<%BANNER%>

Identification of a physical model to evaluate the rutting performance of asphalt mixtures

University of Florida Institutional Repository

PAGE 1

IDENTIFICATION OF A PHYSICAL MODEL TO EVALUATE RUTTING PERFORMANCE OF ASPHALT MIXTURES By CHRISTOS ANDREA DRAKOS A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2003

PAGE 2

Copyright 2003 by Christos Andrea Drakos

PAGE 3

To my parents – Andreas and Olga, my nephew Andreas, and my niece Fotini.

PAGE 4

iv ACKNOWLEDGMENTS I would like to acknowledge those individuals who were involved in the advancement of this research. First I would like to thank my advisor and mentor Dr. Rey Roque for willingly sharing his knowledge a nd experiences through constant support and advice. Acknowledgments should also be pa id to my graduate committee members – Dr. Bjorn Birgisson, Dr. Mang Tia, and Dr. Byr on Ruth – who were always available to discuss ideas and lend valuable advice. Also, I would like to thank my external committee member Dr. Wayne Losano for his tec hnical writing guidance and assistance. Special thanks and appreciation go to th e Florida Department of Transportation, and more specifically to the bituminous materials group – Greg Sholar, Howie Moseley, Susan Andrews, Frank Suarez, Shanna J ohnson, and Stephen Browning – for their support throughout the project. Their help has been invaluable. From the University of Florida, I would like to thank Mr. George Lopp, Sungho Kim, Marc Novak and Edward Roske for their technical assistance in the lab. I want to thank my parents, Andreas and Olga Drakou, for being a source of inspiration and support throughout my studies in the United States. Finally I would like to thank my friends Maria Alvey, Eri Messa ritaki, Kally Kanellis, Alexis Klironomos, and Maria Nikolou for their true friendship.

PAGE 5

v TABLE OF CONTENTS Page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES...........................................................................................................viii LIST OF FIGURES.............................................................................................................x ABSTRACT.....................................................................................................................xi v 1 INTRODUCTION........................................................................................................1 1.1 Background.............................................................................................................1 1.2 Problem Statement..................................................................................................3 1.3 Hypothesis..............................................................................................................4 1.4 Objectives...............................................................................................................4 1.5 Scope...................................................................................................................... .5 1.6 Research Approach.................................................................................................6 2 LITERATURE REVIEW.............................................................................................7 2.1 Overview.................................................................................................................7 2.2 Permanent Deformation..........................................................................................8 2.2.1 Consolidation Rutting...................................................................................8 2.2.2 Instability Rutting.........................................................................................9 2.3 Tire-pavement In terface Stresses..........................................................................10 2.4 Predicting Mixture Performance...........................................................................18 2.5 Accelerated Pavement Testers..............................................................................19 2.6 Superpave™ Shear Tester....................................................................................20 2.7 Torture Tests.........................................................................................................22 2.7.1 Hamburg Wheel-Tracking Device.............................................................22 2.7.2 French Pavement Rutting Tester................................................................24 2.8 Asphalt Pavement Analyzer.................................................................................26 2.9 Summary...............................................................................................................28

PAGE 6

vi 3 APA LOADING MECHANISM................................................................................29 3.1 Overview...............................................................................................................29 3.2 Limitations of Loaded Wheel Testers...................................................................30 3.3 New APA Loading Mechanism............................................................................32 3.3.1 Development of a Tire Model....................................................................32 3.3.2 Verification of the Tire Model...................................................................35 3.3.3 Concept Loading Strip................................................................................38 3.4 Preliminary Contact Stress Calculations..............................................................39 3.4.1 APA Pressurized Hose Stresses..................................................................39 3.4.3 Loading Strip FEM.....................................................................................41 3.5 Measured Contact Stresses in the APA................................................................43 3.5.1 Measurement System..................................................................................43 3.5.2 Hose-Specimen Interface Stresses..............................................................44 3.5.3 Loading Strip-Specimen Interface Stresses................................................47 3.6 Summary...............................................................................................................48 4 STRESS ANALYSES................................................................................................51 4.1 Overview...............................................................................................................51 4.2 Pavement Stress Analyses....................................................................................52 4.2.1 Multi-Layer Elastic Stress Analyses..........................................................52 4.2.2 BISAR Results............................................................................................54 4.2.3 Finite Element Stress Analyses..................................................................60 4.2.3-1 Loading the FEM.............................................................................63 4.2.3-2 FEM Results.....................................................................................65 4.3 APA Stress Analyses............................................................................................66 4.4 Summary...............................................................................................................70 5 MATERIALS AND TESTING METHODS..............................................................71 5.1 Overview...............................................................................................................71 5.2 Materials...............................................................................................................72 5.3 Mixture Preparation..............................................................................................77 5.3.1 Aggregate Preparation and Batching..........................................................77 5.3.2 Mixing........................................................................................................77 5.3.4 Short-Term Oven Aging (STOA) and Compaction...................................78 5.4 Asphalt Pavement Analyzer Procedure................................................................79 5.4.1 Surface Profile Measurement.....................................................................81 5.4.2 APA Hose Testing Procedure.....................................................................82 5.4.3 APA Loading Strip Testing Procedure.......................................................83 5.5 Summary...............................................................................................................83

PAGE 7

vii 6 DATA ANALYSIS METHODOLOGY....................................................................85 6.1 Overview...............................................................................................................85 6.2 Digitizing the Measured Profile............................................................................85 6.3 Rut Depth Calculations.........................................................................................87 6.4 Area Calculation...................................................................................................89 6.5 Summary...............................................................................................................92 7 APA TEST RESULTS...............................................................................................93 7.1 Overview...............................................................................................................93 7.2 Field Results.........................................................................................................93 7.3 Absolute Rut Depth..............................................................................................95 7.4 Differential Rut Depth..........................................................................................95 7.5 Rut-Depth Findings..............................................................................................99 7.6 Area Change.......................................................................................................100 7.7 Unidirectional Loading.......................................................................................101 7.8 HVS Mixture......................................................................................................104 7.9 Discussion...........................................................................................................106 8 CONCLUSIONS AND RECOMMENDATIONS...................................................110 8.1 Conclusions.........................................................................................................110 8.2 Recommendations...............................................................................................110 APPENDIX A AGGREGATE AND MIXTURE VOLUMETRIC PROPERTIES.........................111 B STATISTICAL ANALYSIS RESULTS..................................................................115 C APA RESULTS........................................................................................................122 REFERENCE LIST.........................................................................................................126 BIOGRAPHICAL SKETCH...........................................................................................131

PAGE 8

viii LIST OF TABLES Table page 3-1. Material properties used in the tire FEM...................................................................35 4-1. Material properties and layer th icknesses of FEM pavement structure.....................63 5-1. Field location of selected mixtures............................................................................71 5-2. Aggregate types and sources fo r the selected FDOT mixtures.................................72 5-3. Aggregate sources and modified blends for the laboratory mixtures........................73 6-1. Example of a digitized deformation profile from one location.................................87 7-1. Field rutting data....................................................................................................... .94 A-1. Gradations and specific grav ity of aggregates for Project 1...................................112 A-2. Gradations and specific grav ity of aggregates for Project 7...................................112 A-3. Gradations and specific gravity of aggregates for the HVS mixture......................113 A-4. Batch weight for Project 1......................................................................................113 A-5. Batch weight for Project 7......................................................................................114 A-6. Mixture volumetric properties................................................................................114 B-1. Statistical analyses for Project 1 and Project 7 results at 7% AV with the new APA loading device at 64C..................................................................................116 B-2. Statistical analyses for Project 1 and Project 7 results at 7% AV with the new APA loading device at 70C..................................................................................117 B-3. Statistical analyses for Project 1 and Project 7 results at 7% AV with the original APA loading device at 64C..................................................................................118 B-4. Statistical analyses for Project 1 and Project 7 results at 4% AV with the new APA loading device at 64C..................................................................................119

PAGE 9

ix B-5. Statistical analyses for Project 1 and Project 7 results at 4% AV with the new APA loading device at 70C..................................................................................120 B-6. Statistical analyses for Project 1 and Project 7 results at 4% AV with the original APA loading device at 64C..................................................................................121 C-1. APA test results for Project 1..................................................................................123 C-2. APA test results for Project 7..................................................................................124 C-3. APA test results for the HVS mixture.....................................................................125

PAGE 10

x LIST OF FIGURES Figure page 2-1. Schematic of consolidation rutting...............................................................................9 2-2. Schematic of instability rutting..................................................................................10 2-3. Three-dimensional vertical and latera l contact stress distri butions under radial (R22.5) truck tire at rated load ................................................................................12 2-4. Schematic of the Smithers system used to measure tire contact stresses..................13 2-5. Structural characteristic s of bias-ply and radial truck tires and their effects on the pavement surface......................................................................................................15 2-6. Transverse contact shear stresses meas ured for a bias-ply, radial, and wide-base radial tire at the a ppropriate rated load and inflation pressure.................................16 2-7. Vertical contact stresses measured for a bias-ply, radial, and wi de-base radial tire at the appropriate rated load and inflation pressure.................................................17 2-8. The SST test chamber................................................................................................20 2-9. The Hamburg wheel-tracking machine.....................................................................23 2-10. The French pavement rutting tester.........................................................................25 3-1. Contact imprints of the rubbe r hoses with asphalt beam sample..............................31 3-2. Structural characteri stics of a radial tire....................................................................33 3-3. Schematic cross-section of a typical radial tire.........................................................34 3-4. Finite element representation of th e tread structure of a radial tire...........................34 3-5. Measured and predicted vertical stre ss distribution at surface of steel bed...............36 3-6. Measured and predicted transverse st ress distribution at surface of steel bed..........37 3-7. Schematic of the loading strip...................................................................................38 3-8. Tekscan pressure measurement system.....................................................................40

PAGE 11

xi 3-9. Graphical interpretation of verti cal stresses under the pressurized hose...................41 3-10. Finite element mode l of the loading strip................................................................42 3-11. Contact stress measuring a pparatus setup and calibration.......................................44 3-12. Close-up picture of th e pressurized hose test..........................................................45 3-13. Vertical stress distributi on under the pressurized hose...........................................46 3-14. Close-up picture of the loading strip test.................................................................47 3-15. Vertical stress distribut ion under the loading strip..................................................49 3-16. Lateral stress distributi on under the loading strip...................................................50 4-1. Load configuration used in BISAR to represent measured stresses under bias-ply truck tire ..................................................................................................................53 4-2. Load configuration used in BISAR to re present measured stresses under radial truck tire...................................................................................................................54 4-3. Maximum shear stress distribution............................................................................55 4-4. BISAR sign convention and maximum shear stress angle .....................................56 4-5. Schematic of the maximum shea r stress direction representation.............................56 4-6. Magnitude and direction of maximum sh ear stresses under radial tire load.............58 4-7. Magnitude and direction of maximum sh ear stresses under bias-ply tire load.........59 4-8. Three-dimensional finite element mesh used in the pavement response analysis.....61 4-9. Plan view of the contact area of the th ree-dimensional mesh used in the pavement response analysis......................................................................................................62 4-10. Definition of the shape functions.............................................................................64 4-11. Cross-section view of surface elements w ith nodal forces for the radial-tire load........................................................................................................................... 64 4-12. Maximum shear stress magnitude (psi) and direction under the modeled radial-tire load..........................................................................................................65 4-13. Top view of the finite element model for the APA mold and specimen.................67 4-14. Three-dimensional finite element model for the APA mold and specimen............67

PAGE 12

xii 4-15. Maximum shear stress magnitude (psi) and direction under the modeled loading strip load...................................................................................................................69 4-16. Maximum shear stress magnitude (p si) and direction under the modeled pressurized hose load...............................................................................................69 5-1. Gradation chart for JMF and labora tory blend for Project 1 (9.5mm maximum nominal size)............................................................................................................74 5-2. Gradation chart for JMF and labora tory blend for Project 7 (12.5mm maximum nominal size)............................................................................................................75 5-3. Gradation chart for laboratory blend for the HVS coarse-graded mixture (12.5mm maximum nominal size)...........................................................................................76 5-4. Pine Gyratory Compactor..........................................................................................78 5-5. Original APA measuring plate..................................................................................80 5-6. New measuring plate with elongated slits.................................................................80 5-8. Recording the deformed shape of the contour gauge................................................82 6-1. Grafula3 screen shot..................................................................................................86 6-2. Deformation profile for a specimen tested with the pressurized hose.......................88 6-3. Area change interpretation.........................................................................................89 6-4. Initial surface profile and area calculation................................................................91 6-5. Final surface profile and area calculation..................................................................91 7-1. Measured field rut depth per million ESAL..............................................................94 7-2. Absolute rut depth measurements for Pr ojects 1 and 7 with the two loading devices at 4% AV.....................................................................................................96 7-3. Absolute rut depth measurements for Pr ojects 1 and 7 with the two loading devices at 7% AV.....................................................................................................97 7-4. Differential rut depth measurements for Projects 1 and 7 with the two loading devices at 4% AV.....................................................................................................98 7-5. Differential rut depth measurements for Projects 1 and 7 with the two loading devices at 7% AV.....................................................................................................99 7-6. Area Change measurements for Projects 1 a nd 7 with the two loading devices at 4% AV................................................................................................................102

PAGE 13

xiii 7-7. Area Change measurements for Projects 1 a nd 7 with the two loading devices at 7% AV................................................................................................................103 7-8. Unidirectional loading in the APA..........................................................................104 7-9. Absolute rut depth measurements for the HVS mixture with the two loading devices at two AV leve ls – 4% and 7%.................................................................105 7-10. Differential rut depth measurements for the HVS mixture with the two loading devices at two AV leve ls – 4% and 7%.................................................................106 7-11. Schematic of the initia l hose-specimen contact area.............................................107 7-12. Schematic of a hypothetical hose-sp ecimen contact area after 4000 cycles.........108

PAGE 14

xiv Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy IDENTIFICATION OF A PHYSICAL MODEL TO EVALUATE THE RUTTING PERFORMANCE OF ASPHALT MIXTURES By Christos Andrea Drakos August 2003 Chair: Reynaldo Roque Cochair: Bjorn Birgisson Major Department: Civil and Coastal Engineering Near-surface rutting has become a costly m ode of failure for interstate highways. Even though the pavement is structur ally sound, the hazard of water ponding and hydroplaning urges the transportation agencies to rehabilitate the deformed sections. This study was intended to identify a physical model that can provide reliable predictions about a mixture’s ability to resist permanent deformation. Analyses performed with the elastic la yer analysis program BISAR and the FEM code ADINA provided information on the paveme nt’s response to actual tire loading. The analyses provided evidence that stress st ates in the pavement are dependent on tire structure. Furthermore, it was found that ra dial truck tires induce severe near-surface stress states that have been identified as key factors in the mechanism of instability rutting. The APA is a laboratory torture test that subjects a specimen to an accelerated loading sequence. The end result (rut depth) can be then correlated to the rutting

PAGE 15

xv performance of the mixture in the field. Howeve r, the ability of the te st to replicate field conditions in the laboratory determines the reliab ility of the results. It was shown that the APA loading mechanism, the pressurized hos e, was not capturing the critical lateral stresses found to be detrimental to HMA pavements. Based on the tire study results, a new APA lo ading device was introduced to better replicate the stresses found under radial tires Contact stress measurements under the two loading devices – pressurized hose and loading strip – showed that the loading strip was able to reproduce the lateral stresses found under individual ribs on a radial tire tread. Subsequent finite element modeling also s howed that the loading strip appeared to generate similar shear stress pa tterns to those found under th e modeled radial-tire load. A new method was developed to measure de formations, where a contour gauge is used to record and store the entire surface profile of the sample throughout the progress of the test. The area-change parameter was introduced to calculate the volumetric changes in the sample. Based on the area-cha nge parameter we can calculate whether the specimen is failing primarily due to shear instability or because of excessive consolidation. The introduction of the new loading devi ce and the area-change parameter provided valuable information about the mixtures be havior. Test performed at low air void content, to control consolida tion rutting, showed that both lo ading devices – loading strip and pressurized hose – were ab le to provide accurate pred ictions about the mixture’s susceptibility to instability rutting.

PAGE 16

1 CHAPTER 1 INTRODUCTION 1.1 Background A major distress mode of flexible pavements is permanent deformation, also known as rutting Rutting is characterized by a depression that forms in the wheel paths and can be the result of permanent reduction in vol ume (consolidation/traffic densification), permanent movement of the material at consta nt volume (plastic defo rmation/shear), or a combination of the two. This mode of failu re reduces serviceability and creates the hazard of hydroplaning because of accumu lated water in the wheel-path ruts. Rehabilitation of rutted pavements usually involves asphalt concrete (AC) overlay, recycling, or replacement of all structural layers. The Superpave™ mix design and analysis method was developed more than a decade ago under the Strategic Highway Resear ch Program (SHRP) [Leahy et al. 1994]. Many agencies in North America – including th e Florida Department of Transportation – have adopted the Superpave method of perf ormance-grade (PG) binder specification and the volumetric mixture design method. Although the Superpave volumetric design procedure has resulted in some improvements over the Marshall method of mixture design, it is still devoid of a general strength test that would determine the mixture’s suitability for resistance to rutting and crack ing. The industry has expressed the need for a simple ‘pass–fail’ type of test to co mplement the Superpave volumetric mix design method, especially for use on design–build or warranty projects.

PAGE 17

2 Numerous performance prediction models – numerical and physical – have been implemented to classify an asphalt mixture’s ability to resist rutti ng. In an effort to control this type of distress, many institutions and agencies are searching for a simple performance test that would indicate the rutting potential of hot-mix asphalt (HMA). For this purpose the suitability of various loaded-w heel testers (LWT), as a physical model, is being examined throughout the country. The LWTs provide an accelerated performance evaluation by subjecting the designed mi x to repeated loading under various environmental conditions (moisture and temper ature). Some of the most popular devices used are the Georgia Loaded Wheel Tester (GLWT), Asphalt Pavement Analyzer (APA), Hamburg Wheel Tracking Device (HWTD), a nd the French Pavement Rutting Tester (FPRT) [Cooley et al. 2000, Federal Hi ghway Administration (FHWA) 1998]. A series of studies on the suitability of wheel testers to predict the rutting performance of asphalt mixtures gave mixed results [Epps et al. 1997, Bonaquist et al. 1998]. These types of physical models are cla ssified as empirical or performance-related tests because they do not measure a fundamental property that can be used to explain and identify the mechanisms resulting in surface distress. The models’ accuracy relies on how well (realistically) conditions have been simulated in the lab. Pavement performance has been negatively in fluenced by the change in traffic type and volume in recent years. More than 98% of trucks are now using radial tires that can withstand higher inflation pressures and higher loads. Studies have shown that radial tires induce high lateral stresses that cause tensile stress es at the surface of the pavement [Myers 1997]. Furthermore, it was shown that tire structure has a si gnificant influence on contact stresses; in fact, stre ss states induced by radial a nd wide-base radial tires were

PAGE 18

3 determined to be potentially more detrimental to the pavement’s surface than stress states induced by bias ply tires [Myers et al. 1999] Thus, it is important that the load characterization in the physical model captures the complexity of the load experienced in the field. In most LWTs, the loading device – in the form of a wheel or a pressurized hose – is tracked back and forth over a testing sample to induce rutting. The load follows the same path in both directions (no wander). In recent tests with the Heavy Vehicle Simulator (HVS), researchers have found that there is a differen ce in the rate of accumulated permanent deformation between lo ading in unidirectional and bi-directional mode [Harvey and Popescu 2000, Tia et al. 2001]. The unidirectional loading had a higher rate of rutting (reached failure in fewer cycles) over the same pavement sections than the bi-directional loading. 1.2 Problem Statement Factors that influence the amount of rutting or contri bute to the pavement’s resistance to failure have not been clearly identified. Theref ore, the lack of confidence in existing physical models has inhibited thei r application for prediction of pavement rutting. This study will evaluate the im portance of capturing the specific loading, environmental, structural, and constructi on characteristics on the development of a reliable physical model. Ther e is a need to identify this set of conditions – realistic contact stresses (loading), thermal gradient s in asphalt layer (environment/structural), and test specimen conditioning (construction) – that may dictate performance. Traditionally, mixture evaluation includes average conditions that do not typify the critical condition or capture the key factors that le ad to this kind of failure.

PAGE 19

4 1.3 Hypothesis Rutting performance of mixtures may be governed by a set of critical field conditions – compaction, load, temperature, a nd sequence of loading. Therefore, it is essential to identify the critical design conditi on (mechanism) that must be replicated in a laboratory physical model. A physical mode l that will employ th is set of critical conditions might produce more reliable resu lts for mixture rutting performance. 1.4 Objectives The main objective of this study is to identify the loading, environmental, and construction (density) factors that are critical/essential in defining the mechanism of rutting. The identification of these conditions will logically lead to the development of a reliable physical model. The current version of the APA was selected for the experimental program; necessary modifications will be made to incorporate new testing procedures, which more realistically simu late traffic and environmental conditions existing on pavements. The primary objectives of this re search study are listed below: Identify the characteristics of a loading device necessary to represent a tire load more realistically. Design and construct a new loading device to induce more realistic contact stresses. Verify the effects of loading charact eristics on rutting performance. Examine the importance of unidirectional versus bi-directional loading in the physical model. Evaluate the importance of density/loadi ng history on rutting performance. Investigate the sensitivity of the physical model to mixtur es with different densities as produced by compaction and/or aggr egate gradation of the mixtures.

PAGE 20

5 Recommend test configurati on and procedure/system fo r mixture evaluation. As envisioned, the procedure will define the magnitude and sequence of loading as well as test-temperature requirements. 1.5 Scope The research focuses on identifying the crit ical conditions that contribute to the mechanism(s) of rutting. Defining the conditi ons that might initiate and propagate rutting will lead to the development of better pe rformance prediction models – physical and numerical. However, it will not be feasible to examine all possible parameters that affect rutting within the limited time. Thus, this research will focus on the effects of the following: Load configuration. A new loading devi ce (rib) will be evaluated against the existing pressurized hose. The contact stresses will be measured for both devices and then used in finite element modeli ng (FEM) to calculate the induced stress states in the specimen. Temperature. Two temperatures – 64 and 70oC – have tentatively been selected for evaluation of mixture’s sensitiv ity to temperature changes. Mixture density. Mixtures will be tested at two levels – 93-94, and 95-96% – of maximum theoretical density (MTD). Three mixtures of known field performance will be used for the initial development and the evaluation/validation of the physical model. The following mixtures were chosen from the Superpave™ monitoring project: Poor field performing mixt ure – I-10 Madison County. Good field performing mixtur e – Turnpike Palm Beach. Good performing mixture – HVS coarse-graded mix. The Pine gyratory compactor will be us ed to prepare 150-mm-diameter by 75-mmthick mixture specimens. In this research, beams will not be considered because of compaction issues and the potential for variab ility that may influence the analysis.

PAGE 21

6 1.6 Research Approach The research is divided into two parts – the analytical and the experimental. The analytical part includes the conceptual design, the finite-e lement analysis (FEA) of contact stresses, and the anal ysis of the stress states w ithin the physical model. Subsequently the experimental part include s the laboratory testi ng with the physical model and data analysis. A research-approach outline is detailed below: Literature Review: examine existing ideas theories, and results published on tire contact stresses, critical stress states, a nd rutting in asphalt co ncrete pavements. Also, review work done on other accelera ted physical models – LWT, Accelerated Loading Facilities (ALF), and HVS – and th e effects of certain parameters – load, temperature, and density – on the reported results. Tire Contact Stresses: desi gn a loading device to induce stresses that would be representative of the actual stresses induced by truck tires, based on the tire contact stresses studies. New Loading Device: construct and test th e new loading device for compatibility and durability issues. Measure the contact stresses under both loading devices – rib and pressurized hose – and compare the st ress distribution. Analyze the measured contact stresses with FEM to calculate th e stress states in the test specimen, and compare the results for the two loading devices. Validate New Loading Device: test two mixtures with known field performance – good and poor – in the modified APA and eval uate the new loading device’s ability to produce reliable results. Compare Loading Devices: test mixtures with known field performance that the existing APA failed to correctly predict. Compare the results from the modified APA to the field performance of the mixtures. Evaluate Factors Affecting Performance Prediction: 1) examine the effects of unidirectional versus bi-directional loading, 2) test at different temperatures and densities to obtain temperature-density trends for rut depth development, 3) examine the effects of sequence of loading. Analyze the test data and establish the test conditions that ar e critical in predicting rutting performance. Validate Proposed Test C onditions: test mixtures with known performance and evaluate the reliability of the results.

PAGE 22

7 CHAPTER 2 LITERATURE REVIEW 2.1 Overview In recent years permanent deformation in HMA pavements has generated much concern as many states have experienced an in crease in the severity and extent of this type of failure. Research sugge sts that the steady increase in truck-tire pressure and axle load, which has been noted for the past 20 y ears, altered the tire-pa vement contact stress characteristics. As a result, the pavement surface is exposed to higher stresses than the levels assumed when designing in accord ance with the 1993 AASHTO Guide for Design of Pavement Structures. Concern about high truck-tire pressures and rutting led to a national symposium and sparked interest on tire-p avement interface stresses. Subsequent research concluded that increased truck weights and tire pressures had led to an increase in pavement distress. Measurements of the tire-pavement contact stresses helped identify possible reasons behind surface-initiated crac king and near-surface rutting. Focus also shifted to material selectio n, mix design, and c onstruction practices improvements that could minimize rutting. One of the attempts to improve HMA mixture performance was the development a nd subsequent adoption of the Superpave Volumetric Mixture Design Me thod. Although the Superpave design procedure resulted in some improvements over previous mixture design methods, it has yet to incorporate a general strength test that would determine the mixture’s ability to resist rutting and cracking. The need for a simple ‘pass–fail’ type of test to complement the Superpave

PAGE 23

8 volumetric mix design method forced various agencies to search for a suitable performance test. This chapter reviews some of the literature available on the subjects of permanent deformation, tire-pavement interface stre sses, accelerated pavement testers, and laboratory methods for predicting mixture perf ormance. The literature review focuses more on the three most popular torture te sts – HWTD, FPRT, and APA – with more weight on the leading candidate, th e Asphalt Pavement Analyzer. 2.2 Permanent Deformation Permanent deformation, also known as rutting is an unrecoverable deformation in the form of a depressed channel in the wh eel path of the pavement. Rutting can be attributed to excessive consolidation, formed by an accumulation of permanent deformations caused by repeated heavy loads, or lateral movement of the material, caused by shear failure of the asphalt concrete layer, or a combination of the two. 2.2.1 Consolidation Rutting Consolidation rutting occurs when one or more layers of the pavement structure (usually the subgrade) undergo further densif ication by reduction of ai r voids, or loss of moisture in the case of clay soils. The structur e is especially susceptible to this type of distress when there is insufficient compac tion during pavement construction. A layer with insufficient density is prone to further densification under traffic, especially in hot weather (for asphalt concrete layers) and at intersections where the loads are slowmoving or static. Figure 2-1 shows a schema tic of consolidation rutting which is distinguished by a depression that occurs in the wheel path with no humps on either side of that depression [Huber 1999].

PAGE 24

9 Figure 2-1. Schematic of consolidation rutting. The subgrade (native soil) is the most vulnerable layer for consolidation rutting because it is the weakest material of the paveme nt structure. If the pavement structure fails to reduce the vertic al stress/strain to allowable lim its on the subgrade level, either by improper thickness design or by unexpected in crease of load magnitude, under repeated loading the layer will experience excessi ve consolidation [Huang 1993]. Once the foundation (subgrade) has collapsed, the rema ining yielding layers in the pavement structure will conform to the new contour shape of the supporting layer, resulting in subsidence ruts. These ruts tend to be fair ly wide (30 40 in) with a shallow sloping saucer-shaped cross section [Huber 1999]. 2.2.2 Instability Rutting Instability rutting is strictly an AC layer type of failure, usually within two inches from the surface, and it is attributed to the mixture characteristics of the HMA. The surface material is laterally displaced along shear planes wi thin the AC layer, which shows signs of mixture instabil ity (low shear resistance). Figure 2-2 shows a schematic

PAGE 25

10 of instability rutting which is characterized by longitudinal ruts in the pavement with humps of material on either side of the rut [Huber 1999]. Figure 2-2. Schematic of instability rutting. Asphalt concrete is a bonded granular ma terial mix of aggregates and asphalt cement. Under repeated traffic loads, the aggregates do not deform but rather rigidly translate and rotate within the asphalt binder [Wang et al. 1999]. This effect (aggregate movement) is amplified at low initial compac tion or at high asphalt content. Instability rutting is the surface manifestation of the aggregate skeleton evolution under repeated traffic loading of mixtures with low shear re sistance. Temperature, rate of loading, and magnitude of loading directly affect the performance of the mix and influence the severity of this type of distress. 2.3 Tire-pavement Interface Stresses Engineers have tried for many years to measure the three-dimensional contact stresses between a tire and the pavement. Th e results of such measurements would help tire engineers to design better, more resilien t tires, and would al so enable pavement engineers to analyze the stress states under a tire and evaluate their influence on the structure.

PAGE 26

11 Woodside et al. (1992) developed a devi ce to measure the contact stress patch between the tire and underlying material in th e laboratory. A steelbed transducer array measured normal and tangential contact stresses in static or dynamic mode, under radial tires. The steel plate system was fitted with 12 transducer s and repeatedly measured a strip transversely every 5 mm over the entire contact patch. The device recorded 90 contact stress measurements for each test. Th e contact stresses were then used in the evaluation of durability of su rface chippings on asphalt overlay s. Results concluded that implementation of surface chippings of 1 mm macro-texture may improve skid resistance on pavements. Researchers in South Africa measured th ree-dimensional stresses under bias-ply, radial, and wide-based ra dial truck tires at different load s and inflation pressures with the Vehicle-Road Pressure Transducer Array (V RSPTA) System [De Beer et al. 1997]. The experimental setup consisted of 13 triaxial strain gauge steel pins (spaced 17mm transversely) mounted on a steel plate and fixe d flush with the road surface. Figure 2-3 shows some of the VRSPTA results that rev eal a non-uniform vertical and lateral stress distribution. Dr. Marion Pottinger of Smithers Scientif ic Services, Inc. (Ravenna, Ohio) developed a device to measure the tire-pavemen t contact stresses unde r truck tires. The device measures vertical, transverse, and l ongitudinal forces and displacements under a moving tire using a series of 16 transducers, as shown in Figur e 2-4. The tire is held at one location while the steel bed is moved in th e longitudinal direction, forcing the tire to roll over the transducers that measure disp lacements and stresses. The procedure is repeated by placing the tire at different transv erse positions to acquire a detailed pattern

PAGE 27

12 of the three-dimensional stresses under the ti re. Measurements taken at every 0.14 inches in the transverse and 0.10 inches in the longitudinal direction resulted in about 4000 stress measurements in each of the three ax es under the tire-contact area [Myers et al. 1999]. Vertical contact stress distribution. Lateral contact stress distribution. Figure 2-3. Three-dimensiona l vertical and lateral cont act stress distributions under radial (R22.5) truck tire at rated load [after De Beer et al. 1997].

