NATIONAL

NCHRP
REPORT 646
COOPERATIVE
HIGHWAY
RESEARCH
PROGRAM

Validating the Fatigue

Endurance Limit for
Hot Mix Asphalt
TRANSPORTATION RESEARCH BOARD 2010 EXECUTIVE COMMITTEE*

OFFICERS
CHAIR: Michael R. Morris, Director of Transportation, North Central Texas Council of Governments, Arlington
VICE CHAIR: Neil J. Pedersen, Administrator, Maryland State Highway Administration, Baltimore
EXECUTIVE DIRECTOR: Robert E. Skinner, Jr., Transportation Research Board

MEMBERS
J. Barry Barker, Executive Director, Transit Authority of River City, Louisville, KY
Allen D. Biehler, Secretary, Pennsylvania DOT, Harrisburg
Larry L. Brown, Sr., Executive Director, Mississippi DOT, Jackson
Deborah H. Butler, Executive Vice President, Planning, and CIO, Norfolk Southern Corporation, Norfolk, VA
William A.V. Clark, Professor, Department of Geography, University of California, Los Angeles
Nicholas J. Garber, Henry L. Kinnier Professor, Department of Civil Engineering, and Director, Center for Transportation Studies, University of
Virginia, Charlottesville
Jeffrey W. Hamiel, Executive Director, Metropolitan Airports Commission, Minneapolis, MN
Edward A. (Ned) Helme, President, Center for Clean Air Policy, Washington, DC
Randell H. Iwasaki, Director, California DOT, Sacramento
Adib K. Kanafani, Cahill Professor of Civil Engineering, University of California, Berkeley
Susan Martinovich, Director, Nevada DOT, Carson City
Debra L. Miller, Secretary, Kansas DOT, Topeka
Pete K. Rahn, Director, Missouri DOT, Jefferson City
Sandra Rosenbloom, Professor of Planning, University of Arizona, Tucson
Tracy L. Rosser, Vice President, Corporate Traffic, Wal-Mart Stores, Inc., Mandeville, LA
Steven T. Scalzo, Chief Operating Officer, Marine Resources Group, Seattle, WA
Henry G. (Gerry) Schwartz, Jr., Chairman (retired), Jacobs/Sverdrup Civil, Inc., St. Louis, MO
Beverly A. Scott, General Manager and Chief Executive Officer, Metropolitan Atlanta Rapid Transit Authority, Atlanta, GA
David Seltzer, Principal, Mercator Advisors LLC, Philadelphia, PA
Daniel Sperling, Professor of Civil Engineering and Environmental Science and Policy; Director, Institute of Transportation Studies; and Interim
Director, Energy Efficiency Center, University of California, Davis
Douglas W. Stotlar, President and CEO, Con-Way, Inc., Ann Arbor, MI
C. Michael Walton, Ernest H. Cockrell Centennial Chair in Engineering, University of Texas, Austin

EX OFFICIO MEMBERS
Thad Allen (Adm., U.S. Coast Guard), Commandant, U.S. Coast Guard, U.S. Department of Homeland Security, Washington, DC
Peter H. Appel, Administrator, Research and Innovative Technology Administration, U.S.DOT
J. Randolph Babbitt, Administrator, Federal Aviation Administration, U.S.DOT
Rebecca M. Brewster, President and COO, American Transportation Research Institute, Smyrna, GA
George Bugliarello, President Emeritus and University Professor, Polytechnic Institute of New York University, Brooklyn; Foreign Secretary,
National Academy of Engineering, Washington, DC
Anne S. Ferro, Administrator, Federal Motor Carrier Safety Administration, U.S.DOT
LeRoy Gishi, Chief, Division of Transportation, Bureau of Indian Affairs, U.S. Department of the Interior, Washington, DC
Edward R. Hamberger, President and CEO, Association of American Railroads, Washington, DC
John C. Horsley, Executive Director, American Association of State Highway and Transportation Officials, Washington, DC
David T. Matsuda, Deputy Administrator, Maritime Administration, U.S.DOT
Victor M. Mendez, Administrator, Federal Highway Administration, U.S.DOT
William W. Millar, President, American Public Transportation Association, Washington, DC
Cynthia L. Quarterman, Administrator, Pipeline and Hazardous Materials Safety Administration, U.S.DOT
Peter M. Rogoff, Administrator, Federal Transit Administration, U.S.DOT
David L. Strickland, Administrator, National Highway Traffic Safety Administration, U.S.DOT
Joseph C. Szabo, Administrator, Federal Railroad Administration, U.S.DOT
Polly Trottenberg, Assistant Secretary for Transportation Policy, U.S.DOT
Robert L. Van Antwerp (Lt. Gen., U.S. Army), Chief of Engineers and Commanding General, U.S. Army Corps of Engineers, Washington, DC

*Membership as of February 2010.

NATIONAL COOPERATIVE HIGHWAY RESEARCH PROGRAM

NCHRP REPORT 646

Validating the Fatigue Endurance Limit
for Hot Mix Asphalt
Brian D. Prowell
ADVANCED MATERIALS SERVICES, LLC1
Auburn, AL
E. Ray Brown
U.S. ARMY CORPS OF ENGINEERS1
Vicksburg, MS
R. Michael Anderson
ASPHALT INSTITUTE
Lexington, KY
Jo Sias Daniel
Aravind Krishna Swamy
UNIVERSITY OF NEW HAMPSHIRE
Durham, NH
Harold Von Quintus
APPLIED RESEARCH ASSOCIATES, INC.
Austin, TX
Shihui Shen2
WASHINGTON STATE UNIVERSITY
Pullman, WA
Samuel H. Carpenter
UNIVERSITY OF ILLINOIS
Urbana-Champaign, IL
Sudip Bhattacharjee3
ALABAMA A&M UNIVERSITY
Huntsville, AL
Saeed Maghsoodloo
NATIONAL CENTER FOR ASPHALT TECHNOLOGY
Auburn, AL

Formerly 1National Center for Asphalt Technology, 2University of Illinois, and 3University of New Hampshire, respectively.

Subscriber Categories
Materials • Pavements

Research sponsored by the American Association of State Highway and Transportation Officials
in cooperation with the Federal Highway Administration

TRANSPORTATION RESEARCH BOARD

WASHINGTON, D.C.
2010
www.TRB.org
NATIONAL COOPERATIVE HIGHWAY NCHRP REPORT 646
RESEARCH PROGRAM
Systematic, well-designed research provides the most effective Project 09-38
approach to the solution of many problems facing highway ISSN 0077-5614
administrators and engineers. Often, highway problems are of local ISBN 978-0-309-11821-7
interest and can best be studied by highway departments individually Library of Congress Control Number 2010921174
or in cooperation with their state universities and others. However, the © 2010 National Academy of Sciences. All rights reserved.
accelerating growth of highway transportation develops increasingly
complex problems of wide interest to highway authorities. These
problems are best studied through a coordinated program of COPYRIGHT INFORMATION
cooperative research.
Authors herein are responsible for the authenticity of their materials and for obtaining
In recognition of these needs, the highway administrators of the written permissions from publishers or persons who own the copyright to any previously
American Association of State Highway and Transportation Officials published or copyrighted material used herein.
initiated in 1962 an objective national highway research program Cooperative Research Programs (CRP) grants permission to reproduce material in this
employing modern scientific techniques. This program is supported on publication for classroom and not-for-profit purposes. Permission is given with the
understanding that none of the material will be used to imply TRB, AASHTO, FAA, FHWA,
a continuing basis by funds from participating member states of the
FMCSA, FTA, or Transit Development Corporation endorsement of a particular product,
Association and it receives the full cooperation and support of the method, or practice. It is expected that those reproducing the material in this document for
Federal Highway Administration, United States Department of educational and not-for-profit uses will give appropriate acknowledgment of the source of
any reprinted or reproduced material. For other uses of the material, request permission
Transportation.
from CRP.
The Transportation Research Board of the National Academies was
requested by the Association to administer the research program
because of the Board’s recognized objectivity and understanding of
NOTICE
modern research practices. The Board is uniquely suited for this
purpose as it maintains an extensive committee structure from which The project that is the subject of this report was a part of the National Cooperative Highway
Research Program conducted by the Transportation Research Board with the approval of
authorities on any highway transportation subject may be drawn; it the Governing Board of the National Research Council. Such approval reflects the
possesses avenues of communications and cooperation with federal, Governing Board’s judgment that the program concerned is of national importance and
state and local governmental agencies, universities, and industry; its appropriate with respect to both the purposes and resources of the National Research
Council.
relationship to the National Research Council is an insurance of
The members of the technical committee selected to monitor this project and to review this
objectivity; it maintains a full-time research correlation staff of
report were chosen for recognized scholarly competence and with due consideration for the
specialists in highway transportation matters to bring the findings of balance of disciplines appropriate to the project. The opinions and conclusions expressed
research directly to those who are in a position to use them. or implied are those of the research agency that performed the research, and, while they have
been accepted as appropriate by the technical committee, they are not necessarily those of
The program is developed on the basis of research needs identified
the Transportation Research Board, the National Research Council, the American
by chief administrators of the highway and transportation departments Association of State Highway and Transportation Officials, or the Federal Highway
and by committees of AASHTO. Each year, specific areas of research Administration, U.S. Department of Transportation.
needs to be included in the program are proposed to the National Each report is reviewed and accepted for publication by the technical committee according
Research Council and the Board by the American Association of State to procedures established and monitored by the Transportation Research Board Executive
Committee and the Governing Board of the National Research Council.
Highway and Transportation Officials. Research projects to fulfill these
needs are defined by the Board, and qualified research agencies are The Transportation Research Board of the National Academies, the National Research
Council, the Federal Highway Administration, the American Association of State Highway
selected from those that have submitted proposals. Administration and and Transportation Officials, and the individual states participating in the National
surveillance of research contracts are the responsibilities of the National Cooperative Highway Research Program do not endorse products or manufacturers. Trade
Research Council and the Transportation Research Board. or manufacturers’ names appear herein solely because they are considered essential to the
object of this report.
The needs for highway research are many, and the National
Cooperative Highway Research Program can make significant
contributions to the solution of highway transportation problems of
mutual concern to many responsible groups. The program, however, is
intended to complement rather than to substitute for or duplicate other
highway research programs.

Published reports of the

NATIONAL COOPERATIVE HIGHWAY RESEARCH PROGRAM

are available from:

Transportation Research Board

Business Office
500 Fifth Street, NW
Washington, DC 20001

and can be ordered through the Internet at:

http://www.national-academies.org/trb/bookstore
Printed in the United States of America
COOPERATIVE RESEARCH PROGRAMS

CRP STAFF FOR NCHRP REPORT 646

Christopher W. Jenks, Director, Cooperative Research Programs
Crawford F. Jencks, Deputy Director, Cooperative Research Programs
Edward T. Harrigan, Senior Program Officer
Melanie Adcock, Senior Program Assistant
Eileen P. Delaney, Director of Publications
Hilary Freer, Senior Editor

NCHRP PROJECT 09-38 PANEL

Field of Materials and Construction—Area of Bituminous Materials
Linda M. Pierce, Applied Pavement Technology, Inc., Santa Fe, NM (Chair)
Steven W. Krebs, Wisconsin DOT, Madison, WI
Allen Cooley, Burns Cooley Dennis, Inc., Ridgeland, MS
G. William Maupin, Jr., Virginia DOT, Charlottesville, VA
Richard W. May, SemMaterials, LP, Tulsa, OK
Leslie Ann McCarthy, Villanova University, Villanova, PA
Carl L. Monismith, University of California - Berkeley, Berkeley, CA
Amy M. Schutzbach, Illinois DOT, Springfield, IL
Gregory A. Sholar, Florida DOT, Gainesville, FL
Linbing Wang, Virginia Polytechnic Institute and State University, Blacksburg, VA
Travis M. Wombwell, Missouri DOT, Macon, MO
Wei-Shih “Wes” Yang, New York State DOT, Albany, NY
Ernest Bastian, FHWA Liaison
John “Jack” Youtcheff, FHWA Liaison
Frederick Hejl, TRB Liaison
FOREWORD

By Edward T. Harrigan
Staff Officer
Transportation Research Board

This report presents the findings of research performed to investigate the existence of a
fatigue endurance limit for hot mix asphalt (HMA) mixtures, the effect of HMA mixture
characteristics on the endurance limit, and the potential for the limit’s incorporation in
structural design methods for flexible pavements. The report describes the research per-
formed and includes proposed standard practices using various experimental and analyti-
cal procedures for determining the endurance limit of HMA mixtures. Thus, the report will
be of immediate interest to materials and structural design engineers in state highway agen-
cies and engineers in the HMA construction industry.

