

ness

at

-12°

C (M

Pa)

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-27. Distribution of Mastic Low Temperature Stiffness at -12°C

As evident by the results, both binder and filler type affect stiffness of mastics. The mastics with the two neat binders generally exhibit higher stiffness than the modified mastics. Filler GS2 produces a very high stiffness when combined with the two neat binders but not with the two modified binders. Similarly, FAC2 when combined with the two neat binders produces a low stiffness but not when combined with the modified binders. This indicates the binder modification mitigates the stiffening effect of some fillers at low temperatures. The distribution of relative stiffness is displayed in Figure 2-

61

28. Relative stiffness (mastic stiffness divided by binder stiffness) is used to determine the effect of adding fillers to the different binders.

0.0

1.0

2.0

3.0

4.0

5.0

6.0

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

Rel

ativ

e St

iffne

ss a

t -1

2°C

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-28. Distribution of Relative Stiffness at -12°C

For all mastics, stiffness is more than double the binder stiffness. It is difficult to distinguish between binder types on the basis of relative stiffness, indicating the effect of filler type is greater than the effect of binder source and modification. The distribution of mastic m-values is presented in Figure 2-29.

62

0.20

0.22

0.24

0.26

0.28

0.30

0.32

0.34

0.36

0.38

0.40

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

m-v

alue

at

-12

°C

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-29. Distribution of Mastic m-value at -12°C

There is less variability in mastic m-value with filler type than with stiffness. Mastics containing the Valero binder generally exhibit lower m-value than the other mastics, indicating binder source affects mastic m-value. There is moderate variability in m-value with filler type for all binders. The distribution of relative m-values is displayed in Figure 2-30.

63

0.0

0.2

0.4

0.6

0.8

1.0

1.2

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

Rela

tive

m-v

alue

at

-12°

C

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-30. Distribution of Relative m-value at -12°C

Results indicate addition of fillers has potential to both increase and decrease m-value. The mastics with PPA clearly show higher relative m-values than mastics with other binders. Furthermore, the PPA mastics have relative m-values close to one for all fillers indicating PPA mitigates the loss of stress relaxation that occurs upon addition of fillers to binders. Figure 2-31 shows the distribution of mastic crack temperatures as determined by the Asphalt Binder Cracking Device, (ABCD). The ABCD is a new procedure and its relation to field performance is unknown. The results show that the ranking of mastics depends more on the filler type than on the binder stiffness. For example, PPA modified binder when combined with GS2 leads to a very low crack temperature compared to other mastics. However, PPA + GS2 did not exhibit outlier behavior in the BBR test. To evaluate the change between mastic and binder crack temperatures, the mastic crack temperature minus the binder crack temperature (∆ crack temperature) was used. The distribution of ∆ crack temperature is shown in Figure 2-32.

64

-45

-40

-35

-30

-25

-20

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

Crac

k Te

mpe

ratu

re (°

C)

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-31. Distribution of Mastic Crack Temperature

65

-12.0

-10.0

-8.0

-6.0

-4.0

-2.0

0.0

2.0

4.0

6.0

8.0

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

∆Cra

ck T

empe

ratu

re (°

C)

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-32. Distribution of ∆Crack Temperature

In some cases mastics exhibited lower cracking temperatures than the binder they contain and in some cases addition of filler increases cracking temperature. All mastics with Valero irrespective of filler type have a lower crack temperature than the Valero binder. Since the other binders fluctuate between positive and negative ∆ crack temperatures, it appears binder source is significant. Based on the concept that relative stiffness has a major influence, it was used to select mastics for mixture testing. The same procedure was followed as in the case of mastic viscosity. The quartiles for the relative stiffness results are shown in the following table.

66

Table 2-20 Quartiles for the mastics relative stiffness using BBR BBR Relative

Stiffness Mean 4.39 Minimum 2.93 1st Quartile 3.99 Median 4.32 3rd Quartile 4.76 Maximum 5.63

The selected mastics are listed in Table 2-21. The selection process followed the same requirements as indicated before. The corresponding values for the filler properties and the mastic low temperature testing results are also included.

Table 2-21 Selected Mastics for Mixture Low Temperature Performance Based on the Relative

Stiffness from the BBR

Category Mastic FILLER TYPE RV FM CaO MBV Stiffness

(mPa) Relative Stiffness

m-Value

Relative m-

Value

Cracking Temp. (°C)

Δ Cracking

Temp. (°C)

Low

R

eact

ivity

FFAC2 FAC2 26.2 3.9 23.1 0.0 483 2.93 0.29 0.84 -30.55 -2.70 VLH1 LH1 32.2 5.6 43.1 0.6 600 3.52 0.26 0.84 -32.40 -5.05 VCA2 CA2 45.0 5.1 40.0 10.2 638 3.86 0.27 0.87 -30.65 -3.30

PGRQ2 GRQ2 36.5 6.3 1.0 0.7 483 3.94 0.32 1.06 -30.10 -0.10

Med

ium

L

ow

Rea

ctiv

ity FGS1 GS1 40.2 3.8 4.6 14.5 670 4.06 0.29 0.84 -29.40 -1.55

SHL2 HL2 38.1 4.1 32.2 1.0 503 4.16 0.30 0.87 -34.30 -3.10 VBH2 BH2 33.8 4.6 7.0 11.2 717 4.21 0.26 0.85 -33.05 -5.70 FGH1 GH1 42.6 4.1 3.5 2.4 700. 4.2 0.30 0.87 -23.6 4.25

Med

ium

H

igh

Rea

ctiv

ity SDS2 DS2 29.4 4.7 27.0 1.8 541 4.47 0.31 0.89 -34.55 -3.35

PLS2 LS2 35.4 3.7 46.3 3.9 568 4.63 0.30 1.00 -27.60 2.40 PGH2 GH2 38.8 3.8 2.8 31.6 576 4.70 0.29 0.97 -32.15 -2.15 FGS2 GS2 47.0 3.1 7.8 0.9 783 4.74 0.27 0.79 -31.10 -3.25

Hig

h

Rea

ctiv

ity SGH1 GH1 42.6 4.1 3.5 2.4 605 5.00 0.30 0.85 -27.80 3.40

PFS2 FS2 49.1 4.3 45.1 0.0 621 5.07 0.30 0.99 -32.60 -2.60 VGS2 GS2 47.0 3.1 7.8 0.9 883 5.18 0.24 0.78 -32.55 -5.20 SAN1 AN1 41.9 5.2 6.8 24.6 682 5.63 0.28 0.79 -26.20 5.00

Selected Fillers Max 49.1 6.3 46.3 31.6 Selected Fillers Min 26.2 3.1 1.0 0.0 Total Fillers Max 49.1 6.3 50.3 31.6 Total Fillers Min 26.2 3.0 1.0 0.0

The mixtures were tested for low temperature creep to determine their low temperature stiffness at -12ºC. Additionally, mixtures were tested for indirect tensile strength. In this test, the specimen is exposed to indirect tensile loading at a deformation rate of 0.25mm/sec until failure. The fracture strength is then calculated. Results of mixture low temperature stiffness are displayed in Figure 2-33.

67

0

5

10

15

20

25

FFAC

2

FGH

1

FGS1

FGS2

PFS2

PGH

2

PGRQ

2

PLS2

SAN

1

SDS2

SGH

1

SHL2

VBH

2

VCA2

VGS2

VLH

1

Stiff

ness

(Gpa

)

Mastics

CoarseFine

Flint Hills + PPA

PG 70-22

Flint Hills + SBS PG 70-22

Valero PG 64-22

Flint HillsPG 64-22

Legend: 1st Letter: Binder 2nd Letter: Filler 3rd/4th Letter: F: Flint Hills A: Andesite FS: Furnace Slag H: HardV: Valero B: Basalt G: Granite S: SoftP: PPA C: Caliches HL: Hydrated Lime 4th/5th Letter:S: SBS D: Dolomite 1: Source 1

FA: Fly Ash 2: Source 2 Figure 2-33. Distribution of Mixture Stiffness at -12°C

The results indicate low temperature stiffness is not significantly affected by filler or binder type for the Flint Hills neat, SBS modified, and PPA modified binders while the Valero shows considerable variability. Gradation also appears to have an effect, particularly for mixes containing the SBS and V alero binders. Thus, it appears binder source is a significant factor affecting mixture low temperature stiffness. The variability in mixture low temperature stiffness results do not reflect the high variability in mastic low temperature stiffness. An analysis of variance (ANOV A) was conducted to determine the importance of gradation and mastic type on low temperature stiffness. Results are displayed in Table 2-22.

68

Table 2-22. Low Temperature Stiffness of Mixtures ANOVA: Gradation and Mastic Variable F-Value P-Value Gradation 52.15 0.000 Mastic 3.92 0.000

R2= 60.11%

The low P-Value in ANOV A shows the mixture stiffness at -12ºC is significantly influenced by the aggregate gradation and the mastic type. However, , the variation in the aggregate gradation has more impact on the mixture stiffness than the mastic as suggested by the F-values. Additionally, an ANOVA was conducted to determine the relative importance of gradation, binder source, binder modification and filler mineralogy. The analysis shows that mixture aggregate gradation, and binder source are the only significant factors influencing the mixture stiffness as indicated by the low P-value.

Table 2-23. Low Temperature Stiffness ANOVA: Gradation and Mastic Components Variable F-Value P-Value Gradation 55.21 0.000 Binder Source 17.34 0.000 Binder Modification 0.35 0.708 Filler Mineralogy 0.37 0.946

R2 = 62.31% These results suggest that the interaction of the binder type (source) with mixture aggregate gradation is a significant parameter in determining the mixture stiffness, while binder modification is insignificant. It appears that some fillers may negate or even reverse the effect of the binder modification as observed for the PPA binder with soft limestone. The distribution of mixture low temperature strength at -12°C as measured in the fracture test is presented in Figure 2-34.