PAGE 28

13 1 6 T r a n s d u c e r s B e d M o t i o n T i r e R o l l i n g D i r e c t i o n B e d z ,zy ,y x ,x T i r e C o a x i a l L o a d a n d D i s p l a c e m e n t T r a n s d u c e r D e t a i l 1 6 T r a n s d u c e r s B e d M o t i o n T i r e R o l l i n g D i r e c t i o n B e d z ,zy ,y x ,x T i r e C o a x i a l L o a d a n d D i s p l a c e m e n t T r a n s d u c e r D e t a i l Figure 2-4. Schematic of the Smithers system used to measure tire contact stresses. Analysis of Pottinger’s results gave new information about the way tires load the pavement and the factors that influence the co ntact stress distribution. It appears that the most important factor is the tire structure. Bias-ply tires have a more flexible tread that allows the tire to bulge out upon inflation. When the tire is loaded the contact area becomes flat and the bulging is reversed, causing transverse stresses to pull the pavement towards the center of the tire. This is commonly referred to as the pneumatic effect and it is more prevalent in bias-ply tires. A phenomenon referred to as the Poisson effect has a significant bearing on transverse shear stresses. This effect is based upon the principle that, unless restrained, most materials expand laterally when loaded ve rtically. When indivi dual ribs under a tire are loaded they attempt to expand laterally, and transverse stresses are generated when the surface of the pavement tries to restrain th e expansion. This effect induces transverse shear stresses that pull the pavement apart under each rib.

PAGE 29

14 The Poisson effect, which induces tension under each tire rib, is present in both radial and bias-ply tires. However, because of the large pneumatic effect, tensile stresses are eliminated in all but the cen ter rib of the bias-ply tire. The tensile stresses from the Poisson effect dominate the transverse contac t stress distribution in radial tires because the pneumatic effect is negligible. The co mposite of the pneumatic and Poisson effects for each tire type is shown schematically in Figure 2-5 [Roque et al. 1998]. Recent work at the University of Flor ida focused on the effects of tire type, loading, and inflation pressure on measured c ontact stresses under va rious types of truck tires – radial, bias-ply, and wi de-base radial. Stress states induced by radial and widebase radial tires were determined to be pot entially more detrimental to the pavement’s surface than stress states induced by bias-ply ti res. The tire structure caused a significant difference in the distribution of lateral cont act stresses under radial and bias-ply tires, whereas no significant difference was found in the vertical stress distribution. Distributions of lateral and vertical contact stresses unde r the three tire types are shown in Figures 2-6 and 2-7 respectively. Th e studies concluded th at the contact stress distributions measured under ra dial truck tires appear to c ontribute to the prevalence in recent years of surface-initiated wheel path cracking and near-surface rutting [Drakos et al. 2001, Myers et al. 1999].

PAGE 30

15 Radial Tire Bias Ply Tire More Rigid Tread Flexible Wall Flexible Tread More Rigid Wall Composite Effect Poisson Effect Pneumatic Effect Radial Tire Bias Ply Tire More Rigid Tread Flexible Wall More Rigid Tread Flexible Wall Flexible Tread More Rigid Wall Flexible Tread More Rigid Wall Composite Effect Poisson Effect Pneumatic Effect Figure 2-5. Structural characteri stics of bias-ply and radial truck tires and their effects on the pavement surface [after Roque et al. 1998].

PAGE 31

16 -80 -60 -40 -20 0 20 40 60 80 02468101214Lateral Location, X (in)Transverse Contact Shear Stresses, yy (psi) Wide-base Radial Radial Bias ply Figure 2-6. Transverse contact shear stresses measured for a bias-ply, radial, and widebase radial tire at the appropriate rated load and inflation pressure [after Myers et al. 1999].

PAGE 32

17 -300 -250 -200 -150 -100 -50 0 02468101214Lateral Location, X (in)Vertical Cont act Stresses, zz (psi) Wide-base Radial Radial Bias ply Figure 2-7. Vertical co ntact stresses measured for a bias-p ly, radial, and wide-base radial tire at the appropriate rated load and in flation pressure [after Myers et al. 1999].

PAGE 33

18 2.4 Predicting Mixture Performance There are three general type s of laboratory mixture de sign and evaluation tests – performance related, performance based, and torture tests. Performance-related tests measure properties or responses related to mixture performanc e (density, air voids, etc.) but individually are insufficient to drive a performance prediction model. Performancebased tests measure material properties (resil ient modulus, complex modulus, etc.) that can be used in fundamental response models to predict mixture response to imposed truck and environmental loads. Torture tests, also known as index tests are empirical and apply very severe or extreme loading conditi ons on the test specimen to evaluate a type of failure condition in the mix [FHWA 1998]. As stated above, there is a growing need for a simple performance test (SPT) that can accurately predict the mixture’s ability to resist rutting and cracking. M.W. Witczak defined the SPT as follows: A test method(s) that accurately and reliably measures a mixture response characteristic or parameter that is highly correlated to the occurrence of pavement distress (e.g., cracking and rutting) over a diverse range of traffic and climatic conditions. [Witczak et al. 2002, p.1] Based on the definition above, it is not neces sary for the SPT to predict the entire distress or performance history of the HMA mixture, but the te st results must indicate the mixture’s ability to resist fracture and permanent deforma tion under defined conditions. Various agencies are conducting evaluation studie s to identify the most suitable test that would accompany the volumetric design. The following sections discuss some of the most popular candidates: three torture tests and the performance-based Superpave Shear Tester.

PAGE 34

19 2.5 Accelerated Pavement Testers The use of accelerated pavement test ing (APT) for determining mixture performance has increased in the past 20 year s because of APT’s abil ity to apply traffic loads in a short time. APT facilities enable us to evaluate potential materials, designs, and features, in a ‘real’ (actual size) environment. In a review of existing APT facilities, J.B. Metcalf defined the APT as: Full scale accelerated pavement testing is de fined as the controlled application of a prototype wheel loading, at or above the appropriate legal load limit to a prototype or actual, layered, structural pavement system to determine pavement response and performance under a controlled, accele rated, accumulation of damage in a compressed time period. [Metcalf 1998, p. 556] The appeal of the APT facilities is that they give the closest simulation to real condition of an actual in-service pavement in terms of materials and construction procedures, though the effects of aging and th e environment are in most cases limited. Most APT facilities use load systems that approximate actual traffic by incorporating wheel wander and unidirectional loading capabil ities (the wheel does not load in both directions). These facilities are able to work around the clock and produce early results with a high level of credibility. In the search for a simple performance te st, accelerated pavement testing facilities provide reliable mixture-performance informa tion to evaluate results from various LWTs [Bonaquist et al. 1998, Epps et al. 1997, Romero and Stuart 1998]. These results can be used to evaluate the LWT’s ability to capture the true performance characteristics of the HMA.

PAGE 35

20 2.6 Superpave™ Shear Tester The Superpave™ Shear Tester ( SST), previously known as the Simple Shear Tester was produced by the Strategic Highway Re search Project (SHRP) that measures mixture properties to determine pavement pe rformance. The SST is a servo-hydraulic system that can apply axial loads, shear loads, and confinemen t pressures to asphalt concrete specimens at controlled temperat ures while monitoring sample deformation [Shenoy et al. 2001]. The machine has six main components: testing chamber, test control system, environmental system, hydrau lic system, air pressurization system, and measurement transducers. Figure 2-8 is a close-up of the testing chamber. Figure 2-8. The SST test chamber.

PAGE 36

21 The specimens have a diameter of 150 mm (5.90 in) and a height of 50 mm (1.97 in); however, the system can test specimen s with diameters and heights up to 200 mm (7.87 in) with only minor modifications. The e nvironmental system is able to precisely maintain the temperature inside the te sting chamber anywhere between 0 and 70oC. Three tests are usually performed with the SST: a) Simple Shear at Constant Height test (SSCH), b) Frequency Sweep at Constant Height test ( FSCH), and c) Repeated Shear at Constant Height test (R SCH). The American Associ ation of State and Highway Transportation Officials (AAS HTO) Provisional Standard TP7-94 contains a detailed description of the SST test in th e different modes of operation. There is an ongoing evaluation of this test procedure to establish the accuracy and repeatability of results. A study by the Federal Highway Administration (FHWA) showed that the SST tests can accurately di scriminate between different asphalt binders, but are insensitive to aggregate changes [FHW A 1998]. Other resear chers have reported high variability (15-30%) and that their re sults depended on the data analysis method [Romero et al. 1998]. An evaluation study performed in North Ca rolina on three paveme nt sections with known performance showed that the SST was able to rank the mixtures according to their field performance [Tayebali et al. 1999]. A recent study used the SST and the HWTD to test two mixtures – Superpave and Marshall – for rutting and stri pping susceptibility. Both tests – SST and HWDT – gave consiste nt results and showed that the Superpave mix was more resistant to perman ent deformation [Wang et al. 2001].

PAGE 37

22 2.7 Torture Tests Torture testing devices subject a test specimen of the mi xture to repeated loading applied by a traveling wheel. The tests do not require any prepara tion other than the molding of the specimen, and the results us ually report the rut depth on the mixture specimen as a function of the lo ad applications. Some of th e most popular torture testing devices include the APA, the HLWT and th e FPRT, all of which use the same basic principle that the rut depth measured from the specimen can be correlated to field performance. Although the idea is simple and seems to pi npoint a direct measurement that can be an indication of the mixture’s performance, r ecent studies have shown that the tests failed to distinguish between good and poor performi ng mixtures [Collins et al. 1996, Stuart et al. 1997]. The load-transfer mechanism, boundary conditions (confinement), and the ratio of the size of the loading wheel to the aggregate size are some of the reasons that limit the success of LW Ts [FHWA 1998]. Overall, the idea of the loaded wheel tester has a very intu itive appeal because it is simple, relatively low cost, and easy to use. However, the tests correlate the rut-depth measurement directly to field performa nce without identifying any fundamental properties that will help us improve our mixture design. 2.7.1 Hamburg Wheel-Tracking Device The Hamburg wheel-tracking device measur es the combined effects of rutting and moisture damage by rolling a steel wheel acro ss the surface of an as phalt concrete sample immersed in hot water. Esso A.G. of Ha mburg, Germany developed the device in the 1970, and originally call ed it the “Esso Wheel-Tracking Devi ce.” Initially, the City of Hamburg used the testing machine to meas ure rutting susceptibility of HMA. The

PAGE 38

23 sample was subjected to 9,540 wheel passes at 40 or 50oC, and the resulting rut depth was used as a pass/fail criterion to ensure mixture performance [City of Hamburg 1991]. Today, the Hamburg testing machine is manufactured by Helmut-Wind, Inc. of Hamburg, Germany. It can test two as phalt slab samples simultaneously with dimensions of 320 mm in length by 260 mm in width, and thickness of 40, 80, or 120 mm. The samples are prepared to 7 1 % air voids, and are submerged into 50oC water for the test. A steel wheel, measuring 203.5 mm (8 in) in diameter and 47 mm (1.85 in) in width, loads each sample with a fixed load of 705 22 N (158 5 lbs) at a rate of 50 passes per minute [Aschenbrener, Texas Depa rtment of Transportation]. A linear variable differential transformer, with an accuracy of 0.01 mm, continuously measures the rut depth in each slab and prints out the data every 20, 50,100, or 200 wheel passes. Figure 2-9 is a picture of the Hamburg wheel-t racking device without the data-acquisition machine. Figure 2-9. The Hamburg wheel-tracking machine.

PAGE 39

24 Various institutions use different criteria and testing procedures to evaluate mixes in the Hamburg device. The City of Hambur g uses a maximum allowable rut depth of 4 mm at 19,200 wheel passes, whereas the Colo rado Department of Transportation (CDOT) recommends maximum allowable ru t depths of 4 mm at 10,000 and 10 mm at 20,000 cycles [Aschenbrener]. In an evaluation study at the FHWA, resear chers tested the rutting susceptibility of certain mixtures with the full-scale ALF and compared the ALF results to the predicted performance from the HWTD. The study noted that the trac king device could discriminate mixtures with wi dely different binder grades, bu t failed to give consistent results for mixtures with closer grade binders. Furthermore, the research showed that the HWTD was unable to predict a d ecrease in rutting susceptibil ity for mixtures with altered gradation (maximum nominal aggregate size), even though test results from the ALF clearly showed less ruttin g [Stuart and Mogawer 1997, St uart and Izzo 1995]. 2.7.2 French Pavement Rutting Tester The French Pavement Rutting Tester te sts slabs of HMA to evaluate their resistance to permanent deformation. The machine, shown in Figure 2-10, uses an environmental chamber to keep the test temperature at 60oC and loads the sample with a smooth, reciprocating, pneumatic rubber tire inflated to 0.60 0.03 MPa (87 4 psi). Similar to the Hamburg tester, the FPRT can test two slabs simultaneously measuring 500 mm (19.7 in) long, 180 mm (7 in) wide, and 50100 mm (1.97-3.93 in) thick. Hydraulic jacks push the slabs upward to apply th e 5,000 50 N (1,124 11 lb ) load [Cort and Serfass 2000].

PAGE 40

25 Figure 2-10. The French pavement rutting tester. The test sequence includ es 1,000 cycles at 150-25oC to simulate traffic densification and to take initial slab thickness measurements. The slabs are then conditioned for 12 hours at 60 2oC before the start of the test The average rut depth in each slab is measured manually at 3 0, 100, 300, 1,000, and 3,000 cycles when testing 50mm slabs, and at 300, 1,000, 3,000, 10,000, and 30,000 cycles when testing 100-mm slabs. Finally, the average percent rut depth is calculated based on the initial thickness of the slab [Cort and Serfass 2000]. The FHWA evaluation study mentioned in the section above [Stuart and Mogawer 1997] reported results for the FPRT. Similar to the HWTD, the French pavement rutting tester accurately discriminated mixtures with widely different binde r grades, but lacked precision in identifying mi xtures with distinctly different gradations.

PAGE 41

26 2.8 Asphalt Pavement Analyzer In 1985 the Georgia Department of Tran sportation (GDOT), in association with the Georgia Institute of T echnology (GT), initiated Resear ch Project No. 8503 for the development of a pass/fail laboratory test for the rutting resistance of HMA. The prototype GLWT was a modifica tion of the Benedict Slurry Seal tester, originally designed to test slurry seals [Collins et al. 1995]. The initial version of the GLWT consisted of an aluminum wheel attached to a reciprocating arm moved al ong a pressurized hose, crea ting the desirable contact pressure. Constant temperat ure was maintained during test ing by placing the LWT in an airtight room, where an electric heater with a thermostat was used to heat the room to 95oF. The test was performed with 75or 100-psi hose pressure, and 50-, 75or 100-lb load. Rut measurements were taken at 40, 100, 400, 1000 and 4000 cycles [Lai 1986]. To promote the concept of using GLWT as a supplemental strength test to the Superpave design procedure, the device was m odified in 1992 to be able to evaluate rutting potential of samples prepared by th e Superpave gyratory compactor. The new device had the ability to test six gyratory sa mples simultaneously in an environmentally controlled chamber. Other modifications in cluded operation control, adjustable hose pressure (up to 120 psi), and load (up to 250 lbs) [Collins et al. 1996]. The APA is a further modification of the GLWT, first manufactured in 1996 by Pavement Technology, Inc (PTI). Since it is a new generation of the GLWT, it follows the same testing philosophy. Load is a pplied onto a pressurized linear hose by a pneumatic loaded wheel and tracked back and forth over a testing sample to induce rutting. The APA has the additional capability of testing for moisture susceptibility and fatigue cracking while the specimens are submerged in water.

PAGE 42

27 Extensive studies have been conducted to evaluate the ability of the APA to distinguish between mixtures of known performance. Most of these studies tried to establish a relation between rut depths obtai ned in the laboratory tests and the field performance of the mixture. The Florida DOT has performed a series of tests with the APA. Although the device successfully ranked mixtures according to their rutting potential, some variability from test to test and from location to lo cation was found. It was also reported that gyratory samples and beams rut at statisti cally different levels [Choubane 1998]. To reduce variability, PTI installed new pressure regulators and reconf igured the air supply tubing, but a subsequent study indicated that although variability was decreased the middle position consistently yiel ded in higher rut depths than the left or right positions [Sholar 1999]. A similar study compared test results from WesTrack [Epps et al. 1997] to rutting predictions from three LWT devices. The AP A ranked the mixtures according to their WesTrack performance with 89% accuracy [W illiams et al. 1999]. The National Center for Asphalt Technology (NCAT) indicated that the APA was sensitive to mixtures with different asphalt binder and varying gradat ion (ARZ, BRZ, and TRZ) [Kandhal et al. 1999]. The FHWA also conducted a study at Turner-Fairbank Highway Center. Comparison of LWTs test results to the AL F showed that the LWTs were able to distinguish between mixtures that had good and poor performances when those were prepared with the same aggregate gradation and different binder. When the aggregate gradations were varied, none of the LWTs we re able to separate the mixtures, even

PAGE 43

28 though the ALF testing showed that there we re significant differences in pavement performance [Romero and Stuart 1998]. 2.9 Summary The discussion presented in this chapter indicates that no one SPT can reliably predict HMA performance. The SST, although it is a mechanistic-developed test, is a very expensive and complex machine to operate, making it unsuitable for a simple performance test. From the three torture tests reviewed – HWTD, FPRT, and APA – the Asphalt Pavement Analyzer seems to be the leading candidate. Recent evaluation studies showed that the APA has the potential to accurately rank mixtures according to their field performance. However, the APA proved to be unable to capture the difference in performance for mixtures with altered gradation. Based on this literature review, it is e ssential to identify the critical design condition(s) (mechanisms) that might lead to near-surface rutting. We must attempt to replicate these conditions in a laboratory physic al model (torture test). A physical model that will employ this set of critical conditi ons might produce more reliable results for mixture rutting performance.

PAGE 44

29 CHAPTER 3 APA LOADING MECHANISM 3.1 Overview Tire-pavement interface contact-stress studi es indicated the importance of lateral stresses in the development of critical st ress-states near the surface of the pavement [Drakos et al. 2000, Myers et al. 1999]. These st udies have shown that radial tires induce stresses that are more detrimental to pavement s than bias-ply tires and that the difference has been attributed mainly to tire structure. The theory behind torture tests such as the APA is that a specimen is subjected to an accelerated loading sequence in the laboratory and the end-result (rut depth) can be correlated to the rutting performance of the mixture in the field. However, the ability of the test to replicate field c onditions in the laboratory determines the reliability of the results. The hypothesis was that the APA loading mechanism was not capturing the lateral stresses found under radial tires. The idea that the APA hose could not gene rate these lateral stresses was based on initial observations of the contact area between the sample and the hose which was measured at approximately 8 mm. This limited contact area cannot reproduce the Poison’s effect found under each individual rib on the tire tread. This chapter deals with the effort to develop an alte rnate loading device geared to capture some of the complex stress distributions found in the field.

PAGE 45

30 3.2 Limitations of Loaded Wheel Testers Loaded wheel testers operate on the same basic principle: a test specimen of mixture is subjected to repetitive loading by a traversing wheel, and the surface depression in the sample is then measured and reported as a function of load cycles. These types of torture tests are classified as empirical or performance-related tests because they do not measure a fundamental pr operty that can be used to explain and identify the mechanisms resu lting in surface distress. The APA, like most of the LWTs, attempts to replicate field conditions in a controlled laboratory environment. In this sense, good correlation between results from the APA with field performance relies on how well (realistically) conditions have been simulated in the lab. The following issues rais e some considerations on the ability of the APA to approximate field conditions: Loading scale effects. The loaded area under the pressurized hose is very small (narrow) in proportion to the nominal maximum aggregate size [FHWA 1998, Lai et al. 1990]. Boundary conditions. In the APA the test specimens are resting on a metal plate that limits deflections and increases confinement. Load application. Earlier work [De Beer et al. 1997, Myers et al. 1999] showed that radial truck tires induce high latera l stresses that can cause tension on the surface of the pavement [Drakos 2000]. It is believed that the pressurized hose of the APA does not simulate the effects of the stiff treat of the ra dial tire, thus not inducing any lateral stresses. Lai and Lee (1990) evaluated the stiffness e ffects of the pressuri zed hose by testing asphalt samples with a relatively stiff and a relatively soft hose. Figure 3-1 shows the imprints of the contact area used for compar ison purposes. As expected, under the same load, the stiffer hose generated a more el ongated and narrow contac t area whereas the softer hose produced a shorter and wider contact area.

PAGE 46

31 Figure 3-1. Contact imprints of the rubber hos es with asphalt beam sample [after Lai and Lee 1990]. Although the researchers beli eved that the stiffer hos e would generate greater rutting, a series of tests proved that the soft er hose consistently gave slightly higher rut depths. Unfortunately, there are no direct measurements of the contact stresses between the hoses and the asphalt specimen surface. Nonetheless, this finding is of great importance to the development of the new lo ading configuration because it demonstrates the significance of stress dist ribution to the rutting behavior of samples in the APA.

PAGE 47

32 3.3 New APA Loading Mechanism The concept for a new APA loading mechanism is based on the observations and conclusions from the tire-pavement interface stresses studies. Analyses performed with the elastic layer analysis program BISA R and the finite element program ADINA provided information on the pavement’s re sponse under modeled tires from measured contact stresses. Myers et al. (1999), Drakos et al. (2001) and Birgisson et al. (2002) have identified the lateral st resses induced by radial tires as the fundamental cause of stress reversals (tension) and high magnitude shear stresses near the surface of the pavement. These stress states cause a reduc tion in confinement near the pavement’s surface near the edge of the loaded area, wh ich reduces the resist ance to shear stress within the mixture. The hypothesis was that the stiff pressuri zed hose used by the APA to load the specimen does not reproduce the lateral stre sses found under radial truck tires. The objective was to develop a new loading mechanism, modeled af ter a radial truck tire, to replicate these stress conditions in the APA specimen. 3.3.1 Development of a Tire Model The initial task was to devel op a reasonable tire model that represents the structural behavior and response of a typi cal radial truck tire tread. Earlier work by Roque et al. (2000) showed that the radial tire loading behavior can be simulated with a combination of steel and rubber. This step would enable us to estimate the ri ght amount of steel and rubber needed to built a device that captures the loading be havior of the radial tire. Figure 3-2 shows the structural characte ristics of a radial truck tire with the radial plies and the steel strands that run through the rubber, around the tire.

PAGE 48

33 Figure 3-2. Structural charac teristics of a radial tire. Radial and bias-ply tires are totally different from a struct ural point of view and the actual structural make-up of these tires is proprietary information not available to the general public. However, Smithers Scientif ic Services, Inc. provided some basic response data regarding the beha vior of typical ra dial truck tires a nd their structural characteristics that were used along with a ba sic knowledge of the stru ctural behavior of radial truck tires to develop a two-dimensional model of a ra dial truck tire tread [Myers 2000]. As previously discussed, the structural beha vior of radial truc k tires is governed by a wall structure of very low stiffness and a very stiff tread struct ure resulting from the steel strands used to reinfor ce the tread. The cross section of a typical radial tire is illustrated in Figure 3-3 and shows a tread ar ea that is 8.0 inches wide and 1.44 inches high. The steel reinforcement was concentr ated in an area that is 0.33 inches high, approximately 0.93 inches above the outer surface of the tire. Steel Cable Radial Plies

PAGE 49

34 Figure 3-3. Schematic cross-sect ion of a typical radial tire. Following the guidelines from Roque et al. (2000), the model was constructed with the MSC/Patran pre-processor software and it is illustrated in Figure 3-4. The steel strands were modeled as a so lid strip of steel, and the connection between the steel strands and the rim was modeled as a pin connection at either end of the steel strip used to represent the strands. Tabl e 3-1 presents the modulus and Poisson’s ratio values used for the tire rubber and the st eel strip. Finally, the ABAQUS finite element program was used to run the elastic analysis and retrieve the deformation and stress distributions under the tire model. Figure 3-4. Finite element re presentation of the tread stru cture of a radial tire.

PAGE 50

35 Table 3-1. Material properti es used in the tire FEM. Tire Part Material Elastic Mo dulus, E (psi) Poisson's Ratio, Reinforcing Beads Steel 2.90E+07 0.15 Tire Tread Rubber 1.16E+03 0.48 Tire Grooves Air 9.80E-06 0.49 3.3.2 Verification of the Tire Model As mentioned in Section 2.3, Dr. Marion Pottinger of the Smithers Scientific Services, Inc. successfully measured the tire-p avement contact stresses under truck tires. Pottinger’s device measured vertical, tr ansverse, and longitudinal forces and displacements under a moving tire using a series of 16 transducers on a steel plate. To verify the accuracy of the model, Pottinger’s contact stress measurements were compared to stresses predicted under the FE M tire on a steel foundation. The final thickness of the st eel strip was determined by varying the thickness until the predicted stress response of the modeled tire matched the measured response of the real tire. Thus, the FEM tire matched the overall stiffness and stress-distribution behavior of the tire tread. It was determined that a 0.1-inch-thick steel strip embedded in the modeled tire tread resulted in the same structural response as the steel-strand reinforcement in the actual tire. Figures 3-5 and 3-6 – verti cal and transverse stress distribution respectively – indicate that the tire model pr edicted both vertical and transv erse contact stresses similar to those measured under the real tire. Alt hough there is some variation in magnitude, the tire model accurately captured th e patterns of both the vertic al and transverse contact stress distributions. Figure 3-6 is particularly important because it demonstrates the model’s ability to capture the transverse contact stress reversals under the individual tire ribs. As stated

PAGE 51

36 earlier, these transverse stress es were found to be detrimenta l to the top-down cracking and near-surface rutting performance of HMA. The next step was to build an individual rib replica that would serve as the load transfer mechanism in the APA. -250 -200 -150 -100 -50 0 02468101214 Lateral Location, X (in)Vertical Stress, Z (psi) Measured FEM Predicted Figure 3-5. Measured and predic ted vertical stress distributi on at surface of steel bed.

PAGE 52

37 -80 -60 -40 -20 0 20 40 60 80 02468101214 Lateral Location, X (in)Transverse Stresses, Y (psi) Measured FEM Predicted Figure 3-6. Measured and predicted transverse stress distribution at surface of steel bed.

PAGE 53

38 3.3.3 Concept Loading Strip In the previous section, analyses showed that the FEM tire was able to capture the complex stress distribution measured under a ra dial tire. The idea for the APA loading mechanism was to substitute the pressurized hose with a steel-rubber configuration based on the tire finite element model. Figure 37 shows a schematic of the concept device, called the loading strip where a thin rectangul ar steel plate (14 gauge ) is attached on top of a medium-durometer (45-55) rubber. The so lid steel wheel applies the load on the thin steel plate that distributes the stresses on the sample through the rubber part of the device. Figure 3-7. Schematic of the loading strip. Ideally, the loading strip st ress-distribution behavior would represent that of a single rib from the radial ti re tread. The magnitude of the applied stresses was anticipated to be significantly lower; however, the stress-distribution pattern was expected to be similar. The steel plate would uniformly distribut e the stresses to the rubber and also increase the stiffness of th e device, whereas the rubber member would apply the vertical load and also create the Poisson’s effect that induces lateral stresses as found under radial tires. 18” 1.40” 12” 1” 0.078” Side View Top View Front View 1.25”

PAGE 54

39 3.4 Preliminary Contact Stress Calculations The preliminary contact stress calculati on was the first confirmation that the loading characteristics of the two devices were different. It was important, however, that the average vertical stress ( z_avg) under the loading strip di d not exceed that of the pressurized hose. This ensured that the main difference in the cont act stress distribution was the presence of lateral stre sses under the loading strip. Initially, carbon paper was pl aced under the APA hose to measure the contact area, but at static mode the imprin t was not clear. Then, at a hardware demonstration at the DOT, a technician measured the actual contact stresses using a pressure-sensitive mat. For the loading strip, a finite element model wa s used to predict the stress distribution at the rubber-specimen interface before it was physically constructed. 3.4.1 APA Pressurized Hose Stresses Tekscan Inc. provided an initial estimate of the vertical stre ss distribution under the pressurized hose when a technician visited th e FDOT for a presentation of the company’s Pressure Measurement System (PMS). Th e Tekscan PMS is an extremely thin (~0.1 mm), flexible tactile force sensor that is cap able of measuring pressures from 0-2 psi (015 kPa) to 0-25 ksi (0-175 Mpa). Figure 3-8 sh ows the wide range of shapes, sizes, and spatial resolutions (sensor spacing) of availa ble sensors. Sensing locations within a matrix can be as small as 0.0009 square inches (0.140 mm2); therefore, a one-squarecentimeter area can contain an array of 170 of these locations [Tekscan, Inc.]. The Virtual System Architecture (VSA) integrat es the sensors into a uniform whole and displays the information on a computer screen.

PAGE 55

40 Figure 3-8. Tekscan pressure measurement system. Tekscan's products function in both static and dynamic measurement environments, allowing the development of load profiles and peak load attainment. During the demonstration, the technician placed the sens or mat under each of the three pressurized hoses in the APA, and recorded the verti cal contact stresses while the machine was operating (dynamic mode). This data-acq uisition method provide d a vertical stress profile under the hose for the en tire run (back and forth) throughout the specimen. Since neither UF nor the FDOT owns a Tekscan measurement system, the access to the data was limited. The recorded vertical stresses revealed th at their distribution is not even along the specimen. Figure 3-9 illustrates the measur ed vertical stress distribution under the pressurized hose (100 psi) with a wheel load of 100 lb. The color gradient indicates the stress intensity ranging from light gray (low pressure, <10 psi) to black (high pressure, >80 psi). The stress distribution in Figure 39 shows two dark-shaded peaks (high stress)

PAGE 56

41 at the edges of the contact ar ea. Initially this anomaly was attributed to the uneven surface of the specimen and, more specificall y, from large aggregate that might be bridging the hose over some small gaps. La ter on, based on the hose contact stresses measured on a steel bed, it appeared that th ese peaks were a hose-structure phenomenon that will be discussed in greater extent in the fo llowing sections. Figure 3-9. Graphical interpre tation of vertical stresses under the pressurized hose. The software approximates the contact area based on the number of cells that report pressure. In the case of the APA, there wa s some residual stress from the lowering arm (the frame that holds the hoses in place) th at was touching the mat. Analysis of the spreadsheet provided by the Tekscan technician yielded to an estimated contact area of 1.54 in2 (993.5 mm2), which gives an average vert ical stress of 64.9 psi (477.8 kPa). 3.4.3 Loading Strip FEM Before fabrication, the concept device was modeled in finite elements to estimate the stress distribution at the loading stripspecimen interface. Fi gure 3-10 shows a side view and a top view of the three-dimensiona l model used for the stress analyses. An estimate of the stress distributi on under the loading strip was im portant to ensure that the

PAGE 57

42 average vertical stress ( z_avg) under the loading stri p would not exceed the z_avg under the pressurized hose. Figure 3-10. Finite element model of the loading strip. The finite element model was constructe d with the MSC/Patran pre-processor software and the elastic analyses run with the ABAQUS engine. The model consisted of 20-node ‘brick’ elements, and the material pr operties used for the steel and rubber parts of the loading strip are the same as in the tire model (Table 3-1). Because of the rectangular shape of the r ubber on the loading strip, it was easy to assume that the width of the contact area w ould be constant at 1.25 inches. Thus, the only requirement to approximate the contact area was to estimate the length of the pressure patch. It was clear that the cont act area under the loadi ng strip would greatly exceed the initial contact area of the APA hos e. The initial contact area under the hose differs from the final because, as the material deforms, the hose ‘sinks’ into the material, increasing the contact area.

