Digital Mix Design For Performance Optimization of Asphalt Mixture

Digital Mix Design for Performance Optimization of
Asphalt Mixture
Ying Li
Dissertation submitted to the faculty of
Virginia Polytechnic Institute and State University
in partial fulfillment of the requirements for the degree of
Doctor of Science
In
Civil Engineering
Wang, Linbing
Abbas, Montasir Mahgoub
Dove, Joseph E
Hajj, Muhammad R
Easterling, William S
November 21, 2014
Blacksburg, Virginia
Keywords: Discrete Element Method, Aggregate Gradation, Asphalt Content, Visco-elastic
Model, Binder Film, X-Ray Tomography, Compaction Temperature, Pore Water Pressure
Copyright 2014, Ying Li

Digital Mix Design for Performance Optimization of
Asphalt Mixture
Ying Li
(ABSTRACT)
Asphalt mix design includes the determination of a gradation, asphalt content, other volumetric
properties, the evaluation of mechanical properties and moisture damage potentials. In this study,
a computational method is developed to aid mix design. Discrete element method (DEM) was
used to simulate the formation of skeleton and voids structures of asphalt concrete of different
gradations of aggregates. The optimum gradation could be determined by manipulating the
particle locations and orientations and placing smaller particles in the voids among larger
particles. This method aims at an optimum gradation, which has been achieved through
experimental methods. However, this method takes the mechanical properties or performance of
the mixture into consideration, such as inter-aggregate contacts and local stability. A simple
visco-elastic model was applied to model the contacts between asphalt binder and aggregates.
The surface texture of an aggregate particle can be taken into consideration in the inter-particle
contact model. The void content before compactions was used to judge the relative merits of a
gradation. Once a gradation is selected, the Voids in Mineral Aggregate (VMA) can be
determined. For a certain air void content, the mastics volume or the binder volume or the
asphalt content can be determined via a digital compression test. The surface area of all the
aggregates and the film thickness can be then calculated. The asphalt content can also be
determined using an alternative approach that is based on modeling the inter-particle contact with
iii
an asphalt binder layer. In this study, considering the necessity of preservation of the compaction
temperature, the effect of various temperatures on Hot Mix Asphalt (HMA) samples properties
has been evaluated. As well, to evaluate the effect of this parameter on different grading, two
different grading have been used and samples were compacted at various temperatures. Air voids
also influence pore water pressure and shrinkage of asphalt binder and mixture significantly. The
shrinkage is measured on a digital model that represents beams in a steel mold and is defined as
the linear autogenous deformation at horizontal direction.

iv
ACKNOWLEDGEMENT
I would like to gratefully and sincerely thank Dr. Linbing Wang for his guidance,
understanding, patience, and most importantly, his friendship during my graduate studies at
Virginia Tech. His mentorship was paramount in providing a well-rounded experience consistent
my long-term career goals. He encouraged me to not only grow as an experimentalist and an
engineer but also as an instructor and an independent thinker. I am not sure many graduate
students are given the opportunity to develop their own individuality and self-sufficiency by
being allowed to work with such independence. For everything you’ve done for me, Dr. Wang, I
thank you. I would also like to thank all of the members of Dr. Wang’s research group. These
friends and co-workers provided for some much needed humor and entertainment in what could
have otherwise been a somewhat stressful laboratory environment.
I would like to thank the Department of Civil Engineering at Virginia Tech, especially those
members of my doctoral committee for their input, valuable discussions and accessibility.
Finally, and most importantly, I would like to thank my parents Yuan Li and Li Wang. I
thank my parents, for their faith in me and allowing me to be as ambitious as I wanted. It was
under their watchful eye that I gained so much drive and an ability to tackle challenges head on.
v
Table of Contents
1. INTRODUCTION .......................................................................................................................................... 1
2. DETERMINATION OF GRADATION OF AGGREGATE ...................................................................... 5
Literature Review .................................................................................................................................5
Background of Particle Packing ........................................................................................................... 5
Current Technology Review .................................................................................................................. 9
Summary of Current Technologies ...................................................................................................... 23
Simulation of Aggregates Packing Process in PFC3d ......................................................................26

Determination of Optimal Gradation ...............................................................................................30
Summary..............................................................................................................................................34
3 DETERMINATION OF ASPHALT CONTENT ....................................................................................... 35
Literature Review ...............................................................................................................................35
Binder Content Determination Methods ............................................................................................. 36
Ignition Method Test Procedures ........................................................................................................ 40
Procedure of Determination of Asphalt Content Using DEM .........................................................41

Results and Analysis ...........................................................................................................................43
Calculation of Film Thickness by Using Selected Gradation and Asphalt Content ......................45
Calculation of Specific Surface Area Factor. ...................................................................................... 45
Calculation of Film Thickness ............................................................................................................ 47
Summary..............................................................................................................................................48
4 EFFECT OF FILM THICKNESS ON ASPHALT MIXTURE................................................................. 50
Contact Model..................................................................................................................................... 51
Numerical Methods ............................................................................................................................. 56
Build a Digital Model by Using DEM................................................................................................59

User Defined Contact Model in PFC3d .............................................................................................. 59
Simple Viscoelastic Model .................................................................................................................. 62
Experiment ..........................................................................................................................................63
vi
Simulation ............................................................................................................................................78
Summary..............................................................................................................................................84
5 EFFECT OF AGGREGATE SHAPE ON ASPHALT MIXTURE ........................................................... 86
Materials and Methods .......................................................................................................................88
Aggregate ............................................................................................................................................ 88
Image Measurements of Coarse Aggregate ........................................................................................ 92
Calculation of Particle Index of Coarse Aggregates Based on DEM Simulations .........................94

Simulated Indirect Tensile Tests on Digital Samples with Selected Grading Aggregates Using DEM
..............................................................................................................................................................96
Simulated Compression Tests on Digital Samples with Ten Aggregates in Each of Them Using DEM
..............................................................................................................................................................98
Summary of DEM Properties .............................................................................................................. 98
Results and Analysis ......................................................................................................................... 101
Compression Tests Using X-ray Image Scanner ............................................................................101

Equipment and Specimens Preparation ............................................................................................ 101
Test and Results Analysis .................................................................................................................. 104
Summary............................................................................................................................................106
6 EVALUATION OF COMPACTION TEMPERATURE SENSITIVITY .............................................. 108
Literature Review .............................................................................................................................108
Aggregate Grading............................................................................................................................109
Asphalt Binder Selection ..................................................................................................................110
DEM Simulation Process of Superpave Gyratory Compactor (SGC) Compaction ....................111
Result and Analysis ...........................................................................................................................115
Analysis of Density............................................................................................................................ 115
Analysis of Percent of Air Voids (VA)................................................................................................ 116
Analysis of Percent Voids Filled with Asphalt (VFA) ........................................................................ 117
Analysis of Flow................................................................................................................................ 118
Summary............................................................................................................................................119
7 EVALUATION OF MOISTURE DAMAGE POTENTIAL .................................................................... 121
vii
Literature Review .............................................................................................................................121

Procedure of Numerical Calculation of Pore Water Pressure .......................................................... 123
Simulated Measurement of Pore Water Pressure in Asphalt Mixture .........................................126

Digital Model and Properties Used in DEM Simulation .................................................................. 126
Result and Analysis ........................................................................................................................... 128
Influence of Entrained Air Voids on Pore Water Pressure and Shrinkage on Asphalt Concrete132
Numerical Analysis ........................................................................................................................... 132
Traditional Method of Experiment Test............................................................................................. 135
Simulated Shrinkage Process Using DEM ........................................................................................ 139
Result and Analysis ........................................................................................................................... 141
Summary............................................................................................................................................144
8 CONCLUSIONS .......................................................................................................................................... 145
REFERENCES................................................................................................................................................ 148
viii
LIST OF FIGURES
FIGURE 2.1: Gradex 2000 Particle Size Analyzer……………….……….………….……….…10
FIGURE 2.2: Gradex 2000 Sieve Configuration…………….………………….……….………12
FIGURE 2.3: Gradex 2000 Brush Cleaning Attachment…….……………………….…….……12
FIGURE 2.4: Product File Information from Gradex 2000……….………………….…….……14
FIGURE 2.5: Schematic of the VDG-40 Video Grader…….……………………….…….……17
FIGURE 2.6(a): Mixture Contains Aggregate Larger Than 4.75mm (State I) …………………28
FIGURE 2.6(b): Mixture Contains Aggregate Larger Than 2.36 mm (State II) …………..……29
FIGURE 2.6(c): Mixture Contains Aggregate Larger Than 1.18 mm (State III) ………….……29
FIGURE 2.7: Five Gradation Curve VS. Maximum Density Line………………………………32
FIGURE 2.8(a): Mixture Model with Selected Gradation Aggregates and Asphalt Binder……33
FIGURE 2.8(b): View of Contact Force of the Mixture at Equilibrium State………………...…34
FIGURE 3.1: Stress VS. Strain of Mixture 1………………….…………………………………44
FIGURE 3.2: Asphalt Content VS. Peak Strength…………………….…………………………44
FIGURE 4.1: Flowchart of a PFC3D Program Code Using UDM………………………………60
FIGURE 4.2: Asphalt Mixture Includes Two Aggregates and Asphalt Binder…………………61
FIGURE 4.3: Skyscan 1174 System…………………..…………………………………………64
FIGURE 4.4: Testing Stage of the Microscopy System (59)…………………….………………65
FIGURE 4.5: Crystal Aggregates Used in Compression Tests…………………..………………66
FIGURE 4.6: Asphalt Mixture Sample Used in Compression Tests…………………………….66
FIGURE 4.7(a): X-ray Image of Sample 1………………………………………………………68
FIGURE 4.7(b): Resistant Force vs. Deformation of Sample 1…………………………………68

ix
FIGURE 4.10(a): X-ray Image of Sample 4…………………….………………………………71
FIGURE 4.10(b): Resistant Force vs. Deformation of Sample 4….……………………………71
FIGURE 4.11(a): X-ray Image of Sample 5……………………….……………………………72
FIGURE 4.11(b): Resistant Force vs. Deformation of Sample 5….……………………………72
FIGURE 4.12(a): X-ray Image of Sample 6……………………………………………………73
FIGURE 4.12(b): Resistant Force vs. Deformation of Sample 6…….…………………………73
FIGURE 4.13(a): X-ray Image of Sample 7……………….……………………………………74
FIGURE 4.13(b): Resistant Force vs. Deformation of Sample 7…….….………………………74
FIGURE 4.14(a): X-ray Image of Sample 8…………………………….………………………75
FIGURE 4.14(b): Resistant Force vs. Deformation of Sample 8……….….……………………75
FIGURE 4.15(a): X-ray Image of Sample 9……………….……………………………………76
FIGURE 4.15(b): Resistant Force vs. Deformation of Sample 9…….……….…………………76
FIGURE 4.16(a): X-ray Image of Sample 10……………………………………………………77
FIGURE 4.16(b): Resistant Force vs. Deformation of Sample 10………………………………77
FIGURE 4.17: Relationship between Resistant Force and Film Thickness……………………78
FIGURE 4.18: Summary of Peak Axial Stresses……………………………………...…………79
FIGURE 4.19(a): Two Aggregates and a 10µm Film Layer in Compression Test………………79
FIGURE 4.19(b): Detailed View of Film Layer, Film Thickness = 10µm………………………79
FIGURE 4.19(c): Axial Stress in Compression Test, Film Thickness = 10µm………………80
x
FIGURE 5.1(a): Cubic-shaped Aggregate………………………….…………………………89
FIGURE 5.1(b): Detailed View of Cubic-shaped Aggregates in Simulation…………………89
FIGURE 5.2(a): Rod-shaped Aggregate………………………………………………………89
FIGURE 5.2(b): Detailed View of Rod-shaped Aggregates in Simulation……………………90
FIGURE 5.3(a): Blade-shaped Aggregate………………………….…………………………90
FIGURE 5.3(b): Detailed View of Blade-shaped Aggregates in Simulation…………………91
FIGURE 5.4(a): Hexagonal Disk-shaped Aggregate…………………………………………91
FIGURE 5.4(b): Detailed View of Hexagonal Disk-shaped Aggregates in Simulation………92
FIGURE 5.5: Tamping Rod and Sleeve ………………………………………………………..95
FIGURE 5.6: Effect of Aggregate Shape on Indirect Tensile Strength………………………98
FIGURE 5.7: Digital Sample Include 10 Sphere Aggregates…………………………………99
FIGURE 5.8: Digital Sample Include 10 Cubic Aggregates…………………………………100

xi
FIGURE 5.9: Digital Sample Include 10 Random Aggregates………………………………100
FIGURE 5.10: Skyscan 1174 System…………………………………………………………102
FIGURE 5.11: Testing Stage of the Microscopy System………………………………………102
FIGURE 5.12: Crystal Particles and Aggregates Used in Compression Test…………………103
FIGURE 5.13: Asphalt Mixture Specimen Used in Compression Test………………………103
FIGURE 5.14: X-ray Images of Specimens with Asphalt Binder and (a) Ten Cubic Aggregates,
(b)Ten Sphere Aggregates, and (c) Ten Random Aggregates…………………..105
FIGURE 5.15: Stress vs. Strain of One Specimen with Ten Sphere Aggregates………………106
FIGURE 6.1: Viscosity of Bitumen for Determination of Compaction Temperature………….110
FIGURE 6.2: DEM Simulation Process of SGC…….…………………………………………114
FIGURE 6.3: Density vs. Compaction Temperature…………………………………………116
FIGURE 6.4: Air voids vs. Compaction Temperature……………………………………….…117
FIGURE 6.5: VFA vs. Compaction Temperature………………………………………………118
FIGURE 6.6: Flow vs. Compaction Temperature………………………………………………119
FIGURE 7.1: Mechanical Elements Introduced in Ball–ball Contacts in Normal, Tangential and
Rotational Directions in DEM……...……….………………….…………………123
FIGURE 7.2: Schematic Illustrations of Measurement Sphere and Cross Section Areas……124
FIGURE 7.3: Numerical Specimen for Undrained Triaxial Compression Test………………127
FIGURE 7.4: Pore Water Pressure vs. Dynamic Load…………………………………………129
FIGURE 7.5: Pore Water Pressure vs. Elastic Modulus in Compression Tests, air void = 8%...130
FIGURE 7.6: Pore Water Pressure vs. Elastic Modulus with Varied Poisson Ratios…………130
FIGURE 7.7: Hydraulic Conductivity vs. Air Voids…………………………………………...131
FIGURE 7.8: Pore Water Pressure vs. Air Voids………………………………………………132

xii
FIGURE 7.9: Equipment for Measuring Linear Autogenous Deformation of Cement Paste …136
FIGURE 7.10: Measuring System with Cast-in Nails…………………………………………137
FIGURE 7.11: Measuring System with Moveable End Plates in 40×40×160mm3 Beam……137
FIGURE 7.12: Measuring System with Horizontal Cast-in Bar in 150×150×1000mm3 Beam..138
FIGURE 7.13: Measuring System with Vertical Cast-in Bar in 270×270×100mm3 Slabs…….138
FIGURE 7.14: Measuring System with Flexible Tubes (diameter 100 mm and length 375 mm)
and vertical length measurement………………………………………………139
FIGURE 7.15: Shrinkage of Asphalt Concrete without Air Entraining Mixture……………….141
FIGURE 7.16: Pore Water Pressure of Asphalt Concrete without Air entraining Mixture……142
FIGURE 7.17: Shrinkage of Asphalt Mixture with Varied Air Voids…………………………143
FIGURE 7.18: Pore Water Pressure of Asphalt Mixture with Varied Air Voids………………144
xiii
LIST OF TABLES
TABLE 2.1: Grading Results vs. Standard Dry Sieve Analysis with Gradex 2000……………14
TABLE 2.2: Summary of Current Technology…………………………………………………..23
TABLE 2.3: Capability Summary of the Technologies Reviewed………………………………24
TABLE 2.4: Gradation of Five Models………………………………………………………….30
TABLE 2.5: Calculated fv and Porosity of Five Models………………………………………31
TABLE 3.1: Minimum Sample Weight…………………………………………………………41
TABLE 3.2: Gradation Information……………………………………………………………...42
TABLE 3.3: Asphalt Content and Contact Information…………………………………………42
TABLE 3.4: Contact Properties Applied in Simulations………………………………………43
TABLE 3.5: Calculated and Surface Area Factors………………………………………………47
TABLE 4.1: Properties Applied in Simulations…………………………………………………63
TABLE 4.2: Experiment Specifics……………………………………………………………….67
TABLE 4.3: Parameters and Results of Compression Test………………………………………67
TABLE 5.1: Aggregate Gradations………………………………………………………………88
TABLE 5.2: Flatness Ratio and Elongation Ratio of Four Different-shaped-aggregates………89
TABLE 5.3: Aggregate Geometric Characteristics………………………………………………93
TABLE 5.4: Measured Particle Index (PI) for Coarse Aggregate……………………………95
TABLE 5.5: Peak Normal Strengths Achieved in Indirect Tensile Test………………………97
TABLE 5.6: Geometry Information of the Samples Used in Simulated Compression Test……99
TABLE 5.7: Contact Model and Mechanic Property…………………………………………99
TABLE 5.8: Peak Normal Strengths Achieved in Compression Test on Digital Samples……101
TABLE 5.9: Weights of Aggregates and Asphalt Binder………………………………………104

xiv
TABLE 5.10: Loading Information of Compression Test……………………………………105
TABLE 5.11: Peak Normal Strength of Each Specimen achieved in Compression Test………105
TABLE 6.1: Gradations of Aggregates Used in Simulation……………………………………109
TABLE 6.2: Properties of Asphalt Binder Used in DEM Simulation………………………….110
TABLE 6.3: Data Measured at a Serial of Modeling Temperatures……………………………115
TABLE 7.1: Input Parameters of the Simulated Specimen……………………………………127
TABLE 7.2: Elastic Modulus and Pore Water Pressure………………………………………129
TABLE 7.3: Contacts Properties & Other Governing Parameters Specified in DEM Model…140
1
1. INTRODUCTION
Asphalt concrete is a composite material consisting of aggregates, asphalt binder and air
voids. The constitutive behavior of the material depends largely on the interaction between
aggregates and asphalt binder. The aggregate skeleton determines the load carrying
mechanism and the asphalt binder serves as an adherent, which enables the mixture to hold
tensile and shear stresses (1). Discrete Element Method (DEM) can be used to analyze the
individual characteristic effects of each component in the asphalt mixture, such as the
influence of asphalt content, gradation and relative slippage among aggregate particles. Given
the proper internal structure and constitutive contact models, the simulation could avoid a
huge amount of sample preparation work in lab and provide preliminary assessment of the
behavior of asphalt mixtures, aiding the mix design by providing fundamental understanding
of the behavior.
DEM was developed by Cundall and Strack for simulating movement and interaction of
assemblies of rigid particles (2). In their original BALL model, the movement and interaction
of unbound assemblies of rigid spherical particles subjected to external stress was simulated.
Later developments of this approach were implemented in software PFC3d (Itasca Consulting
Group Inc. 2003). In PFC3d, a gradation of three-dimensional (3D) particles are assembled
together to form a skeleton structure, and inter-particle contacts can be assessed.
Major works presented in this study include: packing aggregates and determining of the
optimal gradation through DEM simulation; determining the optimal asphalt content by
simulating compression tests on asphalt mixtures via a computational method; and estimating
2
the film thickness based on determined gradation and asphalt content. In order to represent
viscoelastic properties of asphalt binder, a simple visco-elastic model and parallel bonds
were applied to the contacts between aggregates and asphalt binder in PFC3d.
DEM was used to simulate the formation of the skeleton and the void structures in
asphalt concrete for different gradations of aggregates before compaction. The particle
surface texture can be included in the inter-aggregate contact model. The void content before
compactions may be used to judge the relative merits of a gradation. It is assumed that an
optimum gradation can be achieved, and the structure is most stable if the number of
inter-aggregate contacts of coarse aggregate reaches the maximum value. The optimum
gradation could be determined by manipulating the particle locations and orientations and
placing smaller particles in the voids among larger particles. For the same type of aggregates,
one criterion of judging the rationality of computational simulation method is whether the
computational method could yield the same gradation as obtained in the traditional mix
design. The computational method can take the mechanical properties or performance of the
mixture into consideration, such as inter-aggregate contacts and local stability.
Once a gradation is selected, the Voids in Mineral Aggregate (VMA) can be determined.
For a certain target air void content, the asphalt content can be determined. If the asphalt
content is known, then the average binder film thickness, which is related to the total surface
area of both aggregates and filler, can be conveniently determined. As the binder layer might
affect the plastic/permanent deformation of particles in contact, this layer’s thickness might
be directly related to the rutting performance of the mixture. In addition, if the asphalt film
thickness is too small, aging and wearing of the asphalt binder will reduce its fatigue life.
3
Therefore estimation of the film thickness is important and can be conveniently obtained via
the computational method.
Mineral aggregate constitutes approximately 95% of hot-mix asphalt (HMA) by weight.
The mineral aggregate is made up predominantly of coarse aggregate. Aggregate
characteristics such as particle size, shape, and texture influence the performance and service
ability of hot-mix asphalt pavement. Flat and elongated particles tend to break during mixing,
compaction, and under traffic. Therefore, aggregate shape is one of the important properties
that must be considered in the mix design of asphalt pavements to avoid premature pavement
failure. In this study, DEM has been used to simulate standard particle index test and indirect
tensile test in order to evaluate the effect of aggregate shape on the strength of asphalt
mixtures.
Due to HMA’s high durability, solid production, temperature control, moisture and quick
provision for traffic crossing, it has still caught attention by many people. Determination of
volumetric parameters of HMA samples is of vital importance for quality control of
constructed roads. The compaction temperature is one of the major issues in HMA and also
one of important criteria in the process of producing good quality hot mix asphalt. Also, the
temperature is a key factor in the control of bitumen viscosity, which affects its ability to coat
and provide adequate lubrication for the aggregates and slides with each other. In this study,
compaction temperature changes from 120 oC to 165 oC. Changes of density, air voids, and
flow of asphalt mixture which result from the increasing compaction temperature are
evaluated.
4
Previous work has revealed that the pore water pressure (PWP) reduction of materials
such as paste, mortar and concrete reflects rather well the stiffness evolution and volume
change of such materials. Early PWP reductions are caused by two interacting mechanisms:
Vacuum effect caused by thermal shrinkage and meniscus effect (capillary tension). Air voids
will alter the early age PWP, and thus, the above mentioned properties, mainly because air
voids act as buffer for both mechanisms. This is because they may grow as long as the
stiffness of the asphalt concrete is low, which results in PWP release. The study presents and
discusses some results from a DEM based test program of asphalt binder with and without air
entraining admixture (AEM) as well as concretes with and without AEM, conducted in order
to evaluate the influence of air voids on the mentioned properties, both from a fundamental
and practical point of view.

