Asphalt
Asphalt
Asphalt
176
ANALYSIS OF ASPHALT
PAVEMENT MATERIALS AND
SYSTEMS
EMERGING METHODS
PROCEEDINGS OF THE SYMPOSIUM
ON THE MECHANICS OF FLEXIBLE PAVEMENTS
SPONSORED BY
Pavements Committee of the Geo-Institute of the American Society of Civil
Engineers
The Inelastic Committee and the Granular Materials Committee of the ASCE
Mechanics Division
EDITED BY
Linbing Wang
Eyad Masad
GEO
INSTITUTE
ASCE
Published by the American Society of Civil Engineers
Library of Congress Cataloging-in-Publication Data on file.
www.pubs.asce.org
Any statements expressed in these materials are those of the individual authors and do not
necessarily represent the views of ASCE, which takes no responsibility for any statement
made herein. No reference made in this publication to any specific method, product,
process, or service constitutes or implies an endorsement, recommendation, or warranty
thereof by ASCE. The materials are for general information only and do not represent a
standard of ASCE, nor are they intended as a reference in purchase specifications, contracts,
regulations, statutes, or any other legal document. ASCE makes no representation or
warranty of any kind, whether express or implied, concerning the accuracy, completeness,
suitability, or utility of any information, apparatus, product, or process discussed in this
publication, and assumes no liability therefore. This information should not be used without
first securing competent advice with respect to its suitability for any general or specific
application. Anyone utilizing this information assumes all liability arising from such use,
including but not limited to infringement of any patent or patents.
ASCE and American Society of Civil Engineers—Registered in U.S. Patent and Trademark
Office.
Copyright © 2007 by the American Society of Civil Engineers. All Rights Reserved.
ISBN 13: 978-0-7844-0924-4
ISBN 10: 0-7844-0924-2
Manufactured in the United States of America.
Each paper published in this GSP was evaluated by peer reviewers and the editors.
The review comments were sent to the authors and they have been addressed to the
reviewers and the editors' satisfaction. The ASCE Geo-Institute Pavements
Committee and the reviewers are sincerely acknowledged for their time and efforts.
The papers in this GSP include eight papers that were presented in the symposium on
Mechanics of Flexible Pavements at the 15th U.S. National Congress of Theoretical
and Applied Mechanics, held at Boulder, Colorado, June 25-30 2006 and five papers
submitted for publication only. The symposium was supported by the Geo-Institute
Pavements Committee, the Inelastic Committee and the Granular Materials
Committee of the ASCE Mechanics Division.
The editors of this GSP would like to thank the Board of Governors of the Geo
Institute for their approving the symposium and the special publication.
iii
Geotechnical Special Publications
1 Terzaghi Lectures 48 Soil Suction Applications in Geotechnical
2 Geotechnical Aspects of Stiff and Hard Engineering
Clays 49 Soil Improvement for Earthquake Hazard
3 Landslide Dams: Processes, Risk, and Mitigation
Mitigation 50 Foundation Upgrading and Repair for
7 Timber Bulkheads Infrastructure Improvement
9 Foundations & Excavations in Decomposed 51 Performance of Deep Foundations Under
Rock of the Piedmont Province Seismic Loading
11 Dynamic Response of Pile Foundations - 52 Landslides Under Static and Dynamic
Experiment, Analysis and Observation Conditions-Analysis, Monitoring, and
14 Geotechnical Aspects ofKarst Terrains Mitigation
15 Measured Performance Shallow Foundations 53 Landfill Closures-Environmental
16 Special Topics in Foundations Protection and Land Recovery
17 Soil Properties Evaluation from Centrifugal 54 Earthquake Design and Performance of
Models Solid Waste Landfills
18 Geosynthetics for Soil Improvement 55 Earthquake-Induced Movements and
19 Mine Induced Subsidence: Effects on Seismic Remediation of Existing
Engineered Structures Foundations and Abutments
21 Hydraulic Fill Structures 56 Static and Dynamic Properties of Gravelly
22 Foundation Engineering Soils
23 Predicted and Observed Axial Behavior of 57 Verification of Geotechnical Grouting
Piles 58 Uncertainty in the Geologic Environment
24 Resilient Moduli of Soils: Laboratory 59 Engineered Contaminated Soils and
Conditions Interaction of Soil Geomembranes
25 Design and Performance of Earth Retaining 60 Analysis and Design of Retaining
Structures Structures Against Earthquakes
27 Geotechnical Engineering Congress 61 Measuring and Modeling Time Dependent
28 Detection of and Construction at the Soil Behavior
Soil/Rock Interface 62 Case Histories of Geophysics Applied to
29 Recent Advances in Instrumentation, Data Civil Engineering and Public Policy
Acquisition and Testing in Soil Dynamics 63 Design with Residual Materials:
32 Embankment of Dams-James L. Sherard Geotechnical and Construction
Contributions Considerations
33 Excavation and Support for the Urban 64 Observation and Modeling in Numerical
Infrastructure Analysis and Model Tests in Dynamic Soil-
34 Piles Under Dynamic Loads Structure Interaction Problems
35 Geotechnical Practice in Dam Rehabilitation 65 Dredging and Management of Dredged
37 Advances in Site Characterization: Data Material
Acquisition, Data Management and Data 66 Grouting: Compaction, Remediation and
Interpretation Testing
39 Unsaturated Soils 67 Spatial Analysis in Soil Dynamics and
40 Vertical and Horizontal Deformations of Earthquake Engineering
Foundations and Embankments 68 Unsaturated Soil Engineering Practice
41 Predicted and Measured Behavior of Five 69 Ground Improvement, Ground
Spread Footings on Sand Reinforcement, Ground Treatment:
42 Serviceability of Earth Retaining Structures Developments 1987-1997
43 Fracture Mechanics Applied to Geotechnical 70 Seismic Analysis and Design for Soil-Pile-
Engineering Structure Interactions
44 Ground Failures Under Seismic Conditions 71 In Situ Remediation of the Geoenvironment
45 In Situ Deep Soil Improvement 72 Degradation of Natural Building Stone
46 Geoenvironmen12000 73 Innovative Design and Construction for
47 Geo-Environmental Issues Facing the Foundations and Substructures Subject to
Americas Freezing and Frost
iv
74 Guidelines of Engineering Practice for 107 Soil Dynamics and Liquefaction 2000
Braced and Tied-Back Excavations 108 Use of Geophysical Methods in
75 Geotechnical Earthquake Engineering and Construction
Soil Dynamics HI 109 Educational Issues in Geotechnical
76 Geosynthetics in Foundation Reinforcement Engineering
and Erosion Control Systems 110 Computer Simulation of Earthquake
77 Stability of Natural Slopes in the Coastal Effects
Plain 111 Judgment and Innovation: The Heritage
78 Filtration and Drainage in and Future of the Geotechnical
Geotechnical/Geoenvironmental Engineering Profession
Engineering 112 Soft Ground Technology
79 Recycled Materials in Geotechnical 114 Soils Magic
Applications 115 Expansive Clay Soils and Vegetative
80 Grouts and Grouting: A Potpourri of Influence on Shallow Foundations
Projects 116 Deep Foundations 2002: An International
81 Soil Improvement for Big Digs Perspective on Theory, Design,
82 Risk-Based Corrective Action and Construction, and Performance
Brownfields Restorations 117 Discrete Element Methods: Numerical
83 Design and Construction of Earth Modeling ofDiscontinua
Retaining Systems 118 A History of Progress: Selected U.S. Papers
84 Effects of Construction on Structures in Geotechnical Engineering
85 Application of Geotechnical Principles in 119 Soil Behavior and Soft Ground
Pavement Engineering Construction
86 Big Digs Around the World 120 Grouting and Ground Treatment
87 Jacked Tunnel Design and Construction 121 Probabilistic Site Characterization at the
88 Analysis, Design, Construction, and Testing National Geotechnical Experimentation
of Deep Foundations Sites
89 Recent Advances in the Characterization of 122 Sinkholes and the Engineering and
Transportation Geo-Materials Environmental Impacts ofKarst
90 Geo-Engineering for Underground 123 Recent Advances in Materials
Facilities Characterization and Modeling of
91 Special Geotechnical Testing: Central Pavement Systems
Artery/Tunnel Project in Boston, 124 GeoSupport 2004: Drilled Shafts,
Massachusetts Micropiling, Deep Mixing, Remedial and
94 Performance Confirmation of Constructed Specialty Foundation Systems
Geotechnical Facilities 125 Current Practices and Future Trends in
95 Soil-Cement and Other Construction Deep Foundations
Practices in Geotechnical Engineering 126 Geotechnical Engineering for
96 Numerical Methods in Geotechnical Transportation Projects
Engineering: Recent Developments 127 Recycled Materials in Geotechnics
97 Innovations and Applications in 128 Soil Constitutive Models: Evaluation,
Geotechnical Site Characterization Selection, and Calibration
98 Pavement Subgrade, Unbound Materials, 129 Advances in Designing and Testing Deep
and Nondestructive Testing Foundations
99 Advances in Unsaturated Geotechnics 130 Advances in Pavement Engineering
100 New Technological and Design 131 Contemporary Issues in Foundation
Developments in Deep Foundations Engineering
101 Slope Stability 2000 132 Advances in Deep Foundations: In
102 Trends in Rock Mechanics Memory of Michael W. O'Neill
103 Advances in Transportation and 133 Earthquake Engineering and Soil
Geoenvironmental Systems Using Dynamics
Geosynthetics 134 Soil Dynamics Symposium in Honor of
104 Advances in Grouting and Ground Professor Richard D. Woods
Modification 135 Erosion of Soils and Scour of Foundations
105 Environmental Geotechnics 136 Innovations in Grouting and Soil
106 Geotechnical Measurements: Lab & Field Improvement
v
137 Legal and Liability Issues in Geotechnical 170 Probabilistic Applications in Geotechnical
Engineering Engineering
138 Site Characterization and Modeling 171 Advances in Shallow Foundations
139 Calibration of Constitutive Models 172 Soil Improvement
140 Slopes and Retaining Structures Under 173 Advances in Measurement and Modeling
Seismic and Static Conditions of Soil Behavior
141 International Perspectives on Soil 174 Designing Our Underground Space
Reinforcement Applications 175 Field Measurements in Geomechanics
142 Waste Containment and Remediation 2007
143 Geomechanics: Testing, Modeling, and
Simulation
144 Sinkholes and the Engineering and
Environmental Impacts ofKarst
145 Seismic Performance and Simulation of
Pile Foundations in Liquefied and
Laterally Spreading Ground
146 Asphalt Concrete: Simulation, Modeling
and Experimental Characterization
147 Unsaturated Soils 2006
148 Advances in Unsaturated Soil, Seepage,
and Environmental Geotechnics
149 Site and Geomaterial Characterization
150 Soil and Rock Behavior and Modeling
151 Advances in Earth Structures: Research to
Practice
152 Ground Modification and Seismic
Mitigation
153 Foundation Analysis and Design:
Innovative Methods
154 Pavement Mechanics and Performance
155 Underground Construction and Ground
Movement
156 Geomechanics II: Testing, Modeling, and
Simulation
157 Computer Applications in Geotechnical
Engineering
158 Contemporary Issues in Deep Foundations
159 Case Studies in Earth Retaining Structures
160 Dynamic Response and Soil Properties
161 Embankments, Dams, and Slopes: Lessons
from the New Orleans Levee Failures and
Other Issues
162 Problematic Soils and Rocks and In Situ
Characterization
163 Geoenvironmental Engineering
164 Innovative Applications of Geophysics in
Civil Engineering
165 Geosynthetics in Reinforcement and
Hydraulic Applications
166 Educational Activities in Geotechnical
Engineering
167 Geotechnics of Soil Erosion
168 Grouting for Ground Improvement:
Innovative Concepts and Applications
169 Soil and Material Inputs for
Mechanistic-Empirical Pavement Design
vi
Contents
vii
This page intentionally left blank
STRESS CONCENTRATION FACTOR AS A PERFORMANCE INDICATOR
FOR ASPHALT MIXES
ABSTRACT: This paper presents some recent developments in using Finite Element
Method (FEM) and the 3D microstructure of asphalt concrete reconstructed from x-
ray tomography imaging to evaluate its performance. Asphalt concrete is modeled as
a material composed of an effective solid and void. The stress concentration factor
and its distributions are computed and correlated to the overall resistance of the
mixtures against rutting and fatigue, and the internal structural quantities. The
correlation of the stress concentration factor with the field performance of the
WesTrack mixes indicates that mixes with large stress concentration factors, as
expected, may be more prone to rutting and fatigue cracking than those with smaller
stress concentration factors.
INTRODUCTION
1
Associate Professor, Department of Civil and Environmental Engineering, Virginia Tech,
Blacksburg,VA24061.
2
Graduate Research Assistant, Department of Civil and Environmental Engineering, Virginia Tech,
Blacksburg,VA24061.
3
Professor, Institute of Hydraulic Structures, Department of Hydraulic Engineering, Tsinghua
University Beijing, 100084, P.R. China.
4
Associate Professor, Department of Civil and Environmental Engineering, Virginia Tech,
Blacksburg, VA
1
2 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
deformation and overall stiffness, but encounters difficulties in dealing with localized
deformations such as rutting, and cracking initiation and propagation. The localized
phenomena are sensitive to the microstructure of the material and the properties of
the constituents. With recent developments in microstructure characterization and
especially the X-ray Computerized Tomography (XCT) technique (Shashidhar, 1999;
Wang et al, 2001; Wang et al., 2003; Tashman et al, 2002), the three-dimensional
(3D) microstructure of asphalt concrete can be characterized, allowing a better
understanding of the localized phenomena. For example, a cylindrical asphalt
concrete specimen subjected to a uniform axial tension on the boundary could be only
considered subject to a uniform tension on every cross section, if no microstructural
information could be incorporated. Nevertheless, the stress in the specimen is
significantly localized and related to the void size and spatial distributions, the
relative stiffness of aggregates and binder/mastics, and the aggregate shape and
configuration. Due to the complicated microstructure, the stress in any specific
location within the specimen could be either shielded (smaller than the stress on the
boundary) or magnified (larger than the stress on the boundary). The localized
phenomena would take place at locations where either the constituent is soft or the
stress is significantly magnified (stress concentration). A study to assess how stress
concentration and/or strain localization are affected by mix characteristics, such as
gradation, would provide some insightful information on the performance of the mix.
Among the three constituents, voids have the most distinct features from binder and
aggregates as voids can be assumed to have zero stiffness. If aggregates coated with
asphalt binder are considered as an effective solid, asphalt concrete can be modeled
as a material consisting of only the effective solids and voids. However, the
simplification would not be true if the void structure of asphalt concrete is not
considered. Without microstructural considerations, it would make no difference for
two mixes of different gradations but of the same asphalt content and the same void
content. The void structure is related to the aggregate gradation as well as the
compaction methods and efforts. It was discovered that the void size distribution
might be a mirror image of the aggregate gradation and carry the information of mix
design. Figure 1 presents the aggregate gradation and the void size distribution of the
three mixes of the WesTrack project (Epps, 1997). The similarity between the
aggregate gradation and the corresponding void size distribution indicates that void
size distribution carries the effect of aggregate gradation and is a comprehensive
indicator of the mixture's internal structure.
The properties of the effective solid are also important. The effective solid is
composed of the aggregates coated with asphalt binder. Theoretically, it should be a
viscoelastic or a viscoplastic material. At low temperature or under fast loading and
small stress conditions, the effective solid may be modeled as an elastic material.
These simplifications reduce the asphalt concrete microstructure to an elastic
continuum weakened by the randomly distributed air voids related to mix
characteristics (although certain percentage air void, i.e. 4%, is required to avoid
bleeding or plastic mixes, above that level, a mix will be weakened more with larger
air void content). With these simplifications, the localized stress distributions in an
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS :
The paper presents the development of FEM simulation of the internal stress
distribution of asphalt concrete subject to a uniform external tensile stress or
compressive stress. The simulation makes use of the actual 3D void structure
obtained from X-ray tomography imaging. A simple parameter, the stress
concentration factor is defined as an indicator of a mixture's potential performance.
The stress concentration factor is correlated with the internal structure and the field
performance of the three mixes of the WesTrack project. The correlations indicate
that the stress concentration factor may provide a promising indicator of mixture
performance.
Figure 2 presents a simple case: an elastic plate weakened by a single cavity. The
plate is subjected to a uniformly distributed tensile stress on two boundaries. The
elasticity solution indicates that the largest stress is at the edge of the void, and the
ratio between the largest stress and the uniform far field stress on the boundaries, the
stress concentration factor 5C, equals to (Wu, 1987):
4 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
I I l-ol ||
FIG. 2 Void Shape and Stress Concentration Factor
a. Mean Solid Path is the Average b. Average Size of Defect is the Average
Spacing among the Cracked Surfaces Segment Length in Different
or Voids in Different Orientations Orientations of All the Defects or Voids
FIG.3 Damaged /Void Surfaces, Spacing among the Surfaces and Definitions of
Mean Solid Path and Average Size of Defect
METHODOLOGY
In this study, the randomly distributed void structure is obtained from x-ray
tomography imaging. X-ray tomography imaging is an effective tool to characterize
the 3D microstructure, defects, and inhomogeneity of materials. Interested readers are
referred to (Kak, 1999) for more details about its mechanisms. Figure 4 graphically
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 5
illustrates the procedures used for this study. Figure 4.a presents a 3D rendering of an
asphalt concrete core obtained through x-ray tomography imaging, where a series of
slices of images were stacked. Figure 4.b presents the plane view of a small block of
the inside of the core. Simulation will be performed on such blocks randomly
sampled from the core so that the boundary effect at the specimen edges can be
eliminated and the computational time reduced. Figure 4.c presents the binary
counterpart of the block presented in Figure 4.b, where voids and the effective solid
are respectively represented as black and white. It should be noted that an image is
actually the visualization of an array showing the intensities of pixels, the smallest
element in the image. Figure 5 presents the image overlapped by the pixel grid. For
each pixel, its coordinates (X, Y) are recorded. If black represents voids, the pixels
that have pixel values of zero are void pixels and their locations can be determined.
slices. Each voxel is modeled as an eight-node element. The simulation follows two
steps. Step one simulates the case when the uniform voxel grid is subject to a
uniformly distributed tensile stress at the boundary (Figure 6a). Step two simulates
the removal of the void voxels (Figure 6b). The stress concentration factor is defined
as the ratio between the maximum tensile stress and the uniformly distributed tensile
stress on the boundaries.
**- **'r-
08.OT.SJ for 3001
FIG. 6 Illustration of the Voxel Concept and the Two Step Simulation
In this study, the size of each element (voxel) is 0.29X0.29X1.8mm. For the
problem studied, 50X50X10=25000 elements, representing a rectangular block of
15X15X18 mm3 were used for the simulation. This selection addresses both
representativeness and computational time. Too small size simulation may cause
numerical instability. Larger size simulation can be achieved but at much slower
speed. For the purpose of this study, many blocks from a specimen are sampled for
the simulation so that a statistical evaluation of the stress concentration factors can be
obtained, and the inhomogeneous void structure or the mix's structure can be
evaluated.
Simulation Results
The above method was applied to the evaluation of the stress concentration factors of
the three mixes: the coarse mix, the fine mix and the fine plus mix of the WesTrack
project.
Figure 6b also presents the distribution of the Von Mises stresses of one of the
simulations, where clearly the effective shear stress (the Von Mises stress) around the
voids is much larger. The mean stress concentration factors for the three mixes are
presented in Figure 7. Clearly, the coarse mix has the largest stress concentration
factor while the fine mix has the smallest stress concentration factor with the fine plus
mix in the middle. It should be noted that the total air void contents of the three mixes
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 7
are the same. So the significant differences in stress concentration factor should be
the result of the void size and spatial distributions that reflect the aggregate gradation
and compaction effects. If the stress concentration factor is used to rank the
performance of the three mixes, they will give a consistent ranking with the field
performance (Epps, 1997).
Statistical Analysis
mixes. The Student t-tests indicate that the coarse mix and the fine mix are
significantly different; the fine mix and fine plus mix are significantly different; but
the coarse mix and the fine plus mix are not significantly different. In addition the
standard deviations of the stress concentration factor for the three mixes also indicate
that the coarse mix has the largest variation, indicating more significant
inhomogeneitv of its void structure.
Obviously, the uni-axial tension analysis is only one measure of the stress
concentration factors. Due to the factor that stress concentration is also related to the
orientation of the irregularly shaped voids, the stress concentration factor due to the
extension/compression in other orientations should also be evaluated. In addition, the
mix's response to shear forces on the boundary may better differentiate the structures
in terms rutting. These studies are currently in progress and will be presented in a
separate paper. Nevertheless, a better mixture structure should have a smaller stress
concentration factor in terms of any of the rational stress boundary conditions.
Figure 9a and Equations (2) and (3) illustrate the concept of effective stress. In Figure
9a, if the cross-section has no or some void area, the stresses are respectively equal to
Where F is the total force applied normal to the cross-section; A is the cross-section
area; Av is the area of air voids contained in A\ an is the nominal normal stress; ereis
^4
the effective stress; and <p is the area fraction of voids (<p = —- , in continuum
A
damage mechanics, this parameter is named as damage scalar).
Generally, the area fractions of air voids in different orientations at the same point
are different. In some orientations, the area fractions may be significantly larger than
those in other orientations. Representation of this orientational variation of the area
fractions requires the use of a second order tensor (pir In continuum damage
mechanics it is named as damage tensor (Voyiadjis, 1999). Its typical application can
be found in the following equation (Murakami, 1988).
where <j.r and crw are the effective stress tensor and Cauchy stress tensor
respectively, while Mijkl is a symmetric fourth rank tensor-the damage effect tensor.
By using their vector formats of the tensors, Murakami (1988) showed that:
a. Effective Stress
b. Spacing/Size Ratio
represent these distributions. For correlation analysis, only the mean solid path is
used. Figure lOb presents the correlation between the mean solid path and the stress
concentration factor. The correlation is also very well and makes sense. Again, more
data points are needed to verify this correlation.