PAGE 58

43 Based on the assumption that the contact area would be greater under the loading strip, the FEM model was analyzed at a higher lo ad level (150 lb) to calculate the extent of the pressure patch and the average verti cal stress. The FEM predicted the pressurepatch length to 4.5 inches; thus the resultant area was 5.6 in2 and the average stress approximately 26 psi. 3.5 Measured Contact Stresses in the APA At the beginning of this chapter the hypothe sis was that the pressurized hose of the APA does not capture the essent ial lateral stress distribution found under radial tires. A new concept loading device (loa ding strip) was designed and tested with the help of numerical modeling that would simu late real tire stress distribu tion. In order to verify the above hypothesis, both loading devices – pres surized hose and the lo ading strip – were sent to the Smithers Scientific Services plant in Ravenna, Ohio, to measure the actual contact stresses at the loading de vice-specimen interface. 3.5.1 Measurement System As mentioned in Section 2.3, Smithers Scient ific Services, Inc. developed the Flat Surface Tire Dynamics Machine (FSTDM) to measure contact stresses at the tirepavement interface. The device measures vert ical, transverse, and longitudinal forces and displacements under a moving tire by us ing a series of 16 transducers. Dr. Pottinger fabricated custom end-restrain ts and a loading foot that allowed load control to within 1 lb, to accommodate the pressurized hose and the loading strip on the FSTDM. Figure 3-11 shows a picture of th e FSTDM with the loading foot during calibration. The 500-lb cell was calibrated using a pedal force transducer, as seen in the lower right-hand corner of the picture.

PAGE 59

44 Figure 3-11. Contact stress measur ing apparatus setup and calibration. The loading strip was tested with three different loads – 110-, 130-, and 150-lb – whereas the pressurized hose was tested at two load levels – 1 00-, and 120-lb. The loading foot with the steel wheel remained stationary, while the bed with the loading device (hose/loading strip) m oved in the longitudinal direction. The movement of the bed forced the steel wheel over the loadi ng device and the transducers measured the displacements and stresses at the contact interface. 3.5.2 Hose-Specimen Interface Stresses Results from the APA pressurized hose c ontact stresses verified the initial hypothesis that the contact area under the hose is too narrow to produce any significant lateral stresses. Figure 3-12 s hows the pressurized hose, whic h is attached to the moving bed, and the concave steel wheel loading th e hose directly above the transducers.

PAGE 60

45 Figure 3-12. Close-up picture of the pressurized hose test. In his report, Dr Pottinger stated that th e narrow (8mm) contact area was not wide enough to record any lateral stress on the transd ucers. Figure 3-13 illustrates the vertical stress distribution under the hose. Similar to the Tekscan results (Figure 3-9), the measured vertical stresses show two humps at each side of where the steel wheel loads the hose, caused by the semi-rigi d structure of the hose.

PAGE 61

46 -140.0 -120.0 -100.0 -80.0 -60.0 -40.0 -20.0 0.0 0123456789 Lateral Location, X (in)Vertical Stress, Z (psi) 100 lbs 120 lbs Figure 3-13. Vertical stress dist ribution under the pressurized hose.

PAGE 62

47 3.5.3 Loading Strip-Specimen Interface Stresses Smithers Scientific Services measured the contact stresses under the loading strip for three load levels – 110-, 130-, and 150-lb. Figure 3-14 is a picture of the steel wheel applying pressure on the loading strip whic h is fixed on the moving bed. Dr. Pottinger noticed that in the case of the loading stri p, the solid wheel had to be centered over the loading strip to avoid asymmetric stress dist ribution. Contrary to the solid wheel, the concave wheel acts as a channel that c ontinuously aligns the rubber hose with the traversing movement of the loading arm. Figure 3-14. Close-up picture of the loading strip test. Figure 3-15 illustrates the vertical stress di stribution under the loading strip for the three load levels. Unlike the pressurized hose results, the verti cal stress distribution under the loading strip resembles that of an elastic material with the stress peaking in the middle of the normal distribution. As expected, the magnitude of the vertical stresses is

PAGE 63

48 much lower under the loading strip due to the increase of the contact area. The highest measured vertical stress under the loading strip for the high load (150 lb) was 35 psi, whereas the pressurized hose recorded 130 psi vertical stress for the 100-lb load. Figure 3-16 shows that the transverse st ress distribution under the loading strip accurately captures the Poisson’s effect found under individual tire ribs. The Poisson’s effect states that, unless restrained, most materials expand laterally when loaded vertically. When individual ribs under a tire are loaded they attempt to expand laterally, and the surface of the pavement tries to restra in the expansion thus generating transverse stresses. Similar to the tire ribs but lower in magnitude, the loading strip induces lateral stresses that change sign (direction) at opposite side s of the loading strip. 3.6 Summary The discussion in this chapter focused on the contact-stress distribution between the specimen and the pressurized hose in the APA and the effort to develop an alternative loading mechanism to capture the complex ity of the actual ti re-pavement interface stresses. Actual contact-stress measurements verified that the limited initial contact area under the pressurized hose coul d not induce lateral stresses on the surface of the HMA sample. However, results from the proposed loading device (loading strip) showed that the distribution of the lateral contact st resses closely resembles that found under individual tire ribs.

PAGE 64

49 -40.0 -35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 0.0 0123456789 Lateral Location, X (in)Vertical Stress, Z (psi) 110 lbs 130 lbs 150 lbs Figure 3-15. Vertical stress di stribution under the loading strip.

PAGE 65

50 -15.0 -10.0 -5.0 0.0 5.0 10.0 15.0 00.250.50.7511.25 Transverse Location, Y (in)Transverse Contact Stress, Y (psi) 110 lbs 130 lbs 150 lbs Figure 3-16. Lateral stress dist ribution under the loading strip.

PAGE 66

51 CHAPTER 4 STRESS ANALYSES 4.1 Overview Recent experimental studies revealed that tire contact stresses are distributed in a highly non-uniform manner and differ significan tly for various tire types [De Beer et al. 1997, Marshek et al. 1986, and Myers et al. 1999]. These stresses include not only vertical normal stresses, but also transverse and longitudinal surface shear stresses. One proposed hypothesis on the mechanism behind inst ability rutting is that radial tires, with their complex non-uniform loading, may be inf licting significant stress states in the HMA that are not predicted with traditional uniform vertical loading patterns [Drakos 2000]. Elastic layer and finite element analyses of asphalt pavements for three load cases – radial tire load, bias-ply tire load, and uniformly distributed vertical load – showed that radial-tire loads induce more severe stress states near the surface of the pavement. The measured contact stresses under th e two loading mechanisms – hose and loading strip – also revealed some expected di fferences in the contac t-stress distribution. Lateral stresses under the load ing strip resembled the distri bution found under a single rib on the radial tire tread. Finite element modeling of the APA te st showed that the loading strip, similar to the radial tires, induced so me tension near the surface of the specimen. These stress states tend to induce shear stresse s that shove the material away from the loaded area similar to instability rutting.

PAGE 67

52 4.2 Pavement Stress Analyses To evaluate the effect of the contact stresses found under various tires, the measured tire-pavement interface stresses were applied as a load on a pavement structure and the resulting stress states were analyzed. In the past, finite elements have not been used to model three-dimensional tire loads due to the complexity of modeling a radial-tire load in three dimensions. Instead, the el astic multi-layer analysis program BISAR [De Jong et al. 1973] was used to pred ict the pavement responses. Typical pavement structures consist of a thin asphalt concrete layer over a base course, which rests on the semi-infinite subgr ade. To produce an accurate model of the non-uniform load and provide adequate bounda ry conditions requires a large number of elements, and the associated amount of memo ry is not available on current PCs. The University of Florida recently purchased a Silicon Graphics Interface (SGI) multiprocessor computer with extensive memory a nd faster computing time than the average PC that made the three-dimensional finite elem ent analysis of HMA pavements possible. 4.2.1 Multi-Layer Elastic Stress Analyses The initial approach to model three-di mensional tire contact stress was to approximate the complex loading conditions wi th uniform circular loads. BISAR can apply circular uniform loads with a single ve rtical stress and one stress in the lateral direction of a specific angle. A series of small uniform circul ar loads of varying vertical ( z) and lateral ( x & y) stresses would represent th e non-uniform tire stresses. Figures 4-1 and 4-2 illustrate the load conf iguration for the bias-ply and radial tires respectively. It took 209 load circles to simulate the la teral and non-uniform vertical stress distribution under the bias-ply tire, and 145 load circles to represent the radial tire. The magnitude, orientation, and location of the vertical and lateral stresses of the

PAGE 68

53 individual circular loads used to represent the bias-ply and ra dial tire contact stresses are listed elsewhere [Drakos 2000]. Figure 4-1. Load configuration used in BISA R to represent measured stresses under biasply truck tire [Drakos 2000]. For the bias-ply tire, Figure 4-1 illustrate s the existence of a significant transverse stress component near the edge of the load, which tends to pull the pavement in towards the center of the tire. This developm ent can be explained by the overwhelming pneumatic effect that is induced by bias-ply tires as explained in section 2.3. On the other hand, Figure 4-2 shows that the radial tire creates a ‘pushing outward’ effect under each individual rib. This trend is attributed to the dominance of the Poisson’s effect in radial tires.

PAGE 69

54 Figure 4-2. Load configuration used in BISAR to represent measured stresses under radial truck tire [Drakos 2000]. 4.2.2 BISAR Results Figure 4-3 illustrates the maximum shear st ress distribution for the modeled radial and bias-ply truck tires. The responses were predicted along the surface of the pavement at 0.2-inch increments. High shear stress va lues were calculated under the left-most rib of the modeled radial load, whereas much lo wer values developed under the bias-ply tire load. The high shear stresses pred icted under the modeled radial tire hinted that shear planes might be developing under the load, sim ilar to a bearing-capacity type of failure. To investigate this hypothesis the direction of the predicte d maximum shear stress, at each output point under the tire loa d, was plotted as a vector.

PAGE 70

55 0 50 100 150 200 250 300 350 400 450 500 012345678910111213 Lateral Location, X (in)Maximum Shear Stress, max (psi) Bias-Ply Tire Load Radial Tire Load Figure 4-3. Maximum sh ear stress distribution. Angle was defined as the smallest angle formed between the maximum shear stress plane and the horizontal. Figure 4-4 demonstrates the sign convention used by the analysis program and the calculation of angle which was used to plot the direction of the shear stress plane. Figure 4-5 shows the two equal and opposite maximum shear stresses that act on a particular element. In this case, the sm allest angle is formed between the negative maximum shear stress and the horizontal plan e. Thus, the plotted directional arrow (vector) would represent the direction of the negative maximum shear stress.

PAGE 71

56 Figure 4-4. BISAR sign convention and maximum shear stress angle Figure 4-5. Schematic of the maximum shear stress direction representation.

PAGE 72

57 Figure 4-6 shows the magnitude and di rection of the maximum shear stress distribution along a vertical section for the mode led radial tire load. The arrows indicate the direction of the maximum shear stress clos er to the horizontal, and the contour plot (shaded area) in the backgr ound specifies the magnitude of the shear stress. The direction of the shear stresses under the right-most rib of th e radial tire indicates the formation of shear planes that tend to ‘shove’ the material away from the tire. At the same location, the contour plot of the predicted maximum shear stress magnitude indicates that shear stresses are at their highest value. Figure 4-7 illustrates the magnitude and direction of the maximum shear stress distribution under the modeled bias-ply tire lo ad. Unlike the radial tire load responses, the maximum shear stress direction for the bi as-ply load appears to be ‘pulling the pavement inwards.’ The orientation of the di rectional arrows under th e bias-ply tire load is pointed inwards, towards the tire. When compared with the radial tire load results, the magnitude of the predicted maximum shear stress for the bias-ply tire load was significantly lower. In an effort to isolate the effect of tireinduced lateral stresses, the horizontal stress component in the BISAR input files was set to zero so that only vertical stress was applied. The results of the an alyses performed without the lateral tire contact stresses revealed that, for the radial tire load, th e direction of the maximum shear stress was reversed. In contrast to that results for the bias-ply tire load did not give any indication of stress reversal.

PAGE 73

58 Figure 4-6. Magnitude and dire ction of maximum shear stre sses under radial tire load.

PAGE 74

59 Figure 4-7. Magnitude and dire ction of maximum shear stresse s under bias-ply tire load.

PAGE 75

60 Analyses performed with the elastic layer analysis program BISAR provided information on the pavement’s response unde r modeled tires from measured contact stresses. Even though the magnitude of the es timated stresses is sometimes exaggerated due to a discontinuity problem at the edge of the circular load [Jacobs 1995], the overall effect of the non-uniform loading showed th at the near-surface stress distribution was highly dependent on tire structure. Lateral st resses induced by radial tires seem to cause a reduction in confinement near the paveme nt’s surface, which reduces the mixture’s resistance to shear stress. The combination of high magnitude and outward direction of the maximum shear stresses is believed to creat e the critical stress conditions contributing to the permanent deformation of the pavement. 4.2.3 Finite Element Stress Analyses The three-dimensional finite element mode l was constructed and analyzed using the commercial finite element code ADINA [B athe 2001]. The finite element model consisted of 30,204 nine-node elements of varying dimensions. Elements under the loaded area were refined to 0.3 by 0.4 inches in the horizontal plane and 0.2 inches in the vertical plane. To overcome the limitations a ssociated with building a traditional mesh, contact surfaces were introduced to ‘bond’ fi ne-graded mesh onto a coarse-graded mesh. This allowed for the introduction of coarse me shes at distances farther away from the loaded area where the change in stress was mo re gradual. The use of contact surfaces was further justified based on the primary area of interest, the near-surface area under the loaded tire, thus negating any possible negati ve numerical effects of ‘far-away’ contact surfaces. Figures 4-8 and 4-9 show the finite elemen t model with the dimensions used in the analysis. The mesh extended 60 inches in ea ch horizontal direction and 72 inches in the

PAGE 76

61 vertical direction. Although these dimensions may seem inadequate for a finite element model, an initial assessment showed that stresse s near the tire footprint (area of interest) were not affected by the ex tent of the boundaries. Figure 4-8. Three-dimensiona l finite element mesh used in the pavement response analysis. The boundary conditions for the four sides (faces) of the FEM were fixed in the horizontal (X and Y) direction and free in th e vertical (Z) direc tion, whereas the bottom of the FEM was fixed in all directions. The model consisted of 260,455 nodes, giving it over one million degrees of freedom. The memory required for analysis exceeded 1700 megabytes and took over four hours to complete on a single processor.

PAGE 77

62 Figure 4-9. Plan view of th e contact area of the three-dimensional mesh used in the pavement response analysis. The structure used in the analysis was a typical three-laye r pavement – asphalt concrete, base, and subgrade – with thickne sses 8, 12, and 52 inches respectively. Each layer was assumed to be isotropic, homoge nous, and linear elastic. Table 4-1 shows a summary of the material properties used in the structure. The low asphalt concrete modulus corresponds to a warm summer day fo r a new pavement – the most critical time for the onset of instability rutting – while base and subgr ade modulus values represent typical materials used in Florida. The Poi sson’s ratio was selected to ensure minimal volumetric changes, as would be expe cted from a single moving tire load.

PAGE 78

63 Table 4-1. Material proper ties and layer thicknesses of FEM pavement structure. Layer Modulus (psi) Poisson’s ratio Thickness (inches) Asphalt concrete 100,000 0.45 8 Base 40,000 0.45 12 Foundation 15,000 0.45 52 4.2.3-1 Loading the FEM Dr. Pottinger measured the tire-pavement contact stresses in a fine grid – 0.1 by 0.15 inches – making it almost im possible to load the mesh di rectly with the recorded stresses. An approximation method was used to convert and redistribute the measured contact stresses to nodal forces. The appropr iate force for each element was determined by converting each uniform stress to an equivalent concentrated force. The forces were then converted to nodal forces with the he lp of shape functions and applied to the respective node [Cook 1995]. Figure 4-10 illustrates the shape-function procedure to redistribute an element force to nodal forces. Parameters and define the position of the element load relative to the element’s center, and their value ranges from zero to one. The shape functions are then calculated for each node based on the and values. Finally, the nodal forces (F1 to F8) are the product of the element force (P) with the respective nodal shape function (N1 to N8). Figure 4-11 is a cross section view of the surface elements with the resulting nodal forces for the modeled radial -tire load. The nodal loads ar e non-uniform and vary in magnitude and direction to simulate the contact stresses found under a radial tire.

PAGE 79

64 5 12 6 8 4 7 3 = Element Node = Element Force) ( ) (_ load center ele load center eleY Y X X ) 1 )( 1 ( 2 1 8 ) 1 )( 1 ( 2 1 7 ) 1 )( 1 ( 2 1 6 ) 1 )( 1 ( 2 1 5 ) 8 ( 2 1 ) 7 ( 2 1 ) 1 )( 1 ( 4 1 4 ) 7 ( 2 1 ) 6 ( 2 1 ) 1 )( 1 ( 4 1 3 ) 6 ( 2 1 ) 5 ( 2 1 ) 1 )( 1 ( 4 1 2 ) 8 ( 2 1 ) 5 ( 2 1 ) 1 )( 1 ( 4 1 12 2 2 2 N N N N N N N NShape Functions P P 5 12 6 8 4 7 3 = Element Node = Element Force) ( ) (_ load center ele load center eleY Y X X ) 1 )( 1 ( 2 1 8 ) 1 )( 1 ( 2 1 7 ) 1 )( 1 ( 2 1 6 ) 1 )( 1 ( 2 1 5 ) 8 ( 2 1 ) 7 ( 2 1 ) 1 )( 1 ( 4 1 4 ) 7 ( 2 1 ) 6 ( 2 1 ) 1 )( 1 ( 4 1 3 ) 6 ( 2 1 ) 5 ( 2 1 ) 1 )( 1 ( 4 1 2 ) 8 ( 2 1 ) 5 ( 2 1 ) 1 )( 1 ( 4 1 12 2 2 2 N N N N N N N NShape Functions P P Figure 4-10. Definition of the shape functions. 0 9.3 inches 0 9.3 inches Figure 4-11. Cross-section view of surface elem ents with nodal forces for the radial-tire load.

PAGE 80

65 4.2.3-2 FEM Results Figure 4-12 shows the magnitude a nd direction of the maximum shear stress distribution along a vertical se ction under the modeled radial -tire load. The distribution of shear stresses is similar to those predic ted using BISAR, except the magnitudes of the maximum shear stresses are lower, ranging fr om 50 to 60 psi compared to values in excess of 100 psi for the BISAR predictions The difference in magnitude can be attributed to the approxima tion method used to convert th e measured point stresses to circular loads, and to the BISAR overestim ation problems mentioned above. However, the key finding is that the overall trend – the formation of the shear planes – remains the same between the two different modeling techniques. Figure 4-12. Maximum shear stress magnit ude (psi) and direction under the modeled radial-tire load. The plotted vectors indicate th e formation of shear planes under the loaded area. Also, it can be seen from the direction of the shear stresses under the fi rst rib of the radial tire that shear planes formed tend to shove the material away from the tire, and the contour plot of the predicted maximum shear stress magnitude indicates that the shear stresses form planes that match the directional arrows.

PAGE 81

66 4.3 APA Stress Analyses The measured contact stresses betwee n the HMA specimen and the two APA loading mechanisms – pressurized hose and th e loading strip – were modeled with finite elements to evaluate the e ffects of the different load ing conditions. The primary objective was to examine whether the loading strip could induce similar stress states in the modeled HMA specimen, as the radial ti re induced in the modeled pavement. The three-dimensional finite element model was constructed using the MSC/PATRAN pre-processor software to build the model geometry and to define the mesh and the HKS/ABAQUS software for the actual analysis [Hibbitt, Karlsson and Sorensen, Inc. 1997]. To build the mesh around the curved surfaces we used an automatic mesh-generating option in PATRAN ca lled paver. The paver is best suited for trimmed surfaces, such as surfaces with holes or cutouts. In this case, the paver meshing algorithm generated quadrilateral elements perpendicular to the curved surfaces and transitional elements to connect to the free e dges. At first, the mesh was generated in two-dimensional space and then the ‘sweep’ action was used to extrude the elements into three-dimensional ‘brick’ elements. Figure 4-13 shows the initial two-dimensional model for the APA mold and the HMA specimen. For practical purposes, only one of the two cylindrical-sample slots was used in the model. Furthermore, the mode l was separated into two main parts – the plastic mold (E = 400000 psi, = 0.4) and the asphalt concrete sample (E = 100000 psi, = 0.4). Figure 4-14 shows the three-dimens ional geometry and mesh definition of the two solids.

PAGE 82

67 Figure 4-13. Top view of the finite elem ent model for the APA mold and specimen. Figure 4-14. Three-dimensional finite elem ent model for the APA mold and specimen.

PAGE 83

68 As mentioned in Section 3.5, Dr. Pottinge r measured the contact stresses under the two loading devices – the orig inal pressurized hose and the new loading strip. The objective in this section was to load the model with the measured contact stresses and analyze the predicted stress states in the spec imen. Once again, the measured stresses were converted to nodal forces with the he lp of shape functions and applied to the respective node (Section 4.2.3-1). Figures 4-15 and 4-16 show the predicted magnitude and direction of the maximum shear stress ( max) distribution, along a vertical se ction, for the loading strip and pressurized hose loading conditions respectively. The range of the max magnitude under the loading strip (3-13 psi) is much lower th an that predicted unde r the pressurized hose (10-80 psi). This magnitude difference can be attributed to the hi gher vertical stresses measured under the pressurized hose (Section 3.5.2). The important finding of this an alysis was the pattern of the max distribution throughout the modeled specimen. Unlike the di stribution under the pressurized hose, the modeled loading strip showed that the max magnitude peaks near the surface of the specimen, under the loaded area. Furtherm ore, the magnitude contour plots for the loading strip condition indicate the existence of shear planes under the load, whereas the same is not true for the modeled pressu rized hose load. The combination of the max magnitude distribution with the max direction under the load ing strip resembles the pattern found under individual ribs for the m odeled radial-tire load (Figure 4-12).

PAGE 84

69 Figure 4-15. Maximum shear stress magnit ude (psi) and direction under the modeled loading strip load. Figure 4-16. Maximum shear stress magnit ude (psi) and direction under the modeled pressurized hose load.

PAGE 85

70 4.4 Summary Analyses performed with th e elastic layer analysis pr ogram BISAR and the finite element program ADINA provided information on the pavement’s response under modeled tires from measured contact stresses. The analyses provided evidence that the radial truck tires induce higher shear stresses near the surface of the pavement than the bias-ply tire. The direction of the maximu m shear stresses under the modeled radial tire load appeared to shove the material away fr om the load, something that was not observed under the modeled bias-ply tire. The measured contact stresses under the two APA loading devices were used to load the APA finite element model. The anal ysis showed that the pressurized hose load produced higher magnitude stresses. However, the loading strip replicated (to some extent) the critical stre ss states identified under th e radial-tire load.

PAGE 86

71 CHAPTER 5 MATERIALS AND TESTING METHODS 5.1 Overview Two of the three mixtures selected for th is study have been placed in Florida in 1998, (see Table 5-1), and the FDOT has been monitoring their field performance ever since. The Job Mix Formula (JMF) of the original FDOT mixtur es had a Reclaimed Asphalt Pavement (RAP) component of 15%-20% that formed part of the aggregate constituent. However, the RAP material wa s no longer available at the time of this research so the percentages of the other aggr egates were adjusted to maintain the same gradation for each mix. Table 5-1. Field locatio n of selected mixtures. FDOT Project No. UF Project No. Placement Date Route County 2134391 1 Jan-98 I-10 Madison 2325941 7 Sep-98 Turnpike Palm Beach The mixtures were designed and produced according to the Superpave Volumetric Mix Design procedure and the samples compacted with the Pine Gyratory Compactor to 150-mm diameter by 115-mm height gyratory specimens. Finally, the samples were tested with the original a nd modified Asphalt Pavement An alyzer test procedures to evaluate their rutting performance.

PAGE 87

72 5.2 Materials Table 5-2 lists the original aggregate types and producers used in the FDOT monitoring projects. Project 1 is a 9. 5-mm nominal maximum-size coarse-graded mixture, Project 7 is a 12.5-mm nominal maximum-size fine-graded mixture, and the HVS mixture is a 12.5-mm nomi nal maximum-size coarse-graded mixture. The last column on the right shows the JMF-blend percen tage for each aggregate type. As stated earlier, the Superpave project mixtures – Project 1 & Project 7 – included RAP (milled material) which was not available at the time of this study. Table 5-2. Aggregate types and sour ces for the selected FDOT mixtures. Milled material---20 # 89 Stone51GA 185Martin Marrietta45 W-10 Screenings20GA 185Martin Marrietta25 M-10 Screenings21GA 185Martin Marrietta10 Milled material---20 S1A Stone4187-339White Rock Quarries20 S1B Stone5187-339White Rock Quarries10 Asphalt Screenings2087-339White Rock Quarries50 S1A Stone4187-089Rinker13.0 S1B Stone5187-089Rinker55.0 Asphalt Screenings2029-361Anderson32.0 HVS Pit No.Producer JMF % 197051A Project No. Mix No.Material FDOT Code 7980139A Table 5-3 shows the new aggregate blends that were obtained from the same sources and adjusted without the RAP to re produce the original JMF. After careful selection, the gradations of the new blends resembled the original JMF for each project – Project 1 & Project 7. Figures 5-1 to 5-3 illustrate the grad ation of the laboratory blend in comparison with the original (fie ld) JMF on a 0.45-power chart.

PAGE 88

73 Table 5-3. Aggregate sour ces and modified blends fo r the laboratory mixtures. # 89 Stone51GA 185Martin Marrietta50.0 W-10 Screenings20GA 185Martin Marrietta18.5 M-10 Screenings21GA 185Martin Marrietta31.5 S1A Stone4187-339White Rock Quarries24.5 S1B Stone5187-339White Rock Quarries12.5 Asphalt Screenings2087-339White Rock Quarries63.0 S1A Stone4187-089Rinker13.0 S1B Stone5187-089Rinker55.0 Asphalt Screenings2029-361Anderson32.0 HVS Pit No.Producer 197051A 7980139A JMF % Project No. Mix No.Material FDOT Code The asphalt binder used for the study is an AC-30 binder with a PG-67-22 grading. It is distributed by Coastal Petroleum Comp any in Jacksonville (subsidiary of Coastal Caribbean Oils & Minerals, Ltd) a nd is a common binder in Florida.

PAGE 89

74 0 10 20 30 40 50 60 70 80 90 100 Sieve SizePercent Passing #200 #100 #50 #30 #16 #8 #4 " "#100 Original JMF Laboratory Blend Figure 5-1. Gradation chart for JMF and laboratory blend for Projec t 1 (9.5mm maximum nominal size).

PAGE 90

75 0 10 20 30 40 50 60 70 80 90 100 Sieve SizePercent Passing #200 #100 #50 #30 #16 #8 #4 " "#100 Original JMF Laboratory Blend Figure 5-2. Gradation chart for JMF and laboratory blend for Project 7 ( 12.5mm maximum nominal size).

PAGE 91

76 0 10 20 30 40 50 60 70 80 90 100 Sieve SizePercent Passing Original JMF#200 #100 #50 #30 #16 #8 #4 " "#100 Figure 5-3. Gradation chart for laborato ry blend for the HVS coarse-graded mixt ure (12.5mm maximum nominal size).

PAGE 92

77 5.3 Mixture Preparation All test specimens for the evaluation te sts were prepared with the Superpave Volumetric Mix Design proce dure. The design procedure uses volumetric properties – air voids (AV), voids in mineral aggregate (VMA), and voids fille d with asphalt (VFA) – as the primary criteria to select the optim um asphalt content (% AC) for the specified aggregate blends. The Superpave criteria vary with the specified nominal maximum aggregate size for the JMF and the expected traffic level. The two Superpave monitoring project mixtur es used in this study – Projects 1 & 7 – have been used in various research pr ojects at UF [Asiamah 2002, Darku 2003] and there is ample information about their volumetri c properties. For this reason, the values for Rice Specific Gravity, optimum asphalt cont ent, and Ndes were not recalculated. Further information about the volumetric prope rties of all three mixtures is included in Appendix A. Outlined below is the procedure followed to produce the 75-mm-thick cylindrical specimens used in the APA. 5.3.1 Aggregate Preparation and Batching The virgin material was dried in the ove n (230 300F) for at least 12 hours and then allowed to cool down to room temperature. The material was sieved and separated into its individual particle sizes – ”, ”, ”, #4, #8, #16, #30, #50, #100, #200, & pan. The aggregates were batched in 4500g samples (Pine Gyratory Compactor) in accordance with the JMF. Tables showi ng batch weights for the aggregates are given in Appendix A. 5.3.2 Mixing The aggregates, asphalt binder and the mixing equipment – mixing bucket and spatulas – were placed in the oven (300F) for about two hours. Aggregate blend and asphalt binder were mixed in the bucket for about 5 minutes or until the aggregates were th oroughly coated with the asphalt.

PAGE 93

78 5.3.4 Short-Term Oven Aging (STOA) and Compaction Before compaction, the samples were spr ead in a pan and placed in the oven (275F) for STOA. While in the oven, the mix was stirred after one hour to achieve uniform aging. After the STOA, the 4500g samples were compacted in the Pine Gyratory Compactor (Figure 5-4) to Ndes. Rice sp ecific gravity information is given in Appendix A. Compacted specimens were allowed to c ool for a minimum of 24 hours at room temperature and then were sa wed down to 75mm thickness. The Bulk Specific Gravity (Gmb) was then determined in accordance with ASTM D1189 and D2726 for each specimen. The percent air voids for each specimen was computed from the Gmm and the Gmb. Figure 5-4. Pine Gyratory Compactor.

PAGE 94

79 5.4 Asphalt Pavement Analyzer Procedure As we have seen, the APA tests the rutti ng susceptibility or rutting resistance of HMA. The original configuration of the m achine creates the desirable contact pressure with a concave aluminum wheel attached to a reciprocating arm moved along a pressurized hose, whereas a steel wheel lo ads the loading strip for the modified configuration. The device is able to test either a 75mm x 125mm x 300mm beam specimen, or a 150mm diameter by 75mm thic k cylindrical specimen. For this study, only cylindrical specimens were evaluated. The APA test procedure was slightly m odified to incorporate a new way of recording and analyzing the test results. Inst ead of using the roller dial gauge to measure a single (the lowest) point on the specimen, the new method uses a contour gauge that captures the entire surface prof ile of the sample. In case the material fails, the new method (contour gauge) enables the user to monitor the rate and the mode (consolidation/instability) at whic h the material is failing. Figure 5-5 shows the original measuring pl ate with the 3-inch slits where the dial gauge is dragged to locate the lowest (highest dial reading) spot on the specimen. This method is limited to a single measurement that denotes the maximum amount of permanent deformation on the specimen and there is no way to distinguish if that deformation is due to cons olidation or plastic flow. Figure 5-6 illustrates the aluminum plat e used with the new method of measuring the specimen deformation. The slits on th e new plate are 5 inch es wide at the two extreme locations (marked with an E on Figure 5-6) and 5 inches at the middle location (marked with an M on Figure 5-6). The cont our gauge adjusts to the surface profile of the specimen, which is (the surface profile) recorded and di gitized for further analysis.