5
2. DETERMINATION OF GRADATION OF AGGREGATE
Literature Review
Background of Particle Packing
The impact of aggregate gradation on the performance of asphalt mixtures has been
extensively studied through the years. Several studies have shown that aggregate gradation
affects the resistance of asphalt mixture to cracking, rutting, ageing and moisture damage (1).
Furthermore, the way that aggregate particles, bitumen binder and air voids interact with each
other influences how a mixture responds under different loading conditions (3). With a better
understanding of these interactions, future mix design can be optimized by mixing the
available materials in the best way.
Asphalt mixture microstructure depends mainly on their size distribution, shape and
angularity. Bitumen binder flows around the particles, forming a film around them and
binding all the components together. Aggregates´ mechanical properties, such as resistance to
abrasion and strength, are dependent primarily on their mineral composition. However, good
processing procedures can significantly improve the quality of the final aggregate by the
elimination of weaker rock layers and crushing particles to reveal rough surfaces. Aggregate
gradation is the distribution of particle sizes expressed as a percent of the total weight.
Gradation can be determined by sieve analysis and expressed as total percent passing various
sieve sizes.
Experimental studies on granular materials showed the existence of stress-transmitting
paths enclosing virtually stress free regions. In particulate materials then, the load is
6
transferred through chains of particles and other smaller particles play the secondary role in
preventing the main chain from buckling (4). Based on the observations mentioned above,
two sub-structures within the aggregate particles have been defined: the Primary Structure,
range of sizes which due to their volume fraction provides the load bearing capacity of the
mixture; and the Secondary Structure, range of sizes smaller than the first one, which
provides stability to the Primary Structure (5). Each substructure can be identified by the
packing theory for spherical particles. Results from field and laboratory tests can be used to
validate the relationship between the aggregate content of Primary Structure and rutting
performance. The model also proposes a distribution system of the bitumen binder. Former
researches show that the thickness of the film around both structures effects not only
permanent deformation but also the resistance to moisture damage of asphalt mixtures. Early
studies proposed that the best gradation for HMA is the one that gives the densest packing,
increasing stability through increased inter-particle contact (6). Sufficient air void content is
necessary for the durability and workability of the mixture, so there still are some air voids in
the mixture to avoid bleeding or rutting.
The determination of the grading of an aggregate blend or stockpile is one of the oldest
test procedures used in the hot mix asphalt (HMA) and aggregate industries. The control of
aggregate grading by the HMA industry begins during aggregate production and ends when
the grading of the finished HMA is determined. Sieve analyses are used in many different
applications, some of which are the determination of aggregate grading for use in HMA mix
design procedures (Marshall, Hveem, and Superpave), the process control of produced
material, and the particle size control and monitoring of quarried materials. In the HMA
7
industry, sieve analyses are often performed on plant produced samples to determine the
proximity of the produced aggregate blend to the job mix formula aggregate grading. The
number of grading tests performed for a given project depends on many factors such as the
number of stockpiles, type of HMA plant, tonnage of HMA produced, process control
methods used, and acceptance procedures outlined by the owner. Furthermore, the grading of
the aggregate blend in HMA can be considered one of the most important mix parameters
since it affects, in varying degrees, the stiffness, stability, durability, permeability,
workability, fatigue resistance, frictional resistance, moisture susceptibility, and volumetric
properties of the produced mixture. Therefore, it is very critical that the grading analysis of
the produced material be done in a timely and accurate manner, so as to yield the true result,
and to allow for the adjustment of the aggregate grading, if necessary, to ensure a quality
product. Although aggregate grading analysis is one of the oldest test procedures, it remains a
very time consuming and labor intensive task.
There is a need today within the HMA industry for a rapid, automated method to
determine aggregate grading. This automated method should reduce the time required to
perform a grading test as well as reduce the amount of technician time required. The need to
reduce the test time arises from the fact that grading testing of aggregate samples is used for
process control during the production of HMA and requires a significant amount of time for
an HMA project. A typical project may require from 5-10 grading tests per day depending on
the specification requirements, number of aggregates, and plant production rate. If the test
time for grading testing can be reduced by 50 percent (approximately 15 minutes), several
hours of total test time per day can be saved.

8
Today, the most widely accepted and accurate method for aggregate grading
determination is through a washed or wet sieve analysis, as outlined in ASTM C117 (7) and
ASTM C136 (8). The main reason for performing a wet sieve analysis is to obtain a more
accurate indication of the true grading, primarily on the material passing the 0.075 mm sieve.
As one might expect, when the amount of material passing the 0.075 mm sieve is low, the
washed and the dry sieving procedures often yield very comparable results; however, when
the percent passing is high the results can vary considerably.
A method of determining the gradation of an aggregate blend in a more efficient manner
is greatly needed in the HMA and aggregate industries. By increasing the efficiency of the
grading analysis, there is a potential for greater control to be obtained over the produced
product by the increased ability to perform more grading analysis throughout production.
Besides aggregate stockpile grading, the grading of the finished HMA mixture must also
be determined for control and acceptance purposes. To obtain the aggregate grading either the
aggregate must be recovered from the HMA or a belt sample be taken and analyzed. The
technician time involved is approximately one-half to one hour for each test. Add this to the
time required to determine stockpile grading and several hours of technician time can easily
be expended on a relatively small amount of process control. The need to automate the testing
procedure is evident.
If grading testing can be automated, the resulting reduction in time requirements would
allow the producer to choose one of two options. With the increase in available technician
time, more testing could be performed to allow for better process control. Or if desired, fewer
technicians would be needed to perform the same level of process control.

9
Current Technology Review
In reviewing current technologies for possible application to aggregate grading testing,
many possibilities were investigated. Several industries make use of grading tests for
manufacturing and/or quality control of products. For example, both the food and drug
industries make wide use of grading testing to assist them during production. In addition, the
in-line measurement of particle size is used by heavy industries such as power generation and
steel manufacturing to help control plant emissions.
The HMA and aggregate production industries have continued to pursue new technology
aimed at increasing the accuracy of grading testing while decreasing the time required. Air
separation techniques are now being used at aggregate production facilities to “air wash”
aggregates. Washing of the coarse aggregate in this manner creates less waste and reduces
environmental concerns.
In the preparation of this report, all of these technologies were carefully reviewed. Each
of these was judged against the four criteria of reduced test time, reduced technician time,
affordability, and the ability to obtain an accurate grading. The methods reviewed are
discussed individually in the following paragraphs.
Gradex 2000
The Gradex 2000 Particle Size Analyzer, manufactured by Rotex, Inc., located in
Cincinnati, Ohio, has been used for a number of years in many industries, such as the
pharmaceutical, ceramic, food products, etc. (9). However, its application in the hot mix
10
asphalt and aggregate industries as a means of determining the grading of an aggregate blend
has only recently been evaluated.
The automatic grading device, shown in FIGURE 2.1, accomplishes grading analysis by
automating the conventional sieve analysis procedure which has been used for many years
with good results. The grading device consists of a set of standard 203 mm diameter round
U.S. sieves, an automatic feeding/loading assembly (optional), a collection pan, and a 0.1
gram resolution electronic balance. The device, as first built, was capable of holding six
full-height sieves or nine half-height sieves, or any combination thereof.
FIGURE 2.1 Gradex 2000 Particle Size Analyzer (9)
A computer system, completed with operational software for the grading device, is
11
interfaced with the device to allow for operator control and retrieval of grading analysis data.
Operation of the device is very simple and requires a minimum amount of labor. Firstly the
operator completes all of the sample information using the provided computer software. This
information includes items such as the operator’s name, test specimen identification, sieving
time, etc. Next, a sample is loaded into the top of the device, either manually through the feed
chute, or by the automatic feeding/loading assembly, which is sold as an attachment. Once
the sample is loaded, the operator initiates the grading analysis by the touch of a key. The
device then sieves the material for the specified sieving time (approximately 10 minutes) in a
rotary or tapping motion, which is currently used by many mechanical sifters or shakers in
operation today. After sieving is complete, the sieves are emptied, as shown in FIGURE 2.2,
into the collection pan, which rests upon the electronic balance. This process begins with the
pan and proceeds through the sieves from fine to coarse, until the last sieve is emptied. After
each individual sieve is emptied into the collection pan, a brush, shown in FIGURE 2.3,
operated through air pressure, cleans the inside of the sieve to remove any particles which
may have become lodged during the sieving process. After the individual sieve has been
cleaned, it is returned to its appropriate location in the sieve stack. Following the completion
of the sieving, weighing, and cleaning procedures, the results of the grading analysis can be
viewed on the computer screen or saved to a data file. A sample of the product file
information is shown in FIGURE 2.4. As is shown, information such as sample identification,
sieves used for analysis, and tabular results showing the percent retained and percent passing
are provided for the user.

12
FIGURE 2.2 Gradex 2000 Sieve Configuration (9)
FIGURE 2.3 Gradex 2000 Brush Cleaning Attachment (9)
In order to determine accuracy of the Gradex 2000 to determine the grading of aggregate
13
samples, two blended aggregate mixes were prepared and tested in the device. The results of
this preliminary testing with the Gradex 2000 are shown in TABLE 2.1. None of the results
shown in TABLE 2.1 include sample washing prior to the grading determination. The results
indicate that the Gradex 2000 is accurate at determining the percent passing for most sieves,
but it has a tendency to underestimate the percent passing the 0.075 mm sieve. However,
some of the underestimating on the 0.075 mm sieve is believed to have been caused by sieve
blinding, a problem that can be remedied. Note also that the Gradex 2000 is limited in the
number of sieves that it can hold. The sieve stack can consist of either six full height sieves or
nine half height sieves. Mixture 1 was tested using a stack of six full height sieves. Mixture 2
was tested using a stack of nine half height sieves. The ninth sieve in this stack was a
“placeholder” (a sieve frame with no wire). Placeholder sieves can be used to make the
Gradex 2000 even more flexible. A placeholder sieve was therefore inserted into the stack as
the ninth sieve. The placeholder does not measure an aggregate fraction size, but simply
allows the Gradex 2000 to hold the stack properly.

14
FIGURE 2.4 Product File Information from Gradex 2000 (9)
TABLE 2.1 Grading Results vs. Standard Dry Sieve Analysis with Gradex 2000 (9)
Percent Passing
Sieve size
Mixture 1 Mixture 2
(mm)
Standard Test Gradex 2000 Standard Test Gradex 2000
25.0 100.0 100.0 100.0 100.0
19.0 100.0 99.8 100.0 99.7
12.5 78.8 78.9 89.2 89.9
9.5 24.7 24.1 49.0 49.1
4.75 1.1 0.5 25.6 25.3
2.36 1.0 - 20.3 20.2
1.18 1.0 - 17.3 -
0.600 0.9 0.4 14.5 14.1
0.300 0.9 - 13.5 -
0.150 0.8 - 10.8 -
0.075 0.5 0.1 5.0 1.2
A shorter testing time is possible if the sample is not washed prior to testing. If an
accurate measure of the percent finer than the 0.075 mm sieve is required, the sample can be
15
washed prior to using the Gradex 2000. However, in certain instances, an apparent reduction
in both test time and required technician time can be realized while obtaining an accurate
measure of the percent passing the 0.075 mm sieve. This arises from the ability to
automatically test several samples in sequential order. Therefore, if a technician has several
aggregate gradations to perform, all of the preliminary work such as washing the sample (or
extracting the asphalt binder) could be performed during the working day and the resulting
aggregate samples could be set to sequentially test during the night. When the technician
returned the next day, the computer would have all of the grading testing completed. In this
scenario an accurate measure of the percent passing the 0.075 mm sieve is achieved, the
technician time involved is reduced (he/she doesn’t have to empty and weigh the contents of
each sieve), and the grading testing time is removed from the normal work day. At the time of
this study the cost for the Gradex 2000 varies, depending upon equipment configuration,
from $17,000 to $25,000.
VDG-40 Video Grader
The VDG-40 Video Grader was developed by the Laboratoire Central des Ponts et
Chaussèes (the French Public Works Laboratory) and is currently marketed in North America
by a Canadian Company. It has recently become the standard French test method for the
determination of aggregate grading as well as percent flat and elongated particles. The
equipment consists of a chute through which the aggregate is fed, a vibrator and rotating
drum for separating particles, a light source, and a bank of photosensitive cells (10). A
schematic of the machine is shown in FIGURE 2.5. A personal computer is used to collect
and analyze the data. The VDG-40 can be used to continuously monitor the grading of an
16
aggregate stream. Cost for the standard VDG-40 Video Grader is approximately $50,000.
Prior to sample testing, the VDG-40 must be calibrated. This is done by preparing two
blended aggregate samples, one corresponding to the minimum grading specification and the
second to the maximum grading specification. These samples are tested in the machine
individually to allow it to establish the grading envelope. Once the grading envelope has been
determined for the given aggregate blend, testing of samples can begin (11).
Firstly, all particles that smaller than 1.0 mm must be removed from the sample when
you try to test a sample using VDG-40. This is due to the limits of the machine. It is capable
of testing aggregate sizes ranging from 1.0 to 50.0 mm although it can be adapted to handle
particles up to 63.0 mm. The separation of particles smaller than 1.0 mm involves washing
and “scalping” the sample with a 1- mm sieve. The total weight of the sample, the weight of
the sample portion finer than the 1- mm sieve, and the weight of the sample portion coarser
than the 1- mm sieve must all be entered into the computer by the operator. The fraction of
the sample larger than the 1-mm sieve can then be introduced into the machine. The particles
are fed through the chute and onto the vibrator which helps to separate the particles. The
particles then pass over the drum and fall through the light curtain established by the light
source and light-sensitive cells. The resolution of the light sensitive cells is 0.2 mm
horizontally and 0.4 mm vertically. When the aggregate particles pass through the light, the
computer determines a two-dimensional shape for the particles. The computer software can
then determine a three-dimensional shape for the aggregate particles using the principle of an
ellipsoid of revolution. In addition, the computer also determines flat and elongated
characteristics. A sample takes approximately eight minutes to be tested once it has been
17
introduced into the VDG-40 (10).
FIGURE 2.5 Schematic of the VDG-40 Video Grader (10)
The VDG-40 is automated, but suffers from lack of ability to test particles smaller than
1.0 mm. In addition, the washing and scalping needed prior to testing increases the overall
test and technician time involved, thus increasing the cost for testing.
Full-Scale Particle Size Analyzer
In 1972, a report was completed at the University of Texas-Austin describing a full-scale
particle size analyzer (12). The analyzer, based on sedimentation theory, consisted of an inner
and outer ring. The outer ring was 6.1 m tall and 90 cm in diameter. The basic purpose for the
outer ring was to hold the water and to provide structural support for the inner ring. The inner
ring was 5.2 m tall and had an outside diameter of 76 cm. Two gates, one at the top and one at
the bottom were used to hold the sample until introduction into the column and to collect it
once it had settled in the column. The inner ring was suspended in the water by a scale
linkage connecting it to the outer ring. This scale linkage served as means of transferring the
18
force applied to either the upper or lower gate to load a cell. This mechanism allowed for the
sample mass to be determined on the upper gate prior to its introduction. It also allowed for
the mass of the sample to be recorded as a function of time as it settled onto the bottom gate.
No cost estimate was included in the report, but any such estimate would certainly be
outdated. Indeed the ideas used for this full-scale particle size analyzer could benefit from
technology developed since 1972.
Operation of the full-scale analyzer involved tarring the mass of the inner column and
water, introducing the sample onto the top gate and recording its mass, opening the top gate
to introduce the aggregate sample into the water column, and recording the mass of the
sample particles as a function of time as they settled onto the bottom gate. The report
concluded that a particle size ranging from 0.43-37.5 mm could be separated and analyzed
(12).
Comparison of this method against the criteria outlined earlier is difficult. It appears that
a version of this equipment making use of current technology would be automated and
reduces both test time and technician time requirements. The cost for the system is unknown,
but would most certainly be relatively expensive. It also has limits on particle sizes that can
be analyzed and it will not produce an accurate measure of the percent passing the 0.075 mm
sieve. Although the aggregate is washed in the water column, the 0.075 mm particles cannot
be detected by the system because of their small sizes.
Fractionating Water Column
This method was evaluated at the University of Waterloo (13) and is a smaller, modified
version of the full-scale particle size analyzer previously discussed. It consists of a 1.5 m long
19
by 77 mm inner diameter Plexiglas tube that is filled with water. Near the bottom of the tube
a constant intensity light shines through the water column and impinges on a bank of
light-sensitive photocells. The photocells are connected to a computer via electrical circuitry
and an analog to digital (A/D) converter board. When the intensity of the light shining on the
photocells changes, the photocells respond by changing their voltage output. This changing
voltage is registered and converted by the A/D board and then recorded by the computer.
Estimated cost for the equipment including the computer is $5,000-$10,000.
To determine the grading of an aggregate sample, the sample is put into the column at
the top. The particles descend in the water according to their sizes as modeled by Stoke’s law.
As the particles pass through the light curtain established by the light source and photocells,
the light intensity is altered and the varying voltages are recorded by the computer. By
recording the voltage as a function of time, the computer can calculate the grading of the
aggregate sample. This method has been shown to produce fairly good results when
compared with standard sieve analysis. The method does have a tendency to slightly
underestimate the coarse fraction and overestimate the fine fraction.
This method is automated and takes approximately three minutes to test a sample. It
therefore meets the criteria for reduced test time and technician involvement. The method is
also simple and the equipment relatively inexpensive. The disadvantage of this system is that
it is currently limited in the particle sizes that can be accommodated. The system is designed
to determine the grading of particles ranging from 2.38 mm to 0.075 mm. This limitation
means that the grading of the coarse aggregate fraction is not determined, nor is an accurate
measure of the percent finer than the 0.075 mm obtained. The researchers do offer
20
alternatives for solving these problems. The retained 2.38 mm fraction could firstly be
separated out and its grading determined by some other method (conventional sieve analysis
for example) while the minus 2.38-mm, retained 0.075 mm fraction grading could be
determined by the Fractionating Water Column. This would lengthen the test time. Aljassar
and Haas (13) also suggest that the minus 0.075 mm fraction could possibly be handled by
the water column with some modifications to the current technology. The size limitations of
this method make it unattractive at present, but research on this method is ongoing and should
be monitored for significant progress towards solving the problem.
Laboratory Particle Size Analyzers
Another method that can be used to determine particle size uses laser diffraction/Mie
scattering theory. Machines based on this method consist of an optical bench which contains
the lasers, mirrors, and other optical equipment, a sample chamber, and a computer for
control, data acquisition, and displaying results. The cost varies according to machine
capability and manufacturer, but is approximately $40,000.
To use a particle size analyzer the sample is firstly put into suspension in a fluid,
normally water and then placed in the sample chamber. The analysis is done by shining laser
light through the suspended sample and measuring the intensity of the scattered light as a
function of the scattering angle. The resulting light intensity function is then analyzed by the
software to determine what distribution of particle sizes would scatter the light in a similar
pattern (14). The operator only needs to aid the computer in properly calibrating the system,
and to supply the proper information to the computer for sample identification. When the
analysis is complete, the computer displays the results in any one of the various formats
21
selected by the user.
Particle size analyzers are automated and can typically analyze a sample in under 10
minutes with a minimal amount of technician time required. The test does provide an
accurate measure of the percent passing the 0.075 mm sieve. However, the equipment is
expensive and limited to the measurement of only small particle sizes. Typical equipment can
measure particles ranging in size from 0.1 to 2000 µm. This inability to handle larger
particles makes these systems unsuitable for determining the grading of complete aggregate
samples. However, this type of system could accurately determine the grading of the finer
portion of an aggregate grading.
Imaging Analysis
Imaging analysis to determine aggregate grading is now in the early stages of
development and holds promise for the future. Several different projects are under way to
develop the equipment and methods.
The imaging analysis process involves using a camera to “look” at the aggregate and
determine its grading. This process can be broken into five separate components: Image
generation and capture, and image coding, image reconstruction, image enhancement, and
image analysis. Image generation and capture is the process of creating the image with the
camera. This can sometimes be difficult because some information may be lost in the process
due to shadows, particle overlap, and other uncontrollable conditions. Image coding refers to
the techniques for storing the image. For example, most systems store the image such that
each screen element (pixel) is associated with an x-y coordinate system and brightness. Image
reconstruction is used to correct problems in the image such as distortion. Image

22
enhancement is used to make the digitized image clearer. For example, enhancement of the
image is used to separate particles that overlap. Image analysis is when the computer actually
determines the parameters sought. In this step, the two-dimensional shape of the particles is
determined (15).
In a practical sense, grading analysis by imaging techniques uses a photographic image
in a digitized form. The computer software uses the digitized photograph and chooses the
boundary lines for each aggregate particle according to the given algorithms. Having
identified the boundaries of each of the particles, the computer then determines the
two-dimensional shape of each particle from which it estimates the three-dimensional shape.
This projection of two dimensions into three is again dependent of software algorithms. Once
the number of particles and their sizes have been determined, the computer calculates the
grading of the sample in question.
Laboratory work in imaging analysis has been performed at Purdue University where the
techniques were used to differentiate siliceous and calcareous sands (15). Work has also been
performed at the University of Arkansas-Little Rock (UALR) on techniques to identify
aggregate shape and grading using imaging analysis. This work could hold promise for future
laboratory aggregate testing methods. The cost of a system similar to this is highly variable
with estimates ranging from $15,000 to $50,000.
One full-scale imaging analysis study worth noting is being field tested by Felix Alba,
Associates of Murray, Utah (16, 17). The study involves the real-time determination of
aggregate grading. Each of the cold feed bins is instrumented so that the grading from each is
constantly monitored using imaging analysis techniques. When the information is sent to the
23
control computer, it checks to make sure that the current blend of aggregates being used by
the plant will meet the job-mix-formula. If not, the computer can alter the feeding rate of any
of the cold feed bins in order to alter the total aggregate blend. In this fashion, the aggregate
blend being fed to the plant can be kept in constant control. The system has shown
remarkable promise and is expected to cost between $70,000 and $100,000. This process
offers obvious advantages. However, it is unable to determine the amount of dust present in
the aggregate.
Image analysis is a fully automated system that holds potential for determining
aggregate grading in a short period of time or on a continuous basis and all but eliminates the
need for a technician. However, the current cost is high and no accurate measure of the
percent passing the 0.075 mm sieve is possible. A considerable amount of work remains to be
done before this method can be used in practice.
Summary of Current Technologies
TABLE 2.2 shows a summary of the methods reviewed along with approximate costs
and availability. Each of the methods has advantages and disadvantages when compared to
the others. None of them can be employed immediately to accurately determine a complete
grading from the coarsest sieves through the 0.075 mm sieve.
TABLE 2.2 Summary of Current Technology

Technology Estimates Availability Estimates Cost
(Years)
Gradex 2000 Immediate $17,000-25,000
VDG-40 Video Grader Immediate $50,000
Full-Scale Particle Size Analyzer Unknown Unknown
24
Fractionating Water Column 1-2 $5,000-10,000

Laboratory Particle Size Analyzer Immediate $40,000
Laboratory Image Analysis 5-10 $15,000-50,000
Full Scale Imaging Analysis 1-5 $70,000-100,000
In reviewing the current technologies available for adaptability to grading testing, it is
obvious that none technology currently exists to fulfill the criteria of reducing test time,
reducing technician involvement, providing an accurate measure of the grading, including the
percent passing the 0.075 mm sieve, and being affordable. TABLE 2.3 lists each of the
methods reviewed and how well they meet the first three criteria. While several of the
systems are automated and reduce both the testing time and required technician time, they
suffer from the inability to measure the range of particle sizes used in HMA mixes. Some
technologies can measure larger particles while others measure the smaller sizes. None of the
systems currently offers adequate means for accurately determining the amount of material
passing the 0.075 mm sieve present in an aggregate sample.
TABLE 2.3 Capability Summary of the Technologies Reviewed