Many well-constructed flexible pavements with a thick HMA structure have been in
service for 40 or more years without any evidence of bottom-up fatigue cracking. This field
experience suggests that an endurance limit, that is, a level of strain below which fatigue damage
does not occur for any number of load repetitions, is a valid concept for HMA mixtures.
The concept of an endurance limit is widely recognized in many areas of materials science,
especially that of ferrous metals. The endurance limit is usually calculated from the relationship
of strain to load repetitions to failure. Defining an endurance limit for HMA mixtures will
result in more efficient structural design of flexible pavements built with mixtures of
varying properties. For instance, small increases in the binder content of HMA mixtures
provide longer fatigue lives (presumably because of a higher strain level for the endurance
limit of these mixtures). Other factors likely to determine the value of the fatigue endurance
limit for a given HMA mixture are the incorporation of a modifier in the asphalt binder, the
aggregate type and gradation, the asphalt binder grade, and the mixture’s volumetric
properties.
Previous research suggested that the fatigue behavior of flexible pavements is consistent
with the existence of an endurance limit with an approximate value of 70 microstrains.
However, few laboratory studies corroborate this value. Moreover, pavement design
approaches, including the Mechanistic-Empirical Pavement Design Guide (MEPDG)
developed in NCHRP Projects 1-37A and 1-40, do not fully incorporate the endurance limit
concept. This is because, to date, research into the fatigue of HMA mixtures has been
limited to strain levels well above the hypothesized value of 70 microstrains.
The objectives of this research were to (1) test the hypothesis that there is an endurance
limit in the fatigue behavior of HMA mixtures, (2) measure the value of the endurance limit
for a representative range of HMA mixtures, and (3) recommend a procedure to incorporate
the effects of the endurance limit into mechanistic pavement design methods. The research
was performed by the National Center for Asphalt Technology, Auburn University,
Auburn, AL, with the assistance of the following organizations: Applied Research Associates,
Inc., Round Rock, TX; the Asphalt Institute, Lexington, KY; the University of Illinois,
Urbana-Champaign, IL; and the University of New Hampshire, Durham, NH.
The report fully documents the design and conduct of an extensive laboratory program
of beam fatigue and uniaxial tension testing that experimentally confirmed the existence of
an HMA fatigue endurance limit and quantified how the value of the limit is influenced by
HMA mixture and binder properties. Based on these results, a practical definition of the
endurance limit was developed, along with a methodology to estimate the endurance limit
in the laboratory. Analysis of in-service pavements by the research team also demonstrated
the existence of the endurance limit and indicated that polymer modification of asphalt
binders improves the fatigue performance of HMA mixtures and flexible pavements.
Finally, sensitivity analyses were conducted that indicated that the value of the endurance
limit can affect the recommended thickness of perpetual pavements designed with the
MEPDG and PerRoad methodologies significantly.
This report includes six appendices as follows:

• Appendix A: A Proposed Standard Practice for Predicting the Endurance Limit of Hot Mix
Asphalt (HMA) for Long-Life Pavement Design;
• Appendix B: A Proposed Standard Practice for Predicting the Endurance Limit of Hot Mix
Asphalt (HMA) by Pseudo Strain Approach;
• Appendix C: A Proposed Standard Practice for Extrapolating Long-Life Beam Fatigue Tests
Using the Ratio of Dissipated Energy Change (RDEC);
• Appendix D: A summary of results of the beam fatigue testing accomplished during the project;
• Appendix E: An analytical method for construction of a damage characteristic curve through
calculation of pseudo strains; and
• Appendix F: A proposed design for an interlaboratory study to develop a precision statement
for AASHTO T321, Standard Method of Test for Determining the Fatigue Life of Compacted
Hot-Mix Asphalt (HMA) Subjected to Repeated Flexural Bending.

The proposed standard practices are under consideration for possible adoption by the
AASHTO Highway Subcommittee on Materials and the AASHTO Joint Technical Commit-
tee on Pavements.
CONTENTS

1 Summary
5 Chapter 1 Introduction and Research Approach
5 Introduction
6 Research Problem Statement
6 Objectives
6 Scope
7 Chapter 2 State of Practice
7 Arguments for the Existence of the Endurance Limit
9 Factors Affecting Fatigue Life
9 Strategies to Produce Long-Life Pavements
12 Laboratory Fatigue Tests and Analysis Methods
13 Laboratory Studies to Quantify the Endurance Limit
13 Modeling Fatigue and Relationship to Field Performance
15 Chapter 3 Research Plan
15 Introduction
16 Materials
17 Test Methods
20 Chapter 4 Beam Fatigue Test Results and Analyses
20 Extrapolation Methods to Predict Fatigue Life
32 Existence of the Endurance Limit
41 Estimate of Precision of Beam Fatigue Tests
42 Indirect Tensile Strength as a Surrogate for Endurance
Limit Determination
44 Chapter 5 Uniaxial Tension Results and Analyses
44 Test Specimens
44 Dynamic Modulus and Phase Angle Master Curves
49 Damage Characteristic Curve
53 Evaluation of Endurance Limit
59 Chapter 6 Examination of LTPP Database for Indications
of an Endurance Limit
59 Introduction
59 Including the Endurance Limit Design Premise into
Mechanistic-Empirical-Based Pavement Design Procedures
62 Defining the Endurance Limit—A Survivability Analysis
66 Updated Survivability Analysis Using LTPP Data
 73 Chapter 7 Sensitivity of Pavement Thickness
to the Endurance Limit
73 Summary of Predicted Endurance Limits
73 Estimate of Shift Factors between Laboratory Tests and
Field Performance
83 Sensitivity of Mechanistic-Empirical Pavement Design Methods
to the Endurance Limit
88 Considerations for Incorporating the Endurance Limit
into M-E Design Procedures
91 Chapter 8 Conclusions and Recommendations
91 Conclusions
91 Recommendations
93 References
97 APPENDIX A Proposed Standard Practice for Predicting the
Endurance Limit of Hot Mix Asphalt (HMA) for
Long-Life Pavement Design
106 APPENDIX B Proposed Standard Practice for Predicting the
Endurance Limit of Hot Mix Asphalt (HMA) by
Pseudo Strain Approach
112 APPENDIX C Proposed Standard Practice for Extrapolating
Long-Life Beam Fatigue Tests Using the Ratio
of Dissipated Energy Change (RDEC)
117 APPENDIX D NCHRP 9-38 Beam Fatigue Data
122 APPENDIX E Construction of Characteristic Curve
127 APPENDIX F NCHRP 9-38 Beam Fatigue Round Robin
AUTHOR ACKNOWLEDGMENTS
The research described herein was performed under NCHRP Project 9-38 by the National Center for
Asphalt Technology (NCAT) at Auburn University. Ray Brown served as the Principal Investigator and
Brian Prowell and Michael Anderson served as Co-Principal Investigators. Brown and Prowell were
employed by NCAT when this research began.
The research team included the Asphalt Institute; the University of New Hampshire; Applied Research
Associates, Inc.; the University of Illinois, and, later, Advanced Materials Services, LLC. The Asphalt Insti-
tute assisted in beam fatigue testing in Phases I and II of the project. Jo Daniel oversaw uniaxial tension
testing and analyses at the University of New Hampshire. Harold Von Quintus with Applied Research
Associates, Inc., performed survival analyses of in-service pavements and provided guidance on the
MEPDG. Samuel Carpenter oversaw beam fatigue testing in Phase II of the study with Shihui Shen at the
University of Illinois and provided expertise on analyses using the ratio of dissipated energy. Brian Prowell
was primarily responsible for the technical content of the remaining sections and for assembling the final
report. Pamela Turner supervised laboratory testing at NCAT.
The project team appreciates the support and technical assistance of Bor-Wen Tsai of the University of
California, Berkeley; Rich May of SEM Materials; and Bill Maupin, Jr., of the Virginia Transportation
Research Council. These three agencies volunteered to participate in the beam fatigue mini round-robin
to help develop a better understanding of the variability of beam fatigue tests. They also assisted in the
evaluation of the draft protocols for determining the endurance limit.
Other individuals provided invaluable assistance to the project. Bor-Wen Tsai provided significant
technical assistance with the single- and three-stage Weibull method. Chris Wagner (FHWA) assisted
with the MEPDG. Buzz Powell provided technical data from the NCAT Test Track. David Timm and
Richard Willis provided advice and guidance for calculating shift factors based on field strain measure-
ments and incorporating the endurance limit into M-E pavement design.
 1

SUMMARY

Validating the Fatigue Endurance Limit

for Hot Mix Asphalt

Hot-mix asphalt (HMA) pavements have been designed primarily to resist rutting of
the subgrade and bottom-up fatigue cracking. In classical pavement design, as design load
applications increase, pavement thickness also must increase. There is a growing belief that
bottom-up fatigue cracking does not occur for thick pavements. The concept of the HMA
fatigue endurance limit—a level of strain below which there is no cumulative damage over
an indefinite number of load cycles—is proposed to explain this occurrence. Therefore, addi-
tional pavement thickness, greater than that required to keep strain levels at the bottom of
the HMA layer below the endurance limit, would not provide additional life. This concept
has significant design and economic implications.
Research in the 1970s observed that the log-log relationship between strain and bending
cycles for a number of HMA mixes converged below 70 micro-strain (ms) at a high number
of loading cycles. Field studies in the United Kingdom recommended an HMA thickness
range between 7.9 in. and 15.4 in. for a long-life pavement depending on such factors as
binder stiffness. The stiffness of thick pavements was observed to increase with time, most
likely due to binder aging. Other studies have documented the absence of bottom-up fatigue
cracking in thick pavements and the common occurrence of top-down cracking.
In theory, samples tested at a strain level below the endurance limit should last for an
indefinite or infinite number of loading cycles. It is impossible to test samples to an infinite
number of cycles. Therefore, a practical definition of the endurance limit, or a laboratory
life representative of the endurance limit was needed. A capacity analysis indicated that at
minimum safe spacing, a lane carrying 100% trucks 24 hours a day, 7 days a week for 40 years
could carry a maximum of 329,376,000 trucks that would produce 1,317,504,000 axle rep-
etitions (neglecting the steer axle). However, a more likely traffic stream would produce a
maximum of approximately 500 million axle load repetitions.
Studies have indicated that a shift factor, ranging from 4 to 100, must be applied to relate
laboratory and field fatigue performance. There are many reasons that probably lead to the
need for a shift factor with two primary factors being rest periods and healing. A shift factor
of 10 was recommended in the Strategic Highway Research Program (SHRP) by Leahy et al.
(46). Considering this shift factor, laboratory testing to 50 million cycles would equate to
approximately 500 million loading cycles in the field or approximately the maximum prob-
able loading in a 40-year period. This was considered a practical target for evaluating param-
eters indicating an endurance limit.
An experimental plan was developed using a single aggregate gradation, two levels of
binder content (optimum and optimum plus 0.7%), and two binder grades (PG 67-22 and
PG 76-22). The aggregates, gradations, optimum binder content, and binder grades/sources
matched those used in the base layers of the 2003 National Center for Asphalt Technology
(NCAT) Test Track structural sections. Using the same mixtures as the NCAT Test Track
2

allowed calibration of the fatigue shift factor. Samples at optimum asphalt content were
compacted to 7 ± 0.5% air voids while samples at optimum plus asphalt content were com-
pacted to 3.3 ± 0.5% air voids. Beam fatigue and uniaxial tension testing were performed on
the four mixtures. In Phase II, beam fatigue tests were performed on two additional mixes
at optimum asphalt content using a PG 58-28 and PG 64-22.
Beam fatigue testing was conducted according to AASHTO T321. At least two replicates
were tested at 800, 400, 200, or 100 ms until the fatigue lives of two replicates at a given strain
level exceeded 50 million cycles. The time to test a single sample to 50 million load cycles is
approximately 50 days. At this point, a log-log regression was performed between strain and
fatigue life using all of the data for which samples failed in less than 50 million cycles. The
strain level that corresponded to a fatigue life of 50 million cycles was predicted. Two addi-
tional beams were tested at this strain level.
A number of extrapolation techniques were considered to predict the number of cycles to
failure (Nf) for beams that are not tested to failure, including the exponential model described
in AASHTO T321, a logarithmic regression, ratio of dissipated energy change, and single-
and three-stage Weibull models. The extrapolations were evaluated using samples that had
long fatigue lives—in excess of 12 million cycles—but failed before 50 million cycles. The
logarithmic regression and ratio of dissipated energy change consistently overestimated Nf,
often by several orders of magnitude. The exponential model consistently underestimated
Nf if the entire loading history to failure was not used in the calculation. This suggested
the exponential model was not a good technique for extrapolation from a number of load-
ing cycles less than Nf. Extrapolations using the single-stage Weibull model were gener-
ally distributed around the line of equality and provided the best prediction. In one case,
the three-stage Weibull model provided a more accurate prediction of Nf and it always
provided a better fit to the stiffness versus loading cycle data.
For each mixture, log-log regression plots were created using the data for samples that
failed in less than 50 million cycles. Prediction limits were determined for lower strain levels.
The extrapolated Nf for samples that did not fail within 50 million cycles indicated fatigue
lives that were longer than those indicated by the prediction limits from the regression of sam-
ples tested at “normal” strain levels. This deviation indicates the existence of an endurance
limit for HMA.
A standard practice was developed to predict the endurance limit based on tests con-
ducted at normal strain levels (above the endurance limit) based on statistical prediction
limits. The estimate of the endurance limit is the one-sided, 95% confidence lower predic-
tion limit for the strain level that produces a fatigue life of 50 million cycles. Confirmation
tests at the predicted strain level are recommended. The predicted endurance limits deter-
mined using this methodology ranged from 75 to 200 ms. Stiffer binders tended to produce
higher endurance limits. Optimum plus binder content combined with lower sample air
voids produced a slight increase in the endurance limit.
A mini round-robin was conducted to assess the variability of beam fatigue tests at normal
strain levels and extrapolations at low strain levels. Nf varies significantly depending on the
strain level at which the samples are tested. The log base 10 transformation of Nf was used in
the round robin analyses to produce a relatively constant variability across the range of strain
levels evaluated. On a log basis for normal strain fatigue tests, the repeatability (within-lab)
standard deviation was determined to be 0.248 and the reproducibility (between-lab) stan-
dard deviation was determined to be 0.318. This results in within- and between-lab coeffi-
cients of variation of 5.4% and 6.8%, respectively. The single-stage Weibull extrapolations
had the lowest variability within and between labs.
Uniaxial tension tests were performed on cylindrical samples cored and sawed from speci-
mens compacted in the Superpave gyratory compactor. Testing included complex modulus,
monotonic, and fatigue tests in uniaxial tension. Analysis of these data was done using vis-
coelastic and continuum damage mechanics principles to identify the fatigue endurance limit
 3