PAGE 95

80 Figure 5-5. Original APA measuring plate. Figure 5-6. New measuring plate with elongated slits. E EM E E M

PAGE 96

81 5.4.1 Surface Profile Measurement As mentioned above, the new measur ement system (contour gauge) was implemented to record the entire surface prof ile of the specimen. Figure 5-7 shows the contour gauge recording the surface profile at the middle location of the sample. The rods are pushed downwards until they come in contact with the specimen, forcing the contour gauge to assume the shap e of the specimen’s surface. Figure 5-7. Contour gauge measuring the surface profile of the specimen.

PAGE 97

82 Figure 5-8 shows the contour gauge on th e custom fabricated holder where the shape of the recorded surface pr ofile, from each location on the measuring plate, is traced on a card. The card slides beneath the c ontour-gauge rods, whereas two PVC strips restrain the card from any lateral movement. Figure 5-8. Recording the deform ed shape of the contour gauge. 5.4.2 APA Hose Testing Procedure The steps for the APA testing procedure using the pressurized hose are outlined below: Preheat the specimen in the APA test cham ber to 64C (147F) for a minimum of 6 hours but not more than 24 hours before the test. Set the hose pressure gauge reading to 1005psi. Calibrate each wheel with the load cell to read a load of 1005lb. Secure the preheated, molded specimen in the APA, close the chamber doors and allow 10 minutes for the temperature to stabilize. Apply 25 load cycles and then take initial (datum) measurements.

PAGE 98

83 Place the specimen back in the APA, close the chamber doors and allow 10 minutes for the temperature to stabilize. Restart the APA and continue rut testing. Repeat measurement procedure at 1000, 2000, 4000, and 8000 cycles. 5.4.3 APA Loading Strip Testing Procedure Outlined below are the steps for the new APA testing procedure: Preheat the specimen in the APA test cham ber to 64C (147F) for a minimum of 6 hours but not more than 24 hours before the test. Calibrate the steel wheel with the load cell to re ad a load of 1505lb. Secure the preheated, molded specimen in the APA, close the chamber doors and allow 10 minutes for the temperature to stabilize. Apply 25 load cycles and then take initial (datum) measurements. Place the specimen back in the APA, close the chamber doors and allow 10 minutes for the temperature to stabilize. Restart the APA and continue rut testing. Repeat measurement procedure at 1000, 2000, 4000, and 8000 cycles. 5.5 Summary The mixtures used in this project contai ned material that was no longer available (milled material), so the aggregate blend wa s reconfigured to match the original JMF gradation. Overall, the la boratory engineered mixtures had similar gradations and volumetric properties with the original HMA. The new testing method for the APA alters the data-recording pr ocedure to gather more information about the sample’s performa nce. Whereas the current FDOT procedure requires four rut-depth measurements per specimen, the new method records approximately 1300 response points per specimen for the analysis. The next chapter

PAGE 99

84 describes the data analysis methods as well as the valuable inform ation from the extra measurements.

PAGE 100

85 CHAPTER 6 DATA ANALYSIS METHODOLOGY 6.1 Overview The data-recording method in the APA test procedure was modified to incorporate a more detailed way of analyzing the test re sults. Instead of the dial gauge, used to measure a single (the lowest) point on the specimen, the new method uses a contour gauge to capture the entire surface profile of the sample. The contour gauge’s rods are pushed downwards until they come in contac t with the specimen, forcing the contour gauge to assume the shape of the specimen’s surface. For each of the three locations on the measuring plate, the specimen’s surface deformation profile is traced on a card and scanned as a bitmap image for further analysis. This chapter will focus on the methods used to analyze the data recorded with the contour gauge. The new data-recording me thod (contour gauge) enables the user to calculate more than just th e highest deformation point. With the new method it is possible to monitor the rate and the mode (cons olidation/instability) at which the material is failing. 6.2 Digitizing the Measured Profile Each specimen requires 15 cards – five m easurements at three locations – to capture its rutting profile throughout the 8000cycle test. To proceed with the data analysis, the cards are scanned as bitmap imag es (.bmp) and digitized with the help of a program called Grafula3 [Wishnevsky 2003]. Graf ula3 allows the user to digitize points

PAGE 101

86 from an image of a graph, scanned from originals such as articles, monographs, and diagrams. Figure 6-1 shows a screen shot of the basi c window of the program. Grafula3 reads the bitmap image and, with the help of user-s et coordinate axes, it automatically digitizes the selected part of the graph. The data table records the loca tion of each point in Cartesian coordinates that can be transferred into a separate (Excel) spreadsheet for further analysis. Figure 6-1. Grafula3 screen shot.

PAGE 102

87 6.3 Rut Depth Calculations The digitized data from the measured surf aced profiles is imported in a spreadsheet and organized per number of remaining cycles. Table 6-1 shows part of a typical table that describes the deformati on characteristics of one speci men (at one location) in the course of the test. The cycles are meas ured in reverse order, so the 8000-cycle measurement is considered to be the initial reference profile (datum). Table 6-1. Example of a digitized defo rmation profile from one location. XYXYXYXYXY 0.5360.8160.5230.8360.5430.8300.4960.8030.5230.809 0.6090.8160.5890.8230.6030.8500.5690.8160.5560.823 0.6760.8430.6560.8230.6820.8630.6360.8430.6090.842 0.7220.8630.7090.8430.7620.8430.7090.8360.6690.856 0.7750.8630.7950.8430.8220.8300.7890.8160.7490.842 0.8550.8430.8350.8360.8950.8300.8490.8100.8020.823 0.9080.8300.9020.8300.9550.8160.9280.8030.8550.816 0.9950.8230.9680.8231.0280.8230.9820.7970.9020.809 1.0680.8161.0480.8161.1010.8231.0550.7970.9820.803 1.1410.8231.1080.8231.1880.8101.1210.7971.0410.796 1.2080.8161.1610.8231.2610.8101.1940.7771.1210.803 1.2880.8101.2340.8101.3270.8231.2680.7901.1880.776 1.3610.8231.3070.8031.4010.8231.3340.7831.2680.783 1.4340.8231.3810.8231.4740.8101.4070.7771.3410.783 1.4870.8231.4600.8161.5600.8101.4800.7701.4210.783 1.5740.8101.5540.8231.6330.8101.5540.7771.4870.769 REMAINING CYCLES 0 4000 800070006000 Figure 6-2 illustrates the deformation profile of a specimen tested with the original pressurized hose. An interesting point on th is graph is the way the deformation profile changes with the progress of the test. Appare ntly the material does not only consolidate, but it also heaves to the sides of the loaded area. This is something that the traditional way of measuring the rut-dept h results could not show.

PAGE 103

88 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 0123456 Lateral Location, X (in)Depth, Z (in) Datum 1000 Cycles 2000 Cycles 4000 Cycles 8000 Cycles Figure 6-2. Deformation profile for a sp ecimen tested with the pressurized hose. The traditional way of calculating the rut depth for an APA specimen is to take two measurements – the lowest point at the beginning and the lowe st point at the end of the test – and report the differe nce after 8000 cycles. For th e purpose of this study, the traditional way of measuring rut depth will be identified as the Absolute Rut Depth (ARD). The Differential Rut Depth (DRD) is defined as the difference of the lowest point at the beginning of the te st and the highest point record ed at the end of the test. Figure 6-2 shows a graphical in terpretation of ARD and DRD. ARD DRD

PAGE 104

89 6.4 Area Calculation In most cases, permanent deformation of asphalt mixtures in the field is a combination of two mechanisms – reduction of air voids and shear deformation. Figure 6-2 proves that the same combination of fa ilure mechanisms applies for HMA specimens tested in the APA. The objective here is to determine which of the two modes of deformation contributes the most in failing the material. Figure 6-3 illustrates the theory behind the area change calculation. Assume the schematic in Figure 6-3 represents a pavement section that is experiencing excessive rutting. If there was a way to calculate the two shaded areas – A1 and A2 – it would be possible to determine if the permanent deform ation was primarily due to plastic flow or primarily due to consolidation. In the case that the material fails due to shear deformation, the magnitude of A1 and A2 would be equal because the material is shoved to the side. With the same logic, if the ma terial fails primarily due to consolidation the magnitude of A1 would be lower than A2. Figure 6-3. Area change interpretation.

PAGE 105

90 To analyze the APA test results, the data was transferred to MathCAD and the LOESS function [Cleveland 1979] was used to fit a polynomial to the original surfaceprofile data. The LOESS f unction is best described as a locally-wei ghted polynomial regression function. The polynomial is fit us ing weighted least squares, giving more weight to points near the point whose res ponse is being estimated and less weight to points farther away. It is a computationa l intensive method; however, this is not a problem in our current computing environmen t unless the data sets are very large. Once the LOESS function is fitted for the set of surface profile data, the polynomial is integrated over a certain interval to calcula te the area under the curve. Figures 6-4 and 6-5 show a graph of the initial and final prof ile with the smoothed (fitted) polynomial and the calculated areas. Based on the discussion above, the failure mode is primarily consolidation if the initial area (Ai) is less than the final area (Af). If the Ai is greater or equal to Af then permanent deformation is prim arily due to plastic flow. A simple way to determine the effect of the area change was to calculate the percent area change ( A). A positive A means that the mixture is experiencing instability rutting, whereas a negative A indicates that the mixture is deforming primarily due to consolidation. Therefore: 100 A A A Change Area %i f i (6-1) ion Consolidat Primarily 0 Change Area % If y Instabilit Primarily 0 Change Area % If

PAGE 106

91 45 5 55 0034 4 ) ( dx x fit AREAi Figure 6-4. Initial surface profile and area calculation. 45 5 55 0062 4 ) ( dx x fit AREAf Figure 6-5. Final surface pr ofile and area calculation.

PAGE 107

92 6.5 Summary The new method of recording data for the APA test provides a plethora of information that can be further analyzed to determine the leading mode of failure. The two new parameters – differential rut depth a nd percent area change that are calculated from the recorded surface profiles provide in sight about the mixture’ s ability to resist instability rutting. The next chapter discu sses the APA tests result s and the effects of each factor – loading mechanism, comp action level, and temperature.

PAGE 108

93 CHAPTER 7 APA TEST RESULTS 7.1 Overview Three mixtures – Project 1, Project 7, a nd the HVS mixture – were tested with the modified and original APA loading de vices. The new method for measuring deformations, recording the entire surface profile was used in both tests. Tests run with the loading strip were performed at two te mperatures – 64C and 70C – whereas tests with the pressurized hose were run at 64C. Also, the mixtures were prepared and tested at two air void content levels – 4%AV and 7%AV – to evaluate the effects of compaction on the test’s ability to predict permanent deformation. Field measurements were available fo r the two Superpave monitoring project mixtures – Project 1 and Project 7. For th is reason, the APA results from the various testing procedures were compared to evaluate each test method’s reliability in predicting performance. Appendix B contains the resu lts of the one-tailed t-tests performed to examine the statistical significance of the differe nces. Initially, Proj ect 5 was selected as the third mixture to be included in the te st sequence, but the material was no longer available so it was substituted with the HVS coarse-graded mixture. 7.2 Field Results The field rut depths were meas ured using a transverse profil er at thirty locations of each project. Table 7-1 shows the average accu mulated rut depths tw o years after the end of construction and opening of the pavements to traffic. The measured rut depths show that Project 1 experienced hi gher rutting while Project 7 s hows relatively lower rutting

PAGE 109

94 within the same period. Figure 7-1 compares the field rut depth per ESAL between the two Superpave projects. Looking at the normalized rut depth, it is furt her established that Project 7 performed better in the field, in te rms of rutting resistan ce, than Project 1. Table 7-1. Field rutting data. 15.11.48 72.52.99 Project No. Avg. Field Rut Depth After Year 2 (mm) Estimated ESAL at Year 2 (million) 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 Project 1Project 7 MixtureRut Depth/ESAL*106, (mm) Figure 7-1. Measured field rut depth per million ESAL.

PAGE 110

95 7.3 Absolute Rut Depth Absolute rut depth is the measured differen ce between the lowest point of the initial surface profile and the lowest point of the final surface profile. This is the traditional way of measuring the specimen’ s performance in the APA. Various agencies suggest that the criterion for good fi eld-rutting performance is to keep ARD less than 8 mm [Kandhal and Cooley, Jr. 2002]. Figure 7-2 compares the absolute rut dept h results between Proj ect 1 and Project 7 for the three test methods – new APA 64C, new APA 70C, and the original APA – at four percent air void content (4%AV). Results from the new and original APA tests at 64C did not show any significant difference in performance between the two mixtures. However, the new APA tests at 70C showed that the ARD for Project 1 increased whereas the ARD for Project 7 remained at the same level. Figure 7-3 shows ARD results for the same two projects tested at 7% AV. In the case of 7%AV, results from all three tests showed significant difference betw een the performances for the two mixtures. 7.4 Differential Rut Depth The differential rut depth is defined as th e difference of the lowest point at the beginning of the test and the highest point record ed at the end of the test. The function of this parameter is to incorporate the instabil ity characteristics of the material into the rutting prediction. Unlike the ARD, the DRD includes the dilated portion of the deformed material into the measurement. Figure 7-4 compares the differential rut de pth results between Pr oject 1 and Project 7 for the three test methods – new APA 64C, new APA 70C, and the original APA – at 4% air void content (4%AV). Similar to th e ARD results, the new and original APA tests at 64C did not show any significant difference in pe rformance between the two

PAGE 111

96 mixtures. Once again, the new APA tests at 70C showed that the DRD for Project 1 increased whereas the DRD for Project 7 remained at the same level. Since the binder is the same for both projects, the difference in DRD suggests that the new APA might be able to account for the effect of aggregate st ructure in the mixture’s ability to resist rutting, something other studies [Kandhal and Cooley, Jr. 2002, Romero and Stuart 1998] showed that the APA was not able to do. Figure 7-5 shows DRD results for the same two projects tested at 7%AV. Once again, the 7%AV results from all three tests showed significant difference between the performances for the two mixtures. 0 1 2 3 4 5 6 7 8 9 10Absolute Rut Depth, ARD (mm) New APA 64CNew APA 70COriginal APA 64C Project 7 Project 1 Figure 7-2. Absolute rut depth measurements for Projects 1 and 7 with the two loading devices at 4% AV.

PAGE 112

97 0 1 2 3 4 5 6 7 8 9 10Absolute Rut Depth, ARD (mm) New APA 64CNew APA 70COriginal APA 64C Project 7 Project 1 Figure 7-3. Absolute rut depth measurements for Projects 1 and 7 with the two loading devices at 7% AV.

PAGE 113

98 0 1 2 3 4 5 6 7 8 9 10Differential Rut Depth, DRD (mm) New APA 64CNew APA 70COriginal APA 64C Project 7 Project 1 Figure 7-4. Differential rut depth measurements for Projects 1 and 7 with the two loading devices at 4% AV.

PAGE 114

99 0 1 2 3 4 5 6 7 8 9 10Differential Rut Depth, DRD (mm) New APA 64CNew APA 70COriginal APA 64C Project 7 Project 1 Figure 7-5. Differential rut depth measurements for Projects 1 and 7 with the two loading devices at 7% AV. 7.5 Rut-Depth Findings The two Superpave monitoring project mixtures were tested at tw o density levels – 4%AV and 7%AV – with the new and original lo ading devices. Tests with the loading strip were run at two temperatures – 64 C and 70C. The new measurement system (surface profile) was used to record results fo r all tests. The key rut-depth findings are the following: Based on absolute rut depth measurements at 4%AV, none of the tests were able to distinguish the better performing mixtur e between Project 1 and Project 7. Absolute rut depth results at 7%AV for all test methods – new APA 64C, new APA 70C, and the original APA – show ed that Project 7 performed better.

PAGE 115

100 The differential rut depth measurements at 4%AV did not how any difference in the results at 64C (original and new APA). There was, however, difference in the DRD results for the new APA 70C. All test methods showed a difference for the DRD at 7%AV. Results from the two Superpave projects tested at 7%AV showed that the new and original APA test methods were able to diffe rentiate the two mixtur es according to their field performance. The issue, however, is whether the performance prediction based on rut depth measurement is adequate to describe the mixture’s ability to resist permanent deformation. It is known that resistance to consolidation is not necessarily related to resistance to shear instability. Tests perf ormed at 7%AV can not conclusively determine whether the mixture is failing primarily due to instability or because of excessive consolidation. The same mixture that fails at 7%AV might demonstrate adequate performance at a higher density level. At 4%AV it is easy to assume that most of the measured rutting will be associated with instability. However, rut depth result s at 4%AV from the new and original APA at 64C did not distinguish between the mixtures. Thus, there is a need to identify a measure or a parameter that is uniquely a ssociated with mixture shear instability. 7.6 Area Change As discussed earlier (Secti on 6.4), calculating the area change between the initial and final surface profiles enables us to dete rmine the predominant mode of permanent deformation – consolidation or instability – of HMA mixture. The failure mode is primarily consolidation if Ai is less than Af (negative percent area change), whereas the failure mode is considered to be primarily consolidation if Af is less than Ai (positive percent change).

PAGE 116

101 Field observations, reported from the Supe rpave monitoring proj ect, show that Project 1 is experiencing highe r rutting than Project 7 and th at the failure mode for the Project 1 rutted sections appears to be in stability [Roque and Villi ers 2003]. Figure 7-6 shows the percent area change for the two mi xtures, at 4%AV, calculated for the new and original APA test methods. All three met hods – new APA 64C, new APA 70C, and the original APA – predicted positive area change (i nstability) for Project 1 and negative area change (consolidation) for Project 7. Figure 7-7 shows the percent area cha nge for the two mixtures, at 7%AV, calculated for the new and orig inal APA test methods. Re sults for the 7% air void content showed that all three methods pred icted positive area change (instability) for Project 1 and negative area chan ge (consolidation) for Project 7. However, statistical analysis for the Original APA results show ed no significant difference between the two projects. 7.7 Unidirectional Loading One of the objectives of this study was to ev aluate the feasibility and the effects of unidirectional (load is applie d in one direction only) loading in the APA. The new loading device was used to test the two Superp ave projects in unidirectional-load mode at two temperatures – 64C and 70C. The unidirectional loading re vealed many issues unknown at the time of the testplanning stage. In unidirectional mode, the APA can test only one sample per mold (the back sample position) because the pneumatic load is applied halfway into the first (front) sample. Figure 7-8 shows the load sequen ce in unidirectional mode. The pneumatic controller initiates the loading sequence at th e beginning of the front sample but the load

PAGE 117

102 appears to peak three inches in to the load path. Because of the inefficient nature of the test procedure and the limited amount of da ta, this part of the study is inconclusive. -5 -4 -3 -2 -1 0 1 2 3 4 5Area Change (%) Negative Area Change Primarily Consolidation Positive Area Change Primarily Instability New APA 64CNew APA 70COriginal APA 64C Project 7 Project 1 Figure 7-6. Area Change measurements for Pr ojects 1 and 7 with th e two loading devices at 4% AV.

PAGE 118

103 -5 -4 -3 -2 -1 0 1 2 3 4 5Area Change (%) Negative Area Change Primarily Consolidation Positive Area Change Primarily Instability New APA 64CNew APA 70COriginal APA 64C Project 7 Project 1 Figure 7-7. Area Change measurements for Pr ojects 1 and 7 with th e two loading devices at 7% AV.

PAGE 119

104 FORWARD RETURN Load=0% Unload Load=100% FrontBack FORWARD RETURN Load=0% Unload Load=100% FrontBack FORWARD RETURN Load=0% Unload Load=100% FORWARD RETURN Load=0% Unload Load=100% FrontBack Figure 7-8. Unidirectional loading in the APA. 7.8 HVS Mixture As mentioned earlier, the HVS coarse-graded mixture was substituted because material for Superpave Project 5, a mixt ure with good field performance, was not available. The HVS coarse-graded mixture was selected because it performed well in tests with the Heavy Vehicle Simulator at the FDOT [Moseley et al. 2003]. Figures 7-9 and 7-10 show the absolute ru t depth and differential rut depth results for the HVS mixture respectively. The mixtur e was tested with the three test methods – new APA 64C, new APA 70C, and the origin al APA at two density levels – 4%AV and 7%AV. All test measurements gave rut-dept h predictions – absolute and differential – below 3mm, which corresponds to the Heavy Vehicle Simulator test observations. The measured rut-depth results – ARD and DRD – from the new APA tests at normal (64C) and high (70C) temperatures were almost identical. This might be

PAGE 120

105 another indication about the ab ility of the new APA loadi ng device to account for the effect of aggregate structur e in rutting prediction. 0 1 2 3 4 5 6 7 8 9 10Absolute Rut Depth, ARD (mm) 4% AV7% AV New APA 70C New APA 64C Original APA 64C Figure 7-9. Absolute rut depth measurements for the HVS mixture with the two loading devices at two AV levels – 4% and 7%.

PAGE 121

106 0 1 2 3 4 5 6 7 8 9 10Differential Rut Depth, DRD (mm) 4% AV7% AV New APA 70C New APA 64C Original APA 64C Figure 7-10. Differential rut depth measur ements for the HVS mixture with the two loading devices at two AV levels – 4% and 7%. 7.9 Discussion The rut-depth findings showed that both de vices were able to distinguish between the two mixtures according to their field perf ormance. However, rut depth by itself is not adequate to determine whether the measured deformation is primarily due to consolidation or because of shear instability. The introduction of the area-change parameter provided a tool to quantify consolidation and shear instability. Both loading devices were able to show the difference in the mode of failure (permanent deformation) for the two mixtur es. Project 1 had a positive area change – primarily instability – and Project 7 had a nega tive area change – primarily consolidation.

PAGE 122

107 The two loading devices were able to disti nguish between the two mi xtures for resistance to shear instability even though the stress di stributions under the two loading mechanisms were found to be very different. The lo ading strip was designed and constructed to simulate stresses – in particular the lateral stresses – found under a radial-tire rib. These lateral stresses were found to be a key factor in the mechanism of instability rutting. Even though the measured stresses unde r the pressurized hose did not show the presence of lateral stresses, the hose was still able to determine that Project 1 failed primarily due to shear instability. The reason behind this phenomenon is the continuously-changing contact area between the hose and the HMA sample. At the beginning of the test the cont act area between the hose and the specimen was measured to be approximately 6-8 mm. Figure 7-11 s hows the initial contact area at the hosespecimen interface. At this stage the stress es induced at the hose-specimen interface are primarily vertical st resses (Section 3.5.2). Figure 7-11. Schematic of the initial hose-specimen contact area.

PAGE 123

108 As the test progresses and the specimen c onsolidates under the vertical stress, the hose ‘sinks’ into the specimen and the c ontact area increases. Figure 7-12 shows a hypothetical contact area at some point beyond 4000 cycles. At this point the specimen is experiencing high shear stresses from the walls of the pressurized hose. Figure 7-12. Schematic of a hypothetical hose-specimen contact area after 4000 cycles. Findings from this study showed that both loading devices – loading strip and pressurized hose – were able to distinguish the better performing mixture based on the area-change calculation. However, the mechanis m that drives the material to instability failure is different for each loading device. Measurements showed that the loading strip induces lateral stresses on the surface of th e specimen at the beginning of the test that remain constant throughout – since the contac t area remains the same. Contrary to that, the pressurized hose does not induce these critical stress states (lateral stresses) until after the specimen is consolidated. Thus, for a material that has good resistance to consolidation, the pressurized hose would be less successful in evaluating the mixture’s ability to resist shear instability. Likewise for a material that has poor resistance to

PAGE 124

109 consolidation, the pr essurized hose could ‘sink-in’ and induce high shear stresses that could lead to the misinterpretation of the mixture’s shear strength.

PAGE 125

110 CHAPTER 8 CONCLUSIONS AND RECOMMENDATIONS 8.1 Conclusions The following conclusions were drawn from this study: The new system (loading strip and profile measurement method) appears to have greater potential of evaluating a mixture’s potential for instability rutting than the original (hose and single rut-de pth measurement) configuration. It is possible to conduct more reliable interpretation of the original APA (pressurized hose) results by using the new system of measuring the entire surface profile of the specimen. The loading strip is a better system in engineering terms, but the APA pressurized hose is more practical and widely available. 8.2 Recommendations The following recommendations are based on the findings and conclusions from this study: The new data-measurement method should be implemented immediately with the existing equipment. For a single-condition test (specimen test ed at one AV level), when testing for instability rutting samples shoul d be prepared at 4% AV. At this point there is not enough evidence to support a move to higher temperature testing for a single-conditi on test (specimen tested at one temperature).

PAGE 126

APPENDIX A AGGREGATE AND MIXTURE VOLUMETRIC PROPERTIES

PAGE 127

112 Table A-1. Gradations and specific gravity of aggregates for Project 1. (#)(mm)(%)(%)(%)(%)(%) 12.5100100100100.0100.0 9.50100100100100.099.0 #44.7532969964.964.0 #82.362738240.340.0 #161.181476028.129.0 #300.601294520.021.0 #500.301153313.614.0 #1000.1516249.18.0 #2000.07503165.85.1 2.6892.6822.7002.6912.667 BlendJMF Sieve SizeSieve #89 Stone W-10Aggregate Blend (%)50.0 18.7 31.3 #89 Stone W-10 M-10Bulk Specific Gravity M-10 Table A-2. Gradations and specific gravity of aggregates for Project 7. (#)(mm)(%)(%)(%)(%)(%) 19.0100100100100.0100.0 12.57410010093.695.0 9.50479210086.088.0 #44.7563710069.170.0 #82.36668757.057.0 #161.18555939.041.0 #300.60543825.730.0 #500.30432215.219.0 #1000.154386.49.0 #2000.0753343.34.2 2.4072.4072.5082.4712.49063.0 S1-B Stone Asphalt ScreensAggregate Blend (%) BlendJMF Bulk Specific Gravity Sieve SizeSieveS1-A StoneS1-A Stone S1-B Stone Asphalt Screens24.5 12.5

PAGE 128

113 Table A-3. Gradations a nd specific gravity of aggregates for the HVS mixture. (#)(mm)(%)(%)(%)(%)(%) 19.0100100100100.0100.0 12.55210010093.893.8 9.50209910089.189.1 #44.755489256.556.5 #82.363117430.130.1 #161.18335620.020.0 #300.60324315.315.3 #500.30222710.010.0 #1000.1522166.56.5 #2000.0752283.63.6 2.3012.3102.3162.3112.31132.0 S1-B Stone Asphalt ScreensAggregate Blend (%) BlendJMF Bulk Specific Gravity Sieve SizeSieveS1-A StoneS1-A Stone S1-B Stone Asphalt Screens13.0 55.0 Table A-4. Batch we ight for Project 1. #89 StoneW-10 GraM-10 Gra#89 StoneW-10 GraM-10 Gra Sieve SizeSieve, mm50%18.7%31.3% 3/4"19.0100100100022503092 1/2"12.5100100100022503092 3/8"9.5100100100022503092 #44.75329699153022843106 #82.3627382220524773345 #161.1814760222826963655 #300.60012945222828473866 #500.30011533222829654035 #1000.1501624222830414162 #2000.0750.42.716.3224130694270 Pan0000225030924500 BATCH WEIGHT, 4500g (Pine) 4500 GradationBatch Weight (g)

PAGE 129

114 Table A-5. Batch we ight for Project 7. S-1-A StoneS-1-B StoneAsphalt ScrnsS-1-A StoneS-1-B StoneAsp Scrns Sieve SizeSieve, mm24.5%12.5%63.0% 3/4"19.0100100100011031665 1/2"12.57410010028711031665 3/8"9.5479210058411481665 #44.75637100103614571665 #82.366687103616312034 #161.185559104716372827 #300.6005438104716433423 #500.3004322105816483876 #1000.150438105816484273 #2000.075333.5106916484401 Pan0000110316654500 BATCH WEIGHT, 4500g (Pine) 4500 GradationBatch Weight (g) Table A-6. Mixture vol umetric properties. Project 1Project 7 Design %AC 5.56.9 Nini 13.7412.96 Ndes 4.074.50 Nmax 2.242.47 Nini 86.2687.04 Ndes 95.9395.50 Nmax 97.7697.53 Nini 24.0023.43 Ndes 15.4715.98 Nmax 13.8614.20 Effective VMA (@ 4% AV) 31.522.8 Nini 42.7444.67 Ndes 73.7371.85 Nmax 83.8782.58 % AC absorption 0.631.68 Effective %AC 4.875.22 Gmm 2.5092.334 Gse 2.7362.573 Gsb 2.6912.470 Effective Film thickness (micrometers) 24.418.6 Dust/Effective AC ratio 1.20.64 % VFA Property % Air viod % Gmm % VMA

PAGE 130

APPENDIX B STATISTICAL ANALYSIS RESULTS

PAGE 131

116Table B-1. Statistical analyses for Pr oject 1 and Project 7 results at 7% AV with the new AP A loading device at 64C. ParameterARD DRDArea Change P value0.00170.0008P<0.0001 P value summary******** Are means signif. different? (P < 0.0 5 YesYesYes Oneor two-tailed P value?One-tailedOne-tailedOne-tailed t, dft=4.682 df=6t=5.470 df=6t=13.43 df=6 ParameterARD DRDArea Change Mean SEM of column A3.860 0.2678 N=45.765 0.3637 N=40.9150 0.09042 N=4 Mean SEM of column B2.155 0.2468 N=43.388 0.2380 N=4-0.4575 0.04768 N=4 Difference between means1.705 0.36422.378 0.43471.373 0.1022 95% confidence interval-2.596 to -0.8139-3.441 to -1.314-1.623 to -1.122 R squared0.78510.8330.9678 ParameterARD DRDArea Change F,DFn, Dfd1.178, 3, 32.336, 3, 33.597, 3, 3 P value0.44810.25210.1605 P value summarynsnsns Are variances significantly different? NoNoNo Unpaired t test How big is the difference? F test to compare variances P1 vs P7 7%AV, 64C, New APA

PAGE 132

117Table B-2. Statistical analyses for Pr oject 1 and Project 7 results at 7% AV with the new AP A loading device at 70C. ParameterARD DRDArea Change P value0.0480.00860.0011 P value summary***** Are means signif. different? (P < 0.0 5 YesYesYes Oneor two-tailed P value?One-tailedOne-tailedOne-tailed t, dft=1.972 df=6t=3.261 df=6t=5.094 df=6 ParameterARD DRDArea Change Mean SEM of column A3.950 0.3479 N=45.728 0.4383 N=40.8350 0.05315 N=4 Mean SEM of column B2.963 0.3600 N=43.808 0.3931 N=4-1.748 0.5042 N=4 Difference between means0.9875 0.50061.920 0.58882.583 0.5070 95% confidence interval-2.213 to 0.2376-3.361 to -0.4793-3.823 to -1.342 R squared0.39340.63930.8122 ParameterARD DRDArea Change F,DFn, Dfd1.071, 3, 31.243, 3, 389.98, 3, 3 P value0.47810.43110.0019 P value summarynsns** Are variances significantly different? NoNoYes Unpaired t test How big is the difference? F test to compare variances P1 vs P7 7%AV, 70C, New APA