69

2.0

2.5

3.0

3.5

4.0

4.5

5.0

5.5

FFAC

2

FGH

1

FGS1

FGS2

PFS2

PGH

2

PGRQ

2

PLS2

SAN

1

SDS2

SGH

1

SHL2

VBH

2

VCA2

VGS2

VLH

1

Stre

ngth

(Mpa

)

Mastics

CoarseFine

Flint Hills + PPA

PG 70-22

Flint Hills + SBS PG 70-22

Valero PG 64-22

Flint HillsPG 64-22

Legend: 1st Letter: Binder 2nd Letter: Filler 3rd/4th Letter: F: Flint Hills A: Andesite FS: Furnace Slag H: HardV: Valero B: Basalt G: Granite S: SoftP: PPA C: Caliches HL: Hydrated Lime 4th/5th Letter:S: SBS D: Dolomite 1: Source 1

FA: Fly Ash 2: Source 2 Figure 2-34. Distribution of Mixture Strength at -12°C

Graphical results indicate substantial variability in low temperature indirect tensile strength among the different mastics. It is interesting that while gradation leads to considerable variability in results, there is not a clear trend with gradation. For some mastics, coarse gradations resulted in higher indirect tensile strength while in others fine gradation is higher. Similar to the analysis of the mixture stiffness at -12 ºC, the mixture fracture strength results were analyzed to evaluate whether the aggregate structure or the mastic are more influential on the mixture low temperature performance. The following Table (2-24) shows the results of the ANOV A.

Table 2-24. Low Temperature Strength ANOVA: Gradation and Mastic Variable F-Value P-Value Gradation 6.13 0.017

Mastic 4.30 0.000 R2 = 46.43%

The ANOV A shows that aggregate gradation and mastic have significant effects on the mixture performance. The mastic is more important than the gradation based on P-Value. The analysis to evaluate the role of each component of the mastic is presented next using gradation, binder source, binder modification, and filler mineralogy.

70

Table 2-25. Low Temperature Strength ANOVA: Gradation and Mastic Components Variable F-Value P-Value Gradation 5.84 0.019 Binder Source 1.38 0.246 Binder Modification 5.08 0.010 Filler Mineralogy 1.96 0.065

R2 = 39.80% The results of the ANOV A in Table 2-25 show that binder modification is the most significant factor as indicated by the P-Value. On the other hand the binder source is the least significant factor, which was not expected as binder source had a significant effect on low temperature stiffness In summary, the initial correlations of low temperature performance indicators show that mastics are important for mixture strength and stiffness; but the importance is more related to the binders than the fillers used in the mastics. This will be further discussed in Chapter 3.

2.5.5 Effects of Fillers on Mastic and Mixture Moisture Damage Mastics were evaluated with respect to their resistance to moisture damage. Two tests were used to evaluate the mastic performance. The first is a water susceptibility test using the European Norm test (EN 1744-4). In this test the mastics were diluted with kerosene to a viscosity of 240 ± 10st. (1stoke = 0.0001m2·s−1). The samples were heated in a hot water bath with an insulating barrier to prevent heat loss and then shaken on an orbital shaker at 500rpm for the amount of time specified in the method. A visual inspection of the material was used to determine whether or not it needed to be filtered. If the water was clear, then the material was determined to be insusceptible to water. If the water was not clear then the sample was filtered and the retained filtrate was dried and weighed. The percent loss after filtration is the reported result (37). Figure 2-35 shows the summary and distribution of the mastics results.

71

Results of Water Susceptibility Test for Mastics

0.0%

5.0%

10.0%

15.0%

20.0%

25.0%

30.0%

AN

1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS2

HL2

Perc

ent M

ass L

oss

Flint Hills PPA SBS Valero Average

Natural Manufactured

Figure 2-35 Results of the Moisture Susceptibility

The results show that only the Valero (low asphaltene) mastics are susceptible to this test. On the other hand mastics with the other three binders show no loss in mass using this procedure. The results of this test were not utilized in ranking the mastics for mixture testing selection. The lack of variability in results of all mastics containing the base binder Flint Hills makes task of distinguishing between the different mastics performance unattainable. The second moisture damage test utilized was the bitumen bond strength test. The test is conducted using a pneumatic asphalt tension testing instrument (PATTI). In this test, a metal stub is adhered to a granite slab surface using a mastic specimen with a film thickness of 2.50mm thickness. The stubs are placed on the asphalt film prior to conditioning. The pull off strength (pressure) is measured for dry samples and for samples submerged in water at 60°C for 24 hours. The water conditioned samples were then removed from the water and conditioned at 25°C for one hour prior to testing. The reported results are the moisture damage ratios (MDR). The MDR is measured by dividing the pull off pressure of moisture-conditioned specimens by that of the dry ones. There appears to be a relatively wide range of pull off strength values. Figure 2-36 provides the distribution of MDR.

72

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

MD

R

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-36 Distribution of Moisture Damage Ratio (MDR)

Results shown in Figure 2-35 shows the neat binders, (Valero and Flint Hills), show the most sensitivity to filler type. It is also interesting that depending on filler type, MDR can be greater than or less than one, indicating for some mastics, moisture conditioning improves pull-off strength, which is not logical. This phenomenon was also observed in the un-filled binders. The increase in pull-off strength with water conditioning could be a results of the long conditioning time and high temperature (60°C). However, future research is needed to fully explain the phenomenon. Results show variability in MDR with filler type, but no clear trend exists. The ratio of the mastic wet pull-off strength divided by the binder wet pull-off strength, (i.e. relative wet pull-off strength), was also examined. The distribution of relative wet pull-off strength is displayed in Figure 2-37.

73

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

No

Fille

r

AN1

BH1

BH2

CA2

DH

1

DS2

GH

1

GH

2

GRQ

2

GS1

GS2

LH1

LS2

FAC2

FAF1 FS

2

HL2

Rela

tive

Wet

Pul

l-of

f St

reng

th

Filler

Flint Hills

PPA

SBS

Valero

Average

Legend: 1st Letter: Filler 2nd / 3rd Letter: A:Andesite FS: Furnace Slag H: HardB: Basalt G: Granite S: SoftC: Caliches GRQ: Gravel Quartzite 4th/5th Letter:D: Dolomite HL: Hydrated Lime 1: Source 1FA: Fly Ash 2: Source 2

Natural Manufactured

Figure 2-37. Distribution of Relative Wet Pull-off Strength

The distribution of relative wet pull-off strength reveals considerable variability with filler type. However, no clear trend with filler type is apparent and the clearest distinction is with binder type. The PPA modified mastics have relative wet pull-off strengths near one for all fillers, indicating little change in wet pull-off strength with addition of fillers. The majority of SBS modified mastics have relative wet pull-off strength ratios above one, indicating the addition of filler to SBS modified binder increases the wet pull-off strength. On the other hand, the majority of Valero mastics exhibit relative wet strengths below one, indicating the mastics with Valero have lower wet pull-off strengths than the binder itself. Mixtures were tested for moisture damage on a limited scale compared to mixture testing for other performance indicators because of lack of string evidence that fillers are as important as binder used in this study. The limited study was conducted using the Flint Hills modified with SBS and neat Valero binders in the mixtures blended with two mastics having high MDRs and two mastics with low MDRs. The mastics were used to produce mixes with coarse and fine gradations. This limited study was conducted to investigate if the fillers show significant effect on the measured moisture damage

74

resistance. Moisture damage of mixture was conducted using wet Hamburg Wheel Tracking Device (HWTD) conducted for 20,000 cycles of loading. The results confirmed the mastic performance in terms of binder dependency. None of the SBS mixtures displayed any stripping. For mixture containing the Valero binder, only fine-graded mixes with soft limestone did not show signs of stripping. Table 2-26 shows test results

Table 2-26 Results of Limited Study on Mixture Moisture Damage Resistance

Binder Mastic Rut at 10k

Cycles (mm) Onset of Stripping

Coar

se

SBS SCA2 2.05 NA

SBS SGH1 1.87 NA

VAL VGS1 3.91 17613

VAL VLS2 4.92 14375

Fine

SBS SCA2 2.17 NA

SBS SGH1 2.16 NA

VAL VGS1 5.63 10915

VAL VLS2 4.83 NA

The graphical presentation of the data in Figure 2-38 shows that the level of damage in the mixtures is dependent on the binder type. The rutting depth after 10,000 passes on the moisture conditioned mixture specimens is used as an indicator for the moisture induced damage.

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

SCA2 SGH1 VGS1 VLS2

Rut

Dep

th a

t 10k

Cyc

les

(mm

)

Coarse Fine

Figure 2-38 Results of the Moisture Related Damage of Mixes Using Limited Mastics

75

The rut depth for the SBS mixtures is constant regardless of the gradation type or the filler type. On the other hand, the neat binder shows some gradation dependence when blended with soft granite (GS1). The influence of the gradation disappears for the soft limestone mix (LS2). An analysis of variance (ANOV A) was conducted to investigate the role of fillers on the mixture moisture induced stability. Two ANOV A analyses were conducted. The first included only gradation and mastic as variables. The second included gradation and mastic components: binder and filler mineralogy. The first ANOV A results are shown in Table 2-27. The results reveal that the gradation of the mixtures has no significant influence on the results. This is confirmed by comparing the results reported in Table 2-26.

Table 2-27 ANOVA Analysis for the Influence of Mixture Gradation Compared to Mastic Type Variable F-Value P-Value Gradation 1.05 0.328

Mastic 8.34 0.004 R2 = 59.52%

Further investigation into the role of the mastic components: binder type (SBS and Neat) and filler (three mineralogies) was conducted. The results are shown in Table 2-28. Table 2-28 Influence of Gradation and Mastic Components on Observed Moisture Damage Indicator

Variable F-Value P-Value Gradation 1.05 0.328 Binder 13.45 0.004 Filler Mineralogy 0.02 0.982

R2 = 59.52% The results show that the moisture resistance of the mixtures is highly dependent on the binder type used, supported by the high P-values for both gradation and filler mineralogy. The results suggest aborting continuation of moisture damage testing of mixtures due to lack of evidence that fillers can have an important influence as compared to the binder modification type using the HWTD.

76

3 INTERPRETATION, APPRAISAL, AND APPLICATION

3.1 Quantifying of Filler Influence on Mixture Performance Indicators

3.1.1 Strategy of Analysis The main objectives of this project are (1) to identify test methods for mineral fillers that can characterize their effects on mastics and mixtures, and (2) recommend specification criteria for fillers that optimize its effects on HMA performance. To achieve these objectives, an analysis approach for the results presented in Chapter 2 was followed to first find correlations between mastic and mixtures properties then use the correlations between mastic and filler properties to define important filler characteristics. If correlations between mastic and mixture properties are found, limits of mastic properties that can result in acceptable mixture performance are defined. These mastic limits are then translated into filler properties, if correlations between filler and mastic properties could be clearly defined. Therefore, the mastic properties were treated as the bridge between filler properties and mixture performance indicators. Only natural fillers were considered in development of correlations in this chapter and also in defining specification criteria when possible. Manufactured fillers are found to exhibit some extreme values in critical filler properties, such as specific gravity and Rigden voids, and to result in unique mastic and mixture properties. The flow chart shown in Figure 3-1 depicts the approach followed for specification criteria development.