Spacing-Size Ratio
The size of and the spacing between voids/defects are important variables in non-
local damage theories. For modeling applications, the size of cracks or voids and the
average spacing among the cracks offer two relevant parameters that are related to
performance. Chudnovsky et al (1987a; 1987b) developed a simple model addressing
the interaction of two cracks, in which the two important model parameters are the
spacing between the two crack tips and the sizes of the cracks. Figure 9b presents the
configuration of two interacting cracks. The stress intensity (Mode I) of the large
crack is affected by the adjacent cracks. The ratio —l— (where Kf is the effective
stress intensity of a crack when subjected to the influence of the adjacent crack while
K\ is the stress intensity without the influence of adjacent cracks) decreases with the
increase o f A / S , which is equivalent to that the larger the crack size and/or the
smaller the spacing, the larger the effective stress intensity. Therefore, under the same
external loading, the material with smaller A18 should have a larger stress intensity
factor. Equations (6) and (7) present the relation between the effective stress intensity
and the crack configuration parameters illustrated in Figure 9b.
Figure lOc presents the correlation between the spacing/size factor and the stress
concentration factor. The correlation is also very good. However, additional data
points are required to verify this correlation.
12 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
CONCLUSION
The stress concentration factor may present a simple performance indicator for
asphalt mixes against both rutting and fatigue. Additional study to consider the
mixture as a three-constituent composite with binder or mastics modeled as
viscoelastic material is recommended. The stress concentration factor is well
correlated to the internal structure parameters such as effective damage, mean solid
path, and the spacing size ratio. The internal structure parameters, by mechanics
theories, are related to material properties and may serve as good internal variables
for modeling and simulation. Additional data shall be needed to widely verify the
presented relations and test validity of using the stress concentration factor as a
performance indicator. However, the good correlations between the stress
concentration factor and the internal structure parameters and the field performance
indicate that stress concentration factor may serve as a promising simple performance
indicator for asphalt concrete.
ACKNOWLEDGEMENT
REFERENCES
1
Tomasini Assistant Professor, Ph.D., P.E., Department of Civil and Environmental
Engineering, Michigan Technological University, 1400 Townsend Drive, Houghton,
Michigan, 49931-1295, [email protected]
2
Associate Professor, Ph.D., P.E., Department of Civil and Environmental
Engineering, University of Illinois at Urbana-Champaign, IL 61801,
[email protected]
3
Visiting Assistant Professor, School of Technology and Department of Mechanical
Engineering-Engineering Mechanics, 1400 Townsend Drive, Houghton, MI 49931,
[email protected]
14
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 15
(E*). A 9.5-mm (nominal maximum aggregate size) mixture was used in this study.
Gyratory specimens were used to produce uniaxial compression specimens and
hollow cylinder tensile test (HCT) specimens. The sand mastic (1.18-mm nominal
maximum aggregate size) had around 14% asphalt content, by weight of the mixture.
The sand mixture was prepared using a specially designed aluminum molds to create
uniaxial compression and HCT specimens. Complex moduli used in this portion of
the study were measured at three temperatures, -20, -10, and 0°C, and four
frequencies, 0.1, 1, 5, and lOHz. In the test, cyclic loads were carefully applied to
keep the response in a linear visco-elastic range (Airey et al. 2003).
MICROSTRUCTURE MODELING OF THE MIXTURE
The discrete element method has received considerable attention in the past 20 years
after its gradual development for studies in rock mechanics (Cundall 1971; Cundall
2000). During this time, the DEM method has been extended to many other fields.
The modeling procedure used in this study adopted a clustered DEM, or microfabric
approach (Buttlar and You 2001). A unique feature of this approach is the cohesive
and adhesive strength for asphalt concrete. Biaxial compressive modulus prediction
using the discrete element modeling approach has been recently demonstrated (You
and Buttlar 2004) to match or exceed the predictive capabilities of empirical modulus
prediction models. A hollow cylinder tensile test was modeled to predict mixture
tensile modulus (You and Buttlar 2005). It was found that the two dimensional (2D)
clustered DEM provided very good estimates of mixture dynamic modulus across a
range of loading temperatures and frequencies without calibration.
In the clustered DEM modeling approach, various material phases (e.g.,
aggregates, mastic) are modeled with clusters of discrete elements. Aggregate and
mastic modulus are significant input parameters in these models, which are obtained
by converting the macro-modulus obtained from laboratory tests to micro-modulus in
the DEM models. A significant model feature is that the cohesive strength of
individual aggregates and the mastic can be considered, along with the adhesion at
the aggregate-mastic interface (You and Buttlar 2005, 2006; and Dai and You 2007).
In addition, aggregate-aggregate interactions are considered in the DEM approach
using detailed morphological information (aggregate shape, contact, force chain
structure, etc.).
figure are from the same specimen image. The difference is the aggregate contents in
the model. The main similarity is that the coarse aggregate structure is the same, i.e.,
all the models include the same aggregate internal 2D structure with minimum
aggregate size of 2.36mm.
model with 0.40 and 0.47 aggregate concentration ratios, there is a modulus increase
of 15% when 7% fine aggregate is included in the mixture, when the mastic modulus
is IGPa. For a similar case, when the mixture includes 11% more fines, the mixture
modulus increased by 18%. It should be noted that when even 1% more fines are
included, the mixture modulus increases by almost 4%. If the mastic modulus is
7.5GPa (at a lower temperature), when the aggregate concentration ratio increased
from 0.40 to 0.41, 0.47, and 0.52, respectively, the mixture modulus increased by
12%, 17%, and 19%, respectively.
Figure 2 also shows the DEM prediction compared with laboratory
measurements versus sand mastic modulus. For the model with an aggregate volume
concentration ratio c=0.47, the average prediction difference is 11%, compared with
16% predicted from the model with a 0.40 concentration ratio. For the model with
aggregate volume concentration ratio of 0.51, the average prediction difference is
10% between the prediction and the measurements. Based upon a paired t-test, there
is a 95% confidence interval (-0.70, 2.38) for the mean difference between the DEM
predictions (c=0.51) and the measurements. Therefore, there is a 95% confidence
that the mean difference between prediction and the measurements across a range of
temperatures and frequencies is statistically insignificant.
Figure 2. DEM predictions of the mixture when aggregate modulus is fixed and
sand mastic modulus varies
For the model with an aggregate volume concentration ratio of 0.52, the
average prediction difference is 10%. Again, the model prediction is much better than
the model with c=0.40 (with low degree of particle-to-particle contact). There is a
95% confidence interval (-1.02, 2.06) for the mean difference. Therefore, the
statistical analysis results indicate that the prediction from this model is acceptable
and is the best among the four models.
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 19
The objective of the examples given here is to aid in the understanding of the
important stiffening effect of aggregate-aggregate contact and aggregate volume
concentration, as well as aggregate gradation effect. Due to the length limitation, it is
very difficult to give a detailed explanation of other results and findings. When a 3D
model is used in the near future, it is possible that the prediction will be further
improved.
In order to study the effect of aggregate modulus to the mixture modulus, the model
in Figure 2 (c=0.51) was used. When a fixed mastic modulus of 5.27GPa, aggregate
modulus was varied from 1 to 120GPa. For this specific model, the mixture modulus
prediction is shown in Figure 3. It can be found that when the aggregate modulus
increases from 40 to 80 GPa, the mixture modulus increases 12%; when aggregate
modulus increases from 20 to 120 GPa, mixture modulus increases 80%. Obviously,
when stiffer aggregate is used, a stiffer mixture is expected. However, many studies
such as the recent, well-known predictions (Witczak et al. 2001; Christensen et al.
2003) for asphalt mixture do not consider the aggregate modulus effect. It should be
noted that the modulus of aggregate is very difficult to measure. Researchers at
University of Illinois measured the modulus of limestone from a local quarry and
found there was more than a two-fold difference in aggregate modulus for the
aggregate sources from different depths in the quarry (You 2003).
Figure 3. Aggregate modulus effect to mixture modulus when aggregate modulus varies
and sand mastic modulus is fixed
20 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
In this paper, the effect of aggregate particle-to-particle interaction was modeled and
the mixture moduli were predicted using the discrete element modeling approach. In
this approach, the material phases (aggregates and mastic) were modeled with
clusters of discrete elements with different properties. The moduli of the aggregate
and mastic were among the input parameters. A compressive test was simulated
considering the heterogeneity of asphalt mixture. When the degree of particle-to-
particle contact in an asphalt mixture increased, a higher mixture modulus was
predicted. Aggregate modulus sensitivity analysis showed that the aggregate modulus
contribution to mixture modulus can be significant. It is anticipated that ongoing 3D
imaging and modeling efforts will have lead further improvement in mixture property
and behavior predictions, and will provide detailed insight towards the roles of coarse
and fine aggregates, their gradation, and other morphological characteristics in the
behavior of asphalt mixtures.
REFERENCES
Aho, B. D., W. R. Vavrik and S. H. Carpenter (2001). Effect of Flat and Elongated
Coarse Aggregate on Field Compaction of Hot-Mix Asphalt, Transportation Research
Board.
Airey, G. D., B. Rahimzadeh and A. C. Collop (2003). "Viscoelastic linearity limits
for bituminous materials." Materials and Structures/Materiaux et Constructions
36(264): 643-647.
Alexander, M. G. and T. I. Milne (1995). Influence of Cement Blend and Aggregate
Type on Stress-Strain Behavior and Elastic Modulus of Concrete, American Concrete
Institute.
Alkio, R. and J. Vuorinen (1996). The Effect of Aggregate Shape on the Wear
Resistance of Asphalt Pavement : A Study With Field Test Slabs: Wear Results Of
The First Winter, Tielaitos, Tielaitoksen Selvityksia.
Buttlar, W. G. and Z. You (2001). "Discrete Element Modeling of Asphalt Concrete:
A Micro-Fabric Approach." Journal of the Transportation Board, National Research
Council, Washington, D.C. 1757.
Christensen, D. W., T. Pellien and R. F. Bonaquist (2003). "Hirsch Models for
Estimating the Modulus of Asphalt Concrete." J. Assoc. Asphalt Paving Tech. 72: 97-
121.
Cundall, P. A. (1971). A Computer Model for Simulating Progressive Large Scale
Movements in Blocky Rock Systems. Proceedings of the Symposium of the
International Society of Rock Mechanics, Nancy, France.
Cundall, P. A. (2000). "A Discontinuous Future for Numerical Modelling in
Geomechanics?" Geotech. Eng 149(1): 41-47.
Dai, Q. and You, Z (2007), Prediction of Creep Stiffness of Asphalt Mixture with
Micromechanical Finite Element and Discrete Element Models, Journal of
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 21
Abstract
This paper presents a microstructure-based finite element model by incorporating elastic
aggregates and viscoelastic sand mastic. The microstructure-based finite element (FE)
approach was used to predict the linear and damage-coupled viscoelastic properties of
reclaimed asphalt mixture. The two-dimensional (2D) microstructure of asphalt mixture
was obtained from the scanned image of a smoothly sawn surface of a reclaimed asphalt
pavement (RAP) mixture specimen. In the microstructure, the sketches of highly irregular
aggregates were converted into polygons. The whole microstructure model was divided
into highly irregular aggregate and sand mastic subdomains. The finite element mesh was
generated within each subdomain. The deformation of the aggregate and mastic
subdomains was connected through the sharing boundary nodes. Linear and damage-
coupled viscoelastic finite element model was developed with displacement-based
incremental formulation. The linear and damage-coupled viscoelastic simulation was
conducted on the image sample of test specimen under different sinusoidal force loading
frequencies. The uniaxial compression simulation results showed creep deformation
constant cyclic force loading amplitude and damage-coupled viscoelastic responses have
larger creep deformation. Simulations under different loading frequencies found
compression strain deceases with loading frequencies due to less relaxation time.
Introduction
A sustainable asphalt mixture is designed to consider the economic, societal, and
environmental factors through life cycle analysis. The use of reclaimed asphalt pavement
(RAP) material has been considered as sustainable asphalt mixtures and has obtained
varying degrees of success in the United States. The asphalt pavement recycling methods
include hot mix recycling, hot in-place recycling, cold in-place recycling and full depth
reclamation. Hot mix recycling is the predominant method of structural recycling and the
mix design procedure is basically similar as the one for new mixtures with the additional
requirement for asphalt pavement analysis (Roberts et al. 1996). A recycled mixture
should be designed to produce an asphalt material having the same properties as that in a
new one. The aged asphalt binder must be mixed with the new asphalt binder or recycling
agent to provide the overall desired binder properties. Recent study have reported
1
Visiting Assistant Professor, School of Technology and Department of Mechanical
Engineering-Engineering Mechanics, Michigan Technological University, Houghton,
1400 Townsend Drive, MI 49931. [email protected]
2
Tomasini Assistant Professor, Dept. of Civil and Environmental Engineering, Michigan
Technological University, 1400 Townsend Drive, Houghton, MI 49931. [email protected]
22
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 23
Buttlar 2006). Image processing technique was used to converted the image into many-
sided polygons using a custom developed macro program to define the microstructure of
asphalt mixture (Buttlar and You 2001). The average of the polygon diameter was chosen
as a threshold to determine which aggregates would be "retained" on a given sieve (the
rest of the aggregates would be "passed" the given sieve), although some other
measurement parameters were also attempted in analyzing the gradation of the aggregates
(You 2003). The polygons used in the micromechanical model were filtered as coarse
aggregates. In the model of the asphalt mixture, the coarse aggregates retained on
2.36mm sieve (No. 8) were counted in the subdomain of aggregate. The fine aggregates
passing 2.36mm were filtered as sand mastic. Elastic properties of aggregates were
measured and applied to aggregate subdomain, and viscoelastic mastic properties were
also calibrated with mastic creep tests (Dai and You 2007). Finite element simulation
was conducted to predict the global mixture behavior by combining aggregate and mastic
properties.
In these equations, Em, rjm and pm are the spring constant, dashpot viscosity and
relaxation time respectively for the mth Maxwell element.
The reduced time (effective time) is defined by using time-temperature
superposition principle as
Where ACT and A£ are incremental stress and strain, K is the incremental stiffness and
Ao** is the residue stress vector.
The incremental formulation of the volumetric behavior is obtained with constant
volumetric strain rate
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 25
where K^ and Km are bulk moduli, and the residual part Acr^ can be expressed in a
recursive relation with the history variable Sm,
For the initial increment, the history variable 5rOT(^1) equals to 3KmRkkpm and
strain rate
and the residual part A<r* can be expressed in the recursive relation
where
26 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
This model incorporates three different nonlinear parameters: he is the nonlinear factor
of the relaxed elastic modulus E9 , /*, measures the nonlinearity effect in the transient
modulus Et , and h2 accounts for the loading rate effect.
Following form (11), a rate-independent damage-coupled viscoelastic model is
proposed by replacing two nonlinear parameters he and hv with damage variables which
would be expressed with damage evolution functions (Assume h2=l to neglect rate-
dependent damage behavior in this study).
where he and h{ are the elastic and viscoelastic damage variables for the rate-
independent failure behavior. These variables are functions of the maximum strain s^,
which is defined as the maximum value over the past history up to the current time £ ,
emm =max(4r)U'e[0,,f] (13)
The elastic damage variable he = 1 - Q measures the relaxed elastic stiffness
reduction, and can be described by using the inelastic damage evolution law in Sadd et
al.(Sadd et al. 2004a; Sadd et al. 2004b),
where the material parameters £0and b are related to the softening strain and damage
evolution rate respectively.
The viscoelastic variable h{ measures the damage effect in the transient modulus,
and is chosen with the following exponential form by (Simo and Ju 1987),
The variable fy will reduce from 1 to p as the maximum strain emao[ increases, and £0 is
also the softening stain.
Following the previous formulation procedures, the incremental formulation of
the volumetric behavior is obtained with constant volumetric strain rate
and the residual part Acr^ can be expressed in a recursive relation with the history
variable Sm,
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 27
For the deviatoric behavior, the formulation of the deviatoric behavior is obtained
with constant deviatoric strain rate
where semwi is the maximum equivalent strain, and the equivalent strain And
The linear and damage-coupled viscoelastic models were defined in the ABAQUS
user material subroutines for mastic subdomains. A displacement-based time-dependent
finite element analysis was conducted by integrating elastic aggregate and linear and
damage-coupled viscoelastic mastic to predict the global behavior of asphalt mixture.
28 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
Fig. 1. The FEM meshes for the aggregate and mastic subdomains
Fig. 2. The FEM simulation results under sinusoidal loading at a frequency of 0.1 Hz
30 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
References
Bahia, H., Zhai, H., Bonnetti, K., and Kose, S. (1999). "Nonlinear viscoelastic and
fatigue properties of asphalt binders." J. Assoc. Asphalt Paving Tech., 68, 1-34.
Bazant, Z. P., Tabbara, M.R., Kazemi, Y., Pijaudier-Cabot, G. (1990). " Random particle
simulation of damage and fracture in participate or fiber-reinforced composites. ."
Damage Mechanics in Engineering Materials, Trans. ASME, AMD.
Budhu, M. S., Ramakrishnan, and Frantziskonis, G. "Modeling of Granular Materials: A
Numerical Model Using Lattices, Mechanics of Deformation and Flow of Particulate
Materials." McNu Conference, Trans. ASCE, Northwestern Univ.
Buttlar, W. G., Bozkurt, D., Al-Khateeb, G. G., and Waldhoff, A. S. (1999).
Understanding asphalt mastic behavior through tnicromechanics (with discussion and
closure), Transportation Research Board.
Buttlar, W. G., and Roque, R. "Evaluation of empirical and theoretical models to
determine asphalt mixture stiffnesses at low temperatures." Baltimore, MD, USA, 99-
141.
Buttlar, W. G., and Roque, R. (1997). Effect of asphalt mixture master compliance
modeling technique on thermal cracking performance evaluation using superpave,
University of Washington, Seattle.
Buttlar, W. G., and You, Z. (2001). "Discrete Element Modeling of Asphalt Concrete: A
Micro-Fabric Approach." Journal of the Transportation Board, National Research
Council, Washington, D.C.9 1757.
Christensen, R. M., and Lo, K. H. (1979). "Solutions for Effective Shear Properties in
Three Phase Sphere and Cylinder Models." J. Mech Phys. Solids, 27, 315-330.
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 31
INTRODUCTION
Mineral fillers have commonly been used in asphalt mixtures to fill the voids
between the larger aggregate particles (Chen 1996, Jo et al. 1998, Jo and Kim 2001).
Generally, aggregates passing the No. 200 sieve are called filler (Ishai and Craus
1996, Anderson and Marasteanu 1999) and are used in percentages by volume in the
mix designs. In fact, if a core of asphalt concrete is observed in cross section, three
components can be visually observed are aggregate (coarse and fine), air voids, and
mastic (Vacin et al. 1997, Buttlar et al. 1998). The asphalt cement occurs not as
asphalt alone, but is intimately mixed with mineral fillers forming the mastic. Such a
distribution points to the important role played by the mastic in the workability and
performance of asphalt concrete.
1
Postdoctoral Research Assistant, Virginia Tech University, Department of Civil and Environmental
Eng., Blacksburg VA 24061, [email protected]
2 Associate Professor, Virginia State University, Department of Civil and Environmental
Eng.,Blacksburg VA 24061, Tel. (540)231-5262, [email protected]
3 Boyd Professor, Louisiana State University, Baton Rouge LA 70803, Tel. (225)578-8668
4 Materials Administrator, Louisiana Transportation Research Center, 4101 Gourrier Avenue
Baton Rouge LA 70808, Tel. (225)767-9109, [email protected]
33
34 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
Although several studies have been reported in the asphalt literature, a good
correlation between any measured physical property and the stiffening of asphalt by
mineral fillers has not been reported yet (Smith and Hesp 2000; Uddin 1999). This
may be attributed to the fact that the stiffening potential of mineral fillers is a
complex function of several factors such as average particle size, gradation, shape of
the particles, presence of agglomerates, degree of dispersion, and asphalt-filler
interface properties. Also, some of these properties interact with each other in
complex ways - for example, the degree of dispersion is related to asphalt-filler
interface properties; and the presence of agglomerates is dependent upon the average
particle size and also the asphalt-filler interface properties (Anderson and Chrismer
1984, Jiang etal. 1997).
OBJECTIVES
There were two major objectives for this study. The first objective was to
estimate the influence of type and amount of mineral fillers on mechanical properties
of asphalt mastics through dynamic rheometer testing (DSR). One binder and three
types of fillers were mixed at five volume fractions and tested at three temperatures.
For the interpretation of the results, "master curves" technique was employed in order
to determine a correlation between the binder and the mastics using these test results.
The second objective was to estimate the complex modulus of mastics by using
binder complex modulus from the DSR testing and the volumetric composition of the
mineral fillers through Hirsch model.
MATERIALS
Fillers
Three fillers were used in this study and they were: Donna Fill, Limestone and
Granite. All the fillers were used to form the mastics in five volume fractions: 5, 10,
15, 20, and 30% (bulk material).
Donna Fill, also known as nepheline syenite granite, is a by-product of
crushed syenite granite rock and has been successfully used in asphalt concrete
mixes, construction of road base and subbase courses, airport runways, and concrete
bridge restoration (Waltham 1994). The addition of this type of granite filler,
improves Marshall stability, % VMA, shear values, film thickness of asphalt on the
aggregates, reduces rutting and improves load-carrying capacity of the pavement
structure.
Usually, the effect of donna fill and other granites as compared to natural
sands used as fines in asphalt mixtures demonstrated an increase in mixture stability,
thicker film thickness retained and a reduction in rutting and stripping susceptibility.
Modifications made to hot mix asphalt with limestone fillers might add years to its
life, because these modifications can reduce stripping, rutting, cracking, and aging
due to complicated chemical reactions. Limestone substantially improves each of
these properties when used alone, and also works well in conjunction with polymer
additives, helping to create pavement systems that will perform to the highest
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 35
expectations for many years. Life cycle cost analysis demonstrated that limestone is
also cost-effective.
Some properties of the above mentioned materials are presented in Table 1.
All fillers were obtained by dry sieving fine aggregate parent rock over a 75 um sieve
and then analyzed with the Microtrac analyzer for particle size distribution. The
results of the analysis (gradations) are presented in Figure 1. It can be seen that their
gradations are very close and therefore the different results of mastics thus formed
were mainly attributed to different material properties.