PAGE 133

118Table B-3. Statistical analyses for Proj ect 1 and Project 7 results at 7% AV with the original APA loading device at 64C. ParameterARD DRDArea Change P valueP<0.00010.00080.1825 P value summary******ns Are means signif. different? (P < 0.0 5 YesYesNo Oneor two-tailed P value?One-tailedOne-tailedOne-tailed t, dft=13.68 df=4t=7.648 df=4t=1.021 df=4 ParameterARD DRDArea Change Mean SEM of column A6.273 0.08253 N=38.037 0.2771 N=31.017 1.474 N=3 Mean SEM of column B4.347 0.1141 N=35.157 0.2550 N=3-0.7933 0.9855 N=3 Difference between means1.927 0.14082.880 0.37661.810 1.773 95% confidence interval-2.318 to -1.536-3.925 to -1.835-6.732 to 3.112 R squared0.97910.9360.2067 ParameterARD DRDArea Change F,DFn, Dfd1.910, 2, 21.182, 2, 22.237, 2, 2 P value0.34360.45840.3089 P value summarynsnsns Are variances significantly different? NoNoNo Unpaired t test How big is the difference? F test to compare variances P1 vs P7 7%AV, 64C, Original APA

PAGE 134

119Table B-4. Statistical analyses for Pr oject 1 and Project 7 results at 4% AV with the new AP A loading device at 64C. ParameterARD DRDArea Change P value0.43060.49080.0001 P value summarynsns*** Are means signif. different? (P < 0.0 5 NoNoYes Oneor two-tailed P value?One-tailedOne-tailedOne-tailed t, dft=0.1826 df=6t=0.02398 df=6t=7.750 df=6 ParameterARD DRDArea Change Mean SEM of column A2.048 0.1946 N=42.195 0.7630 N=40.9850 0.2123 N=4 Mean SEM of column B1.990 0.2477 N=42.175 0.3365 N=4-2.493 0.3953 N=4 Difference between means0.05750 0.31500.02000 0.83393.478 0.4487 95% confidence interval-0.8282 to 0.7132-2.061 to 2.021-4.576 to -2.379 R squared0.0055240.000095850.9092 ParameterARD DRDArea Change F,DFn, Dfd1.621, 3, 35.142, 3, 33.465, 3, 3 P value0.35070.10590.1673 P value summarynsnsns Are variances significantly different? NoNoNo Unpaired t test How big is the difference? F test to compare variances P1 vs P7 4%AV, 64C, New APA

PAGE 135

120Table B-5. Statistical analyses for Pr oject 1 and Project 7 results at 4% AV with the new AP A loading device at 70C. ParameterARD DRDArea Change P value0.02660.00060.0006 P value summary******* Are means signif. different? (P < 0.0 5 YesYesYes Oneor two-tailed P value?One-tailedOne-tailedOne-tailed t, dft=2.402 df=6t=5.701 df=6t=5.836 df=6 ParameterARD DRDArea Change Mean SEM of column A2.590 0.2935 N=44.910 0.3499 N=41.130 0.4950 N=4 Mean SEM of column B1.810 0.1388 N=42.010 0.3693 N=4-2.210 0.2873 N=4 Difference between means0.7800 0.32472.900 0.50873.340 0.5723 95% confidence interval-1.575 to 0.01455-4.145 to -1.655-4.741 to -1.939 R squared0.49020.84410.8502 ParameterARD DRDArea Change F,DFn, Dfd4.472, 3, 31.114, 3, 32.968, 3, 3 P value0.12510.46570.1978 P value summarynsnsns Are variances significantly different? NoNoNo Unpaired t test How big is the difference? F test to compare variances P1 vs P7 4%AV, 70C, New APA

PAGE 136

121Table B-6. Statistical analyses for Proj ect 1 and Project 7 results at 4% AV with the original APA loading device at 64C. ParameterARD DRDArea Change P value0.37750.25950.0237 P value summarynsns* Are means signif. different? (P < 0.0 5 NoNoYes Oneor two-tailed P value?One-tailedOne-tailedOne-tailed t, dft=0.3341 df=4t=0.7064 df=4t=2.827 df=4 ParameterARD DRDArea Change Mean SEM of column A4.403 0.1915 N=35.743 0.1189 N=31.400 0.6352 N=3 Mean SEM of column B4.107 0.8670 N=34.827 1.292 N=3-0.4233 0.1122 N=3 Difference between means0.2967 0.88790.9167 1.2981.823 0.6450 95% confidence interval-2.761 to 2.168-4.519 to 2.686-3.614 to -0.03284 R squared0.027150.11090.6664 ParameterARD DRDArea Change F,DFn, Dfd20.49, 2, 2118.1, 2, 232.08, 2, 2 P value0.04650.00840.0302 P value summary**** Are variances significantly different? YesYesYes Unpaired t test How big is the difference? F test to compare variances P1 vs P7 4%AV, 64C, Original APA

PAGE 137

APPENDIX C APA RESULTS

PAGE 138

123Table C-1. APA test results for Project 1. SampleSampleAppliedTestWheel NumberIDDesignActualLoadTemperatureTracking1 1-2-3 7.5 4.426.630.98 2 1-2-4 6.4 4.194.850.84 3 S1-7-1 7.7 3.275.811.13 4 S1-7-2 7.7 3.565.770.71 5 S1-7-3 6.7 Uni-directional 3.815.36-2.45 6 1-2-7 7.0 4.015.930.81 7 1-2-8 7.2 3.104.450.79 8 1-2-9 7.2 3.896.100.75 9 1-2-10 6.8 4.806.430.99 10 S1-7-4 6.5 Uni-directional 4.977.79-2.95 11 S1-4-1 3.6 1.860.201.23 12 S1-4-2 3.4 1.641.811.45 13 S1-4-3 3.9 2.153.500.71 14 S1-4-4 3.7 2.543.270.55 15 S1-4-9 4.4 Uni-directional 1.983.05-1.91 16 1-4-3 3.8 2.605.021.18 17 1-4-4 4.4 1.784.062.48 18 S1-4-5 3.9 3.165.760.69 19 S1-4-6 4.4 2.824.800.17 20 S1-4-8 3.3 Uni-directional 4.407.39-2.67 21 1-2-2 6.2 6.157.56-0.31 22 1-2-1 7.4 6.438.52-0.60 23 1-2-6 6.5 6.248.033.96 24 1-4-5 3.9 4.045.512.38 25 1-5-1 4.4 4.485.901.61 26 1-5-2 4.4 4.695.820.21 Original APA 100 lbs 100 lbs 64 oC 64 oC Bi-directional Bi-directional 7% 4%Differential Rutting (mm) Area Change (%)New APAAir Voids4% 150 lbs 150 lbsAbsolute Rutting (mm)Project 1 70 oC Bi-directional 64 oC 70 oC Bi-directional Bi-directional 64 oC Bi-directional 7%

PAGE 139

124Table C-2. APA test results for Project 7. SampleSampleAppliedTestWheel NumberIDDesignActualLoadTemperatureTracking1 7-3-4 6.5 2.853.67-0.52 2 7-3-3 6.7 2.093.84-0.54 3 7-2-4 6.1 1.692.77-0.33 4 7-2-5 7.0 1.993.27-0.44 5 S7-7-3 6.7 Uni-directional 2.712.99-1.98 6 7-2-1 8.3 3.394.91-1.94 7 7-2-2 8.3 3.753.80-0.40 8 S7-7-1 6.4 2.443.12-1.81 9 S7-7-2 6.7 2.273.40-2.84 10 S7-7-4 6.0 Uni-directional 3.134.16-3.93 11 S7-4-3 3.8 1.641.41-1.79 12 S7-4-4 3.9 2.652.71-3.62 13 S7-4-5 4.1 2.092.77-2.18 14 S7-4-6 4.1 1.581.81-2.38 15 S7-4-12 3.7 Uni-directional 1.922.37-2.01 16 S7-4-7 4.1 1.812.26-1.74 17 S7-4-8 4.1 1.811.98-2.96 18 S7-4-9 3.9 1.471.02-2.36 19 S7-4-10 4.6 2.152.78-1.78 20 S7-4-11 3.9 Uni-directional 2.102.72-2.33 21 7-3-6 6.4 4.175.360.39 22 7-3-5 6.5 4.565.46-2.75 23 7-3-2 8.0 4.314.65-0.02 24 7-4-1 4.5 4.405.36-0.47 25 S7-4-1 4.1 2.482.37-0.21 26 S7-4-2 3.9 5.446.75-0.59 Project 7 New APAAir VoidsAbsolute Rutting (mm) Differential Rutting (mm) Area Change (%)7%150 lbs 64 oC Bi-directional 70 oC Bi-directional 4%150 lbs 64 oC Bi-directional 70 oC Bi-directional Original APA 7%100 lbs 64 oC Bi-directional 4%100 lbs 64 oC Bi-directional

PAGE 140

125Table C-3. APA test resu lts for the HVS mixture. SampleSampleAppliedTestWheel NumberIDDesignActualLoadTemperatureTracking1 HVS-7-1 7.3 2 HVS-7-2 7.3 3 HVS-7-3 6.9 4 HVS-7-4 6.8 5 HVS-4-1 4.6 6 HVS-4-2 4.9 7 HVS-4-3 4.4 8 HVS-4-4 4.4 9 HVS-7-5 7.2 10 HVS-7-6 7.1 11 HVS-4-5 4.2 12 HVS-4-6 4.1 2.82-1.32 1.33-1.20 4%150 lbs 2.40 1.30 Bi-directional Bi-directional 100 lbs 64 oC 100 lbs 64 oC HVS New APAAbsolute Rutting (mm) Differential Rutting (mm) Area Change (%) Air VoidsBi-directional 7% 4% 150 lbs 7% 64 oC 70 oC 70 oC 64 oC Original APA 1.631.43-2.68 1.781.81-2.74 1.100.98-1.26 Bi-directional 1.241.07-2.09

PAGE 141

126 REFERENCE LIST Aschenbrener, T., “Evaluation of the Ha mburg Wheel-Tracking Device to Predict Moisture Damage in Hot-Mix Asphalt,” Transportation Research Record 1492, Transportation Research Board, Washington, DC, 1995, pp. 193-201. Asiamah, S., “Relationship between Laborat ory Mix Properties and Rutting Resistance for Superpave Mixtures,” Master’s thesis, University of Florida, Gainesville, 2002. Bathe, K. ADINA System 7.5, User’s Manual. ADINA R&D, Inc. Watertown. MA 2001. Bonaquist, R., J. Sherwood, and K. Stuart, “Accelerated Pave ment Testing at the Federal Highway Administration Pavement Testing Facility,” Association of Asphalt Paving Technologists, Vol. 67, 1998, pp. 690-716. Choubane, B., G.C. Page, and J.A. Musselman, “Investigation of the Suitability of the Asphalt Pavement Analyzer for Predicti ng Pavement Rutting,” Research Report FL/DOT/SMO/98-427, Florida De partment of Transportation, Gainesville, October 1998. City of Hamburg, “Tracking Test, Determination of the Tr ack Depth of High-Stability Binding Layers,” Construction Bureau, Civil Engineering Office, Department of City Traffic, Hamburg, Germany, 1991. Cleveland, W.S., “Robust Locally Weighted Regression and Smoothing Scatterplots,” Journal of the American Statistical Association, Vol. 74, 1979, pp. 829-836. Collins, R., D. Watson, and B. Campbell, “Development and Use of the Georgia Loaded Wheel Tester,” Transportation Research R ecord No. 1492, Transportation Research Board, Washington, DC, 1995, pp. 202-207. Collins, R., H. Shami, and J. S. Lai, “Use of Georgia Loaded Wheel Tester to Evaluate Rutting of Asphalt Samples Prepared by Superpave Gyratory Compactor,” Transportation Research Record No. 1545, Transportation Research Board, Washington, DC, 1996, pp. 161-168. Cook, R. Finite Element Modeling For Stress Analysis. John Wiley and Sons, Inc., New York, 1995. Cooley, A.L., P.S. Kandhal, and S.M. Buch anan, “Loaded Wheel Testers in the United States: State of Practice,” Transportation Research Circ ular, Number E-C016, July 2000, ISSN00978515.

PAGE 142

127 Cort, J.F., and J.P. Serfass, “The Fren ch Approach to Asphalt Mixture Design: A Performance-Related System of Specifi cations,” Association of Asphalt Paving Technologists, Vol. 69, 2000, pp. 794-834. Darku, D., “Evaluation of the Superpave Gy ratory Compactor for Assessing the Rutting Resistance of Asphalt Mixtures,” PhD di ssertation, University of Florida, Gainesville, 2003. De Beer, M., C. Fisher, and F. Jooste, “Determination of Pneumatic Tyre/Pavement Interface Contact Stresses Under Movi ng Loads and Some Effects on Pavements With Thin Asphalt Surfacing Layers,” Proceedings of the Eighth International Conference on Asphalt Pavements, Seat tle, Washington, 1997, pp. 179-226. De Jong, D. L., M. G. F. Peatz, and A. R. Korswagen. “Computer Program BISAR, Layered Systems Under Normal and Ta ngential Loads,” Koninklijke/ShellLaboratorium, External Report AMSR. 0006.73, Amsterdam, The Netherlands, 1973. Drakos, C., “Effect of Measured Tire C ontact Stresses on Near Surface Wheel Path Rutting of Asphalt Pavements,” Master’s thesis, University of Florida, Gainesville, 2000. Drakos, C., R. Roque, and B. Birgisson, “E ffects of Measured Ti re Contact Stresses on Near Surface Rutting,” Transportation Res earch Record No. 1764, Transportation Research Board, Washington, DC, 2001, pp. 59-69. Epps, J., C.L. Monismith, S.B. Seeds, S.C. Ashmore, and T.M. Mitchell, “WesTrack Full-Scale Test Track: Interim Findings,” International Symposium on Asphalt Pavements (ISAP), August 1997, http:// www.westrack.com/wt_04.htm, last accessed on 2/5/2003. Federal Highway Administ ration, SUPERPAVE™ Support and Performance Models Management, “Preliminary Recommendations for the Simple Performance Test,” FHWA No. DTFH 61-94-R-00045, Research Report to the U.S. Department of Transportation, University of Ma ryland, College Park, MD, May 1998. Harvey, J., and L. Popescu, “Accelerated Pavement Testing of Rutting Performance of Two CalTrans Overlay Stra tegies,” Transportation Re search Record No. 1716, Transportation Research Board, Washington, DC, 2000, pp. 116-125. Hibbitt, Karlsson and Sorensen, Inc., Computer Program ABAQUS. Version 5.7, Pawtucket, RI, 1997. Huang, Y.H., Pavement Analysis and Design, Prentice-Hall, Englew ood Cliffs, NJ, 1993. Huber, G.A., “Methods To Achieve Rut-Resi stance Durable Pavements,” Synthesis of Highway Practice 274, Transportation Resear ch Board, National Research Council, Washington, DC, 1999.

PAGE 143

128 Jacobs, Maarten M.J., “Crack Growth in Asphaltic Mixes,” PhD dissertation, Delft University of Technology, Delft, The Netherlands, 1995. Kandhal, P.S., and R.B. Mallick, “Evaluat ion of Asphalt Pavement Analyzer for HMA Design,” Report No. 99-4, National Cent er for Asphalt Technology, June 1999. Kandhal, P.S., and L.A. Cooley, Jr., “Evalu ation of Permanent Deformation of Asphalt Mixtures Using Loaded Wh eel Tester,” Report No. 2002-08, National Center for Asphalt Technology, October 2002. Lai, J.S., “Development of a simplified Me thod to Predict Rutting Characteristics of Asphalt Mixtures,” Research Project 8502, Final Report, Georgia Department of Transportation, Atlanta, 1986. Lai, J.S., and T.M. Lee, “Use of a Loaded-W heel Testing Machine to Evaluate Rutting of Asphalt Mixes,” Transportation Research Record No. 1269, Transportation Research Board, Washington, DC, 1990, pp. 115-184. Leahy, R.B., E.T. Harrigan, and H. Von Quin tus, “Validation of Relationships Between Specification Properties and Perfor mance,” Report No. SHRP-A-409, Transportation Research Board, Nationa l Research Council, Washington, DC, 1994. Marshek, K.M., Chen, H.H., Conell, R.B., and C. L. Saraf, “Effect of Truck Tire Inflation Pressure and Axle Load on Flexible and Rigid Pavement Performance,” Transportation Research Record No. 1070, Transportation Research Board, Washington, D.C., 1986, pp. 14-21. Metcalf, J.B., “Accelerated Pavement Testi ng, a Brief Review Di rected Towards Asphalt Interests,” Journal of the Association of Asphalt Paving Technologist, Volume 67, 1998, pp 553-572. Moseley, H., Page, G., Musselman, J., Sholar, G., and P. Upshaw, “Laboratory Mixture and Binder Rutting Study,” Research Report FL/DOT/SMO/03, Florida Department of Transportation, Gainesville, July 2003. Myers, L., “Mechanism of Whell Path Cracki ng that Initiates at the Surface of Asphalt Pavements,” Master’s thesis, Univer sity of Florida, Gainesville, 1997. Myers, L., “Development and Propagation of Surface-Initiated Longitudinal Wheel Path Cracks in Flexible Highway Pavements,” Ph D dissertation, University of Florida, Gainesville, 2000. Myers, L., R. Roque, B. Ruth, and C. Dra kos, “Measurement of Contact Stresses for Different Truck Tire Types to Evaluate Their Influence on Near-Surface Cracking and Rutting,” Transportation Research R ecord No. 1655, Transportation Research Board, Washington, DC, 1999, pp. 175-184.

PAGE 144

129 Romero, P. and W. Mogawer, “Evaluation of the Superpave Shear Tester Using 19-mm Mixtures from the FHWA ALF,” Journa l of the Associati on of Asphalt Paving Technologist, Volume 67, 1998, pp 573-601. Romero, P., and K. Stuart, “Evaluating Acceler ated Rut Testers,” Public Roads, Federal Highway Administration, U.S. Department of Transportation, Vol. 62, No 1, July/August 1998, p.50. Roque, R., Myers, L., and B.E. Ruth, “Loadi ng Characteristics of Modern Truck Tires and Their Effects on Surface Cracking of Asphalt Pavements,” Proceedings of the Fifth International Conferen ce on the Bearing Capacity of Roads and Airfields, Vol. 1, Trondheim, Norway, 1998, pp. 93-102. Roque, R., L. A. Myers, and B. Birgisson, “E valuation of Measured Tire Contact Stresses for the Prediction of Pavement Resp onse and Performance,” Transportation Research Record No. 1716, Transportati on Research Board, Washington, DC, 2000, pp. 73-81. Roque, R., and C. Villiers, “Progress Repor t No.9: Comprehensive Monitoring of Field Performance of Superpave Project,” UF Project No.: 49104504704-12, Department of Civil and Coastal Engineering, Univers ity of Florida, Gainesville, June 2003. Shenoy, A. and P. Romero, “Superpave Shear Tester as a Simple Standardized Measure to Evaluate Aggregate-Aspha lt Mixture Performance,” ASTM – Journal of Testing & Evaluation, Vol. 29, No. 5, Sept. 2001, pp. 50-62. Sholar, G.A., and G.C. Page, “Follow-up Eval uation of the Asphalt Pavement Analyzer,” Research Report FL/DOT/SMO/99-436, Flor ida Department of Transportation, Gainesville, September 1999. Stuart, K.D., and R.P. Izzo, “Correlation of Superpave G*/Sin Delta with Rutting Susceptibility from Laboratory Mixture Tests,” Transportation Research Record No. 1492, Transportation Research Bo ard, Washington, DC, 1995, pp. 176-183. Stuart, K.D., and W.S. Mogawer, “Validati on of Asphalt Binder and Mixture Tests That Predict Rutting Susceptibility Using th e FHWA ALF,” Association of Asphalt Paving Technologists, Vol. 66, 1997, pp. 109-152. Tayebali, A.A., N.P. Khosla, G.A. Malpass, and H.F. Waller, “Eva luation of Superpave Repeated Shear at Constant Height Test To Predict Rutting Potential of Mixes,” Transportation Research Record No. 1681, Transportation Research Board, Washington, DC, 1999, pp. 97-105. Tekscan, Inc. http://www.tekscan.com/t echnology.html, last accessed on 3/15/03. Texas Department of Trans portation, Manual of Testing Procedures, available at http://manuals.dot.state.tx.us/ dynaweb, last accessed on 2/4/2003.

PAGE 145

130 Tia, M., R. Roque, O. Sirin, and H.J. Kim, “Progress Report No.1: Evaluation of Superpave Mixtures With and Wit hout Polymer Modification by Means of Accelerated Pavement Testing,” UF Project No.: 49104504801-12, Department of Civil and Coastal Engineering, Univers ity of Florida, Gainesville, May 2001. Wang, J.N., C.K. Yang, and T.Y. Luo, “Mechanistic Analysis of Asphalt Pavements, Using Superpave Shear Tester a nd Hamburg Wheel-Tracking Device,” Transportation Research Record No. 1767, Transportation Research Board, Washington, DC, 2001, pp. 102-110. Wang, L.B., J.D. Frost, and J.S. Lai, “Non -Invasive Measurement of Permanent Strain Field Resulting from Rutting in Asphalt Concrete,” Transportation Research Record No. 1687, Transportation Research Board, Washington, DC, 1999, pp. 8594. Williams, R.C., and B.D. Prowell, “Compa rison of Laboratory Wheel-Tracking Test Results with WesTrack Performance,” Tr ansportation Research Record No. 1681, Transportation Research Board, Washington, DC, 1999, pp. 121-128. Witczak, M.W., K. Kaloush, T. Pellinen, M. El-Basyouny, and H. Von Quintus, “Simple Performance Test for Superpave Mix De sign,” NCHRP Report 465, Transportation Research Board, Washington, DC, 2002. Wishnevsky, A. http://home.comset.net/wes ik/grafula3/english.htm, last accessed on 6/13/03. Woodside, A., J. Wilson, and G.X. Liu, “The Distribution of Stre sses At the Interface Between Tyre and Road and Their Effect on Surface Chippings,” Proceedings of the Seventh International Conference on Asphalt Pavements, Vol. 3, Nottingham, UK, 1992, pp. 428-442.

PAGE 146

131 BIOGRAPHICAL SKETCH Christos Andrea Drakos wa s born in Lefkosia, Cyprus, on April 29, 1974, to Andreas and Olga (Phoka) Drakos He graduated from Kyk kos B’ Lyceum in Lefkosia, Cyprus, in 1991 and then joined the Cyprus Armed Forces where he served as an army lieutenant until 1993. The first two years of his college career Christos was enrolled at Old Dominion University, Norfolk, VA. In 1995, Christos tr ansferred to the University of Florida and received a Bachelor of Science degree in civil engineering in 1998. During the summer breaks of his undergraduate st udies, Christos worked for Phoenix Construction Company Ltd., a civil engineering firm in Cyprus which specialized in earthworks and road construction. Christos started working as an undergraduate research assistant in the Department of Civil and Coastal Engineering before going on to complete his Master of Engineering degree in 2000. After completing the doctoral program, Christos intends to remain at the University of Florida as a rese arch assistant professor.


Permanent Link: http://ufdc.ufl.edu/UFE0000997/00001

Material Information

Title: Identification of a physical model to evaluate the rutting performance of asphalt mixtures
Physical Description: Mixed Material
Creator: Drakos, Christos Andrea ( Author, Primary )
Publication Date: 2003
Copyright Date: 2003

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0000997:00001

Permanent Link: http://ufdc.ufl.edu/UFE0000997/00001

Material Information

Title: Identification of a physical model to evaluate the rutting performance of asphalt mixtures
Physical Description: Mixed Material
Creator: Drakos, Christos Andrea ( Author, Primary )
Publication Date: 2003
Copyright Date: 2003

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0000997:00001


This item has the following downloads:


Full Text












IDENTIFICATION OF A PHYSICAL MODEL TO EVALUATE RUTTING
PERFORMANCE OF ASPHALT MIXTURES















By

CHRISTOS ANDREA DRAKOS


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2003

































Copyright 2003

by

Christos Andrea Drakos

































To my parents Andreas and Olga, my nephew Andreas, and my niece Fotini.















ACKNOWLEDGMENTS

I would like to acknowledge those individuals who were involved in the

advancement of this research. First I would like to thank my advisor and mentor Dr. Rey

Roque for willingly sharing his knowledge and experiences through constant support and

advice. Acknowledgments should also be paid to my graduate committee members Dr.

Bjorn Birgisson, Dr. Mang Tia, and Dr. Byron Ruth who were always available to

discuss ideas and lend valuable advice. Also, I would like to thank my external

committee member Dr. Wayne Losano for his technical writing guidance and assistance.

Special thanks and appreciation go to the Florida Department of Transportation,

and more specifically to the bituminous materials group Greg Sholar, Howie Moseley,

Susan Andrews, Frank Suarez, Shanna Johnson, and Stephen Browning for their

support throughout the project. Their help has been invaluable. From the University of

Florida, I would like to thank Mr. George Lopp, Sungho Kim, Marc Novak and Edward

Roske for their technical assistance in the lab.

I want to thank my parents, Andreas and Olga Drakou, for being a source of

inspiration and support throughout my studies in the United States. Finally I would like

to thank my friends Maria Alvey, Eri Messaritaki, Kally Kanellis, Alexis Klironomos,

and Maria Nikolou for their true friendship.















TABLE OF CONTENTS
Page

A C K N O W L E D G M E N T S ................................................................................................. iv

L IST O F TA B LE S ........... .... ........ ... .... ...................... .... .... .............. viii

LIST OF FIGU RE S ................................................................. x

ABSTRACT .............. ............................................ xiv

1 IN TRODU CTION .................. .................. ............. .. ......................... ..

1.1 B background ......... ...... ................................................................... ........... 1
1.2 Problem Statem ent ....... .............................. ............ .. .... .................
1 .3 H y p o th e sis .............................................................................................................. 4
1.4 O bjectiv es ...................... .. ............. .. ........... .................................. . 4
1.5 Scope..................................................... . 5
1.6 Research Approach .. ................................. ...... ..............................6


2 LITERATURE REVIEW ...................... .. ........................ .. ....... ................

2 .1 O overview .................. ...................7..........................
2.2 Permanent Deformation .......................................... ............... ...............
2.2.1 C consolidation R utting............................................................................. 8
2 .2 .2 Instability R cutting ............................................................... .....................9
2.3 Tire-pavem ent Interface Stresses .................................... ............................... 10
2.4 Predicting Mixture Performance ..................................................................18
2.5 A accelerated Pavem ent Testers......... ............................................ ................. 19
2.6 SuperpaveTM Shear Tester ............................................................... ...............20
2 .7 T o rtu re T e sts ............................ ................................ ......................................2 2
2.7.1 H am burg W heel-Tracking D evice ........................................... .............22
2.7.2 French Pavem ent Rutting Tester ..................................... ...............24
2.8 A sphalt Pavem ent A nalyzer ........................................ ................................. 26
2 .9 S u m m ary ............................ ...........................










v









3 A PA LO A D IN G M ECH AN ISM ...............................................................................29

3 .1 O v erv iew ......................................................................................................... 2 9
3.2 Limitations of Loaded Wheel Testers ............... .......................................30
3.3 N ew APA Loading M echanism ....................................... ......................... 32
3.3.1 D evelopm ent of a Tire M odel ....................................... ............... 32
3.3.2 Verification of the Tire M odel ............. ............................ .............. 35
3.3.3 C concept Loading Strip..................................................... ............... 38
3.4 Prelim inary Contact Stress Calculations ................................... .................39
3.4.1 APA Pressurized Hose Stresses....................................... ............... 39
3.4.3 Loading Strip FEM .................. ............................... 41
3.5 M measured Contact Stresses in the APA ..................................... ............... 43
3.5.1 M easurem ent System ............................................ ........... ............... 43
3.5.2 Hose-Specimen Interface Stresses.............................44
3.5.3 Loading Strip-Specimen Interface Stresses..................... .... ..........47
3 .6 S u m m a ry ......... ........................................................................................4 8


4 ST R E S S A N A L Y SE S ........................................................................ .................. 5 1

4 .1 O v e rv iew ...........................................................................................5 1
4.2 Pavem ent Stress A nalyses ............................................. ............ ............... 52
4.2.1 M ulti-Layer Elastic Stress Analyses .................................. ............... 52
4.2.2 B ISA R R esults......... ............................... ............... 54
4.2.3 Finite Elem ent Stress Analyses ...................................... ............... 60
4.2.3-1 Loading the FEM .................................... .......................... ......... 63
4.2.3-2 FEM Results.............. ... .. ...................... .............. 65
4 .3 A P A Stress A n aly ses ........................................ .............................................66
4.4 Sum m ary ........................................................................................ ...................70


5 MATERIALS AND TESTING METHODS.............................71

5.1 O overview ......................................................................... ..... ... ... 7 1
5 .2 M a te ria ls ............................................................................................................... 7 2
5.3 M mixture Preparation ................... ......... .............. .. ..... .............. ... 77
5.3.1 Aggregate Preparation and Batching........................................................77
5.3.2 M ixing ....................................... ....... ............ 77
5.3.4 Short-Term Oven Aging (STOA) and Compaction .................................78
5.4 Asphalt Pavement Analyzer Procedure ..................................... ............... 79
5.4.1 Surface Profile M easurem ent .......................................... ............... 81
5.4.2 APA H ose Testing Procedure................................... ...................... 82
5.4.3 APA Loading Strip Testing Procedure...................................................83
5 .5 S u m m ary ......................................................................................8 3












6 DATA ANALYSIS METHODOLOGY ....................................... ............... 85

6.1 O verview .............................................................................. 85
6.2 Digitizing the Measured Profile................................ ...............85
6.3 R ut D epth C alculations........... .................................. ................. ............... 87
6.4 A rea Calculation ............ .................................. .. .. ...... .. ............ 89
6 .5 S u m m a ry .................................................................................................9 2


7 A PA TE ST R E SU L T S ...................................................................... ...................93

7 .1 O v e rv iew ...........................................................................................9 3
7 .2 F ie ld R e su lts ................................................................................................... 9 3
7.3 A absolute R ut D epth ......................... .. .................... ......... ........... 95
7.4 D ifferential Rut D epth .......................................................... ............... 95
7.5 R ut-D epth Findings .............................................. ......... ......... 99
7 .6 A rea C h ang e ...............................................................100
7.7 U nidirectional L oading ...................... .. ... ........... ................... ............... 101
7.8 H V S M mixture ......................... .......... .. .. ........... .. ..... .. 104
7.9 D discussion ......... ........... ...................................106


8 CONCLUSIONS AND RECOMMENDATIONS......... ....... ..............110

8 .1 C o n c lu sio n s ................................................................................................... 1 1 0
8.2 Recommendations ............................................. ... .......... .... ........ 110