77

Figure 3-1 Flow Chart Illustrating Analysis Procedure and Strategy for Accepting Fillers

As depicted in the flow chart, the recommendations are developed by correlating mixture and mastic results. If no relationship between mixture and mastic results exists, further analysis is aborted as the testing conducted in this study could not provide information to clearly determine which mastics property correlates with mixture performance. If a relationship between mastic and mixture results exists, mastic limits that result in acceptable mixture performance are defined. To relate the mastic limits to filler characteristics, a multi-linear regression model relating mastic results to binder and filler properties is used. If the regression model has a high prediction power (high R2 that is logical in trends), two options could be followed for acceptance of filler:

• Conduct mastic testing to quantify the mastic property value and compare to acceptable limit directly, or

• Perform binder and filler tests and use the regression equation to estimate mastic property value and compare to acceptable limit.

Statistical Correlation No Discontinue

Analysis

Measure mastic properties directly using recommended

tests

Use Regression Models to predict mastic properties based on filler and

binder properties

Yes

To accept fillers (based on mastic properties)

Define acceptance limits for Mastic Properties, and derive models for these properties in terms of filler and binder properties

Mastics Testing Results

Mixture Testing Results

Measure filler and binder properties

Check measured or estimate mastic properties with acceptance recommended limits

78

It is recommended that mastics be tested directly when possible, and the mastic regression model only be used in cases when mastic testing is impossible. For manufactured fillers (such as fly ash and cement) mastics will have to be tested as the regression models for estimating mastic properties do not apply for such fillers. Aggregate gradation was found to have a significant effect on many mixture performance indicators as discussed in Chapter 2. Hence, in limit development analysis, it was necessary to sort limits by gradation. For the purpose of this study, a coarse aggregate gradation is defined according to the Asphalt Institute’s definition as any aggregate gradation that crosses below 0.45 power maximum density line at or before sieve #8 (2.36mm). The fine gradation is that which passes above the sieve #8. The following sections give details of the analysis conducted to develop recommendations based on the mixture performance studied.

3.1.2 Workability An attempt to relate mastic and mixture workability was conducted using the correlation between mixture number of gyrations to 92% Gmm and mastic relative viscosity. Relative viscosity was used in place of mastic viscosity as the SBS modified binder produces extreme viscosities but not extreme mixture performance. Analyses were separated by gradation because gradation was found to have a very important effect on workability of mixtures as measured by number of gyrations to 92% Gmm. Results of the correlation are shown in Figure 3-2 for the fine and the coarse aggregate gradations.

y = 2.26x + 26.95R² = 0.40

y = 0.67x + 13.97R² = 0.20

0

10

20

30

40

50

60

0.0 2.0 4.0 6.0 8.0 10.0Number of Gyrations to 92% Gmm

Relative Viscosity

Coarse

Fine

\ One Standard Deviation

Figure 3-2 Correlation between Mixture and Mastic Workability Indicators

Figure 3-2 shows the expected trend that the mixtures are less workable as the relative viscosity of mastics increases. However, for the fine mixtures the increase in number of gyrations to reach 92% Gmm with increasing relative viscosity is minimal. The results also suggest that the coarse mixes are more sensitive (higher slope) to mastic viscosity

79

than fine mixes. As indicated in Chapter 2 ANOV A analysis, the gradation of aggregates in this study demonstrated more influence on workability than the other factors. An extensive literature search was conducted to determine what limits can be used to define satisfactory workability limits of mixtures. No such limits for the number of gyrations to 92%Gmm could be found in the literature. The most comprehensive comparison between laboratory measured and field measured compactability found was a study by Leiva and West (40). Leiva and West evaluated the ability of multiple laboratory measured compactability indicators to predict actual mixture compactability in the field. Laboratory measured compactability indicators included compaction energy index (CEI), number of gyrations to reach 92%Gmm, compaction slope, locking point, and Bailey Method ratios. The accumulated compaction pressure (ACP) was used to indicate compactability measured in the field. Authors concluded that a simple relationship between laboratory and field compactability does not exist. The laboratory measured number of gyrations to 92%Gmm, (used as the indicator of mixture compactabiliy in this study), correlated weakly (R2 of 0.40) with the ACP measured at post-construction density. Hence, the authors did not recommend limits on number of gyrations to reach 92%Gmm. Rather, they developed a multiple linear regression model to incorporate the multiple factors found to affect mixture compactability. Since no limits for number of gyrations to reach 92%Gmm exist in the literature, limits are based on the results of mixture workability testing conducted within the scope of this study by establishing the limit at the average number of gyrations to reach 92%Gmm measured for coarse mixture plus one standard deviation measured for all coarse mixtures.. Because all fine mixtures exhibit low values of number of gyrations to reach 92%Gmm, the results for the fine mixtures were excluded from the limits analysis. Using this strategy, the corresponding limit for number of gyrations to reach 92%Gmm is 43 gyrations. That is, mixture with number of gyrations to reach 92%Gmm less than 43 gyrations are deemed satisfactory. The corresponding mastic limit was determined based on the linear relationship between mixture number of gyrations to reach 92%Gmm and relative viscosity presented in Figure 3-2. To increase the reliability of the limit and account for variability in the results, the relative viscosity limit was developed based on the corresponding relative viscosity minus one standard deviation, rounded to the nearest 0.5. Using the trend line minus one standard deviation, the limiting relative viscosity corresponding to the proposed mixture limit is 5.0. A summary of the proposed limits is provided in Table 3-1.

Table 3-1 Proposed Workability Limits

Maximum N92 43

Maximum Relative Viscosity

5.0

80

In case mastics cannot be tested, a multiple linear regression analysis was conducted to develop a model relating mastic viscosity to binder and filler properties. Results of the best subsets analysis are displayed below. Best Subsets Regression: Mastic Viscosity as a function of Binder Viscosity, RV, CaO, and FM Response is Mastic Viscosity B i n d e r V C Mallows i R a F Vars R-Sq R-Sq(adj) Cp S s V O M 1 43.4 42.3 50.6 1630.4 X 1 25.0 23.5 82.6 1876.3 X 2 68.4 67.2 9.0 1229.7 X X 2 53.8 51.9 34.6 1488.7 X X 3 70.8 69.0 6.8 1194.8 X X X 3 70.8 69.0 6.9 1195.2 X X X 4 73.0 70.7 5.0 1161.1 X X X X The best model where all model parameters were found to be significant included Rigden voids and binder viscosity. The R2 for this model is 67.2% when using these parameters. The model for binder viscosity at 135°C is displayed below. Regression Analysis: Mastic Viscosity as a function of Binder Viscosity and RV The regression equation is Mastic Viscosity = - 8244 + 4.68 Binder Viscosity + 205 RV Predictor Coef SE Coef T P Constant -8244 1368 -6.03 0.000 Binder Viscosity 4.6842 0.5706 8.21 0.000 RV 204.83 32.85 6.24 0.000 S = 1229.72 R-Sq = 68.4% R-Sq(adj) = 67.2%

The regression analysis indicates that RV is the only filler property found to have a significant influence on mastic viscosity. In order to incorporate a filler in the mixture it is proposed to test the viscosity of the binder and mastic made with filler to determine relative viscosity to assure the measured relative viscosity is below the limit assigned if a coarse gradation is being used. Alternatively, if one is unable to measure mastic viscosity, binder viscosity and filler RV can be used in the above regression to predict the mastic viscosity. If the predicted relative viscosity is below the proposed limit, the mastic could be deemed acceptable. Otherwise, filler or binder adjustments could be made to achieve an acceptable relative viscosity. It should be noted due to the fairly weak correlation between mastic and mixture workability indicators and the existence of outliers, the proposed limits should be considered as preliminary guidelines.

81

3.1.3 Permanent Deformation The mixture flow number (FN) results were correlated with the measured mastic Jnr. Coarse and fine mixtures were separated as the ANOVA analysis conducted in Chapter 2 indicated that gradation has a significant effect on FN results. Additionally, coarse mixtures with SBS and AN1 or GRQ2 fillers were not included as they are exhibited extreme unrealistic performance beyond the observed and expected trends. Results of the correlation are displayed in Figures 3-3 and 3-4.

y = -1,208.73x + 1,271.92R² = 0.61

0

200

400

600

800

1000

1200

1400

1600

0.00 0.20 0.40 0.60 0.80 1.00 1.20

Mixture Flow Number

Mastic Jnr (1/kPa)

\ One Standard Deviation

Figure 3-3 Correlation between Mixture Flow Number and Mastic Jnr for Fine Gradation

Figure 3-4 Correlation between Mixture Flow Number and Mastic Jnr for Coarse Gradation

The plots include the average trend lines represented by the center lines and reliability represented as one standard deviation above and below the average trend lines. A clear

82

trend between mixture flow number and mastic Jnr exists for both gradations. Linear models were used to relate mixture flow number and mastic Jnr. It is interesting to observe that the correlation trends for the gradations show similar slope value with a shift factor separating the two gradations. The linear correlations have R2 values of 66% and 61% for coarse and fine mixtures, respectively. Additionally, mixture flow number was correlated to mastic percent recovery at 3.2kPa and 58°C. Results for fine and coarse mixtures are provided in Figures 3-5 and 3-6, respectively.

Figure 3-5 Correlation between Mixture Flow Number and Mastic Percent Recovery for Fine

Gradation

83

Figure 3-6 Correlation between Mixture Flow Number and Mastic Percent Recovery for Coarse

Gradation As evident by Figures 3-5 and 3-6 above, a relationship between mixture flow number and percent recovery exists, but it is not linear. It should be pointed out, however, that the relationship is driven by binder type rather than filler type. Hence, filler influence appears to be less significant than the binder influence, particularly when SBS modification is considered. A multiple linear regression analysis was conducted using filler properties and binder percent recovery as independent parameters to determine if filler properties significantly affect mastic percent recovery. The results of a best subset regression analysis are displayed below. Best Subsets Regression: Mastic % Rec versus Binder % Rec, RV, CaO, FM Response is Mastic % Recovery B i n d e r % C Mallows R R a F Vars R-Sq R-Sq(adj) Cp S e V O M 1 69.1 68.5 3.0 0.090852 X 1 1.4 0.0 114.9 0.16239 X 2 70.1 68.9 3.4 0.090338 X X 2 69.7 68.5 4.0 0.090899 X X 3 71.5 69.7 3.1 0.089085 X X X 3 70.1 68.3 5.3 0.091190 X X X 4 71.5 69.1 5.0 0.089969 X X X X As best subset regression results indicate, 68.5% of the variability in mastic percent recovery can be explained by binder percent recovery alone. Furthermore, regression

84

equations containing filler properties as parameters show filler properties are not significant model parameters. Hence, there is no need to develop a mastic percent recovery specification. To derive filler specification limits to ensure acceptable rutting performance of mixture, the definition of what is acceptable was taken from previous studies. In the NCHRP Project 9-43, minimum flow numbers are proposed based on flow number data collected on several field projects (42). These flow number limits, which are based on traffic level, are displayed in Table 3-2.