Technology Reduced Reduced Accurate Measure of %
Test Time Technician Time Passing 0.075 mm Sieve
Gradex 2000 Maybe Yes No
VDG-40 Video Grader No No No
Full-Scale Particle Size
Yes Yes No
Analyzer
Fractionating Water
Yes Yes No
Column
Laboratory Particle Size
Yes Yes No
Analyzer
Laboratory Image
Yes Yes No
Analysis
Full Scale Imaging
Yes Yes No
Analysis
25
Gradex 2000
The main problem with developing an automated grading analysis system seems to be in
determining how to get an accurate measure of the percent passing the 0.075 mm sieve within
the time allotted. If this requirement is waived, then the Gradex 2000 can quickly be adapted
for use. Contractors and laboratories could use the same sieves they currently employ and the
only added expense would be the Gradex 2000 and computer. If an accurate measure of the
percent passing the 0.075 mm sieve is required, the Gradex 2000 still presents the best
alternative. Although the actual testing time involved would not be reduced, the automation
makes it possible to reduce the technician time required.
VDG-40 Video Grader
The VDG-40 Video Grader is one of the only two technologies reviewed that offers the
potential ability to obtain an accurate measure of the percent passing the 0.075 mm sieve.
However, if the percent passing the 0.075 mm sieve is coating the coarser aggregate particles,
the ability of accurately determining the percent passing the 0.075 mm sieve material is
reduced. Additionally, this method does not reduce test time or technician time requirements
and is expensive.
Full-Scale Particle Size Analyzer
This method does have the potential to reduce test time and technician involvement, but
its cost is estimated to be high and it does not offer an accurate measure of the percent
passing the 0.075 mm sieve. In addition, it is limited in the particle size that can be evaluated.
26
Fractionating Water Column
At the present time, the method lacks the ability to produce an accurate measure of the
percent passing the 0.075 mm sieve. In addition, it is limited in the particle size that can be
evaluated.
Laboratory Particle Size Analyzer
At present the equipment is relatively expensive. This method is also limited in the
particle size that can be evaluated. Additionally, the equipment does not physically measure
the grading, but rather statistically estimates the grading. This equipment is also somewhat
sensitive and requires a clean, stable environment that may not be available in some HMA
production laboratories.
Imaging Analysis
Both laboratory and full-scale imaging analysis are in their infant states for grading
applications. The equipment is expensive and cannot offer an accurate measure of the percent
passing the 0.075 mm sieve. Research efforts are ongoing and should be closely monitored
for improvements.
Simulation of Aggregates Packing Process in PFC3d
Because different sizes and distributions of aggregate particles have different
sensitivities to the changes of VMA, their effects on volumetric properties should be analyzed
separately. For this analysis two assumptions are made:
1) All particles are assumed spherical.
2) The aggregate particles are uniformly distributed within the total volume.
27
The evaluation of aggregate packing begins with the comparison of two different states
of aggregate structures: State (I), a compacted structure of Aggregate 1 (Primary Structure),
and State (II), a compacted structure using both Aggregate 1 and Aggregate 2 (Secondary
Structure). For the purpose of simplicity, it’s assumed that newly added Aggregate 2 has the
size equal to or smaller than that of Aggregate 1. In the State (I) structure, the following
volumetric relation holds:
V1=Va1+Vv1 (1)
P1=Vv1/V1 = (V1-Va1)/V1 (2)
V1=Va1/(1-P1) (3)
Where V1 = total volume at state (I); Va1 = total aggregate volume at State (I); Vv1 = total air
void volume at State (I); and P1 = aggregate porosity or VMA at State (I).
In this study a gradation weighing factor fv was defined as the percentage of void change
by volume because of the addition of unit aggregate (Equation (4)), which is an indication of
the resulting volumetric impact when smaller aggregates are added into the structure (7).
fv=(Vv2-Vv1)/Va2 (4)
Vv1=V1P1 (5)
Vv2=V2P2 (6)
Where V2 = total volume at State (II); Va2 = the added aggregate volume at State (II)
compared with State (I); Vv2 = total air void volume at State (II); and P2 = aggregate porosity
or VMA at State (II).
The development of the fv parameter provides a straightforward method to estimate the
VMA of an aggregate structure. This method includes an iterative process starting from an
28
aggregate structure with only one size of aggregates. When smaller size aggregates are added
in, the porosity of the new structure can be determined as
P2 = Vv2/V2 = (fvVa2 + Vv1) / [V1 + (1+fv)Va2] (7)
Repeating the same procedure, smaller size aggregates are added into the mixed
aggregate structure and the corresponding fv values are determined. Once all fv values for each
sieve size are determined, Equation (8) will be used to predict the VMA (or porosity) of the
HMA mixtures:
∑𝑛 𝑓𝑣𝑖 𝑉𝑎𝑖
𝑃 = ∑𝑛 𝑖=1 (8)
𝑖=1(1+𝑓𝑣𝑖 )𝑉𝑎𝑖
Where fvi = fv value for the ith sieve size of the gradation; Vai = percentage of aggregate
retained on the ith sieve size; and P = porosity or VMA of the aggregate structure.
FIGURE 2.6(a) Mixture Contains Aggregate Larger Than 4.75mm (State I).
29
FIGURE 2.6(b) Mixture Contains Aggregate Larger Than 2.36 mm (State II).
FIGURE 2.6(c) Mixture Contains Aggregate Larger Than 1.18 mm (State III).
The determination of fv values in the DEM simulation follows an iteration process as follows:
1. Generate a basic model with only large aggregates, and calculate the volume and porosity
of this structure. The basic model, which includes aggregates retained on 4.75 mm sieve
size, is shown in FIGURE 2.6(a). The mixture reaches equilibrium after gravity and
loading cycles are applied. The contact force and unbalanced force are recorded at
equilibrium state.
30
2. Add an appropriate amount of finer aggregates into the structure, mix and compact the
structure to equilibrium state, and calculate the volume and porosity of the new structure.
FIGURE 2.6(b) and FIGURE 2.6(c) present the aggregate structures when particles of
sizes 2.36–4.75 mm and 1.18–2.36 mm are added into the system consecutively.
3. Use Equation (8) to calculate the fv value for this portion of aggregate.
Determination of Optimal Gradation
In this session, five aggregate gradations tabulated in TABLE 2.4, are adopted for the
evaluation. Consequently, there are five models conducted in PFC3d for different gradations.
Based on evaluation and calculation method described above, the value of fv and VMA are
calculated, shown in TABLE 2.5. As we can see, Model 1 with the smallest porosity has the
largest density. While largest density as sole criterion may not be rational, it is widely used in
experimental approaches, and therefore adopted for the computational method. Nevertheless,
other criteria will be identified and used for future work.
TABLE 2.4 Gradation of Five Models

Percentage passing (%)
ID 19 12.5 9.5 4.75 2.36 1.18 0.6 0.3 0.15 0.075
mm mm mm mm mm mm mm mm mm mm
1 100 93 80 49 32 23 17 12 8 5.2
2 100 93 83 55 35 24 18 13 7 5.4
3 100 98 85 53 32 20 14 10 8 6.1
4 100 99 87 51 29 20 14 10 6 4.6
5 100 97 82 52 30 17 11 8 6 5.1
31
TABLE 2.5 Calculated fv and Porosity of Five Models

fv Sieve size (mm) Porosity
ID 19 12.5 9.5 4.75 2.36 1.18 0.6 0.3 0.15 0.075 (%)
1 0.426 0.426 0.426 0.426 0.194 -0.396 -0.396 -0.419 -0.600 -1.000 16.3
2 0.432 0.432 0.432 0.432 0.198 -0.404 -0.404 -0.421 -0.600 -1.000 16.7
3 0.420 0.420 0.420 0.420 0.190 -0.388 -0.388 -0.417 -0.566 -1.000 16.5
4 0.441 0.441 0.441 0.441 0.199 -0.396 -0.396 -0.416 -0.600 -1.000 17.0
5 0.462 0.462 0.462 0.462 0.208 -0.404 -0.404 -0.415 -0.634 -1.000 17.7
The FHWA Advisory suggests using the 0.45 power curve to evaluate and adjust
aggregate gradations. The most convenient way to evaluate gradation is to plot the combined
gradation on a plot on which the FHWA 0.45 power gradation line has been graphed, shown
in FIGURE 2.7. Maximum density line is also plotted in FIGURE 2.7. Based on Goode and
Lufsey’s recommendation (6), this maximum density line is considered to be the most
realistic one for 3/4 inch nominal maximum aggregate size (NMAS). It’s very important to
have consensus among designers on this issue. A gradation in close proximity to the
maximum density line generally gives low VMA, resulting in a less durable mix that is more
sensitive to slight variations in asphalt content during production. To avoid these difficulties,
gradations are usually shifted away from the maximum density line. It’s clearly shown in
FIGURE 2.7 that gradation of Model 1 satisfies this criterion. To sum up, gradation of Model
1 has the lowest porosity and is adequately away from the maximum density line, so Model 1
is considered as the optimal gradation (Gradation 1).

32
FIGURE 2.7 Five Gradation Curve VS. Maximum Density Line.
The following asphalt mixture is based on selected gradation (Model 1). The gradation
has a NMAS of 3/4 inch and a primary control sieve size of 2.36 mm, and belongs to
coarse-graded group. As shown in FIGURE 2.8(a), orange balls represent aggregates and
smaller yellow balls represent asphalt binder. A complete consideration of fine particles in
DEM is almost impossible, because it not only significantly increase the computational time
and cost but also affects the system’s capability to reach equilibrium. Assuming that smaller
particles mainly fill in the voids created by coarse aggregates, only the particles retained on
1.18 mm sieve were simulated in a cylinder (rad = 40 mm, height = 160 mm) under a
unidirectional gravitational force. Then, balls are compacted by the assigned gravitational
force to their stable position. An equilibrium state is reached when the unbalanced force
reaches 1% of the maximum contact force. Maximum contact force = 9.837×108 Newton, is
shown in FIGURE 2.8(b), mean unbalanced force = 9.468×106Newton. The equilibrium state
33
is reached in FIGURE 2.8(b) since unbalanced force/ maximum contact force = 0.009625 ≈
0.01.
Based on findings by Shen and Yu (18), the contact stiffness kn of 30 GN/m, the rigid
wall stiffness of 300 GN/m, and the inter-aggregate contact friction coefficient of μ=0.5 are
calibrated parameters that can reasonably simulate the volumetric properties of the aggregate
structure using the spherical DEM model. While these parameters are varying with different
types of aggregate, it is acceptable for this qualitative evaluation (mix design for a selected
aggregates).
FIGURE 2.8(a) Mixture Model with Selected Gradation Aggregates and Asphalt Binder.
34
Maximum Contact Force

= 9.837x108 N
FIGURE 2.8(b) View of Contact Force of the Mixture at Equilibrium State.
Summary
Gradation is determined based on a simulated packing process. The evaluation of
aggregate packing begins with the comparison of two different states of aggregate structures.
We assume the newly added aggregates are always equal in size to or smaller than aggregates
in former state. After gravity and loading cycles are applied, the contact force and unbalanced
force were recorded. Then finer aggregates were added into the structure. After mixing and
compacting the structure into equilibrium, volume and porosity of the new structure were
calculated. By comparing five digital models, an optimal model which has the lowest
porosity was selected.

35
3 DETERMINATION OF ASPHALT CONTENT
Literature Review
The binder content in an asphalt mixture is one of the critical factors determining the
quality of the mixture. Using a mixture with too much binder can lead to pavement rutting
and bleeding, whereas having a mixture with too little binder can lead to pavement durability
problems, including raveling and cracking. The Virginia Department of Transportation
(VDOT) now accepts hot mix asphalts (HMA) based on their volumetric properties (voids in
the mineral aggregate (VMA) and percent air voids) and binder content (the mass of binder
expressed as a percentage of the total mass of the mix), so accurate determination of the
binder content is essential. Historically, the binder content of a mixture has most commonly
been determined by one of several extraction methods in which a solvent is used to remove
binder from a known mass of HMA. The main advantages of these methods are as follows: (a)
no calibration factor needs to be determined and (b) the properties of the binder and
aggregate recovered from the HMA may be tested after extraction. The main limitations
associated with solvent extraction methods are related to the length of the test time, possible
alteration of binder properties and the hazardous nature of the solvents. The ignition oven is
one alternate to solvent extraction. This oven uses high temperature to burn the asphalt off the
aggregate. The procedure terminates when the weight of the sample stabilizes, indicating that
there is no more binder to ignite. In 1999, VDOT began using the ignition oven on a few trial
projects; usage increased substantially after that because of the advantages of the method.
Those advantages include the relatively fast test time and avoidance of the use of hazardous
chemicals.
36
The ignition method has generally worked well on HMAs, except those containing
dolomitic aggregates and some limestone sources containing pyrite. These types of
aggregates, however, are prevalent in some areas of the state, especially dolomites. With
these “problematic aggregates,” the high temperatures experienced during ignition cause
chemical changes within the aggregates, which result in variable mass loss. In some cases,
the sample weight does not stabilize at or below the threshold level, and the ignition test
continues for excessive periods of time. This variability in the aggregate mass loss causes
difficulties in determining a consistently accurate calibration (or correction) factor and in
obtaining consistent and repeatable test results from multiple samples of the same material.
Because of this problem, ignition ovens are not used at all in some parts of the state or are
only used for those mixes that do not contain dolomite.
Binder Content Determination Methods
Historically, the binder content has most commonly been determined by one of the
several extraction methods in which a solvent is used to remove the binder from a known
mass of HMA. The solvents used were typically chlorinated solvents such as methylene
chloride, trichloroethylene or trichloroethane. The main advantages of these methods are: (a)
no calibration factor needs to be determined and (b) the properties of the binder and
aggregate extracted from the HMA may be tested after extraction.
Another method for binder content determination is the nuclear gauge method. The
nuclear gauge uses radiation to detect the presence of hydrogen in the binder. While relatively
quick and easy to conduct, this test does not offer the possibility of testing the binder and
37
aggregate since they are not separated in the process. Another disadvantage of using the
nuclear gauge is the requirement for obtaining a Nuclear Regulatory Commission license.
The nuclear asphalt content gauge requires calibration for each mixture tested, which can be
time consuming.
In the 1980s, the conventionally used chlorinated solvents for asphalt extractions began
to fall into disfavor. Due to concerns over potential carcinogenicity and environmental
impacts, some solvents became difficult to obtain and dispose of after use. Costs also
increased greatly. The asphalt industry, which relied heavily on solvent extractions to
determine binder contents, needed to find alternatives to chlorinated solvents. In 1990, the
National Center for Asphalt Technology (NCAT) began development of an ignition oven to
burn off the asphalt from a sample of hot mix. Eventually, this effort led to the development
of the NCAT ignition oven (19). This developmental effort built on work from the late 1960s
by Antrim and Busching showed that the asphalt cement in a mixture could be completely
combusted by heating to 843oC (1550oF) with an excess of oxygen. At this high temperature,
some aggregate mass was also lost; more mass loss occurred with limestone than with gneiss
in the early work.
NCAT built upon this earlier work, but substituted a muffle furnace for the special
furnace Antrim and Busching had used. NCAT also used a lower burn temperature (593oC or
1100oF) to reduce the aggregate mass loss. NCAT was motivated to reduce the test time, so
they experimented with different ignition temperatures and burn times to optimize the two
factors (19).
Since the method does not use potentially hazardous solvents, it has become quite
38
popular across the country. The oven also does not require nuclear regulatory licensing, so it
is easier to implement than the nuclear asphalt content gauge. In addition, the aggregate
remaining after ignition is generally (but not always) suitable for gradation analysis, which
cannot be conducted with the nuclear gauge. At least one study (20) has demonstrated that the
ignition oven is potentially far less expensive to operate than the nuclear gauge or various
solvent extractions. Many states have adopted the oven for use and find it similar to or fewer
variables than alternate binder content determination methods. (21, 22, 23, 24, 25, 26)
Currently, the ignition oven test is described in both AASHTO and ASTM standards.
The ignition oven uses high temperature to burn the asphalt off the aggregate; the typical test
is conducted at 538oC or 1000oF. (ASTM specifies a test temperature of 540°C or 1000°F in
keeping with their policy of using rationalized unit conversion, so-called “hard conversions.”)
During a typical ignition test, the sample is heated continuously until three consecutive
readings of the mass of the sample, taken at one minute intervals, do not change by more than
a specified threshold level, typically 0.01%. To facilitate this determination, many ovens have
internal balances to provide continuous readout of the sample weight. All calculations of the
asphalt content must account for the calibration factor, which is the difference between the
known binder content and ignition oven test result. The calibration factor should be
determined independently for each mix. The results of two properly conducted tests by the
same operator on the same size samples of HMA should not differ by more than 0.196 %.
If the calibration factor is higher than 1.0%, as is the case for some high mass-loss
aggregates, the test temperature should be lowered to 482oC (900oF). The calibration factors
for most limestone aggregates are typically less than 0.2%. For some dolomitic aggregates,
39
however, the factor can be over 2%. (According to AASHTO T308 (27), a correction factor
greater than 1.0% is considered unusually high.) The correction factor has been observed to
decrease when the ignition temperature was decreased from 538ºC (1000ºF) to 482ºC (900ºF).
(28, 29, 30)
In a conventional ignition oven, the furnace chamber is heated with a radiant heat source.
However, there is an alternate type of ignition oven, called the infrared oven. In this oven, an
infrared heating element is used to heat the sample. Hurley, Prowell (31) and Prowell (32)
compared the results obtained using both types of ovens. They found the correction factor
due to aggregate weight loss for the infrared oven to be generally smaller than that for the
conventional oven. They reported the results from the infrared and conventional oven were
not statistically different. Research conducted by Williams and Hall (33) reached a similar
conclusion that the accuracy of both types of ovens is statistically similar. However, they did
not observe that infrared heating limited the over-heating of the aggregate. They discovered
that the peak temperature recorded by the infrared oven was often greater than the peak
temperature recorded by the conventional oven for the same type and size of sample.
Changes in aggregate properties resulting from binder ignition were observed by
Mallick et al. (34). During a study involving different types of aggregate (granite, limerock,
gravel and traprock), the authors found that for the same duration of test, the ignition process
has a greater effect on the properties of aggregate when an asphalt - aggregate mix (rather
than an unbound aggregate sample) is placed in a preheated ignition oven.
Similar research was conducted by Prowell and Carter (35). In their study, the impact of
the ignition test on the aggregate properties was evaluated. In six of ten cases, they noticed
40
significant differences between the bulk specific gravity before and after the test. They also
noted that aggregate recovered using the ignition furnace appears to be unsuitable for sand
equivalency testing. The results of fine aggregate angularity testing were significantly
different between the virgin and burnt samples in three of ten cases. They also observed that
accurate results may be obtained for gradation analysis and flat and elongated particle
measurements performed on aggregates recovered from the ignition oven test.
During a typical ignition test, the pre-set oven temperature is often exceeded due to the binder
ignition. Rogers et al. (36) conducted a study with a modified ignition oven, where a
thermo-kinetics infrared thermometer was placed in the oven door. The temperature of the
flame created during the ignition process was measured. They noticed up to a 200ºC
difference between the flame and oven temperatures. Temperatures above 565ºC were
consistently measured in the oven, but in no case did the temperature exceed 750ºC. In
addition, differences in flame temperature due to differences in binder content were observed.
Ignition Method Test Procedures
Currently, the ASTM D6307 "Standard test method for asphalt content of hot mix
asphalt by the ignition method" (37), and the "Virginia test method for determination of
asphalt content from asphalt paving mixtures by the ignition method" (30) are the two widely
known test procedures of ignition tests. In both procedures, the test temperature is specified at
538 ºC for HMA and 578 ºC for aggregate only calibration, and the sample size is determined
by the nominal maximum size (defined as one size smaller than the: first sieve that retains
aggregate) of the aggregate according to TABLE 3.1.

41
In the ASTM procedure, two samples are required per calibration, no specific guideline
of the calibration of RAP mixes is provided. In the VDOT procedure, a total of six samples
(two at design asphalt content, two at 0.5 percent below, arid two at 0.5 percent above the
design asphalt content) are required for mixture calibration. Four samples are required for
aggregate only calibration. The method provides a guideline for the calibration of RAP
mixtures.
TABLE 3.1 Minimum Sample Weight

Nominal maximum aggregate size Minimum sample weight (g)
(mm)
4.75 (No.4) 1200
9.5 (3/8 in.) 1200
12.5 (1/2 in.) 1500
19.0 (3/4 in.) 2000
25.0 (1 in.) 3000
37.5 (1 ½ in.) 4000
Procedure of Determination of Asphalt Content Using DEM
In addition to gradation, the content of asphalt is another important factor in the design
of asphalt concrete. Ten sample mixtures were prepared with same gradation but different
asphalt content, as summarized in TABLE 3.2 and TABLE 3.3. The asphalt content ranges
from 3.9% to 5.9%. In a common sense, the asphalt mixture that reaches highest peak
strength is the strongest. As compressive load is applied, mixtures become weaker, and cracks
initiate and propagate.
Here the contacts between aggregates and asphalt binder are considered as the
evaluation factor to determine whether the mixture is damaged because of cracking. A simple
42
visco-elastic model and parallel bonds were applied on these contacts. The type of asphalt
binder used in the simulations is PG 64-28. In order to simulate the properties of this type of
asphalt, some important parameters were selected and shown in TABLE 3.4. The following
procedures were followed to determine the exact optimal asphalt content value based on
DEM method.
1. Calculate gradation and surface area. Since aggregates in simulations are represented by
balls with regular shape and sizes, gradation and surface area of the mixture could be
easily calculated.
2. Record contact number between aggregates and asphalt binder. TABLE 3.3 shows contact
information in ten mixtures at initial state and after load applied, respectively. The
information could be used to determine whether the mixture is cracked. It is assumed that
the mixture will stop working due to cracking when the contact number decreases to
50%~60% of that at the initial state.
3. Compare peak strengths when the mixtures cracked. The model with the highest strength
was considered as the optimum mix design.
TABLE 3.2 Gradation Information

Sieve size (mm) 19 12.5 9.5 4.75 2.36 1.18 0.6 0.3 0.15 0.075
Percentage passing (%) 100 93 80 49 32 23 17 12 8 5.2
TABLE 3.3 Asphalt Content and Contact Information

Mixture Number of Contacts Number of Contacts Asphalt Peak strength
ID at initial state after compression content
1 19746 11774 5.3% 84.81 MPa
2 18400 10428 4.6% 47.82 MPa
3 17632 9628 3.9% 30.87 MPa
43
4 19156 11034 5.0% 72.01 MPa

5 17980 10019 4.3% 39.55 MPa
6 20246 12037 5.7% 58.36 MPa
7 20499 11558 5.9% 38.97 MPa
8 20371 11476 5.8% 40.01 MPa
9 20135 11377 5.6% 69.82 MPa
10 19960 11292 5.5% 79.44 MPa
TABLE 3.4 Contact Properties Applied in Simulations

Properties for viscoelastic
Properties for parallel bonds:
contact model:
15 -3
Normal stiffness = 30*10 Nm ; Viscoelastic normal stiffness
15 -3
Shear stiffness = 30*10 Nm ; = 0.36*106 N/m
Normal strength = 190 MPa; Viscoelastic shear stiffness
Shear strength = 190MPa; = 0.12*106 N/m
Radius multiplier = real parallel bond radius/radius
-9 Shear viscosity = 0.414*106 N/m
= 30x10
Results and Analysis
Peak strengths were recorded in TABLE 3.3. From TABLE 3.3, Mixture 1 which has
asphalt content equal to 5.3% is the optimal one. The Stress-Strain figure of Mixture 1 is
shown in FIGURE 3.1. A relationship between the asphalt content and the peak strengths of
ten mixtures is shown in FIGURE 3.2; Mixture 1 with the asphalt content of 5.3% locates at
the turning point on the curve.

44
84.81
Stress (MPa) 90
80
70
60
50
40
30
20
10
0 Strain (%)
0 2 4 6 8 10
FIGURE 3.1 Stress VS. Strain of Mixture 1
Peak Strength (MPa)

90 84.81
80 79.44
70 72.01 69.82
60 58.36
50 47.82
40.01
40 39.55
38.97
30 30.87
20
10
0
Asphalt
3.50% 4.00% 4.50% 5.00% 5.50% 6.00% 6.50% content
FIGURE 3.2 Asphalt Content VS. Peak Strength.