of the PG 67-22 optimum, PG 67-22 optimum plus, PG 76-22 optimum, and PG 76-22 opti-
mum plus mixtures. Frequency sweeps at various temperatures were run to measure the
complex dynamic modulus of each mix. The dynamic modulus and phase angle master
curves were constructed from these data and the relaxation master curve obtained through
linear viscoelastic conversion. The damage characteristic curve of each mix was obtained by
running uniaxial monotonic tests to failure or by running constant amplitude fatigue tests
to failure. The characteristic damage curve was used to predict the number of cycles to fail-
ure at different strain amplitudes to determine the fatigue endurance limit of the mixture.
The estimated endurance limits from this method varied depending on whether the gener-
alized power law or exponential model was used to fit the data in the characteristic curve.
The estimated endurance limits for the generalized power law ranged from 164 to 261 ms,
while those for the exponential model ranged from 47 to 96 ms. The relative rankings of the
binders evaluated in Phase I appeared to be reversed from the beam fatigue tests.
A second methodology was employed where fatigue tests with increasing strain amplitude
were run in uniaxial tension to directly identify the fatigue endurance limit of the mixtures.
The stress versus strain loading history forms a hysteresis loop, the area of which is the dis-
sipated energy per cycle. The use of pseudo strain instead of engineering strain in constitu-
tive analysis removes the hysteretic effect of viscoelasticity. If the induced strain levels are
low enough not to induce damage (e.g., below the endurance limit) then the hysteresis loop
collapses since the dissipated energy is only due to the viscoelastic response of the material.
The estimated endurance limit using this methodology ranged from 115 to 250 ms. However,
differences were observed between the strain applied through the cross head and the induced
strain measured on the samples. These differences made it difficult to precisely increment the
induced strain in order to accurately determine the endurance limit.
Analyses were performed of LTPP data for indications of the endurance limit. A 1995
survivability analysis of data from the general pavement studies (GPS) experiments known
as GPS-1 and GPS-2 indicated an endurance limit of approximately 65 ms. However,
when the analysis was updated, and LTPP special pavement studies (SPS) data were added
from SPS-1, no endurance limit was indicated. During the period between the original and
updated analyses, LTPP changed the definition of longitudinal cracking in the wheel path.
This may have resulted in cracks previously identified as top-down cracking to be reclassi-
fied as bottom-up fatigue cracks. Forensic investigations on the thicker pavements in the
study, exhibiting cracking, are required to identify the type of cracking. Still, separation
was indicated between the survivability of pavements with tensile strains at the bottom of
the HMA layer less than 150 ms and those equal to or greater than 150 ms. This may indi-
cate that neglecting top-down cracking, the endurance limit may be less than 150 ms. Addi-
tional cracking data were presented, which demonstrates the improved fatigue performance
of mixes containing polymer modified binders, similar to that indicated in the beam fatigue
testing program.
Analyses were conducted on data from the structural sections of the 2003 NCAT Test
Track to estimate the shift factor between laboratory and field performance. Three of the
eight structural sections exhibited fatigue cracking in excess of 40% of the wheel-path area
or approximately 20% of the total lane area during the test track loading cycle (application
of 10 million equivalent single-axle loads [ESALs]). Forensic investigations indicated that
the cracking in one of these sections resulted from debonding of the HMA layers. A fourth
section had limited cracking and cumulative damage was estimated at 0.7 at the end of the
loading cycles.
All of the structural sections were instrumented to measure strain at the bottom of the
HMA layer. The measured strain data, pavement temperatures, and axle repetitions were
used with the fatigue transfer functions developed as part of the beam fatigue testing to
calculate incremental damage on an hourly basis. Shift factors between 4.2 and 75.8 were
calculated between laboratory and field performance. The shift factor of 4.2 was determined
4

for a polymer modified section with a total asphalt thickness of only 4.8 in. Similar calcula-
tions were performed with PerRoad and the Mechanistic-Empirical Pavement Design Guide
(MEPDG). The shift factors determined using PerRoad ranged from 6.7 to 45.0. Damage in
the MEPDG was calculated using the nationally calibrated fatigue model, not the transfer
function developed as part of the beam fatigue testing. Shift factors for two sections ana-
lyzed with the MEPDG ranged from 0.7 to 1.0. The predicted cracking based on 90% reli-
ability was much higher. Overall, it was concluded that the assumption of a shift factor of
10 between laboratory and field performance was reasonable.
Analyses were conducted using the MEPDG and PerRoad to determine their sensitivity
to the measured fatigue endurance limit. Analyses were conducted with both the NCAT Test
Track traffic (very limited range of axle weights) and the MEPDG default truck traffic clas-
sification No. 1 for principal arterials. The perpetual pavement thickness determined with
both programs was sensitive to the measured endurance, a 50 ms change in the endurance
limit resulted in approximately a 7- to 8-in. change in pavement thickness for the MEPDG
or a 4-in. change in pavement thickness for PerRoad.
Using the endurance limits predicted from beam fatigue tests conducted as part of the study
and the default traffic classification, the perpetual pavement thickness determined with Per-
Road was approximately the same as the 20- and 40-year conventional (no endurance limit)
MEPDG or 1993 AASHTO Pavement Design Guide pavement thicknesses. The MEPDG
perpetual thickness was approximately 50% thicker. The overall conclusions at the end of a
20- or 40-year period were significantly different. With the conventional designs, the pave-
ments would have failed in bottom-up fatigue with cracking over 20% of the lane area at
90% reliability while no cracking would be expected if the endurance limit was considered.
The implementation of the endurance limit as a single value for a given mix appears rea-
sonable for M-E design programs that use equivalent temperature concepts, such as PerRoad.
For the MEPDG, however, temperature as a function of depth is calculated on an hourly basis
throughout the design life or analysis period. This can result in higher peak temperatures,
with corresponding higher strains at the bottom of the asphalt layer, than equivalent temper-
ature methods.
Research conducted outside this study indicates that the endurance limit based on lab-
oratory testing varies as a function of temperature. One interpretation of this may be that
a mix’s ability to heal and therefore its fatigue “capacity” are greater at higher temperatures.
A field study of data from the NCAT Test Track indicates pavements can withstand a distri-
bution of strains without incurring cumulative damage. Both of these concepts have merit
and both are recommended for future research aimed at better implementing the endurance
limit in the MEPDG.
 5

CHAPTER 1

Introduction and Research Approach

Introduction mix properties that will improve the resistance to bottom-up

cracking will also improve the resistance to top-down crack-
Fatigue cracking has been an issue in the design and per- ing. Whether it is top-down cracking or bottom-up cracking,
formance of hot mix asphalt (HMA) pavements ever since the likely cause of fatigue distress is tensile strains in the HMA
hot mix asphalt pavements began to be used. It has long been at critical points in the pavement structure.
understood that pavement structures that are too thin fail in There is now a much wider use of polymer modified asphalts
fatigue under repeated loads. This type of structural failure than in past years. There have been very few fatigue studies on
results from fatigue of the HMA mixtures and results in the mixtures containing modified asphalts, and it is not clear how
formation of alligator cracking. much improvement, if any, can be obtained from modified
There are a number of properties in the mixture that can mixtures.
be adjusted to improve the resistance to fatigue but the most Typically, fatigue data are analyzed by plotting the number
important property by far is the thickness of the HMA and its of cycles to failure versus the maximum tensile strain or stress
resulting effect on the tensile strain at the bottom of the HMA. in the HMA specimens. It has been shown that plotting these
It is this strain at the bottom of the asphalt that is the primary data on a log-log (log of cycles to failure versus log of applied
cause of the bottom-up fatigue cracking. Ultimately, a method stress or strain) plot will provide approximately a straight line.
is needed to determine the expected strain at the bottom of the It is very time consuming to conduct fatigue tests at very high
asphalt layer more accurately and to determine the effect of numbers of cycles so, generally, researchers have applied fairly
this expected strain on the fatigue resistance of the mixture. high stress or strain values so that the number of cycles to fail-
Determination of fatigue life based on laboratory tests gener- ure will not be so high as to provide excessive test time. It is
ally does not provide an accurate predictor of what is observed generally believed that approximately 10 specimens must be
in the field and a shift factor must be applied. There are many tested for each mixture being evaluated to provide a suitable
reasons that probably lead to the need for a shift factor, includ- relationship between applied stress or strain and number of
ing: aging, rest periods, healing, densification under traffic, cycles to failure.
temperature fluctuations, concept of constant stress versus Since most of the testing has been done at higher stress
constant strain, simulation of field compaction versus labora- or strain levels, there has not been much work to look at the
tory compaction, and property changes due to other environ- expected performance at lower stress or strain levels even
mental conditions such as aging of the binder and moisture though these lower stress levels are typical in the field. Gener-
damage. The shift factor can be significant (e.g.,15 to 20 times ally, a best-fit line is determined for the data at these relatively
or more of the laboratory estimate of the fatigue life). high strain or stress levels and the performance of mixtures
Much research was performed in the 1960s and 1970s to at lower stress or strain levels is extrapolated from the data.
study fatigue of HMA. Until recent years it was believed that Hence, there is very little understanding of the actual per-
fatigue cracking always started at the bottom of the HMA formance that would be expected at these lower levels.
layers and propagated up through the HMA layers, eventually Some work has indicated that there might be a level of stress
reaching the surface. However, recently there is evidence that or strain below which no damage occurs to the test specimen.
some projects have exhibited fatigue failure due to top-down The stress or strain level below which no fatigue damage orig-
cracking. Generally, it is believed that any modification to the inating from the bottom of the pavement structure occurs has
6

been termed fatigue endurance limit. If a pavement is designed Objectives

and constructed so that under repeated traffic loads no dam-
age occurs, then that pavement should last indefinitely without The objectives of this study are to
a structural failure. This pavement will still require overlays on
some regular basis to maintain the surface in good operating 1. Test the hypothesis that there is an endurance limit in the
condition, but the pavement structure should provide a very fatigue behavior of HMA mixtures and measure its value
long life. for a representative range of HMA mixtures.
2. Recommend a procedure to incorporate the effects of
the endurance limit into mechanistic pavement design
Research Problem Statement methods.
Work is needed to determine if there is an endurance limit
for HMA and, if so, how this information should be used in Scope
pavement design. One would not want to design all pave-
ments, regardless of traffic level, so that the endurance limit This study included a literature review, laboratory testing,
is not exceeded since this would result in the same design and analysis of field data. A mix design, used to construct struc-
thickness regardless of traffic level. For very high traffic lev- tural sections at the 2003 National Center for Asphalt Technol-
els it might be desirable to design the pavement so that the ogy (NCAT) Test Track, was replicated. Four versions of the
endurance limit is not exceeded, but for low traffic levels it mix design were evaluated encompassing the effect of increased
is likely that this approach would be too expensive. How- asphalt binder content and polymer modification. The mix-
ever, a similar concept could be employed for pavements tures with higher asphalt binder contents were prepared at
subject to very low traffic levels to ensure a long life. As always, higher densities to simulate the improvement in compaction
the pavement should be designed to provide the lowest life- expected in the field. Beam fatigue and uniaxial tension test-
cycle costs. ing were conducted to determine fatigue life. Beam fatigue
Hence, there is a need first of all to determine if there is testing was conducted at progressively lower strain levels until
an endurance limit for HMA. Once the endurance limit is a fatigue life in excess of 50 million cycles was achieved. Data
determined it is important to determine how this endurance from the Long-Term Pavement Performance (LTPP) study
limit would fit into the new mechanistic pavement design were analyzed to determine if they supported the endurance
procedures. limit concept.
 7