Table 3-2 Minimum Flow Number at 600 kPa (42) Traffic Level

(Million ESALs) Minimum Flow

Number @ 600kPa < 3 -----------

3 to < 10 53 10 to < 30 105

≥ 30 415 It is important to note that the results reported in NCHRP 9-43 are based on testing at 600kPa, which is three times higher than the stress level applied in this study. As discussed in Chapter 2, 200kPa was selected for this study based on NCHRP 9-43 recommendations for unconfined tests, which suggests use of stress levels between 70kPa and 210kPa. Due to the differences in stress levels, there is a need for a transfer function between stress levels. In the Airfield Asphalt Pavement Technology Program (AAPTP) Project 04-02: PG Binder Grade Selection for Airfield Pavements, the following equation relating the number of load repetitions and stress level was developed through the combination of equations to predict rutting used in the Mechanistic-Empirical Pavement Design Guide (MEPDG) and other mathematical relations used in the MEPDG (41):

Equation 3-1 Using the above equation, the flow number limits presented in Table 3-2 can be derived for a stress level of 200kPa to arrive at the minimum flow numbers in Table 3-3, which are applicable to the results of this study.

Table 3-3 Minimum Flow Numbers at 200 kPa

Traffic Level (Million ESALs)

Minimum Flow Number @ 200kPa

< 3 ----------- 3 to < 10 530

10 to < 30 1,050 ≥ 30 4,120

85

It should also be noted that in the NCHRP 9-43 project testing was conducted at the 50 percent reliability performance grade temperature determined using LTPP Bind at a depth of 20mm without traffic volume or speed adjustments. In this study all testing was conducted at 58°C irrespective of performance grade. Based on these limits, three (or 19%) of coarse mixtures and eight (or 50%) of fine mixtures do not pass the criterion for traffic levels between 3 million and 10 million ESALs. It is important to note that the mixtures were prepared using a mix design designated for use by the Wisconsin Department of Transportation for less than 10 million ESALs. Therefore, mastic limits were developed based on the criterion for traffic levels greater than or equal to 3 million and less than 10 million ESALs. Mastic Jnr limits were derived based on the average trend line minus one standard deviation for Jnr to increase the reliability of the limits and account for any variability in the results. Using the models relating mastic and mixture results for coarse and fine aggregate gradations, the resulting mastic limits are presented in Table 3-4.

Table 3-4 Maximum Values for Mastic Jnr at 3.2kPa Testing Separated by Gradation

Mixture Gradation

Maximum Mastic Jnr at 3.2kPa (1/kPa)

Fine 0.40 Coarse 0.55

The mastic limit for Jnr is lowest for fine aggregate gradations and thus most conservative. Therefore the recommended mastic limit, irrespective of gradation follows the fine aggregate gradation limit. To determine which filler properties have significant influence on mastics Jnr, a best subsets regression analysis was conducting using both binder Jnr and filler properties to predict mastic Jnr. Results of the best subsets analysis are presented below. Best Subsets Regression: Mastic Jnr versus Binder Jnr, RV, CaO, FM Response is Mastic Jnr B i n d e r J C Mallows n R a F Vars R-Sq R-Sq(adj) Cp S r V O M 1 50.3 49.3 49.6 0.17257 X 1 24.7 23.2 99.9 0.21243 X 2 74.9 73.9 3.2 0.12377 X X 2 54.2 52.3 44.0 0.16737 X X 3 75.9 74.4 3.4 0.12269 X X X 3 75.1 73.5 4.9 0.12466 X X X 4 76.1 74.0 5.0 0.12352 X X X X

86

The best model from the subsets analysis includes the binder Jnr and the Rigden Voids (RV) as displayed below. Regression Analysis: Mastic Jnr versus Binder Jnr, RV The regression equation is Mastic Jnr = 1.01 + 0.160 Binder Jnr - 0.0230 RV Predictor Coef SE Coef T P Constant 1.0120 0.1334 7.59 0.000 Binder Jnr 0.16002 0.01614 9.91 0.000 RV -0.022953 0.003306 -6.94 0.000 S = 0.123770 R-Sq = 74.9% R-Sq(adj) = 73.9% Regression results show the mastic Jnr model has An R2 of 73.9%, indicating these properties are sufficient to adequately predict mastic Jnr. Accordingly, in order to develop limits for filler properties to ensure acceptable mixture resistance to rutting, it is proposed that either mastic Jnr be measured, or the above regression model with the required filler and binder properties be used to predict the mastic Jnr. In either case the Jnr value should be lower than 0.40 (1/kPa).

3.1.4 Fatigue Mixture fatigue life (Nf45) and mastic G*sinδ were correlated as shown in Figure 3-7. Coarse and fine gradations were separated as gradation has a significant effect on mixture fatigue life as indicated by the ANOV A results in Chapter 2.

Figure 3-7 Correlation between Mixture Fatigue Life and Mastic G*sinδ

87

Results shown indicate no trend exists between mastic G*sinδ and mixture fatigue results. The ANOV A results reported in Chapter 2 clearly indicate that the mixture gradation, binder modification, filler mineralogy, and binder source are significant factors. However, when comparing the mixture performance to the mastic performance, no relationship can be defined. It is speculated the lack of relationship is more due to testing method used in characterizing the mastic and mixture fatigue performance. Therefore, although the ANOV A analysis indicates that the changes in the mixture performance are significantly associated with the variability in its components, the used tools to evaluate the mixtures and mastics could not find a relationship between the fatigue indicators. The mixture fatigue life (number of cycles corresponding to a 45% reduction in dynamic modulus) was also compared to mastic relative fatigue life. The comparison is shown in Figure 3-8.

Figure 3-8 Comparison of Mixture Fatigue indicator and Mastic Relative Fatigue Life The results in Figure 3-8 also show no relationship between mixture and mastic fatigue performance. The results in both figures confirm to some extent what was discussed in Chapter 2, regarding the observation that filler effect on mastic relative fatigue life was clear. As shown previously in Figure 2-24, mastic relative fatigue lives varied from 0.1 to 2.1, showing a clear filler effect. To further confirm this finding, a multiple linear regression analysis was conducted to determine which filler properties affect mastic fatigue. Results of a best subsets analysis are presented below.

88

Best Subsets Regression: Relative Fatigue Life versus RV, CaO, FM Response is Relative Fatigue Life C Mallows R a F Vars R-Sq R-Sq(adj) Cp S V O M 1 8.0 6.1 0.3 0.48653 X 1 0.9 0.0 4.0 0.50483 X 2 8.4 4.7 2.0 0.49028 X X 2 8.0 4.2 2.3 0.49143 X X 3 8.5 2.8 4.0 0.49519 X X X The best subset analysis show that the best model is has a very low R2 of 6.1%. This indicates that no reasonable prediction model of relative fatigue life using filler properties could be found. It should be noted the lack of correlation between mastic relative fatigue life and mixture fatigue life could be due to the use of stress controlled loading in mixture testing and strain controlled testing of mastics. However, the inability to quantify the filler effect in the mastic overshadows this discrepancy. In summary, the results of mastics and mixtures do not provide a clear trend or relationship that can be used to derive filler or mastic limits for acceptable fatigue resistance of mixtures. It can, therefore, be only concluded that results collected in this study are not sufficient to define role of fillers in mixture fatigue resistance. Further development in the tools used to characterize the fatigue performance of mixtures and mastics are needed in the future.

3.1.5 Low Temperature Cracking Resistance Mixture and mastic results were compared to see if a relationship can be found to determine appropriate limits to ensure acceptable low temperature cracking performance of mixtures. Mixture performance was measured by low temperature stiffness and strength. As indicated in Chapter 2, for low temperature stiffness of mixtures, fillers could not be identified as important variables, and only gradation and base binder source were found to have a statistically significant influence. Hence, in attempting to relate mastic and mixture performance indicators, the focus was on mixture low temperature strength. Because gradation was found to have a significant effect on strength, coarse and fine mixtures were analyzed separately. Figure 3-9 shows that there is a moderate correlation, (R2 of 45%), between mixture strength and mastic relative stiffness for coarse mixtures only. This correlation provides evidence that there is some filler effects on mixture low temperature performance. Figure 3-10 depicts data to confirm that for fine graded mixtures, there is no specific trend relating mastic and mixture indicators.

89

y = 0.51x + 2.02R² = 0.45

3.0

3.5

4.0

4.5

5.0

5.5

6.0

3.00 4.00 5.00 6.00

Mix

ture

Low

Tem

pera

ture

St

reng

th(M

Pa)

Mastic Relative Stiffness

±One Standard Deviation

Figure 3-9 Correlation between Mixture Strength and Mastic Relative Stiffness for Coarse Mixtures

Figure 3-10 Correlation between Mixture Strength and Mastic Relative Stiffness for Fine Mixtures A multiple linear regression analysis was conducted in an attempt to define the role of fillers in increasing the mastic relative stiffness. Results are shown below.