45
Calculation of Film Thickness by Using Selected Gradation and Asphalt Content
Calculation of Specific Surface Area Factor.
The Hveem–Edwards’s method (38) is used to estimate surface area. The specific
surface area of aggregate is determined from sieve analysis based on the total percentage
passing a stipulated group of sieves as
SA = 0.01SPP i x CP i (9)
where SA = specific surface area of the aggregate in m2/kg (ft2/lb); SPPi = percent of
aggregate by weight passing the ith sieve; CPi = surface area factor (a coefficient for
“passing”) in m2/kg (ft2/lb); i = sieve index (i = 0,1…N); N = 9 for the standard set (ASTM
C136-06) with the following diameters: D0=19.0 mm, D1=12.5 mm, D2 = 9.50 mm, D3 = 4.75
mm, D4 = 2.36 mm, D5 = 1.18 mm, D6 = 0.600 mm, D7 = 0.300 mm, D8= 0.150 mm, and D9
= 0.075 mm.
The surface area factors CPi have been tabulated but the background research data on
them are not available. To derive them, one can start from specific surface of aggregate
fraction that passed the (i -1)th sieve and retained in the ith sieve. By definition, the specific
surface is a surface-to-weight ratio. The specific surface of particles contained in the ith sieve
can be expressed as the average specific surface of the endpoints of the sieve range:
CRi = 3 / (gi-1 Di-1) +3 / (gi Di) (10)
where CRi = specific surface of the particles with diameter in the range (Di, Di-1), retained in
the ith sieve (a coefficient for “retained”), m2/kg (ft2/lb); gi= effective density of aggregate in
kg/m3; Gsbi = bulk specific gravity of aggregates contained in the ith sieve; gw= density of
46
water; Ksph = sphericity coefficient = (surface area of spherical shape / surface area of
irregular shape): 1 for sphere, 0.806 for cube, 0.874 for cylinder (h=d), etc. Sphericity values
relate the surface area of the non-spherical shape to a spherical shape based on the
assumption that both particles have the same volume.
Once the relationship between the specific surface of particles retained in the sieve and
the sieve diameter is established, the specific surface area of the aggregate based on the
percentage retained in a group of sieves can be expressed as
SA = 0.01SPR i x CR i (11)
Where PRi = percent of aggregate by weight retained in the ith sieve; i = sieve index (i = 1,
2…N+1); N = 9 for the same standard set of ASTM sieves.
Since the percent of aggregate retained in the ith sieve, PRi, is related to the percent
passing the (i-1)th and ith sieves by:
PR i = PP i - 1 – PP i (12)
Equation (11) can be rewritten as:
SA = 0.01[(PP0 – PP1) CR1 + (PP1 – PP2) CR2 + … (PPN – PPN+1) CRN+1] (13)
Which is same as:
SA = 0.01[PP0 CR1 + PP1 (CR2 – CR1) + PP2 (CR3 – CR2) + …PPN (CRN+1 – CRN)] (14)
Then the unknown surface area factors for the percentage passing the sieves (CPi), can
be related with specific surface of particles retained (CRi), as:
CP0 = CR1 CPi = CRi+1 – CRi (i = 1, 2…N), (15)
Substituting Equation (10), we finally get the presumable equation for conventional
surface area factors:

47
CP0 = 3 / (g0 D0) + 3 / (g1 D1) CPi = 3 / (gi+1 Di+1) – 3 / (gi-1 Di-1), (i = 1, 2…N) (16)
TABLE 3.5 lists the surface area factors calculated using Equation (16). These results
can be used to calculate film thickness. Combine with gradation listed in TABLE 3.2, the
specific surface area has been calculated to be 5.92 m2/kg.
TABLE 3.5 Calculated and Surface Area Factors

Sieve Sieve size Effective density of aggregate Calculated surface
index, i (mm), Di retained, x 1000 kg/m3 area factor, m2/kg
0 19 2.34 0.410
1 12.5 2.34 0.410
2 9.5 2.34 0.415
3 4.75 2.34 0.418
4 2.36 2.34 0.827
5 1.18 2.34 1.694
6 0.6 2.34 3.012
7 0.3 2.44 6.160
8 0.15 2.44 12.295
9 0.075 2.44 32.787
Calculation of Film Thickness
The asphalt binder specific gravity is 1.03 g/mL, selected asphalt content is 5.3%, and
the bulk density of the mixture is 2.340 g/mL. Film thickness can be calculated by Equation
(17).
TF=1000×Vasp / (SA×W) (17)
Where,
TF = Average film thickness (micrometer);
Vasp = Effective volume of asphalt binder (liters);

48
SA = Surface area of the aggregate (m2 per kg of aggregate);
W = Weight of aggregate (kg).
Let’s take a one liter sample as example:
Total volume of asphalt binder = total weight × asphalt content / asphalt binder specific
gravity = 2340 g×0.053/1.03 = 120.4 mL
If we ignore absorbed asphalt, Vasp = 120.4 mL = 0.1204 L.
W = (1-0.053) ×2340 = 2215.98 g = 2.21598 kg
TF = 1000×0.1204 / (5.92×2.21598) = 9.2 micrometers
So the optimal film thickness calculated through determined gradation and asphalt
content is 9.2 micrometers.
Summary
In this session, a computational method is developed to aid mix design, especially for
determine the asphalt content. Discrete element method (DEM) was used to simulate the
formation of skeleton and voids structures of asphalt concrete of different asphalt contents.
This method takes the mechanical properties or performance of the mixture into consideration,
such as inter-aggregate contacts and local stability. A simple visco-elastic model was applied
to model the contacts between asphalt binder and aggregates. The surface texture of an
aggregate particle can be taken into consideration in the inter-particle contact model. For a
certain air void content, the mastics volume or the binder volume or the asphalt content can
be determined via a digital compression test. The surface area of all the aggregates and the
film thickness can be then calculated.
Ten digital mixtures were built in software PFC3d, and the asphalt content ranges from
49
3.9% to 5.9%. During compressive loading being applied, the number of contacts between
aggregates and asphalt binder were recorded. This number was used as a judgment to tell
whether the mixtures were damaged due to cracking. The mixture that reached the highest
peak strength was considered as optimal. Therefore, optimal asphalt content determined by
this process was 5.3%. The procedure described in this study helps to improve the
experimental method to determine the optimum asphalt content by adding the capability of
internal deformation mechanism (breaking of bonds and loss of contacts).

50
4 EFFECT OF FILM THICKNESS ON ASPHALT MIXTURE
Literature Review
Asphalt mixture is a composite material which consists of interspersed aggregates,
asphalt binder and air voids. The constitutive behavior of the material depends largely on the
interaction between the aggregates and asphalt binder. The aggregate skeleton determines the
load carrying mechanism and the asphalt binder serves as an adherent which enables the mix
to hold tensile and shear stresses. A comprehensive understanding of the mechanical behavior
of aggregate-binder system is critical for research in the area of the deformation mechanism
of asphalt mixtures. To study the constitutive behavior of an aggregate-binder system, the
contact between aggregates coated with asphalt binder is the key problem to solve. A contact
model which properly shows the relationship between the contact force and relative
movement is needed. Combined with the mechanical model, numerical tools could be
incorporated into the study, such as the Finite Element Method (FEM) and Discrete Element
Method (DEM).
Both of the FEM and the DEM are utilized in the simulation of mechanical materials.
The former is based on the continuum approach and has been successful in capturing the
stress-strain distribution within the asphalt mixtures and its effect on the stiffness anisotropy.
The latter based on the discrete approach could analyze the individual characteristic effects of
each component in the mixture, such as the influence of shape, gradation and relative
slippage of aggregates. The proper representation of the asphalt mixture internal structure is
possible through the use of image analysis techniques. These techniques are utilized to
accurately capture the actual microstructure of the asphalt mixture. Such images offer a
51
realistic representation of the internal structure of asphalt mixtures and could be used to
analyze the deformation of the composite material. Given the proper internal structure and
constitutive contact models, the simulation could avoid huge amount of sample preparation
work in lab and provide reliable results in the research of the asphalt mixtures.
Contact Model
Asphalt mixture can be considered to be three-phase composites consisting of asphalt,
aggregates and air voids. Because of the complex heterogeneous nature of the material, the
macro load carrying behavior depends on many micro-phenomena that occur at the
aggregate/binder level. Some important micro behaviors are related to binder properties
including volume percentage, elastic moduli, time-dependent response, aging hardening,
micro-cracking, and de-bonding from aggregates. Other microstructural features include
aggregate size, shape, texture and packing geometry. Because of these issues it appears that a
micromechanical model would be best suited to properly simulate such a material.
Furthermore, micromechanics offer the possibility to more accurately predict asphalt
performance and to relate such behavior to particular mix parameters such as binder
properties, aggregate gradation, and sample compaction.
Van der Poel (1954) attempted to model the behavior of asphalt mixture by calculating
the rigidity of a concentrated solution of elastic spheres in an elastic medium, using a method
developed for dilute dispersions by Frohlich and Sack (39). The model was reported to give
reasonable agreement with experimental measurements of dynamic mix stiffness up to a
volume fraction of 60 % aggregate.

52
Hills (1973) attempted to develop theoretical models for the long time creep behavior of
asphalt mixes (40). His modeling approach described the internal structure in terms of a
characteristic asphalt thickness, and the evolution of this state variable as a function of the
macroscopic straining of the material. The macroscopic strain of a mix was assumed to be
accommodated, on a microscopic scale, by displacements of adjacent aggregate particles in
both shear and compression. These displacements were assumed to be independent of each
other. It was further postulated that the macroscopic strain of the mix was uniquely related to
the shear displacements of adjacent aggregate particles. The microscopic shear resistance of
the asphalt was considered to be controlled by the thickness of the asphalt. The effect of
compression was taken into account by changing the thickness of the asphalt. The asphalt
was modeled as an incompressible, linear viscous fluid. The general forms of the constitutive
equations were given in terms of the ‘stiffness’ of asphalt, Sbit, in the region Sbit < 5 MPa.
Hills' model was based on an empirical estimation of the evolution of film thickness under
deformation rather than by analysis of the deformation of the film. The model depended on
curve fitting to experimental data to obtain numerical data.
Cheuang at al (1997) used the isolated model for micromechanical modeling of the stage
I sintering process, when the dominant mechanism of contact deformation is rate-independent
plasticity or power law creep (41). A compatible strain rate field is assumed for describing the
motion of individual particles. The constraints on the motion of each contact are determined
by analysis of the corresponding deformation mechanism. Using kinematics bounds, the
macroscopic deformation behavior can then be estimated by integrating the microscopic
deformation behavior, assuming that individual contacts deform in isolation. They assumed
53
that all of the aggregate particles are separated by thin films of asphalt; the constraints on the
relative motion of the particles will largely be determined by the deformation behavior of the
bitumen films until they become so thin that their stiffness becomes comparable with the
stiffness of the aggregate particles. They also assumed that asphalt existing in larger
quantities of particles within the random aggregate skeleton will have only secondary effects
on the deformation behavior of the mix. The assumption of isolated contacts means asphalt
displaced by the deformation of each contact flows into an air void rather than into an
adjacent contact. It is therefore necessary for the volume of asphalt displaced to be small
compared with the volume of voids. Analysis based on the assumption of isolated contact
behavior provides a theoretical limiting case solution regarding the macroscopic behavior of
the idealized mix, corresponding to the case when microscopic deformation occurs at all
distributed thin film bitumen contacts.
Uddin (1998) presented a micromechanical analysis method for calculating the creep
compliance of asphaltic mixes on a microscopic level using laboratory viscoelastic
characterization of the binder and elastic material properties of the aggregates at a given
temperature (42). The micromechanical model is based upon the “method of cells” (MOC)
developed to predict viscoelastic response of resin matrix composites. The properties of the
aggregate are assumed to be linear elastic and can be described by the elastic constitutive
relationship. A time-stepping algorithm was developed for a viscoelastic material with a
Prony series representation of the time-dependent properties. And the micromechanical
model was incorporated in a microcomputer program which calculates the viscoelastic
response of the mix and predicts the mix stiffness. It is reported that reasonably good
54
agreement is found between the predicted modulus and measured modulus of the mix if
proper percent air voids are considered in the micromechanical model.
Zhong and Chang (1999) applied a micromechanics approach to investigate the
inter-particle behavior of two particles connected by binder (43). The model is based on the
premises that the inter-particle binder initially contains micro-cracks. As a result of external
loading, these micro-cracks propagate and grow. Thus, binders are weakened and fail. Theory
of fracture mechanics was employed to model the propagation and growth of the
micro-cracks. The contact law is then incorporated in the analysis for the overall damage
behavior of material using a discrete element method. Using this model, the stress-strain
behaviors under uniaxial and biaxial conditions were simulated. And it is reported that a
reasonable agreement is found between the predictions and experimental results.
Levenberg and Uzan (2003) developed a triaxial cross-anisotropic viscoelastic-
viscoplastic constitutive model for asphalt aggregate mixtures for the small-strain domain
(44). The model follows the concept of strain decomposition by separately analyzing the
viscoelastic and viscoplastic strain components. In order to calibrate the model, advanced
triaxial testing was employed which included both hydrostatic and uniaxial creep and
recovery cycles. The test data is presented and analyzed along with the derivation of all
model parameters. It is reported that the proposed model correlates extremely well with the
entire test-data in both axial and radial directions.
Mazzotti and Savoia (2003) proposed an isotropic model for creep damage of concrete
under uniaxial compression, where the combined effect of nonlinear viscous strain evolution
and crack nucleation and propagation at high stress levels are considered (45). Strain splitting
55
assumption is used for creep and damage contributions. Creep is modeled by a modified
version of solidification theory. In the modeling of damage of concrete, a damage index based
on positive strains is introduced. In particular cases, the proposed model reduces to linear
viscoelasticity for long time low stress levels whereas, for very high stresses, tertiary creep
causing failure at a finite time can be described. The effect of strength variation with time is
also included. The model is numerically implemented to perform time integration of
nonlinear equations by means of a modified version of exponential algorithm. The model was
validated through comparison with experimental results and numerical examples are also
presented, where the roles of concrete ageing and strength variation with time are
investigated.
Huang et al (2004) presented a temperature dependent viscoplastic model that
incorporated temperature and loading rate into the Hierarchical Single Surface plasticity
based model (46). The model was able to reflect the nonlinear plasticity, as well as the
temperature and loading rate dependencies of the asphalt mixtures. Triaxial compression,
triaxial extension, and axial creep tests at three temperatures 28°C, 40°C and 60°C were
performed to calibrate the material properties. And they proposed an algorithm to compare
the numerical analysis obtained from the model and experimental results. It is reported a
reasonable agreement was observed from the back calculation and the experimental results.
Chehab and Kim (2005) reported a viscoelastic-plastic continuum damage (VEPCD)
model, which is developed and validated under the auspices of the National Cooperative
Highway Research Program 9-19 project, entitled “Advanced Mixture Characterization for
Superpave Support and Performance Models Management” (47). The VEPCD model was
56
able to characterize the viscoelastic and viscoplastic responses of asphalt concrete in addition
to micro-cracking. They validated the model under thermal loading conditions that are
distinctively different from the mechanical loading conditions used in model development
and calibration. Measured responses and fracture parameters from thermal strain restrained
specimen tensile strength tests were conducted to compare with those predicted using the
VEPCD model, the viscoelastic continuum damage model, and the linear viscoelastic model.
It is reported that the ability of the VEPCD model to accurately characterize the tensile
behavior of asphalt concrete under thermally induced loading was confirmed.
Numerical Methods
Sepehr et al. (1994) conducted a finite element analysis of a pavement structure with the
asphalt layer represented by an idealized microstructure (48). Air voids were simulated by
imposing small stiffness values to some of the elements. This analysis revealed significant
information about the influence of the micro-structural properties on the macroscopic
behavior of the pavement. It was observed that increasing the air voids from 1% to 5%
resulted in 1.2% increase in pavement surface deflection. It was also shown that reducing the
asphalt binder stiffness from 1000MPa to 250MPa resulted in 2.25% increase in the
deflection. The influence of aggregate shape on pavement behavior was also investigated and
2% decrease in surface deflection was reported when sphere shaped aggregates were replaced
by aggregates of sharp edges.
Bahia et al. (1999) conducted finite element analyses using an idealized internal
structure of asphalt mixture, whereby aggregates were represented by circular objects (49).
57
Binder and aggregate elements were considered linear elastic materials. The objective of this
analysis was to evaluate the deformation and strain distribution in asphalt mixes and its
relation to the nonlinear mix behavior. It was observed that when applying 1% strain, the
binder shear strain could be as high as 46%. Therefore, they suggested that within an actual
mixture, a wider range of film thickness would exist and a wider range of strain magnitude
could be realized within the binder domain.
Abbas et al. (2004) incorporated a nonlinear viscoelastic material model in a FEM to
analyze the asphalt concrete microstructure (50). The viscoelastic behavior of the asphalt
mastic was defined using a mechanistic model, which was numerically solved using a
convolution integral approach. To account for the asphalt binder non-linearity, the mechanical
parameters were updated during the analysis according to the strain level within each element.
The analysis was also used to analytically study some of the discrepancies observed between
laboratory measurements of the asphalt mixture E* and G*.
Saadeh et al. (2003) extended the comparison between E* and G* to include both
experimental and numerical measurements (51). They studied the main factors causing
discrepancies between the axial and shear tests, which include the type of loading (stress
versus strain-controlled), direction and reversal of principal stresses, and stress and strain
distribution within the specimen. Using finite element simulations of these tests, they
reported a Poisson’s ratio between 1.0 and 1.5, depending on the test frequency, which is
relatively smaller than that obtained experimentally. Therefore, they concluded that the FEM
only captures the effect of the stress and strain distribution and does not capture the effect of
the rest of these factors. It should be noted that the finite element analyses they considered
58
included linear viscoelastic mastics and elastic aggregates.
Rothenburg et al. (1992) presented a micromechanical model of asphalt mixture based
on discrete element techniques, whereby aggregates were modeled as polygons and the
inter-particle forces were described using a mechanistic viscoelastic model (52). The
objective of this work was to relate the asphalt mixture internal structure to its deformation
characteristics. Particles were considered as plane elements that interact by means of contact
forces. Simulation of angular aggregate particles with an arbitrary gradation was described by
a particle generation algorithm. It was found that the complex performance of the granular
matrix is the main reason for the nonlinear trends in mechanical response of asphalt mixture,
and to a large degree its susceptibility to rutting in field situations. Simulations indicated that
mix strength increased with the increase in the fraction of contacts and that at least 30% of
contacts should be bound to get nonzero strength. At least 20% of the contacts must be
nonzero friction to develop shear resistance.
Chang and Meegoda (1997) used the DEM to describe different types of
aggregate-aggregate and asphalt-aggregate contacts (53). They utilized mechanistic models to
simulate the viscoelastic behavior of asphalt binder. They incorporated the Mohr-Coulomb
failure criterion to account for the sliding of asphalt-coated particles due to rotation. Both
macroscopic and microscopic behaviors were monitored during simulation. It was reported
that the stress-strain behavior was properly captured even in the post-peak region.
Buttlar and You (2001) used DEM to simulate the behavior of asphalt mixture in the
indirect tension test (IDT) (54). They described the internal structure using clusters of circular
particles. A linear contact model along with a bonding and sliding capabilities was used to
59
define the particles’ interaction. Their observations pointed out the significant contribution of
aggregate interaction in accurately simulating the stiffness of asphalt mixtures, and hence,
suggested the need for utilizing realistic aggregate shapes in microstructure models.
Given the proper mechanical contact model, the numerical simulation could avoid huge
amount of sample preparation work in the lab.
Build a Digital Model by Using DEM
Several methodologies were used to represent particle geometry and the mechanical
interaction between particles. For example, particle geometry has been represented using
image based models (54, 55), and elastic, visco-elastic and cohesive models have been used
to represent mechanical behaviors. How the user-defined contact model especially simple
visco-elastic model represented particular materials and contact properties in PFC3d were
introduced in following sections.
User Defined Contact Model in PFC3d
FIGURE 4.1 plots the flowchart of DEM simulation using User-defined contact model
(UDM) in PFC3d (56). The input is done via the PFC interface, but the UDM dictates the
behavior of all interactions within specific models. The UDM calculates the
fore-displacement law each timestep. During each timestep relative motion of particles and
contact forces are extracted from different models and applied in these models to dictate the
force-displacement behavior between two particles in contact (57).
The constitutive model used in PFC3d comprises three parts:
 a contact model;
60
 a bond model;
 a friction slip model.
FIGURE 4.1 Flowchart of a PFC3D Program Code Using UDM (56)
The contact model is described in terms of a normal secant contact stiffness Kn and a
shear tangent contact stiffness Ks between the two contacting entities either wall-particle or
particle-particle. Note that the in-plane and out-of-plane shear parameters are assumed to be
the same. PFC3d allows particles to be bonded together at contacts. If the value of the tensile
normal contact force equals or exceeds the normal contact bond strength, the bond breaks,
and both the normal and shear contact forces are set to zero. If the value of the shear contact
force equals or exceeds the shear contact bond strength, the bond breaks and the slip model
are activated. Parameters which were defined in PFC3d are the normal and shear contact
stiffness Kn, Ks, the normal and shear bond strengths Pbn, Pbs and the coefficient of friction
61
between contacting particles.
Discrete element modeling was used to simulate compressive tests of asphalt mixtures
with different film thicknesses. FIGURE 4.2 shows a model representing an asphalt mixture
includes two aggregates and asphalt binder. Each aggregate is covered by thousands small
particles which represent asphalt binder around aggregates. Between the two aggregates,
several layers made of even smaller blue balls represent film. Each layer of film in the model
contains 60*60=3600 particles with size:
 radius of Big aggregate = 500µm;
 radius of yellow balls = 15~30µm;
 radius of blue balls in film layers = 5µm.
FIGURE 4.2 Asphalt Mixture Includes Two Aggregates and Asphalt Binder.
62
Simple Viscoelastic Model
Asphalt mixture was modeled in PFC3d. Material properties were specified by applying
corresponding values of relative parameters. Such as normal stiffness, shear stiffness, friction
ratio, etc. Parallel bonds were built between adjacent particles. Cracking within the structure
was modeled by allowing bonds breakage. Elastic contact properties were used to investigate
the effect of random variations in internal sample geometry. A simple viscoelastic model was
used to introduce time dependent shear and normal contact stiffness and an elastic contact
was assumed for normal contact.
Consider a simple viscoelastic model in which the shear behavior consists of a spring in
series with a dashpot (58). The total shear velocity, 𝑢̇ s, can be decomposed into elastic and
viscous parts: Viscoelastic properties had been applied to film clump.
𝑢𝑠̇ = 𝑢𝑠̇ 𝑒 + 𝑢𝑠̇ 𝑣 (18)
For the two-dimensional case, the shear velocities and forces can be taken as scalars.
Taking Fs° and Fs' as the shear forces before and after one time step, respectively, we can
express the components of shear velocity as
𝑒 𝐹𝑠 ′ − F𝑠 °
𝑢𝑠̇ = − (19)
𝑘𝑠 ∆𝑡
and
𝐹𝑠 ′ + F𝑠 °
𝑢𝑠̇ 𝑣 = − (20)
2𝜂
where ks is the shear stiffness and η is the viscosity. Substituting these expressions into
Equation (18) and rearranging:

63
1 ∆𝑡
F𝑠 ° (𝑘 − 2𝜂) − 𝑢𝑠̇ ∆𝑡
𝐹𝑠 ′ = 𝑠
1 ∆𝑡 (21)
+ 2𝜂
𝑘𝑠
For the three-dimensional case, the shear force and relative shear velocity are both
vectors, and the constitutive equation becomes
° 1 ∆𝑡
′
𝐹𝑖𝑠 (𝑘 − 2𝜂) − 𝑢𝑖𝑠̇ ∆𝑡
𝐹𝑖𝑠 = 𝑠
1 ∆𝑡 (22)
+ 2𝜂
𝑘𝑠
Equation (22) is encoded into a contact model accessed with the “model viscous”
command. The required properties are as follows. Parameters of the contact model used in
simulations are summarized in TABLE 4.1.
 vis kn: viscoelastic normal stiffness
 vis ks: viscoelastic shear stiffness
 vis viscosity: shear viscosity

TABLE 4.1 Properties Applied in Simulations
Properties for viscoelastic contact
Properties for parallel bonds:
model:
15 -3
Normal stiffness = 30*10 Nm ; Viscoelastic normal stiffness
15 -3
Shear stiffness = 30*10 Nm ; = 0.36*106 N/m
Normal strength = 190 MP; Viscoelastic shear stiffness
Shear strength = 190MP; = 0.12*106 N/m
Radius multiplier = real parallel bond radius/radius
Shear viscosity = 0.414*106 N/m
= 30*10-9
Experiment
The Skyscan 1174 cabinet X-ray tomography system was used in the experiment to
verify the normal compliance model introduced before. The Skyscan 1174 system, shown in
FIGURE 4.3, is a compact, cost efficient micro X-ray scanner for nondestructive
64
three-dimensional microscopy. It’s supplied with software for system control, X-ray
radiography, three-dimension (3D)-reconstruction, two-dimensional (2D) /3D image analysis
and 3D realistic visualization. The SkyScan-1174 scanner supports variable magnification (6
to 30mm field of view), adjustable source energy (20 to 50KV) and flexible image format.
The material testing stage of Skyscan system, shown in FIGURE 4.4, can perform
compression, tension and torsion test. The loading-displacement or the stress-strain curve can
be saved as an image or text file. The testing sample can be held under specific loading for
scanning. The testing stage applies displacement to the top and bottom of the sample in equal
amount but in opposite directions. This keeps the central part of the sample relatively static
for scanning purpose.
FIGURE 4.3 Skyscan 1174 System.