CHAPTER 2

State of Practice

A review of the existing literature was conducted to answer pavement thickness, greater than that required to keep
the following specific questions: strains below the endurance limit, would not provide addi-
tional life. This concept has significant design and eco-
1. What is an endurance limit? nomic implications.
2. What field or laboratory studies support the existence of The concept of the endurance limit was originally devel-
an endurance limit? oped for metals (4, 5). Barret et al. (5) describe the endurance
3. What HMA material factors affect fatigue life and conse- limit for metals as being a stress below which, for uncracked
quently might affect the endurance limit for different materials, the plot of stress versus cycles to failure becomes
materials? essentially horizontal and fatigue does not occur. Figure 2.1
4. What are the best methods of measuring fatigue life in the illustrates the theoretical concept of the endurance limit, as it
laboratory? would be applied to HMA.
5. What analysis methods should be used to analyze fatigue The concept of the endurance limit was first implemented
data in order to identify the endurance limit for a given for paving materials by the Portland Cement Association.
mixture? An examination of fatigue tests conducted by various
researchers on Portland cement concrete beams, discussed
A tremendous volume of literature was identified related to in Baladi and Snyder (6) and Huang (7), indicated that if the
the factors affecting fatigue life, methods of measuring fatigue stress ratio is kept below 0.45, the concrete will have an
life, and analyzing fatigue data. Three very good summaries of essentially infinite fatigue life. The stress ratio was defined
this literature were identified, one by Epps and Monismith (1) as the ratio of stress induced in the concrete pavement to the
produced in the early 1970s and two additional summaries concrete’s modulus of rupture. A maximum of 10 million
produced as part of the Strategic Highway Research Program cycles to failure was used for the majority of this testing (one
(SHRP) (2, 3). Therefore, an attempt was not made to sum- sample was tested to approximately 20 million cycles) (Fig-
marize all of the references related to factors affecting fatigue ure 2.2).
life and fatigue measurement. Monismith and McLean (8) first proposed an endurance
limit of 70 micro-strain (ms) for asphalt pavements. It was
observed that the log-log relationship between strain and
Arguments for the Existence of
bending cycles converged below 70 ms at approximately
the Endurance Limit
5 million cycles. Maupin and Freeman (9) noted a similar
Pavements have been designed primarily to resist rutting convergence. Using low-strain design principles, Monismith
of the subgrade and bottom-up fatigue cracking. In classi- and McLean (8) designed a pavement structure that increased
cal pavement design, as design load applications increase, the fatigue life of the pavement from 12 to approximately
pavement thickness must also increase. There is a growing 19-plus years.
belief that for thick pavements, bottom-up fatigue cracking In the field, Nunn (10) in the United Kingdom (UK) and
does not occur. The concept of an endurance limit has been Nishizawa et al. (11) in Japan proposed concepts for long-life
developed, representing a strain level resulting from a com- pavements for which classical bottom-up fatigue cracking
bination of HMA stiffness and thickness, below which would not occur. Nunn (10) defines long-life pavements as
bottom-up cracking will not initiate. Therefore, additional those that last at least 40 years without structural strengthen-
8

1000

Target Micro Strain

100

Endurance Limit

10

1 1

10

1,

1,

10
10

10

10

10

1,

10
00

00

00
0,
0

,0

0,

,0

,0
0

0,

00
0

00

00

0,

00
0

00

00
0,
0

,0

,0
0

00
00

0,

00
00
0

,0
0

00
Number of Cycles to 50% Stiffness

Figure 2.1. Idealized concept of the endurance limit.

ing. The UK’s pavement design system was based on experi- • Surface initiated cracking was common in high-traffic
mental roads that had carried up to 20 million standard axles. pavements, but there was little evidence of bottom-up
When this study was conducted, these relationships were being fatigue. Surface initiated cracks tended to stop at a depth of
extrapolated to more than 200 million standard axles. Nunn 4 in. (100 mm).
(10) evaluated the most heavily traveled pavements in the UK, • It was observed that the stiffness of thick pavements was
most of which had carried in excess of 100 million standard increasing with time, most likely due to binder aging. This
axles to evaluate the then current design system. Nunn (10) would not tend to occur if the pavement was weakening due
concluded the following: to accumulated damage.
• A minimum thickness for a long-life pavement was recom-
• For pavements in excess of 7.1 in. (180 mm) thick, rutting mended as 7.9 in. with a maximum thickness of 15.4 in. This
tended to occur in the HMA layers. range is based on a variety of factors such as binder stiffness.

Source: Huang, Yang, H., Pavement Analysis and Design, 2nd ed., © 2004, p. 317. Reprinted by
permission of Pearson Education, Inc. Upper Saddle River, NJ.

Figure 2.2. Illustration of endurance limit for concrete

pavements (7).
 9

Nishizawa et al. (11) reported an endurance limit of 200 ms void levels were targeted, resulting in relationships between
based on the analysis of in-service pavements in Japan. Simi- fatigue life and both air void content and asphalt content.
larly, strain levels at the bottom of the asphalt layer of between Voids filled with binder (VFB) have been a typical parame-
96 and 158 ms were calculated based on back-calculated stiff- ter used in fatigue life prediction equations. Harvey and Tsai
ness data from the falling-weight deflectometer for a long-life (15) caution against the use of VFB since various combina-
pavement in Kansas (12). Others (13, 14) report similar find- tions of air voids and asphalt content will produce the same
ings, particularly, the absence of bottom-up fatigue cracking VFB. Maupin and Freeman (9) note that there was little
in thick pavements and the common occurrence of top-down increase in fatigue life resulting from an increase of 0.5%
cracking. binder, but significant increases were seen with an increase
of 1.0% binder.
Factors Affecting Fatigue Life
Strategies to Produce
A significant amount of fatigue research was conducted in
Long-Life Pavements
the 1960s and 1970s. Epps and Monismith (1) provide a
summary of the effects resulting from binder stiffness, A number of strategies have been put forth to promote the
asphalt content, aggregate type, aggregate gradation, and air likelihood of constructing a long-life pavement, including:
void content. Table 2.1 indicates the relative affects of these polymer modification, rich bottom layers, and high-modulus
components. The authors conclude that binder stiffness and asphalt bases.
air void content have a larger influence on fatigue life than
aggregate type and gradation. The SHRP A-404 research (3)
Polymer Modification
noted that angular aggregates tended to produce both
stiffer mixes and longer fatigue lives. Harvey and Tsai (15) Fatigue testing and analyses of asphalt mixtures made
evaluated the effect of air voids and asphalt content on with modified asphalt binders has been performed in sev-
fatigue life. In most previous evaluations, a constant com- eral studies. In 1988, Goodrich presented an early study on the
paction effort was used to produce samples. This results in fatigue performance of polymer modified mixes (17). In this
air voids being highly correlated with the asphalt content of study, three unmodified asphalt binders with different tem-
a given mix in which case there is little relationship between perature susceptibilities and two modified asphalt binders—
asphalt content and fatigue life. In this instance, specific air produced using one base asphalt and two levels of modification

Table 2.1. Factors affecting the stiffness and fatigue behavior of hot mix
asphalt (16 in 1).

Factor Change in Effect of Change in Factor

Factor On Stiffness On Fatigue On Fatigue
Life in Life in
Controlled Controlled
Stress Mode of Strain Mode
Test of Test
Asphalt Penetration decrease increase increase decrease
Asphalt Content increase increasea increasea increaseb
Aggregate Type increase in increase increase decrease
rough
texture and
angularity
Aggregate Gradation open to increase increase decreased
dense
Air Void Content decrease increase increase increased
c
Temperature decrease increase increase decrease
Notes:
a
Reaches optimum level above that required by stability considerations.
b
No significant amount of data; conflicting considerations of increase in stiffness and
reduction of strain in asphalt make this speculative.
c
Approaches upper limit at temperature below freezing.
d
No significant amount of data.
10

(not identified)—were evaluated using mixture fatigue tests. Testing was conducted on mixtures using 10 Hz sinusoidal
The intent was to correlate binder properties with mixture loading with a peak-to-peak strain of 800 ms. The test tem-
fatigue performance. Laboratory testing was conducted using perature was selected for each mixture as the temperature
flexural beam fatigue at 25°C and 1.67 Hz loading frequency where the intermediate temperature requirement was met
in controlled stress mode. Tests were conducted at an initial (G*sin δ = 5000 kPa). So, unlike other studies in which the
strain level of 400 ms, and findings indicated that the fatigue temperature was fixed (usually at 20°C), the viscous compo-
lives of the two modified asphalt binders were an order of nent of the shear modulus was fixed.
magnitude greater than the fatigue life of one of the unmod- A brief examination of the data in the report indicates that
ified asphalt binders (produced from the same source as the the fatigue lives of the modified mixtures could be signifi-
base asphalt used to create the modified asphalt binders). The cantly different. Mixtures made with a PG 82-22 asphalt
base asphalt properties appear to be important in the perform- binder produced using a radial styrene-butadiene-styrene
ance of the modified asphalt binders. Mixtures made with one (SBS) modifier had fatigue lives at 24°C that were two to five
unmodified asphalt binder with low temperature susceptibil- times the fatigue lives (tested at 32°C) of mixtures made with
ity, had approximately two to three times the fatigue life of the an unmodified (oxidized) PG 82-22 asphalt binder. Since the
polymer modified mixtures. G*sin δ value was the same for each of these mixtures at their
During the Strategic Highway Research Program (SHRP), respective test temperatures, the researchers considered the
the A-003A contractor evaluated the use of the flexural beam intermediate temperature criterion in the PG binder specifi-
fatigue test as a mixture performance test for fatigue. The cation to be inadequate for assessing the fatigue characteris-
modified asphalt mixtures experiment (MAME), described tics of asphalt binders.
in SHRP Report A-404, was performed to determine if the A temperature equivalency experiment conducted during
fatigue characteristics of modified mixtures could be evalu- SHRP (18) indicated a strong relationship between temper-
ated using the flexural beam fatigue test (18). Asphalt mixtures ature and fatigue life at a given strain level, with fatigue life
were made using one aggregate source, three asphalt binders decreasing as the temperature decreased. Thus, it could be
(AAF-1, AAG-1, and AAK-1), and three modifiers (identified hypothesized that the difference in fatigue life between the
as M-405, M-415, and M-416). Test results indicated that SBS-radial modified PG 82-22 mixtures and the oxidized
the addition of modifier M-405 to each of the three asphalt PG 82-22 mixtures would be greater if the test temperatures
binders decreased the fatigue life compared to the unmod- had been equal.
ified asphalt mixes. The addition of modifiers M-415 and Monismith et al. (20) reported on the development of the
M-416 had a negative effect on the fatigue life of mixes made design and specifications for the California I-710 rehabilita-
with AAG-1, but substantially increased the fatigue life (by tion. In the study, AR-8000 (roughly equivalent to a PG 64-10
approximately three to five times) of mixtures made with or PG 64-16) and PBA-6a (PG 64-40) asphalt binders were
AAK-1 compared to the unmodified mixtures. A validation used to prepare mixtures for testing. When tested at 20°C
study was performed using slab wheel track tests on mixtures using the procedure described in AASHTO T321, the mea-
made with AAG-1 and the three modifiers. Although the sured fatigue life of the PBA-6a mixtures was approximately
results were similar for the M-405 modifier (decrease in an order of magnitude (10 times) greater than the fatigue
fatigue life), the M-415 and M-416 modifiers resulted in an life of the AR-8000 mixtures. This relationship seemed to be
increase in fatigue life from the slab wheel track test. This was affected by the applied strain resulting in an increased dif-
contrary to the findings of the flexural beam fatigue tests. ference between the two sets of fatigue lives at higher strain
Shortly after the implementation of the Superpave levels.
performance-graded asphalt binder tests and specification, Lee et al. (21) reported on a laboratory evaluation of the
users recognized that the properties of modified asphalt effects of aggregate gradation and binder type on mixture
binders may not be characterized properly using the Super- fatigue life. Using uniaxial tension fatigue test results (con-
pave binder tests. This led to the funding of NCHRP Project ducted at 25°C) and a viscoelastic fatigue model, the authors
9-10, “Superpave Protocols for Modified Asphalt Binders,” calculated that mixtures made with an SBS-modified PG 76-22
in 1996. The research (19) was conducted by the University asphalt binder had 10 times greater fatigue life than mixtures
of Wisconsin-Madison, Asphalt Institute, and NCAT, under made with an unmodified asphalt binder, regardless of aggre-
the direction of Hussain Bahia. gate gradation.
The NCHRP 9-10 research evaluated the effectiveness of the Research performed in 2002 by the Asphalt Institute for
intermediate temperature stiffness requirement (G*sin δ ≤ the Asphalt Pavement Alliance evaluated the possibility of a
5000 kPa) by performing flexural beam fatigue tests on four fatigue endurance limit by testing mixtures made with two
aggregate structures (gravel and limestone, coarse and fine asphalt binders (PG 64-22 and PG 76-22) and two asphalt
gradation) using nine different modified asphalt binders. binder contents at various strain levels. The results, shown in
 11

1.E+09

1.E+08

Cycles to Failure
1.E+07 64-22 Opt
64-22 Opt+
1.E+06
76-22 Opt
1.E+05 76-22 Opt+

1.E+04

1.E+03
10 100 1,000
Strain (x10-6)

Figure 2.3. Fatigue life comparison for modified and unmodified

mixes (22).