Asphalt binder
The asphalt binder used in this study was a PG64-22 unmodified base asphalt,
primarily used in paving for both new construction and pavement rehabilitation and
in both dense-graded and open graded Hot Mix Asphalt (HMA). It is also used for
sealing of edges and cracks of asphalt pavements.
Mastics
36 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
Mixtures of asphalt binder and filler, called mastics in this paper, are often assumed
to behave as simple viscoelastic materials or simply, binders stiffened by filler.
Stiffening of asphalt binder is only one way in which the addition of mineral filler
changes the properties of this material. In fact, mastics behave quite similarly to
asphalt binders, in that, they flow with time and are temperature dependent. Thus,
both loading time and test temperature must be taken into account when
characterizing the flow behavior of mastics.
EXPERIMENTAL STUDY
Sample preparation
All mineral fillers were thoroughly mixed and dried for one hour in an oven at
160°C before being used in the testing. The asphalt mastics were prepared by mixing
the binder with 3 types of filler having particle sizes smaller than the 0.075 (mm)
sieve opening (#200 mesh). Measured quantities of 25 (g) of binder were heated at
150° C for 10 minutes and then mixed with the fillers in five volume fractions of 5,
10, 15, 20, and 30% of bulk material. After being heated, the binder was removed
from the oven and the heated filler was added to it and was mixed until a
homogeneous mixture was obtained; then the mixture was poured into silicon molds
to make samples for the shear rheometer measurements - single and frequency sweep
oscillation at either constant strain or constant stress.
DSR testing
The rheological properties of the binder and mastics were measured using an
Automated Dynamic Shear Rheometer (ADSR), which can be used in stress and
strain control modes. Two types of tests were conducted: single shear oscillation at
constant strain (12%) and frequency sweep at constant stress (2500 Pa). The behavior
of the materials was evaluated in terms of |G*|/sin8. as a function of temperature and
loading time. Samples were tested at temperatures of 46° C, 55° C, and 64° C and
loading frequencies ranging from 0.1 (Hz) to 100 (Hz).
The dynamic shear rheometer (DSR) (AASHTO 2002) is currently being used by
the DOTs for determination of the viscoelastic properties of asphalts following the
recommendations of the Strategic Highway Research Program (SHRP) (Anderson et
al. 2000, Stuart et al 2000). The Superpave specification parameter |G*|/sinS was
identified as the term to be used for medium to high temperature performance grading
of asphalt binder for their rutting resistance (Bahia 1995, Gubler et al. 1999, Bahia et
al. 2001).
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 37
In this study all the samples were tested in the strain mode, at a constant strain
of 12% and a frequency of 10 (rad/s) or 1.592 (Hz), in accordance with SHRP
specifications. The plates selected for testing the samples had 25 (mm) in diameter
with a 1 (mm) gap between the plates. The shear complex modulus G* and phase
angle 8 were calculated automatically as part of the operation of the rheometer using
proprietary computer software supplied by the equipment manufacturer.
It can be seen from Figure 2 that the mineral filler type influences the
rheological properties of mastics, in that the moduli of mastics containing granite and
limestone were larger than those containing donna fill. The graph shows the volume
fractions of fillers without air voids. Volume fractions without air voids are presented
in Table 2.
The three types of filler had different effects on the same asphalt due to the
filler properties and the physical-chemical reaction between fillers and asphalt
(Shashidhar and Romero 1998). The increase in (|G*|/sin8) for granite and limestone
was due to the fact that both granite and limestone have larger moduli of elasticity
than donna fill.
38 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
f r = f7aT (1)
and it is only a function of temperature for a certain material. The Williams-Landel-
Ferry (WLF) equation has been widely used to characterize the temperature shift
factors for asphalt cement. This equation could accurately describe the shift factors
for asphalt cement above a characteristic temperature, called the reference
temperature Tr:
Figure 4 shows the master curves developed for the frequency sweeps of the samples
presented above. The shifting was conducted using Williams-Landel-Ferry (WLF)
equation, normalized to those at test temperature of 46 °C.
Figure 4. Master Curves for Pure Binder and Mastics (30% filler)
Hirsch model
One of the purposes of this paper was to adapt the Hirsch model (Christensen et al
2003) for estimating the complex shear modulus G* of asphalt concrete to estimating
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 41
the complex shear modulus of mastics by using the binder complex modulus and
volumetric composition of the mineral fillers. The following equation was employed
for the estimation of the mastics complex modulus, in terms of the complex shear
modulus, G*, VFA, and VMA:
Values for G* binder were determined experimentally using the DSR oscillation at
constant strain test.
Regarding the Hirsch model adapted for estimating the complex shear
modulus of mastics G* by using the binder complex modulus and volume fractions of
the mineral fillers, Figures 5 through 7 show that there is a good correlation between
the predicted values and the measured values (all R2 being around 0.8 or higher). For
this study VFA was taken as 100% and VMA was replaced by the filler volume
fraction and calculated without any air voids (see Table 2).
References
AASHTO Standard T315-02 (2002). "Standard Test Method for Determining the
Rheological Properties of Asphalt Binder Using a Dynamic Shear Rheometer
(DSR)."
Anderson, D. A., and. Chrismer, S. M. (1984). "Evaluation of tests for
characterizing the stiffening potential of baghouse dust in asphalt mixes." TRR
968, TRB, National Research Council, Washington, D.C., 31-37.
Anderson, D. A. and Marasteanu, M.O. (1999). "Physical hardening of asphalt
binders relative to their glass transition temperatures." TRR 1661 (1547),
National Research Council, Washington, D.C., 27-35.
Anderson, D.A., Christensen, D.W., and Bahia, H. (1991). "Physical properties of
asphalt cement and the development of performance related specifications."
AAPT Proceedings, 60,437-475.
Anderson et al. (2000). "Factors affecting variability in strategic highway research
program binder tests." TRR 1728 (1360), 28-36.
Bahia, H.U. (1995). "Critical evaluation of asphalt modification using the SHRP
concepts" TRR 1488, 82-88.
Bahia, H.U. et al. (2001). "Characterization of modified asphalt binders in
Superpave mix design." TRR, NCHRP Report 459,124-128.
Buttlar et al. (1999). "Understanding asphalt mastic behavior through
micromechanics" TRR, 1681, 157-169.
Chen, J.S. (1996). "Analysis of tensile failure properties of asphalt-mineral filler
mastics." Journal of Materials in Civil Engineering, 10(4), 256-262.
44 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
Christensen, D.W., Pellinen, T., and Bonaquist, R.F. (2003). "Hirsch model for
estimating the modulus of asphalt concrete." AAPT, 72, 184-204.
Druta, C. (2006). "A micromechanical approach for predicting the complex shear
modulus and accumulated shear strain of asphalt mixtures from binder and
mastics." Dissertation, Louisiana State University, Baton Rouge, Louisiana.
Gubler, R. et al. (1999). "Investigation of the system filler and asphalt binders by
Theological means." AAPT, 68, 284-304.
Hoare, T.R. and Hesp, S.A.(2000). "Low temperature fracture testing of asphalt
binders." TRR, 1728(1234), 36-42.
Ishai, I. and Craus, J. (1996). "Effects of some aggregate and filler characteristics on
behavior and durability of asphalt paving mixtures." TRR 1530, National Research
Council, Washington, D.C., 75-85.
Jiang, R.-B., Lin, J.D., and Lin, D.F. (1997). "Rheology of asphaltic binders and their
effects on asphalt concrete." TRR, 1535, 74-81.
Jo, D., Kim, R., and Lee, H. (1998). "Effects of aging on viscoelastic properties of
asphalt-aggregate mixtures." TRR, 1630(0334), 21-27.
Jo, D. and Kim, R. (2001). "Laboratory evaluation of fatigue damage growth and
healing of asphalt concrete mixtures using the impact resonance method", ASCE
Journal of Materials in Civil Eng., 13(6), 434-440.
Roberts, F.L. et al. (1996). "Hot-mix asphalt materials, mixture design, and
construction." NAPA Research and Education Foundation, Second Ed., 76-84.
Shashidhar, N. and Romero, P. (1998). "Factors affecting the stiffening potential of
mineral fillers." TRR, 1638(0989), 94-100.
Smith, B.J. and Hesp, S.A.M. (2000). "Crack pinning in asphalt mastic and concrete:
regular fatigue studies." TRR, 1728(1233), 75-81.
Stuart, K.D., Mogawer, W. S., and Romero, P. (2000). "Evaluation of the Superpave
asphalt binder specification for high-temperature pavement performance." AAPT,
69, 148-176.
Uddin, W. (1999). "A micromechanical model for prediction of creep compliance and
viscoelastic analysis of asphalt pavements." TRB, 78th Annual Meeting,
Washington, D.C., CD Version.
Vacin, O.J., Stastna, J., and Zanzotto, L. (1997). "Creep compliance of polymer
modified asphalt, asphalt mastic and hot mix asphalt." TRB, 82nd Annual Meeting,
Washington, D.C., CD Version.
Waltham, T.(2003). "Foundations of engineering geology." First Ed.,E & FN Spon,
44-50.
FUNDAMENTAL MECHANICS OF ASPHALT COMPACTION THROUGH
FEM AND DEM MODELING
Linbing Wang1, Bing Zhang2 and Dong Wang3 and Zhongqi Yue4
ABSTRACT
The compaction of asphalt concrete is the last step in controlling the quality of pavement.
Tremendous information on the mixture's characteristics, segregation, and fundamental
properties of asphalt concrete demonstrates itself in the compaction process. An
understanding of the fundamental mechanisms of compaction will help better control the
quality of asphalt concrete pavement. This paper presents the investigation of the
compaction mechanics using both Finite Element Method (FEM) and Discrete Element
Method (DEM). The FEM simulation adopted a porous viscoplasticity model, which
considers the compaction in the view of its air void reduction resulted from the squeezing
of aggregates and mastics. The DEM simulations provide a view of considering the
compaction a process in which aggregate particles translate and rotate to positions
forming denser packing. By this approach, particle shape and binder stifmess factors can
be considered. The simulation results indicate that both approaches can describe the
compaction phenomena consistently with field observations and empirical experience.
INTRODUCTION
Compaction of asphalt concrete is one of the most important steps in controlling the
quality of pavement construction. More importantly, many problems demonstrate
themselves during compaction: some mixes (may indicate a good mix) may not be
easy to compact compared to others; too thin layer thickness may prevent effective
compaction; a weaker mastic may make compaction too easy but it doesn't mean
good quality; a soft base or subbase may cause difficulties for compaction
(compaction energy is absorbed by the base or subbase); the non-uniformity of the
base or subbase support causes non-uniform compaction if operation parameters do
not vary correspondingly; and a change of environmental conditions such as
temperature and wind speed results in change of viscosity and therefore compaction
1
Associate Professor, Virginia Tech, Blacksburg VA 24061, email: [email protected]
2
Ph.D. Candidate, Virginia Tech, Blacksburg, VA 24061
3
Graduate Research Assistant, Virginia Tech, Blacksburg, VA 24061
4
Associate Professor, University of Hong Kong, Hong Kong
45
46 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
Factors that affect compaction can be generally classified into three categories:
Properties of the Materials, Environmental Variables and Boundary Conditions, and
Compaction Equipment and Operation Control. The properties of the materials
include the gradation, shape, angularity and texture of aggregate, and the grade, the
amount and the temperature sensitivity of the binder. The environmental variables
and boundary conditions include air and base temperatures, the humidity, sunshine or
cloudiness and wind velocity, layer thickness, stiffness of the underlying layer, and
mix laydown temperature. Layer thickness is one of the most important variables that
affect other influencing factors such as the rate of cooling of the asphalt mixtures, and
the particle movement during compaction. Compaction equipment and operation
control include types of roller, rolling patterns, number of passes, and rolling speed.
These factors can be adjusted during the compaction. An understanding of the
compaction mechanism plus the intelligent compactor technique will certainly
enhance the compaction quality significantly (intelligent compaction requires a
fundamental understanding of the compaction mechanism to make it wise; wrong
understanding would make it unwise). The factors in the three categories are listed as
follows (see Figure 1 for an illustration as well):
OBJECTIVE
The objective of this research is to investigate how different influencing factors affect
compaction through modeling using both FEM and DEM.
Numerical Approaches
There are several approaches in mechanics to model the compaction of asphalt
concrete. Fundamentally there are two major groups of approach, those in the
continuum mechanics branch and those in the micromechanics branch. The numerical
method in the continuum approach uses mainly the FEM while the micromechanics
approach uses the DEM. Analytical approach, although can offer some general guides
in understanding the force and displacement relations, it cannot model the
compaction process quantitatively and the deforming process as asphalt concrete is
not a linear elastic/viscoelastic material.
Analytical Solution
The compaction problem can be theoretically modeled as a rolling contact problem
(for example, the deforming behavior at the end of compaction). There are quite few
48 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
analytical solutions in the elasticity domain and viscoelasticity domain to solve the
contact problem. Examples include the Hertz solution and the Mindlin solution.
These are elasticity solutions but cannot be converted into viscoelasticity solutions
through the correspondence theory as conveniently because the contacting area is
changing.
Both Hertz and Mindlin solutions indicate that a stiff material would deform (elastic
deformation) less or more difficult to compact. The Mindlin solution also indicates
that the surface adhesion or cohesion play a role in the deformation of the material
(allowing for tangent forces applied on the surfaces). Nevertheless none of them can
explain the mechanism of compaction well enough as they are elastic solutions and
the displacements/ deformations will be completely recovered if the forces are
removed. The viscoelasticity solution can explain the time-dependent deformation
mechanisms to certain degree. It, however, still cannot explain the internal change of
the material during the compaction, the unrecoverable deformations, the reduction of
air voids and so on.
FEM Modeling
There are few studies in the continuum regime for investigating the compaction
mechanisms. The use of mixture theory, to certain degree explains the void reduction
during the compaction process.
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 49
One of the more popular methods is using a viscoplasticity model in conjunction with
FEM numerical simulations. Henny and Huerne (2002) studied the simulation of
compaction process with an analogy between hot asphalt mixtures and wet soils. A
"Critical State" material model based on soil mechanics was used as a fundamental
model to describe the behavior of asphalt mixture in his study. He analyzed the
influence of the compaction rate on mechanical properties of asphalt mixtures with
the help of FEM simulation. A non-standard FEM approach, the Arbitrary
Lagrangian Eulerian (ALE) method, was used to simulate the behavior of mixture
during compaction considering the fact that the behavior of the material during this
stage is between a solid and a liquid, hi this paper we will use a continuum type of
porous viscoplasticity model to simulate the compaction process. Improvement may
be made through incorporation of microstructures (Hu et al, 2005;Zhang et al., 2006).
Yield function:
Plastic flow:
Strain hardening:
Matrix material (fully dense):
Rate dependent yielding can be modeled in ABAQUS and the failure can be modeled
in ABAQUS explicit. Temperature dependent material parameters can be defined as a
tabular function of temperature. The model gives reasonable results forf< 10%.
Notation
(Ji , GI The total stress and strain tensor
50 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
FEM Simulations
Model
In the compaction simulation model (FIG. 2), the roller was assumed to be rigid and
the porous plasticity model was used for Asphalt Concrete (AC). Besides the constant
loading applied on the roller, a vibration force with smaller magnitude was also
applied on the roller. The simulation was conducted using ABAQUS [1995].
Variables
Force applied on the roller: IF, 2F, 3F, 4F
Vibration of the roller: 0.02F*sin (62.8T)
Thickness of the AC section: l.Ot, 1.5t, 2.0t, 2.5t
Where F=5000N, t=30mm, and T is the loading time.
(a) (b)
(c) (d)
FIG. 3 Element volume change for different roller forces: (a) IF, (b) 2F, (c) 3F and (d)
4F. The volume change was obtained at the bottom, the mid height and other
two points between them
(a) (b)
FIG. 4 Change of element volume under different roller pressures at (a) bottom and
(b) mid height of a typical section
52 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
Besides the reduction of the volume of asphalt concrete, the WF was also obtained
for the same asphalt layers, i.e. lower portion of the asphalt concrete in compaction.
FIG. 5 shows the evolution of the WF under different levels of compaction pressure.
In general, larger compaction force will make the WF reduce quickly. However, at
certain level, e.g. 3F and 4F, the effect of the increased compaction force is not so
obvious. We may also conclude from the plot that the compaction pressure should be
between 2F and 3F for compaction processes with 12 passages. It may require less
passages if the compaction forces are larger.
(a) (b)
(c) (d)
FIG. 5 Change of WF for different roller forces: (a) IF, (b) 2F, (c) 3F and (d) 4F. The
WF was obtained at the bottom, the mid height and other two points between
The changes of WF at bottom and mid height of the AC section against the roller
passage were plotted in FIG. 6. It can be seen that the targeted compaction air void
content (or density) can be achieved much earlier (4-6 passages) for forces 3F and/or
above in this case.
(a) (b)
FIG. 6 Change of WF under different roller pressures at (a) bottom and (b) mid height
FIG. 7 shows the final EVOL and WF (when additional passages would not further
reduce the air void or increase the density) under different compaction pressures. For
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 53
both the EVOL and WF, the increase of the compaction pressure will lead to
minimal difference in air void reduction if the pressure reached 3F.
(a) (b)
FIG. 7 Final EVOL and WF under different compaction pressure
FIG. 8(a) shows the change of EVOL at the bottom of the AC section for different
thickness. Generally, the volume decreases of asphalt concrete are in a similar
manner for different thickness. At mid height (FIG. 8(b)), changes of air void volume
vary more largely, indicating a potential of less compaction efficiency for the middle
layer of asphalt concrete. It should be noted that the values in this plot are not from
the same depth (absolute values) to the surfaces. This simulation can help determine
the optimal layer thickness for compaction. For example, from Figure 8b, it may be
concluded that 2t-2.5t (60mm-75mm) are too high for effective compaction.
(a) (b)
FIG. 8 Change of EVOL for asphalt concrete with different thickness at: (a) bottom and
(b) mid height of the compaction section
The evolution of WF shown in FIG. 9 also indicates that the compaction becomes
difficult when the thickness of the asphalt concrete layer increases. At the bottom of
the asphalt concrete section, the WF had almost no change for thickness 2.0t and
2.5t (60mm-75mm), indicating too larger thickness. At the mid height, the
compaction pressure applied can meet the compaction requirement for thickness less
than 2.0t.
(a) (b)
FIG. 9 Change of WF for asphalt concrete with different thickness at: (a) bottom and
(b) mid height of the compaction section
54 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
FIG. 10 shows the deformation profile of the centerline. For different thickness, due to
the lateral extension of the AC section in both directions (transversely and
longitudinally), the evolution of the deformation profile and the maximum
deformation at the compaction center are not proportional to the thickness. The
maximum deformation is for thickness around 2.0t (FIG.l 1), indicating the optimal
layer thickness for the compaction in terms of effectiveness and costs.
(a) (b)
FIG. 10 Deformation profiles: (a) different thickness of the compaction section at
compaction pressure 3F, and (b) different compaction pressures (thickness
LOt)
(a) fb)
FIG. 11 Center maximum deformations versus (a) thickness and (b) compaction pressure
It can be seen from FIG. 12 that the effect of base/subbase stiffness on compaction is
greater at lower part of the asphalt concrete layer than that at the upper part.
From the evolution of the WF (Figure 13), the stiffness of the base/subbase seems to
have similar effects on the change of WF.
If we compare the evolution of EVOL at bottom, mid height and top of the asphalt
concrete layer, the same trend was obtained for all the stiffness ratio variations (FIG.
14).
This also can be seen from comparison of the WF along the thickness of the asphalt
concrete layer (FIG. 15).
It is interesting to see that deformation at top of the asphalt concrete layer is smallest
for the least stiff subbase (FIG. 16). From the EVOL and the WF evolution, it has
least volume change and void reduction. We may conclude that this is due to the
deformation of the base/subbase, which causes difficulties the compaction of asphalt
concrete layer.
56 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
FIG. 16 final deformation profiles at top of the asphalt concrete layer for stiffness ratio:
(a) 5:1, (b) 1:1 and (c) 1:2
FIG. 17 maximum deformation of the asphalt concrete layer at center of the compaction
versus stiffness ration between subbase material and the asphalt layer
DEM MECHANISMS
Compared with FEM, DEM has its own characteristics and advantages. For example,
after applying the image technique (Fu, 2005), a realistic representation of the
internal structure allows modeling the deformation of composite materials more
accurately. FEM, as a continuum method, can successfully capture the stress-strain
distribution within the asphalt mixtures and its effect on the stiffness anisotropy
(Abbas, 2004). However, it lacks the ability to account for the slippage between the
aggregate particles, which has been cited as one of the most important mechanisms
resulting in permanent deformation or rutting. Such a limitation can be overcome by
DEM, which could consider the movement and behavior of those individual particles.
For these reasons, continuum codes with interface elements are restrictive in their
applicability. A class of computer programs collectively described as discrete element
58 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
2. Modal methods are similar to the distinct element method in the case of rigid
blocks but, for deformable bodies, modal superposition is used. This method appears
to be better-suited for loosely-packed discontinua; in dynamic simulation of dense
packing. Eigenmodes are apparently not revised to account for additional contact
constraints by this approach. A representative code is CICE.
3. Discontinuous deformation analysis assumes contacts are rigid bodies, and bodies
may be rigid or deformable. The condition of no-penetration is achieved by an
iterative scheme; the deformability comes from superposition of strain modes. The
relevant computer program is DDA.
DEM SIMULATIONS
With the descriptions in the previous section, we can see the differences between
FEM and DEM approaches. We can also appreciate the unique capability of the DEM
approach, where the material structure can be better incorporated into simulations.
Figure 18 illustrates the DEM model for investigating the compaction mechanism.
Figures 18a and 18b show the roller's moving while Figure 18c illustrates the
boundary walls. In the following explanations, Figure 18c may be referred in the next
sections for the wall locations where different boundary conditions can be applied.
a. b. c.
FIG. 18 Illustration of the DEM Model
FIG. 20 Particle Rotations: a) against X-axis; b) against Y-axis; and c) against Z -axis
(the first 1000 steps are ball-generation process while the compaction starts at step
1001).
In this evaluation, we chose to keep the ks (shear stifmess) value of the model a constant
and change the kn (normal stifmess) value, and monitor the peak forces in the z direction
against the roller wall (Id=l 1, Figure 18c) in the same compaction process, where the
roller is gradually rolling into the pavement through a displacement controlled algorithm.