65
FIGURE 4.4 Testing Stage of the Microscopy System (59).
Samples were tested under uniaxial compression loads applied on top and bottom stage.
With the help of X-ray scanner, the parameters needed in the compliance model were
measured by the tools provided by the software. The testing stage is displacement controlled.
The displacement speed was chosen according to the allowable range of the testing device,
and was set at 17.5μm/s. The resistant force was monitored and recorded during the loading.
When the force reaches the maximum allowable value of the testing stage, the displacement
application will stop. For the sample in this study, the application of displacement was
stopped when two elastic particles started to contact to each other through the asphalt layer.
Prior to each test, the testing stage was calibrated according to the procedure recommended
by the manufacturer. Force displacement data was stored as a text file for each sample. The
specification of asphalt binder is PG 64-22 coming from the lab of the Virginia Tech
Transportation Institute and the temperature at which the experiments are conducted is
around 25-27Co given by thermometer. The specifics of the experiment are listed in the
66
TABLE 4.2.
Ten compression tests are performed on samples that include two aggregates and asphalt
binder by using X-ray CT. The samples are made with two sphere aggregates and PG 64-22
binder (see FIGURE 4.5). Material of the aggregates is crystal (see FIGURE 4.6). Shown in
TABLE 4.2 are the load and other parameters of the experiment.
FIGURE 4.5 Crystal Aggregates Used in Compression Tests
FIGURE 4.6 Asphalt Mixture Sample Used in Compression Tests
FIGURE 4.7 - FIGURE 4.16 show the measurements and X-ray images of the ten
samples. Their film thicknesses are summarized in TABLE 4.3. After compressive load
applied, results of resistant force-deformation are shown in FIGURE 4.7 – FIGURE 4.16.
67
Since the samples are made manually, Sample 3 has the same film thickness with Sample 1.
From TABLE 4.3, peak resistant force of Sample 1 and Sample 3 are almost the same, so the
results are reasonable.
A relationship between film thickness and resistant force is plotted in FIGURE 4.17.
Based on current results, when compressive load keeps constant and all asphalt mixtures
move same displacement, the one with thinnest asphalt film had highest axial stress and
resistant force. That means the mixture includes aggregates and binder will become stiffer if
the film thicknesses decrease. The experimental results presents here agree well with my
simulation results and theoretical analysis in my former reports.
TABLE 4.2 Experiment Specifics

Inner diameter of the chamber 10mm
Diameter of particle 12.7mm
Column diameter 9mm
Material of particle Polyvinyl Chloride (E=2400-4100 MPa, υ = 0.41)
Asphalt binder PG 64-22
Test temperature Room Temperature
Loading speed 17.5 µm/s
TABLE 4.3 Parameters and Results of Compression Test

Load applied Speed Final Deformation Peak Resistant Film Thickness
ID
(N/step) (μm/s) (μm) Force (N) (μm)
1 0.22 17.5 10010.0 35.5 72
2 0.22 17.5 10010.0 37.6 36
3 0.22 17.5 10010.0 35.6 72
4 0.22 17.5 10010.0 38.8 30
5 0.22 17.5 10010.0 36.8 45
6 0.22 17.5 10010.0 41.1 21
7 0.22 17.5 10010.0 35.9 60
8 0.22 17.5 10010.0 36.4 47
9 0.22 17.5 10010.0 38.1 33
10 0.22 17.5 10010.0 35.2 90
68
FIGURE 4.7(a) X-ray Image of Sample 1
Resistant force
Deformation
FIGURE 4.7(b) Resistant Force vs. Deformation of Sample 1
69
Resistant force
Deformation
70
Resistant force
Deformation
71
Resistant force
Deformation

72
Resistant force
Deformation

73
Resistant force
Deformation
74
Resistant force
Deformation
75
Resistant force
Deformation
76
Resistant force
Deformation
77
Resistant force
Deformation
78
Ressistant force (N)

42
41 41.1
40
39
38.8
38 38.1
37.6
37
36.8
36.4
36 35.9 35.5
35.2
35 35.6
Film thickess
34 (μm)
20 30 40 50 60 70 80 90 100
FIGURE 4.17 Relationship between Resistant Force and Film Thickness
Simulation
The model analyzes a system with two elastic particles bonded by a thin viscoelastic
binder layer. Peak axial stresses are summarized in FIGURE 4.18. The only difference in
comparison is film thickness. FIGURE 4.19 - FIGURE 4.23 show models simulated in
PFC3d with varied thickness of 10µm to 20µm, 30µm, 40µm, 50µm, and axial stress of each
case. The peak stress of asphalt mixture was found to be as a power-law function of film
thickness. As Figure 8 shows, peak stress decreases with the increase of film thickness.
Comparing FIGURE 4.18 with FIGURE 4.17, simulation results agree reasonably with
experiment results.
79
FIGURE 4.18 Summary of Peak Axial Stresses
Case 1, Film Thickness = 10µm
FIGURE 4.19(a) Two Aggregates and a 10µm Film Layer in Compression Test
FIGURE 4.19(b) Detailed View of Film Layer, Film Thickness = 10µm

80
Axial Stress (MPa)

800
700
600
500
400
300
200
100
0 Timestep
300 400 500 600 700 800 900 1000 1100 1200 1300
FIGURE 4.19(c) Axial Stress in Compression Test, Film Thickness = 10µm

81
Axial Stress (MPa)

600
500
400
300
200
100
0
300 500 700 900 Timestep

82
Axial Stress (MPa)

600
500
400
300
200
100
0
300 500 700 900 Timestep

83
Axial Stress (MPa)

500
400
300
200
100
0
300 500 700 900 1100 Timestep

84
Axial Stress (MPa)

400
300
200
100
0
300 400 500 600 700 800 900 1000 Timestep
Summary
DEM has been used to simulate the behavior of an idealized asphalt mixture with
different film thicknesses under compressive load. Aggregates and asphalt binder were
simulated by balls generated in PFC3d. Corresponding behaviors were achieved by applying
properties to different clumps and contacts in PFC3d. These clumps represent aggregates,
surrounded asphalt binder, and asphalt film layer. UDM associated with material properties
provides a plate for performing visco-elastic behavior. A simple linear visco-elastic model
was introduced to give time dependent shear and normal contact stiffness. Asphalt binder of
asphalt mixture was successfully simulated using simple viscoelastic contact model. Results
of axial stresses of models simulated in PFC3d with varied thickness of 10µm to 20µm,
30µm, 40µm, 50µm showed peak stress of asphalt mixture is a power-law function of film
thickness. Peak stress decreases with the increase of film thickness.
Ten compression tests which is a validation of numerical analysis and DEM simulations
were conducted using samples with different binder film thicknesses. The Skyscan 1174
85
cabinet X-ray tomography system was used in the experiments. Samples were tested under
uniaxial compression loads. With the help of X-ray scanner, the parameters needed in the
compliance model were measured. Compression tests were done by applying displacement to
the top and bottom of the sample in equal amount but in opposite directions. This
displacement controlled testing stage output displacement versus resistant force results.
Based on the experiment results, it’s clearly that by moving same displacement, the resistant
force of thinner binder film system is larger than the mixture with thicker binder film.
Both simulation results and experiment results agree reasonably with theoretical analysis.
When compressive load keeps constant and all asphalt mixtures move same displacement, the one
with thinnest asphalt film had highest axial stress and resistant force. This means the mixture
includes aggregates and binder will become stiffer if the film thicknesses decrease.
86
5 EFFECT OF AGGREGATE SHAPE ON ASPHALT MIXTURE
Literature Review
Aggregates consist of approximately 95% of hot-mix asphalt (HMA) by weight.
Aggregate morphological characteristics such as particle size, shape, and texture influence
the performance and serviceability of hot-mix asphalt pavement (60). Flat and elongated
particles tend to break during mixing and compaction in mixture production and under traffic
loads in service. Therefore, aggregate shape is one of the important properties that must be
considered in the mix design of asphalt pavements to avoid premature pavement failure. The
recent Strategic Highway Research Program (SHRP) targeted primarily asphalt binder
properties and their contribution to pavement performance. Detailed investigations of
aggregate contributions to pavement performance were excluded from the program.
There is no standard test that is capable to quantify aggregate morphological
characteristics. Conflicting results have been reported on how aggregate shape influences the
quality of HMA mixtures. For example, Shklarsky and Livneh (61) concluded that replacing
uncrushed coarse aggregates with crushed coarse aggregates cannot significantly improve the
properties of asphalt mixture. Krutz and Sebaaly (62) found a direct correlation between the
rutting potential of HMA mixtures and the shape and texture of coarse aggregate particles. Li
and Kett (63) concluded in their study that flat and elongated particles can be used in a
mixture without adverse effect on its strength. Some mixes with flaky aggregates have been
found to exhibit higher fatigue life than mixes with nonflaky aggregates. The percentage of
crushed coarse particles had a significant effect on laboratory permanent deformation
properties. As the percentage of crushed coarse particles decreased, the rutting potential of
87
the HMA mixtures increased. Huber and Heiman (64) found that crushed aggregate
containing 19% flat and elongated particles did not adversely affect the volumetric properties
of HMA mixtures. Stephens and Sinha (65) presented data on the effect of aggregate shape,
and recommended blends of regular particles, flat particles, and rod-like particles to achieve
optimum strength. They discussed the effect of crushing, aggregate size, and particle shape.
They emphasized that asphalt usage can be reduced by increasing the size of crushed
aggregate used in a mix. These conflicting statements are due to the lack of understanding on
the effect of aggregate shape on engineering properties of a HMA mix. Visual examination is
the most common method of judging aggregate shape. Because of the tedious task of making
numerous readings, the civil engineer generally finds it impractical to identify particle shape
visually. Various innovative methods are available to facilitate the quantification for
aggregate shape. One of the most effective methods is digital image processing and analysis.
This method provides the capability of quick and accurate measurement of the characteristics
of aggregate particles. For example, Yue (66) used a digital image processing technique to
quantify the distribution, orientation, and shape of coarse aggregate. Their quantitative results
indicated that microstructural characteristics of asphalt concrete mixtures can be accurately
measured using a digital image processing technique. Kuo (67) also used three-dimensional
image analysis techniques of aggregate particles to measure the shape, size, and diameter of
the particles. Therefore, the objectives of this study are (1) to quantify the morphological
characteristics of coarse aggregate, and (2) to evaluate the engineering properties of HMA
mixtures made of aggregate with different shapes.

88
Materials and Methods
Aggregate
The following procedure was performed for evaluating the effect of aggregate shape on
strength of asphalt mixtures. Coarse aggregates refer to aggregate particles retained on the
4.75 mm sieve, whereas fine aggregates are aggregate particles passing through the 4.75 mm
sieve. TABLE 5.1 tabulates the gradation curve.
TABLE 5.1 Aggregate Gradations

Sevier size (mm) 19 12.5 9.5 4.75 2.36 1.18 0.6 0.3 0.15 0.075
Cumulative percentage by 100 93 80 49 32 23 17 12 8 5.2
weight (%)
The coarse size fractions are evaluated for the influence of aggregate shape on
engineering properties of a HMA mix. Aggregate shape is determined from the longest
diameter (dL), the intermediate diameter (dI), and the shortest diameter (dS). The selection
process was carefully conducted by an image analyzer, as discussed in the next section. The
elongation ratio and the flatness ratio were used to define the aggregate shape as shown in
FIGURE 5.1 – FIGURE 5.3. Four different aggregate shapes are as follows: hexagonal disk,
blade, rod, and cube; their flatness ratio and elongation ratio are summarized in TABLE 5.2.
Only cubical aggregates have flatness ratio and elongation ratio greater than 2/3, whereas the
other three types of aggregates are flaky and oblate (i.e., either the flatness ratio or elongation
ratio are greater than 2/3). The hexagonal disk-shaped aggregate is flaky and oblate, the
rod-shaped is elongated, and the blade-shaped is both flaky and elongated.

89
TABLE 5.2 Flatness Ratio and Elongation Ratio of Four Different-shaped-aggregates

Aggregate shape Flatness Ratio (dS/dI) Elongation Ratio (dI/dL)
Cubic >2/3 >2/3
Rod >2/3 <2/3
Blade <2/3 <2/3
Hexagonal disk <2/3 >2/3
FIGURE 5.1(a) Cubic-shaped Aggregate
FIGURE 5.1(b) Detailed view of Cubic-shaped Aggregates in Simulation
FIGURE 5.2(a) Rod-shaped Aggregate

90
FIGURE 5.2(b) Detailed View of Rod-shaped Aggregates in Simulation
FIGURE 5.3(a) Blade-shaped Aggregate

91
FIGURE 5.3(b) Detailed View of Blade-shaped Aggregates in Simulation
FIGURE 5.4(a) Hexagonal Disk-shaped Aggregate

92
FIGURE 5.4(b) Detailed View of Hexagonal Disk-shaped Aggregates in Simulation
Image Measurements of Coarse Aggregate
Image analysis techniques were used in this study to characterize the morphological
characteristics of coarse aggregate particles. The system consists of two major assemblies,
including a microscope with a scanner and a rack of modules. The device named Skyscan
1174 cabinet X-ray tomography system is versatile software capable of providing full
measurements of coarse aggregate.
Aggregate particles were attached to adhesive clear plastic trays with two perpendicular
faces; then, the sample tray was rotated 90 ° to establish two orthogonal planes of
measurement. Particles were placed on a light box that illuminated the sample and made
definite contact between aggregate and background. The parameters of length, width, and
thickness were obtained by measuring the two orthogonal planes. These parameters provided
a direct method for determining the flatness and the elongation ratios of the particles.
93
Additionally, the image analysis method provided other shape indices that could be related to
the effects of aggregates on the properties of a HMA mix. This image analysis method was
more time efficient than the ASTM Test Method for Flat Particles, Elongated Particles, or
Flat and Elongated Particles in Coarse Aggregate (ASTM D4791) (68). The imaging
processing technique also provided more information.
The mean value for each aggregate size is listed in TABLE 5.3. For cubical particles, the
flatness ratio is 0.80 and elongation 0.78, and both values are larger than 2/3. Note that, in
cubical limestone, particles in the size range 19 to 12.5 mm are the most spherical, and those
in the size range 9.5 to 4.75 mm are relatively flat and elongated. Data in TABLE 5.3 indicate
that the greater the shape factor is, the more nearly cubical the aggregate is. For aggregates
used in pavement construction, the shape factor is generally between 0.3 and 0.8. Aggregate
used in this study falls within this range. The sphericity value represents the roundness of an
aggregate regardless of its thickness. As expected sphericity value ranges between 0.5 and 0.9,
and cubical particles have higher sphericity values than the others.
TABLE 5.3 Aggregate Geometric Characteristics

Shape Size dL dI dS Elongation Flatness Shape Sphericity
(mm) (mm) (mm) (mm) ratio ratio factor
Cubic 19-12.5 18.32 15.02 11.28 0.82 0.78 0.68 0.80
12.5-9.5 14.05 11.34 9.08 0.81 0.80 0.72 0.80
9.5-4.75 9.58 7.39 5.74 0.77 0.75 0.68 0.77
Average 0.80 0.78 0.69 0.79
Rod 19-12.5 26.58 14.75 11.88 0.55 0.81 0.60 0.63
12.5-9.5 18.69 10.89 8.88 0.58 0.82 0.62 0.65
9.5-4.75 13.85 9.06 7.18 0.65 0.79 0.64 0.70
Average 0.59 0.81 0.62 0.66
Hexagonal 19-12.5 20.87 17.98 10.32 0.86 0.57 0.53 0.75
Disk 12.5-9.5 15.36 13.02 8.01 0.85 0.62 0.57 0.76
9.5-4.75 11.01 8.98 5.32 0.82 0.59 0.54 0.73
94
Average 0.84 0.59 0.55 0.75

Blade 19-12.5 29.31 17.38 11.32 0.59 0.65 0.50 0.61
12.5-9.5 20.96 12.87 5.38 0.61 0.42 0.33 0.54
9.5-4.75 15.85 7.98 3.25 0.50 0.41 0.29 0.47
Average 0.43 0.49 0.37 0.54
Note: Elongation ratio= dI/dL; Flatness ratio= dS/dI; Shape factor= dS/√𝑑𝐼 ⋅ 𝑑𝐿 ); Sphericity= dS⋅ 3√𝑑𝐼 ⋅ 𝑑𝐿 .
Calculation of Particle Index of Coarse Aggregates Based on DEM Simulations
The combined effects of particle shape and surface texture of an aggregate are
determined in accordance with simulated ASTM D 3398 (69). FIGURE 5.5 shows the
equipment of the test. Simulated tests to measure particle index (PI) were done by using
DEM; results are summarized in TABLE 5.4. The simulation requirements for this test
consists basically of a cylindrical steel mold 152 mm (6 in.) in diameter by 178 mm (7 in.)
high, and a steel rod 16 mm (5/8 in.) in diameter by 610 mm (24 in.) long with the tamping
end rounded to a hemispherical tip. A clean, washed, oven-dried, and single-size aggregate
fraction was used for this test. The mold was filled in three equal layers, with each layer
compacted with 10 well-distributed blows of the tamping rod. Each tamp consisted of a drop
with the tamping rod from 51 mm (2 in.) above the surface of the layer being compacted.
This procedure was repeated using the same material but applying 50 blows on each of the
three layers. The weight of the contents of the mold in each case was determined and the
corresponding percentage of voids was calculated using the bulk specific gravity of each
aggregate fraction. The particle index (PI) is derived using the following equation:
PI = 1.25V10 – 0.25V50 – 32 (23)
Where
V10 = percent voids in the aggregate compacted with 10 blows per layer;
95
V50 = percent voids in the aggregate compacted with 50 blows per layer.
FIGURE 5.5 Tamping Rod and Sleeve (69)
TABLE 5.4 Measured Particle Index (PI) for Coarse Aggregate

Shape Size Range (mm) Percentage (%) PI for each size Weighted value
Cubic 19-12.5 62.7 17.6 11.0
12.7-9.5 14.5 15.7 2.3
9.5-4.75 22.8 15.2 3.5
PI 16.8
Rod 19-12.5 62.7 16.3 10.2
12.7-9.5 14.5 14.9 2.2
9.5-4.75 22.8 14.0 3.2
PI 15.6
Hexagonal disk 19-12.5 62.7 15.1 9.5
12.7-9.5 14.5 14.1 2.0
9.5-4.75 22.8 13.3 3.1
96
PI 14.6
Blade 19-12.5 62.7 12.8 8.0
12.7-9.5 14.5 12.3 1.8
9.5-4.75 22.8 11.2 2.6
PI 12.4
Simulated Indirect Tensile Tests on Digital Samples with Selected Grading Aggregates
Using DEM
The indirect tensile strength of HMA is used often to evaluate the relative quality of
materials. The repeated-load indirect tension test for determining the resilient modulus was
conducted by applying compressive loads according to ASTM D4123 (70). The load was
applied vertically in the vertical diameter plane of a cylindrical specimen of asphalt concrete
through a curved loading strip. The resulting horizontal deformation was measured and used
to calculate the total resilient modulus (Mr) as follows:

𝑃∙(𝑣+0.27)
𝑀𝑟 = (24)
𝑡∙𝐻
Where
P = repeated load,
v = Poisson’s ratio,
t = thickness of specimen,
H = total recoverable horizontal deformation.
Poisson’s ratio was calculated using the measured recoverable vertical and horizontal
deformation. The tensile strength (ST) is calculated as follows:

1.96∙𝑃𝑢𝑙𝑡
𝑆𝑇 = (25)
𝜋∙𝑡∙𝐷
Where
97
Pult = ultimate applied load required to cause to the specimen to fail,
t = thickness of specimen,
D = diameter of specimen.
Conventionally, the indirect tensile test is conducted to indicate the internal resistance in
a mix. Mixes were prepared to observe the effect of aggregate shape on the strength of the
mixes. TABLE 5.5 shows peak strength of mixtures include different aggregates. FIGURE
5.6 shows the test results for different aggregate shapes vs. PI values. A trend was observed
for indirect tensile strength. Coarse aggregates with lower PI provide lower stiffness than
those with higher PI. Blade and hexagonal disk particles impede compaction and thus may
prevent the development of satisfactory properties in HMA. In compacted mixtures, cubical
particles provide great interlocking internal friction between aggregates and, resulting in
greater mechanical stability than flat, thin, and/or elongated particles. The PI value is a
quantitative measure of the aggregate shape that influences the characteristics of HMA
mixtures.
TABLE 5.5 Peak Normal Strengths Achieved in Indirect Tensile Test

Aggregate Shape Loading Rate Temperature Specimen Peak Normal
(mm/min) (oC) Diameter (mm) Strength (MPa)
Cubic 50 25 100 1.69
Rod 50 25 100 1.40
Hexagonal disk 50 25 100 1.35
Blade 50 25 100 1.31
98
1.8
Indirect tensile strength (MPa) Cubic

1.7
1.6
1.5
Rod
1.4 Hexagonal Disk
Blade
1.3
1.2
PI
12 13 14 15 16 17
FIGURE 5.6 Effect of Aggregate Shape on Indirect Tensile Strength
Simulated Compression Tests on Digital Samples with Ten Aggregates in Each of Them
Using DEM
Summary of DEM Properties
Different form the indirection tensile test, digital model of the samples used in the
compression test are built on a smaller scale. Each sample is created by asphalt binder and 10
aggregates. Same compression load has been applied on 36 digital samples, 12 of them
include only sphere aggregates, 12 of them include only cubic aggregates, and the other 12
include randomly shaped aggregates. These samples are true to size with the specimens that
are tested by an X-ray scanner, Skyscan 1174 system. Details of the X-ray aided compression
experiments are summarized in next session.
FIGURE 5.7, 5.8 and 5.9 show the views of the digital samples in DEM simulation.
Geometry information of the samples could be found in TABLE 5.6. Two different contact
models are applied to aggregates and asphalt binders (see TABLE 5.7). Small yellows balls
99
represent asphalt binder and orange clusters represent aggregates. Though 10 aggregates are
included in each sample, some of them are located more close to center, so it’s hard to see all
of them on the pictures.
TABLE 5.6 Geometry Information of the Samples used in Simulated Compression Test
Sample Aggregate
Height = 16mm Sphere Cubic Random
Longest diameter = 4mm;
Diameter = 8mm Diameter = 4mm Side length = 3mm
Shortest diameter = 2mm
TABLE 5.7 Contact Model and Mechanic Property

Aggregate (Elastic) Asphalt binder (Viscoelastic)
Density: 2.34 g/ml Shear viscosity = 0.414*106 N/m
Normal stiffness Viscoelastic normal stiffness = 0.36*106 N/m
Shear stiffness Viscoelastic shear stiffness = 0.12*106 N/m
FIGURE 5.7 Digital Sample Include 10 Sphere Aggregates

100
FIGURE 5.8 Digital Sample Include 10 Cubic Aggregates
FIGURE 5.9 Digital Sample Include 10 Random Aggregates

101
Results and Analysis
TABLE 5.8 Peak Normal Strengths Achieved in Compression Test on Digital Samples
with Ten Aggregates in Each of Them
Aggregate Loading Rate Temperature Peak Normal Strength
o
Shape (μm/s) ( C) (MPa)
Sphere 17.5 25 1.12
Cubic 17.5 25 1.68
Random 17.5 25 1.34
Similar with the results of indirect tensile test, specimen with cubic aggregates reached
the greatest peak normal strength. The values of the strength of the digital specimens in both
of the simulated tests are pretty close to each other. This means contact models and physical
properties applied to the digital models in PFC3d perfectly represent what real asphalt binder
and aggregate should have. These results are also verified by experiments in the following
session.
Compression Tests Using X-ray Image Scanner
Equipment and Specimens Preparation
The Skyscan 1174 cabinet X-ray tomography system was used in the experiment to
verify the simulation results from DEM models introduced before. The Skyscan 1174 system,
shown in FIGURE 5.10, is a compact, cost efficient micro X-ray scanner for nondestructive
three-dimensional microscopy. It’s supplied with software for system control, X-ray
radiography, t hree-dimension (3D)-reconstruction, two-dimensional (2D) /3D image analysis
and 3D realistic visualization. The SkyScan-1174 scanner supports variable magnification (6

102
to 30mm field of view), adjustable source energy (20 to 50KV) and flexible image format.
The material testing stage of Skyscan system, shown in FIGURE 5.11, can perform
compression, tension and torsion test. The loading-displacement or the stress-strain curve can
be saved as an image or text file. The testing sample can be held under specific loading for
scanning. The testing stage applies displacement to the top and bottom of the sample in equal
amount but in opposite directions. This keeps the central part of the sample relatively static
for scanning purpose.
FIGURE 5.10 Skyscan 1174 System
FIGURE 5.11 Testing Stage of the Microscopy System