APPENDIX

A AGGREGATE AND MIXTURE VOLUMETRIC PROPERTIES ............... 11......111

B STATISTICAL ANALYSIS RESULTS ...............................................115

C APA RESULTS ............................... ............................... ............ 122

REFERENCE LIST .................... ......................................126

BIOGRAPHICAL SKETCH ............................... ................................ 131
















LIST OF TABLES


Table page

3-1. M material properties used in the tire FEM ................. ............................................. 35

4-1. Material properties and layer thicknesses of FEM pavement structure..................... 63

5-1. Field location of selected m ixtures ................................................. ..... .......... 71

5-2. Aggregate types and sources for the selected FDOT mixtures. .............................72

5-3. Aggregate sources and modified blends for the laboratory mixtures......................73

6-1. Example of a digitized deformation profile from one location. ..............................87

7-1. Field rutting data ................................. ... .......................................... 94

A-1. Gradations and specific gravity of aggregates for Project 1................................. 112

A-2. Gradations and specific gravity of aggregates for Project 7................................ 112

A-3. Gradations and specific gravity of aggregates for the HVS mixture..................113

A-4. Batch weight for Project 1........................................................................113

A -5. Batch w eight for Project 7. ....................................................................... 114

A -6. M ixture volum etric properties. ........................................................................ ...114

B-1. Statistical analyses for Project 1 and Project 7 results at 7% AV with the new
APA loading device at 64C. ............ ....... ............. ................. 116

B-2. Statistical analyses for Project 1 and Project 7 results at 7% AV with the new
A PA loading device at 70 C ...................................................................... .... 117

B-3. Statistical analyses for Project 1 and Project 7 results at 7% AV with the original
APA loading device at 64C. ............ ....... ............. ................. 118

B-4. Statistical analyses for Project 1 and Project 7 results at 4% AV with the new
A PA loading device at 64 C ...................................................................... .... 119









B-5. Statistical analyses for Project 1 and Project 7 results at 4% AV with the new
A PA loading device at 70 C ...................................................... .....................120

B-6. Statistical analyses for Project 1 and Project 7 results at 4% AV with the original
A PA loading device at 64 C ...................................................... .....................121

C-1. A PA test results for Project 1...................................................... ............... 123

C-2. A PA test results for Project 7...................................................... ............... 124

C-3. APA test results for the HVS mixture .............. .......... .................... 125
















LIST OF FIGURES


Figure p

2-1. Schem atic of consolidation rutting. ........................................ .......................... 9

2-2. Schem atic of instability rutting.......................................................... ............... 10

2-3. Three-dimensional vertical and lateral contact stress distributions under radial
(R22.5) truck tire at rated load .........................................................................12

2-4. Schematic of the Smithers system used to measure tire contact stresses ................13

2-5. Structural characteristics of bias-ply and radial truck tires and their effects on the
pavem ent surface .................. ................................ ........ ... ........ .... 15

2-6. Transverse contact shear stresses measured for a bias-ply, radial, and wide-base
radial tire at the appropriate rated load and inflation pressure................................16

2-7. Vertical contact stresses measured for a bias-ply, radial, and wide-base radial tire
at the appropriate rated load and inflation pressure. ..............................................17

2-8. T he S ST test cham ber ............................................................................... .... ........20

2-9. The Hamburg wheel-tracking machine. ....................................... ............... 23

2-10. The French pavement rutting tester. .............................................. ............... 25

3-1. Contact imprints of the rubber hoses with asphalt beam sample. ............................31

3-2. Structural characteristics of a radial tire. ........................................ ............... 33

3-3. Schematic cross-section of a typical radial tire. .............................. ......... ...... .34

3-4. Finite element representation of the tread structure of a radial tire...........................34

3-5. Measured and predicted vertical stress distribution at surface of steel bed..............36

3-6. Measured and predicted transverse stress distribution at surface of steel bed. .........37

3-7. Schem atic of the loading strip. ........................................................ ..................38

3-8. Tekscan pressure measure ent system .......................................... ............... 40









3-9. Graphical interpretation of vertical stresses under the pressurized hose .................41

3-10. Finite element model of the loading strip ....................... ................... .......... 42

3-11. Contact stress measuring apparatus setup and calibration................... ............ 44

3-12. Close-up picture of the pressurized hose test. ................................. ............... 45

3-13. Vertical stress distribution under the pressurized hose. ........................................46

3-14. Close-up picture of the loading strip test............... ........................ ................ ..... 47

3-15. Vertical stress distribution under the loading strip ......... ............................. 49

3-16. Lateral stress distribution under the loading strip. .............................................50

4-1. Load configuration used in BISAR to represent measured stresses under bias-ply
tru ck tire ............................................................................5 3

4-2. Load configuration used in BISAR to represent measured stresses under radial
truck tire. ............................................................................54

4-3. M aximum shear stress distribution........................ .............. ... ............... 55

4-4. BISAR sign convention and maximum shear stress angle a ...................................56

4-5. Schematic of the maximum shear stress direction representation...........................56

4-6. Magnitude and direction of maximum shear stresses under radial tire load ............58

4-7. Magnitude and direction of maximum shear stresses under bias-ply tire load. ........59

4-8. Three-dimensional finite element mesh used in the pavement response analysis.....61

4-9. Plan view of the contact area of the three-dimensional mesh used in the pavement
response analy sis. ................................................... ................. 62

4-10. D definition of the shape functions....................................... .......................... 64

4-11. Cross-section view of surface elements with nodal forces for the radial-tire
load..................... ....................................... 64

4-12. Maximum shear stress magnitude (psi) and direction under the modeled
radial-tire load ..................................................... ................. 65

4-13. Top view of the finite element model for the APA mold and specimen ...............67

4-14. Three-dimensional finite element model for the APA mold and specimen. ...........67









4-15. Maximum shear stress magnitude (psi) and direction under the modeled loading
strip load .............................................................................69

4-16. Maximum shear stress magnitude (psi) and direction under the modeled
pressurized hose load. ................................... .............. ...... ............ 69

5-1. Gradation chart for JMF and laboratory blend for Project 1 (9.5mm maximum
nom final size). ...................................................... ................. 74

5-2. Gradation chart for JMF and laboratory blend for Project 7 (12.5mm maximum
nom final size). ...................................................... ................. 75

5-3. Gradation chart for laboratory blend for the HVS coarse-graded mixture (12.5mm
m axim um nom inal size). ............................................... ............................... 76

5-4 P ine G yratory C om pactor ............................................................... .....................78

5-5. Original A PA m easuring plate. ............................................................................ 80

5-6. New measuring plate with elongated slits. ..................................... ...............80

5-8. Recording the deformed shape of the contour gauge. .............................................82

6-1. G rafula3 screen shot. ...................... .................... ................. .... ...... 86

6-2. Deformation profile for a specimen tested with the pressurized hose .......................88

6-3. A rea change interpretation................................................ ............................. 89

6-4. Initial surface profile and area calculation. .................................... .................91

6-5. Final surface profile and area calculation................... .............. ...............91

7-1. M measured field rut depth per million ESAL. .................................. .................94

7-2. Absolute rut depth measurements for Projects 1 and 7 with the two loading
devices at 4% AV. ........................................ .......................96

7-3. Absolute rut depth measurements for Projects 1 and 7 with the two loading
devices at 7% AV. ........................................ .......................97

7-4. Differential rut depth measurements for Projects 1 and 7 with the two loading
devices at 4% AV. ........................................ .......................98

7-5. Differential rut depth measurements for Projects 1 and 7 with the two loading
devices at 7% AV. ........................................ .......... .............99

7-6. Area Change measurements for Projects 1 and 7 with the two loading devices
at 4% A V ................................................................... ........................... 102









7-7. Area Change measurements for Projects 1 and 7 with the two loading devices
at 7% AV ................................... ................................. ......... 103

7-8. Unidirectional loading in the APA. .............................................. ............... 104

7-9. Absolute rut depth measurements for the HVS mixture with the two loading
devices at two AV levels 4% and 7%. ...................................................... 105

7-10. Differential rut depth measurements for the HVS mixture with the two loading
devices at two AV levels 4% and 7%. ...................................................... 106

7-11. Schematic of the initial hose-specimen contact area ................... ...... .......... 107

7-12. Schematic of a hypothetical hose-specimen contact area after 4000 cycles. ........108















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

IDENTIFICATION OF A PHYSICAL MODEL TO EVALUATE THE RUTTING
PERFORMANCE OF ASPHALT MIXTURES

By

Christos Andrea Drakos

August 2003

Chair: Reynaldo Roque
Cochair: Bjorn Birgisson
Major Department: Civil and Coastal Engineering

Near-surface rutting has become a costly mode of failure for interstate highways.

Even though the pavement is structurally sound, the hazard of water ponding and

hydroplaning urges the transportation agencies to rehabilitate the deformed sections.

This study was intended to identify a physical model that can provide reliable predictions

about a mixture's ability to resist permanent deformation.

Analyses performed with the elastic layer analysis program BISAR and the FEM

code ADINA provided information on the pavement's response to actual tire loading.

The analyses provided evidence that stress states in the pavement are dependent on tire

structure. Furthermore, it was found that radial truck tires induce severe near-surface

stress states that have been identified as key factors in the mechanism of instability

rutting.

The APA is a laboratory torture test that subjects a specimen to an accelerated

loading sequence. The end result (rut depth) can be then correlated to the rutting









performance of the mixture in the field. However, the ability of the test to replicate field

conditions in the laboratory determines the reliability of the results. It was shown that the

APA loading mechanism, the pressurized hose, was not capturing the critical lateral

stresses found to be detrimental to HMA pavements.

Based on the tire study results, a new APA loading device was introduced to better

replicate the stresses found under radial tires. Contact stress measurements under the two

loading devices pressurized hose and loading strip showed that the loading strip was

able to reproduce the lateral stresses found under individual ribs on a radial tire tread.

Subsequent finite element modeling also showed that the loading strip appeared to

generate similar shear stress patterns to those found under the modeled radial-tire load.

A new method was developed to measure deformations, where a contour gauge is

used to record and store the entire surface profile of the sample throughout the progress

of the test. The area-change parameter was introduced to calculate the volumetric

changes in the sample. Based on the area-change parameter we can calculate whether the

specimen is failing primarily due to shear instability or because of excessive

consolidation.

The introduction of the new loading device and the area-change parameter provided

valuable information about the mixtures behavior. Test performed at low air void

content, to control consolidation rutting, showed that both loading devices loading strip

and pressurized hose were able to provide accurate predictions about the mixture's

susceptibility to instability rutting.














CHAPTER 1
INTRODUCTION

1.1 Background

A major distress mode of flexible pavements is permanent deformation, also known

as rutting. Rutting is characterized by a depression that forms in the wheel paths and can

be the result of permanent reduction in volume (consolidation/traffic densification),

permanent movement of the material at constant volume (plastic deformation/shear), or a

combination of the two. This mode of failure reduces serviceability and creates the

hazard of hydroplaning because of accumulated water in the wheel-path ruts.

Rehabilitation of rutted pavements usually involves asphalt concrete (AC) overlay,

recycling, or replacement of all structural layers.

The SuperpaveTM mix design and analysis method was developed more than a

decade ago under the Strategic Highway Research Program (SHRP) [Leahy et al. 1994].

Many agencies in North America including the Florida Department of Transportation -

have adopted the Superpave method of performance-grade (PG) binder specification and

the volumetric mixture design method. Although the Superpave volumetric design

procedure has resulted in some improvements over the Marshall method of mixture

design, it is still devoid of a general strength test that would determine the mixture's

suitability for resistance to rutting and cracking. The industry has expressed the need for

a simple 'pass-fail' type of test to complement the Superpave volumetric mix design

method, especially for use on design-build or warranty projects.









Numerous performance prediction models numerical and physical have been

implemented to classify an asphalt mixture's ability to resist rutting. In an effort to

control this type of distress, many institutions and agencies are searching for a simple

performance test that would indicate the rutting potential of hot-mix asphalt (HMA). For

this purpose the suitability of various loaded-wheel testers (LWT), as a physical model, is

being examined throughout the country. The LWTs provide an accelerated performance

evaluation by subjecting the designed mix to repeated loading under various

environmental conditions (moisture and temperature). Some of the most popular devices

used are the Georgia Loaded Wheel Tester (GLWT), Asphalt Pavement Analyzer (APA),

Hamburg Wheel Tracking Device (HWTD), and the French Pavement Rutting Tester

(FPRT) [Cooley et al. 2000, Federal Highway Administration (FHWA) 1998].

A series of studies on the suitability of wheel testers to predict the rutting

performance of asphalt mixtures gave mixed results [Epps et al. 1997, Bonaquist et al.

1998]. These types of physical models are classified as empirical or performance-related

tests because they do not measure a fundamental property that can be used to explain and

identify the mechanisms resulting in surface distress. The models' accuracy relies on

how well (realistically) conditions have been simulated in the lab.

Pavement performance has been negatively influenced by the change in traffic type

and volume in recent years. More than 98% of trucks are now using radial tires that can

withstand higher inflation pressures and higher loads. Studies have shown that radial

tires induce high lateral stresses that cause tensile stresses at the surface of the pavement

[Myers 1997]. Furthermore, it was shown that tire structure has a significant influence on

contact stresses; in fact, stress states induced by radial and wide-base radial tires were









determined to be potentially more detrimental to the pavement's surface than stress states

induced by bias ply tires [Myers et al. 1999]. Thus, it is important that the load

characterization in the physical model captures the complexity of the load experienced in

the field.

In most LWTs, the loading device in the form of a wheel or a pressurized hose -

is tracked back and forth over a testing sample to induce rutting. The load follows the

same path in both directions (no wander). In recent tests with the Heavy Vehicle

Simulator (HVS), researchers have found that there is a difference in the rate of

accumulated permanent deformation between loading in unidirectional and bi-directional

mode [Harvey and Popescu 2000, Tia et al. 2001]. The unidirectional loading had a

higher rate of rutting (reached failure in fewer cycles) over the same pavement sections

than the bi-directional loading.

1.2 Problem Statement

Factors that influence the amount of rutting or contribute to the pavement's

resistance to failure have not been clearly identified. Therefore, the lack of confidence in

existing physical models has inhibited their application for prediction of pavement

rutting. This study will evaluate the importance of capturing the specific loading,

environmental, structural, and construction characteristics on the development of a

reliable physical model. There is a need to identify this set of conditions realistic

contact stresses (loading), thermal gradients in asphalt layer (environment/structural), and

test specimen conditioning (construction) that may dictate performance. Traditionally,

mixture evaluation includes average conditions that do not typify the critical condition or

capture the key factors that lead to this kind of failure.









1.3 Hypothesis

Rutting performance of mixtures may be governed by a set of critical field

conditions compaction, load, temperature, and sequence of loading. Therefore, it is

essential to identify the critical design condition (mechanism) that must be replicated in a

laboratory physical model. A physical model that will employ this set of critical

conditions might produce more reliable results for mixture rutting performance.

1.4 Objectives

The main objective of this study is to identify the loading, environmental, and

construction (density) factors that are critical/essential in defining the mechanism of

rutting. The identification of these conditions will logically lead to the development of a

reliable physical model. The current version of the APA was selected for the

experimental program; necessary modifications will be made to incorporate new testing

procedures, which more realistically simulate traffic and environmental conditions

existing on pavements.

The primary objectives of this research study are listed below:

* Identify the characteristics of a loading device necessary to represent a tire load
more realistically.

* Design and construct a new loading device to induce more realistic contact stresses.

* Verify the effects of loading characteristics on rutting performance.

* Examine the importance of unidirectional versus bi-directional loading in the
physical model.

* Evaluate the importance of density/loading history on rutting performance.

* Investigate the sensitivity of the physical model to mixtures with different densities
as produced by compaction and/or aggregate gradation of the mixtures.









* Recommend test configuration and procedure/system for mixture evaluation. As
envisioned, the procedure will define the magnitude and sequence of loading as
well as test-temperature requirements.

1.5 Scope

The research focuses on identifying the critical conditions that contribute to the

mechanisms) of rutting. Defining the conditions that might initiate and propagate rutting

will lead to the development of better performance prediction models physical and

numerical. However, it will not be feasible to examine all possible parameters that affect

rutting within the limited time. Thus, this research will focus on the effects of the

following:

* Load configuration. A new loading device (rib) will be evaluated against the
existing pressurized hose. The contact stresses will be measured for both devices
and then used in finite element modeling (FEM) to calculate the induced stress
states in the specimen.

* Temperature. Two temperatures 64 and 70C have tentatively been selected for
evaluation of mixture's sensitivity to temperature changes.

* Mixture density. Mixtures will be tested at two levels 93-94, and 95-96% of
maximum theoretical density (MTD).

* Three mixtures of known field performance will be used for the initial development
and the evaluation/validation of the physical model. The following mixtures were
chosen from the SuperpaveTM monitoring project:

Poor field performing mixture I-10 Madison County.

Good field performing mixture Turnpike Palm Beach.

Good performing mixture HVS coarse-graded mix.

The Pine gyratory compactor will be used to prepare 150-mm-diameter by 75-mm-

thick mixture specimens. In this research, beams will not be considered because of

compaction issues and the potential for variability that may influence the analysis.









1.6 Research Approach

The research is divided into two parts the analytical and the experimental. The

analytical part includes the conceptual design, the finite-element analysis (FEA) of

contact stresses, and the analysis of the stress states within the physical model.

Subsequently the experimental part includes the laboratory testing with the physical

model and data analysis. A research-approach outline is detailed below:

* Literature Review: examine existing ideas, theories, and results published on tire
contact stresses, critical stress states, and rutting in asphalt concrete pavements.
Also, review work done on other accelerated physical models LWT, Accelerated
Loading Facilities (ALF), and HVS and the effects of certain parameters load,
temperature, and density on the reported results.

* Tire Contact Stresses: design a loading device to induce stresses that would be
representative of the actual stresses induced by truck tires, based on the tire contact
stresses studies.

* New Loading Device: construct and test the new loading device for compatibility
and durability issues. Measure the contact stresses under both loading devices rib
and pressurized hose and compare the stress distribution. Analyze the measured
contact stresses with FEM to calculate the stress states in the test specimen, and
compare the results for the two loading devices.

* Validate New Loading Device: test two mixtures with known field performance -
good and poor in the modified APA and evaluate the new loading device's ability
to produce reliable results.

* Compare Loading Devices: test mixtures with known field performance that the
existing APA failed to correctly predict. Compare the results from the modified
APA to the field performance of the mixtures.

* Evaluate Factors Affecting Performance Prediction: 1) examine the effects of
unidirectional versus bi-directional loading, 2) test at different temperatures and
densities to obtain temperature-density trends for rut depth development, 3)
examine the effects of sequence of loading. Analyze the test data and establish the
test conditions that are critical in predicting rutting performance.

* Validate Proposed Test Conditions: test mixtures with known performance and
evaluate the reliability of the results.














CHAPTER 2
LITERATURE REVIEW

2.1 Overview

In recent years permanent deformation in HMA pavements has generated much

concern as many states have experienced an increase in the severity and extent of this

type of failure. Research suggests that the steady increase in truck-tire pressure and axle

load, which has been noted for the past 20 years, altered the tire-pavement contact stress

characteristics. As a result, the pavement surface is exposed to higher stresses than the

levels assumed when designing in accordance with the 1993 AASHTO Guide for Design

of Pavement Structures.

Concern about high truck-tire pressures and rutting led to a national symposium

and sparked interest on tire-pavement interface stresses. Subsequent research concluded

that increased truck weights and tire pressures had led to an increase in pavement

distress. Measurements of the tire-pavement contact stresses helped identify possible

reasons behind surface-initiated cracking and near-surface rutting.

Focus also shifted to material selection, mix design, and construction practices

improvements that could minimize rutting. One of the attempts to improve HMA

mixture performance was the development and subsequent adoption of the Superpave

Volumetric Mixture Design Method. Although the Superpave design procedure resulted

in some improvements over previous mixture design methods, it has yet to incorporate a

general strength test that would determine the mixture's ability to resist rutting and

cracking. The need for a simple 'pass-fail' type of test to complement the Superpave









volumetric mix design method forced various agencies to search for a suitable

performance test.

This chapter reviews some of the literature available on the subjects of permanent

deformation, tire-pavement interface stresses, accelerated pavement testers, and

laboratory methods for predicting mixture performance. The literature review focuses

more on the three most popular torture tests HWTD, FPRT, and APA with more

weight on the leading candidate, the Asphalt Pavement Analyzer.

2.2 Permanent Deformation

Permanent deformation, also known as rutting, is an unrecoverable deformation in

the form of a depressed channel in the wheel path of the pavement. Rutting can be

attributed to excessive consolidation, formed by an accumulation of permanent

deformations caused by repeated heavy loads, or lateral movement of the material, caused

by shear failure of the asphalt concrete layer, or a combination of the two.

2.2.1 Consolidation Rutting

Consolidation rutting occurs when one or more layers of the pavement structure

(usually the subgrade) undergo further densification by reduction of air voids, or loss of

moisture in the case of clay soils. The structure is especially susceptible to this type of

distress when there is insufficient compaction during pavement construction. A layer

with insufficient density is prone to further densification under traffic, especially in hot

weather (for asphalt concrete layers) and at intersections where the loads are slow-

moving or static. Figure 2-1 shows a schematic of consolidation rutting which is

distinguished by a depression that occurs in the wheel path with no humps on either side

of that depression [Huber 1999].




















subgrade
weak subgrade or underlying layer su
deformation


Figure 2-1. Schematic of consolidation rutting.

The subgrade (native soil) is the most vulnerable layer for consolidation rutting

because it is the weakest material of the pavement structure. If the pavement structure

fails to reduce the vertical stress/strain to allowable limits on the subgrade level, either by

improper thickness design or by unexpected increase of load magnitude, under repeated

loading the layer will experience excessive consolidation [Huang 1993]. Once the

foundation (subgrade) has collapsed, the remaining yielding layers in the pavement

structure will conform to the new contour shape of the supporting layer, resulting in

subsidence ruts. These ruts tend to be fairly wide (30 40 in) with a shallow sloping

saucer-shaped cross section [Huber 1999].

2.2.2 Instability Rutting

Instability rutting is strictly an AC layer type of failure, usually within two inches

from the surface, and it is attributed to the mixture characteristics of the HMA. The

surface material is laterally displaced along shear planes within the AC layer, which

shows signs of mixture instability (low shear resistance). Figure 2-2 shows a schematic









of instability rutting which is characterized by longitudinal ruts in the pavement with

humps of material on either side of the rut [Huber 1999].




original
profile






shear plane



Figure 2-2. Schematic of instability rutting.

Asphalt concrete is a bonded granular material mix of aggregates and asphalt

cement. Under repeated traffic loads, the aggregates do not deform but rather rigidly

translate and rotate within the asphalt binder [Wang et al. 1999]. This effect (aggregate

movement) is amplified at low initial compaction or at high asphalt content. Instability

rutting is the surface manifestation of the aggregate skeleton evolution under repeated

traffic loading of mixtures with low shear resistance. Temperature, rate of loading, and

magnitude of loading directly affect the performance of the mix and influence the

severity of this type of distress.

2.3 Tire-pavement Interface Stresses

Engineers have tried for many years to measure the three-dimensional contact

stresses between a tire and the pavement. The results of such measurements would help

tire engineers to design better, more resilient tires, and would also enable pavement

engineers to analyze the stress states under a tire and evaluate their influence on the

structure.









Woodside et al. (1992) developed a device to measure the contact stress patch

between the tire and underlying material in the laboratory. A steel-bed transducer array

measured normal and tangential contact stresses, in static or dynamic mode, under radial

tires. The steel plate system was fitted with 12 transducers and repeatedly measured a

strip transversely every 5 mm over the entire contact patch. The device recorded 90

contact stress measurements for each test. The contact stresses were then used in the

evaluation of durability of surface chippings on asphalt overlays. Results concluded that

implementation of surface chippings of 1 mm macro-texture may improve skid resistance

on pavements.

Researchers in South Africa measured three-dimensional stresses under bias-ply,

radial, and wide-based radial truck tires at different loads and inflation pressures with the

Vehicle-Road Pressure Transducer Array (VRSPTA) System [De Beer et al. 1997]. The

experimental setup consisted of 13 triaxial strain gauge steel pins (spaced 17mm

transversely) mounted on a steel plate and fixed flush with the road surface. Figure 2-3

shows some of the VRSPTA results that reveal a non-uniform vertical and lateral stress

distribution.

Dr. Marion Pottinger of Smithers Scientific Services, Inc. (Ravenna, Ohio)

developed a device to measure the tire-pavement contact stresses under truck tires. The

device measures vertical, transverse, and longitudinal forces and displacements under a

moving tire using a series of 16 transducers, as shown in Figure 2-4. The tire is held at

one location while the steel bed is moved in the longitudinal direction, forcing the tire to

roll over the transducers that measure displacements and stresses. The procedure is

repeated by placing the tire at different transverse positions to acquire a detailed pattern







12


of the three-dimensional stresses under the tire. Measurements taken at every 0.14 inches


in the transverse and 0.10 inches in the longitudinal direction resulted in about 4000


stress measurements in each of the three axes under the tire-contact area [Myers et al.


1999].




In- ,-' -. -
I N






_)
5 1 5-------- l





t:::'' 0 I




> from -_

Longiludinal rear Lateral

Vertical contact stress distribution.


o 0.2
0 _
-- 01

G.
a)


T -0 1

ra -0.2
-J


-I


AI -r


Lateral


Lateral contact stress distribution.


10Lon

Longitudinal


Figure 2-3. Three-dimensional vertical and lateral contact stress distributions under
radial (R22.5) truck tire at rated load [after De Beer et al. 1997].













Bed Motion

16 Transducers


Coaxial Load and
Displacement
Transducer Detail


Figure 2-4. Schematic of the Smithers system used to measure tire contact stresses.

Analysis of Pottinger's results gave new information about the way tires load the

pavement and the factors that influence the contact stress distribution. It appears that the

most important factor is the tire structure. Bias-ply tires have a more flexible tread that

allows the tire to bulge out upon inflation. When the tire is loaded the contact area

becomes flat and the bulging is reversed, causing transverse stresses to pull the pavement

towards the center of the tire. This is commonly referred to as thepneumatic effect and it

is more prevalent in bias-ply tires.

A phenomenon referred to as the Poisson effect has a significant bearing on

transverse shear stresses. This effect is based upon the principle that, unless restrained,

most materials expand laterally when loaded vertically. When individual ribs under a tire

are loaded they attempt to expand laterally, and transverse stresses are generated when

the surface of the pavement tries to restrain the expansion. This effect induces transverse

shear stresses that pull the pavement apart under each rib.


O'y, 8y


(x, 8x


oTz, 8z









The Poisson effect, which induces tension under each tire rib, is present in both

radial and bias-ply tires. However, because of the large pneumatic effect, tensile stresses

are eliminated in all but the center rib of the bias-ply tire. The tensile stresses from the

Poisson effect dominate the transverse contact stress distribution in radial tires because

the pneumatic effect is negligible. The composite of the pneumatic and Poisson effects

for each tire type is shown schematically in Figure 2-5 [Roque et al. 1998].

Recent work at the University of Florida focused on the effects of tire type,

loading, and inflation pressure on measured contact stresses under various types of truck

tires radial, bias-ply, and wide-base radial. Stress states induced by radial and wide-

base radial tires were determined to be potentially more detrimental to the pavement's

surface than stress states induced by bias-ply tires. The tire structure caused a significant

difference in the distribution of lateral contact stresses under radial and bias-ply tires,

whereas no significant difference was found in the vertical stress distribution.

Distributions of lateral and vertical contact stresses under the three tire types are

shown in Figures 2-6 and 2-7 respectively. The studies concluded that the contact stress

distributions measured under radial truck tires appear to contribute to the prevalence in

recent years of surface-initiated wheel path cracking and near-surface rutting [Drakos et

al. 2001, Myers et al. 1999].








Bias Ply Tire


Flexible Wall

More Rigid
Tread -


Flexible
NTread \


More Rigid
Wall


Pneumatic
Effect


Poisson
Effect





Composite
Effect


Figure 2-5. Structural characteristics of bias-ply and radial truck tires and their effects on
the pavement surface [after Roque et al. 1998].


Radial Tire


IIII














80
Wide-base Radial
E- Radial
S60 -
S- Bias ply


40


20


0








-60
S-40------ -I-- ^ --






-80
0 2 4 6 8 10 12 14

Lateral Location, X (in)




Figure 2-6. Transverse contact shear stresses measured for a bias-ply, radial, and wide-
base radial tire at the appropriate rated load and inflation pressure [after Myers
et al. 1999].









-300



-250


N
-200



-150



S -100
..
w_


Wide-base Radial
-- Radial
-Bias ply


.




-I
I I I 'II


I II II
I I-


,I II II I I
------------ { ^ B ^ l --- l|-- Jl --- l^- ----


Lateral Location, X (in)


Figure 2-7. Vertical contact stresses measured for a bias-ply, radial, and wide-base radial
tire at the appropriate rated load and inflation pressure [after Myers et al.
1999].









2.4 Predicting Mixture Performance

There are three general types of laboratory mixture design and evaluation tests -

performance related, performance based, and torture tests. Performance-related tests

measure properties or responses related to mixture performance (density, air voids, etc.)

but individually are insufficient to drive a performance prediction model. Performance-

based tests measure material properties (resilient modulus, complex modulus, etc.) that

can be used in fundamental response models to predict mixture response to imposed truck

and environmental loads. Torture tests, also known as index tests, are empirical and

apply very severe or extreme loading conditions on the test specimen to evaluate a type

of failure condition in the mix [FHWA 1998].

As stated above, there is a growing need for a simple performance test (SPT) that

can accurately predict the mixture's ability to resist rutting and cracking. M.W. Witczak

defined the SPT as follows:

A test methods) that accurately and reliably measures a mixture response
characteristic or parameter that is highly correlated to the occurrence of pavement
distress (e.g., cracking and rutting) over a diverse range of traffic and climatic
conditions. [Witczak et al. 2002, p.1]

Based on the definition above, it is not necessary for the SPT to predict the entire

distress or performance history of the HMA mixture, but the test results must indicate the

mixture's ability to resist fracture and permanent deformation under defined conditions.

Various agencies are conducting evaluation studies to identify the most suitable test that

would accompany the volumetric design. The following sections discuss some of the

most popular candidates: three torture tests and the performance-based Superpave Shear

Tester.









2.5 Accelerated Pavement Testers

The use of accelerated pavement testing (APT) for determining mixture

performance has increased in the past 20 years because of APT's ability to apply traffic

loads in a short time. APT facilities enable us to evaluate potential materials, designs,

and features, in a 'real' (actual size) environment. In a review of existing APT facilities,

J.B. Metcalf defined the APT as:

Full scale accelerated pavement testing is defined as the controlled application of a
prototype wheel loading, at or above the appropriate legal load limit to a prototype
or actual, layered, structural pavement system to determine pavement response and
performance under a controlled, accelerated, accumulation of damage in a
compressed time period. [Metcalf 1998, p. 556]

The appeal of the APT facilities is that they give the closest simulation to real

condition of an actual in-service pavement, in terms of materials and construction

procedures, though the effects of aging and the environment are in most cases limited.

Most APT facilities use load systems that approximate actual traffic by incorporating

wheel wander and unidirectional loading capabilities (the wheel does not load in both

directions). These facilities are able to work around the clock and produce early results

with a high level of credibility.