Figure 21 plots the peak forces for three cases when the contact normal stifmess
increases. It may be concluded from Figure 21 that as the normal stifmess of the
contacting particles increases, the peak z-forces on the roller increases. This means more
efforts are needed if the same compaction is to achieve. It should be noted that the
original porosity or air void content for each of the three cases is kept the same.
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 61
FIG. 21 Peak forces required to press the roller into pavement for the same rolling
process
Results-Temperature Effect
Obviously, when the mixture temperature drops during transportation, the stiffness of the
binder/mastics will increase resulting in an increase of the normal contact stiffness. From
Figure 21, it may be inducted that the lower the temperature the more difficult to
compact. Another situation that involves the temperature effect is when the surface
temperature drops during the compaction, forming a relatively stiffer layer at the surface.
It is anticipated the larger the temperature drop, the thicker and stiffer the surface hard
layer.
The surface temperature drop effect is investigated through the simulation of the same
compaction process in the particle contact property investigation, keeping the others the
same while changing the stiffness of the top layer. Two cases are considered: case one
considers a smaller top hard layer (1/5 of the asphalt layer) while case two considers the
relatively larger thickness (2/5 of the asphalt layer) of the top colder/harder layer. The
normal contact stiffness change is also related to the degree of temperature drop. The
monitored peak z-forces (for parallel bonding) are presented in Figure 22. From Figure
22, it may be concluded that the peak z-force decreases as the stiffness of the top layer
becomes larger and as time goes longer, which means the stiffer layer becomes thicker,
the compaction effort will increase; larger compaction load/effort may be needed.
62 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
Both FEM and DEM simulations of the compaction process have been conducted for
understanding the mechanics of compaction. Both approaches have demonstrated
their uniqueness in the interpretation of the fundamental kinematics of compaction.
The FEM approach can model the air void reduction process with reasonable
accuracy. The DEM approach on the other hand can provide a fundamental
understanding of the particle kinematics, and relative binder/mastics to aggregate
stiffness. The simulation results indicate that both methods, if appropriately
calibrated, can serve as guide for selection of compaction parameters such as layer
thickness, compactor weight/vibration, number of passages and temperature. These
methods can also be used to evaluate the compaction effectiveness.
REFERENCES
ABAQUS (1995). User's Manual. Hibbitt, Karlsson & Sorensen, Inc.,Pawtucket, R.I.
Cundall, P.A. and Strack, O.D.L. (1979). A distinct element model for granular
assemblies. Geotechnique, 29: 47-65.
Henny ter Huerne (2002). Possibilities for material characterization and FEM
simulation of compaction process of asphalt pavement. Heron, 45.
Hu, RL, Yue, ZQ, Tham, LG, and Wang, LC, (2005). Digital image analysis of
dynamic compaction effects on clayey fills, Journal of Geotechnical and
Geoenvironmental Engineering, ASCE, 131(11): 1411-1422.
Zhang, B., Wang, LB., and Tumay, M. (2006) "An Evaluation of the Stress Non-
uniformity due to the Heterogeneity of AC in the Indirect Tensile Test", ASCE
Geotechnical Special Publication, 146: 29^43.
ANALYSIS OF NONLINEAR VISCOELASTIC PROPERTIES
OF ASPHALT MIXTURES
ABSTRACT
This study presents characterization of the nonlinear viscoelastic behavior of
hot mix asphalt (HMA) at different temperatures and strain levels using the
Schapery nonlinear viscoelastic model. This model is employed to describe
experimental measurements of two asphalt mixes under several combined
temperatures and strain levels. The master curve is created for each strain level
using time temperature superposition principle (TTSP) with a reference
temperature of 40°C. The measurements at strain levels higher than 0.01% are
used to determine the nonlinear viscoelastic parameters in the model. The FE
model with the calibrated time-dependent and nonlinear material parameters is
used to simulate the creep experimental tests, and good predictions are shown.
KEY WORDS: nonlinear viscoelasticity, hot mix asphalt, Schapery theory, finite
element
1 INTRODUCTION
The nonlinear behavior of asphalt mixtures can be caused by the rotation and
slippage of aggregates and the localized high strains in the binder phase (Kose et
al., 2000). Abbas et al. (2004) developed an incremental nonlinear viscoelastic
model to simulate the behavior of asphalt binders within the asphalt mix
microstructure using finite element (FE) analysis. The results confirmed that the
high strains in the binder phase can cause nonlinear viscoelastic response of the
mix.
Graduate Research Assistant, Department of Civil Engineering, Texas A&M University, College
Station, TX 77845-3136, Tel:979 845 8308, Fax: 979 845 0278, Email :[email protected]
2
Associate Professor, Department of Civil Engineering, Texas A&M University, College Station,
TX 77845-3136, Tel: 979 845 8308, Fax: 979 845 0278 , Email: [email protected].
3
Assistant Professor, Department of Mechanical Engineering, Texas A&M University, College
Station, TX 77845-3123, Tel: 979 458 3579, Fax: 979 845 3081, Email: [email protected]
4
Professor, Department of Civil Engineering, University of Wisconsin-Madison, Madison, Wl
53706-1691, Tel: 608 265 448 IFax: 608 262 5199Email: bahiafaengr.wise.edu
64
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 65 65
where Dn is the n{ coefficient of Prony series and A,n is the n retardation time.
In this paper, the recursive-iterative integration approach developed by
Haj-Ali and Muliana (2004) is used to implement the Schapery nonlinear
viscoelastic model. The recurrence formula allows bypassing the need to store
entire strain histories at the material levels. The linear strain formulation is used
within the recursive approach to give the trial solutions. The stress corrector
scheme is added to minimize error arising from the linearization. This will
enhance convergence and divergence. Furthermore, the recursive-iterative
approach permits using a larger time increment, while giving accurate solution.
The strain response for isotropic materials can be decoupled into deviatoric
and volumetric parts and it can be presented as:
where G and K are shear modulus and bulk modulus, respectively. J and
B are shear compliance and bulk compliance, respectively. Sv is the deviatoric
stress and akk is the volumetric stress. Applying the Schapery integral
constitutive model, the deviatoric and volumetric strain can be expressed as:
66 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
where etj is the deviatoric strain and e^ is the volumetric strain. J0 and B0
are instantaneous elastic shear compliance and instantaneous elastic bulk
compliance, respectively. AJ and A# are transient shear compliance and
transient bulk compliance, respectively.
Assuming the Poisson's ratio v to be time independent, the instantaneous
elastic shear compliance, instantaneous elastic bulk compliances, transient shear
and bulk compliances can be represented as:
Substituting Eqs. (3) and (7) into (5) and (6), the deviatoric and volumetric can be
written in terms of hereditary integral formulation and as follows:
For implementation in finite element method, the incremental shear and bulk
strains are derived and shown as:
The variables q\~* and q^* are the shear and volumetric hereditary integrals,
respectively for every Prony series term n at previous time t - A* . The
hereditary integrals are updated at the end of every converged time increment,
which will be used for the next time increment. The formulation of shear and
volumetric hereditary integrals are:
The shear and volumetric strain increments can be determined from Eqs. (10)
and (11) provided that the stresses are given. This algorithm will be
implemented in the displacement based FE framework, in which strains are the
given variables. The current shear and volumetric stresses and the current
nonlinear parameters can not be determined directly, because the nonlinear
parameters are dependent on the current stress and vise versa. Hence, the
iterative algorithm is added to solve for the current stress state, in which the
nonlinear parameters are assumed at the beginning of each time increment
g'a = 8a" ; « = W»2 and Ayr' = Ayr ''*
Then, the trial stresses can be determined as follows:
where J and B ' have the same forms as Eqs. (12) and (13), respectively, but
the nonlinear parameters are assumed to be functions of the last converged stress
state.
In this paper, the iterative scheme is used to calculate the correct stress state
from the current strain increment. In the iterative scheme algorithm, the residual
strain should be defined and it can be determined by calculating the current strain.
The residual strain equation can be written as:
where A^J. is provided from the structural level. The Newton-Raphson typed
iterative algorithm is used to minimize the strain residual in Eq. (18).
3. EXPERIMENTAL MEASUREMENTS
The dynamic frequency sweep test was used to characterize the linear
viscoelastic coefficients and nonlinear parameters. This test applies repeated
shear loading on the upper surface of a HMA specimen with a height of
approximately 50 mm and a diameter of about 150 mm. This test was conducted
under several temperatures and strain levels. The temperatures, frequencies and
strain levels were as follows:
68 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
The time-temperature shifting is used to obtain the master curve at each of the
strain levels and a reference temperature of 40°C. Figures 1 and 2 show the
relationship between temperature and time-temperature factor aT for the fine and
coarse mixes, respectively. The nonlinear parameters (gig2) are obtained by
vertical shifting of the master curves at all strain levels to a reference strain of
0.01%. Figure 3 shows the relationship between gig2 and strain levels for the
fine mix. As expected, the nonlinear parameter gi g2 increases with an increase in
strain level. The master curves at the different strain levels are also shifted
horizontally to the reference strain of 0.01% in order to determine the time-strain
shift factors (as) and obtain the long-term HMA behavior. The long-term linear
viscoelastic Prony coefficients are obtained by fitting all the data shifted
horizontally to the 0.01% strain. Figure 4 presents the long-term master curves for
the fine mix. The results show that the Prony series can fit the temperature-strain
master curve very well.
Temperature (n )
FIG. 4. The long-term master curve of strain level 0.01% for fine HMA
mixes.
4. MODEL VERIFICATION
Inverse analysis is conducted to determine the ability of the parameters
obtained in establishing the master curve and used in FE analysis to match the
experimental measurements at different combinations of temperatures and strain
levels. The time-temperature shift factors, nonlinear viscoelastic parameters and
the long-term linear viscoelastic coefficients obtained from the analysis of the
experimental tests are used as input parameters to the material subroutine of the
finite element. Examples of the results for the fine and coarse mixtures are
shown in Figures 5 and 6. In general, the numerical results have very good
agreement with the experimental measurements. The differences between the
model and measurements are attributed to the inherent errors in determining the
model parameters. For example, aT in Figures 1 and 2 vary slightly as a
function of strain, while aT is taken as the average of all strain levels in the FE
model.
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 71
FIG. 5. The verification of strain level 0.04% for fine HMA mixes.
FIG. 6. The verification of strain level 0.04% for coarse HMA mixes.
5. SUMMARY OF FINDINGS
The nonlinear viscoelastic response of asphalt mixtures is described in this
study using the Schapery's single integral model. The model parameters were
obtained by analyzing the nonlinear viscoelastic response of two asphalt mixes
tested at different temperatures, frequencies and strain levels. The time-strain
72 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
shift factors were obtained by shifting the master curves at the different strain
levels horizontally to the reference strain within the linear range. The nonlinear
parameters were obtained by vertical shifting of the master curves at all strain
levels to the same reference strain. The long-term linear viscoelastic coefficients
were determined by fitting the Prony series to the data shifted horizontally at the
reference strain. The time-temperature shift factors were determined and found
to vary only slightly as a function of strain level.
Finite element analysis was conducted to simulate the experimental
measurements. The analysis showed that the FE model matched the
experimental measurements at different combinations of temperatures and strain
levels. The nonlinear viscoelastic material subroutine implemented in finite
element can be used to model the time dependent and stress dependent response of
asphalt pavements under different environmental and loading conditions. In
addition, research is currently underway to use the model in representing the
nonlinear viscoelastic response of asphalt binders.
6. REFERENCES
Abbas, A., Papagiannakis, T., and Masad, E. (2004). "Linear and Non-Linear
Viscoelastic Analysis of the Microstructure of Asphalt Concretes," Journal of
Materials in Civil Engineering, ASCE, Vol. 16, No. 2, pp. 133 - 139.
Christensen, R.M. (1968). "On Obtaining Solutions in Nonlinear
Viscoelasticity." Journal of Applied Mechanics, Vol. 35, pp. 129—133.
Haj-Ali, R.M. and Muliana, A.H. (2004). "Numerical Finite Element
Formulation of the Schapery Nonlinear Viscoelastic Material Model."
International Journal for Numerical Methods in Engineering, Vol. 59, pp.
25-45
Kose, S., Guler, M., Bahia H., and Masad, E. (2000). "Distribution of Strains
within Hot-Mix Asphalt Binders." Journal of the Transportation Research
Board No. 1728, pp. 21-27.
Masad, E. and Niranjanan, S. (2002). "Microstructural Finite Element Analysis
of the Influence of Localized Strain Distribution on Asphalt Mix Properties,"
Journal of Engineering Mechanics, ASCE, Vol. 129, No. 10, pp. 1105-1114.
Schapery, R.A. (1969). "On the Characterization of Nonlinear Viscoelastic
Materials." Polymer Engineering and Science, Vol. 9, pp. 295-310.
Schapery R. A. (2000). "Nonlinear Viscoelastic Solids." International Journal
of Solids and Structures, pp. 359-366.
THE TEMPERATURE DEPENDENT
GENERALIZED KUHN MODEL FOR
ASPHALT CONCRETE
Vassilis P. Panoskaltsis
ABSTRACT
An internal variable formulation for the introduction of the effects of temperature in
viscoelastic models is presented. A new model, the temperature-dependent
generalized Kuhn model of viscoelasticity, is developed by using this powerful and
novel approach. An alternative formulation, through the use of the creep function
and the time-temperature superposition principle, is also presented. The new model's
predictions are compared to experimental data.
INTRODUCTION
In this paper we develop a new temperature-dependent viscoelastic model for the
description and prediction of the behavior of asphalt concrete. The new model can be
also used for the analysis of pavements in cases in which the influence of temperature
is important. Furthermore, the incorporation of temperature is essential for modeling
the fatigue behavior of the material. The temperature-dependent viscoelastic model
developed here, is based on a viscoelastic model recently proposed by Panoskaltsis
and co-workers (Panoskaltsis et al. 2006), named by them "The generalized Kuhn
model of viscoelasticity". In order to incorporate temperature effects in the model,
the classical approach, having been referred to in the literature as "the time-
temperature superposition principle", is used in this work. Moreover, a novel
approach for the incorporation of temperature is presented. This approach is based on
the concept of internal variables of inelasticity and as it will be shown here it is very
powerful. The paper is organized as follows: In the first section a brief overview of
the generalized Kuhn model is presented. In the second section the new temperature-
73
74 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
dependent model is developed. In the third section the internal variable approach for
the introduction of temperature in viscoelastic models is developed. In the last
section the capabilities of the new model to predict experimental results are shown.
where A, B, C and a are parameters and 0 < a < 1. For a = 0 the modified Kuhn
model is obtained. The creep compliance of the discrete generalized Kuhn model
(rheological representation) is given by the following truncated Dirichlet (or, as
alternatively called, Prony) series
The creep function JN(t) has as a limit the function J(t) for N -> oo. Observe that
JN(t) is the creep function of a generalized Kelvin-Voigt model, shown in Fig. 1,
consisting of a spring with compliance A in series with N +1 Kelvin-Voigt units. In
the discrete GKM the stiffnesses of the springs are not constant, unlike the discrete
modified Kuhn model. The stiffness of the spring and the viscosity coefficient of the
dashpot in the mth Kelvin-Voigt element, denoted by Em and rjm respectively, are
and
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 75
m
The retardation time in each Kelvin-Voigt unit is rm = — = — . We will call these
E C
m
Kelvin-Voigt units with the special properties, Kuhn units.
The generalized Kuhn model shares the remarkable property of the MKM, i.e. that no
matter how many Kuhn elements are used the number of parameters remains the
same. The internal variable representation of the discrete GKM is given by Eqs. (5)
and (6), which represent the evolution equations of the internal variables (interpreted
here as the inelastic strains) in multi dimensions.
where qdm and Om are the deviatoric and volumetric components respectively of the
viscous strain tensor qm of the m'k Kuhn element, i.e.
76 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
*„=<*.. (8)
where 1 is the identity tensor of rank 2, s is the deviator of the stress tensor a and tr
stands for the trace operator. The superimposed dot in Eqs. (5) and (6) indicates time
derivative. The total viscous strain tensor ev for the discrete GKM is
The total strain tensor is the sum of elastic and viscous strain tensors, i.e.
This decomposition is well justified in the small strain framework, which is used in
this work. Eqs. (5) to (10) give a complete characterization of the GKM in an
internal variable form. Both the Theological representation of the discrete GKM,
described by Eq. (2), and its internal variable representation are equivalent forms of
the model. However, it should be noted that the internal variable form is
advantageous because it can be also expressed in multi-dimensions and it can be
extended into the nonlinear domain in a relatively easy manner.
The loss tangent -which is a measure of energy dissipation in viscoelasticity- of the
discrete generalized Kuhn model is given by the following relation (Panoskaltsis et
al. 2006)
where CD is the frequency of excitation. For the derivation of the expression for the
loss tangent the creep function and its Fourier transform are used. The discrete
generalized Kuhn model's loss tangent predictions of experimental results of asphalt
concrete are very good (Panoskaltsis et al. 2006).
where tT is the time to obtain a value of the creep function at temperature T, and tT is
the time to obtain the same value of the creep function at temperature T0 (the
reference temperature). For asphalt concrete, based on numerous experiments, the
logarithm of the shift factor aT is observed to be varying linearly with temperature,
as shown in Fig. 3 (see e.g. FHWA 1978, Huang 1993). The slope of the curve is
given by
A qualitative plot of creep functions with time on a linear scale, for asphalt concrete,
is given in Fig. 4. In the figure, T0 represents the reference temperature, Th a higher
than the reference temperature and Tt a lower than the reference temperature.
We assume that the creep function of the discrete generalized Kuhn model, given by
Eq. (2), is valid at the reference temperature TQ9 we denote it as J(t;T0)rnd we
rewrite Eq. (2) as
The problem at hand is to find the expression for the creep function at any other
temperature T. For this, we proceed as follows: Let us assume that at the current time
t -at which the creep value at the reference temperature is V (Fig. 4) - we want to
find the creep value at a lower temperature Tl. From Fig. 4 we see that this is equal
to '&' (on the creep curve corresponding to temperature T,). This value is equal to
the value V on the creep curve for the reference temperature T0. The value 'c'
occurred at an earlier time equal to ^ . Therefore, the value V can be obtained from
Eq. (15) if we substitute for time the value tTo. The relation between t and tTo is given
in Eq. (14) (t.T in Eq. (14) is equal to t). By solving for tT we obtain from Eq. (14):
Substituting this value in Eq. (15) we obtain the value 'c' and therefore the value 'b\
The same exactly arguments hold if we want to obtain the creep value ' d ' at a higher
temperature Th. Therefore, the creep function at any temperature T, denoted as
J(t\T), is obtained if we substitute Eq. (16) in Eq. (15), i.e.
where
and
As it is seen from the above equations the parameters Bl and C{ are temperature
dependent.
Since we assumed that the creep function given by Eq. (2) or equivalently by Eq.
(15) holds at the reference temperature TQ, the expression for the loss tangent in Eq.
(11), which was derived by using Eq. (2), will also hold at the reference temperature
r o . hi order to obtain the loss tangent at any temperature T we should use Eq. (17).
Since Eq. (17) has the same exactly form as Eq. (2) the expression for the loss tangent
80 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
at any temperature Twill have the same form as Eq. (11) with the parameters B and C
replaced by the temperature dependent parameters Bl and C,, which are determined
from Eqs. (19) and (18) respectively. Therefore, the loss tangent at any temperature
T will be
From this expression it is seen that when the temperature increases the viscosity
coefficient decreases, which is in accordance to experimental results. It is assumed
that the stiffnesses of the springs remain independent of the temperature. The relation
between the relaxation times at temperatures T and TO will be
As it was mentioned earlier the internal variable formulation of the discrete GKM is
equivalent to its description via the creep function (Panoskaltsis et al. 2006).
The evolution equations of the internal variables, given by Eqs. (5) and (6), can be
written as follows at the reference temperature T0
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 81
Using Eq. (29) in Eqs. (27) and (28) the following equations are obtained,
82 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
where rfm and r*m represent the viscous parameter and relaxation time respectively at
temperature Tand they are given by Eqs. (32) and (33).
Eq. (32) shows the relation between the viscosity coefficient of the mth Kuhn element
at temperature Tand its value at the reference temperature T0. This is identical to the
relationship obtained in Eq. (21). Similarly, Eq. (33) shows the relation between the
relaxation time of the mth Kuhn element at temperature Tand its value at the reference
temperature T0 and this is identical to the relationship obtained in Eq. (22). Eqs.
(30), (31), (32) and (33) constitute the internal variable formulation of the discrete
generalized Kuhn model at any temperature T. The internal variable formulation is
more general than the formulation based on the creep function, since it also holds in
the multidimensional spaces of stress and strain tensors. Moreover, it does not
depend on the existence of a closed form for the creep function and it can be used in
cases of nonlinearities. It should be also mentioned that the internal variable
formulation is amenable to a finite element implementation of the model.
RESULTS
The loss tangent given by Eq. (20) is used to fit the experimental results reported by
Cerni (2001). Bitumen specimens are subjected to frequency sweep experiments at
various temperatures and the plots of phase angle 6 (in degrees) versus frequency (in
Hz) at temperatures 6°C, 15°C and 25°C are obtained and shown in Fig. 5. Log scale
has been used in the frequency axis. The parameters are estimated from model fit of
the experimental results. The parameter estimation is done by nonlinear optimization
techniques and programming is done in MATLAB. The objective function used for
the parameter estimation is given by Eq. (34),
where ltmi and //exp,,. are ith model and experimental loss tangent values and M is the
total number of experimental data points. Fig. 5 shows the model fit at 6°C, 15°C,
and 25°C. The values of the parameters are listed in Table 1. Fifteen Kuhn elements
were used. It is noted that the values of the parameters A, a and r are temperature
independent as they should.
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 83
Temperature A B C a r
Ti=6°C 1.15 Bi=25.10 Ci=0.0196 0.43 1.01
T2=15°C 1.15 B2=34.30 C2=0.0404 0.43 1.01
T3=25°C 1.15 B3=48.52 C3=0.0907 0.43 1.01
For the validation of the model the following procedure is followed. Using the values
of the parameters obtained for temperatures 6°C and 15°C, the values for the
temperature dependent parameters B and C corresponding to 25°C are calculated.