Samples were tested under uniaxial compression loads applied on top and bottom stages.
With the help of X-ray scanner, the parameters needed in the compliance model were
measured by the tools provided by the software. The testing stage is displacement controlled.
The displacement speed was set at 17.5μm/s, chosen according to the allowable range of the
103
testing device, and was set at 17.5μm/s. The resistant force was monitored and recorded
during the loading. When the force reaches the maximum allowable value of the testing stage,
the displacement application will stop. For the sample in this study, the application of
displacement was stopped when two elastic particles started to contact to each other through
the asphalt layer. Prior to each test, the testing stage was calibrated according to the
procedure recommended by the manufacturer. Stress vs. Strain data was stored as a text file
for each sample. The PG 64-22 asphalt binder is used and the test temperature is 25-27Co
according to the thermometer.
Compression tests are performed on samples that include ten aggregates and asphalt
binder by using X-ray CT. The samples are made with aggregates with different shapes and
PG 64-22 binder (see FIGURE 5.12 and FIGURE 5.13).
FIGURE 5.12 Crystal Particles and Aggregates Used in Compression Test
FIGURE 5.13 Asphalt Mixture Specimen Used in Compression Test

104
Test and Results Analysis
There are 36 specimens (12 with sphere aggregates, 12 with cubic aggregates, and 12
with random aggregates) tested under compression load by using Skyscan 1174 X-ray
scanner. Weight of the aggregates and asphalt binder used in each specimen are summarized
in TABLE 5.9. These specimens were taken from a freezer to and mounted to a stage placed
as vertically as possible. Once mounted and imaged, the stage was first pulled using a tension
load, in order to make sure the specimen had adequate space to compress properly, then the
compression test was performed. Throughout the compression test, X-ray images of the
specimens were taken to show the positions of the aggregates within the specimens (See
FIGURE 5.14). The compression test software was also calibrated to the numbers shown in
TABLE 5.10. TABLE 5.11 tabulated the test results, and FIGURE 5.15 plots the stress-strain
relationship.
TABLE 5.9 Weights of Aggregates and Asphalt Binder

Sample Cubic Asphalt Binder Sphere Asphalt Binder Random Asphalt Binder
Aggregate in Random Aggregate in Random Aggregate in Random
(g) Specimens (g) (g) Specimens (g) (g) Specimens (g)
1 0.7 0.4 0.4 0.3 0.5 0.3
2 0.8 0.3 0.4 0.3 0.4 0.1
3 0.8 0.4 0.4 0.5 0.5 0.3
4 0.7 0.3 0.4 0.3 0.3 0.3
5 0.7 0.5 0.4 0.3 0.4 0.2
6 0.7 0.7 0.4 0.4 0.6 0.5
7 0.7 0.5 0.4 0.4 0.5 0.4
8 0.8 0.7 0.4 0.4 0.3 0.2
9 0.7 0.9 0.4 0.4 0.3 0.4
10 0.7 0.5 0.4 0.4 0.9 0.4
11 0.7 0.6 0.4 0.3 0.5 0.5
12 0.7 0.3 0.4 0.4 0.6 0.3
105
TABLE 5.10 Loading Information of Compression Test

Force (N/step) 0.22
Torque (N*cm/step) 0.08
Speed (μm/s) 17.5
Angle (deg/sec) 52.80
Linearization (%) 2
Symmetry (%) 110
TABLE 5.11 Peak Normal Strength of Each Specimen Achieved in Compression Test
Peak Normal Strength of Peak Normal Strength of Peak Normal Strength of
Sample
Specimens with Sphere Specimens with Cubic Specimens with Random
Aggregates (MPa) Aggregates (MPa) Aggregates (MPa)
1 1.139 1.656 1.269
2 1.135 1.785 1.386
3 1.068 1.761 1.268
4 1.136 1.699 1.234
5 1.132 1.640 1.359
6 1.119 1.627 1.215
7 1.116 1.642 1.247
8 1.109 1.683 1.354
9 1.118 1.605 1.238
10 1.110 1.651 1.275
11 1.133 1.647 1.263
12 1.114 1.684 1.346
Mean Peak Strength 1.119 (MPa) 1.673 (MPa) 1.288 (MPa)
(a) (b) (c)

FIGURE 5.14 X-ray Images of Specimens with Asphalt Binder and (a) Ten Cubic
Aggregates, (b) Ten Sphere Aggregates, and (c) Ten Random Aggregates
106
1200000
1000000
Stress (Pascals)
800000
600000
400000
200000
0 Strain
0 0.01 0.02 0.03 0.04 0.05 0.06
FIGURE 5.15 Stress vs. Strain of One Specimen with Ten Sphere Aggregates
There is a good agreement between simulations and experiments. Specimens with cubic
aggregates have the greatest peak normal strength in both the simulation test and
experimental test. However, there are some discrepancies. In the simulation the specimen was
able to reach its ultimate strength, which was hard in the experiment due to the allowed
loading of the specimen by the testing software. Finally, the simulation was much more
detailed and spiky because the PFC3d software could tell exactly how the contact force
changed between particles, rather than the experiment software, which could only read the
stress and strain of the total specimen.
Summary
The indirect tensile strength of HMA with four different shapes of aggregate was
evaluated in simulated test to evaluate the relative quality of materials. The image analyzer
was shown to be a useful tool for quantifying the morphological characteristics of coarse
aggregates. Image evaluation provided quantitative indices such as geometric measurement

107
and angle rotation of granular materials. Data showed that the morphological characteristics
of coarse aggregate correlated well with the results of other indirect tests such as the particle
index. Cubical particles possessed the best resistance over the other shapes. The particle
index (PI) was shown to be an adequate measure of the combined contribution of particle
shape, angularity, and surface texture to the stability of an aggregate. The PI value correlated
well to aggregate geometric characteristics including elongation ratio, flatness ratio, shape
factor, and sphericity. The particle shape determined how aggregate was packed into a dense
configuration and also determined the internal resistance of a mix. The morphological
characteristics of coarse aggregates found from both digital test and image analysis were in
good agreement with the engineering properties of HMA mixtures. This study presents a
precise method to evaluate the aggregate characteristics in a HMA mix and demonstrates
their effects on pavement performance. The results of this study would provide useful
guidelines for highway engineers to construct a long-lasting pavement.

108
6 EVALUATION OF COMPACTION TEMPERATURE SENSITIVITY
Literature Review
One of the most traditional constructed materials is “Hot Mix Asphalt (HMA)”.
Mechanical performance of HMA depends on solid production, temperature control, and
moisture, it is still being paid attention by many people. Determination of volumetric
parameters of HMA samples has vital importance for quality control of constructed roads.
The compaction temperature is one of the parameters that can change the HMA volumetric
parameters. In HMA pavement construction, compaction is defined as the process by which
the volume of air in a HMA mixture is reduced through the application of external forces.
Compaction is an essential factor in the design and subsequent production of asphalt mixtures.
The expulsion of air enables the mix to occupy a smaller space, thereby increasing the unit
weight or density of the mix.
The compaction temperature influences workability which is related to the achieved
density of the mixture. According to the ASTM D6926, the compaction temperature is the
temperature at which bituminous should be heated to produce viscosities of 0.28 ± 0.03 pa·s.
(71). This requirement was based on experience with conventional asphalt binders. In general,
the binder in modified asphalt mixture is stiffer than in conventional mixtures, therefor, there
is a need for a higher compaction temperature. However, previous studies (Bahia 2000;
Huner and Brown 2001) on the effect of compaction temperature on the volumetric properties
of asphalt mixture reported that specimens could have the same volumetric properties over a
very wide range of compaction temperatures (72, 73). Azari et al. (2003) also suggested that a
temperature range from 119 oC to 159 oC could be used for modified mixtures with the
109
limestion - Novophalt binder (74).
The compaction temperature is one of the major issues in HMA and also one of
important criteria in the process of producing good quality of hot mix asphalt. Also, the
temperature is a key factor in the control of bitumen viscosity, which affects its ability to coat
and provide adequate lubrication for aggregates.
Viscosity is simply a measure of a fluid resistance to flow and is described by the
following equation:
τ
μ= (26)
γ
Where: μ = viscosity (in cgs units of poise); poise = dyne-sec/cm2 = g/cm-sec (the SI unit of
viscosity is the Pa·s = N-sec/m2 = 10 poise)
τ = shear stress
γ = shear rate
Aggregate Grading
In this study two mix designs were used and the gradation of the aggregates has been
shown in the TABLE 6.1. Gradation of surface course layer is the optimal one selected in the
chapter “Determination of Gradation”.
TABLE 6.1 Gradations of Aggregates Used in Simulation

Sieve size (mm) 25 19 12.5 9.5 4.75 2.36 1.18 0.6 0.3 0.15 0.075
Surface Course Layer,
Percentage passing sieve size (%) 100 93 80 49 32 23 17 12 8 5
Binder Course Layer,
Percentage passing sieve size (%) 100 95 69 50 35 10 5
110
Asphalt Binder Selection
Considering this fact that the most consuming asphalt binder is the bitumen of 60-70, in
this study asphalt binder PG 64-22 was used and FIGURE 6.1 shows the temperature and
viscosity tested by Freddy L. Roberts and Prithvi S. Kandhal, et al (75). TABLE 6.2 shows
the properties of asphalt binder specified in DEM simulation.

Viscosity, Centistokes
Temperature oc
FIGURE 6.1 Viscosity-Temperature of Bitumen for Determination of Compaction

Temperatures (75)
TABLE 6.2 Properties of Asphalt Binder Used in DEM Simulation

Density at 25 oC 1.017 (g/cm3)
Penetration of asphalt binder at 25 oC (100g, 5sec) 80 (d-mm)
Softening point of asphalt binder 50 (oC)
Ductility at 25 oC +100
Solubility in trichloroethylene 99.7 (% by mass)
Flash and fire points by cleveland open cup 296 (oC)
Kinematic viscosity at 120 oC 884 (Pa·s)
o
Kinematic viscosity at 125 C 690 (Pa·s)
o
o
o
o
111

DEM Simulation Process of Superpave Gyratory Compactor (SGC) Compaction
Gyratory compaction is the commonly used compaction method to prepare asphalt
mixture specimen. In the DEM simulation of gyratory compaction, the vertical pressure was
set at 600 kPa and the angle of gyration was set at 1.25º. The gyration was applied at a rate of
30 revolutions per minute. The gyratory kneading action in PFC3d was applied by a rotary
compression plate which is controlled by two separate DEM simulation engines: rotation
plate engine and constant pressure engine. The rotation plate engine applies a constant
rotation speed on the compression plate around a specific axis. The function of constant
pressure engine is to apply a specific pressure to the underneath particles by specified.
The procedures of the DEM simulation process is summarized as follows:
1) Generate compaction cylinder and funnel (FIGURE 6.2a);
2) Calculate particle numbers of each particle size according to gradation curve;
3) Randomly generate particles in specific space (FIGURE 6.2b);
4) Pack particles under gravity force until it is stable (FIGURE 6.2c and FIGURE 6.2d);
5) Generate compression plate and gyratory compact asphalt mixture by constant
pressure (FIGURE 6.2e and FIGURE 6.2f);
6) Record the position of compression plate and spheres during compaction process;
7) Calculate air void content of the mixture.

112
(a)
(b)
113
(c)
(d)
114
(e)
(f)
FIGURE 6.2 DEM Simulation Process of SGC
115
Result and Analysis
Simulated compactions were done at different temperatures; several results were
recorded (see TABLE 6.3). In this study, the temperature of the asphalt binder during mixing
with aggregate that with continuous grading in mixer increases from 120 oC to 160 oC.
TABLE 6.3 Data Measured at a Serial of Modeling Temperatures

Stiffness Air voids Voids filled with
Density (g/cm3) Flow (mm)
T of asphalt (VA, %) asphalt (VFA, %)
o
( C) binder Binder Surface Binder Surface Binder Surface Binder Surface
(kg/mm) course course course course course course course course
120 407.19 2.199 2.175 5.4 5.0 66.6 (VFA,
67.1%) 2.76 2.76
125 416.32 2.204 2.184 5.2 4.8 67.9 68.3 2.79 2.91
130 429.28 2.208 2.192 4.9 4.5 69.6 69.9 2.83 3.00
135 443.97 2.216 2.198 4.7 4.3 70.7 71.2 2.88 3.08
140 466.54 2.222 2.203 4.5 4.2 71.8 72.2 2.91 3.17
145 481.66 2.229 2.208 4.4 4.1 72.7 73.1 2.95 3.24
150 499.58 2.227 2.207 4.2 4.0 73.3 73.7 2.98 3.29
155 519.09 2.226 2.207 4.1 3.9 73.8 74.3 3.03 3.35
160 541.08 2.225 2.206 4.0 3.8 74.4 74.7 3.05 3.40
Analysis of Density
FIGURE 6.3 shows the following results: Density of binder course layer is slightly
greater than surface course layer. Density of binder course layer and surface course layer is
being simultaneously increased with increasing compaction temperature, which is as a result
of asphalt cement viscosity due to temperature increase and subsequently, the location of
aggregates beside each other becomes denser.
The highest density of binder course and surface course is being occurred over the
temperature of 145 °C. High correlation of binder course and surface course indicates the
116
strong relationship between density and compaction temperature. (For the surface course
aggregate, R2 = 0.936 and for the binder course aggregate R2 = 0.951).
Density (g/cm3)
2.24
2.23
2.22
2.21 Binder course

Y=1.681*X^0.055
2.2
Surface course
Y=1.634*X^0.059
2.19
2.18
2.17
120 125 130 135 140 145 150 155 160 165
Temperature oC
FIGURE 6.3 Density vs. Compaction Temperature
Analysis of Percent of Air Voids (VA)
Air voids (VA) of aggregate of binder course layer is higher comparing to the VA of
aggregate of surface course layer. Furthermore, FIGURE 6.4 shows high R2 value for binder
course layer and surface course layer is indicator of the strong relationship between air void
and compaction temperature. (For the surface course layer aggregate, R2= 0.970 and for the
binder course layer aggregate, R2= 0.986) The higher compaction temperature results in the
change of asphalt binder viscosity and more asphalt binder dispersion on the surface. It
117
makes a thin film of asphalt binder becomes enough for covering the coarse aggregates, and
finally since the volume of asphalt binder is constant, consequently, the temperature increase
only decreases the percent air void.
Air voids (VA, %)

6
4
Binder course
3 Y=575.1*X^(-0.97)
2 Surface course
Y=810.6*X^(-1.05)
0
120 125 130 135 140 145 150 155 160 165
Temperature oC
FIGURE 6.4 Air voids vs. Compaction Temperature
Analysis of Percent Voids Filled with Asphalt (VFA)
FIGURE 6.5 shows that voids filled with asphalt (VFA) simultaneously increases with
the increase of compaction temperature. The amount of VFA in the binder course is more
than that in the surface course, as the optimum asphalt content in the surface course is greater
than that in the binder course. However, due to the lack of inequality of air void percent
difference of surface course comparing to the binder course at different temperatures, the
amount of surface course VFA are not exactly equal to the binder course VFA. Also, the
118
drawn curves for binder course and surface course show an attended move of VFA versus the
increase of temperature. This figure also indicates a strong relationship between VFA and
compaction temperature (because of the high correlation of binder course and surface course,
for the surface course aggregate, R2 = 0.990 and for the binder course aggregate R2 = 0.970).
Voids filled with asphalt (VFA, %)

76
75
74
73
72
Surface course
71 Y=10.91*X^(0.380)
70 Binder course
Y=10.74*X^(0.378)
69
68
67
66
120 125 130 135 140 145 150 155 160 165
Temperature oC
FIGURE 6.5 VFA vs. Compaction Temperature
Analysis of Flow
The quantity of Marshall Flow in surface course and binder course increases with the
increase of compaction temperature. FIGURE 6.6 indicates that the difference between flows
of surface course and binder course becomes greater at higher temperatures. This figure
119
shows a strong relationship between flow and compaction temperature (because of the high
R2 value for binder course = 0.966 and 0.989 for surface course). According to the minimum
and maximum authorized quantity for flow in the Asphalt Institute Manual Series MS-2
(2-3.5mm) (76), FIGURE 6.6 shows that all of the attained flow quantities for surface course
and binder course are within the authorized limitation.
Flow (mm)
4
3.5
2.5
Binder course
2 Y=0.154*X^(0.608)
Surface course
1.5 Y=0.278*X^(0.478)
0.5
0
120 125 130 135 140 145 150 155 160 165
Temperature oC
FIGURE 6.6 Flow vs. Compaction Temperature
Summary
This study indicates that the temperature has significant impression on hot mix asphalt
properties. From the discovered high correlation value (R2 over 0.9) in the drawn figures, we
can realize that there is a strong relation between various parameters of asphalt mixture
samples and compaction temperature. So, observing the compaction temperature during the
120
performing hot mix asphalt, can help us to avoid from considerable changes which there are
in the asphalt specifications.
Increasing compaction temperature makes density to be increased, but there is a
limitation for this increase and afterwards density will be decreased.
Considering the mentioned minimum and maximum for VA and VFA in MS-2 and other
references, the amount of authorized temperature parallel with them can be achieved.
Increasing the temperature makes flow to be increased. The difference between flows of
surface course and binder course becomes greater at higher temperatures.

121
7 EVALUATION OF MOISTURE DAMAGE POTENTIAL
Literature Review
The liquefaction behaviors of concrete have received a considerable amount of attention.
Because flow-type landslides are strongly associated with the liquefaction phenomenon (77,
78, 79), the extensive geo-technical tests, especially the undrained triaxial compression tests,
have been performed under both dynamic stress-loading and static conditions. It was
concluded that the collapse of loose meta-stable structures due to the compressive
deformation produced excessive pore-water pressure resulting in the liquefaction (80). Since
the triaxial compression tests cannot investigate the undrained shear behaviour subjected to a
large shear displacement, the studies using the ring-type direct shear testing apparatus have
also been conducted. Sassa proposed the sliding surface liquefaction that occurred within the
shear zone where particles were crushed and comminuted (77). Note that the sliding surface
liquefaction takes place even when the specimens are medium-dense through dense, whereas
the liquefaction can only occur in the loose specimens. Okada et al. (2004) stated that the
crushed and comminuted particles can form the less-permeable shear zone, and it hindered
the dissipation of generated excessive pore-water pressure in the ring shear tests (81). Cundall
and Strack (1979) proposed a discrete numerical method, to model the shear behavior of
granular assemblies. In discrete element method, the contacts between ball and ball (in three
dimensions) or disc and disc (in two dimensions) are modeled by the mechanical elements of
elastic spring and viscous damper (FIGURE 7.1) (82). Gravity flow of granular materials
were numerically simulated by two-dimensional discrete element method. Although the
discrete element method has difficulties in tuning up the setting parameters and it consumes
122
rather too much time with limited number of elements, it is a very strong tool to investigate
non-continuum problems.
Previous work also has revealed that the pore water pressure (PWP) reduction of
cement-based materials such as paste, mortar and concrete reflects rather well the stiffness
evolution and volume change of such materials from time zero and through the time of set
(83). Furthermore, it has revealed that PWP is a good indicator for susceptibility of “plastic
shrinkage cracking” and for friction against panels in forming slips (84). Early PWP
reductions are caused by two interacting mechanisms: vacuum effect caused by chemical
shrinkage and meniscus effect (capillary tension). Air voids will alter the early age PWP. This
is because they may grow as long as the stiffness of the paste-mortar-concrete is low, which
results in PWP release.
Initially and in a sealed condition (no moisture exchange with the surroundings), PWP
corresponds to the hydrostatic pressure of the liquid paste-mortar-concrete given by the
height over the point of measuring and the density. PWP will decrease and reach the water
pressure given by the height over the point of measuring when the paste-mortar-concrete is
able to support its own weight (point of self-support, PSS). Note that PSS occurs before time
of initial set (83). In the time before PSS the body is too week to resist the total volume
change (chemical shrinkage). Beyond PSS, the total volume change will be resisted, more
and more as the stiffness increases. It results in compression force in the solid and tension
force of the pore water, forming so-called vacuum effect. Beyond final setting (i.e. beyond
the time period considered here), emptying of capillary pores (self-desiccation) forms menisci,
resulting in capillary tension.

123
The buffer effect and the fact that the amount of air pores in cement paste varies from
nominally zero in pure cement paste, to 20% in the cement paste of concrete with air
entraining mixture (AEM), suggest that air voids play an important role in the present period
in time. The study presents and discusses some results from a limited test program of cement
paste with and without AEM as well as concretes with and without AEM, conducted in order
to evaluate the influence of air voids on the mentioned properties, both from a fundamental
and practical point of view.
FIGURE 7.1 Mechanical Elements Introduced in Ball–ball Contacts in Normal,

Tangential and Rotational Directions in DEM
Procedure of Numerical Calculation of Pore Water Pressure

124
FIGURE 7.2 Schematic Illustrations of Measurement Sphere and Cross Section Areas
The detailed procedure of the fluid-coupled discrete element method is highlighted. First,
the measurement sphere, which assigns the pore water pressure, is introduced, and then the
calculation scheme is shown. The measurement sphere surrounds and coincides with the
center of each ball element with the radius R (FIGURE 7.2). The radius of the measurement
sphere is twice of that of the ball element, i.e. 2R. The mean velocity and center position of
the ball and wall elements that intersect a given measurement sphere are calculated as
follows:
∑𝑁 𝑉𝑖 ∑𝑁 𝑋𝑖
̅𝑖 =
𝑉 , 𝑋̅𝑖 = (𝑖 = 1,2,3) (27)
𝑁 𝑁
in which N is the total number of ball and wall elements contained in the measurement sphere,
Vi is the translational velocity, and Xi is centroid location of ball and wall elements. The
relative velocity and position of ball and wall elements from the mean values could be written
as:
𝑉̃𝑖 = 𝑉𝑖 − 𝑉
̅𝑖 , 𝑋̃𝑖 = 𝑋𝑖 − 𝑋̅𝑖 (𝑖 = 1,2,3) (28)
In this model, the relative velocity is estimated by using strain rate tensor 𝛽̇𝑖𝑗 . It’s
expressed as:
𝑉̃𝑖′ = 𝛽̇𝑖𝑗 𝑋̃𝑗 (𝑖 = 1,2,3) (29)

125
The strain rate tensor 𝛽̇𝑖𝑗 that minimizes the difference between relative velocity 𝑉̃𝑖
and the estimated relative 𝑉̃𝑖′ is computed using the least square method, in which the
following nine equations are resolved. In this calculation, LU decomposition based on the
Crout’s algorithm, and backward and forward substitutions were conducted.
∑ 𝛽̇𝑖𝑘 𝑋̃𝑘 𝑋̃𝑗 = ∑ 𝑉̃𝑖 𝑉̃𝑗 (𝑖, 𝑗 = 1,2,3) (30)

𝑁 𝑁
The volumetric strain during the increment of time, Δt is obtained like:
̇ + 𝛽̇22 + 𝛽̇33 )∆𝑡