90

Best Subsets Regression: Stiffness versus Binder Low Temp , RV, CaO, FM Response is Mastic Stiffness B i n d e r S t i C Mallows f R a F Vars R-Sq R-Sq(adj) Cp S f V O M 1 38.1 36.9 26.2 69.652 X 1 14.4 12.7 54.7 81.914 X 2 52.5 50.6 10.9 61.616 X X 2 50.8 48.8 13.1 62.763 X X 3 60.6 58.1 3.3 56.750 X X X 3 54.6 51.8 10.4 60.875 X X X 4 60.8 57.5 5.0 57.167 X X X X The best model with all significant parameters, which has an R2 of 58.1%, is displayed below. The model indicates that RV, calcium oxide content (CaO%) and binder low temperature stiffness have significant effects on mastic low temperature stiffness. Regression Analysis: Stiffness versus Binder Low Temp Stiff, RV, CaO The regression equation is Mastic Stiffness = 145 + 2.32 Binder Low Temp Stiff + 4.84 RV - 1.71 CaO Predictor Coef SE Coef T P Constant 144.58 79.45 1.82 0.075 Binder Low Temp Stiff 2.3221 0.3409 6.81 0.000 RV 4.841 1.549 3.12 0.003 CaO -1.7149 0.4964 -3.45 0.001 S = 56.7503 R-Sq = 60.6% R-Sq(adj) = 58.1% This equation can be rearranged in terms of relative stiffness as follows:

145 4.84 1.712.32relativebinder

RV CaOStiffnessStiffness

+ −= +

According to the rearranged model, to maintain the relative stiffness above a limit, RV should be maximized, CaO minimized, and binder stiffness minimized, which agrees with the requirement of upper limit on binder stiffness in the Superpave specification. Thus, the correlation between mixture strength and relative stiffness is consistent with the assumption that a low binder stiffness is desireable.

91

Mixture strength and mastic m-value were also plotted to see if a relationship between mixture strength and mastic m-value exists, as shown in Figure 3-11. The plot reveals no relationship exists, so analyses proceeded with consideration of relative stiffness only. Although a relationship was found between mixture strength and relative mastic stiffness, it was not possible to recommend limits on mastic or filler properties to ensure acceptable mixture low temperature cracking resistance. Cracking resistance is known to depend on stiffness, stress relaxation capability (such as m value), and strength. Since no trends could be found for mixture stiffness, no limits could be proposed. In addition there are no minimum limits of mixture strength cited in the literature for acceptance. The best model for predicting mastic relative stiffness will be included in the recommendation as a guideline for future development.

Figure 3-11 Correlation between Mixture Strength and Mastic m-value

3.1.6 Moisture Damage Resistance The moisture damage evaluation of mastics indicated that mastic performance is highly binder specific with limited influence of filler properties. A limited scale testing plan of mixture was conducted using the Flint Hills modified with SBS and neat Valero binders in the mixtures blended with two mastics with high Moisture Damage Ratios (MDRs) and two mastics with low MDRs. The mixture rut depth at 10,000 cycles measured in the Hamburg Wheel test was measured and correlated to mastic moisture damage ratio (MDR) to see if a relation between mixture and mastic moisture damage resistance exists. This correlation is shown in Figure 3-12 below. As shown in the plot, no relationship between mastic and mixture moisture damage results can be identified.

92

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Rut

Dep

th a

fter

100

00cy

cles

(m

m)

Mastic MDR

Mixture Rut Depth vs. Mastic MDR

Figure 3-12 Correlation between Mixture Rut Depth and Mastic Moisture Damage Ratio (MDR) Despite the minimal filler effects observed in mastic moisture damage testing, an attempt was made to determine filler properties, if any, which affect mastic moisture damage ratio using multiple linear regression analysis. Variables in the analysis included binder MDR and filler properties. No correlation could be identified as evident by the best subsets regression results shown below. Best Subsets Regression: MDR (Moisture Damage Ratio) of mastic as a function of Binder MDR, RV, CaO, and FM Response is MDR (Moisture Damage Ratio) B i n d e r M C Mallows D R a F Vars R-Sq R-Sq(adj) Cp S R V O M 1 3.5 1.6 0.1 0.18600 X 1 2.0 0.1 0.8 0.18744 X 2 5.5 1.7 1.1 0.18590 X X 2 3.9 0.0 1.9 0.18750 X X 3 5.7 0.0 3.0 0.18770 X X X 3 5.6 0.0 3.1 0.18781 X X X 4 5.7 0.0 5.0 0.18967 X X X X

93

The best model obtained, has a very low R2 of 1.7%, indicating the measured filler properties and binder MDR are incapable of reasonably predicting mastic MDR. It is therefore concluded that moisture damage testing of mastics and mixtures do not reveal a significant filler influence. Furthermore, no relation between mastic and mixture moisture damage testing results could be found. It could only be speculate that a larger sample size, or better performance indicators, are needed in order to better define if fillers gave important impact on moisture damage.

3.2 Assessment of Filler Test Methods Based on the analysis of the mixture performance measures, the value of the Rigden Voids (RV), the specific gravity, and the CaO content were identified as the main filler characteristics that could be found to influence mixture performance indicators measured in the study. Based on the literature, size distribution as measured by the Fineness Modulus (FM) was also cited repeatedly as an important filler characteristic. Although the RV measure is known to be affected by the FM, the FM was kept in this part of the study. The assessment of the following testing systems that are used to measure each of the four filler properties was conducted:

1- The Rigden voids test to determine the fractional voids of the fillers. 2- The Helium pycnometer test to determine the specific gravity of the fillers. 3- The Laser diffraction to determine the fineness modulus. 4- The X-Ray Florescence to determine the Calcium Oxide (CaO) content.

The sensitivity analysis was focused on evaluating the repeatability of the results collected at various laboratories. Additionally, a questionnaire was sent to these laboratories to generate a qualitative evaluation of the testing methods with respect to ease of use, time to produce data, and level of complexity. The initial plan was to have six labs participate in testing six selected fillers. It was very challenging to conduct this task because not enough laboratories could participate for all testing systems due to unavailability of equipment. In particular, the X-ray florescence testing device and the Laser diffraction device are not widely available and of high cost. However, the Rigden Voids and the Helium pycnometer devices were simple and small enough to be shipped to various labs.

3.2.1 Fillers Selected for Sensitivity Analysis The selection of the fillers for this task was based on the distribution of the property which showed the most influence, which is the RV. The selection was also partially based on the mineralogical types. Five natural fillers and one manufactured filler that are widely available and commonly used in practice were included. Some of the fillers that showed extreme results and are not widely used, such as Andesite, Caliches, and Gravel Quartzite fillers were avoided in the study. For the manufactured fillers, most showed extreme values with respect to the Rigden voids test with the exception of the Steel

94

Furnace Slag and the non specification Fly Ash. The Slag was selected because it is more widely used and there was enough of it to conduct the testing. Table 3-5 shows the list of fillers selected, the mineralogical type, and the corresponding measured characteristics. To put values in perspective, the ranges in values measured for all fillers tested in the study is also listed in the table.

Table 3-5 List of Selected Fillers and the Corresponding Initial Values Measured.

Code Filler Type Rigden Voids (%)

Specific Gravity FM CaO (%)

LS2 Soft Limestone 35.40 2.62 3.68 46.30 DS2 Soft Dolomite 29.40 2.70 4.73 27.00 GH1 Hard Granite 42.60 2.66 4.06 3.50 GS2 Soft Granite 47.00 2.40 3.13 7.80 BH2 Hard Basalt 33.80 2.77 4.65 7.00 FS1 Steel Furnace Slag 40.00 2.80 4.87 50.30

Selection MAXIMUM values 47.00 2.80 4.87 50.30 Selection MINIMUM values 29.40 2.40 3.13 3.50 Maximum values in the study 49.10 2.89 6.32 50.30 Minimum values in the study 26.20 2.14 2.98 0.95

The plots in Figure 3-13 through Figure 3-16 show the ranking of the selected fillers with respect to the filler population incorporated in the study.

Average = 36.02%

20

25

30

35

40

45

50

55

FAC

1FA

C2

LS1

DH

2FA

N2

DS

2G

RQ

1C

M1

FAF1

LH1

BH

1B

H2

DS

1LS

2G

RQ

2B

V1

HL2

GH

B1

GH

2H

L1FS

1G

S1

CA

1FA

N1

AN

1G

H1

DH

1C

A2

GS

2FS

2

RV(

%)

Minimum = 26.20%Maximum = 49.10%

Figure 3-13 Rigden Voids Distribution Including Selected Fillers for Task 6 (Circled)

95

Average = 2.61

2.0

2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

2.9

3.0

FAF1

FAC

2FA

N1

GS

2FA

C1

CA

2G

H2

FAN

2G

RQ

2G

S1

DH

1C

A1

AN

1D

H2

LS2

GH

B1

LS1

LH1

GH

1G

RQ

1H

L1D

S2

DS

1B

H1

BH

2B

V1

HL2

FS1

CM

1FS

2

SGMinimum = 2.14Maximum = 2.89

Figure 3-14 Specific Gravity Values Distribution Including Selected Fillers for Task 6 (Circled)

Average = 4.45

2.0

2.5

3.0

3.5

4.0

4.5

5.0

5.5

6.0

6.5

7.0

FAN

2G

S2

FAN

1FA

C1

LS2

GH

2G

S1

FAC

2G

H1

HL2

GH

B1

CM

1FA

F1LS

1FS

2G

RQ

1D

H2

BV

1B

H2

DS

2H

L1FS

1C

A1

DH

1C

A2

AN

1B

H1

LH1

DS

1G

RQ

2

FM

Minimum = 2.98Maximum =6.32

Figure 3-15 Fineness Modulus Values Distribution Including Selected Fillers for Task 6 (Circled)

96

Average = 23.03

0

10

20

30

40

50

60

GR

Q2

GH

2G

H1

GS

1G

HB

1FA

F1A

N1

BH

2G

S2

BH

1B

V1

CM

1FA

N1

FAC

2FA

N2

FAC

1D

H1

DS

1D

S2

HL1

DH

2H

L2G

RQ

1C

A2

LH1

CA

1FS

2LS

2LS

1FS

1

CaO

(%)

Minimum = 0.95%Maximum = 50.3%

Figure 3-16 Fineness Modulus Values Distribution Including Selected Fillers for Task 6 (Circled)

Figures 3-13 through 3-16 show that the six fillers selected cover a good level of spread over the range of measured filler properties. This allows for a reasonable evaluation of the test procedures at different levels. The fillers were sampled and split to produce a representative sample for each test. Split samples were packaged and shipped to the participating laboratories. This ensured that all laboratories and operators test similar specimens with minimal error due to different sampling practices. The fillers were labeled in numbers from one to six such that the filler type and mineralogies would be anonymous to ensure results would not be influenced by the operators experience.

3.2.2 Results from the Sensitive Testing The data collected in this task is summarized in this section. Only a subset of the intended number of laboratories participated.

1- RIGDEN VOIDS è 4 LABS OUT OF 6 PARTICIPATED. 2- FINENESS MODULUS è 2 LABS OUT 6 PARTICIPATED.