These calculated values are compared with the parameter values which have been
obtained by the fitting process. For this, the slope ft of the variation of the log of the
shift factor aT with temperature is calculated from Eq. (18) (treating Tj as the
reference temperature), i.e.
84 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
where TI, 7^ C/ and C? are given in Table 1. From Eq. (35), the value of the slope
J3 is calculated as J3 = -0.035. Clearly, instead of Eq. (18), Eq. (19) could have been
used.
The values of the parameters B and C for 25°C are calculated from the following
equations, using the obtained value for /?.
These calculated (predicted) values are almost equal to the fitted values B3 and €3
given in Table 1. This provides an excellent validation of the model, since the
predicted values match the fitted ones. Therefore, the model is able to predict the
variation of the loss tangent with the frequency of excitation at different temperatures
by computing the corresponding to the temperatures parameters B and C, following
the procedure just shown, and then using Eq. (20). We note that as is seen in Fig. 5,
the loss tangent values are increasing with temperature, which means that the
dissipated energy is higher at higher temperatures.
The model is validated against a second set of experiments reported by Corte (2001).
In this case, dynamic shear tests were conducted at various temperatures on four
different asphalt concrete mixes with penetration grades (PG) 10/20, 25/35, 35/50,
and 50/70. Phase angle variation (in degrees) with temperature at 7.8 Hz has been
recorded. In this case for each concrete mix Eq. (20) is expressed as a function of the
temperature, and the parameters A, B, C, a,r,J3, with the help of Eqs. (18) and (19).
B and C are the values of the parameters at the reference temperature, taken here to
be 0°C. The objective function is given by Eq. (34). The parameters are tabulated in
Table 2 and the number of Kuhn elements used for the fitting is 15. Fig. 6 shows the
experimental results and the model's fit to the experimental data. It is evident from
the figure that the generalized Kuhn model fits phase angle variation with respect to
temperature with very good accuracy. The values of the parameters Bj and C/ at any
other temperature can be obtained by using Eqs. (19) and (18) respectively and thus
the phase angle value at this temperature is computed with the use of Eq. (20), for any
of those four asphalt mixes and for frequency equal to 7.8 Hz.
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 85
Figure 6. Model fit for phase angle versus temperature at 7.8 Hz.
Experimental data is from Cortc (2001).
Again, we observe from Fig. 6 that for all asphalt concrete mixes the phase angle
(and hence the dissipated energy) increases with temperature, as it is expected.
86 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
CONCLUSIONS
REFERENCES
FHWA. (1978). "Predictive design procedures". VESYS Users Manual, Report No.
FHWA-RD-77-154, Federal Highway Administration.
Lubliner, J., and Panoskaltsis, V. P. (1992). "The modified Kuhn model of linear
viscoelasticity." InternationalJournal of Solids and Structures, 29 (24), 3099-
3112.
Panoskaltsis, V. P., Papoulia, K. D., Bahuguna, S., and Korovajchuk, I. (2006). "The
generalized Kuhn model of linear viscoelasticity." In review.
Panoskaltsis, V. P., Bahuguna, S., Papoulia, K. D., and Lubliner, J. (1999). "Finite
element analysis of rate dependence and failure of concrete." Proceedings of the
European Conference on Computational Methods, H. A. Mang et al. eds., 31
August - 3 September, 1999, Munich, Germany.
Myung Goo Jeong1, Amara Loulizi2, P.E., and Gerardo W. Flintsch3, M. ASCE
1
Graduate Research Assistant (Former GRA at Virginia Tech, 2003-2005), Arizona State University,
Department of Civil and Environmental Eng., PO Box 875306, Tempe, AZ 85287-5306,
mgj [email protected]
2
Assistant Professor, Ecole Nationale d'ingenieur de Tunis, ENIT Departement Genie Civil B.P. 37 Le
Belvedere, 1002 Tunis, Tunisia, [email protected]
3
Associate Professor, Charles Via, JR. Department of Civil and Environmental Engineering and ,
Director, Center for Sustainable Transportation Infrastructure, Virginia Tech Transportation Institute,
3500 Transportation Research Plaza, Blacksburg, VA 24061-0105, [email protected]
87
88 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
INTRODUCTION
Determining the viscoelastic properties of hot mix asphalt (HMA) is important for
several applications, including modeling flexible pavement response to truck and
environmental loadings, predicting the performance of the material in terms of
specific distress types (e.g., low-temperature cracking and rutting), and for
specifications as quality control measures (e.g., limiting a function of the dynamic
modulus and the phase angle to a specified passing value).
Rheologists have studied the viscoelastic properties of polymers for many years, and
the theory governing viscoelastic behavior of these materials is well established (1).
Among the tests used by rheologists to determine viscoelastic parameters are static
creep compliance and dynamic modulus tests. These tests are now gaining wide
acceptance among highway agencies to measure viscoelastic properties of HMA.
Asphalt material researchers have proven that the interconversion between those
viscoelastic properties measured on asphalt mixtures using the theory worker
relatively well at low temperatures (i.e., linear condition) (2).
In 2005, the Virginia Department of Transportation (VDOT) and the Virginia Tech
Transportation Institute reviewed existing methods for determining the moduli of
HMA and identified simple laboratory tests that accurately describe the constitutive
behavior of HMA used in the Commonwealth of Virginia (3). The performed tests
evaluated included uniaxial static creep compliance and dynamic modulus. The
results from that study presented an opportunity to verify whether existing
interconversion methods between viscoelastic properties could be applied to the
HMA at wider temperature ranges. This would eliminate the need to perform one of
the two tests as the results from one test could be used to calculate the properties
obtained from the other. This paper presents the measured data on both mixes, the
procedure used to convert from one property to another and a comparison of the
converted property with the measured data.
EXPERIMENTAL PROGRAM
Two typical HMA mixtures used in Virginia, SM-9.5A and BM-25.0, were used in
this study. The SM-9.5A mix is a surface mix with a 12.5-mm maximum nominal
aggregate size and a PG64-22 binder. The BM-25.0 is a base mix with a 25.0-mm
maximum nominal aggregate size and a PG64-22 binder. The mixes were designed
according to VDOT specifications (4), which follow the SUPERPAVE procedures
with some minor odifications. Table 1 shows the job mix formula with the source of
the aggregates for both mixes. Table 2 presents the average aggregate gradation for
both mixes.
All specimens prepared for testing were compacted using a Gyratory Compactor.
The specimens were compacted to achieve approximately 4% ± 1 % voids in total mix
(VTM) for the molded specimen. Therefore, the specimens were compacted by
fixing the height rather than fixing the number of required gyrations. Once a
specimen was extracted from the TGC, its bulk density (Gmb) was measured using the
AASHTO T166 procedure (5), and then it was cored and cut to the final specimen
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 89
dimensions of 102 mm by 152 mm. Another bulk density was then performed on the
final specimen since it is known that there is a slight difference in the measured VTM
between the compacted mold specimen and the final cored and cut specimen.
Table 3 presents the Gmb and VTM values for all prepared specimens after
extraction from the TGC and after coring and cutting. For the SM-9.5A mix, the
average VTM for the dynamic modulus specimens were 4.2% and 2.7% right after
extraction from the TGC and after coring and cutting, respectively. The static creep
test specimens had 4.3% and 3.1% VTM after extraction from the TGC and after
90 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
coring and cutting, respectively. For the BM-25.0 mix, on average, the VTM for the
dynamic modulus specimens were 5.0% and 3.7% right after extraction from the
TGC and after coring and cutting, respectively. The static creep test specimens had
5.0% and 3.6% VTM after extraction from the TGC and after coring and cutting,
respectively.
TABLE 3. Gmb and VTM for all prepared specimens
Dynamic Modulus Static Creep
Extracted Final Extracted Final
Specimen Specimen Specimen Specimen
VTM VTM VTM VTM
Label Gmb Gmb Label Gmb Gmb
(2) (%) (4) (%) (6) (7) (%) (9) (%)
(1) (3) (5) (8) (10)
S93 2.365 4.1 2.402 2.6 S109 2.363 4.2 2.400 2.7
S94 2.366 4.1 2.402 2.6 S110 2.364 4.2 2.405 2.5
S95 2.365 4.1 2.398 2.8 Sill 2.361 4.3 2.396 2.9
3
ON
S96
S97
S98
2.367
2.368
2.363
4.1
4.0
4.2
2.403
2.427
2.392
2.6
1.6
3.0
S112
S113
S114
2.358
2.370
2.365
4.4
3.9
4.1
2.393
2.406
2.395
3.0
2.5
2.9
1 S101
S102
2.364
2.360
4.2
4.3
2.399
2.388
2.7
3.2
S115
S116
2.356
2.362
4.5
4.3
2.392
2.387
3.1
3.2
S105 2.366 4.1 2.392 3.1 S117 2.356 4.5 2.362 4.2
S107 2.362 4.3 2.388 3.2 S119 2.362 4.3 2.368 4.0
Average 2.365 4.2 2.399 2.7 Average 2.362 4.3 2.390 3.1
B56 2.473 4.9 2.510 3.5 B55 2.464 5.2 2.510 3.5
B62 2.473 4.9 2.511 3.4 B58 2.467 5.1 2.509 3.5
B63 2.470 5.0 2.517 3.2 B60 2.463 5.3 2.493 4.2
O
B64 2.473 4.9 2.496 4.0 B61 2.474 4.9 2.499 3.9
B65 2.465 5.2 2.511 3.5 B78 2.478 4.7 2.517 3.2
i
I/)
B67 2.473 4.9 2.497 4.0 B79 2.471 5.0 2.518 3.2
B68 2.471 5.0 2.502 3.8 B80 2.462 5.3 2.498 4.0
B69 2.471 5.0 2.508 3.6 B81 2.466 5.2 2.514 3.3
B76 2.470 5.0 2.488 4.3 B84 2.470 5.0 2.500 3.9
B77 2.474 4.9 2.513 3.4 B85 2.476 4.8 2.509 3.5
Average 2.471 5.0 2.505 3.7 Average 2.469 5.0 2.507 3.6
The dynamic modulus and static creep tests were performed with the same servo-
hydraulic machine and using the same extensometers to measure the vertical
deformations. Five temperatures were used:-15°C, 5°C, 20°C, 30°C, and 40°C. For
the static creep test, the applied load was sustained for 1,000 seconds, while for the
dynamic modulus test, six frequencies were used: 0.1 Hz, 0.5 Hz, 1 Hz, 5 Hz, 10 Hz,
and 25 Hz. hi both tests, two specimens were tested per temperature; therefore, 10
specimens per mix and per test were tested in total, as is presented in Table 3.
DYNAMIC MODULUS
Table 4 presents the average measured dynamic modulus and the phase angle for
both mixes. As expected, under a constant loading frequency, the magnitude of the
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 91
TABLE 4. Average dynamic modulus (MPa) and phase angle (°) results
-15°C 5°C 20°C 30°C 40°C
|E*| d |E*| d |E*| 6 |E*| d |E*| d
0) (2) (3) (4) (5) (6) (7) (8) (9) (10)
SM-9.5A
25 19,563 1.2 13,833 6.0 11,312 15.1 6,722 24.3 3,484 32.1
10 19,117 2.4 12,058 7.9 9,787 17.3 5,432 25.2 2,521 29.8
5 18,750 3.0 11,313 8.7 8,609 19.2 4,384 27.1 1,985 28.6
1 18,285 3.4 10,118 10.9 5,953 24.2 2,635 28.8 1,226 23.9
0.5 17,528 3.7 8,804 12.8 4,734 29.8 2,054 31.5 709 28.3
0.1 15,940 5.1 7,432 16.3 2,890 33.4 1,316 28.2 532 19.6
BM-25.0
25 24,483 1.3 18,687 8.0 13,395 r!5.1 7,978 23.1 4,029 30.7
10 24,243 2.8 17,431 7.8 11,549 17.1 6,568 24.4 3,122 28.5
5 23,867 3.0 16,485 9.1 10,104 19.1 5,366 26.1 2,843 26.5
1 22,484 3.9 13,993 11.1 6,952 23.9 3,246 29.5 1,708 23.0
0.5 21,428 4.7 12,426 13.5 5,534 29.2 2,132 36.9 956 27.8
0.1 19,991 5.8 10,408 16.6 3,419 33.6 1,316 35.1 725 20.5
92 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
(a) (b)
FIG. 1. Shift factors versus temperature: (a) SM-9.5A and (b) BM-25.0
CREEP COMPLIANCE
Figure 2 shows the measured creep compliances over the 1,000-s loading time at
each of the five tested temperatures for mixes SM-9.5A and BM-25.0. Each curve
represents the average of the two tested specimens at that temperature. A master
curve of the creep compliance was constructed for a reference temperature of 20°C
for both mixes. A visual method with the Excel spreadsheet software was used to
find the shift factors needed to construct the master curve.
(a) (b)
FIG. 2. Creep compliance for both mixes (a) SM-9.5A and (b) BM-25.0
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 93
This visual method was adopted because it produced appropriate results, and other
researchers had not found any significant differences between shift factors
determined manually and those determined by a computer program (8). The shift
factors used to construct the creep compliance master curve are presented in Table 5.
The table presents for comparison the shift factors used for the dynamic modulus data.
It is known that simple rheological material should have the same shift factor for all
viscoelastic properties. However, there are some differences between the shift factors
used for the dynamic modulus data and those used for the creep compliance data.
These differences could be due to experimental errors in both tests, errors with the
nonlinear regression procedure used for the dynamic modulus data, or a departure of
the mix behavior from the linear viscoelastic theory.
Once the creep compliance master curve was constructed, a Prony series
model was fit to the curve. This was performed because this type of fitting is needed
in order to convert the creep data into dynamic modulus data as will be explained in
the next section. The model is given by the following equation:
where
D(tr) = creep compliance at reduced time tr,
D(tr)is the equilibrium (glassy) creep compliance,
DI and Tt = Prony series parameters, and
n = the number of terms used in the series.
A five-term Prony series was found to be suitable to fit the data though there is a
relatively poor agreement in shorter time region. The use of additional terms to the
series would likely improve the fitting. Figure 3 shows the creep compliance master
curve with the Prony series fit for the SM-9.5A mix. Table 6 shows the calculated
parameters for both mixes. It can be observed that the model does not fit the
experimental data very well at the very short times, where it levels out more than the
experimental model.
94 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
FIG. 3. Master curve and Prony series fit for the SM-9.5A mix
TABLE 6. Prony series parameters for the creep compliance master curve
(4)
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 95
where
co - angular frequency
The complex compliance is then determined using Equation 6, and the dynamic
modulus can be calculated as the inverse of the complex compliance as shown by
Equation 7 (9).
Figure 4 shows the calculated dynamic modulus for the SM-9.5A mix using the
procedure described above. The figure also shows the measured dynamic modulus
data for the same mix with the 95% confidence interval at each measurement point.
It was assumed that at each testing temperature and frequency, the measured dynamic
modulus followed a normal distribution with a mean equal to the average of the tested
specimens and a standard deviation equal to the calculated standard deviation
between those specimens. It must be noted that the wide confidence intervals are
partially due to the small number of specimens tested. In other words, the intervals
would have been more accurate if they were computed by a standard deviation of
enough number of replicates; however only limited replicates were available. The
figure shows that the calculated dynamic modulus from the creep compliance data is
within the measured 95% confidence interval at some frequencies but out of this
interval at most frequencies. In most cases, the calculated dynamic modulus was
smaller than the measured one. In addition, the calculated dynamic modulus from the
creep data reached the £« modulus value (dynamic modulus value at infinite
frequency) at a frequency of around only 10 Hz. The calculated Eoofor the SM-9.5A
mix was 5.6 GPa.
The same findings were observed for the BM-25.0 mix as shown in Figure 5. The
calculated Eoofor the BM-25.0 mix was 14.4 GPa. This means that creep data
predicted a mix with weaker dynamic modulus than what was measured. This might
be due to the fact that the creep compliance was performed using a high constant load
of 690 kPa without any confinement. This stress might have caused the material to
be outside its linear viscoelastic region, making the predicted dynamic modulus
smaller than what was measured.
96 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
where
A' = adjustfimction(F(l - n) cos(nn 12)),
A " = adjust function ( F(l - n) sm(nn / 2)),
r°°
F = gamma function (F(w) = J u e~udu ), and
For this study, the storage modulus data was used to calculate the relaxation
modulus. The storage modulus (E") was calculated using the dynamic modulus and
phase angle (E" = |E*| cos8) and then plotted on a logarithmic scale. A sigmoidal
function was fitted to the storage modulus data using curve-fitting software. The
regressed analytical function was used to compute the local logarithmic slope of the
storage modulus (n) over the specified frequency range. The relaxation modulus at
each time was then computed. Figure 6, for example, shows the computed relaxation
modulus for the SM-9.5A mix.
where
E(t) = relaxation modulus
D(t) = creep compliance,
t = time, and
r = an integral variable.
where
E(t) = E}r" and
D(t) = Df.
Practically, lab-determined data are not exactly represented by the power law
function. However, if the data do not perfectly follow a power model but the
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 99
functions behave smoothly, Equation 11 still works well. In this case, the local slope
of the power model can be determined using Equation 12 (11).
Therefore, for this study, the determined relaxation modulus was fitted with a
sigmoidal function, whose derivative was used to find the values of n, and then the
creep compliance was calculated using Equation 11.
Figures 7 and 8 show the computed creep compliances for mixes SM-9.5A and
BM-25.0, respectively. The figures also show the measured creep compliance with a
95% confidence interval. For both mixes, the computed creep compliance was
smaller than the average measured one at almost all loading times. This may confirm
what was previously explained for the predicted dynamic modulus from the creep
compliance.
The measured dynamic modulus predicted smaller creep compliance (stronger mix)
than what was measured. One possible reason is that the creep test was performed
using a high constant stress of 690 kPa, which might have caused the material to be
outside its linear viscoelastic region.
CONCLUSIONS
The results from creep compliance and dynamic modulus testing of two HMA mixes
were used to verify whether existing interconversion methods between viscoelastic
properties could be used on HMA. This would eliminate the need to perform one of
the two tests as the results from one test could be used to calculate the properties
obtained from the other. The uniaxial creep compliance and dynamic modulus of two
typical mixes used in Virginia were measured at various temperatures and creep
compliance master curves, and dynamic modulus master curves were constructed
using the time-temperature superposition principle. The creep compliance data were
used to compute the dynamic modulus and vice versa using existing interconversion
procedures.
The investigation found that the converted properties did fall in most cases within
the 95% confidence interval of the measured data. The discrepancy between the
measured and calculated properties are mainly attributed to the fact that the creep test
was performed using a high, constant stress, which might have brought the material
outside its linear viscoelastic region. The interconversion between the viscoelastic
properties showed very promising results, but more testing with other mixes is
needed to confirm these findings.
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 101
REFERENCES
1. Ferry, J.D., Viscoelastic properties of polymers, 3rd ed., Wiley, New York, 1980.
2. Kim, Y.R. and Y. Lee, "Interrelationships Among Stiffnesses of Asphalt-
Aggregate Mixtures," Journal of the Association of Asphalt Paving
Technologists, Vol. 64, 1995, pp. 575-606.
3. Flintsch, G.W., I.L. Al-Qadi, A. Loulizi, and D. Mokarem, "Laboratory Tests for
Hot-Mix Asphalt Characterization in Virginia," VTRC Report 05-CR22,
Charlottesville, VA, 2005.
4. Virginia Department of Transportation, Road and Bridge Specifications, 2002.
5. American Association of State Highways and Transportation Officials.
AASHTO T166-00, Specific Gravity of Compacted Bituminous Mixtures Using
Saturated Surface-Dry Specimens, Standard Specifications for Transportation
Materials and Methods of Sampling and Testing, Part2 - Tests, 20th Edition,
Washington, DC, 2000, pp. 514^516.
6. Clyne, T. R., Li, X., Marasteanu, M. O., and Skok, E. L. Dynamic and Resilient
Modulus of MN DOT Asphalt Mixtures, Report No. MN/RC-2003-09, 2003,
Minnesota Department of Transportation, Minneapolis, MN.
7. Pellinen, T.K., and M.W. Witczak, "Stress Dependent Master Curve
Construction for Dynamic Modulus," Journal of the Association of Asphalt
Paving Technologists, Vol. 71, 2002, pp. 281-309.
8. Witczak, M.W., R. Roque, D.R. Hiltunen, and W.G. Buttlar, "Modification and
Re-Calibration of Superpave Thermal Cracking Model," NCHRP 9-19 Project
Report, Arizona State University Department of Civil and Environmental
Engineering, Tempe, Arizona, December 2000.
9. Park, S.W. and R.A. Schapery, "Methods of interconversion between linear
viscoelastic material functions. Part I—A numerical method based on Prony
series," International Journal of Solids and Structures, Vol. 36, pp. 1653-1675,
1999.
10. Schapery, R.A. and S.W. Park, "Methods of interconversion between linear
viscoelastic material functions. Part II—An approximate analytical method,"
International Journal of Solids and Structures, Vol. 36, pp. 1677-1699, 1999.
11. Park. S.W. and Y.R. Kim, "Interconversion Between Relaxation Modulus and
Creep Compliance for Viscoelastic Solids," ASCE Journal of Materials in Civil
Engineering, Vol. 11, No. 1, pp. 76-82,1999.
12. Leaderman, H., "Viscoelasticity phenomena in amorphous high polymeric
systems," Rheology, Vol. II, F.R. Eirich, Ed., Academic Press, NY, 1958.
A LABORATORY STUDY ON CRACKING POTENTIAL OF BINDER
COURSE ASPHALT MIXTURES USED IN SEMI-RIGID PAVEMENTS
INTRODUCTION
102
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 103
Reflective cracking is one of the major early distresses commonly found in newly
constructed semi-rigid pavements. Research on how to improve a HMA mix design to
control such reflective cracking problems is urgently needed. Results from a finite
element simulation (Ni et al. 2006) indicate that, when the CTA base course is under
intact condition (no cracks), the overall stress state at the bottom of the binder course
is in compression. However, once the CTA base course develops cracks primarily due
to shrinkage, significantly high tensile stresses and strains will be generated in the
vicinity of the base cracks at the bottom of the HMA binder course. Under such a
tension state, repeated traffic loads will first cause fatigue cracking at the bottom of
the binder course. Fatigue cracks will then gradually propagate to the pavement
surface, and finally result in failure for a surface HMA layer due to the reflective
cracking.