Figure 2.3, indicate that the mixtures made with the PG 76-22 reported in SHRP Report A-404, one asphalt-aggregate
asphalt binder have approximately an order of magnitude (i.e., mixture (RB aggregate and AAG-1 asphalt binder) was used
10 times) greater fatigue life than the mixtures made with the to prepare specimens at three different levels of air voids
PG 64-22 asphalt binder. and two different levels of asphalt binder content (4.5 and
Von Quintus (23) conducted a study to quantify the effects 6.0%). The effect of asphalt binder content on fatigue life was
of polymer modified asphalts. Based on a literature review, Von a primary focus of this study since the 8 × 2 expanded test
Quintus reported that PMA mixtures generally last about 25% program experiment did not evaluate asphalt binder content
longer than conventional mixtures. Some premature failures as a variable. A separate 2 × 2 pilot test program used two
caused concern. However, most of the failures were “ . . . found asphalt binder contents defined as optimum and high—with
to occur prior to the adoption of the Performance Graded (PG) the high asphalt binder content established as 0.6% higher
binder specification and can be traced back to inferior con- than the optimum asphalt binder content. For the 2 × 2
struction (for example, high air voids), inferior materials, pilot test program, a statistical analysis of results from flex-
and/or inadequate design thickness.” The study also notes ural beam fatigue (controlled stress and controlled strain) and
“One of the more important findings from the recent field other tests indicated that asphalt content did not signifi-
experiments is that many of the PMA pavements are not cantly affect fatigue life (18).
exhibiting fatigue cracking or have less load-related crack- The results of the mix design fatigue experiment indicate
ing than the control sections (unmodified mixtures).” that asphalt binder content significantly affected flexural stiff-
In summary, there appears to be significant historical data ness (decreasing by 8% as asphalt binder content increased)
indicating that the laboratory fatigue performance of modi- and fatigue life—increasing the fatigue life by 67% as the
fied asphalt mixtures is greater than mixtures made with asphalt binder content increased from 4.5% to 6%. As in other
unmodified asphalt binders. In some reported cases, modi- experiments, increasing the air void content resulted in a
fied asphalt mixtures have exhibited an order of magnitude decrease in flexural stiffness of 33% and a decrease in fatigue
greater fatigue life compared to unmodified asphalt mix- life by 45% as air voids increased from 4% to 8% (18).
tures. The fatigue characteristics appear to be dependent on Harvey and Tsai (25) conducted a study on the effects of
the base asphalt binder used for modification. asphalt content and air void content on mixture fatigue and
stiffness. Samples were produced at five asphalt contents:
4.0%, 4.5%, 5.0%, 5.5%, and 6.0% by weight of aggregate and
Rich Bottom Layers
three ranges of air voids: 1% to 3%, 4% to 6%, and 7% to 9%.
The concept of a rich bottom layer originated from the In this experiment, a constant compaction effort was not
Australian experience (24) and was explored during SHRP used, so air voids were independent of asphalt content. Con-
experiments. Two potential benefits are created through stant strain tests were performed at two strain levels: 300 and
the use of a rich bottom layer: increased asphalt binder con- 150 ms. Analysis of the data indicated that higher asphalt con-
tent and decreased air voids in the bottom layer (as a result of tent and lower air voids resulted in longer fatigue lives. Lower
easier compaction created by the additional asphalt binder). asphalt content and lower air voids resulted in higher initial
Several known literature sources confirm the rationale stiffness. Instead of using stiffness in mixture fatigue life pre-
for these concepts. In the mix design fatigue experiment diction models, the authors recommended evaluating the
12

laboratory fatigue life and then predicting pavement perform- Testing was conducted on 9.5-mm and 12.5-mm nominal
ance by incorporating the effect of mixture stiffness on the maximum aggregate size (NMAS) mixtures to examine the
predicted pavement strains resulting from layered elastic effect of increasing asphalt binder content. Flexural beam
analysis. Using this methodology, the authors show the poten- fatigue tests conducted at 600 ms indicated a slight increase
tial benefits of a rich bottom layer. It is also important to rec- in fatigue life with the 0.5% higher asphalt content. A sub-
ognize that lower in-place air voids increase stiffness while stantial increase in fatigue life was noted with 1.0% higher
resulting in longer fatigue lives, opposite conventional wis- asphalt content.
dom that suggests that stiffer mixes have shorter fatigue lives.
Monismith et al. (20) reported on the development of
High Modulus Base
the design and specifications for the California I-710 reha-
bilitation. In their study, AR-8000 (roughly equivalent to a Europeans have used stiff binders to produce high modu-
PG 64-10 or PG 64-16) and PBA-6a (PG 64-40) asphalt lus base layers (10, 14, 28, 29). Corte (28) reported on the use
binders were used to prepare mixtures for testing at two com- of high modulus asphalt mixtures in France. The first use of
binations of asphalt binder content and air void content. high-modulus asphalt concrete (HMAC) occurred around
Mixes were prepared with 0.5% higher asphalt binder content 1980. Initially, these mixtures were used for strengthening or
and 3% lower air voids. When tested at 20°C using the proce- rehabilitation where pavement thickness was constrained (for
dure described in AASHTO T321, the measured fatigue life of instance by bridge clearance). The use increased in 1985. It
the mixes with 0.5% higher asphalt binder content (and lower was found that locally available weak aggregates could be
air void contents) was approximately two times the fatigue life used with stiff binders. These mixtures were designed with
of mixes prepared at lower asphalt binder content. relatively high asphalt binder content and low voids (less
Anderson and Bentsen (26) reported on a study evaluating than 6%). Constant strain fatigue tests indicate that HMAC
the influence of voids in mineral aggregate (VMA) on mix- mixes are more fatigue resistant than conventional base
ture performance. Since VMA is related to asphalt binder mixtures. This is believed to be due to the higher asphalt con-
content, mixtures with high VMA had asphalt binder con- tent and lower voids found in the HMAC mixtures. Corte’s
tents that were approximately 1.0% higher than the low VMA findings match the findings in similar studies where lower
mixtures. Flexural beam fatigue testing conducted at 20°C air voids increase stiffness, but also appear to increase fatigue
and 500 ms indicated that the mixes with the higher asphalt life (20, 21). The stiffness of these layers reduce the strain at
binder contents had two times greater laboratory fatigue life the bottom of the asphalt layer using less thickness than
than the low asphalt binder content mixes. conventional asphalts. Cracking can be a problem with these
Harvey et al. (24) reported on California’s experiences with mixes. Corte (28) discusses binder tests to minimize the like-
the design and construction of long-life asphalt pavements. lihood of cracking.
The authors reported that most full-depth asphalt long-life
designs will include a stiff, fatigue-resistant bottom layer. This
Laboratory Fatigue Tests
layer, termed a rich bottom layer, is designed to have a very low
and Analysis Methods
air void content (approximately 0% to 3%). Stiffness of the
layer is a consideration since it is intended to reduce the over- The SHRP A003-A project (2, 3) evaluated seven methods
all thickness of the HMA layers. The low air void content also of measuring laboratory fatigue life. Repeated load flexure
reduces permeability and improves moisture resistance. The and direct tension tests received the highest rankings. A
benefit of rich bottom layers is maximized with a thickness methodology was developed to evaluate fatigue life using
range of 50 mm to 75 mm. Illinois has also adopted this flexural beam fatigue tests conducted in constant strain mode
concept (14). at 10 Hz. Thin pavements are generally subjected to a mode
Generally the stiffness of a mix can be increased with of loading best represented by constant strain. Thick pave-
increased compaction. Further, increased compaction gen- ments are generally represented by a mode of loading most
erally increases fatigue life at the same strain level. Thus, closely represented by constant stress. However, the SHRP
increased compaction specifications for lower lifts result in A003-A researchers recommended constant strain tests for
both lower strains due to increased stiffness and also increased all pavement loading conditions. This recommendation was
fatigue life for a given strain level, producing a more eco- based upon the fact that if fatigue evaluations are made in the
nomical pavement. The unbound layers must be sufficiently context of the pavement structure (e.g. by calculating expected
stiff to allow a high degree of compaction in the bottom HMA strains at the bottom of a given pavement structure), then
layers. constant stress and constant strain tests give similar rank-
A study conducted by Maupin (27) examined the impact ings (3, 25, 30). AASHTO T321 is the current standard for
of asphalt content on durability of Virginia surface mixtures. beam fatigue tests.
 13

The direct tension test was eliminated early in the SHRP fatigue lives in excess of 11 million cycles (37). Tests were also
A003-A research due to difficulties aligning and gripping the conducted that indicated that periodic overloads (loads cre-
specimens (2). However, research under the direction of Kim ating strain levels in excess of the endurance limit) would not
appears to have improved this technique (31–33). Previous substantially reduce the fatigue life where the majority of
researchers noted that only a portion of a sample’s dissipated the load cycles were less than the endurance limit (37). The
energy is most likely causing damage (34, 35). Daniel and Asphalt Institute conducted a study to identify the endurance
Kim (32) developed a method using a characteristic curve limit for the Asphalt Pavement Alliance (38, 39). Beam samples
from which fatigue life can be predicted rapidly from monot- were tested to a maximum of 4 million cycles. Extrapolations
onic uniaxial tension tests. A characteristic curve is generated support the presence of an endurance limit between 70 and
by modeling viscoelastic material behavior using Schapery’s 100 ms.
correspondence principle, continuum damage mechanics, Shen and Carpenter (40) developed a new method for
and work potential theory. This method may be used to more determining/predicting the endurance limit. Their research
rapidly determine the endurance limit of a mixture. Uniaxial indicated a linear relationship between the log of the plateau
tension testing would not require the production of an HMA value of the ratio of dissipated energy curve (previously called
beam and instead would use a sample more closely related to DER) and the log of cycles to 50% initial stiffness for both
those being contemplated for simple performance tests related normal and low (below the endurance limit) strain levels.
to rutting. Further, they believe this methodology can be used to ex-
Maupin and Freeman (9) evaluated five simple tests to pre- trapolate the failure point for tests conducted in as little as
dict fatigue life. Indirect tensile test results were found to be 500,000 cycles. A tentative plateau value of 8.57E-9 was iden-
correlated to the coefficients used in standard equations to tified as indicating the endurance limit. This appears to be a
predict fatigue life in both constant stress and constant strain promising technique for analyzing the endurance limit.
modes of testing. Von Quintus has indicated that a long-life
pavement may be designed where the strain at the bottom
Modeling Fatigue and Relationship
of the asphalt layer is less than 1.0% of the indirect tensile
to Field Performance
strength failure strain. Thus, indirect tensile strength may
also provide a rapid screening tool to evaluate the endurance The NCHRP 1-37A Design Guide has instituted an
limit of a given mixture. enhanced version of the Asphalt Institute Model for fatigue
life (41). The enhancements were developed to better predict
the performance of thin pavements in a constant strain mode
Laboratory Studies to Quantify
of loading. With this model, pavement thickness will contin-
the Endurance Limit
ually increase with increasing design traffic loads. The litera-
As interest in long-life pavements grew, laboratory studies ture indicates a number of factors that must be accounted for
began to try to validate the existence of the endurance limit if an alternate approach for fatigue life determination, which
and develop methods of determining it for a given mixture. incorporates the endurance limit, is developed.
Ghuzlan and Carpenter (34) proposed the use of the dissipated Previous studies indicate a coefficient of variation of
energy ratio (DER) to define the existence of an endurance approximately 40% for beam fatigue tests (3, 30). This vari-
limit. The dissipated energy for a given fatigue cycle is calcu- ability must be accounted for when making fatigue life pre-
lated as the area of the stress-strain hysteresis loop (3, 35). DER dictions and extrapolations. Several methods for accounting
is simply the ratio of dissipated energy from one cycle to the for this variability have been proposed (30, 42–44). For
next. During the course of a fatigue test, three regions of the fatigue-life predictions below the endurance limit, varia-
DER curve versus loading cycles may be identified: an ini- bility of design traffic estimates should also be considered
tial downward trend, a plateau with a nearly constant energy (30, 43). However, if a truly infinite fatigue life is estimated
input, and a failure region where the dissipated energy rapidly by the endurance limit, design traffic reliability may be of
increases which occurs at approximately 40% of the initial lesser importance. Harvey et al. (43) developed a method
stiffness (34, 35). Carpenter et al. (36) conducted additional that incorporates the variability from laboratory fatigue
beam fatigue tests in the range of 70 to 100 ms with samples tests, materials production, and construction. Material pro-
being tested to between 38 and 46 million cycles. They con- duction variability includes asphalt content and in-place air
cluded that low strain testing in the range of 70 ms resulted in voids. Construction variation is represented by variation in
“extraordinarily long fatigue life.” Researchers at the Univer- pavement thickness and subgrade support. Monte Carlo sim-
sity of Illinois, under the direction of Carpenter, have con- ulation is used to estimate the combined affect of material
ducted a number of low strain beam fatigue tests that and construction variation. Savard et al. (44) report on a
indicated a break in fatigue life behavior for samples with French methodology to determine an acceptable strain limit
14

based on variability in the fatigue tests and subgrade sup- Finally, in-service pavements tend to have longer fatigue
port. This methodology could be used to shift the measured lives than those indicated by laboratory tests. Thus, a shift
endurance limit for a mixture from a 50% reliability of the factor is applied to laboratory fatigue life to predict field
laboratory test data to an acceptable level of reliability for fatigue life (41, 43–44, 46–47). The shift factor is believed
the constructed pavement. to account for such things as the affect of rest periods and
Fatigue tests are typically conducted at 20°C. However, healing. Shift factors may range from 10 to 100 (43). Leahy
pavement damage varies with temperature and the result- et al. (46) recommended a shift factor of 10 for up to 10%
ing changes in stiffness of the HMA layers. Methodologies fatigue cracking in the wheel path. Harvey et al. (43) cali-
have been developed to shift fatigue life results at 20°C to brated shift factors for California conditions. The shift fac-
an equivalent annual or monthly temperature (30, 43–45). tor calculated with their equation increases with decreasing
Overloads, or strain levels exceeding the endurance limit, are strain levels. A shift factor of 16.6 would be calculated for
most likely to occur either in the warmest summer months 70 ms. Pierce and Mahoney (47) note that Washington
(37) or when the spring thaw occurs, depending on environ- State uses shift factors between 4 and 10. Smaller shift fac-
mental conditions. tors are used for thicker pavements.
 15