3- CALCIUM CONTENT è 1 LAB OUT OF 6 PARTICIPATED. 4- SPECIFIC GRAVITY è 3 LABS OUT OF 6 PARTICIPATED.

97

The laboratories involved in conducting testing for this task and the details of testing conditions are summarized as follows:

1- University of Wisconsin-Madison a. Two operators b. Tests: Rigden Voids, and Specific Gravity.

2- University of Illinois at Urbana-Champaign a. Two operators. b. Tests: Rigden Voids, Specific Gravity, and Fineness Modules.

3- Mathy Technology and Engineering at Onalaska, Wisconsin a. Two Operators. b. Tests: Rigden Voids, Specific Gravity, Calcium Oxide Content, and

Fineness Modulus. 4- Bloom Companies, LLC at Milwaukee Wisconsin.

a. One Operator. b. Tests: Rigden Voids

Testing was conducted by two operators at most of the labs which allows for the evaluation of operator error as well as inter-laboratory variation. All RV, and specific gravity testing was conducted using the same device. The laser diffraction testing was conducted based on different devices available within the participating laboratories. This is because the laser diffraction device is not as portable. The pooled standard deviation is used to estimate the variability due to changing the laboratory and operators. The pooled standard deviation can be measured using the following formula.

Equation 3-2

The suffices 1, 2, ... k refer to the different series of measurements, in this case it represents the number of laboratory, or the operator. The “s” value represents the standard deviation for each measurement due to the variation of the laboratory, or operator. The following table details the variability between the different laboratories in the results obtained for each filler independently. Table 3-6 shows the coefficient of variation (COV) for each filler property. The COV is calculated using the pooled standard deviation of the results when varying the laboratory and operators, combined. It is the outcome of dividing the pooled standard deviation by the average values for each filler property. For the CaO results only one laboratory was involved but for each Filler 2 operators measured property for two replicate samples.

98

Table 3-6 Variability in Filler Testing between Laboratories Test Filler

Rigden Voids (%) CaO (%) FM SG

Overall COV (%) LS2 0.98% 1.54% 2.12% 0.66% GS2 2.31% 1.76% 1.34% 1.76% GH1 1.53% 0.51% 0.58% 1.49% FS1 0.60% 0.26% 2.75% 1.96% BH2 2.03% 0.07% 0.51% 0.31% DS2 1.46% 0.85% 0.60% 1.52% Max 2.31% 1.76% 2.75% 1.96% Min 0.60% 0.07% 0.51% 0.31%

No. of Labs 4 1 2 3 As shown in the table, all tests appear to be fairly repeatable with the maximum coefficient of variation, observed for the FM value for the steel furnace slag (FS1) filler, at 2.75%. The least variability was observed for the X-ray florescence test to determine the CaO%. However, the low variability for the CaO% can be attributed to the fact that only one laboratory conducted the test. This test is uncommon for the pavement industry and it was challenging to locate additional laboratories to conduct the testing. In general, all the observed COV values due to changing the laboratories is less than 3% indicating that the between-laboratory error is also minimal. To evaluate the operator variability, Table 3-7 shows the COV values calculated for the pooled standard deviation due to varying the operators only for all the tests.

Table 3-7 Variability in Filler Testing between Operators Test Filler

Rigden Voids (%) CaO (%) FM SG

Overall COV (%) LS2 0.74% 0.03% 1.14% 0.30% GS2 0.59% 1.31% 0.72% 0.63% GH1 0.87% 0.38% 0.51% 0.50% FS1 0.40% 0.01% 1.19% 0.59% BH2 0.90% 0.05% 0.48% 0.81% DS2 1.14% 0.01% 0.65% 0.69% Max 1.14% 1.31% 1.19% 0.81% Min 0.40% 0.01% 0.48% 0.30%

No. of Labs 4 1 2 3 Table 3-7 shows that the variability of the test results due to changing the operator is less than the variability due to changing laboratory. The FM for FS1 still shows the maximum variability in the laser diffraction test but only at a COV value of 1.19%. In general, the observed COV values for all the tests were below 2%. In conclusion, the tests included in this sensitivity analysis seem to be highly repeatable. More importantly, the analysis of the mixture results revealed that the filler RV and the

99

specific gravity are the most important filler properties. Therefore, this sensitivity analysis indicates that the filler testing procedures selected in this study are fairly repeatable and are able to distinguish between the different fillers with high accuracy.

3.2.3 Qualitative Assessment of Filler Testing Procedures During the sensitivity analysis of the filler tests, a questionnaire was distributed to the different laboratories. The questionnaire included the following questions regarding the test procedure in order to get feedback on how practical are these methods from the operator stand point.

1- Is the procedure clear enough? 2- What is the time required to complete test? 3- What level of training is required? 4- What is level of simplicity 5- Is the equipment easy to operate? 6- Variability of test result: 7- The operators were given space to include any additional remark

Table 3-8 includes the details of the feedback gathered from the different laboratories. It should be noted that the questionnaire included a detailed procedure to conduct each test as well as the corresponding standard specification. The response from the different operators participating in this study showed that all the testing procedures are clear. The time required to conduct the different tests varied significantly. The fractional voids test using the Rigden voids device was the test that required the least time to obtain results for one test. Six operators reported that the test requires 10min to 20min to conduct one test. Only one operator reported that the test required 30 min to conduct one test. On the other hand, the Helium pycnometer used to determine the specific gravity was reported to take 90min to conduct one test by two of the operators. The same test required 15min to 20min by other three operators. Further communication with the operators revealed that the long time reported by one laboratory includes machine preparation and warm up, while the actual testing time is in the range of 30min. One of the operators reported that the time required to conduct the laser diffraction test is dependent on the filler type. Further communication with the operator revealed that the test time consistently takes 45min. The operator indicated additional time is required however, for post testing clean up depending on the filler type as the finer the filler the more time required to conduct the test. All the tests required minimal to moderate training and were reported to be simple to moderately simple to conduct. The testing devices are also reported to be easy to operate with one exception. One operator of the Rigden voids device indicated it was moderately simple to operate.

100

All the tests were reported to produce repeatable results. One of the operators of the X-Ray Florescence test reported that the repeatability is dependent on the time difference between two consecutive samples. The operator indicated that the test is repeatable if the samples are tested one after the other, but if test was conducted on two different days the repeatability of the test drops. In summary, the tests included in the sensitivity analysis show that the four fillers’ tests included are user friendly and can produce repeatable results after minimal training. It should be mentioned however that the X-ray florescence and laser diffraction devices are costly and not practical for current practice. The results from these tests, however, are less important than the Rigden Voids and Specific Gravity.

Table 3-8 Summary of Questionnaire Showing the Feedback of the Different Operators

Num

ber

of O

pera

tors

1- Is

the p

roce

dure

cle

ar

enou

gh?

2- W

hat i

s the

tim

e re

quir

ed to

com

plet

e te

st?

3- W

hat l

evel

of t

rain

ing

is r

equi

red?

4- W

hat i

s lev

el o

f si

mpl

icity

5- Is

the e

quip

men

t eas

y to

ope

rate

?

6- V

aria

bilit

y of

test

re

sult:

Remarks

Rig

den

Voi

ds

7 7: Yes 6:10-20Mins 1: 30Mins

6: Minimal 1: Moderate

6: Simple 1: Moderate

6: Yes 1: No* 7: Repeatable

* The reason indicated is that it is difficult to cut the filter paper. 1: The device required careful cleaning with continual use.

Hel

ium

Pycn

omet

er

6 6: Yes

3:15-20Mins 1: 40Mins 2: 90Mins

6: Minimal 5: Simple 1: Moderate 6: Yes 6: Repeatable

Lase

r Diff

ract

ion

4 4: Yes

1: 60Mins 1: 30Mins 1: 15Mins 1: Depends on Sample

1: Minimal 3: Moderate

1: Simple 3: Moderate 4: Yes 4: Repeatable

1: Need to determine when the ultrasonic agitations is needed. 1: Test is repeatable when run one sample after the other. Repeatability is affected is samples are run on two different days.

X-ra

y Fl

ores

cenc

e

2 2: Yes 2: 25-30Mins 1: Minimal 1: Moderate

1: Simple 1: Moderate 2: Yes 2: Repeatable

1: Calibration and method development should be conducted by experienced individuals. 1: Some fillers are difficult to prepare for testing.

101

4 CONCLUSIONS AND SUGGESTED FUTURE RESEARCH

This study included an extensive literature search and analysis and a survey to identify filler characteristics that influence performance of mastics and mixtures performance. The study also included a large experimental study in which natural and manufactured fillers from 32 sources were collected and characterized using carefully selected physical and chemical tests. Based on the fillers’ testing results, a subset of 17 fillers were mixed with four asphalt binders and two aggregate gradations to characterize performance related properties of 68 mastics and 32 mixtures. The testing of mastics and mixtures covered five main performance indicators related to workability, rutting, fatigue, low temperature cracking, and moisture damage. For each of these indicators results of filler testing were statistically correlated and trends were identified for relationships between filler properties and mixture performance indicators. Additionally, models to estimate mastic performance in terms of filler and binder properties were developed. The conclusions in this chapter are drawn from the study findings. The findings indicate that for some mixture properties, the analysis of results offer sufficient information to suggest specification criteria for fillers to ensure satisfactory performance of mixtures, while for other critical properties of mixtures further research is needed.

4.1 General Conclusions The following points summarize the general conclusions for the study.

• Although there is no lack of evidence in the published literature that fillers can have important effects on mixture performance, the current asphalt mix design procedures and component specifications used in the United States include only general limits on filler to binder mass ratio. These procedures, including the Superpave system, do not provide sufficient guidance for testing of fillers to address the possible influence of fillers on critical mix performance indicators.

• Based on extensive review of literature and international standards, four primary filler characteristics were identified as critical for defining the influence of fillers on mixture performance indicators. These include fractional voids, size distribution, content of calcium compounds and active clay content. In addition the specific gravity needs to be measured for calculation of the fractional voids. Standard methods for measuring these properties were identified and modified in this study. The procedures are well established and found suitable for practical applications. However, some of the devices are not yet available in the North American market (e.g. Rigden Voids), and others (e.g. X-ray florescence) are relatively costly.

• Mineral fillers currently used in practice are found to vary significantly in their physical and chemical properties as measured by the four selected testing methods. In general, manufactured fillers, including fly ash and slag dust, show a more extreme range of properties compared to natural fillers. The effects of these fillers on mastics and mixture performance indictors cannot be easily estimated.