Therefore, to control the CTA cracks reflected to the surface, a HMA binder course
mixture is desired to be not only anti-reflective cracking but also fatigue cracking
resistant. This study was in an effort to control the reflective cracking problem on
semi-rigid pavements through a proper binder course mixture design and laboratory
characterization of the fatigue and anti-cracking properties of binder course mixtures
evaluated.
The main objective of this study was to evaluate the fatigue and anti-cracking
properties of the binder course mixtures using different asphalt binders and aggregate
gradations. This knowledge will eventually be used to develop a proper binder course
mixture design to improve the reflective cracking problem commonly found in semi-
rigid pavements in China.
To achieve the objective, three asphalt binder types and four aggregate gradations
were selected in this study. HMA mixture characterization was accomplished by
conducting three cracking-related laboratory tests — reflective cracking simulation
test, impact ductility test, and beam fatigue test.
MIXTURE DESIGN
Asphalt Binders
Binder Type Al A2 A3
Test Results Specification Results Specification Results Specification
Penetration at 40-60
51 63.8 60-80 92 80-100
25°C
Flash point, °C 325 260+ 295 260+ 290 260+
Solubility, % 99.88 99.5 99.89 99.5 - -
Viscosity, 60°C, -
- - 222.4 180 272
Pa.s
Ductility at 15°C, 80+
>150 >150 100+ - -
cm
Ductility at 10°C,
26 15+ 50 20+ - -
cm
Ductility at 5°C, - 18+
- - - 21.2
cm
Ductility at 15°C,
after thin-film 47 10+ 94.2 15+ - -
oven test, cm
Ductility at 10°C,
after thin-film 12 4+ 19 6+ - -
oven test, cm
Ductility at 5°C,
after thin-film - - - - 16.8 10+
oven test, cm
Aggregate Gradations
Four aggregate gradations with different nominal maximum aggregate size (NMAS)
were included in this study. Table 2 presents the percent passing of each gradation.
Typical crushed limestone aggregates, currently used in a highway construction
project, were selected for this study, which included crushed coarse and fine
limestone aggregates as well as limestone fines. The name of each gradation in Table
2 is associated with its NMAS. For example, AG-25 has a NMAS of 25 mm. It is
noted that only gradation AG-19 was fine-graded (the gradation curve is above the
Superpave restricted zone), the remaining gradations passed through the restricted
zone.
TABLE 2. Aggregate Gradations
Gradation Type
Sieve Size(mm)
AG-25 AG-19 AG-16 AG-13
31.5 100 100 100 100
25.0 97 100 100 100
19.0 79 98 100 100
16 70 88 95 100
13.2 63 79 84 95
9.5 53 67 70 76.5
4.75 41 53 48 53
2.36 31 40 34 37
1.18 23 29 24.5 26.5
0.6 17 21 17.5 19
0.3 12 14 12.5 13.5
0.15 9 7.5 9.5 10
0.075 5 5.7 6 6
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 105
Mix Design
Marshall mix design was used to determine the optimum asphalt content. The
optimum asphalt cement content for all mix designs was determined from asphalt
concrete mixtures compacted with the Marshall Hammer at 75 blows per face and an
air voids between 3 - 5 percent. Table 3 presents the Marshall mix design results. It is
noted that, in terms of the Stability, mix containing AG-19 gradation was the lowest;
in terms of the Flow, mix with AG-25 was the highest.
Mixtures
% Stability
Grouped by %AC Gmb VTM(%) Flow (O.lmm)
Saturation (kN)
Gradation
AG-25 3.6 2.452 4.2 65.6 12.7 41.2
AG-19 4.0 2.453 3.8 69.9 10.5 38.1
AG-16 4.2 2.453 3.5 72.5 13.6 31.6
AG-13 4.5 2.462 3.2 79.3 14.1 34.6
As stated earlier, three cracking related tests were conducted in this study. Testing
temperature for the reflective cracking simulation test and impact ductility test was
set to be 20 °C, while the beam fatigue test was conducted at 15 °C. The testing
temperature of 15 °C for the beam fatigue test was based on the current China asphalt
pavement construction specification (RIHMC 2005), where a fatigue resistance
parameter (i.e. the allowable tensile stress at the bottom of a HMA layer) is required
to be tested at this temperature. Table 4 presents the test factorial of this study.
Rectangle beam samples were used in all three cracking tests. The beam sample
sizes for the reflective cracking simulation, impact ductility test and beam fatigue test
were 240-mm (long) x 70-mm (wide) x 50-mm (high), 250-mm x 35-mm x 35-mm,
and 381-mm x 63.5-mm x 50-mm, respectively. All test specimens were compacted
106 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
using a kneading compactor. In general, a larger slab was first compacted at a target
air void of 5.0 ±1.0 percent (similar to the initial roadway air voids found for the
binder course layer) and then cut into the proper test sample size using a diamond saw
machine.
TEST DESCRIPTION
Note that performing a RCS test is not only relatively complicated in sample
preparation, but also time-consuming. It may be not suitable for the routine lab use.
Several researches (Huang et al. 1994; Baburamani and Porter 1996; Van Dijk 1975)
indicated that, when a vehicle load passes through a pavement surface on top of a
base crack, due to the high vehicle speed (small loading time), such wheel loading is
similar to an impact load to asphalt materials. Based on this observation, an impact
load based test was also selected for characterizing the anti-reflective cracking
property for the binder course mixtures in this study.
The test setup of Impact Ductility test is showed in Figure 2. The Impact Ductility
test was a displacement-controlled loading test, designed to perform on a MTS-810
loading system with a three-point loading setup. During a test, the beam specimen
was loaded monotonically till fracture failure under a constant cross-head
deformation rate of 500 mm/min at a test temperature of 20 °C. The load and vertical
deformation were continuously recorded and a load-vertical displacement curve was
plotted. An impact ductility value is defined as the area under the loading portion in a
load-vertical deflection curve, up to the maximum applied load. According to the
literature (Huang et al. 1994; Baburamani and Porter 1996), the impact ductility value
represents the total stored energy in a HMA mixture before its fracture failure. The
higher an impact ductility value, the better a HMA mixture resists to the reflective
cracking.
The Beam Fatigue test was conducted according to SHRP M009 protocol at 15°C. It
was a strain-controlled type of test where a beam 318 mm (15 in) long by 63.5 mm
(2.5 in) wide by 50.8 mm (2.0 in) height is subjected to 4-point bending. Different
strain levels are required in the beam fatigue test to generate a range of load
repetitions at failure. The center deflection of the beam was continuously measured
and used in the computation of the stiffness. Failure is defined as the load cycle at
which the specimen exhibits a 50 percent reduction in stiffness. Figure 3 presents the
108 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
test setup for the Beam Fatigue test. In this study, this test was conducted using a
UTM machine with a temperature control chamber.
DISCUSSION OF RESULTS
As stated earlier, three parameters can be obtained from a RCS test result: the
average cracking rate, initial cracking number, and final cracking number. The initial
cracking number is defined as the number of repetitions when the beam starts to
crack. The final cracking number is the number of load repetitions when the beam
fails. The average cracking rate is the ratio between the average crack length
measured from two side walls of a beam sample and the difference of the final and
the initial cracking numbers.
Figure 4 shows reflective cracking development photos for HMA mixtures evaluated
under a RCS test. Note that side walls of asphalt beams were painted into a white
color in order to better trace the development of cracking. As shown in Figure 4,
under the repeated compressive load, all cracks propagated in a zigzag manner,
moving across those weak links inside an asphalt-aggregate structure. In addition, the
crack initiation point does not necessarily start in the middle of a base crack.
(a) Mix BMA1AG.19at 164,955 cycles (b) Mix BMA2AG.,3 at 397,520 cycles
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 109
(c) Mix BMA2AG.19at 149,842 cycles (d) Mix BMA2AG-i6at 211,835 cycles
(e) Mix BMA3AG.19 at 314,814 cycles (f) Mix BMA2AG-25at 184,960 cycles
Figure 5 presents the average RCS results grouped for HMA mixtures using
different binder types but with a same AG-19 gradation. As shown in Figure 5, as the
binder elasticity grade increases, the average cracking rate decreases, and the final
cracking number increases. This indicates that using a high elasticity grade binder in a
HMA mixture generally can delay the reflective cracking propagation rate
significantly.
Figure 6 presents the average RCS results grouped for HMA mixtures in different
gradation groups but with a same A2 binder. Figure 6 illustrates that as the NMAS
increases, the average cracking rate increases, and the initial cracking number
decreases with an exception for the mixture containing the AG-25 gradation. As
shown in Figure 6, even though the mixture with AG-25 gradation had the highest
average crack propagation rate, it also had the highest initial cracking number among
the four mixtures evaluated. This implies that with a large NMAS of 25-mm, a HMA
mixture (e.g. Mix BMA2AG-25 in Figure 4(f)) tends to form a better interlock in the
aggregate structure to resist an initial cracking. On the other hand, due to its low
asphalt content (Table 3), it lacks of binder cohesion. As a result, when an initial
crack is developed, it will propagate at a rapid rate. Overall, by considering the
effects from both the binder content and aggregate structure, a smaller NMAS HMA
mixture tends to have a higher anti-reflective cracking property than a larger NMAS
HMA mixture.
110 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
(a) (b)
FIG. 5. Average RCS results for Mixtures with Different Asphalt Binders
(a) (b)
FIG. 6. Average RCS results for Mixtures with Different Aggregate Gradations
Figure 7 presents the impact ductility results for HMA mixtures evaluated in this
study. A high impact ductility value is desired for a crack-resistant mixture. As the
NMAS increases, the impact ductility value decreases, Figure 7(a). Similarly, as the
binder elasticity grade increases, the impact ductility of a mixture tends to increase,
Figure 7(b). Overall, the impact ductility test results confirmed the results obtained
from the RCS tests as described above.
Figure 8 presents the relationship between the impact ductility and the average
cracking rate of six kinds of mixtures. A fairly good polynomial relationship was
observed. This observation indicates that, in general, the impact ductility test can be
used as a good alternative test to characterize the anti-reflective cracking property for
a HMA mixture.
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 111
(a) (b)
Figures 9(a) and (b) present the beam fatigue test results for mixtures containing
various binders and different aggregate gradations, respectively. As shown in Figure
9(a), the mixtures containing the highest elasticity grade, polymer-modified binder
(A3) displayed a significantly higher fatigue life than mixtures with other two non-
modified, lower elasticity grade binders. This indicates that the polymer modified
binder used in this study can significantly improve a mixture's fatigue cracking
resistance. On the other hand, the fatigue curves for mixtures containing Al and A2
binders were found to cross-link with each other. At a lower tensile strain (smaller
than 400 micro-strains), mixtures with Al binder had a slightly higher fatigue life;
whereas, at a higher tensile strain (higher than 400 micro-strains), the fatigue life for
mixtures containing A2 binder were higher. One possible reason for this
inconsistency may be attributed to sample preparation and handling. However, as
shown in Figure 9(a), the fatigue resistance for mixtures containing Al and A2
112 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
binders were not significantly different from each other under a normal traffic
loading.
For mixtures containing a same binder type but with different NMAS gradations,
those with a smaller NMAS tended to have a higher fatigue resistance. As shown in
Figure 9(b), except the mixture with AG-19, as the NMAS decreases, the fatigue
resistance of a mixture increases. The reason why the mixture with AG-19 appeared
to have a lower fatigue resistance than mixtures with AG-25 will be further discussed
in the following sections. But it should be noted here that AG-19 was the only fine-
graded gradation used in this study.
(a)
(b)
FIG. 9. Beam Fatigue Test Results (a) Mixtures with Different Binders; (b)
Mixtures with various Aggregate Gradations
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 113
The reason for high elasticity grade binder performed better than low elasticity grade
binder in both fatigue and reflective cracking resistance may be explained by the
binder's fatigue factor, G*sin(5), where G* is dynamic shear modulus of asphalt
binder and 8 is phase angle. A dynamic shear rheometer, DSR, was used to
investigate the rheological behavior of the PAV-aged binders. The DSR tests were
performed at 25 °C. The G*sin(6) values for Al, A2, and A3 binders were found to be
9,663-, 7,362-, and 2,005-kPa, respectively. According to Superpave specification,
G*sin(6) for a certain binder at its fatigue temperature should be less than 5,000-kPa.
Note that only A3 had a G*sin(5) value lower than 5,000-kPa. This illustrates that the
polymer-modified binder, A3, had significantly better fatigue cracking resistance than
the two non-modified binders (Al and A2). In addition, it is interesting to note that
the G*sin(8) values had a fairly good relationship to both RCS and impact ductility
test results, which are shown in Figure 10. As the G*sin(5) value increases, the
average cracking rate increases and the impact ductility value decreases. In summary,
binder DSR test results indicate that G*sin(5) obtained at a fatigue temperature is
indicative to both fatigue resistance and anti-cracking property for a HMA mixture.
(a) (b)
FIG. 10. (a) Relationship between binder G*sin(6) values and RCS Test
Results; (b) Relationship between binder G*sin(5) values and Impact
Ductility Test Results
and
where,
114 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
=
PCA or PFA percent by weight passing a given sieve have opening of
width d
=
acA intercept constant for the coarse aggregate
apA = intercept constant for the fine aggregate
d = sieve opening width, mm
=
ncA slope (exponent) for the coarse aggregate
=
npA slope (exponent) for the fine aggregate
In this study, the sieve size of 2.36 mm was selected as a divider for the CA and the
FA portions in the regression analysis. In general, both intercepts of acA and apA
represent the total percent passing on a sieve size of 1mm. The slopes of ncA and n?A
are indicative of the shapes of a gradation curve and can be either greater or smaller
than 0.45. The greater an ncA value, the coarser the CA portion gradation (i.e. more
distances below the 0.45 maximum density line). On the other hand, the higher an n?A
value, the finer the FA portion gradation (i.e. more distances above the 0.45
maximum density line) (Mohammad et al. 2004).
Table 5 presents the gradation analysis results for the four gradations selected in this
study. As shown in Table 5, all regression analyses received a high R2-value,
indicating that Eq.l fits those gradation curves very well. In general, the acA and a?A
results (% passing on 1mm sieve size) agreed well with those percentage values
passing on 1.18mm sieve size listed in Table 2, both illustrating that the AG-19
gradation had the highest fine aggregate percentage among the four gradations
evaluated. It is interesting to note that the HCA values in Table 5 provide an exactly
same ranking as those fatigue test results shown in Figure 9(b). A lower ncA value
was found to be associated with a lower fatigue resistance for a mixture. This
observation implies that a coarser CA portion (i.e. higher ncA value) is desired for a
fatigue resistance mixture. On the other hand, Mohammad et al. (2004) reported that
npA values were found to have good correlation with the rut-resistance of a HMA
mixture. Although rutting was not the topic of this research, however, it may be
pointed out here that laboratory rut resistance test did observe the lowest average rut
depth for mixtures containing the AG-19 gradation (with the highest nFA in Table 5),
followed by mixtures with the AG-25 gradation (with the second highest n?A in Table
5) (Ni et al. 2006).
TABLE 5. Results of Power Law Regression Analyses for four Gradation Types
In summary, the gradation analysis of this study yields two main observations: (1) to
obtain a fatigue resistant mixture, the CA portion of a gradation should be the coarser
the better (i.e. high ncA values); (2) to obtain a rut resistant mixture, the FA portion of
a gradation should be the coarser the better (i.e. high nFA values). Therefore, how to
balance both ncA and npA values in a gradation design is the key to obtain a HMA
mixture with both rutting and fatigue cracking resistance.
CONCLUSIONS
REFERENCES
INTRODUCTION
Elastic modulus is an important property of pavement materials. Different
methods have been proposed by researchers to estimate the elastic modulus based
on laboratory bending tests and empirical equations (Bonnaure et a/., 1997), wave
propagation methods (Cho and Lin, 2001), and the Falling Weight Deflectometer
(FWD).
117
118 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
where, RMSE is the relative root-mean-square error, n is the total number of the
deflection measurement points, df is the backcalculated deflection at point i, and
Di is the measured deflection at point /. When the RMSE value decreases, the
accuracy of the backcalculated elasticity moduli is assumed to increase as the
error between the measured and calculated deflections decreases.
In the LTPP test sections, a RMSE of 3% was used as an acceptable error
(Von Quinrus and Simpson, 2002). In addition, Von Quinrus and Simpson (2002)
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 119
BACKCALCULATION STUDY
The majority of the available research investigates the difference between
the RMSE values from different backcalculation programs (Fwa et al 1997), the
effect of other factors on the quality of the deflection data (Mehta and Roque
2003), the effect of the seed generation on the RMSE values (Fwa and Rani
2005), and the effect of other factors on the FWD data. To the best of the authors'
knowledge, however, the effect of the backcalculated elastic moduli and the
associated RMSE on the strain and stress responses of flexible pavements has not
been discussed so far.
In this paper, to study the effect of the RMSE on the strains and stresses in
flexible pavements, a three-layer pavement section was selected. The flexible
pavement section and the backcalculated elastic moduli were reported by
Anderson (1988) and were shown in Tables 1 and 2. Responses at a total of 11
points (as shown in Table 4) were calculated for each case: responses along the
ground surface at common locations of the velocity sensors used in the FWD test
along the pavement profile (points 1 through 7), at the middle of the AC layer
(point 8), at the bottom of the AC layer (point 9), at the middle of the base layer
(point 10), and at the top of the subgrade (point 11). For all cases, responses were
calculated for a circular load with a radius of 150 mm and a pressure of 690 kPa.
The Poisson's ratio for all layers was equal to 0.35.
The pavement response was calculated using the MultiSmartSD program.
The MultiSmartSD program is a fast and accurate software tool developed by the
Computer Modeling and Simulation Group at the University of Akron, and it is
based on the innovative computational and mathematical techniques for
multilayered elastic systems (Pan 1989a,b; Pan 1990; Pan 1997). The program is
capable of analyzing any pavement system regardless of the number of layers, the
thickness of each layer, the number of response points, and the shape of the
applied pressure at the surface of the pavement.
The RMSE for each case was calculated using Eq. 1 and shown in Table
3. In addition, Table 3 shows the relative errors using the exact and
backcalculated elastic moduli. The relative error is defined as:
Exact - Calculated
RE = xlOO% (2)
Exact
120 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
This definition of the relative error is also used to compare the calculated strains
and stresses based on the exact and backcalculated models. In other words,
Calculated is the response using the backcalculated elastic moduli, and Exact is
the response using the exact elastic moduli.
Pavement responses from the exact and backcalculated elastic moduli
(Cases 1 through 5) are shown in Tables 5 through 10. In addition, Tables 5
through 10 show the relative error associated with each response.
TABLE 10. Response (Top Row) and Relative Error (Bottom Row in
Parenthesis) Using Case 5 Backcalculated La]per Model
Point £k €y &i 0k Oy <*L uz
(/mi/m) (fjanJm) 0«n/m) (kPa) (kPa) (kPa) (mm)
0) (2) (4) (5) (6) (7) (8)
(3)
1 -517? -55,77 -57 JO -685.30 -685.30 -690.00 0.21
(3.472) (3.472) (?<988) (1.085) (1.085) (0.000) (1.367)
2
-733 -26.58 18.21 -68.92 42133 0.00 0.17
(3321) (1.857) (0.793) (6,448) (4570) (0.000) (0.553)
~7J9 -17J3 13*91 -59.33 -85,99 0.00 0.16
3
(5J84) (2,c#i)(3.102) (1.936) $.394) (0.000) (0.588)
4
-3,05 43.70 9M -32.70 •4135 0.00 0.14
(8.05?) (2479) 0.705) (0.946) (2.807) (0.000) (0.677)
5 «J7 40,61 5,68 -15.20 -44J2 0.00 0.13
O»4W) 1X996) (3,964) (0.007) C&545) (0.000) (0.755)
6
1.76 •4J9 331 -4,75 -31,97 0.00 0.12
(1.183) (3.202) C4J21) (3.761) (2331) (0.000) (0.832)
7
2.67 -633 2,08 1.6<) -23,32 0.00 0.11
(1.168) (3363) (4J25) (21.260) (2,138) (0.000) (0.906)
21.22 21.22 -82,82 -70.10 -70.10 -351.94 0.19
8
(5J81) (5,981) (6,270) (1.401) (1.401) (0.685) (0.820)
9 335 335 -21.27 -36.94 -36.94 -10343 0.18
(109.897) (109.897) (0.299) (26.065) ((26.0065) (4495) (0.554)
11.63 1143 -14.49 194.73 194.73 -37,81 0.18
10
(12.7X8 CJM72) (12491) {1.320) (1.320) (5352) (0.636)
11 22,64 22.64 -86.94 -134 434 -7.00 0.18
(5,667) (5467) (1.614) (7375) (7375) (1.180) (0.700)
In Tables 5 through 10, values without parentheses are the pavement
responses using the backcalculated moduli in Table 2 (Cases 1 to 5) while values
in parentheses are the relative errors compared to responses using the original
(exact) pavement profile. Cells that are highlighted with gray show the response
with a relative error higher than 2% whilst values in bold show the response with
a relative error higher than 10%.
The results in Tables 3, and 5 through 9 show that even if the RMSE
value is kept less than 1%, the resulting response of strains and stresses can
largely differ than the exact response. Cases 1 and 2 (Tables 6 and 7,
respectively) have relatively the same RMSE in deflection but have different
relative errors in strains and stresses. In addition, the magnitude and location of
the relative errors vary randomly between Cases 1 and 2. Case 1 showed a relative
error of 19.686% in the horizontal strain (s*, or £y) at the bottom of the AC layer
(point 9) while Case 2 showed a higher relative error of 81.365% in the horizontal
strain at the same point. On the other hand, at the ground surface (point 5), Case 2
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 127
where Ne and NO- are the number of load repetitions to fatigue cracking using the
constant strain and constant stress analysis, respectively, Af and K are material
constants, $ is the tensile strain at the critical location and Es is the stiffness of the
material (i.e. elastic modulus). The constant strain model is applicable to thin AC
layers usually less than 51 mm, whilst the constant stress model is applicable to
thick AC layers usually more than 203 mm. The Shell Model was calibrated and
generalized for any thickness as given below (MEPDG, 2004):
where Nf is the number of load repetitions to fatigue cracking, and F"is a constant
that depends on the layer thickness and the material stiffness.
backcalculated set of elastic moduli and that using the exact set of elastic moduli.