In the search for a simple performance test, accelerated pavement testing facilities

provide reliable mixture-performance information to evaluate results from various LWTs

[Bonaquist et al. 1998, Epps et al. 1997, Romero and Stuart 1998]. These results can be

used to evaluate the LWT's ability to capture the true performance characteristics of the

HMA.









2.6 SuperpaveTM Shear Tester

The SuperpaveTM Shear Tester (SST), previously known as the Simple .\/wr

Tester, was produced by the Strategic Highway Research Project (SHRP) that measures

mixture properties to determine pavement performance. The SST is a servo-hydraulic

system that can apply axial loads, shear loads, and confinement pressures to asphalt

concrete specimens at controlled temperatures while monitoring sample deformation

[Shenoy et al. 2001]. The machine has six main components: testing chamber, test

control system, environmental system, hydraulic system, air pressurization system, and

measurement transducers. Figure 2-8 is a close-up of the testing chamber.


Figure 2-8. The SST test chamber.









The specimens have a diameter of 150 mm (5.90 in) and a height of 50 mm (1.97

in); however, the system can test specimens with diameters and heights up to 200 mm

(7.87 in) with only minor modifications. The environmental system is able to precisely

maintain the temperature inside the testing chamber anywhere between 0 and 70C.

Three tests are usually performed with the SST: a) Simple Shear at Constant Height

test (SSCH), b) Frequency Sweep at Constant Height test (FSCH), and c) Repeated Shear

at Constant Height test (RSCH). The American Association of State and Highway

Transportation Officials (AASHTO) Provisional Standard TP7-94 contains a detailed

description of the SST test in the different modes of operation.

There is an ongoing evaluation of this test procedure to establish the accuracy and

repeatability of results. A study by the Federal Highway Administration (FHWA)

showed that the SST tests can accurately discriminate between different asphalt binders,

but are insensitive to aggregate changes [FHWA 1998]. Other researchers have reported

high variability (15-30%) and that their results depended on the data analysis method

[Romero et al. 1998].

An evaluation study performed in North Carolina on three pavement sections with

known performance showed that the SST was able to rank the mixtures according to their

field performance [Tayebali et al. 1999]. A recent study used the SST and the HWTD to

test two mixtures Superpave and Marshall for rutting and stripping susceptibility.

Both tests SST and HWDT gave consistent results and showed that the Superpave

mix was more resistant to permanent deformation [Wang et al. 2001].









2.7 Torture Tests

Torture testing devices subject a test specimen of the mixture to repeated loading

applied by a traveling wheel. The tests do not require any preparation other than the

molding of the specimen, and the results usually report the rut depth on the mixture

specimen as a function of the load applications. Some of the most popular torture testing

devices include the APA, the HLWT and the FPRT, all of which use the same basic

principle that the rut depth measured from the specimen can be correlated to field

performance.

Although the idea is simple and seems to pinpoint a direct measurement that can be

an indication of the mixture's performance, recent studies have shown that the tests failed

to distinguish between good and poor performing mixtures [Collins et al. 1996, Stuart et

al. 1997]. The load-transfer mechanism, boundary conditions (confinement), and the

ratio of the size of the loading wheel to the aggregate size are some of the reasons that

limit the success of LWTs [FHWA 1998].

Overall, the idea of the loaded wheel tester has a very intuitive appeal because it is

simple, relatively low cost, and easy to use. However, the tests correlate the rut-depth

measurement directly to field performance without identifying any fundamental

properties that will help us improve our mixture design.

2.7.1 Hamburg Wheel-Tracking Device

The Hamburg wheel-tracking device measures the combined effects of rutting and

moisture damage by rolling a steel wheel across the surface of an asphalt concrete sample

immersed in hot water. Esso A.G. of Hamburg, Germany developed the device in the

1970, and originally called it the "Esso Wheel-Tracking Device." Initially, the City of

Hamburg used the testing machine to measure rutting susceptibility of HMA. The









sample was subjected to 9,540 wheel passes at 40 or 500C, and the resulting rut depth was

used as a pass/fail criterion to ensure mixture performance [City of Hamburg 1991].

Today, the Hamburg testing machine is manufactured by Helmut-Wind, Inc. of

Hamburg, Germany. It can test two asphalt slab samples simultaneously with

dimensions of 320 mm in length by 260 mm in width, and thickness of 40, 80, or 120

mm. The samples are prepared to 7 1 % air voids, and are submerged into 500C water

for the test. A steel wheel, measuring 203.5 mm (8 in) in diameter and 47 mm (1.85 in)

in width, loads each sample with a fixed load of 705 22 N (158 5 lbs) at a rate of 50

passes per minute [Aschenbrener, Texas Department of Transportation]. A linear

variable differential transformer, with an accuracy of 0.01 mm, continuously measures

the rut depth in each slab and prints out the data every 20, 50,100, or 200 wheel passes.

Figure 2-9 is a picture of the Hamburg wheel-tracking device without the data-acquisition

machine.


Figure 2-9. The Hamburg wheel-tracking machine.









Various institutions use different criteria and testing procedures to evaluate mixes

in the Hamburg device. The City of Hamburg uses a maximum allowable rut depth of 4

mm at 19,200 wheel passes, whereas the Colorado Department of Transportation

(CDOT) recommends maximum allowable rut depths of 4 mm at 10,000 and 10 mm at

20,000 cycles [Aschenbrener].

In an evaluation study at the FHWA, researchers tested the rutting susceptibility of

certain mixtures with the full-scale ALF and compared the ALF results to the predicted

performance from the HWTD. The study noted that the tracking device could

discriminate mixtures with widely different binder grades, but failed to give consistent

results for mixtures with closer grade binders. Furthermore, the research showed that the

HWTD was unable to predict a decrease in rutting susceptibility for mixtures with altered

gradation (maximum nominal aggregate size), even though test results from the ALF

clearly showed less rutting [Stuart and Mogawer 1997, Stuart and Izzo 1995].

2.7.2 French Pavement Rutting Tester

The French Pavement Rutting Tester tests slabs of HMA to evaluate their

resistance to permanent deformation. The machine, shown in Figure 2-10, uses an

environmental chamber to keep the test temperature at 600C and loads the sample with a

smooth, reciprocating, pneumatic rubber tire inflated to 0.60 0.03 MPa (87 4 psi).

Similar to the Hamburg tester, the FPRT can test two slabs simultaneously measuring 500

mm (19.7 in) long, 180 mm (7 in) wide, and 50-100 mm (1.97-3.93 in) thick. Hydraulic

jacks push the slabs upward to apply the 5,000 50 N (1,124 11 lb) load [Corte and

Serfass 2000].





























Figure 2-10. The French pavement rutting tester.

The test sequence includes 1,000 cycles at 150-250C to simulate traffic

densification and to take initial slab thickness measurements. The slabs are then

conditioned for 12 hours at 60 +20C before the start of the test. The average rut depth in

each slab is measured manually at 30, 100, 300, 1,000, and 3,000 cycles when testing 50-

mm slabs, and at 300, 1,000, 3,000, 10,000, and 30,000 cycles when testing 100-mm

slabs. Finally, the average percent rut depth is calculated based on the initial thickness of

the slab [Corte and Serfass 2000].

The FHWA evaluation study mentioned in the section above [Stuart and Mogawer

1997] reported results for the FPRT. Similar to the HWTD, the French pavement rutting

tester accurately discriminated mixtures with widely different binder grades, but lacked

precision in identifying mixtures with distinctly different gradations.









2.8 Asphalt Pavement Analyzer

In 1985 the Georgia Department of Transportation (GDOT), in association with

the Georgia Institute of Technology (GT), initiated Research Project No. 8503 for the

development of a pass/fail laboratory test for the rutting resistance of HMA. The

prototype GLWT was a modification of the Benedict Slurry Seal tester, originally

designed to test slurry seals [Collins et al. 1995].

The initial version of the GLWT consisted of an aluminum wheel attached to a

reciprocating arm moved along a pressurized hose, creating the desirable contact

pressure. Constant temperature was maintained during testing by placing the LWT in an

airtight room, where an electric heater with a thermostat was used to heat the room to

95F. The test was performed with 75- or 100-psi hose pressure, and 50-, 75- or 100-lb

load. Rut measurements were taken at 40, 100, 400, 1000 and 4000 cycles [Lai 1986].

To promote the concept of using GLWT as a supplemental strength test to the

Superpave design procedure, the device was modified in 1992 to be able to evaluate

rutting potential of samples prepared by the Superpave gyratory compactor. The new

device had the ability to test six gyratory samples simultaneously in an environmentally

controlled chamber. Other modifications included operation control, adjustable hose

pressure (up to 120 psi), and load (up to 250 lbs) [Collins et al. 1996].

The APA is a further modification of the GLWT, first manufactured in 1996 by

Pavement Technology, Inc (PTI). Since it is a new generation of the GLWT, it follows

the same testing philosophy. Load is applied onto a pressurized linear hose by a

pneumatic loaded wheel and tracked back and forth over a testing sample to induce

rutting. The APA has the additional capability of testing for moisture susceptibility and

fatigue cracking while the specimens are submerged in water.









Extensive studies have been conducted to evaluate the ability of the APA to

distinguish between mixtures of known performance. Most of these studies tried to

establish a relation between rut depths obtained in the laboratory tests and the field

performance of the mixture.

The Florida DOT has performed a series of tests with the APA. Although the

device successfully ranked mixtures according to their rutting potential, some variability

from test to test and from location to location was found. It was also reported that

gyratory samples and beams rut at statistically different levels [Choubane 1998]. To

reduce variability, PTI installed new pressure regulators and reconfigured the air supply

tubing, but a subsequent study indicated that although variability was decreased the

middle position consistently yielded in higher rut depths than the left or right positions

[Sholar 1999].

A similar study compared test results from WesTrack [Epps et al. 1997] to rutting

predictions from three LWT devices. The APA ranked the mixtures according to their

WesTrack performance with 89% accuracy [Williams et al. 1999]. The National Center

for Asphalt Technology (NCAT) indicated that the APA was sensitive to mixtures with

different asphalt binder and varying gradation (ARZ, BRZ, and TRZ) [Kandhal et al.

1999].

The FHWA also conducted a study at Tumer-Fairbank Highway Center.

Comparison of LWTs test results to the ALF showed that the LWTs were able to

distinguish between mixtures that had good and poor performances when those were

prepared with the same aggregate gradation and different binder. When the aggregate

gradations were varied, none of the LWTs were able to separate the mixtures, even









though the ALF testing showed that there were significant differences in pavement

performance [Romero and Stuart 1998].

2.9 Summary

The discussion presented in this chapter indicates that no one SPT can reliably

predict HMA performance. The SST, although it is a mechanistic-developed test, is a

very expensive and complex machine to operate, making it unsuitable for a simple

performance test. From the three torture tests reviewed HWTD, FPRT, and APA the

Asphalt Pavement Analyzer seems to be the leading candidate.

Recent evaluation studies showed that the APA has the potential to accurately rank

mixtures according to their field performance. However, the APA proved to be unable to

capture the difference in performance for mixtures with altered gradation.

Based on this literature review, it is essential to identify the critical design

conditions) (mechanisms) that might lead to near-surface rutting. We must attempt to

replicate these conditions in a laboratory physical model (torture test). A physical model

that will employ this set of critical conditions might produce more reliable results for

mixture rutting performance.














CHAPTER 3
APA LOADING MECHANISM

3.1 Overview

Tire-pavement interface contact-stress studies indicated the importance of lateral

stresses in the development of critical stress-states near the surface of the pavement

[Drakos et al. 2000, Myers et al. 1999]. These studies have shown that radial tires induce

stresses that are more detrimental to pavements than bias-ply tires and that the difference

has been attributed mainly to tire structure.

The theory behind torture tests such as the APA is that a specimen is subjected to

an accelerated loading sequence in the laboratory and the end-result (rut depth) can be

correlated to the rutting performance of the mixture in the field. However, the ability of

the test to replicate field conditions in the laboratory determines the reliability of the

results. The hypothesis was that the APA loading mechanism was not capturing the

lateral stresses found under radial tires.

The idea that the APA hose could not generate these lateral stresses was based on

initial observations of the contact area between the sample and the hose which was

measured at approximately 8 mm. This limited contact area cannot reproduce the

Poison's effect found under each individual rib on the tire tread. This chapter deals with

the effort to develop an alternate loading device geared to capture some of the complex

stress distributions found in the field.









3.2 Limitations of Loaded Wheel Testers

Loaded wheel testers operate on the same basic principle: a test specimen of

mixture is subjected to repetitive loading by a traversing wheel, and the surface

depression in the sample is then measured and reported as a function of load cycles.

These types of torture tests are classified as empirical or performance-related tests

because they do not measure a fundamental property that can be used to explain and

identify the mechanisms resulting in surface distress.

The APA, like most of the LWTs, attempts to replicate field conditions in a

controlled laboratory environment. In this sense, good correlation between results from

the APA with field performance relies on how well (realistically) conditions have been

simulated in the lab. The following issues raise some considerations on the ability of the

APA to approximate field conditions:

* Loading scale effects. The loaded area under the pressurized hose is very small
(narrow) in proportion to the nominal maximum aggregate size [FHWA 1998, Lai
et al. 1990].

* Boundary conditions. In the APA the test specimens are resting on a metal plate
that limits deflections and increases confinement.

* Load application. Earlier work [De Beer et al. 1997, Myers et al. 1999] showed
that radial truck tires induce high lateral stresses that can cause tension on the
surface of the pavement [Drakos 2000]. It is believed that the pressurized hose of
the APA does not simulate the effects of the stiff treat of the radial tire, thus not
inducing any lateral stresses.

Lai and Lee (1990) evaluated the stiffness effects of the pressurized hose by testing

asphalt samples with a relatively stiff and a relatively soft hose. Figure 3-1 shows the

imprints of the contact area used for comparison purposes. As expected, under the same

load, the stiffer hose generated a more elongated and narrow contact area whereas the

softer hose produced a shorter and wider contact area.











A. Flexible Hose


r 4 -
-I .Cq


B. Stiffer Hose


Sz r
-. -t -r


Figure 3-1. Contact imprints of the rubber hoses with asphalt beam sample [after Lai and
Lee 1990].

Although the researchers believed that the stiffer hose would generate greater

rutting, a series of tests proved that the softer hose consistently gave slightly higher rut

depths. Unfortunately, there are no direct measurements of the contact stresses between

the hoses and the asphalt specimen surface. Nonetheless, this finding is of great

importance to the development of the new loading configuration because it demonstrates

the significance of stress distribution to the rutting behavior of samples in the APA.









3.3 New APA Loading Mechanism

The concept for a new APA loading mechanism is based on the observations and

conclusions from the tire-pavement interface stresses studies. Analyses performed with

the elastic layer analysis program BISAR and the finite element program ADINA

provided information on the pavement's response under modeled tires from measured

contact stresses. Myers et al. (1999), Drakos et al. (2001), and Birgisson et al. (2002)

have identified the lateral stresses induced by radial tires as the fundamental cause of

stress reversals (tension) and high magnitude shear stresses near the surface of the

pavement. These stress states cause a reduction in confinement near the pavement's

surface near the edge of the loaded area, which reduces the resistance to shear stress

within the mixture.

The hypothesis was that the stiff pressurized hose used by the APA to load the

specimen does not reproduce the lateral stresses found under radial truck tires. The

objective was to develop a new loading mechanism, modeled after a radial truck tire, to

replicate these stress conditions in the APA specimen.

3.3.1 Development of a Tire Model

The initial task was to develop a reasonable tire model that represents the structural

behavior and response of a typical radial truck tire tread. Earlier work by Roque et al.

(2000) showed that the radial tire loading behavior can be simulated with a combination

of steel and rubber. This step would enable us to estimate the right amount of steel and

rubber needed to built a device that captures the loading behavior of the radial tire. Figure

3-2 shows the structural characteristics of a radial truck tire with the radial plies and the

steel strands that run through the rubber, around the tire.

























Radial Plies


Figure 3-2. Structural characteristics of a radial tire.

Radial and bias-ply tires are totally different from a structural point of view and the

actual structural make-up of these tires is proprietary information not available to the

general public. However, Smithers Scientific Services, Inc. provided some basic

response data regarding the behavior of typical radial truck tires and their structural

characteristics that were used along with a basic knowledge of the structural behavior of

radial truck tires to develop a two-dimensional model of a radial truck tire tread [Myers

2000].

As previously discussed, the structural behavior of radial truck tires is governed by

a wall structure of very low stiffness and a very stiff tread structure resulting from the

steel strands used to reinforce the tread. The cross section of a typical radial tire is

illustrated in Figure 3-3 and shows a tread area that is 8.0 inches wide and 1.44 inches

high. The steel reinforcement was concentrated in an area that is 0.33 inches high,

approximately 0.93 inches above the outer surface of the tire.




















Figure 3-3. Schematic cross-section of a typical radial tire.
Following the guidelines from Roque et al. (2000), the model was constructed with
the MSC/Patran pre-processor software and it is illustrated in Figure 3-4. The steel
strands were modeled as a solid strip of steel, and the connection between the steel
strands and the rim was modeled as a pin connection at either end of the steel strip used
to represent the strands. Table 3-1 presents the modulus and Poisson's ratio values used
for the tire rubber and the steel strip. Finally, the ABAQUS finite element program was
used to run the elastic analysis and retrieve the deformation and stress distributions under
the tire model.


.40 28 0 0.05 _-
I P


8.00'


Figure 3-4. Finite element representation of the tread structure of a radial tire.


j


_I


L.00"
1 ,~Io









Table 3-1. Material properties used in the tire FEM.
Tire Part Material Elastic Modulus, E (psi) Poisson's Ratio, v
Reinforcing Beads Steel 2.90E+07 0.15
Tire Tread Rubber 1.16E+03 0.48
Tire Grooves Air 9.80E-06 0.49


3.3.2 Verification of the Tire Model

As mentioned in Section 2.3, Dr. Marion Pottinger of the Smithers Scientific

Services, Inc. successfully measured the tire-pavement contact stresses under truck tires.

Pottinger's device measured vertical, transverse, and longitudinal forces and

displacements under a moving tire using a series of 16 transducers on a steel plate. To

verify the accuracy of the model, Pottinger's contact stress measurements were compared

to stresses predicted under the FEM tire on a steel foundation.

The final thickness of the steel strip was determined by varying the thickness until

the predicted stress response of the modeled tire matched the measured response of the

real tire. Thus, the FEM tire matched the overall stiffness and stress-distribution

behavior of the tire tread. It was determined that a 0.1-inch-thick steel strip embedded in

the modeled tire tread resulted in the same structural response as the steel-strand

reinforcement in the actual tire.

Figures 3-5 and 3-6 vertical and transverse stress distribution respectively -

indicate that the tire model predicted both vertical and transverse contact stresses similar

to those measured under the real tire. Although there is some variation in magnitude, the

tire model accurately captured the patterns of both the vertical and transverse contact

stress distributions.

Figure 3-6 is particularly important because it demonstrates the model's ability to

capture the transverse contact stress reversals under the individual tire ribs. As stated










earlier, these transverse stresses were found to be detrimental to the top-down cracking

and near-surface rutting performance of HMA. The next step was to build an individual

rib replica that would serve as the load transfer mechanism in the APA.


Lateral Location, X (in)
0 2 4 6 8 10 12 14
0




-50
0 ---------------------------------










-100




t -150





-200
Measured
EM Predicted

-250


Figure 3-5. Measured and predicted vertical stress distribution at surface of steel bed.














80



60



40-



B 20










-40





FEM Predicted
-80

0 2 4 6 8 10 12 14

Lateral Location, X (in)
IX
-6 --w ,-------------------
-IXeaue
-IEMPedce


Figure 3-6. Measured and predicted transverse stress distribution at surface of steel bed.









3.3.3 Concept Loading Strip

In the previous section, analyses showed that the FEM tire was able to capture the

complex stress distribution measured under a radial tire. The idea for the APA loading

mechanism was to substitute the pressurized hose with a steel-rubber configuration based

on the tire finite element model. Figure 3-7 shows a schematic of the concept device,

called the loading strip, where a thin rectangular steel plate (14 gauge) is attached on top

of a medium-durometer (45-55) rubber. The solid steel wheel applies the load on the thin

steel plate that distributes the stresses on the sample through the rubber part of the device.

Side View 18" Front View




......-- -- -- -- .. .. .S.. .. . .k--- I
1.25"
Top View

140"


12"



Figure 3-7. Schematic of the loading strip.

Ideally, the loading strip stress-distribution behavior would represent that of a

single rib from the radial tire tread. The magnitude of the applied stresses was

anticipated to be significantly lower; however, the stress-distribution pattern was

expected to be similar. The steel plate would uniformly distribute the stresses to the

rubber and also increase the stiffness of the device, whereas the rubber member would

apply the vertical load and also create the Poisson's effect that induces lateral stresses as

found under radial tires.









3.4 Preliminary Contact Stress Calculations

The preliminary contact stress calculation was the first confirmation that the

loading characteristics of the two devices were different. It was important, however, that

the average vertical stress (cz avg) under the loading strip did not exceed that of the

pressurized hose. This ensured that the main difference in the contact stress distribution

was the presence of lateral stresses under the loading strip.

Initially, carbon paper was placed under the APA hose to measure the contact area,

but at static mode the imprint was not clear. Then, at a hardware demonstration at the

DOT, a technician measured the actual contact stresses using a pressure-sensitive mat.

For the loading strip, a finite element model was used to predict the stress distribution at

the rubber-specimen interface before it was physically constructed.

3.4.1 APA Pressurized Hose Stresses

Tekscan Inc. provided an initial estimate of the vertical stress distribution under the

pressurized hose when a technician visited the FDOT for a presentation of the company's

Pressure Measurement System (PMS). The Tekscan PMS is an extremely thin (-0.1

mm), flexible tactile force sensor that is capable of measuring pressures from 0-2 psi (0-

15 kPa) to 0-25 ksi (0-175 Mpa). Figure 3-8 shows the wide range of shapes, sizes, and

spatial resolutions (sensor spacing) of available sensors. Sensing locations within a

matrix can be as small as 0.0009 square inches (0.140 mm2); therefore, a one-square-

centimeter area can contain an array of 170 of these locations [Tekscan, Inc.]. The

Virtual System Architecture (VSA) integrates the sensors into a uniform whole and

displays the information on a computer screen.





























Figure 3-8. Tekscan pressure measurement system.

Tekscan's products function in both static and dynamic measurement environments,

allowing the development of load profiles and peak load attainment. During the

demonstration, the technician placed the sensor mat under each of the three pressurized

hoses in the APA, and recorded the vertical contact stresses while the machine was

operating (dynamic mode). This data-acquisition method provided a vertical stress

profile under the hose for the entire run (back and forth) throughout the specimen. Since

neither UF nor the FDOT owns a Tekscan measurement system, the access to the data

was limited.

The recorded vertical stresses revealed that their distribution is not even along the

specimen. Figure 3-9 illustrates the measured vertical stress distribution under the

pressurized hose (100 psi) with a wheel load of 100 lb. The color gradient indicates the

stress intensity ranging from light gray (low pressure, <10 psi) to black (high pressure,

>80 psi). The stress distribution in Figure 3-9 shows two dark-shaded peaks (high stress)








at the edges of the contact area. Initially this anomaly was attributed to the uneven

surface of the specimen and, more specifically, from large aggregate that might be

bridging the hose over some small gaps. Later on, based on the hose contact stresses

measured on a steel bed, it appeared that these peaks were a hose-structure phenomenon

that will be discussed in greater extent in the following sections.















Figure 3-9. Graphical interpretation of vertical stresses under the pressurized hose.

The software approximates the contact area based on the number of cells that report

pressure. In the case of the APA, there was some residual stress from the lowering arm

(the frame that holds the hoses in place) that was touching the mat. Analysis of the

spreadsheet provided by the Tekscan technician yielded to an estimated contact area of

1.54 in2 (993.5 mm2), which gives an average vertical stress of 64.9 psi (477.8 kPa).

3.4.3 Loading Strip FEM
Before fabrication, the concept device was modeled in finite elements to estimate

the stress distribution at the loading strip-specimen interface. Figure 3-10 shows a side

view and a top view of the three-dimensional model used for the stress analyses. An

estimate of the stress distribution under the loading strip was important to ensure that the
Figue 39. Gaphcal nteprettio of ertcal stress nde thepresu.'.'..se
The oftwre pproimats te cotac are basd o thenumer o cels.'.. '..r
pressure.~ ~~ ~ ~ ~~~~~~~~~~. .'..'.'..teAAteewa oe eiua tes rmth oern r
(the. ..."e ..thlstehssi pae htwstuhn temt nlsso
spreadsheet ~:.i:::1 :::";.....esantcniinyele o netiae cnat rao
1.54 in2 (93.5 mm2 '' 'i;? iviei:::i an 'vrg "etclsrs f649pi(7. ~
od n :ti : i::::.".









average vertical stress (cz avg) under the loading strip would not exceed the Cz avg under

the pressurized hose.




















Figure 3-10. Finite element model of the loading strip.

The finite element model was constructed with the MSC/Patran pre-processor

software and the elastic analyses run with the ABAQUS engine. The model consisted of

20-node 'brick' elements, and the material properties used for the steel and rubber parts

of the loading strip are the same as in the tire model (Table 3-1).

Because of the rectangular shape of the rubber on the loading strip, it was easy to

assume that the width of the contact area would be constant at 1.25 inches. Thus, the

only requirement to approximate the contact area was to estimate the length of the

pressure patch. It was clear that the contact area under the loading strip would greatly

exceed the initial contact area of the APA hose. The initial contact area under the hose

differs from the final because, as the material deforms, the hose 'sinks' into the material,

increasing the contact area.









Based on the assumption that the contact area would be greater under the loading

strip, the FEM model was analyzed at a higher load level (150 lb) to calculate the extent

of the pressure patch and the average vertical stress. The FEM predicted the pressure-

patch length to 4.5 inches; thus the resultant area was 5.6 in2 and the average stress

approximately 26 psi.

3.5 Measured Contact Stresses in the APA

At the beginning of this chapter the hypothesis was that the pressurized hose of the

APA does not capture the essential lateral stress distribution found under radial tires. A

new concept loading device (loading strip) was designed and tested with the help of

numerical modeling that would simulate real tire stress distribution. In order to verify the

above hypothesis, both loading devices pressurized hose and the loading strip were

sent to the Smithers Scientific Services plant in Ravenna, Ohio, to measure the actual

contact stresses at the loading device-specimen interface.

3.5.1 Measurement System

As mentioned in Section 2.3, Smithers Scientific Services, Inc. developed the Flat

Surface Tire Dynamics Machine (FSTDM) to measure contact stresses at the tire-

pavement interface. The device measures vertical, transverse, and longitudinal forces and

displacements under a moving tire by using a series of 16 transducers.

Dr. Pottinger fabricated custom end-restraints and a loading foot that allowed load

control to within +1 lb, to accommodate the pressurized hose and the loading strip on the

FSTDM. Figure 3-11 shows a picture of the FSTDM with the loading foot during

calibration. The 500-lb cell was calibrated using a pedal force transducer, as seen in the

lower right-hand corner of the picture.

































Figure 3-11. Contact stress measuring apparatus setup and calibration.

The loading strip was tested with three different loads 110-, 130-, and 150-lb -

whereas the pressurized hose was tested at two load levels 100-, and 120-lb. The

loading foot with the steel wheel remained stationary, while the bed with the loading

device (hose/loading strip) moved in the longitudinal direction. The movement of the

bed forced the steel wheel over the loading device and the transducers measured the

displacements and stresses at the contact interface.

3.5.2 Hose-Specimen Interface Stresses

Results from the APA pressurized hose contact stresses verified the initial

hypothesis that the contact area under the hose is too narrow to produce any significant

lateral stresses. Figure 3-12 shows the pressurized hose, which is attached to the moving

bed, and the concave steel wheel loading the hose directly above the transducers.

































Figure 3-12. Close-up picture of the pressurized hose test.

In his report, Dr Pottinger stated that the narrow (8mm) contact area was not wide

enough to record any lateral stress on the transducers. Figure 3-13 illustrates the vertical

stress distribution under the hose. Similar to the Tekscan results (Figure 3-9), the

measured vertical stresses show two humps at each side of where the steel wheel loads

the hose, caused by the semi-rigid structure of the hose.












-140.0



-120.0



-100.0



-80.0



-60.0


-40.0



-20.0



0.0


0 1 2 3 4 5 6 7 8 9

Lateral Location, X (in)


Figure 3-13. Vertical stress distribution under the pressurized hose.









3.5.3 Loading Strip-Specimen Interface Stresses

Smithers Scientific Services measured the contact stresses under the loading strip

for three load levels 110-, 130-, and 150-lb. Figure 3-14 is a picture of the steel wheel

applying pressure on the loading strip which is fixed on the moving bed. Dr. Pottinger

noticed that in the case of the loading strip, the solid wheel had to be centered over the

loading strip to avoid asymmetric stress distribution. Contrary to the solid wheel, the

concave wheel acts as a channel that continuously aligns the rubber hose with the

traversing movement of the loading arm.
























Figure 3-14. Close-up picture of the loading strip test.

Figure 3-15 illustrates the vertical stress distribution under the loading strip for the

three load levels. Unlike the pressurized hose results, the vertical stress distribution

under the loading strip resembles that of an elastic material with the stress peaking in the

middle of the normal distribution. As expected, the magnitude of the vertical stresses is









much lower under the loading strip due to the increase of the contact area. The highest

measured vertical stress under the loading strip for the high load (150 lb) was 35 psi,

whereas the pressurized hose recorded 130 psi vertical stress for the 100-lb load.

Figure 3-16 shows that the transverse stress distribution under the loading strip

accurately captures the Poisson's effect found under individual tire ribs. The Poisson's

effect states that, unless restrained, most materials expand laterally when loaded

vertically. When individual ribs under a tire are loaded they attempt to expand laterally,

and the surface of the pavement tries to restrain the expansion thus generating transverse

stresses. Similar to the tire ribs but lower in magnitude, the loading strip induces lateral

stresses that change sign (direction) at opposite sides of the loading strip.

3.6 Summary

The discussion in this chapter focused on the contact-stress distribution between the

specimen and the pressurized hose in the APA and the effort to develop an alternative

loading mechanism to capture the complexity of the actual tire-pavement interface

stresses. Actual contact-stress measurements verified that the limited initial contact area

under the pressurized hose could not induce lateral stresses on the surface of the HMA

sample. However, results from the proposed loading device (loading strip) showed that

the distribution of the lateral contact stresses closely resembles that found under

individual tire ribs.













-40.0


-35.0


-30.0


-25.0


-20.0


-15.0


-10.0


-5.0


0.0


0 1 2 3 4 5 6 7 8 9

Lateral Location, X (in)


Figure 3-15. Vertical stress distribution under the loading strip.













15.0



10.0



5.0



0.0



-5.0



-10.0



-15.0


0 0.25 0.5 0.75 1


Transverse Location, Y (in)


Figure 3-16. Lateral stress distribution under the loading strip.


1.25














CHAPTER 4
STRESS ANALYSES

4.1 Overview

Recent experimental studies revealed that tire contact stresses are distributed in a

highly non-uniform manner and differ significantly for various tire types [De Beer et al.