CHAPTER 3

Research Plan

Introduction and a single steering axle (three axle groups). Strain traces
indicate that the tandem axle results in two distinct load
Based upon the review of the literature, a controlled labo- repetitions (49). Assuming that one is designing a perpetual
ratory experimental plan was developed. The experimental pavement for the tandem axle load, the steering axles would
plan was developed with the primary objective of testing the have a lower loading so, theoretically, in a perpetual pavement
hypothesis that there is an endurance limit for HMA mixtures. design they would do no damage to the pavement. Thus, each
As a secondary objective, some of the HMA material proper- Class 9 vehicle would provide four load repetitions to the pave-
ties that affect the endurance limit were investigated. ment for a maximum total of 1,317,504,000 axle load repeti-
A working definition of the endurance limit was devel- tions in a 40-year period. This represents a theoretical maxi-
oped as a framework for testing within the experimental plan. mum loading where every truck is fully loaded and the design
Although the endurance limit is defined as an essentially in- lane is at maximum capacity for 24 hours a day, 7 days a week,
finite fatigue life for metal alloys, testing for an infinite life is for a 40-year period. This loading condition would be expected
impractical. The endurance limit must be defined in practi- to be even more severe than a dedicated truck lane. A similar
cal, usable terms if it is to have meaning. For example, the methodology was used by Mahoney to calculate the maxi-
literature has defined 40 to 50 years as a reasonable lifetime to mum number of ESALs expected in a 40-year period (unpub-
be considered as a long lasting, or perpetual, pavement. Hence, lished data).
determining a strain level that results in 40 to 50 years (or In actual mixed traffic streams, the highest percentage of
even more) of pavement life is a very practical way to identify trucks tends to be about 50%, which would reduce the maxi-
the endurance limit. mum number of load repetitions to 940,473,600 or a maxi-
The Highway Capacity Manual states that the maximum mum number of load repetitions of 592,876,800 for 25%
number of passenger cars per hour per lane for a freeway at a trucks. Even the most heavily traveled highways do not
free flow speed of 65 mph is 2,350 (48). In rolling terrain, a maintain traffic streams at capacity 24 hours a day and not
single truck or bus would replace 2.5 passenger cars (48). all trucks are loaded. The fact that all trucks are not fully
Thus, one would expect a maximum of 940 trucks per hour, loaded is illustrated by a Washington DOT study of 10 weigh-
22,560 trucks per day, or a maximum of 329,376,000 trucks in-motion sites over a one-year period, which indicated
in a 40-year period. Such a case might represent a dedicated that the typical number of ESALs for a Class 9 vehicle was
truck lane running at capacity 24 hours a day, 7 days a week, 1.2 (50). If 1.2 “design load” axles were applied per truck for
365 days a year, an unlikely occurrence. By comparison, the the maximum number of trucks per lane in a 40-year period
very heavily traveled section of Interstate 710 in California (329,376,000), a total of 395,251,200 load repetitions would
carried a maximum of 9,650 trucks per day in the design be applied. Also, in winter months in many parts of the
lane (20). By calculating the appropriate heavy-vehicle adjust- country, the pavement stiffness is very high, and this results
ment factor and determining its impact on traffic flows (48), in significantly lower strains. Therefore, it is a reasonable
mixed traffic streams with 25% and 50% trucks would pro- assumption that the maximum possible number of load
duce a maximum of 148,219,200 (10,152 trucks per day) and repetitions expected in a 40-year period is approximately
235,118,400 trucks in a 40-year period, respectively. 500 million. This could be considered as a practical target
Consider, for example, an FHWA Class 9 vehicle or five-axle for evaluating parameters (strain or energy) indicating an
single trailer, which typically consists of two tandem axles endurance limit.
16

Research conducted during the SHRP recommended a shift Materials

factor of 10 between laboratory beam fatigue results and field
performance, equating to 10% cracking in the wheel-path (46). The literature indicated that the primary material proper-
Considering this shift factor, laboratory testing to 50 million ties affecting fatigue life are binder content, binder stiffness,
cycles would equate to approximately 500 million loading and air void content. The literature indicated that aggregate
cycles in the field or approximately the maximum possible gradation, type, shape, and angularity have more limited
loading in a 40-year period. Based on these analyses, a mix effects. Different nominal maximum aggregate sizes, with
that provided 50 million cycles or more of fatigue life in the their corresponding differing minimum VMA requirements,
laboratory was considered to be indicative of a long-life will tend to produce differing volumes of asphalt binder.
pavement.
If pavements were designed to have a strain level at the en- Phase I
durance limit, then all pavements, regardless of traffic, would
be designed with approximately the same thickness of HMA A full-factorial experiment was conducted to evaluate the
(assuming the same underlying support). This approach is existence of an endurance limit and to identify factors affect-
contrary to the way pavements have been designed in the past ing the endurance limit. Two main factors were included in
and is unlikely to be cost effective for future designs. Hence, the experiment. Two additional factors were fixed. These fac-
it is more reasonable, especially for highways with low to tors along with their levels are as follows:
medium traffic, to design a pavement for the expected traffic
during an extended period of years (e.g., 40 to 50) than to • Nominal maximum aggregate size (NMAS)—19.0 mm
simply design at the endurance limit. • Aggregate type—granite
Practically speaking, highways with lower traffic levels can • Asphalt content—optimum and optimum + 0.7%
be designed with less pavement structure and still have long • Binder stiffness—PG 67-22 and PG 76-22
lives. Hence, the amount of traffic has to be a critical element
in the design process. In the past there has been very little The lower lifts of part of the structural experiment from
fatigue testing at low strains (very high cycles to failure) and the 2003 NCAT Test Track were replicated to provide two
this study was designed to identify the relationship between of the mixes for the experiment. The mixes were a 19.0 mm
strain and cycles to failure at these very low strains (high cycles NMAS granite mixture at optimum asphalt content with
to failure). In the process of evaluating low strains, it was felt both neat PG 67-22 and SBS modified PG 76-22 binder. The
that the “endurance limit” would be better identified. average field gradations, shown in Table 3.1, were used as the
Two test procedures were utilized to evaluate the existence target gradation for the laboratory study. Previous research
of the endurance limit: beam fatigue tests and uniaxial ten- on fatigue has indicated that fatigue results are relatively insen-
sion tests. Beam fatigue tests have been the most widely used sitive to gradation. Therefore, it was felt that the use of the
method for testing fatigue in the United States. Uniaxial ten- average gradation would be appropriate.
sion tests have provided an alternative that allowed a more The base mix was placed in four lifts for sections N3 and N4
fundamental analysis by modeling the viscoelastic material and two lifts in the other sections. The asphalt content of the
behavior using Schapery’s correspondence principle, contin- base layer for section N8 was intentionally set at 0.5% above
uum damage mechanics, and work potential theory. optimum. Therefore, it was not included in the average.

Table 3.1. Production gradations for base layers of 2003 NCAT Test Track structural experiment.
N3 - N3 - N4 - N4 -
Sieve Size N1 N2 Upper Lower Upper Lower N5 N6 N7 N8 Average
1" 100 100 100 100 100 100 100 100 100 100 100
3/4" 92 93 100 90 92 88 92 90 90 92 92
1/2" 80 84 84 79 79 77 79 78 78 83 80
3/8" 71 74 75 68 66 66 66 71 71 73 70
No. 4 49 53 57 50 49 49 49 53 53 54 52
No. 8 40 43 48 44 43 42 43 44 44 45 44
No. 16 33 35 42 39 36 36 36 36 36 37 37
No. 30 24 24 33 30 26 28 26 27 27 26 27
No. 50 13 14 20 16 14 16 14 15 15 14 15
No. 100 8 9 11 9 8 9 8 9 9 9 9
No. 200 5.5 5.5 6.7 5.6 5.5 5.5 5.5 5.7 5.7 5.5 6
Asphalt
Content 4.3 4.5 4.3 4.6 4.7 4.4 4.7 5.0 5.0 5.2 4.7
 17

Table 3.2. High and intermediate temperature test Test Methods

data for PG 64-22 binder.
Flexural Beam Fatigue Testing
Test Value, Failure
kPa Temperature, Four-point beam fatigue testing was conducted according
°C to AASHTO T321, “Determining the Fatigue Life of Com-
DSR G*/sin δ, original binder at 64°C 1.702 68.4
DSR G*/sin δ, RTFO residue at 64°C 4.268 69.1 pacted Hot-Mix Asphalt (HMA) Subjected to Repeated Flex-
DSR G* (sin δ), PAV residue at 25°C 2805 20.4 ural Bending.” In this procedure, beam specimens (380-mm
length, 63-mm width, 50-mm height) are loaded under strain-
controlled conditions using sinusoidal loading at 10 Hz. The
The PG 67-22 used at the 2003 NCAT Test Track is a non- literature indicated that beam fatigue tests were the most
standard grade used in the southeastern United States. The high commonly used form of fatigue test in the United States. The
temperature and intermediate temperature binder test data for literature also indicated that beam fatigue tests were sensitive
the neat binder used in the 2003 NCAT Test Track are shown to material properties.
in Table 3.2. The data indicates that the PG 67-22 used at the Testing was conducted in constant strain mode. Although
NCAT Test Track also meets the properties of a PG 64-22. the literature indicated that constant stress tests may be more
Following the procedures developed during SHRP and de- appropriate for thick pavements, it also indicated that pave-
scribed in AASHTO R30, all mixtures underwent short-term ments never perform in a true constant stress manner, whereas
aging for 4 h at 135°C before compaction. This short-term the performance of thick pavements can be approximated
aging procedure allows for absorption of the asphalt binder by constant strain tests. Further, the stiffest mix performs the
into the aggregate and simulates the aging that occurs during best in constant stress testing, but this is usually not the case
production at an HMA facility. in the field. It is felt that mixture stiffness is accounted for in
Sample preparation affects the measured fatigue life. To the analysis when calculating the strain at the bottom of the
reduce variability, all of the samples tested in the study were HMA layer.
mixed and compacted by NCAT. Individual beams were com- Each of the cells in the experimental plan (Table 3.1) was to
pacted using a linear kneading compactor for beam fatigue be tested at six strain levels beginning on the high end of the
testing. Samples were then wet sawed to specified dimensions. range, as follows:
Cylindrical samples were compacted using the Superpave
gyratory compactor for uniaxial tension testing. These sam- • 800 ms,
ples were later cored and sawed to size once they reached the • 400 ms,
University of New Hampshire’s laboratory. Samples were care- • 200 ms,
fully packed for shipping to other laboratories. • 100 ms,
The air void contents of the optimum asphalt content sam- • 70 ms, and
ples were targeted at 7 ± 0.5%. An experiment was conducted • 50 ms.
to assess the expected reduction in air voids, using the same
constant stress compaction effort that would result from the At least two replicates were tested for each cell. Once the fatigue
optimum plus asphalt content. A 3.7% reduction in air voids lives of both replicates at a given strain level exceeded 50 mil-
was observed, resulting in a target air voids content of 3.3 ± lion cycles, the next lower strain level was not tested. AASHTO
0.5% for the optimum plus asphalt content samples. T321 indicates typical strain levels between 250 and 750 ms.
The literature suggests that the endurance limit in the labo-
ratory is on the order of 70 ms (8, 36) and possibly up to 200 ms
Phase II
in the field (11). The 50 ms strain level was added so that at
Additional testing was completed at the end of Phase I to least one strain level would be investigated that was believed
examine the variability of beam fatigue testing and calcula- to be below the endurance limit.
tion of the endurance limit and the affect of binder grade on Two replicate tests were performed at each strain level.
the endurance limit. Two additional binder grades, PG 58-28 This provided a maximum total of 12 data points to fit the
and PG 64-22, were utilized in the previously described mix- relationship between strain and cycles to failure. Ideally, the
ture at optimum asphalt content. research team would have tested three replicates at each strain
To date, a precision statement has not been developed for level. However, there was concern over the additional time
beam fatigue testing. A full round-robin according to ASTM this would take at low strain levels. If the research team were
C802 is beyond the scope of this project. A smaller scale round- assured that the log-log relationships for strain or energy con-
robin was conducted to provide an estimate of the variability cepts remained a straight line at low strain levels, three repli-
of beam fatigue testing. cates would have been preferable (51). However, since this
18