In other words, the ratio is equal to A//& (backcalculated set of elastic moduli) over
Nfe (exact set of elastic moduli).
RUTTING DAMAGE
Rutting in flexible pavement is considered as a functional deterioration.
Rutting is mainly predicted by calculating the vertical strains at the top of the
subgrade and then estimating the allowable load repetitions until a certain rutting
threshold is met. For example, Shook et al. (1982) assumed a rutting depth of 10
mm in their method, while Potter and Donald (1985) assumed 20-30 mm rutting
depth.
Recently, the results from the test sections at MnROAD were used to
develop a method to predict the number of allowable load repetitions until rutting
failure using a rut depth of 13 mm as shown in the following relation (Skok et al.,
2003):
where Nr is the number of allowable load repetitions until rutting failure, and ^ is
the maximum compressive strain at the top of the subgrade layer.
It can be seen, from the above equation, that the vertical strain at the top of
the subgrade layer is very important to predict the lifetime of the pavement due to
rutting. Similar to the fatigue case, the effect of the backcalculated elastic moduli
on the rutting can be studied by finding the ratio between the estimated number of
repeated loads (Nr) using the backcalculated set of elastic moduli and that using
the exact set of elastic moduli. In other words, the ratio is equal to Nrb
(backcalculated set of elastic moduli) over Nre (exact set of elastic moduli).
The effect of the RMSE value and the relative error in pavement moduli
on rutting can be observed from Table 12. It can be seen that even for a very
small RMSE (Cases 1 and 2 where the RMSE is 0.22% and 0.24%, respectively),
the rutting failure prediction based on the backcalculated moduli was
overestimated by 11.7% in Case 2 while it was overestimated by 0.3% in Case 1,
indicating a high sensitivity of the rutting life on the relative error in moduli. In
addition, the results show that as the RMSE (error) increases, the underestimation
of the rutting increases, as can be observed by comparing the rutting results in
Table 12 for Cases 2 to 5.
CONCLUSIONS
This study shows that the use of the RMSE is not enough to secure an
accurate backcalculation of the pavement elastic moduli. Large discrepancies can
exist in the predicted pavement strains and stresses using the backcalculated and
exact elastic moduli. As a result, even a RMSE value less than 1% can
significantly affects the fatigue and rutting predictions in flexible pavements.
The effect of the RMSE is suitable for controlling the fitness of the
backcalculated deflection basin to that measured in the field while the use of the
relative error in the elastic moduli is more appropriate. However, the availability
of laboratory measured elastic moduli does not guarantee more appropriate
backcalculated elastic moduli. This variation adds more uncertainty when dealing
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 131
with data from the FWD test, and should be the future endeavor in pavement
engineering.
ACKNOWLEDGEMENT
We are grateful for the support by ODOT/FHWA under grant ODOT
20943.
REFERENCES
Anderson, M. (1988). "Backcalculation of composite pavement layer moduli."
PhD dissertation, The University of Kentucky, Kentucky.
Bonnaure, F., Gest, G., Gravois, A. and Uge, P. (1977). "A new method of
predicting the stiffness of asphalt paving mixtures." Proceedings of the
Association of Asphalt Paving Technologists, 46, 64—100.
Bonnaure, F., Gravois, A., and Udron, J. (1980). "A new method of predicting the
fatigue life of bituminous mixes." Journal of the Association of Asphalt Paving
Technologist, 49, 499-529.
Cho, Y.S., and Lin, F.B. (2001). "Spectral analysis of surface wave response of
multi-layer thin cement mortar slab structures with finite thickness." NDT and
E International, 34, 115-122.
Daleiden, J.F., Killingsworth, B.M., Simpson, A.L., and Zamora, R.A. (1994).
"Analysis of procedures for establishing in situ subgrade moduli."
Transportation Research Record 1462, TUB, National Research Council,
Washington, D.C., 102-107.
Fwa, T.F., and Rani, T.S. (2005). "Seed modulus generation algorithm for
backcalculation of flexible pavement moduli." Transportation Research
Record 1905, TRB, National Research Council, Washington, D.C., 117-127.
Fwa, T.F., C. Y. Tan, and Chan, W. T. (1997). "Backcalculation analysis of
pavement-layer moduli using genetic algorithms." Transportation Research
Record 1905, TRB, National Research Council, Washington, D.C., 134-142.
Harichandran, R. S., Ramon, C. M., and Baladi, G. Y. (1994). "MICHBACK
user's manual." Department of Civil and Environmental Engineering, Michigan
State University, East Lansing, Michigan.
Mehta, Y., and Roque, R. (2003). "Evaluation of FWD data for determination of
layer moduli of pavements." Journal of Materials in Civil Engineering ,ASCE,
15(1), 25-31.
MEPDG, National Research Council (2004). "Guide for mechanistic-empirical
design (MEPDG)." National Cooperative Highway Research Program
(NCHRP).
Pan, E. (1989a). "Static response of a transversely isotropic and layered
half-space to general dislocation sources." Phys. Earth Planet. Inter., 58,
103-117..
Pan, E. (1989b). "Response of a transversely isotropic and layered half-space to
general surface loads." Phys. Earth Planet. Inter., 54, 353-363.
132 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
ABSTACT: Flexible pavements are commonly used for low to high volume roads
subjected to many wheel load applications and also for airfields subjected to rather
heavy aircraft gear/wheel loads. As the demand for applied wheel loads and number
of load applications increase, it becomes very important to properly characterize the
behavior of unbound aggregate layers and subgrade soils as the pavement foundation
geomaterials. Laboratory studies have shown that resilient responses for these
geomaterials follow nonlinear, stress-dependent behavior under repeated loading.
Therefore, a finite element (FE) type layered elastic analysis is needed to employ
nonlinear resilient material models to predict accurate pavement responses for
mechanistic based pavement design. In this study, modulus models well proven over
the years to adequately describe the nonlinear pavement geomaterial behavior were
programmed in a user material subroutine (UMAT) to perform axisymmetric and
three-dimensional (3D) analyses using the general-purpose ABAQUS FE program.
The results indicated that modulus characterizations of the nonlinear, stress-
dependent base and subgrade layers were essentially needed to reliably predict
accurate pavement responses both in axisymmetric and 3D analyses.
Introduction
1
Associate Professor, PH (217) 333-8637; email: [email protected]
2
Graduate Research Assistant, PH (217) 333-6973; email: [email protected]
Department of Civil and Environmental Engineering, University of Illinois, 205 N. Mathews,
Urbana, IL 61801, USA; FAX (217) 333-1924.
133
134 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
where bulk stress 9 = a\ + a2 + <J3 = a\ + 2a3 (when cr2 = #3), Po is a unit pressure (1
kPa or psi), and KI, K2, and KS are model parameters obtained from triaxial test data.
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 135
For both the nonlinear axisymmetric and 3D FE analyses performed in this study,
stress states applied in the laboratory triaxial conditions (<72 = #3) were essentially
used to determine model parameters.
For fine-grained subgrade soils, the resilient modulus is also dependent upon
the applied stress state. Typically, soil moduli follow a stress-softening type behavior
and the bilinear model proposed by Thompson and Robnett (1979) is one of the most
commonly used MR models to simply capture this behavior. The bilinear MR model
which is primarily a function of the deviator stress is given as follows:
where KI, K2, KS, and IQ are model parameters obtained from triaxial test data.
Representing the bilinear behavior, parameter KI is referred to as the breakpoint
modulus ERI, which is a characteristic soil property.
The load was applied as a uniform pressure of 0.55 MPa over a circular area
of 152 mm radius. Figure l(a) shows the axisymmetric FE mesh for the selected
domain size of 20-times R in the horizontal direction and 140-times R in the vertical
direction. Using this mesh, the predicted pavement responses were in very good
agreement with those of the KENLAYER closed form solutions as listed in Table 2.
Table 2. Predicted Critical Pavement Responses for the Analysis Domain Study
Linear Elastic Analysis
ABAQUS with
Pavement response KENLAYER 8-noded quadrilateral
elements
6 SUIface (mm) -0.927 -0.930
<*r bottom of AC (MPa) 0.777 0.773
dv too of subgrade (MPa) -0.041 -0.041
Svtopofsubgrade(l^e) -936 -933
For the 3D FE model, the same pavement geometry and the input properties
listed in Table 1 were again considered this time using the 20-noded solid element
instead of 8-noded quadrilateral elements. The developed 3D FE mesh is shown in
Figure l(b) for the same domain size of 20-times R in the horizontal direction and
140-times R in the vertical direction. The boundary conditions at the vertical
boundary and symmetry planes were assigned zero horizontal displacements together
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 137
with a restraint condition of zero horizontal and vertical displacements on the bottom
plane. A fine FE mesh was generated in the vicinity of wheel loading to apply an
accurate representation of the circular shape loading and capture the steep stress and
strain gradients.
Before performing nonlinear FE analysis, the linear elastic solutions were
compared from the ABAQUS axisymmetric and 3D models to evaluate differences
between the two pavement models. Table 3 lists predicted critical pavement
responses from the two analyses. Overall, the differences were negligible and the
largest difference was not more than 3% for the surface deflection. The other
responses, such as tensile stresses at bottom of asphalt concrete (AC) and vertical
stresses on top of subgrade, had differences less than 1%. Generally, these
comparisons between the axisymmetric and the 3D analyses are quite acceptable
especially when considering the assumptions for the axisymmetric FE formulations
and the circular shaped mesh discretization representing the wheel loading. Therefore,
the developed 3D pavement FE model was considered accurate enough to study next
the nonlinear pavement foundation considerations.
ABAQUS, ANSYS, and ADINA are the most commonly used general-
purpose FE programs which can provide proper analyses for various engineering
problems. Several researchers have already performed linear and/or nonlinear
pavement analyses using these general-purpose FE programs. Chen et al. (1995) and
Cho et al. (1996) conducted a comprehensive study of various FE pavement analyses.
The results from ABAQUS were comparable to those from other pavement analysis
programs, such as ILLI-PAVE (Raad and Figueroa 1980), GT-PAVE (Tutumluer
1995), and KENLAYER (Huang 2004), and the results yielded the lowest tensile
strains for the linear analysis case. Three types of ABAQUS FE models, i.e., plane
strain, axisymmetric, and 3D, were evaluated to facilitate the selection of appropriate
modeling and corresponding element types for simulating traffic loading effects.
They concluded that the axisymmetric and 3D FE models yielded different results for
pavement responses.
Taciroglu (1998) solved for the pavement responses using 3D ABAQUS
analysis and adopted the K-0 model (Hicks and Monismith 1971) and the Uzan (1985)
model as the nonlinear unbound granular material MR models. The ABAQUS user
material (UMAT) subroutine was formulated to incorporate mainly strain dependent
138 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
modulus models for the unbound aggregates. The nonlinear resilient behavior of
granular materials were adequately modeled and but the solutions often predicted
high asphalt bending stresses, and substantially different granular base stresses and
strains were obtained from the nonlinear model.
The ABAQUS FE program was also used for pavement analysis by Schwartz
(2002) who employed the K-0 model (Hicks and Monismith 1971) in the base course
by using this time the "*HYPOELASTIC material model" inputs in the ABAQUS 3D
modeling framework. The secant resilient modulus values could not be directly used
in nonlinear solutions but were numerically converted to tangent moduli for input as a
function of the first stress invariant, Ij. A tension cut-off was also imposed by
specifying a very small modulus for tensile Ii values. Comparing the linear and
nonlinear solutions, Schwartz (2002) reported that there were up to 25% and 20%
differences between the maximum asphalt tensile stresses and strains, respectively,
for the most extreme case and less than 5% differences of stresses and strains at the
top of the subgrade and surface deflection. He also noted that these differences
would seem acceptable for practical design.
Recently, Saad et al. (2005) examined the dynamic responses of flexible
pavement structures to single wheel traffic loads using the 3D FE analysis program
ADINA. Their study focused on elastoplastic analyses different than the pavement
resilient behavior. The base material was dealt with elastoplastic Drucker-Prager
model as a strong or weak base and the subgrade was simulated by the modified
CamClay model. Elastoplasticity of the base material caused an increase of 46% in
the rutting strain, 28% in the maximum tensile fatigue strain on the bottom of the
asphalt layer, and 30% in the maximum surface deflection. The subgrade nonlinearity
had little impact on the fatigue strain which was less than 1%.
where a{j is the stress tensor, v is Poisson's ratio, E is the Young's modulus, e^ is
the volumetric strain, 8{j is Kronecker delta, and etj is the strain tensor.
Generally, when constitutive properties were updated at the integration points
within each element, the element stiffness matrix, Ke, was computed using the
constitutive relation matrix D.
and new values of the resilient moduli usually controlled the convergence of
nonlinear solutions. The developed UMAX material model subroutine was validated
with the GT-PAVE axisymmetric FE program used specifically for pavement
analysis (Kim and Tutumluer, 2006).
Table 5 lists the critical pavement responses predicted by both linear and
nonlinear axisymmetric analyses using the ABAQUS FE program. Figure 2 shows
the vertical displacements predicted at the centerline of loading in the AC, base, and
subgrade layers as obtained from the three different pavement layer material
characterization cases. The nonlinear characterization of the base layer mainly caused
increases in the predicted responses of up to 18% in the tensile strain at the bottom of
the AC, 29% in the vertical strain on top of the subgrade, and 33% in the surface
deflection. The nonlinearity of subgrade also affected the critical pavement responses;
up to 21% decrease in the vertical strain on top of the subgrade and 23% decrease in
the surface deflection. On the other hand, the nonlinearity of subgrade soils had little
impact on the tensile strain at the bottom of the AC. As listed in Table 5, the
combined nonlinear base and subgrade characterizations yielded the most accurate
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 141
pavement response predictions quite different from the linear elastic case. Note that
these differences, specific to the pavement geometries, layer material properties, and
the loading conditions studied here, were given as examples to demonstrate the
important effects of nonlinear pavement foundation behavior.
Table 6 lists the critical pavement responses predicted by both linear and
nonlinear 3D analyses using the ABAQUS FE program with more responses
presented here than listed in Table 5. The nonlinear FE analyses resulted in
considerably different pavement response predictions when compared to those from
the linear analyses. The largest difference was for the case of the nonlinear base and
linear subgrade material. The nonlinear characterization of the base layer caused
increases in the predicted responses of up to 29% in the tensile strain at the bottom of
the AC, 30% in the vertical strain on top of the subgrade, and 36% in the surface
deflection. The subgrade nonlinear characteristics resulted in 21% decrease in the
vertical strain and 22% decrease in the surface deflection. Similar to the results of the
axisymmetric FE analyses, the nonlinearity of subgrade soils had little impact on the
tensile strain at bottom of the AC.
References
Brown, S.F. and Pappin, J.W. (1981). "Analysis of Pavements with Granular Bases."
Transportation Research Record 810, TRB, National Research Council,
Washington, D.C., 17-23.
Chen, D.H., Zaman, M., Laguros, J., and Soltani, A. (1995). "Assessment of
Computer Programs for Analysis of Flexible Pavement Structures."
Transportation Research Record 14829 TRB, National Research Council,
Washington, D.C., 123-133.
Cho, Y., McCullough, B.F., and Weissmann, J. (1996). "Considerations on Finite-
element Method Application in Pavement Structural Analysis."
Transportation Research Record 1539, TRB, National Research Council,
Washington, D.C., 96-101.
Duncan, J.M., Monismith, C.L., and Wilson, E.L. (1968). "Finite Element Analyses
of Pavements." Highway Research Record 228, TRB, National Research
Council, Washington, D.C., 18-33.
Hibbit, Karlsson & Sorensen, Inc. (2005). ABAQUS/'standard User's Manual
Version 6.5. Hibbit, Karlsson & Sorensen, Inc., Pawtucket, Rhode Island.
144 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
Hicks, R.G. and Monismith, C.L. (1971). "Factors Influencing the Resilient
Properties of Granular Materials." Transportation Research Record 345, TRB,
National Research Council, Washington, D.C., 15-31.
Huang, Y.H. (2004). Pavement Analysis and Design, 2nd Edition, Pearson Prentice
Hall, Pearson Education Inc., Upper Saddle River, New Jersey
Kim, M. and Tutumluer, E. (2006). "Modeling Nonlinear, Stress-Dependent
Pavement Foundation Behavior Using A General-Purpose Finite Element
Program." In ASCE Geotechnical Special Publication 154, entitled, Pavement
Mechanics and Testing, Edited by B. Huang, R. Meier, J. Prozzi, and E.
Tutumluer, Proceedings Book of the ASCE Geo-Institute Geoshanghai
Conference, Shanghai, P.R. of China.
Raad, L., and Figueroa, J.L. (1980). "Load Response of Transportation Support
Systems." Journal of Transportation Engineering, ASCE, 106(1), 111-128.
Saad, B., Mitri, H., and Poorooshasb, H. (2005). "Three-dimensional Dynamic
Analysis of Flexible Conventional Pavement Foundation." Journal of
Transportation Engineering, ASCE, 131(6), 460-469.
Schwartz, C.W. (2002). Effect of Stress-dependent Base layer on the Superposition of
Flexible Pavement Solutions, The International Journal of Geomechanics, 2(3),
331-352.
Taciroglu, E. (1998). "Constitutive Modeling of the Resilient Response of Granular
Solids." Ph.D. thesis, University of Illinois at Urbana-Champaign, Urbana, IL.
Thompson, M.R. and Elliott, R.P. (1985). "ILLI-PAVE Based Response Algorithms
for Design of Conventional Flexible Pavements." Transportation Research
Record 1043, TRB, National Research Council, Washington, D.C., 50-57.
Thompson, M.R. and Robnett, Q.L. (1979). "Resilient Properites of Subgrade Soils."
Journal of Transportation Engineering, ASCE, Vol. 105, No.TEl.
Tutumluer, E. (1995). Predicting Behavior of Flexible Pavements with Granular
Bases. Ph.D. thesis, Georgia Institute of Technology, Atlanta.
Uzan, J. (1985). "Characterization of Granular Material," Transportation Research
Record 1022, TRB, National Research Council, Washington D.C., 52-59.
EVALUATION OF GEOGRID BENEFITS USING MONOTONIC AND
REPEATED LOAD TRIAXIAL TESTS
ABSTRACT
A series of monotonic and repeated load resilient modulus triaxial tests were
conducted on unreinforced and geogrid reinforced crushed limestone samples to
evaluate the effects of the stiffness, location, and number of geogrid layer/s on the
strength properties and stress strain response parameters of those samples. Five
different types of geogrids were used, and for each geogrid type, four different cases
were investigated. For each monotonic triaxial test, three response parameters were
selected to assess the benefits of geogrid improvement namely, Esio/0, ES2%, and USS.
Statistical analyses were conducted on the results using ANOVA and post ANOVA
LSM. The results of these analyses indicated that generally geogrid with higher
stiffness moduli exhibited greater improvement. In addition, double layers had always
the maximum improvement, while the minimum benefit was observed for single
geogrid layer placed at the sample mid-height. The cyclic triaxial test results were
used to determine resilient moduli for the different samples. These results showed
that the geogrid reinforcement did not enhance the resilient properties of the
reinforced samples.
INTRODUCTION
Research Associate, Louisiana Transportation Research Center, Louisiana State University, Baton
Rouge,
LA 70808.
2
Research Assistant Professor, Louisiana Transportation Research Center, Louisiana State University,
Baton Rouge, LA 70808.
3
Professor, Department of Civil and Environmental Engineering, Louisiana Transportation Research
Center, Louisiana State University, Baton Rouge, LA 70808.
145
146 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
pavements, bridge abutments, and many other applications. Many studies have been
conducted to characterize the behavior of reinforced soils in large-scale model
experiments (Haas et al., 1988; Barksdale et al., 1989; Miura et al., 1990; Webster,
1993; Cancelli et al., 1996; Collin et al., 1996; and Perkins and Ismeik, 1997; Berg et
al., 2000; Perkins, 2001). These studies showed that geogrids can mechanically
improve the overall strength and stability of reinforced structures.
TESTING PROGRAM
The testing program for this study included conducting a series of drained
montonic and repeated load triaxial tests on reinforced and unreinfoerced samples.
All tests were performed using the MTS closed loop, servo hydraulic triaxial testing
machine. The applied load was measured using a load cell installed inside the triaxial
cell. This type of the set up reduces the equipment compliance errors and also
alignment errors. The capacity of the load cell used was ± 5000 Ibf. The axial
displacement measurements were made using two Linearly Variable Differential
Transducer's (LVDT) placed between the top platen and base of the cell to reduce the
amount of extraneous axial deformation measured compared to external LDVTs. Air
was used as the confining fluid to the specimens.
Materials
Crushed limestone
The tested material was taken from a selected crushed limestone sample
typically used in construction of base course layers in Louisiana. The material has a
nominal size of 25.4 mm, and a DIG and D6o sizes of 0.18 mm and 6 mm respectively,
giving it a uniformity coefficient of 30. In addition, it is classified as A-l-a and GW-
GC according to the American Association of State Highway and Transportation
(AASHTO) classification system, and the Unified Soil Classification System (USCS).
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 147
The crushed limestone has a maximum dry unit weight 17.2 kN/m3 (139.2 lb/ft3) and
an optimum moisture content of 7.0%, as measured by the standard proctor test.
Geogrid
Sample Preparation
shaped load-pulse. These load pulses have a 0.1 sec load duration and 0.9 sec rest
period. Cyclic triaxial tests were conducted on unreinofrced samples, and samples
reinforced with one of the five geogrid types used in this study. Three different
placement cases were investigated for each reinforcement type; namely, single layer
placed at the sample mid-height, single layer placed at the one third of the sample
height from top, and double layers placed at one and two thirds of the sample height
from top. Three replicate samples were tested for each case.