1997, Marshek et al. 1986, and Myers et al. 1999]. These stresses include not only

vertical normal stresses, but also transverse and longitudinal surface shear stresses. One

proposed hypothesis on the mechanism behind instability rutting is that radial tires, with

their complex non-uniform loading, may be inflicting significant stress states in the HMA

that are not predicted with traditional uniform vertical loading patterns [Drakos 2000].

Elastic layer and finite element analyses of asphalt pavements for three load cases -

radial tire load, bias-ply tire load, and uniformly distributed vertical load showed that

radial-tire loads induce more severe stress states near the surface of the pavement.

The measured contact stresses under the two loading mechanisms hose and

loading strip also revealed some expected differences in the contact-stress distribution.

Lateral stresses under the loading strip resembled the distribution found under a single rib

on the radial tire tread. Finite element modeling of the APA test showed that the loading

strip, similar to the radial tires, induced some tension near the surface of the specimen.

These stress states tend to induce shear stresses that shove the material away from the

loaded area similar to instability rutting.









4.2 Pavement Stress Analyses

To evaluate the effect of the contact stresses found under various tires, the

measured tire-pavement interface stresses were applied as a load on a pavement structure

and the resulting stress states were analyzed. In the past, finite elements have not been

used to model three-dimensional tire loads due to the complexity of modeling a radial-tire

load in three dimensions. Instead, the elastic multi-layer analysis program BISAR [De

Jong et al. 1973] was used to predict the pavement responses.

Typical pavement structures consist of a thin asphalt concrete layer over a base

course, which rests on the semi-infinite subgrade. To produce an accurate model of the

non-uniform load and provide adequate boundary conditions requires a large number of

elements, and the associated amount of memory is not available on current PCs. The

University of Florida recently purchased a Silicon Graphics Interface (SGI) multi-

processor computer with extensive memory and faster computing time than the average

PC that made the three-dimensional finite element analysis of HMA pavements possible.

4.2.1 Multi-Layer Elastic Stress Analyses

The initial approach to model three-dimensional tire contact stress was to

approximate the complex loading conditions with uniform circular loads. BISAR can

apply circular uniform loads with a single vertical stress and one stress in the lateral

direction of a specific angle. A series of small uniform circular loads of varying vertical

(cz) and lateral (cx & cy) stresses would represent the non-uniform tire stresses.

Figures 4-1 and 4-2 illustrate the load configuration for the bias-ply and radial tires

respectively. It took 209 load circles to simulate the lateral and non-uniform vertical

stress distribution under the bias-ply tire, and 145 load circles to represent the radial tire.

The magnitude, orientation, and location of the vertical and lateral stresses of the








individual circular loads used to represent the bias-ply and radial tire contact stresses are
listed elsewhere [Drakos 2000].


Y (in)


Rib 1


ee
ee
oe



Ge


Ge
r> r
11 L


Rib 2

Q0
00
00
S
00
00

00
00
00
00
oo
00
00
00@


(0,) Origin


Rib 3

00
0
00

00
00
00
00
Go




00
00
00
@








s0


X (in)
0
0

0

X@


Rib 4


Q^


ote
4->;
~30
)S



N

^0


Rib 5



ee00
00

O,
4-I 11





(7e



(7.751I)


Figure 4-1. Load configuration used in BISAR to represent measured stresses under bias-
ply truck tire [Drakos 2000].
For the bias-ply tire, Figure 4-1 illustrates the existence of a significant transverse
stress component near the edge of the load, which tends to pull the pavement in towards
the center of the tire. This development can be explained by the overwhelming
pneumatic effect that is induced by bias-ply tires as explained in section 2.3. On the
other hand, Figure 4-2 shows that the radial tire creates a 'pushing outward' effect under
each individual rib. This trend is attributed to the dominance of the Poisson's effect in
radial tires.
















Y linl


10,8.341 -,


Rib 1






e





0
i fl NA 8C
(slZ_^ 1*.0


(0)) Origin


Rib2
0
00
00

00
00
80


00

80
00
00
0


Rib 3









(i


Go


Rib 4
00
00
00
00
00
00

00
00

00
00
00
00
(S0


Figure 4-2. Load configuration used in BISAR to represent measured stresses under
radial truck tire [Drakos 2000].
4.2.2 BISAR Results
Figure 4-3 illustrates the maximum shear stress distribution for the modeled radial

and bias-ply truck tires. The responses were predicted along the surface of the pavement

at 0.2-inch increments. High shear stress values were calculated under the left-most rib

of the modeled radial load, whereas much lower values developed under the bias-ply tire

load.

The high shear stresses predicted under the modeled radial tire hinted that shear

planes might be developing under the load, similar to a bearing-capacity type of failure.

To investigate this hypothesis the direction of the predicted maximum shear stress, at

each output point under the tire load, was plotted as a vector.


Rib 5



0)
e /-








GI



(8.66,)











500

450

400 Bias-Ply Tire Load
400
t Radial Tire Load

S350

S300

250

200::




100 '

50
50



0 1 2 3 4 5 6 7 8 9 10 11 12 13
Lateral Location, X (in)



Figure 4-3. Maximum shear stress distribution.

Angle a was defined as the smallest angle formed between the maximum shear

stress plane and the horizontal. Figure 4-4 demonstrates the sign convention used by the

analysis program and the calculation of angle a, which was used to plot the direction of

the shear stress plane.

Figure 4-5 shows the two equal and opposite maximum shear stresses that act on a

particular element. In this case, the smallest angle is formed between the negative

maximum shear stress and the horizontal plane. Thus, the plotted directional arrow

(vector) would represent the direction of the negative maximum shear stress.











n= (-)



S(-)
W< (+)



When :Txz< 0 & iYH < if7


BISAR Sign Convention

Compression = Negative
Tension = Positive
Clockwise M = Negative
Moment @ Point Outside
of the Element




+ s
Positive
Moment


0 if a > cy
k= -1 if ar < cy, and i< 0
1 if av < c7 and T> 0


Figure 4-4. BISAR sign convention and maximum shear stress angle a.


Tire Load
_J


*


Element1 mx Horizontal
Output Point -.-. ------ Plane
P T aM n


Figure 4-5. Schematic of the maximum shear stress direction representation.


The Load


Plotted Shear
Direction









Figure 4-6 shows the magnitude and direction of the maximum shear stress

distribution along a vertical section for the modeled radial tire load. The arrows indicate

the direction of the maximum shear stress closer to the horizontal, and the contour plot

(shaded area) in the background specifies the magnitude of the shear stress. The

direction of the shear stresses under the right-most rib of the radial tire indicates the

formation of shear planes that tend to 'shove' the material away from the tire. At the

same location, the contour plot of the predicted maximum shear stress magnitude

indicates that shear stresses are at their highest value.

Figure 4-7 illustrates the magnitude and direction of the maximum shear stress

distribution under the modeled bias-ply tire load. Unlike the radial tire load responses,

the maximum shear stress direction for the bias-ply load appears to be 'pulling the

pavement inwards.' The orientation of the directional arrows under the bias-ply tire load

is pointed inwards, towards the tire. When compared with the radial tire load results, the

magnitude of the predicted maximum shear stress for the bias-ply tire load was

significantly lower.

In an effort to isolate the effect of tire-induced lateral stresses, the horizontal stress

component in the BISAR input files was set to zero so that only vertical stress was

applied. The results of the analyses performed without the lateral tire contact stresses

revealed that, for the radial tire load, the direction of the maximum shear stress was

reversed. In contrast to that, results for the bias-ply tire load did not give any indication

of stress reversal.






58

Radial Tire Load
Y LateralStxess
VefticaiSess I Ini.IIIn
-------------
| :I

t


0


-0.2


-0.4


-0.6
N
-0.8

-1


-1.2


-1.4


Rib #


t=244"


0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Lateral Location, X (in)


Figure 4-6. Magnitude and direction of maximum shear stresses under radial tire load.


1z1~

OFi
'iiiiiiiiiiiiiiii, .... ,,, ,,, ,' iiiiiiii !ii
i i iiiiiii i
.. L"""""""""""__. _

[] U
_q~aKb, 7_
YIJJ 7 raIr


A~ 77 yu~"*rrr
r-J J~t J







59



Bias-ply Tire Load
SLateralStress f Loi o I l II I

F VerticalStress I LaI









Rib#l t=l" 1























-1.2 -, -- 4- 4






0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
N I












-1.4 ,. ,- -, 7- ,









Analyses performed with the elastic layer analysis program BISAR provided

information on the pavement's response under modeled tires from measured contact

stresses. Even though the magnitude of the estimated stresses is sometimes exaggerated

due to a discontinuity problem at the edge of the circular load [Jacobs 1995], the overall

effect of the non-uniform loading showed that the near-surface stress distribution was

highly dependent on tire structure. Lateral stresses induced by radial tires seem to cause

a reduction in confinement near the pavement's surface, which reduces the mixture's

resistance to shear stress. The combination of high magnitude and outward direction of

the maximum shear stresses is believed to create the critical stress conditions contributing

to the permanent deformation of the pavement.

4.2.3 Finite Element Stress Analyses

The three-dimensional finite element model was constructed and analyzed using the

commercial finite element code ADINA [Bathe 2001]. The finite element model

consisted of 30,204 nine-node elements of varying dimensions. Elements under the

loaded area were refined to 0.3 by 0.4 inches in the horizontal plane and 0.2 inches in the

vertical plane. To overcome the limitations associated with building a traditional mesh,

contact surfaces were introduced to 'bond' fine-graded mesh onto a coarse-graded mesh.

This allowed for the introduction of coarse meshes at distances farther away from the

loaded area where the change in stress was more gradual. The use of contact surfaces

was further justified based on the primary area of interest, the near-surface area under the

loaded tire, thus negating any possible negative numerical effects of 'far-away' contact

surfaces.

Figures 4-8 and 4-9 show the finite element model with the dimensions used in the

analysis. The mesh extended 60 inches in each horizontal direction and 72 inches in the









vertical direction. Although these dimensions may seem inadequate for a finite element

model, an initial assessment showed that stresses near the tire footprint (area of interest)

were not affected by the extent of the boundaries.


Figure 4-8. Three-dimensional finite element mesh used in the pavement response
analysis.

The boundary conditions for the four sides (faces) of the FEM were fixed in the

horizontal (X and Y) direction and free in the vertical (Z) direction, whereas the bottom

of the FEM was fixed in all directions. The model consisted of 260,455 nodes, giving it

over one million degrees of freedom. The memory required for analysis exceeded 1700

megabytes and took over four hours to complete on a single processor.












-F-


Vf t7


I/I


'Ii


I I


0 10 20


30
Inches


40 50 60


Figure 4-9. Plan view of the contact area of the three-dimensional mesh used in the
pavement response analysis.

The structure used in the analysis was a typical three-layer pavement asphalt

concrete, base, and subgrade with thicknesses 8, 12, and 52 inches respectively. Each

layer was assumed to be isotropic, homogenous, and linear elastic. Table 4-1 shows a

summary of the material properties used in the structure. The low asphalt concrete

modulus corresponds to a warm summer day for a new pavement the most critical time

for the onset of instability rutting while base and subgrade modulus values represent

typical materials used in Florida. The Poisson's ratio was selected to ensure minimal

volumetric changes, as would be expected from a single moving tire load.


t


CI


""''''"'


' """"""""""""""""""" '


\


I/ / // /^
I / /// /T.
/ // / X-< /^,
/ /l / ",^
III /^ / /
/ll / / ^


TTt~


\


~-rrr I


/


B


B









Table 4-1. Material properties and layer thicknesses of FEM pavement structure.
Layer Modulus (psi) Poisson's ratio Thickness (inches)

Asphalt concrete 100,000 0.45 8

Base 40,000 0.45 12

Foundation 15,000 0.45 52

4.2.3-1 Loading the FEM

Dr. Pottinger measured the tire-pavement contact stresses in a fine grid 0.1 by

0.15 inches making it almost impossible to load the mesh directly with the recorded

stresses. An approximation method was used to convert and redistribute the measured

contact stresses to nodal forces. The appropriate force for each element was determined

by converting each uniform stress to an equivalent concentrated force. The forces were

then converted to nodal forces with the help of shape functions and applied to the

respective node [Cook 1995].

Figure 4-10 illustrates the shape-function procedure to redistribute an element force

to nodal forces. Parameters and f define the position of the element load relative to the

element's center, and their value ranges from zero to one. The shape functions are then

calculated for each node based on the and f values. Finally, the nodal forces (Fl to F8)

are the product of the element force (P) with the respective nodal shape function (N1 to

N8).

Figure 4-11 is a cross section view of the surface elements with the resulting nodal

forces for the modeled radial-tire load. The nodal loads are non-uniform and vary in

magnitude and direction to simulate the contact stresses found under a radial tire.










18


3
,


P

5 i 7


-^-------------



1 6 2

O = Element Node
P = Element Force

= (Xele center Xload)

1 = (Y"ele center Yload)


Figure 4-10. Definition of the shape functions.


Ifll\\1 I1111M\


Shape Functions

1 1 1
N= (1- )(1-)- (N5) (N8)
4 2 2
1 1 1
N2= (1 +)(1- 7)- (N5)- (N6)
4 2 2
1 1 1
N3 = (1 + )(1 + ) -(N6) (N7)
4 2 2
1 1 1
N4 = (1-)(1+ 7)- (N7)- (N8)
4 2 2
1
N5 = (1- 2)(1-7)
2
N6 = 1(1+)(1- 72)
2
1
N7 = (1-2)(1+7)
2
1
N8 = )(1- 2)
2


111)ii \ f111n U


inches


Figure 4-11. Cross-section view of surface elements with nodal forces for the radial-tire
load.


I ITJT _IT ITI-I ITI-I Tl`Tj











4.2.3-2 FEM Results

Figure 4-12 shows the magnitude and direction of the maximum shear stress

distribution along a vertical section under the modeled radial-tire load. The distribution

of shear stresses is similar to those predicted using BISAR, except the magnitudes of the

maximum shear stresses are lower, ranging from 50 to 60 psi compared to values in


excess of 100 psi for the BISAR predictions. The difference in magnitude can be

attributed to the approximation method used to convert the measured point stresses to

circular loads, and to the BISAR overestimation problems mentioned above. However,

the key finding is that the overall trend the formation of the shear planes remains the

same between the two different modeling techniques.




Rib #1 Rib #2 Rib #3 Rib #4 Rib 15


N 10
C 15
S 20
25


'( r.

A`


. .,~ 1:. .


. ,

,. .,


I:


50
40
30
20
- 10
- 0


0 3 6 9 12
Lateral Location, X (in)



Figure 4-12. Maximum shear stress magnitude (psi) and direction under the modeled
radial-tire load.

The plotted vectors indicate the formation of shear planes under the loaded area.


Also, it can be seen from the direction of the shear stresses under the first rib of the radial

tire that shear planes formed tend to shove the material away from the tire, and the

contour plot of the predicted maximum shear stress magnitude indicates that the shear

stresses form planes that match the directional arrows.









4.3 APA Stress Analyses

The measured contact stresses between the HMA specimen and the two APA

loading mechanisms pressurized hose and the loading strip were modeled with finite

elements to evaluate the effects of the different loading conditions. The primary

objective was to examine whether the loading strip could induce similar stress states in

the modeled HMA specimen, as the radial tire induced in the modeled pavement.

The three-dimensional finite element model was constructed using the

MSC/PATRAN pre-processor software to build the model geometry and to define the

mesh and the HKS/ABAQUS software for the actual analysis [Hibbitt, Karlsson and

Sorensen, Inc. 1997]. To build the mesh around the curved surfaces we used an

automatic mesh-generating option in PATRAN called paver. The paver is best suited for

trimmed surfaces, such as surfaces with holes or cutouts. In this case, the paver meshing

algorithm generated quadrilateral elements perpendicular to the curved surfaces and

transitional elements to connect to the free edges. At first, the mesh was generated in

two-dimensional space and then the 'sweep' action was used to extrude the elements into

three-dimensional 'brick' elements.

Figure 4-13 shows the initial two-dimensional model for the APA mold and the

HMA specimen. For practical purposes, only one of the two cylindrical-sample slots was

used in the model. Furthermore, the model was separated into two main parts the

plastic mold (E = 400000 psi, v = 0.4) and the asphalt concrete sample (E = 100000 psi, v

= 0.4). Figure 4-14 shows the three-dimensional geometry and mesh definition of the

two solids.
















x






* L



' .I J


'-I-


~-4-












I>'r'
___ '


Figure 4-13. Top view of the finite element model for the APA mold and specimen.































Figure 4-14. Three-dimensional finite element model for the APA mold and specimen.









As mentioned in Section 3.5, Dr. Pottinger measured the contact stresses under the

two loading devices the original pressurized hose and the new loading strip. The

objective in this section was to load the model with the measured contact stresses and

analyze the predicted stress states in the specimen. Once again, the measured stresses

were converted to nodal forces with the help of shape functions and applied to the

respective node (Section 4.2.3-1).

Figures 4-15 and 4-16 show the predicted magnitude and direction of the maximum

shear stress (Tmax) distribution, along a vertical section, for the loading strip and

pressurized hose loading conditions respectively. The range of the Tmax magnitude under

the loading strip (3-13 psi) is much lower than that predicted under the pressurized hose

(10-80 psi). This magnitude difference can be attributed to the higher vertical stresses

measured under the pressurized hose (Section 3.5.2).

The important finding of this analysis was the pattern of the Tmax distribution

throughout the modeled specimen. Unlike the distribution under the pressurized hose, the

modeled loading strip showed that the Tmax magnitude peaks near the surface of the

specimen, under the loaded area. Furthermore, the magnitude contour plots for the

loading strip condition indicate the existence of shear planes under the load, whereas the

same is not true for the modeled pressurized hose load. The combination of the Tmax

magnitude distribution with the Tmax direction under the loading strip resembles the

pattern found under individual ribs for the modeled radial-tire load (Figure 4-12).











0



-0.1



E -0.2
N


-0.3



-0.4



-0.5
2 2.5 3 3.5 4 4.5
Lateral Location, X (in)



Figure 4-15. Maximum shear stress magnitude (psi) and direction under the modeled
loading strip load.

0



-0.1



-0.2
N


-0.3



-0.4



-0.5
2 2.5 3 3.5 4 4.5
Lateral Location, X (in)



Figure 4-16. Maximum shear stress magnitude (psi) and direction under the modeled
pressurized hose load.









4.4 Summary

Analyses performed with the elastic layer analysis program BISAR and the finite

element program ADINA provided information on the pavement's response under

modeled tires from measured contact stresses. The analyses provided evidence that the

radial truck tires induce higher shear stresses near the surface of the pavement than the

bias-ply tire. The direction of the maximum shear stresses under the modeled radial tire

load appeared to shove the material away from the load, something that was not observed

under the modeled bias-ply tire.

The measured contact stresses under the two APA loading devices were used to

load the APA finite element model. The analysis showed that the pressurized hose load

produced higher magnitude stresses. However, the loading strip replicated (to some

extent) the critical stress states identified under the radial-tire load.














CHAPTER 5
MATERIALS AND TESTING METHODS

5.1 Overview

Two of the three mixtures selected for this study have been placed in Florida in

1998, (see Table 5-1), and the FDOT has been monitoring their field performance ever

since. The Job Mix Formula (JMF) of the original FDOT mixtures had a Reclaimed

Asphalt Pavement (RAP) component of 15%-20% that formed part of the aggregate

constituent. However, the RAP material was no longer available at the time of this

research so the percentages of the other aggregates were adjusted to maintain the same

gradation for each mix.

Table 5-1. Field location of selected mixtures.
FDOT UF Project Placement
Route County
Project No. No. Date

2134391 1 Jan-98 I-10 Madison

2325941 7 Sep-98 Turnpike Palm Beach



The mixtures were designed and produced according to the Superpave Volumetric

Mix Design procedure and the samples compacted with the Pine Gyratory Compactor to

150-mm diameter by 115-mm height gyratory specimens. Finally, the samples were

tested with the original and modified Asphalt Pavement Analyzer test procedures to

evaluate their rutting performance.










5.2 Materials

Table 5-2 lists the original aggregate types and producers used in the FDOT

monitoring projects. Project 1 is a 9.5-mm nominal maximum-size coarse-graded

mixture, Project 7 is a 12.5-mm nominal maximum-size fine-graded mixture, and the

HVS mixture is a 12.5-mm nominal maximum-size coarse-graded mixture. The last

column on the right shows the JMF-blend percentage for each aggregate type. As stated

earlier, the Superpave project mixtures Project 1 & Project 7 included RAP (milled

material) which was not available at the time of this study.

Table 5-2. Aggregate types and sources for the selected FDOT mixtures.
Project FDOT JMF
Project Mix No. Material FDOT Pit No. Producer F
No. Code %
Milled material 20
# 89 Stone 51 GA 185 Martin Marrietta 45
1 97051A
W-10 Screenings 20 GA 185 Martin Marrietta 25
M-10 Screenings 21 GA 185 Martin Marrietta 10
Milled material -20
7 980139A S1A Stone 41 87-339 White Rock Quarries 20
S1B Stone 51 87-339 White Rock Quarries 10
Asphalt Screenings 20 87-339 White Rock Quarries 50
S1A Stone 41 87-089 Rinker 13.0
HVS S1B Stone 51 87-089 Rinker 55.0
Asphalt Screenings 20 29-361 Anderson 32.0



Table 5-3 shows the new aggregate blends that were obtained from the same

sources and adjusted without the RAP to reproduce the original JMF. After careful

selection, the gradations of the new blends resembled the original JMF for each project -

Project 1 & Project 7. Figures 5-1 to 5-3 illustrate the gradation of the laboratory blend

in comparison with the original (field) JMF on a 0.45-power chart.










Table 5-3. Aggregate sources and modified blends for the laboratory mixtures.
Project FDOT JMF
Project Mix No. Material FDOT Pit No. Producer F
No. Code %
# 89 Stone 51 GA 185 Martin Marrietta 50.0
1 97051A W-10 Screenings 20 GA 185 Martin Marrietta 18.5
M-10 Screenings 21 GA 185 Martin Marrietta 31.5
S1A Stone 41 87-339 White Rock Quarries 24.5
7 980139A S1B Stone 51 87-339 White Rock Quarries 12.5
Asphalt Screenings 20 87-339 White Rock Quarries 63.0
S1A Stone 41 87-089 Rinker 13.0
HVS S1B Stone 51 87-089 Rinker 55.0
Asphalt Screenings 20 29-361 Anderson 32.0



The asphalt binder used for the study is an AC-30 binder with a PG-67-22 grading.

It is distributed by Coastal Petroleum Company in Jacksonville (subsidiary of Coastal


Caribbean Oils & Minerals, Ltd) and is a common binder in Florida.






















































0 0 0 0 D 0
00 tl em : : \C
Sieve Size


Figure 5-1. Gradation chart for JMF and laboratory blend for Project 1 (9.5mm maximum nominal size).






















































0 0 0 0 D 0
00 tl em : : \C
Sieve Size


Figure 5-2. Gradation chart for JMF and laboratory blend for Project 7 (12.5mm maximum nominal size).























































0 0 0
00 %


Sieve Size


Figure 5-3. Gradation chart for laboratory blend for the HVS coarse-graded mixture (12.5mm maximum nominal size).









5.3 Mixture Preparation

All test specimens for the evaluation tests were prepared with the Superpave

Volumetric Mix Design procedure. The design procedure uses volumetric properties -

air voids (AV), voids in mineral aggregate (VMA), and voids filled with asphalt (VFA) -

as the primary criteria to select the optimum asphalt content (% AC) for the specified

aggregate blends. The Superpave criteria vary with the specified nominal maximum

aggregate size for the JMF and the expected traffic level.

The two Superpave monitoring project mixtures used in this study Projects 1 & 7

- have been used in various research projects at UF [Asiamah 2002, Darku 2003] and

there is ample information about their volumetric properties. For this reason, the values

for Rice Specific Gravity, optimum asphalt content, and Ndes were not recalculated.

Further information about the volumetric properties of all three mixtures is included in

Appendix A. Outlined below is the procedure followed to produce the 75-mm-thick

cylindrical specimens used in the APA.

5.3.1 Aggregate Preparation and Batching

* The virgin material was dried in the oven (230 3000F) for at least 12 hours and
then allowed to cool down to room temperature.

* The material was sieved and separated into its individual particle sizes 34", 1/2",
/", #4, #8, #16, #30, #50, #100, #200, & pan.

* The aggregates were batched in 4500g samples (Pine Gyratory Compactor) in
accordance with the JMF. Tables showing batch weights for the aggregates are
given in Appendix A.

5.3.2 Mixing

* The aggregates, asphalt binder and the mixing equipment mixing bucket and
spatulas were placed in the oven (3000F) for about two hours.

* Aggregate blend and asphalt binder were mixed in the bucket for about 5 minutes
or until the aggregates were thoroughly coated with the asphalt.










5.3.4 Short-Term Oven Aging (STOA) and Compaction

* Before compaction, the samples were spread in a pan and placed in the oven
(275F) for STOA. While in the oven, the mix was stirred after one hour to
achieve uniform aging.

* After the STOA, the 4500g samples were compacted in the Pine Gyratory
Compactor (Figure 5-4) to Ndes. Rice specific gravity information is given in
Appendix A.

* Compacted specimens were allowed to cool for a minimum of 24 hours at room
temperature and then were sawed down to 75mm thickness.

* The Bulk Specific Gravity (Gmb) was then determined in accordance with ASTM
D1189 and D2726 for each specimen.

* The percent air voids for each specimen was computed from the Gmm and the
Gmb.

















U 1r
m iP
I


Figure 5-4. Pine Gyratory Compactor.


1
1









5.4 Asphalt Pavement Analyzer Procedure

As we have seen, the APA tests the rutting susceptibility or rutting resistance of

HMA. The original configuration of the machine creates the desirable contact pressure

with a concave aluminum wheel attached to a reciprocating arm moved along a

pressurized hose, whereas a steel wheel loads the loading strip for the modified

configuration. The device is able to test either a 75mm x 125mm x 300mm beam

specimen, or a 150mm diameter by 75mm thick cylindrical specimen. For this study,

only cylindrical specimens were evaluated.

The APA test procedure was slightly modified to incorporate a new way of

recording and analyzing the test results. Instead of using the roller dial gauge to measure

a single (the lowest) point on the specimen, the new method uses a contour gauge that

captures the entire surface profile of the sample. In case the material fails, the new

method (contour gauge) enables the user to monitor the rate and the mode

(consolidation/instability) at which the material is failing.

Figure 5-5 shows the original measuring plate with the 3-inch slits where the dial

gauge is dragged to locate the lowest (highest dial reading) spot on the specimen. This

method is limited to a single measurement that denotes the maximum amount of

permanent deformation on the specimen and there is no way to distinguish if that

deformation is due to consolidation or plastic flow.

Figure 5-6 illustrates the aluminum plate used with the new method of measuring

the specimen deformation. The slits on the new plate are 5 inches wide at the two

extreme locations (marked with an E on Figure 5-6) and 512 inches at the middle location

(marked with an M on Figure 5-6). The contour gauge adjusts to the surface profile of

the specimen, which is (the surface profile) recorded and digitized for further analysis.




























Figure 5-5. Original APA measuring plate.


Figure 5-6. New measuring plate with elongated slits.









5.4.1 Surface Profile Measurement

As mentioned above, the new measurement system (contour gauge) was

implemented to record the entire surface profile of the specimen. Figure 5-7 shows the

contour gauge recording the surface profile at the middle location of the sample. The

rods are pushed downwards until they come in contact with the specimen, forcing the

contour gauge to assume the shape of the specimen's surface.


Figure 5-7. Contour gauge measuring the surface profile of the specimen.









Figure 5-8 shows the contour gauge on the custom fabricated holder where the

shape of the recorded surface profile, from each location on the measuring plate, is traced

on a card. The card slides beneath the contour-gauge rods, whereas two PVC strips

restrain the card from any lateral movement.





















Figure 5-8. Recording the deformed shape of the contour gauge.

5.4.2 APA Hose Testing Procedure

The steps for the APA testing procedure using the pressurized hose are outlined

below:

* Preheat the specimen in the APA test chamber to 640C (147F) for a minimum of 6
hours but not more than 24 hours before the test.

* Set the hose pressure gauge reading to 100+5psi.

* Calibrate each wheel with the load cell to read a load of 100+51b.

* Secure the preheated, molded specimen in the APA, close the chamber doors and
allow 10 minutes for the temperature to stabilize.

* Apply 25 load cycles and then take initial (datum) measurements.









* Place the specimen back in the APA, close the chamber doors and allow 10 minutes
for the temperature to stabilize.

* Restart the APA and continue rut testing.

* Repeat measurement procedure at 1000, 2000, 4000, and 8000 cycles.

5.4.3 APA Loading Strip Testing Procedure

Outlined below are the steps for the new APA testing procedure:

* Preheat the specimen in the APA test chamber to 640C (147F) for a minimum of 6
hours but not more than 24 hours before the test.

* Calibrate the steel wheel with the load cell to read a load of 150+51b.

* Secure the preheated, molded specimen in the APA, close the chamber doors and
allow 10 minutes for the temperature to stabilize.

* Apply 25 load cycles and then take initial (datum) measurements.

* Place the specimen back in the APA, close the chamber doors and allow 10 minutes
for the temperature to stabilize.

* Restart the APA and continue rut testing.

* Repeat measurement procedure at 1000, 2000, 4000, and 8000 cycles.



5.5 Summary

The mixtures used in this project contained material that was no longer available

(milled material), so the aggregate blend was reconfigured to match the original JMF

gradation. Overall, the laboratory engineered mixtures had similar gradations and

volumetric properties with the original HMA.

The new testing method for the APA alters the data-recording procedure to gather

more information about the sample's performance. Whereas the current FDOT procedure

requires four rut-depth measurements per specimen, the new method records

approximately 1300 response points per specimen for the analysis. The next chapter







84


describes the data analysis methods as well as the valuable information from the extra

measurements.














CHAPTER 6
DATA ANALYSIS METHODOLOGY

6.1 Overview

The data-recording method in the APA test procedure was modified to incorporate

a more detailed way of analyzing the test results. Instead of the dial gauge, used to

measure a single (the lowest) point on the specimen, the new method uses a contour

gauge to capture the entire surface profile of the sample. The contour gauge's rods are

pushed downwards until they come in contact with the specimen, forcing the contour

gauge to assume the shape of the specimen's surface. For each of the three locations on

the measuring plate, the specimen's surface deformation profile is traced on a card and

scanned as a bitmap image for further analysis.

This chapter will focus on the methods used to analyze the data recorded with the

contour gauge. The new data-recording method (contour gauge) enables the user to

calculate more than just the highest deformation point. With the new method it is

possible to monitor the rate and the mode (consolidation/instability) at which the material

is failing.

6.2 Digitizing the Measured Profile

Each specimen requires 15 cards five measurements at three locations to

capture its rutting profile throughout the 8000-cycle test. To proceed with the data

analysis, the cards are scanned as bitmap images (.bmp) and digitized with the help of a

program called Grafula3 [Wishnevsky 2003]. Grafula3 allows the user to digitize points