study was trying to identify a break or curve in those relation- be found by utilizing the time-temperature superposition
ships, it was felt that fewer points at more levels provided principle and the concept of reduced time.
more information. Chehab et al. (33) demonstrated that the viscoelastic time-
Testing was conducted to failure (a reduction in stiffness temperature shift factors are applicable to mixtures with grow-
of 50%) or a minimum of 50 million cycles. Since the goal ing damage. Therefore, the shift factors determined from
of this study was to determine the existence of an endurance complex modulus master curve construction can be used to
limit, the strain levels were being altered to better define the shift the characteristic curves at various temperatures to a
endurance limit. For instance, the PG 64-22 mix at optimum single reference temperature. Complex modulus (frequency
asphalt content tested at 100 ms had fatigue lives in excess sweep) testing was conducted at five temperatures, −10°C,
of 50 million cycles, but when tested at 200 ms, failed prior 0°C, 10°C, 20°C, and 30°C to develop the master curve.
to 50 million cycles (average 20,445,922 cycles). Therefore, it Uniaxial frequency sweep testing was conducted with a
was decided that it was more informative to perform tests at mean stress of zero to prevent the accumulation of perma-
an intermediate strain level between 100 and 200 ms instead nent deformation. It is interesting to note that Daniel and
of conducting tests at strain levels less than 100 ms. In this Kim (32) recommend the following for testing:
example, 170 ms was selected as the point where the log-log
relationship between strain and cycles to failure, developed at Ms levels of 50–70 should be targeted at each frequency-
temperature combination to ensure that the linear viscoelastic
higher strain levels (800 to 200 ms), predicted a fatigue life of response is measured and that damage is not induced in the
50 million cycles. specimens.
Three beam fatigue devices were used to conduct the test-
ing. The study began with NCAT using a single IPC Global The ms levels noted by Daniel and Kim correspond to the
beam fatigue device and the Asphalt Institute using a Cox & anticipated level of the endurance limit. Following the fre-
Sons beam fatigue fixture in an Interlaken hydraulic load quency sweep tests, the same samples were loaded in monoto-
frame. NCAT later added a second IPC Global beam fatigue nic tension to failure. The strain rate will be chosen to prevent
device. The Asphalt Institute had some difficulties testing at the occurrence of a brittle failure. The monotonic tension tests
low strain levels and testing to greater than 10 million cycles will be used to develop the characteristic curve.
to failure with their Interlaken hydraulic load frame. Conse- Once the characteristic curve and viscoelastic shift factors
quently, the Asphalt Institute also obtained an IPC Global are known, the behavior of the mix at other temperatures and
beam fatigue device. Rutgers University also tested a PG 67-22 loading rates/amplitudes can be predicted. The number of
at optimum plus asphalt content beam at 200 ms using an cycles to failure for different amplitudes and temperatures were
IPC Global beam fatigue device. then predicted using the characteristic curve, and the shift
In Phase II, two of the labs used a Cox & Sons fixture in a factors were determined from complex modulus testing.
servo-hydraulic frame and the remaining four labs used IPC Selected continuous cycles to failure tests were performed
Global’s pneumatic system to conduct the beam fatigue tests. to verify the predicted values. The continuous cycles to failure
Testing in Phase II was conducted at 800 and 400 ms and the test consists of a constant crosshead strain amplitude haver-
strain level representing the average of the predicted endurance sine loading applied continuously to the specimen in the ten-
limit for all of the labs testing a given mix. sile direction until failure occurs. Frequencies of 1 Hz and
10 Hz are used for the fatigue testing. The amplitude is chosen
Uniaxial Testing to achieve failure of the specimen at a desired number of cycles
based on the fact that the higher the amplitude, the faster the
A methodology by which the material response under var- specimen will fail. For this study, tests at 10 Hz were used for
ious uniaxial tensile testing conditions (type of loading and the verification fatigue tests to allow comparison with the beam
temperature) can be predicted from the material response fatigue results. Because of machine compliance, even when
obtained from a single testing condition has been proposed constant strain tests are conducted, the sample receives a mixed
by Daniel and Kim (32). The basis of this methodology is in a mode of loading comparable to real pavements.
characteristic curve that describes the reduction in material Due to limitations in computer memory, and the need for
integrity as damage increases. The characteristic curve is gen- a reasonably fast data acquisition rate to capture the neces-
erated by modeling the viscoelastic material behavior using sary information, only snapshots of data can be acquired
Schapery’s correspondence principle, continuum damage during the damage tests. In the continuous cyclic fatigue tests,
mechanics, and work potential theory. The characteristic curve one-second snapshots of data at a rate of 100 points per cycle
at any combination of temperature and loading conditions (1,000 points per second for the 10 Hz loading frequency)
(cyclic versus monotonic, amplitude/rate, frequency) where were collected on a logarithmic scale up to a time increment
viscoelastic behavior dominates the material response can of 2 to 10 minutes, depending upon the projected failure time
 19

of the specimen. If specimens are expected to fail in a shorter

amount of time, the time between successive snapshots was
reduced in an attempt to acquire data close to the actual fail-
ure point and to adequately describe the changing material
behavior as damage grows in the specimen.
Samples 150 mm tall by 75 mm in diameter were cored
from gyratory samples for testing. Prior to testing, steel end
plates were glued to the specimen using Devcon Plastic Steel
epoxy. A gluing jig was used to minimize any eccentricities
due to unparallel specimen ends. Four loose core type linear
variable differential transformers (LVDTs) were mounted to
the specimen surface at 90° radial intervals using a 100-mm
gage length. Additionally, two spring-loaded LVDTs were
mounted 180° apart to measure the plate-to-plate defor-
mations. The ram and LVDT deformations and load cell
measurements were collected using a National Instruments
data acquisition board and Labview software.
Testing was performed using a closed-loop servo-hydraulic
testing system. A 8.9 kN (2,000-lb) or 89 kN (20,000-lb) load
cell was used depending upon the anticipated testing loads.
The temperature was controlled with an environmental cham-
ber that uses liquid nitrogen for cooling and a feedback system
that maintains the temperature during testing. An example of
a failed uniaxial tension fatigue sample is shown in Figure 3.1.

Indirect Tensile Testing

The literature indicates that parameters from the indirect
tensile strength test, AASHTO T322, may be correlated with
parameters related to the endurance limit. This test was con-
sidered as a possible surrogate test, which could be conducted
more expediently, than the long-duration beam fatigue tests.
Indirect tensile tests were conducted on the Phase I mixes. Figure 3.1. Uniaxial tension fatigue sample.
20

CHAPTER 4

Beam Fatigue Test Results and Analyses

This chapter describes the beam fatigue testing conducted was determined at Ndesign = 80 gyrations. In addition, the
to confirm the existence of the endurance limit. One of the samples’ air void contents were reduced from 7.0 ± 0.5% to
most important aspects of this research is a practical defini- 3.3 ± 0.5% for the optimum plus samples. The voids for the
tion for the endurance limit. Asphalt mixtures simply cannot optimum plus samples were reduced to simulate the expected
be tested for an infinite fatigue life in the laboratory. Testing improved densification in the field. Advantages of optimum
at 10 Hz, approximately one million load repetitions can be plus or rich bottom layers are believed to include better com-
applied to a beam fatigue sample in a given day. The primary pactability, greater resistance to fatigue damage, and improved
goal of this testing was to confirm the existence of a fatigue moisture susceptibility.
endurance limit. In order to accomplish this goal, it was nec-
essary to develop a method for estimating the endurance limit
Extrapolation Methods
through accelerated testing in a reasonable period of time. A
to Predict Fatigue Life
secondary goal was an estimate of the variability associated
with beam fatigue testing and its potential impact on pave- As discussed previously, it was decided prior to the start of
ment thickness design. testing that beam fatigue tests would be terminated at 50 mil-
The first portion of Chapter 4 discusses methods for extrap- lion cycles. If a shift factor of 10 was applied to the test results
olating the fatigue failure point for strain levels that did not from a sample tested to 50 million cycles, it would then be
result in failure in less than 50 million loading cycles. The estimated that the pavement could withstand 500 million
methods were applied to samples that had fatigue lives in loading cycles at the corresponding strain level. Based on
excess of 10 million, but less than 50 million loading cycles. capacity analysis of a lane, this then represents a reasonable
The second portion of the chapter presents the data collected maximum number of loading cycles that might occur in a
in Phase I, as well as additional binder grades tested with 40-year period. For the practical definition of the endurance
the same mixture in Phase II. Evidence of the existence of a limit, a 40-year life was considered to be indicative of a long-
fatigue endurance limit is presented for each of these mixtures. life pavement. It takes approximately 50 days to test a single
The third portion of the chapter describes a limited round- sample to 50 million cycles. Additional analyses will be dis-
robin conducted to assess the variability of fatigue testing and cussed later to evaluate the existence of a theoretical or truly
the prediction of the endurance limit. Finally, indirect tensile infinite life endurance limit.
tests were investigated as a surrogate for beam fatigue tests. For samples that failed in less than 50 million cycles at 50%
As discussed in Chapter 3, a single gradation and aggregate of initial stiffness, the number of cycles to failure was deter-
type was used for all of the testing. A full-factorial experiment, mined from the data acquisition software controlling the test.
shown in Table 4.1, was conducted in Phase I to evaluate the However, if the test was terminated prior to reaching 50%
existence of an endurance limit and to identify factors affect- of initial stiffness, either due to an equipment problem or
ing the endurance limit. Two main factors were included in to reaching 50 million cycles, an extrapolation procedure was
the experiment—binder grade and asphalt content. Binder used to estimate the number of loading cycles, Nf, correspon-
grade was varied at two levels: PG 67-22 and PG 76-22. As ding to 50% of initial stiffness. Ideally, a method of extrap-
noted previously, the PG 67-22 also met the requirements olation would be identified that could be used to shorten
of a PG 64-22. Asphalt content was varied at two levels: opti- the beam fatigue testing procedure used to determine the
mum and optimum plus 0.7%. Optimum asphalt content endurance limit. Then, samples could be tested to 4 million
 21

Table 4.1. Experimental design.

NMAS (mm) Binder Content Granite
PG 67-22 PG 76-22
19.0 Optimum X X
Optimum + X X

cycles as done in the Asphalt Institute study (38, 39) or pos- lation: exponential model, logarithmic model, Weibull func-
sibly 10 million cycles and the fatigue life at, or close to, the tion, three-stage Weibull function, and ratio of dissipated
endurance limit predicted. energy change (RDEC). Each of these is discussed below, and
In Phase I, testing was conducted at progressively lower examples are provided of the predicted fatigue lives.
strain levels until two samples at a given strain level reached
50 million cycles without reaching 50% of their initial stiff-
Fatigue Life Extrapolation Using
ness (failure). Instead of testing at a lower strain level below AASHTO T321 Exponential Model
that providing a fatigue life of at least 50 million cycles, sam-
ples were tested at the strain level predicted to provide a fatigue AASHTO T321 specifies an exponential model (Equation 2)
life of 50 million cycles. The goal of this additional point was for the calculation of cycles to 50% initial stiffness, as follows:
to help define the transition from “normal” strain test to
tests below the apparent fatigue endurance limit or “low” S = Ae bn (2)
strain tests.
For the PG 67-22 mix at optimum asphalt content, the where,
data from 800 through 200 ms were used to estimate the S = sample stiffness (MPa),
strain level that would result in a fatigue life of 50 million A = constant,
cycles. A linear regression was performed between the Log10 b = constant, and
of ms and the Log10 of loading cycles to 50% initial stiffness. n = number of load cycles.
The R2 = 99.6 for Equation 1. Using Equation 1, it was deter-
mined that a strain level of 166 ms should produce a fatigue The constants are determined by regression analysis of load-
life of 50 million cycles. This was rounded to 170 ms for ing cycles versus the natural logarithm of the flexural stiff-
testing purposes. Testing was conducted at this strain level ness. The number of cycles to failure is determined by solving
to better define the endurance limit. Equation 2 for 50% of initial stiffness. In this study, for sam-
ples tested to less than 50 million cycles, the number of cycles
N f = 1020.6 × ε −5.81 (1) reported to reach 50% of the initial stiffness are the actual
number of cycles recorded by the test equipment, not the num-
where, ber of cycles determined using Equation 2. No discussion is
Nf = number of cycles to 50% of initial stiffness and provided in AASHTO T321 regarding whether or not all of
ε = constant strain used in beam fatigue test (ms). the data (particularly the initial data) should be used when
solving for the constants in Equation 2 (52).
When testing the PG 67-22 mix at optimum asphalt content Sample 5 of the PG 67-22 mix at optimum asphalt content,
at 170 ms, the first replicate failed in 34.7 million cycles. The which was tested at 170 ms, was selected as an example. It was
second replicate was at 55% of its initial stiffness at 50 million desirable to select a sample that had as long a fatigue life as pos-
cycles. Therefore, testing was extended to see if the failure point sible and had reached 50% of initial stiffness within 50 million
could be determined. However, at 60 million cycles, Sample 23 cycles. It was felt that an extrapolation method that worked
still retained 53% of its initial stiffness. Therefore, testing was well at high strain levels may not prove to be as accurate at
suspended at 60 million cycles. strain levels closer to the anticipated endurance limit. Fig-
For the PG 67-22 mix at optimum asphalt content, samples ure 4.1 shows fits to the loading cycle versus sample stiffness
tested at 200 and 170 ms were used to investigate extrapola- data determined using Equation 2. The coefficients for Equa-
tion techniques. These strain levels and similar strain levels tion 2 were fit using the data up to 4 million cycles, 10 mil-
for the other mixes used in Phase I provided long fatigue lives lion cycles, and failure (34.7 million cycles). As can be seen
(in excess of 10 million cycles) while still having a defined fail- from Figure 4.1, when all of the data up to failure was used
ure point that could be used to investigate the accuracy of the to fit the model, the model provides a reasonable estimate
extrapolation. Five techniques were investigated for extrapo- of fatigue life.
22

PG 67-22 Sample 5 at 170 ms

5000
y = 3970 .6e-5E-08x y = 3927 .2e-2E-08x y = 3890 .5e-1E-08x
4500 R² = 0.4992 R² = 0.4776 R² = 0.7259
4000

Flexural Stiffness, MPa

3500
3000
2500
2000
1500
1000
500
0
0

10

20

30

40

50
,0

,0

,0

,0

,0
00

00

00

00

00
,0

,0

,0

,
00

00
00

00

00
0

0
Loading Cycles

50% Initial Expon. (4.00E+06) Expon. (1.00E+07) Expon. (To Failure)

Figure 4.1. Examples of fatigue life estimates using the exponential model.