102

• Mineral fillers are found to cause significant changes in performance indicators of the resulting mastics and mixtures. The changes are in most cases highly filler and binder specific. Additionally, the mass ratio of filler to binder used in this study (1:1) is not found sufficient to estimate effects of fillers on mixture behavior. Therefore, it could be concluded that the current practice of limiting dust to binder ratio by mass is insufficient to ensure acceptable influence of fillers on mixture or mastic performance. It was clearly shown that specific filler characteristics measured by Rigden Voids (packing characteristics) vary significantly between natural fillers and have important influence on mastic and mixture behavior. Varying mass ratio was not studied, therefore, further study is required to evaluate the role of the filler concentration on mastic performance

• The study found that natural fillers, which are fillers produced in quarries during

rock processing or collected in baghouse collectors during HMA production, have significantly different effects than manufactured fillers, which are by products such as fly ash and slag dust. The natural fillers appear to influence mastic and mixture performance using a uniform physio-chemical mechanism. Therefore, a regression analysis to obtain a generalized prediction model for some mastic properties was achievable. The manufactured fillers, on the other hand, showed unique influence on the mastic and mixtures studied in this report. A generalized trend could not be obtained and thus it is recommended that these fillers be tested thoroughly after mixing with intended binders before introducing into the mix.

4.2 Specific Conclusions by Performance Indicators Based on the correlations and models developed for the specific mixture performance indicators, the following conclusions for each indicator could be drawn: • Workability Indictors: Mastic viscosity was successfully related to mixture

workability, as measured by the number of cycles to 92% Gmm. Although N92 values were found to be highly dependent on aggregate gradation, using multi-linear regression, the Rigden Voids value was identified as the filler property that has important influence on mastic viscosity and mixture workability. Tentative maximum limits for mastic relative viscosity were defined to ensure acceptable mixture compactability for coarse graded mixtures. Limits for mixtures were estimated from the project data set since no specific guidance could be found in the literature. In cases where mastic testing is not possible, a statistical model to estimate mastic relative viscosity based on Rigden voids of filler and binder viscosity is proposed. The model can be used to check that the relative viscosity of mastic is below the maximum proposed limit.

• Rutting Resistance Indictors: Mastic rutting resistance indicators were

successfully related to mixture rutting resistance as measured by the flow number (FN). Mastic non-recoverable compliance (Jnr) was found to have an influence on mixture FN that is statistically significant but not as important as the aggregate

103

gradation. Tentative maximum limits for mastic Jnr were defined to ensure acceptable mixture FN values for coarse and fine graded mixtures. Acceptable limits for mixtures were derived based on most recent studies focused on mixture rutting indicators. A statistical model to estimate mastic Jnr based on filler Rigden Voids and binder Jnr is proposed to ensure acceptable FN values.

• Fatigue Resistance Indictors: Mastic fatigue resistance indicators measured in this study were found to vary significantly based on fillers used. The role of fillers in mastic fatigue was found highly dependent on binder chemistry, and more importantly, binder modification type. Mixture fatigue was found to be affected mostly by gradation, binder modification, and only marginally by filler or base binder properties. It is therefore concluded that results of this study are not sufficient to define role of fillers in mixture fatigue resistance. Further development in the tools used to characterize fatigue performance of mixtures and mastics are needed before role of fillers in fatigue performance could be specifically defined, and specification criterion be derived.

• Low Temperature Cracking Resistance Indictors: For low temperature stiffness of mixtures, gradation and base binder source are the only variables that could be found to be statistically significant. In mixture strength however, it was found that mastic properties, specifically the mastic relative stiffness, have significant effects and only slightly less important than gradation. Mastic low temperature stiffness and creep rate (m) were found to be sensitive to type of filler and binder modification. It was found, as expected, that all fillers increase binder stiffness, but fillers can increase or decrease (m) value, depending on modification type. The statistical analysis found that the mastic low temperature stiffness is dependent on the Rigden voids and Calcium Oxide (CaO) of the filler. A model for estimating mastic stiffness at low temperatures as a function of filler and binder properties is proposed to ensure acceptable mastic influence on mixture strength. However, no strength limit is proposed.

• Moisture Damage Resistance Indictors: Mastic results showed that mastic

moisture resistance indictors are highly binder specific with limited influence of filler properties. Therefore, testing of mixture for moisture damage was conducted on a limited scale as compared to other performance indicators. The results showed that mixtures’ moisture resistance is highly dependent on mastic performance but that dependency is mainly due to the effects of binder type used, rather than the fillers used in mastics. Mixture moisture damage testing was terminated due to lack of evidence that fillers can have an important influence as compared to the binder modification type. This finding cannot be generalized due to the fact that only four binders with highly variable properties were used in the study.

• The assessment of testing of selected filler characterization methods included four

tests. The results, although limited in scope indicate that all tests are highly repeatable. In particular, test methods for the filler Rigden voids and the specific

104

gravity, which have been identified as the most important filler properties for mixture performance, were found to be highly repeatable and are able to distinguish between the different fillers with high accuracy. The tests were rated by the operators as user friendly and can produce repeatable results after minimal training. The other two tests studied included the X-ray florescence and laser diffraction, although showed good repeatability are found to be costly and potential not practical for routine testing of fillers.

4.3 Recommended specification criteria for mineral fillers that optimize HMA performance.

Due to the high filler-binder interactions measured in this study, the specification criteria proposed are based on mastic properties rather than filler properties. Table 4-1 is proposed as the framework for controlling effects of fillers in Hot Mix Asphalt to ensure acceptable performance. The table lists the mixture performance indicators that were considered, the mastic properties measured, and the proposed limits. In case mastic properties cannot be measured, the table includes the best fit models derived from this project to estimate mastic properties in terms of filler and binder properties. Due to the limited number of binders used in this study, it is high recommended that these models be verified, and modified if needed, in future studies.

Table 4-1 Mastic property and limits to ensure proper mixture performance

Performance Indicator Mastic Property Mastic

Limit Mastic Models

Workability Relative Viscosity @135ºC <5.0 Mastic Viscosity = - 8244 + 4.68 Binder

Viscosity + 205 Rigden Voids

Rutting Jnr @3.2kPa and

58ºC (1/kPa)

< 0.40 Mastic Jnr = 1.01 + 0.160 Binder Jnr - 0.0230 Rigden Voids

Low Temperature Relative Stiffness @60 Seconds and

@-12ºC No Limits

Mastic Stiffness = 145 + 2.32 Binder Low Temp Stiff + 4.84 Rigden Voids – 1.71

CaO%

Moisture Damage Pull off strength No Limits No acceptable Model

Fatigue G*sin(delta),

Number of Cycles to Failure

No Limits No acceptable Model

The specification limits proposed are preliminary and should be measured against local experience.

105

4.4 Suggested Future Research

• The study used one filler to binder ratio by mass. Although the ratio selected is commonly used in practice, the role of volume fraction of filler on performance was estimated indirectly in this study. A more in-depth evaluation of the effect of volume fraction on fatigue and low temperature cracking could be very useful to define limits that ensure satisfactory mixture performance.

• This study could not generated sufficient information to critically define the role of fillers in mixture fatigue and low temperature mixture stiffness performance. It was found that binders selected for the study have a much higher influence than the fillers. This could be caused by the limited size of the binder set and/or the extreme difference in binder behavior. A broader selection of binders could prove to be more informative.

• Moisture damage results were significantly biased by one of the modified asphalts used in the study. A study using more un-modified asphalts with the fillers selected could shed more light on the possible role of fillers in moisture damage resistance. Past research has claimed calcium bearing fillers improve moisture damage resistance but this could not be quantified in this study.

• The effect of filler volume fraction on the measured mastic and mixture performance needs to be studied further. This study indicated the binder modification may interact with the filler used. This interaction may enhance the role of the modifier or reduce it. The effect of volume fraction needs to be studied to determine how this will affect the interaction between filler and modification.

• The analysis in this study showed that manufactured fillers have unique influence on the mastics and mixtures. However, each manufactured filler appears to interact uniquely with different binders and modifiers. More detailed investigation of the manufactured fillers is required for a better understanding of their influence on mastic and HMA performance.

106

5 REFERENCES

1- Standard Specification for Mineral Filler for Bituminous Paving Mixtures. ASTM D 242-04. 2004.

2- Ridgen, P .J. The Use of Fillers in Bituminous Road Surfacing – A Study of Filler-Binder System in Relation to Filler Characteristics. Journal of Society of Chemical Industry, No. 66, 1947, pp. 9- 299.

3- Anderson, D.A. and W.H. Goetz. Mechanical Behavior and Reinforcement of Mineral Filler-Asphalt Mixtures. Proceedings. Association of Asphalt Paving Technologists, Volume 42, 1973, pp. 37–66.

4- Anderson, D.A. Guidelines for the Use of Dust in Hot Mix Asphalt Concrete Mixtures. Proceedings of the Association of Asphalt Paving Technologists, Volume 56, 1987, pp. 492-516.

5- Rigden Voids and Wilhelmi Ring and Ball Tests to Quantify the Role of Filler in Soil Treated with Emulsion. Road Materials and Pavements Design Journal, 2005, pp. 1-20.

6- European Committee for Standardization, . EN 1097-4, “Tests for Mechanical and Physical Properties of Aggregates, Part 4, Determination of the Voids of Dry Compacted Filler. Brussels: CEN; 1999.

7- Anderson, D.A., J.P . Tarris. Adding Dust Collector Fines to Asphalt Paving Mixtures. In Transportation Research Record: Journal of the Transportation Research Board, No. 252, TRB, National Research Council, Washington, D.C., 1982, pp. 90.

8- Smith, B.J., S. Hesp. Crack Pinning in Asphalt Mastic and Concrete: Regular Fatigue Studies. In Transportation Research Record: Journal of the Transportation Research Board, No. 1728, TRB, National Research Council, Washington, D.C., 2000, pp. 75-81.

9- Winniford, R. S. The Rheology of Asphalt-Filler Systems as Shown by the Microviscometer.. American Society for Testing and Materials. STP309 Pp. 109-120. 1961.

10- Heukelom, W. Role of Filler in Bituminous Mixes. Journal of Association of Asphalt Paving Technologists, Vol. 34, 1965.

11- Heukelom, W., and Wija, P . W. O. Viscosity of dispersions governed by concentration and rate of shear. Journal of Asphalt Paving Technologists .Vol. 40 1971.