𝜀𝑣 = 𝜀̇𝑣 ∆𝑡 = (𝛽11 (31)
The changes in pore-water pressure due to the volume change is given by the product of
volumetric strain and the modulus of compressibility of water, and it is applied as forces to
the centers of ball and wall elements, which are within a given measurement sphere as
follows:
𝐹𝑏𝑎𝑙𝑙 = 𝑃𝐴′ 𝑆 = (𝑃𝐴 + ∆𝑃)𝑆 = (𝑃𝐴 + 𝐸𝑤 𝜀𝑣 )𝜋𝑟 2 (32)
𝐹𝑤𝑎𝑙𝑙 = 𝑃𝐴′ 𝑆 ′ = (𝑃𝐴 + ∆𝑃)𝑆 ′ = (𝑃𝐴 + 𝐸𝑤 𝜀𝑣 )𝜋[(2𝑟)2 − 𝐿2 ] (33)
in which PA is the pore-water pressure of the measurement sphere A, ΔP is incremental
changes in pore-water pressure within Δt, S and S' are the areas on which pore-water pressure
acts (FIGURE 7.2), Ew is the modulus of compressibility of water. The fluid-flow due to the
difference of pore-water pressure values between the neighboring measurement spheres is
given based on the Darcy’s law by:
2
𝑃𝐵′ − 𝑃𝐴′
∆𝑄𝐴𝐵 = 𝑘 ∙ 𝜋𝑟 ∆𝑡 (34)
𝜌 ∙ 𝑔 ∙ 𝐿𝐴𝐵
in which PA' and PB' are the pore-water pressure of the measurement spheres A and B, and g
is the gravitational acceleration, LAB is the distance between the measurement spheres A and
B. The total change in pore-water pressure (PA) because of the fluid-flow is given by:
126
𝐸𝑤 ∆𝑄𝐴𝐵
𝑃𝐴 = ∑ ∆𝑃𝐴𝐵 = ∑ (35)
𝑉𝐴
𝑁 𝑁
where VA is the volume of the measurement sphere A.
Simulated Measurement of Pore Water Pressure in Asphalt Mixture
Digital Model and Properties Used in DEM Simulation
A dynamic load was applied on the asphalt mixture model. The simulated asphalt
mixture follows an elastic model. The numerical specimens were cylindrical in shape (0.10 m
in diameter and 0.20 m in high) formed by three wall elements, including a circumferential
wall cylindrical in shape, upper end and lower end walls (FIGURE 7.3). These wall elements
were not deformed during the numerical simulation. The upper and lower end walls moved
vertically, and the circumferential wall cylindrical in shape contracted or expanded radially.
The ball elements had the uniform size of 0.015 m in diameter. The packing of the ball
elements was conducted such that certain numbers of ball elements with a reduced diameter
of 0.0031 m (0.625 times of 0.015 m) were created to effectively make assembly at the
positions determined by the pseudo-random numbers within the cylindrical numerical space
surrounded by wall elements. After that, the diameter of ball elements created within the wall
elements was restored to 0.015 m to form the designated specimens. Initial confining stress of
20 kPa was given to normally (isotropic) consolidate the numerical specimens, then
compressed with a constant axial speed of 0.01 m/sec (the upper end wall elements moved
down vertically). The input parameters used are given in TABLE 7.1. The bulk permeability
used in the simulation was the input value that was selected to be similar to the one found in
127
the sandy soils. The bulk permeability was constant all through in the simulation, neglecting
the changes in the void ratio in microscopic scale due to deformation of the assembly of ball
elements during compression.
FIGURE 7.3 Numerical Specimen for Undrained Triaxial Compression Test
TABLE 7.1 Input Parameters of The Simulated Specimen

Normal elastic spring stiffness 5.00×106 N/m
Tangential elastic spring stiffness 1.25×106 N/m
Rotational elastic spring stiffness 7.03×101 Nm
Normal viscous damping coefficient 2.27×105 Ns/m
Tangential viscous damping coefficient 1.13×105 Ns/m
Rotational viscous damping coefficient 1.00×10−3 Nms
Frictional coefficient of wall element 0.01
Passion radio 0.30
128
Confining pressure 2.00×104 Pa

Modulus of compressibility of water 2.10×106 Pa
Axial loading speed 1.00×10−2 m/s
Bulk permeability 1.00×10−3 m/s
Thermal expansion at 25oC 12×10−6K-1
Thermal conductivity at 25oC 0.75 W/(m.K)
The simulated asphalt mixture follows an elastic model; properties of the model are
summarized in TABLE 7.1.
Result and Analysis
The change of pore water pressure is linear along with dynamic load growth (see
FIGUER 7.4). Both positive and negative pore water pressures were produced in the process
of dynamic load applying repeatedly. Such an erosion of high pore water pressure on the
pavement will spray outward a mortar with a scoured material from crevice, resulting in peel
off and loose of asphalt surface. Then the performance of the mixture will rapidly decline in
the overall structure, and damage will occur.

129
Pore Water Pressure (MPa)
Dynamic Load (MPa)
FIGURE 7.4 Pore Water Pressure vs. Dynamic Load
As a result, calculated pore water pressure decreases along with an increasing elastic
modulus (see FIGURE 7.5). Air voids of the model equals to 8%. Pore water pressure that
calculated from a model with a 2500 MPa elastic modulus equals to 0.31 MPa, this value
increases to 0.59, 0.92 and 1.8 MPa when the elastic modulus was changed to 1500, 1000 and
500 MPa, respectively (See TABLE 7.2).
TABLE 7.2 Elastic Modulus and Pore Water Pressure

Elastic Modulus (MPa) Air Voids Pore Water Pressure (MPa)
2500 0.08 0.31
500 0.08 1.80
1000 0.08 0.92
1500 0.08 0.59
130
2
1.8 1.8
1.6
1.4
1.2
1
0.92
0.8
0.6 0.59
0.4
0.31
0.2
0 Elastic Modulus (MPa)
0 500 1000 1500 2000 2500 3000
FIGURE 7.5 Pore Water Pressure vs. Elastic Modulus in Compression Tests, Air void =
8%
1.1
1
0.9 E=1500MPa
0.8 E=2000MPa
E=2500MPa
0.7
0.6
0.5 Poisson Ratio

0.18 0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34 0.36 0.38 0.4 0.42
FIGURE 7.6 Pore Water Pressure vs. Elastic Modulus with Varied Poisson Ratios
In order to analyze how Poisson ratio effect pore water pressure on asphalt mixture,
different Poisson ratios have been set to the digital model. Compare with influence of elastic
modulus, Poisson ratio rarely affects pore water pressure (see FIGURE 7.6).
As we known, there is pore water in the pavement structure. Under high traffic loads,
pore water pressure becomes an important reason of moisture damage. Many properties of
131
asphalt mixtures are closely related to air voids. Shown in FIGURE 7.7, Hydraulic
conductivity is very small when air voids is smaller than 8%, and become pretty high when
air voids is larger than 15%. That means water can flow freely and be discharged easily when
air voids is larger than 15%. So air voids significantly affect pore water pressure; moisture
damage rarely exists when air voids is smaller than 8% or larger than 15%. It’s clearly in
FIGURE 7.8 that pore water pressure changed visibly when air voids equals to 8%.
Hydraulic conductivity (mL/minute)
600
500
400
300
200
100
0 Air Voids
0 0.05 0.1 0.15
FIGURE 7.7 Hydraulic Conductivity vs. Air Voids

132
0.84
0.82
0.8
0.78
0.76
0.74
0.72
0.7
0.68
0.66 Air Voids

0.01 0.1 1
FIGURE 7.8 Pore Water Pressure vs. Air Voids.
Influence of Entrained Air Voids on Pore Water Pressure and Shrinkage on Asphalt
Concrete
Numerical Analysis
Shrinkage of the asphalt binder in asphalt concrete is restrained by the stiff,
non-shrinking aggregates. Measurements of asphalt concrete shrinkage presented are in fact
the result of the interaction between a shrinking asphalt binder, whose mechanical properties
change during hardening, and the inert inclusions. A challenge in the modeling of autogenous
shrinkage is to calculate shrinkage of asphalt concrete based on shrinkage of the
corresponding asphalt binder. If this would be accomplished, it would save experiments on
asphalt mixture. In fact, the autogenous deformation of mixtures can be predicted from the
133
autogenous shrinkage of the corresponding asphalt binder. There is no agreement on the
feasibility of this approach. According to Tazawa et al. (2000), autogenous shrinkage of
asphalt concrete at 28 days can be calculated from autogenous deformation of asphalt binder
with a simple composite model (85). On the other hand, Hammer et al. (2002) excluded that
shrinkage of a concrete mixture can be calculated from shrinkage of an asphalt binder using
composite models (86).
In the following calculations, asphalt concrete is assumed to consist of two phases,
namely aggregate particles dispersed in an asphalt binder matrix. If the shrinkage of the
asphalt binder is greater than that of the aggregates, the aggregates restrain the shrinkage of
the asphalt binder. According to this approach, asphalt concrete shrinkage is a function of the
asphalt binder shrinkage, the aggregate shrinkage, the stiffness of the asphalt binder and of
the aggregate, and the aggregate volume concentration.
Some attempts (Pickett 1956, Hansen & Nielsen 1965, Hobbs 1969) have been made to
obtain expressions for the dependence of the asphalt concrete shrinkage upon the aggregate
volume concentration and the properties of the constituents (87, 88, 89). However, exact
solutions for asphalt concrete shrinkage cannot be obtained because of the geometric
complexity of concrete and the many interacting particles. In order to obtain a solution,
simplified assumptions are made and as a consequence the solutions are only approximate.
Analytical models for asphalt concrete shrinkage are discussed in (Hobbs 1974) (90).
Basic assumptions are as follows:
1) Asphalt concrete consists of two homogenous phases, aggregates and asphalt binder.
2) Aggregate and asphalt binder behave elastically.

134
3) Elastic properties are not influenced by shrinkage, i.e. micro-craking of the asphalt binder,
if it occurs, does not reduce the stiffness.
4) The total volume occupied by shrinkage cracks is independent of aggregate grading.
In the following, Pickett’s model (Pickett 1956) and Hobbs’ model (Hobbs 1969) will be
discussed and applied to quantify autogenous shrinkage of asphalt concrete (87, 89).
Pickett’s model
Pickett (1956) derived an expression for the effects of aggregates on concrete shrinkage
(87). The formula is derived by considering the restraining effect of one small spherical
aggregate particle embedded in a large body of shrinking concrete. The concrete surrounding
the aggregate particle is considered as a homogeneous material and both the particle and the
concrete are assumed to be elastic. As further particles are added, the elastic properties and
the shrinkage are recalculated, while the body is still considered to remain homogenous.
Integration and a number of assumptions lead to the expression:
𝜀𝐶 = 𝜀𝑃 ∙ (1 − ∅𝐴 )𝛼𝑃 (36)
where 𝜀𝐶 (m/m) is the shrinkage of concrete, 𝜀𝑃 (m/m) is the shrinkage of the asphalt
binder, ∅𝐴 (m3/m3) the volume fraction of the aggregates and 𝛼𝑃 is a parameter defined as:
3 ∙ (1 − 𝑉𝐶 )
𝛼𝑃 = (37)
1 + 𝑉𝑐 + 2 ∙ (1 − 2𝑉𝐴 ) ∙ 𝐸𝐶 ⁄𝐸𝐴
where 𝑉𝐶 and 𝑉𝐴 are the Poisson’s ratio of the concrete and of the aggregates, 𝐸𝐶 and 𝐸𝐴
(MPa) are the elastic moduli of the concrete and of the aggregates, respectively.
According to Equation (1) and (2) 𝜀𝐶 = 𝜀𝑃 only when the ratio 𝐸𝐶 /𝐸𝐴 becomes ∞.
Hobbs’ model
Hobbs’ model (89, 90), also called C&CA model, was developed for drying shrinkage,
135
where the elastic modulus of the paste is constant. According to the model, the shrinkage of
the concrete, 𝜀𝐶 (m/m), is equal to:

𝜀𝑃 ∙ (1 − ∅𝐴 ) ∙ (𝐺𝑃 + 𝐺𝐴 ) + 2 ∙ 𝜀𝐴 ∙ ∅𝐴 ∙ 𝐺𝐴
𝜀𝐶 = (38)
𝐺𝑃 + 𝐺𝐴 + ∅𝐴 ∙ (𝐺𝐴 − 𝐺𝑃 )
where 𝜀𝑃 (m/m) is the shrinkage of the cement paste, 𝜀𝐴 (m/m) is the shrinkage of the
aggregates, 𝐺𝐴 (MPa) is the shear modulus of the aggregates, and 𝐺𝑃 (MPa) the shear
modulus of the paste.
This model is more general than Pickett’s, taking into account the possible shrinkage of
aggregates upon drying. Another difference is that the elastic modulus of the paste is used in
the computation in Hobb’s model, whereas the modulus of concrete is used in Pickett’s
model.
Traditional Method of Experiment Test
Autogenous deformation of concrete varies enormously in magnitude and even in sign
between different publications (Hammer et al. 2002) (86). Since concrete is the material
which is used in practical application, the principal aim of autogenous deformation
measurements on cement paste is to understand and possibly predict autogenous deformation
of concrete made with the same paste. However, this approach suffers of a series of
drawbacks of which one of the most serious lies in the measuring methods.
Cement paste
Linear measurement of autogenous deformation is frequently performed by placing the
cement paste in a rigid mold with low friction. The length change of the asphalt binder is
recorded by a displacement transducer at the end of the specimen, see FIGURE 7.9.
136
FIGURE 7.9 Equipment for Measuring Linear Autogenous Deformation of Cement

Paste (91)
Concrete
Different linear methods have been used to measure autogenous deformation of concrete
on beams, slabs or cylinders (FIGURES 7.10-7.14):
1) Cast-in nails through a hole in the middle of the end plates with the nail heads embedded
in a concrete beam (FIGURE 7.10).
2) Moveable endplates with plugs in a concrete beam (FIGURE 7.11).
3) Horizontal transverse cast-in bars through a concrete beam (FIGURE 7.12).
4) Vertical cast-in bars in a concrete slab (FIGURE 7.13).
5) Cast-in strain gage (Hanehara et al. 1999).
6) Metal plates placed on top of cylinders (FIGURE 7.14).
For all the measuring systems presented, the main problems to overcome are ensuring a
good contact between the measuring points and the concrete, minimizing restraint and
moisture loss and keeping the temperature constant. The systems presented in FIGURES
137
7.9-7.12, solve these problems in a similar way: embedment of rods or bars in the fresh cast,
to which later the measuring points are fixed; reduction of friction on the lower face with
low-friction foils; moisture loss is reduced by a plastic or aluminum foil on the upper face
after casting; constant temperature is obtained with temperature-controlled baths or by
circulation of cooling liquid in the mold. The system in FIGURE 7.14, on the other hand,
presents different solutions and is analogous to the one designed for cement paste.
FIGURE 7.10 Measuring System with Cast-in Nails (92)
FIGURE 7.11 Measuring System with Moveable End Plates in a 40×40×160 mm3 Beam
(93)
138
FIGURE 7.12 Measuring System with Horizontal Cast-in Bars in a 150×150×1000 mm3
Beam (94)
FIGURE 7.13 Measuring System with Vertical Cast-in Bars in a 270×270×100 mm3 Slab
(95)
139
FIGURE 7.14 Measuring System with Flexible Tubes (diameter 100 mm and length 375
mm) and Vertical Length Measurement (96)
As pointed out in the previous section about cement paste, linear measurements of
autogenous deformation on a concrete mix before setting are not only associated with a large
scatter, but are also intrinsically ambiguous, since the length is not defined in a fluid system.
Start of the measurement should coincide with setting, but exact determination of the
fluid-solid transition is difficult and subjective. A possible approach, which was followed in
the measurements is to start measuring before setting and to register at the same time the
self-induced stress of a restrained specimen. The deformations are then zeroed at the moment
when a stress is recorded in the restrained specimen. This ensures that only the deformations
measured on a solid system, which is able to exert an external stress, are taken into account.
Simulated Shrinkage Process Using DEM
The shrinkage is measured on a digital model that represents beams in a steel mold with
140
length/thickness/height of 280/100/100 mm. The shrinkage is defined as the linear
autogenous deformation at horizontal direction, it’s measured as horizontal movement of two
“nails” placed centric in both ends of the specimen. The nails are made of 3mm thick steel
rods with a 15 × 15mm steel plate soldered to the end of each rod, placed 30mm in the
concrete. The other end is wedged and screwed into inductive displacement transducers
through a hole in the ends of the mold. The transducers are fixed to the mold. All data
(shrinkage and pore water pressure) were recorded every second minute. The water loss may
also be recorded.
Different with experiment method, material used in simulation is asphalt concrete
instead of cement concrete. PG 64-22 asphalt binder is used. Elastic and simple viscoelastic
contact models are adopted in DEM simulations. Contacts properties and other governing
parameters in the DEM model in PFC 3d are summarized in TABLE 7.3.
TABLE 7.3 Contacts Properties and Other Governing Parameters Specified in DEM
Model
Normal elastic spring stiffness 5.00×106 N/m
Tangential elastic spring stiffness 1.25×106 N/m
Rotational elastic spring stiffness 7.03×101 Nm
Normal viscous damping coefficient 2.27×105 Ns/m
Tangential viscous damping coefficient 1.13×105 Ns/m
Rotational viscous damping coefficient 1.00×10−3 Nms
Frictional coefficient of wall element 0.01
Passion radio 0.30
Confining pressure 2.00×104 Pa
Modulus of compressibility of water 2.10×106 Pa
Axial loading speed 1.00×10−2 m/s
Bulk permeability 1.00×10−3 m/s
Thermal expansion at 25oC 12×10−6K-1
Thermal conductivity at 25oC 0.75 W/(m.K)
141
Result and Analysis
12
10
Shringkage (10-3)
6
No Air 1
4
No Air 2
2
Time
0
(hour)
0 5 10 15 20
FIGURE 7.15 Shrinkage of Asphalt Concrete without Air Entraining Mixture
Firstly, pore water pressure and deformation are measured on two asphalt concrete
samples, one with AEM and the other one without AEM. Tests on asphalt mixture without
AEM were repeated once, results of shrinkage of the tests are shown in FIGURE 7.15.
It is a difficult task to generate a stable air pore system in the asphalt concrete and to
reach the target air content of 8-15%. The air content was calculated from the difference in
density assuming that the asphalt binder sample without AEM contains no air. The asphalt
binder sample showed sufficient fluidity to allow good compaction during filling. The asphalt
binder sample did not show significant bleeding, and thus, the deformation test was
performed without rotation.

142
Pore Water Pressure (mm water column) 500
-500
-1000
No Air 1
-1500
No Air 2
-2000
-2500
Time
-3000
(hour)
0 2 4 6 8 10 12 14 16 18 20
FIGURE 7.16 Pore Water Pressure of Asphalt Binder without Air Entraining Mixture
Pore water pressures of the asphalt binder specimen are measured from simulations.
Pore pressure at a given point can be measured by the height of water in a standpipe located
at that point. Pore pressures are often indicated in this way on diagrams.
𝑢
ℎ𝑤 = (39)
𝛾𝑤
where u is the pore water pressure, 𝛾𝑤 is unit weight of water, and the height of the water
column is the pressure head (hw). FIGURE 7.16 shows the results of the tests on specimens
without air entraining mixture.
Comparing previous work on asphalt mixture without AEM and with varied air voids,
FIGURE 7.17 and FIGURE 7.18 show that the influence of air voids on shrinkage are as
follows: higher rate of shrinkage before the “knee point”, (i.e. in the few hours before
autogenous shrinkage rate becomes significantly lower than later), earlier arrival of the
knee-point, and lower rate of shrinkage beyond the knee-point. Since the time of full collapse
depends on the stiffness evolution of the asphalt binder, the simple explanation can be that
143
presence of air voids results in a lower stiffness. This suggestion should be further explored
by more systematic and accurate measurements.
Another feature is that shrinkage of the asphalt binder without AEM increases distinctly
along with the duration of time beyond the knee-point, while shrinkage of the asphalt binder
with AEM does not. This is probably directly caused by the buffer effect of air pores. The
vacuum effect drives the PWP downwards, but the air pores will release the pressure through
expansion. The result is reduced rate of PWP accompanied with reduced shrinkage rate.
16
14
12
Shringkage (10-3)
10
6 Air Voids = 2.9%

Air Voids = 3.7%
4 Air Voids = 4.5%
2
Time
0
(hour)
0 2 4 6 8 10 12 14 16 18
FIGURE 7.17 Shrinkage of Asphalt Binder with Varied Air Voids
144
Pore Water Pressure (mm water column) 500

Air content = 2.9%
0 Air content = 3.7%
Air content = 4.5%
-500
-1000
-1500
-2000
-2500 Time
0 2 4 6 8 10 12 14 16 18 (hour)
FIGURE 7.18 Pore Water Pressure of Asphalt Binder with Varied Air Voids
Summary
Air voids in asphalt binder seems to contribute to an increased rate of autogenous
shrinkage in the time before the “knee-point”, and an earlier appearance of the knee-point.
Through the period shortly after, air pores alter the pore water pressure evolution in that they
act as buffer, and thereby reduce the pressure decrease, and the subsequent autogenous
shrinkage as well as the friction against panels in slip forming.
The scatter observed may be related to inaccuracy in measurement, but also to the
differences in air content. It can be seen that there is a fairly good coincidence between the
knee-point and the point when PWP bends of as expected, which suggests that inaccuracy in
measurements does not dominate. This would in turn cause faster hydration, which may
explain both earlier bending off of and higher PWP rate.

145
8 CONCLUSIONS
DEM has been used to simulate the behavior of asphalt mixture with different gradations,
different film thicknesses, and different asphalt contents under compressive loads. Optimal
gradation and optimum values of asphalt content and film thickness can be determined
through DEM simulations.
Gradation is determined based on a simulated packing process. The evaluation of
aggregate packing begins with the comparison of two different states of aggregate structures.
We assume the newly added aggregates are always equal in size to or smaller than aggregates
in a former state. After gravity and loading cycles are applied, the contact force and
unbalanced force were recorded. Then finer aggregates were added into the structure. After
mixing and compacting the structure into equilibrium, volume and porosity of the new
structure were calculated. By comparing five digital models, an optimal model which has the
lowest porosity was selected.
The asphalt content was determined using an approach that is based on modeling the
inter-aggregate contact with a binder layer. The total surface area is related to the gradation,
particle shape, angularity and texture of the aggregates. Once the gradation is determined, the
total surface area, asphalt content and the binder film thickness can be calculated. Ten digital
mixtures were built in software PFC3d, and the asphalt content ranges from 3.9% to 5.9%. As
compressive loading is applied, the number of contacts between aggregates and asphalt
binder were recorded. This number was used as a judgment to tell whether the mixtures were
damaged due to cracking. The mixture that reached the highest peak strength was considered
as optimal. Therefore, optimal asphalt content determined by this process was 5.3%. The
146
procedure described in this study determines the optimum asphalt content by adding the
capability of internal deformation mechanism (breaking of bonds and loss of contacts),
eliminating laborious and tedious laboratory experiments.
Based on optimal gradation and optimal asphalt content determined by computational
methods, a particular mixture was simulated with sphere aggregates and asphalt binder. A
simple visco-elastic model was applied to contacts between aggregates and asphalt binder in
PFC3d. The thickness of the asphalt binder layer was calculated to be 9.2 micrometers
associated, consistent with data in literature.
The indirect tensile strength of HMA with four different shapes of aggregate was
evaluated in simulated test to evaluate the relative quality of materials. The image analyzer
was a useful tool for quantifying the morphological characteristics of coarse aggregates.
Image evaluation provided quantitative indices, such as geometric measurement and angle
rotation of granular materials. Data showed that the morphological characteristics of coarse
aggregates correlated well with the results of other indirect tests such as the particle index.
Cubical particles possessed the best resistance over the other shapes. The particle index (PI)
was shown to be an adequate measure of the combined contribution of particle shape,
angularity, and surface texture to the stability of an aggregate. The PI value correlated well to
aggregate geometric characteristics including elongation ratio, flatness ratio, shape factor, and
sphericity. The particle shape determined how aggregate was packed into a dense
configuration and also determined the internal resistance of a mix. The morphological
characteristics of coarse aggregates found from both digital test and image analysis were in
good agreement with the engineering properties of HMA mixtures. This study presents a
147
precise method to evaluate the aggregate characteristics in a HMA mix and demonstrates
their effects on pavement performance. The results of this study would provide useful
guidelines for highway engineers to construct a long-lasting pavement.
In this study, considering the necessity of preservation of the compaction temperature,
the effect of various temperatures on Hot Mix Asphalt (HMA) samples properties has been
evaluated. As well, to evaluate the effect of this parameter on different grading, two different
grading have been used and samples were compacted at various temperatures.
Based on the results of asphalt mixture without AEM and with varied air voids, one
could conclude that the influence of air voids on shrinkage are as follows: higher rate of
shrinkage before the knee point, earlier arrival of the knee-point, and lower rate of shrinkage
beyond the knee-point. Another feature is that shrinkage of the asphalt binder without AEM
increases distinctly along with the duration of time beyond the knee-point, while shrinkage of
the asphalt binder with AEM does not. This is probably directly caused by the buffer effect of
air pores. The vacuum effect drives the PWP downwards, but the air pores will release the
pressure through expansion. The result is reduced rate of PWP accompanied with reduced
shrinkage rate.
148
REFERENCES
1. Roberts, F.L., Prithvi S. Kandhal, E. Ray Brown, D.Y. Lee and Thomas D. Kennedy. Hot Mix Asphalt
Materials, Mixture Design and Construction. First Edition, NAPA Education Foundation, Maryland,
1991.
2. Cundall, P. A., and Strack, O. D. L. A discrete numerical model for granular assemblies. Geotechnique,
Vol. 59, 1979, pp. 47-51.
3. Nukunya, B. Aggregate Gradation Characterization Factors and Their Relationships to Fracture Energy
and Failure Strain of Asphalt Mixtures. Journal of the Association of Asphalt Paving Technologist, Vol.
33, 2002, pp. 56-67
4. Cundall, P.A. and O.D.L. Strack, A Discrete Numerical Model for Granular Assemblies.
Geotechnique,Vol. 29, 1979, pp. 47-65.
5. Bernardita Lira, Denis Jelagin, “Gradation-based framework for asphalt mixture.” Materials and
Structures, 2012, pp 21-38.
6. Goode, J. E., and Lufsey, L. A. “A New Graphical Chart for Evaluating Aggregate Gradation.”
Association of Asphalt Paving Teclmologists, 1962.
7. American Society of Testing and Materials (ASTM) C117: Standard Test Method for Materials Finer
than 75 μm (No. 200) Sieve in Mineral Aggregates by Washing. Philadelphia, PA. 1997.
8. American Society of Testing and Materials (ASTM) C136: Standard Test Method for Sieve Analysis of
Fine and Coarse Aggregates. Philadelphia, PA. 1997.
9. Gradex 2000 Particle Size Analyzer Catalog 229, ROTEX, INC., Cincinnati, OH
10. Sales Literature on the VDG-40, Obtained from EMACO, Ltd., Montreal, Canada
149
11. Standard for Granulometric Analysis of Aggregate and Determination of Flatness and Elongation of
Particles by Video Granulometer, Laboratoire Central des Ponts et Chaussées (the French Public
Works Laboratory), 1993.
12. Harlow, R. L., “Design of a Particle Size Analyzer for Gravel Sized Aggregate.” Master’s Thesis,
University of Texas at Austin, 1972.
13. Aljassar, A., and Haas, R., “Toward Automating Size-Grading Analysis of Mineral Aggregate.”
Transportation Research Record 1437 Transportation Research Board, Washington, D.C., 1994, pp.
35-42.
14. Surgrue, S., and Row, G., “True Resolution in Laser Diffraction Particle Sizing: Who needs it and
where does it come from?” Technical Monograph, Coulter Corporation.
15. Thomas, T. W. White, T. D., and Kuczek, T., “Siliceous Content Determination of Sands Using
Automatic Image Analysis.” Transportation Research Record 1437 Transportation Research Board,
Washington, D.C., 1994, pp. 51-58.
16. Alba, Felix, “Phase I Final Report for Developments of a Prototype for On-Line Real Time
Measurement and Control of Aggregate Gradation in Asphalt Plants.” Report presented to the
SHRP-IDEA Program Manager, May 1992.
17. Alba, Felix, “Phase II (Stage 2) Report for Developments of a Prototype for On-Line Real-Time
Measurement and Control of Aggregate Gradation in Asphalt Plants.” Report presented to the
SHRP-IDEA Program Manager, September 1993.
18. Yu, H., and S. Shen Investigation of aggregate packing using discrete element modeling. Geotechnical
Special Publication, 2010, pp.33-58