The stress-strain curves were obtained from the monotonic triaxial tests for
unreinforced samples and samples reinforced with geogrid BX1100, BX1200, BX
6100, and BX 6200. To quantitively evaluate the improvement due to the
reinforcement under monotonic loading; three response parameters were first
determined for each triaxial test, namely, the secant elastic moduli at 1% strain level
(Esio/0), the secant elastic moduli at 2% strain level (Es2o/0), and the ultimate shear
strength (USS). These response parameters were chosen to assess the reinforcement
influence on the behavior of the tested material at different strain levels. An
improvement factor (IM) of each of the three response parameters was then calculated
for each reinforced sample as the ratio of that parameter to the corresponding average
value of the unreinforced samples. Figures la to Ic present the average improvement
factor of each reinforced case. It is clear that the improvement in all three parameters
due to the inclusion of geogrid reinforcement increased with increasing the stiffness
of the geogrid. However, this improvement was dependent on the location and
number of geogrid layer/s. In addition, the improvement in the USS and ES2% was
more apparent than that in Esio/0, with a maximum improvement in the ultimate shear
strength values (a maximum value of IM-USS of 2.5 was observed).
(a)
(b)
(c)
FIG. 1. Improvement in Response Factors:
a)Eslo/0 b)Es2o/0c)USS
150 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
The Type III Tests of fixed effects output were used to determine the
significance of the effects. The ANOVA results of the three analyses are presented
Table 1. It can be seen that at a 95% confidence level the type of geogrid and geogrid
placement case, had a significant effect on the improvement in Esio/0, E &%, USS.
While the interaction between the geogrid type and placement case had a significant
effect only on the improvement in the ES2%, USS. The significance of interaction
indicates that the behaviors of the two main effects (placement case and stifmess) are
inconsistent, which means that they do not increase and decrease by the same amount.
However, a closer look at the F-value reveals that the influence of the interaction
effect is small compared to the main effects.
Based on the ANOVA analyses result, Post ANOVA Least Square Means (LSM)
analyses were conducted to compare all of the different geogrid type-location cases.
Tukey adjustment was used in this analysis since it provides tests for all pairwise
comparisons at a well balance of the type I and type II errors when compared to other
adjustments available. Saxton's macro was conducted to convert the results of the
Post ANOVA-LSM analyses to letter groupings. The results of this grouping are
shown in Tables 2 and 3. Table 2 presents the grouping of the geogrid type effect on
Esi%, Es2%, and USS. It is observed that the maximum and minimum improvement in
all of the three response parameters was achieved when using geogrids BX-1500 and
BX-6100 respectively. While, the improvement in Esio/0 due to geogrid types BX-
1200 and BX-6200 was statistically indistinguishable from each other. However, the
effect of geogrid type BX-6200 on the improvement in Es2°/0 and USS were
distinguishable from BX-1200 but indistinguishable from BX-1100. In addition
Table 3 shows that the effect of the placement case on the improvement in Esio/0
values are significantly different from one another and rank from the highest to the
lowest values as follows: double > upper one third > lower one third > middle. While
the effect of the locations on the Es2o/0 USS had similar trend except that geogrids with
locations at the upper and lower one third were statistically indistinguishable from
each other.
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 151
Esio/0 Parameter
Geocrid Type Estimate Standard Error Letter Grouo
BX-1500 1.7413 0.01978 A
BX-6200 1.5705 0.01978 B
BX-1200 1.5681 0.01978 B
BX-1100 1.2658 0.01978 C
BX-6100 1.1817 0.01978 D
E^o/. Par ameter
BX-1500 1.9013 0.01532 A
BX-1200 1.6640 0.01532 B
BX-6200 1.5888 0.01532 C
BX-1100 1.5526 0.01532 C
BX-6100 1.2847 0.01532 D
USS Parameter
BX-1500 1.9215 0.01438 A
BX-1200 1.7038 0.01438 B
BX-6200 1.5810 0.01438 C
BX-1100 1.5682 0.01438 C
BX-6100 1.4986 0.01438 D
152 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
E$i% Parameter
Location Estimate Standard Error Letter Group
Double Layers 1.6049 0.01769 A
Upper one third 1.4993 0.01769 B
Lower one third 1.4286 0.01769 C
Middle 1.3291 0.01769 D
ES2% Parameter
Double Layers 1.9187 0.01371 A
Upper one third 1.6378 0.01371 B
Lower one third 1.5813 0.01371 C
Middle 1.2554 0.01371 D
USS Parameter
Double Layers 2.0358 0.01286 A
Upper one third 1.6633 0.01286 B
Lower one third 1.6302 0.01286 B
Middle 1.2892 0.01286 C
(2)
Where
R
: is the resilient modulus; pa: is the atmospheric pressure (101.3 kPa); 0: is the
bulk stress and equal to o} + 2a^, T°ct: is the octahedral shear stress and equal
The resilient modulus was then calculated for the different unreinforced and
reinforced samples at stress state that was selected based on previous finite element
study (Nazzal et al., 2006). Figure 2 presents the results of this calculation. It is
noticed that for samples reinforced with one layer a slight improvement in the
resilient modulus calculated at the selected stress state was detected for all tests.
However, no conclusion can be drawn since this improvement lies within the margin
of error of the calculated value.
FIG. 2. Resilient Modulus Values at Stress State Obtained From Finite Element
Analysis
CONCLUSIONS
REFERENCES
Ashmawy, A.K, Bourdeau, P.L., Drnevich, V.P., Dysli, M., (1999) "Cyclic Response
of Geotextile- Reinforced Soil," Journal of Soils and Foundations, Vol. 39, No. 1,
February, pp.
Barksdale, R.D., Brown, S. F., and Chan, F., 1989, "Potential benefits of geosynthetics
in flexible pavement systems", National Cooperative Highway research Program
Report No. 315, Transportation Research Board, National Research Council,
Washington, DC, USA, 56 p.
Berg, R. R., Christopher, B.R., and Perkins, S.W., 2000, "Geosynthetic reinforcement
of the aggregate base course of flexible pavement structures", GMA White paper II,
Geosynthetic material Assoiation, Roseville, MN, USA, 130 p.
Cancelli, A., and Montanelli, F., 1996, "In-ground test for geosynthetic reinforced
flexible paved roads", Proc. Of the Conference Geosynthetics '99, Boston, MA, USA,
Vol. 2, pp. 863-878.
Cancelli, A., Montanelli, F., Rimoldi, P., and Zhao, A, 1996, "Full scale laboratory
testing on geosynthetics reinforced paved roads", Earth reinforcement, Ochiai, H.,
Yasufuku, N., and Omine, K., Editors, Balkema, Proceedings of the International
Symposium on Earth Reinforcement, Fukuoka, Kyushu, Japan, November, 1996, pp.
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 155
573-578.
Haas, R., Walls, J. and Carroll, R.G. (1988). "Geogrid Reinforcement of Granular
Bases in Flexible Pavements," Transportation Research Record 1188, pp. 19-27.
Miura, N., Sakai, A., Taesiri, Y., Yamanouchi, T., and Yasuhara, K., 1990, "Polymer
grid reinforced pavement on soft clay grounds", geotextiles and geomembranes, Vol. 9,
No. 1, pp. 99-123.
Moghaddas-Nejad, F. and Small, J.C. (1996). "Effects of Geogrid Reinforcement in
Model Track Tests on Pavements", Journal of Transportation Engineering, Vol. 122,
No. 6, pp. 468-474.
Moghaddas-Nejad, F. and Small, J.C. (2003), "Resilient and Permanent Characteristics
ofReinforced Granular Materials by Repeated Load Triaxial Tests", ASTM
Geotechnical Testing Journal, Vol. 26, No. 2, pp. 152-166.
Nazzal M., Abu-Farsakh, M. Y., and Mohammad, L. (2006). Numerical Analyses of
Geogrid Reinforced Flexible Pavements. Proc., GeoCongress Conference, Atlanta,
NCHRP (2004), NCHRP Project 1-37A, Development ofNCHRP 1-37A Design Guide,
Using Mechanistic Principles to Improve Pavement Design, http://www.NCHRP 1-
37A Designdesignguide.com/.
Perkins, S.W. and Ismeik, M. (1997a). "A Synthesis and Evaluation of Geosynthetic
Reinforced Base Layers in Flexible Pavements: Part I," Geosynthetics International,
Vol. 4, No. 6, pp. 549- 605.
Perkins, S.W. and Ismeik, M. (1997b). "A Synthesis and Evaluation of Geosynthetic
Reinforced Base Layers in Flexible Pavements: Part II," Geosynthetics International,
Vol. 4, No. 6, pp. 549- 605.
Perkins, S.W., Christopher, B., Cuelho, E.L., Eiksund, G.R., Hoff, I., Schwartz, C.W.,
Svano, G., and Want, A. (2004). "Development of design methods for geosynthetic
reinforced flexible pavements". A report prepared for the U.S. Department of
transportation, Federal Highway Administration, FHWA/DTFH61-01-X-00068.
PERMANENT DEFORMATION OF SUBGRADE SOILS
INTRODUCTION
156
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 157
pavements and water will infiltrate the subgrade easily. However, few research
projects on permanent deformation in the subgrade soils have been conducted in
China. Test procedures and prediction models for permanent deformation with the
consideration of the specific conditions of China have not been established. Because
asphalt pavement properties, axle loads and soil properties in China may be different
from those elsewhere, the procedures and models developed in other countries may
not be used directly in China. As a result, permanent deformation of subgrade soils
has not been incorporated into the design of asphalt pavements in China.
The pioneering research work in the area of permanent deformation was credited to
the efforts of Monismith et al (1975). The power model proposed by Monismith is
well accepted in predicting the amount of rutting contributed by the subgrade. It is
expressed as
ep=ANb (1)
where
ep = permanent or plastic strain;
N = number of load repetitions;
A, b = material parameters (regression coefficients from test data)
Besides Monismith, many researchers have proposed models to predict the
permanent deformation of soils. These include a hyperbolic curve model
(Barksdale,1972), a polynomial model (Allen and Deen, 1986), a VESYS model
(Kenis, 1978), an incremental model (Bonaquist and Witczak, 1996) and a shakedown
model (Muhanna et al, 1998; Werkmeister et al, 2001). Zhao et al (2004) found that
the power model proposed by Monismith (1975) can be used to predict the permanent
deformation of most soils and established the regression formula for the coefficients
in the power model through repeated load tests for eight soils.
Models developed by Tseng and Lytton (1989) were used as the starting point in
MEPDG 2002 Design Guide to estimate the permanent deformation of unbound
materials. The basic relationship is given by
158 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
where
€o, /3 andp, material constants.
€r, resilient strain imposed in laboratory tests to obtain material properties
e0, /Sandyo.
j8ri, calibration factor.
The models for subgrade soils developed by Tseng and Lytton (1989) were given
by
(3c)
where
Wc = Water content (%)
ad = Deviator stress (kPa)
(70 = Bulk stress (kPa)
Er = Resilient modulus of the layer/sub-layer (kPa)
Extensive studies of the Ayres's modified models in the calibration process also
resulted in several unfavorable conditions. Through modifications, the final form of
the permanent deformation model for unbound materials used in the 2002 Design
Guide is given by
The testing procedures adopted by researchers are presented in Table 1. Uzan (1998)
conducted a survey on the moisture content of the subgrade in Israel. He found that
most of the moisture contents were 1.2 to 1.3 times of plastic limits (PL). So the
moisture content he adopted in permanent deformation tests was 1.22 times of the
plastic limit. The survey conducted by Elliott (1992) showed that the moisture
contents were about 100 to 120 percent of the optimum moisture contents (OMC) in
most cases. Few other researchers have made efforts to discuss the determination of
the test procedure. Puppala (1999) pointed out that the procedure for permanent
deformation tests of subgrade soils need to be studied further according to the specific
condition of highways.
The testing program for measuring the resilient modulus of subgrade soils can be
referred when the program of permanent deformation test is determined. The major
specifications about the test of the resilient modulus include T307-99, LTPP P46-96,
NCHRP 1-28, and NCHRP 1-28A. The test configurations in the four specifications
are presented in Table 2.
Moisture Content
Table 1 shows that in some research the moisture contents adopted are all greater
than the OMC. But in other research the moisture contents of less than, greater than
and equal to the OMC are adopted at the same time. Surveys have shown that the
moisture contents of subgrade soils in China, which have been kept in balance, are
often greater than the OMC. To make the design of asphalt pavement safer, the
moisture contents used by Elliott et al (1998), 105% and 110% and 120% of OMC are
adopted.
Dry Density
The Technical Standard of Highway Engineering of China requires that the ratio of
the dry density to the maximum dry density (MOD) should be 94% to 96% and 93%
to 96% for the soils which are 0 to 80 cm and 80 to 150 cm from the top of the
subgrade, respectively. But the ratio may be less than 93% in the construction of
subgrades, so 91% and 96% are recommended to be used.
TABLE 1. Summary of the Test Procedures byr Various Permanent Deformation Studies
Behzadi
Monismith et al Muhann et al Uzan Puppala Suppakit Zhao
Configurations Yandell
(1975) (1998) (1998) (1999) (2003) (2004)
(1996)
Half Half
Load Shape Rectangle Half sine Half sine Half sine
sine sine
Load duration
j\
(second)
0.1 0.5 0.1 0.1 0.1 0.1 0.1
Rest Period
2.9 1 0.9 0.9 0.9 0.9 0.9
(second)
No. of Applications 10 or 3 or
10 or 100 10 10 10 10 or 100
(in thousands) 100 2.5
Confining Pressure Oand
34.5 15 to 50 20.7 a b 21
(kPa) 20
Deviator Stress 25 to 90% of 40 and 0.2, 0.4 and 0.6
34.5 to 138 34.5 to 345 c 14tol66
(kPa) strength 75 times of strength
Fine and coarse
Soil Tested Silty clay Silty clay A-5 and A-6 Eight soils
grained soils
Dry Density
90 to 95 90 to 100 90 95 95
(%ofMMC)
OMC, 105, 110 and 120% of
Moisture Content 16 to 20% 16% 1.22PL
OMC±2.5% OMC
where a=the confining pressure is 21, 35 70, 105, 140 kPa for a sand and 0, 21, 42 kPa for a clay;
b=the confining pressure is 0, 21, 42 kPa for a clay and a silty clay and 21, 48 , 97 kPa for a sand and coarse-grained soils;
c = the deviator stress is 21, 35, 52.5, 70, 105, 140, 210, 280 kPa for a sand and 14, 28, 42, 56, 70 kPa for a clay.
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 161
Load shape Half sine Half sine Half sine Half sine
Load duration 0.1 0.1 0.1 0.2
(second)
Rest Period 0.9 3 0.9 0.9 0.8
(second)
Confining Fine-grained 41.4,27.6, 41.4,27.6, 0 55.2,41.4,
Pressure soils 13.8 13.8 27.6, 13.8
(kPa) Coarse-grained 41.4,27.6, 41 .4, 27. 6, 13.8, 20.7, 13.8, 27.6,
soils 13.8 13.8 27.6,41.4 41.4,55.2,
82.8
Deviator Fine-grained 12.4, 24.8, 12.4, 24.8, 13.8, 27.6, 27.6, 48.3,
Stress soils 37.3, 49.7, 37.3, 49.7, 41.4,55.2, 69, 96.6
(KPa) 62.1 62.1 69
Coarse-grained 12.4, 24.8, 12.4, 24.8, 13.8, 20.7, 6.9, 13.8,
soils 37.3, 49.7, 37.3, 49.7, 27.6,41.4, 20.7, 27.6,
62.1 62.1 55.2 41.4,55.2,
82.8, 110.4,
124.2, 165.6,
248.4
Deviator Stress
The compressive stress at the top of the subgrade has been calculated for many
asphalt pavements used in China. Results show that the compressive stresses are 10 to
180 kPa, with typical values of 10 to 30 kPa in most cases.
The field test conducted in an expressway by Zhang et al (2005) showed that when
the axle load was 17.4 KN, the compressive stress at 30 cm below the top of the
subgrade was 12 kPa.
To determine the upper limit of the deviator stress, the compressive stress at the top
of the subgrade is calculated for the pavement shown in Table 3, which may represent
the weakest pavement used in China. The compressive stress is 45 kPa under the
standard axle load of 10 KN. If the axle load is 50% greater than the standard axle load
it becomes 68 kPa, which can represent the largest compressive stress in the subgrade
in China.
Based on the discussion above, referring to the deviator stresses in NCHRP 1-28A,
three deviator stress levels are selected. They are 28, 48 and 69 KPa for fine-grained
soils and 28, 55 and 83 KPa for coarse-grained soils.
Confining Pressure
Several researchers (Allen and Thompson, 1974; Brown and Hyde, 1975) explored
162 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEM
S
the effect of variable (cyclic) confining pressures on the deformation of subgrade soils.
Similar results for resilient and permanent strain were obtained from cyclic and
constant confining pressures. This suggests that a fixed confining pressure could be
used in repeated loading tests to get deformation data. Elliott el al (1998) and Zhao et
al (2004) found that the confining pressure was not a critical factor influencing
permanent deformation. It can also be seen from Table 1 that the fixed confining
pressure of 21 kPa is used by about half of the researchers.
The confining pressure in the subgrade is calculated for many asphalt pavements
used in China. Results show that, whether subgrade is made up of fine or coarse
grained soils, the range of the confining pressure is from 0 to 60 kPa. In most cases, it
is 10 to 30 kPa.
Table 2 also presents the confining pressures adopted in some major specifications
for the resilient modulus test in USA. Table 2 shows that the confining pressure of
27.6 kPa is adopted in all of the four specifications. 27.6 kPa lies between 10 to 30
kPa and is close to 21 kPa. So the confining pressure of 28 kPa, which is the integer
being most close to 27.6 kPa, is adopted in this study.
asphalt macadam ^
Friction course 3 Q 25
mixture
Graded crushed
Base course 2Q 30Q Q 2J
stone
Sub base course sand 20 200 0.25
subgrade 30 0.35
Table 1 shows that a repeated load frequency of 1 HZ, a rest period of 0.9 second
and a load duration of 0.1 second are adopted by all of the researchers after 1998. This
stress pulse is produced by the vehicle running at the velocity of 80 km/h. But in
NCHRP 1-28A they are considered to be only adapted to pavement materials. The
subgrade is under the pavement, thus the duration of the stress pulse caused by
vehicles will be closer to 0.2 second in the subgrade as compared to 0.1 second in the
pavement. So a repeated load frequency of 1 HZ, a rest period of 0.8 second and a
load duration of 0.2 second are adopted in NCHRP 1-28A.
Zhang et al (2005) measured the compressive stress distribution in the subgrade of
an expressway under moving vehicles. When the weight of the rear axle was 9.8 kN
and the moving velocity was 60 km/h, the compressive stress distribution in subgrade
was shown in Fig. 1. The higher and lower peaks in Fig.l are caused by the rear axle
and front axle, respectively. It can be calculated that the load duration in the subgrade
is about 0.3 to 0.4 second. If the velocity of the vehicle is 80 km/h the load duration
will be 0.2 to 0.3 second.
Based on the discussions above, a repeated load frequency of 1 HZ, a rest period of
ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS 163
0.8 second and a load duration of 0.2 second are adopted in this study.
Finally, the following procedure parameters are selected for the permanent
deformation test of subgrade soils:
(1) Testing Configuration
Load Frequency: 1HZ.
Load Duration: 0.2 second.
Rest Period: 0.8 second.
Confining Pressure: 28 kPa.
Deviator Stress: 28, 48 and 69 kPa for fine-grained soils; 28, 55and 83 kPa for
coarse-grained soils.
Number of Load Application: 10,000
(2) Compaction Specification
Compaction Method: Static compactor.
Compacted Dry Density: 91 and 96% of MDD.
Compacted Moisture Content: 105, 110 and 120% of OMC.
TEST RESULTS
Repeated load tests are conducted for sand, silty clay and clay in UTM. Test
specimens are 200 mm in height and 100 mm in diameter. The basic properties of the
soils are listed in Table 4.
164 ANALYSIS OF ASPHALT PAVEMENT MATERIALS AND SYSTEMS
When the deviator stress is equal to 28 kPa, the permanent deformation increases
with time at first and then decreases for some tests. Obviously this situation doesn't
exist in reality. Therefore, these test results are not used in the analysis.
Elliott (1998) found that the initial data points couldn't completely represent soils'
behavior because the loading system was being tuned to apply for the correct deviator
stress during the first 10 to 30 cycles of load. Test results given by the authors show
that an excessively large portion of the total permanent deformation is developed
during the first 10 to 20 cycles of load. In the construction of highways, after the
sugrade is completed, it will be compacted further by the rollers and trucks during the
construction of the pavement. So the first 20 cycles of load can be included in this
further compaction and the permanent deformation produced in these cycles will not
contribute to the rut on the surface of the pavement.
For the reasons above, permanent deformation produced in the first 20 cycles of
load are ignored. Some test results are plotted in Appendix II.
It should be noted that, the work condition with the moisture content of 105% of
OMC and the confining pressure of 28 kPa is named "105-28". In the figures in
Appendix II, this work condition is noted as 1. Similarly, the work condition "105-48"
is noted as 2, "105-68" is noted as 3, "110-28" is noted as 4, "110-48" is noted as 5,
"110-69" is noted as 6, "120-28" is noted as 7, "120-48" is noted as 8 and "120-69" is
noted as 9.
Analysis of the test results shows that the relationship between the permanent strain
and the number of load applications is close to the function given in Equation 2. The
correlation coefficient R2 in any work condition is greater than 0.99. This proves that
the basic relationship shown in Equation 2 can be used to predict the permanent
deformation of subgrade soils. However the test results are not in accordance with the
results calculated from any one of the models used by Tseng and Lytton, Ayre et al, or
MEPDG Design guide.
According to the test results, the coefficients in the formula are given as
The regression coefficient R2 in these formulas is all greater than critical values.
Subgrade soils play a critical role in the initiation and propagation of permanent
deformation of pavements and directly influence the pavement performance. Based
on comprehensive literature review, the basic configurations of permanent
deformation tests for subgrade soils have been established according to the specific
condition in China. The results of repeated load tests conducted to three soils are in
well accordance with the basic relationships between the permanent strain and the
number of load applications proposed by Tseng and Lytton. The regression models for
sand, silty clay and clay are found.
APPENDIX L NOTATION
FIG H-3. Development of Permanent Deformation (silty clay with dry density of
96%ofMDD)
REFERENCES:
171
This page intentionally left blank
Author Index
Page number refers to the first page of paper
Li, Qingbin, 1
173