12- Craus, J., I, Ishai and A. Sides. Some Physico-Chemical Aspects on the Effect of the Filler on the Properties and Behavior of Bituminous Paving Mixtures. Proceedings from the Asphalt Paving Technologists, Volume 47, 1978, pp. 558-588.

107

13- Kandhal, P .S. Evaluation of Baghouse Fines in Bituminous Paving Mixtures”. Journal of the Association of Asphalt Paving Technologists. Volume 50, 1981. pp 150-210.

14- Dukatz, E., and D. Anderson. The Effect of Various Fillers on the Mechanical Behavior of Asphalt and Asphaltic Concrete. Proceedings, Association of Asphalt Paving Technologists. Volume 49. 1980.

15- Anderson, D.A., J. P . Tarris, and J. D. Brock. Dust Collector Fines and Their Influence on Mixture Design. Asphalt Paving Technology. Proceedings of the Association of Asphalt Paving Technologists. Technical Sessions in Kansas City, Misouri. Volume 51. 1982, pp. 363-397.

16- Shashidhar, N. R. Romero. Factors Affecting the Stiffening Potential of Mineral Fillers. In Transportation Research Record: Journal of the Transportation Research Board, No. 1638, TRB, National Research Council, Washington, D.C., 1998, pp. 94.

17- Shashidar, N., S.P. Needham, B.H. Chollar. Prediction of the Performance of Mineral Fillers in Stone Matrix Asphalt. 1999. Pp. 222-245.

18- Mogawer, W.S. Effects of Mineral Fillers on Properties of Stone Matrix Asphalt Mixtures. In Transportation Research Record: Journal of the Transportation Research Board, No. 1530, TRB, National Research Council, Washington, D.C., 1996, pp. 86-94.

19- Kandhal, P .S., C.Y. Lynn, F. Parker. Tests for Plastic Fines in Aggregates Related to Stripping in Asphalt Paving Mixtures. NCAT Report 98-03. Auburn University, Auburn, Alabama. 1998.

20- Harris, B.M.and K.D. Stuart. Analysis of Mineral Fillers and Mastics Used in Stone Matrix Asphalt. Journal of the Association of Asphalt Paving Technologists, Volume 64, 1995, pp. 54-95.

21- Kallas, B.F. and Puzinauskas, V .P ., A Study of Mineral Fillers in Asphalt Paving Mixtures. Proceedings of the Association of Asphalt Paving Technologists, Volume 10, 1961, pp. 493–528.

22- Tunniclif, D.G. Review of Mineral Fillers. Journal of Association of Asphalts Paving Technologists, Vol. 3, 1962, pp. 118-150.

23- Anderson, D.A., R. Dongre, Effects of Minus No. 200-Sized Aggregate on Fracture Behavior of Dense-Graded Hot-Mix Asphalt. ASTM Special Technical Publication, 1992, No. 1147 pp. 154-176.

24- Anderson, D.A. Rheological Properties of Mineral Filler-Asphalt Mastics and its Importance to Pavement Performance. ASTM Special Technical Publication, 1992, No. 1147, pp. 131-153.

25- Brown, E.R, J.E. Haddock, and C. Crawford. Investigation of Stone Matrix Asphalt Mortars. In Transportation Research Record: Journal of the Transportation Research Board, No. 1530, TRB, National Research Council, Washington, DC, 1996, pp. 95–102

108

26- Mogawer, W.S., and K.D. Stuart. “Effects of Mineral Fillers on Properties of Stone Matrix Asphalt Mixtures,” Transportation Research Record 1530, Transportation Research Board, National Research Council, Washington, DC, 1996; pp. 86–94

27- Cooley, L., M. Stroup-Gardiner, E.R. Brown, D. Hanson, and M. Fletcher. Characterization of Asphalt-Filler Mortars with Superpave Binder Tests. Journal of the Association of Asphalt Paving Technologists, Volume 67, 1998, pp. 42- 65.

28- Kandhal, P .S., C. Y . Lynn. Characterization Tests for Mineral Fillers Related to Performance of Asphalt Paving Mixtures, In Transportation Research Record: Journal of the Transportation Research Board, No. 1638, TRB, National Research Council, 1998, pp. 101-110.

29- Kim, Y ., D. N. Little, and R.L. Lytton. Fatigue and Healing Characterization of Asphalt Mixtures. Journal of Materials in Civil Engineering, ASCE, Volume 15, No. 1, 2003, pp. 75-83.

30- Kim, Y .R., D.N. Little, I. Song. Effect of Mineral Fillers on Fatigue Resistance and Fundamental Material Characteristics: Mechanistic Evaluation. In Transportation Research Record: Journal of the Transportation Research Board, No. 1832, TRB, National Research Council, Washington, D.C., 2003, pp. 1-8.

31- Kim, Y .R, D.N. Little, R.L. Lytton. Effects of Moisture Damage on Material Properties and Fatigue Resistance of Asphalt Mixtures. In Transportation Research Record: Journal of the Transportation Research Board, TRB, National Research Council, Washington, D.C., 2004, pp. 48-54.

32- Vansteenkiste, S., “Properties of Fillers”, Belgian Road Research Centre, 2005. Presentation.

33- Determination of Acid Insoluble Residue in Limestone and Dolostone. ASTM D3042 (Minnesota DOT modified), 2000.

34- European Committee for Standardization, EN 459-2, “Building Lime, Part 2, Test Methods”. Brussels: CEN; 2001.

35- European Committee for Standardization, EN 1744-1, “Tests for Chemical Properties of Aggregates – Part 1: Chemical Analysis”. Brussels: CEN; 1998.

36- Panalytical Epsilon 5 X-Ray Fluorescence Spectrometer technical information. Web Page: http://www.panalytical.com/index.cfm?pid=259.

37- European Committee for Standardization, EN 1744-4, “Tests for Chemical Properties of Aggregates – Part 4: Determination of Water Susceptibility of Fillers for Bituminous Mixtures”. Brussels: CEN; 2005.

38- Faheem, A., H. Wen, L. Stephensen, and H. Bahia. “Effect of Mineral Filler on Characteristics of Asphalt Binders”. Journal of the Association of Asphalt Paving Technologists, Vol. 77, 2008

39- (a)Faheem, A., H. Bahia. “Conceptual Phenomenological Model for Interaction of Asphalt Binders with Mineral Fillers”. Journal of the Association of Asphalt Paving Technologists, Vol. 78, 2009

109

(b) Faheem, A., Bahia H, “Modeling of Asphalt Mastic in Terms of Filler-Bitumen Interaction”, Journal of Road Materials and Pavements Design, Volume 11 – Special Issue, 2010

40- Leiva-Villacorta, V . and West, R., “Relationships Between Laboratory Measured Characteristics of HMA and Field Compactability”, Journal of the Association of Asphalt Paving Technologists, Vol. 77, pp. 183-220, 2008.

41- Advanced Asphalt Technologist, LLC., “PG Binder Grade Selection for Airfield Pavements” Airfield Asphalt Pavement Technology Program (AAPTP) Project 04-02: Final Report, 2008.

42- Advanced Asphalt Technologies, LLC. “Mix Design Practices for Warm Mix Asphalt- Participant Workbook- Appendix to AASHTO R35, Simple Mix Design Considerations & Methods for Warm Mix Asphalt.” National Cooperative Highway Research Program (NCHRP) on Project NCHRP 9-43”, October 2010.

43- Ishai, I., J. Craus, A. Sides. A Model for Relating Filler Properties to Optimal Behavior of Bituminous Mixtures. Proceedings from the Association of Asphalt Paving Technologists, Volume 49, 1980, pp 416.

44- Bolk, N., J. Van Der Heide, M. Van Zantvliet. Basic Research into the Effect of Filler on the Mechanical Properties of Dense Asphaltic Concrete. Journal of the Asphalt Paving Technologists, Volume 51, 1982, pp.398.

45- Taybebali, A., G. Malpass, N. Khosla. Effect of Mineral Filler Type and Amount on Design and Performance of Asphalt Concrete Mixtures. In Transportation Research Record: Journal of the Transportation Research Board, No. 1609, TRB, National Research Council, Washington, D.C., 1998, pp. 36.

46- Ishai, I., J. Craus. Effect of the Filler on Aggregate-Bitumen Adhesion Properties in Bituminous Mixtures, Proceedings from the Association of Asphalt Paving Technologists, Volume 46, 1977, pp 228.

47- Warden, W.B., S.B. Hudson, H.C. Howell. Evaluation of Mineral Filler in Terms of Practical Pavement Performance. Asphalt Pavement Technologists, Volume 28, 1959

48- Rao, B.K., B.R. Sen. Evaluation of Mineral Fillers for Asphalt Paving Mixture, Indian Institute of Technology, pp. 100-106.

49- Lee, A.R., Rigden, P .J., Use of Mechanical Tests in Design of Bituminous Road-Surfacing Mixtures - Dense Tar Surfacings. Journal of the Society of Chemical Industry, Volume 64, Issue 6, 1945, pp. 153-161.

50- Sarsam, S.I.. A Study of Mineral Filler in Dense Graded Asphalt Concrete. Indian Highways, Volume 12, Issue 12, 1984, pp 5-10.

51- Ishai, I., and J. Craus. Effects of Some Aggregate and Filler Characteristics on Behaviour and Durability of Asphalt Paving Mixtures. In Transportation Research Record: Journal of the Transportation Research Board, No. 1530, TRB, National Research Council, Washington, D.C., 1996, pp. 75-85.

110

52- Silva, H., J. Pais, P . Pereira, D. Gardete. Influence of the Mastic Binder in the Bituminous Mixtures Behaviour. Belfast, UK: Fourth International Conference on Maintenance and Rehabilitation of Pavements and Technological Control, 2005.

53- Tsohos, G., Influence of Filler on the Rheological and Mechanical Properties of Bitumen. Civil Engineering for Practicing and Design Engineering, Volume 5, Issue 6, 1968, pp. 453-475.

54- Kavussi A and Hicks R.G., Properties of Bituminous Mixtures Containing Different Fillers, Journal of the Association of Asphalt Paving Technologists, Volume 66, 1997, USA, pp. 153-186.

55- Bissada, A.F., A.A. Anani. Evaluation of Rubberized Limestone Filler in Asphalt Paving Mixtures. Data for Decision Makers. Rubber Chemistry and Technology, Volume 57, 1983.

56- Venkataraman, T.S., V. Venkatasubramanian. Effect of Fillers on the Rheological Response of Sheet Asphalt Mixtures. Highway Research Bulletin, New Delhi, 1977