150
19. Brown, E. R., Murphy N. E., Yu L. and Mager S., “Historical Development of Asphalt Content by the
Ignition Method,” Asphalt Paving Technology, Association of Asphalt Paving Technologists, St. Paul,
MN, 1995, pp. 1-35.
20. Behrens, M. L., Dvorak B. I. and Woldt W. E., “Comparison of Asphalt Extraction Procedures:
Implications of Hidden Environmental and Liability Costs,” Transportation Research Record 1661,
Transportation Research Board, Washington, DC, 1999, pp. 46-53.
21. Brown, E. R. and Mager S., “Asphalt Content by Ignition Round Robin Study,” Report No. 95-3,
National Center for Asphalt Technology, Auburn, AL, 1995.
22. McKeen, R. G., “Asphalt Content by Ignition: Round Robin Experiment,” New Mexico State
Highway and Transportation Department, Santa Fe, NM, December 1997, pp. 13-14.
23. Burati, J.L. and Connell J., “Study of Alternative Procedures for Controlling and Accepting Asphalt
Mixtures,” FHWA-SC-96-05, Washington, DC, 1996.
24. Howard, C. P. and Burati J. L., “Study of the NCAT Ignition Oven for Control of Asphalt Mixtures,”
FHWA-SC-96-06, Washington, DC, 1996.
25. Lee, R. E. and Tahmoressi M., “Using Ignition Ovens to Determine Asphalt Content and to Prepare
Aggregates for Sieve Analysis,” Texas DOT, Austin, TX, 1996.
26. Mallick, R. and Brown E. R., “Limited Round Robin Asphalt Content Test Using Troxler Furnace,”
Report No. 99-6, National Center for Asphalt Technology, Auburn, AL, 1999.
27. American Society of Testing and Materials (ASTM) T308: Determining the Asphalt Binder Content of
Hot Mix Asphalt (HMA) by the Ignition Method. Philadelphia, PA. 2013
28. Schroer J., “Asphalt Content of Dolomite Mixtures Using the Troxler NTO,” Missouri DOT, August
21, 2002.
151
29. Prowell B. D. and Schreck R., “Asphalt Content by the Ignition Method: Virginia’s Experience,” Eight
International Conference on Asphalt Pavements, University of Washington, Seattle, August 10-14,
1997.
30. “Asphalt Content by the Ignition Method, VTM 102” Orientation School, Virginia DOT, Virginia
Transportation Research Council (VTRC), Virginia Asphalt Association (VAA), Virginia, April 1995.
31. Hurley G. C. and Prowell B. D., “Evaluation of Infrared Ignition Furnance for Determination of
Asphalt Content,” Transportation Research Record 1861, Transportation Research Board, Washington,
DC, 2003, pp. 44-50.
32. Prowell B. D., “Evaluation of Infrared Ignition Furnace for Determination of Asphalt Content,” Report
No. 02-05, National Center for Asphalt Technology, Auburn, AL, 2002, pp. 5-10.
33. Williams S. G. and Hall K. D., “A Comparison of Ovens Used in Determination Binder Content by the
Ignition Method.” Journal of Association of Asphalt Paving Technologists, Proceedings of the
technical Sessions, Vol. 70, 2002, pp. 712-728.
34. Mallick R., Brown E. R. and McCauley N., “Effect of Ignition Test for Asphalt Concrete on Aggregate
Properties.” presented at Transportation Research Board 77th Annual Meeting, Washington D.C., 1998,
pp. 1-15.
35. Prowell B. D. and Carter C. B., “Evaluation of the Effect on Aggregate Properties of Samples
Extracted Using the Ignition Furnance.” Virginia DOT, Interim Report, VTRC 00-IR1, April 2000, pp.
14-17.
36. Rogers C., Williams D., Gorman B., Sahsuvar H., Greenbury B. and Masliwec T., “Ignition Oven
Problems with Quartz Rich Aggregates Materials.” International Center for Aggregates Research,
Twelfth Annual Symposium Proceedings, Denver, CO, April 4-7, 2004.

152
37. American Society of Testing and Materials (ASTM) D6307: Standard Test Method for Asphalt
Content of Hot-Mix Asphalt by Ignition Method. Philadelphia, PA. 1998
38. Lavin, P. “Asphalt Pavements: A Practical Guide to Design, Production and Maintenance for
Engineers and Architects”. 2003
39. Van der Poel, C. A “General System Describing the Visco-Elastic Properties of Bitumen and Its
Relation to Routine Test Data”. J. Appl. Chem., Vol.4, May 1954, pp. 221-236
40. Hills, J.F. “The Creep of Asphalt Mixes.” Journal of the Institute of Petroleum, Vol. 59, 1973.
41. Cheung, C.Y. and Cebon, D. (1997). “Thin Film Deformation Behavior of Power Law Creeping
Materials.” Journal of Engineering Mechanics, Vol. 123-11, pp. 1138-1152.
42. Uddin, W., “A Micromechanical Model for Prediction of Creep Compliance and Viscoelastic Analysis
of Asphalt Pavements.” Transportation Research Record 1596, Transportation Research Board,
Washington, DC, 1998, pp. 43-56.
43. Zhong, Xiaoxiong., Chang, Ching S. “Micromechanical Modeling for Behavior of Cementitious
Granular Materials”. Journal of Engineering Mechanics, November, 1999.
44. Levenberg,E., Uzan, J. , “Triaxial Small-Strain Viscoelastic Modeling of Asphalt Aggregate Mixes.”
Mechanics of Time-Dependent Materials, 2004, Vol 8, pp. 365–384.
45. Mazzotti, Claudio and Savoia, Marco. “Nonlinear Creep Damage Model for Concrete under Uniaxial
Compression.” Journal of Engineering Mechanics., September 2003, pp.1065-1075.
46. Huang, Baoshan., Mohammad, Louay N. and G. Wije Wathugala. “Application of a Temperature
Dependent Visco-plastic Hierarchical Single Surface Model for Asphalt Mixtures.” Journal of
Materials in Civil Engineering, March/April, 2004, pp.147-154.

153
47. Ghassan R. Chehab, and Y. Richard Kim. “Viscoelastoplastic Continuum Damage Model Application
to Thermal Cracking of Asphalt Concrete.” Journal of materials in civil engineering, 2005,
pp.384-392
48. Sepehr, K., Svec, O.J., Yue, Z.Q., and El Hussein, H.M. “Finite Element Modelling of Asphalt
Concrete Microstructure.” Computer Aided Assessment and Control of Localized Damage Proceedings
of the International Conference, 1994, p 225.
49. Bahia, H.U., Zhai, H., Bonnetti, K., and Kose, S. “Non-Linear Viscoelastic and Fatigue Properties of
Asphalt Binders.” Journal of the Association of Asphalt Paving Technologists, AAPT, 1999, Vol. 68,
pp. 1-34.
50. Abbas, A., Papagiannakis, A.T. and Masad, E. “Linear and Non-Linear Viscoelastic Analysis of the
Microstructure of Asphalt Concretes.” Journal of Materials in Civil Engineering, ASCE, 2004, Vol. 16,
pp. 133-139.
51. Saadeh, S., Masad, E., Stuart, K., Abbas, A., Papagainnakis, T., and Al-Khateeb, G. “Comparative
Analysis of Axial and Shear Viscoelastic Properties of Asphalt Mixes.” Journal of the Association of
Asphalt Paving Technologists, AAPT, 2003, Vol. 72, pp. 122-153.
52. Rothenburg, L., Bogobowicz, A., Haas, R., Jung, F.W., and Kennepohl, G. “Micromechanical
Modelling of Asphalt Concrete in Connection with Pavement Rutting Problems.” Proceedings of the
7th International Conference on Asphalt Pavements, 1992, pp. 230-245.
53. Chang, G. k., Meegoda, N. J. “Simulation of the Behaviour of Asphalt Concrete using Discrete
Element Method.” In: 2nd International Conference on Discrete Element Methods, M.I.T 1993
154
54. Buttlar, W. G., and You, Z. “Discrete element modelling of asphalt concrete: A microfabric approach.”
Transportation Research Record 1883, Transportation Research Board, Washington, DC, 2001, pp.
13-21.
55. Adhikari, S., and You, Z. “3D microstructural models for asphalt mixtures using X-ray computed
tomography images.” International Journal of Pavement Research and Technology, 2008, pp. 94–99.
56. Anderson, T. Balakrishnan, H. Parulkar, G. and Peterson, L. “Integrated Open Flow–GMPLS control
plane: an overlay model for software defined packet over optical networks.” Optics Express, Vol. 19,
Issue 26, 2011, pp. B421-B428
57. H Cleland, M Suzuki, N Ito, 2010, “Efficient Implementations of Molecular Dynamics Simulations for
Lennard-Jones Systems.” Progress of Theoretical Physics, Vol. 126, No. 2, 2011, pp. 203-235.
58. Itasca, PFC3d 4.0 Manual: Background and Theory.
59. Skyscan 1174 X-ray microtomograph – user guides.
60. Kim, Y.R., Shah, K.A, and Khosla, N.P, “Influence of Test Parameters in SHRP P07 Procedure on
Resilient Moduli of Asphalt Concrete Field Cores”, Transportation Research Record 1353,
Transportation Research Board, Washington, DC, 1992, pp.82-89.
61. Shklarsky, E. and Livneh, M. “The use of gravels for bituminous mixtures.” Proceedings of
Association of Asphalt Paving Technologists, 33, 23-65, 1964.
62. Krutz, N.C., and Sebaaly, P.E. “Effect of aggregate gradation on permanent deformation of asphaltic
concrete.” Proceedings of Association of Asphalt Paving Technologists, 1993
63. Li, M.C., and Kett, I. “Influence of coarse aggregate shape on the strength of asphalt concrete
mixtures.” Highway Research Record 178, 93-106, 1967.

155
64. Huber, G.A., and Heiman, G.H. “Effect of asphalt concrete parameters on rutting performance: a field
investigation.” Proceedings of Association of Asphalt Paving Technologists, Vol. 56, 1987, pp. 33-61.
65. Stephens, J.E., and Sinha, K.C. “Influence of aggregate shape on bituminous mix character.” Journal
of Association of Asphalt Paving Technologists, 47, 1978, pp. 434-456.
66. Yue, Z., Bekking, W., and Morin, I. “Application of digital image processing to quantitative study of
asphalt concrete microstructure.” Transportation Research Record 1492, Transportation Research
Board, Washington, DC, 1995, pp. 53-60.
67. Kuo, C.Y., Rollings, R.S., and Lynch, L.N. “Morphological study of coarse aggregates using image
analysis.” Journal of Materials in Civil Engineering, 10, 1998, pp. 135-142.
68. American Society of Testing and Materials (ASTM) D4791: Standard Test Method for Flat Particles,
Elongated Particles, or Flat and Elongated Particles in Coarse Aggregate. Philadelphia, PA. 2010
69. American Society of Testing and Materials (ASTM) D3398: Standard Test Method for Index of
Aggregate Particle Shape and Texture. Philadelphia, PA. 2000
70. American Society of Testing and Materials (ASTM) D4123: Standard Test Method for Indirect
Tension Test for Resilient Modulus of Bituminous Mixtures. Philadelphia, PA. 1995
71. American Society of Testing and Materials (ASTM) D6926: Standard Practice for Preparation of
Bituminous Specimens Using Marshall Apparatus. Philadelphia, PA. 2004
72. Bahia, H.U. “Recommendation s for Mixing and Compaction Temperatures of Modified Binders.”
National Cooperative Highway Research Program Report 9-10, Transportation Research Council,
Washington, D.C., 2000.
73. Huner, M.H., and Brown, E.R. “Effects of Re-heating and Compaction Temperature on Hot Mix
Asphalt Volumetric.” National Center for Asphalt Technology Report 01-04, Auburn, Ala, 2001.
156
74. Azari, H., McCuen, R.H., and Stuart, K.D. “Optimum Compaction Temperature for Modified Binders.”
Journal of Transportation Engineering, 129(5), 2003, pp. 531-537
75. Kennedy, T. W., Robbert F.L, and Mcgennis R. “effect of compaction temperature and effort on the
engineering properties of asphalt concrete mixtures.” Symposium on placement and compaction of
asphalt mixtures, 1982.
76. “Mix Design Method for Asphalt Concrete and Other Hot Mix Types.” The Asphalt Institute, Manual
series No 2 (MS-2), College Park, MD, 1979.
77. Sassa, K. “Motion of landslides and debris flows—prediction of hazard area.” In: Report of
Grant-in-Aid for Scientific Research, Japanese Ministry of Education, Science and Culture, 1988
78. Ishihara, K., Okusa, S., Oyagi, N., and Ischuk, A. “Liquefaction-induced flow slide in the collapsible
loess deposit in Soviet Tajik.” Soil Found Vol. 30, No. 4, 1990, pp. 73–89.
79. Eckersley, J.D., “Flowslides in stockpiled coal.” Eng Geol, Vol. 22, 1985, pp. 13–22.
80. Hutchinson, J.N., “A sliding-consolidation model for flow slides.” Can Geotech, Vol. J32, 1986, pp.
610–623
81. Okada, Y., Sassa, K., and Fukuoka, H., “Excess pore pressure and grain crushing of sands by means of
undrained and naturally drained ring-shear tests.” Eng Geol, Vol. 75, 2004, pp. 325–343.
82. Cundall, P.A., Strack, ODL., “A discrete numerical model for granular assemblies.” Géotechnique, Vol.
29, No. 1, 1979, pp. 47–65.
83. Hammer, T.A., “On the strain capacity and cracking mechanisms of high strength concrete at very
early age.” Proceedings of the sixth international conference on creep, shrinkage and durability of
mechanics of concrete and other quasi-brittle materials, Cambridge (MA), 2001, pp. 657–662.
157
84. Fossa, K.T., “Slipforming of Vertical Concrete Structures - Friction between concrete and slipform
panel, Dissertation, Department of Structural Engineering, The Norwegian University of technology,
2001.
85. Tazawa, E., Sato, R., Sakai, E., and Miyazawa, S. “Work of JCI committee on autogenous shrinkage.”
RILEM Workshop on Shrinkage of Concrete, Paris, 2000, pp. 21-40.
86. Hammer, T.A., Bjøntegaard, Ø., and Sellevold, E.J. “Internal curing – Role of Absorbed Water in
Aggregates.” Presented at ACI Fall 2002 Convention, Session High-Performance Structural
Lightweight Concrete, Phoenix, AZ, 2002.
87. Pickett, G. “Effect of aggregate on shrinkage of concrete and a hypothesis concerning shrinkage.” ACI
Journal, Vol. 52, No. 5, 1956, pp. 581-590.
88. Hansen, T.C., Nielsen, K.E.C., “Influence of aggregate properties on concrete shrinkage.” ACI Journal,
Vol. 62, No. 7, 1965, pp. 783-794.
89. Hobbs, D.W., “Bulk Modulus Shrinkage and Thermal Expansion of a Two Phase Material.” Nature,
Vol. 222, 1969, pp. 849-851.
90. Hobbs, D.W., “Influence of aggregate restraint on the shrinkage of concrete.” ACI Journal, Vol. 71,
1974, pp. 445-450.
91. Baron, J., and Buil, M. “Comments on the paper ‘Mechanical features of chemical shrinkage of
cement pastes’ bySetter, N and Roy, D.M., Cem Concr Res Vol. 9, No. 4, 1979, pp. 545-547.
92. Bjontegaard, Ø. “Thermal dilation and autogenous deformation as driving forces to self-induced
stresses in high performance concrete.” Dissertation, NTNU Division of Structural Engineering,
Trondheim, Norway, 1999.

158
93. Morioka, M., Hori, A., Hagiwara, H., Sakai, E., and Daimon, M. “Measurement of autogenous length
changes by laser sensors equipped with digital computer systems.” Workshop Autoshrink’98, London,
1999, pp. 191-200.
94. Lokhorst, S.J “Deformational behavior of concrete influenced by hydration related changes of the
microstructure.” Research Report, Delft University of Technology, Delft, Netherlands, 1998.
95. Holt, E.E, and Leivo, M.T. “Autogenous shrinkage at very early ages.” Workshop Autoshrink’98,
London, 1999, pp. 133-140.
96. Hansen, K.K, and Jensen, O.M. “Equipment for Measuring Autogenous RH-Change and Autogenous
Deformation in Cement Paste and Concrete.” Proc. Int. Res. Sem. on Self-desiccation and its
Importance in Concrete Technology, Lund, Sweden, 1997, pp. 27-30.

Digital Mix Design For Performance Optimization of Asphalt Mixture

Uploaded by

Copyright:

Available Formats

Digital Mix Design For Performance Optimization of Asphalt Mixture

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Digital Mix Design For Performance Optimization of Asphalt Mixture

Uploaded by

Copyright:

Available Formats

Digital Mix Design for Performance Optimization of

Dissertation submitted to the faculty of

Virginia Polytechnic Institute and State University

in partial fulfillment of the requirements for the degree of

Abbas, Montasir Mahgoub

November 21, 2014

Keywords: Discrete Element Method, Aggregate Gradation, Asphalt Content, Visco-elastic

Copyright 2014, Ying Li

gradations of aggregates. The optimum gradation could be determined by manipulating the

the linear autogenous deformation at horizontal direction.

my long-term career goals. He encouraged me to not only grow as an experimentalist and an

have otherwise been a somewhat stressful laboratory environment.

Current Technology Review .................................................................................................................. 9

Summary of Current Technologies ...................................................................................................... 23

Simulation of Aggregates Packing Process in PFC3d ......................................................................26

Ignition Method Test Procedures ........................................................................................................ 40

Procedure of Determination of Asphalt Content Using DEM .........................................................41

Calculation of Film Thickness ............................................................................................................ 47

Numerical Methods ............................................................................................................................. 56

Build a Digital Model by Using DEM................................................................................................59

Simple Viscoelastic Model .................................................................................................................. 62

Image Measurements of Coarse Aggregate ........................................................................................ 92

Calculation of Particle Index of Coarse Aggregates Based on DEM Simulations .........................94

Results and Analysis ......................................................................................................................... 101

Compression Tests Using X-ray Image Scanner ............................................................................101

Test and Results Analysis .................................................................................................................. 104

Analysis of Percent of Air Voids (VA)................................................................................................ 116

Analysis of Percent Voids Filled with Asphalt (VFA) ........................................................................ 117

Analysis of Flow................................................................................................................................ 118

Literature Review .............................................................................................................................121

Simulated Measurement of Pore Water Pressure in Asphalt Mixture .........................................126

Result and Analysis ........................................................................................................................... 128

Traditional Method of Experiment Test............................................................................................. 135

Simulated Shrinkage Process Using DEM ........................................................................................ 139

Result and Analysis ........................................................................................................................... 141

FIGURE 2.1: Gradex 2000 Particle Size Analyzer……………….……….………….……….…10

FIGURE 2.2: Gradex 2000 Sieve Configuration…………….………………….……….………12

FIGURE 2.3: Gradex 2000 Brush Cleaning Attachment…….……………………….…….……12

FIGURE 2.4: Product File Information from Gradex 2000……….………………….…….……14

FIGURE 2.5: Schematic of the VDG-40 Video Grader…….……………………….…….……17

FIGURE 2.7: Five Gradation Curve VS. Maximum Density Line………………………………32

FIGURE 2.8(b): View of Contact Force of the Mixture at Equilibrium State………………...…34

FIGURE 3.1: Stress VS. Strain of Mixture 1………………….…………………………………44

FIGURE 3.2: Asphalt Content VS. Peak Strength…………………….…………………………44

FIGURE 4.1: Flowchart of a PFC3D Program Code Using UDM………………………………60

FIGURE 4.3: Skyscan 1174 System…………………..…………………………………………64

FIGURE 4.4: Testing Stage of the Microscopy System (59)…………………….………………65

FIGURE 4.5: Crystal Aggregates Used in Compression Tests…………………..………………66

FIGURE 4.6: Asphalt Mixture Sample Used in Compression Tests…………………………….66

FIGURE 4.7(a): X-ray Image of Sample 1………………………………………………………68

FIGURE 4.7(b): Resistant Force vs. Deformation of Sample 1…………………………………68

FIGURE 4.8(a): X-ray Image of Sample 2………………………………………………………69

FIGURE 4.8(b): Resistant Force vs. Deformation of Sample 2…………………………………69

FIGURE 4.9(a): X-ray Image of Sample 3………………………………………………………70

FIGURE 4.9(b): Resistant Force vs. Deformation of Sample 3…………………………………70

FIGURE 4.10(a): X-ray Image of Sample 4…………………….………………………………71

FIGURE 4.10(b): Resistant Force vs. Deformation of Sample 4….……………………………71

FIGURE 4.11(a): X-ray Image of Sample 5……………………….……………………………72

FIGURE 4.11(b): Resistant Force vs. Deformation of Sample 5….……………………………72