Fatigue Life Extrapolation Using Natural When all of the fatigue data are used to fit a logarithmic
Logarithm of Loading Cycles versus Stiffness model, the slope of the fitted line at higher numbers of loading
cycles may be flatter than the actual data. This leads to an over-
A logarithmic model (Equation 3) using the natural loga- estimation of the fatigue life. This is illustrated in Figure 4.2 for
rithm of loading cycles versus stiffness was evaluated as one Sample 5 of the PG 67-22 mix at optimum asphalt content.
alternative to the exponential model. The fits to the logarithmic model using just the first 10 million
cycles and using all of the data are indistinguishable on the
S = α + β × ln (n ) (3) plot. Note that in Figure 4.2 the logarithmic model provides
very high R2 values, but the fitted model does not match the
where, experimental data at a high number of cycles.
S = the sample stiffness at loading cycle n, and However, by eliminating a portion of the early loading
α and β are regression constants. cycles, a good match to the data can generally be obtained,

5000
Data to 4 million Cycles Data to 10 million Cycles
4500 y = -91.789Ln(x) + 4865.2 y = -91.483Ln(x) + 4862.4
R2 = 0.9714 R2 = 0.9763
4000
Flexural Stiffness, MPa

3500
3000
2500
2000
1500 To Failure
1000 y = -106.17Ln(x) + 5005.6
R2 = 0.9108
500
0
0

10

20

30

40

50
,0

,0

,0

,0
00
00

00

00

0
0,

0,
,0

,0

,0
0

00
0

00

00
0

Loading Cycles

50% Initial Log. (4.00E+06) Log. (1.00E+07) Log. (To Failure)

Figure 4.2. Logarithmic model fits for PG 67-22 at optimum, Sample 5.

 23

8,000
Logarithmic Model using All Data Logarithmic Model Starting at 1 million Cyles
7,000 y = -235.ln(x) + 7766. y = -405ln(x) + 10226
R² = 0.965 R² = 0.965
6,000

Flexural Stiffness, MPa

5,000

4,000

3,000

2,000

1,000

0
0 2,000,000 4,000,000 6,000,000 8,000,000 10,000,000 12,000,000 14,000,000
Loading Cycles

Log. (All Data) Log. (Starting 1E+6 cycles)

Figure 4.3. Logarithmic model fits for PG 67-22 at optimum, Sample 6.

particularly at low strain levels. In Figure 4.3, logarithmic where,

models were fit to the data from Sample 6 of the PG 67-22 mix α = 10 raised to the power of the intercept from regression
at optimum asphalt content, which was tested at 200 ms. In of log (S) versus log (n), and
Figure 4.3, logarithmic models are shown including all of the β = the slope from regression of log (S) versus log (n).
initial loading cycles and excluding the first million loading
cycles. This provides a better fit to data, but still would tend The power model has a similar shape to the logarithmic model.
to overestimate the fatigue life. Further, the number of early Shen (54) also reported the need to eliminate a number of ini-
loading cycles that are not included must be determined by tial cycles to obtain a good fit to the slope at high numbers of
trial and error. Note that all of the logarithmic models shown cycles. Failure to eliminate some of the initial cycles results
in Figures 4.2 and 4.3 provide high R2 values, even when the in an overestimation of the fatigue life. Additional discussion
fit to the data at a high number of cycles is not very good. This on the use of the power model and its application to the ratio
suggests that R2 values alone are not adequate to evaluate of dissipated energy will be provided later in this report.
extrapolation models. It is believed that the poor fit at a high
number of cycles results from the fact that the data are col- Fatigue Life Extrapolation Using
lected using a logarithmic progression. That is, the sampling the Weibull Survivor Function
rate is high, every 10 cycles, when the test is initiated but may
be every million cycles at a high number of cycles. Thus, there Often, failure data can be modeled using a Weibull distri-
are more data points to fit in the early portion of the curve. bution. The Weibull function is commonly used in reliability
Rowe and Bouldin (53), when examining fits from the expo- engineering to estimate survival life. Tsai et al. (55) applied
nential model, concluded that fatigue data should be taken the Weibull survivor function to HMA beam fatigue data.
with every 5% reduction in stiffness. The generalized equation for the Weibull function is given by
Equation 5.
Fatigue Life Extrapolation Using γ
⎛ t −δ ⎞
R ( t ) = exp− ⎜
⎝ θ − δ ⎟⎠
the Power Model (5)
The ratio of dissipated energy, developed by Shen and
Carpenter (40) also requires that the number of cycles to 50% where,
of initial stiffness be calculated in order to determine the plateau R(t) = the reliability at time t where t might be time or
values. Shen (54) recommends a power model (Equation 4) for another life parameter such as loading cycles,
the extrapolation of stiffness versus loading cycles, as follows: γ = the slope,
δ = the minimum life, and
S = α + nβ (4) θ = the characteristic life.
24

Tsai et al. (55) applied a specialized case of the Weibull func- Figure 4.4 shows an example of the data from the two
tion where the minimum life, δ, was assumed to be 0. In this 100 ms samples from the PG 67-22 at optimum mixture in
case, the characteristic life = 1/λ and the Weibull function the form of Equation 7. Tsai et al. (55) observed that the
simplifies to Equation 6. Since the beam fatigue loading cycles concave down shape, exhibited by Sample 13, “implied that
are applied at a constant frequency of 10 Hz, loading cycles, n, the fatigue damage rate is slowed down and flattens out with
can be substituted for time, t. increased repetitions and thus causes no further damage after
a certain number of repetitions.” This behavior is believed
S ( t ) = exp ( − λ × n γ ) (6) to be indicative of the endurance limit.

where,
S(t) = probability of survival until time t, Fatigue Life Extrapolation Using
λ = scale parameter (intercept), and Three-Stage Weibull Function
γ = shape parameter (slope).
In the previous section, the Weibull survivor function was
The stiffness ratio can be used to characterize fatigue dam- presented as a method for modeling the fatigue life of beam
age. The stiffness ratio is the stiffness measured at cycle n, fatigue tests. In later sections, it will be demonstrated that the
divided by the initial stiffness, determined at the 50th cycle. single-stage Weibull function generally provides a good esti-
Tsai (56) states that at a given cycle n, the beam being tested mate of a sample’s fatigue life and is reproducible when calcu-
has a probability of survival past cycle n equal to the SR lated by different laboratories. There are, however, cases for
times 100%. Thus, SR(n) can be substituted for S(t). Tsai (56) which the single-stage Weibull function apparently under-
presents the derivation of Equation 7, which allows the scale predicts fatigue life. Sample 13 in Figure 4.4 is one such exam-
and shape parameters for laboratory beam fatigue data to be ple (but Sample 13 is not specifically labeled on the plot). It
determined by linear regression. should be noted that with the exception of two examples
analyzed by Tsai, the three-stage Weibull analyses were not
ln ( − ln ( SRn )) = ln ( λ ) + γ × ln (n ) (7) conducted until after the completion of all of the Phase I
and II testing.
where, To improve upon the accuracy of the single-stage Weibull
SRn = stiffness ratio or stiffness at cycle n divided by the function, Tsai et al. (57) developed a methodology for fitting
initial stiffness. a three-stage Weibull curve. Tsai et al. (57) theorized that a

0
y = 0.243x -5.816
-1
R² = 0.82

-2
Ln(-Ln(Stiffness Ratio))

-3

-4

y = 0.4014x -8.2141
-5
R² = 0.72
-6

-7

-8

-9

-10
0 2 4 6 8 10 12 14 16 18 20
Ln (Cycles)
Sample 4 Sample 13 Linear (Sample 4) Linear (Sample 13)

Figure 4.4. Weibull survivor function for PG 67-22 at optimum 100 ms samples.
 25

plot of loading cycles versus stiffness ratio could be divided 1

into three stages: initial heating and temperature equilibrium, ⎡ α 3 ⎛ β 3 ⎞ β3 ⎤ β 2 − β3

n2 = γ 1 + ⎢ × ⎜ ⎟ ⎥ (11)
crack initiation, and crack propagation. In the case of low ⎢⎣ α 2 ⎝ β 2 ⎠ ⎥⎦
strain tests (below the endurance limit), the third stage does
not appear to represent crack propagation, but rather con-
⎛ β ⎞ β
cave down stage with a reduced rate of damage. γ 2 = ⎜ 1 − 3 ⎟ × n2 + 3 × γ 1 (12)
⎝ β2 ⎠ β2
A Weibull function is fit to each of these stages as shown in
Equation 8:
Tsai et al. (57) applied a genetic algorithm to resolve the six
unknown parameters. A genetic algorithm requires a param-
SR1 = e ( )
− α1 ×nβ1
for 0 ≤ n < n1 eter definition, in this case Equations 8 through 12, and a fit-
SR2 = e ( )
− α 2 ×(n− γ 1 )β2 ness function. Residual sum of squares between the measured
for n1 ≤ n < n2 (8) and fitted ln(-ln(SR)) for each cycle is used as the fitness func-
SR3 = e ( )
− α3 ×(n− γ 2 )β3 tion. A “good gene” is defined as an optimum set of param-
for n2 ≤ n < n3
eters to minimize the fitness function. A set of input ranges is
first determined for α1, α2, α3, β1, β2, and β3. The input ranges
Coefficients α1, α2, α3, β1, β2, β3, n1, and n2 are illustrated in are determined by visual inspection of the data. Simple linear
Figure 4.5. regressions are performed for each stage to determine slopes
Using a series of mathematical manipulations (57), n1, γ1, and offsets. An example is shown in Figure 4.5. Tolerances are
n2, and γ2 can be calculated sequentially as follows: applied to the parameters determined by inspection to set ini-
tial ranges for each coefficient. The input ranges and test data
1 are entered into a FORTRAN program, N3stage.exe, devel-
⎡ α 2 ⎛ β 2 ⎞ β2 ⎤ β2 −β1 oped by Tsai. The program randomly generates a set of param-
n1 = ⎢ × ⎜ ⎟ ⎥ (9)
⎢⎣ α1 ⎝ β1 ⎠ ⎥⎦ eters or “genes” within the input ranges. The fitness parameter
is calculated for each set of genes and the sets of genes are
ranked. Good genes are mated and bad genes are discarded
⎛ β ⎞
γ 1 = ⎜ 1 − 2 ⎟ × n1 (10) and replaced with new genes. The cycle of producing genes,
⎝ β1 ⎠ ranking genes by the residual sum of squares, mating and

PG 76-22 Optimum Plus Sample 9 200 ms

0
y = 0.0667x -2.6913
R² = 0.7617
-1
y = 0.1769x -4.1281
R² = 0.9075 3
Ln( -Ln(Stiffness Ratio))

1
-2
y = 0.5427x -6.6613 2

R² = 0.547 1
-3 Stage 1
Stage 2
1
Intercept = Lnα3
Stage 3
-4
1 Intercept = Lnα2

Intercept = Lnα1 LN (n1) Ln (n2)

-5
Stage 1 Stage 2 Stage 3
Initial Heating
-6
0 2 4 6 8 10 12 14 16 18 20
Ln (Loading Cycles)

Figure 4.5. Three-stage Weibull curve definitions.

26

PG 76-22 Optimum Plus Sample 9 200 ms

0

-1

Ln (-Ln(Stiffness Ratio))

-2

-3
Raw Data
Fit
-4

-5

-6
0 2 4 6 8 10 12 14 16 18 20
Ln (Loading Cycles)

Figure 4.6. Three-stage Weibull fit to transformed data.

replacing genes, continues until the specified number of gen- for Long-Life Pavement Design. The NCHRP 9-38 research
erations is complete (58). The N3stage program typically takes team developed a Microsoft Excel spreadsheet to solve the
30 to 60 minutes to complete 750 generations, depending on three-stage Weibull parameters, which produces similar results
the size of the data set. The complete calculation procedure to the N3stage program.
is described in Appendix A, Proposed Standard Practice for Figures 4.6 and 4.7 show examples of the three-stage
Predicting the Endurance Limit of Hot Mix Asphalt (HMA) Weibull fit. This methodology provides a good fit to both

PG 76-22 Optimum Plus Sample 9 200 ms

6,000

5,000

4,000
Stiffness, MPa

3,000

2,000

1,000

0
0 10,000,000 20,000,000 30,000,000 40,000,000 50,000,000 60,000,000
Loading Cycles
Raw Data 3-Stage Weibull Fit

Figure 4.7. Three-stage Weibull fit to stiffness data.

 27

normal and low strain fatigue data. In some cases, only one two data points, that is, the average ratio of dissipated energy
or two stages are fit, even if three stages are initially identified. change per loading cycle. This is written as follows:

DEa − DEb
Ratio of Dissipated Energy Change (RDEC) RDECa = (13)
DEa ⴱ (b − a )
Dissipated energy is a measure of the energy that is lost to
the material or altered through mechanical work, heat gener- where,
ation, or damage to the sample. Other researchers have used RDECa = the average ratio of dissipated energy change at
cumulative dissipated energy to define damage within a spec- cycle a, comparing to next cycle b;
imen, assuming that all of the dissipated energy is responsi- a, b = load cycle a and b, respectively (the cycle count
ble for the damage. The approach suggested by Ghuzlan and between cycle a and b for RDEC calculation will
Carpenter (34) considers that only a portion of the dissipated vary depending on the data acquisition soft-
energy is responsible for actual damage. ware); and
DEa, DEb = the dissipated energy produced in load cycle a,
Typically, three regions are observed in the RDEC analy-
and b, respectively.
sis as shown in Figure 4.8. Region I represents the initial
“settling” of the sample where the rate of change of dissipated
The dissipated energy for each loading cycle is determined
energy decreases. In Region II, the rate of change of dissipated
by measuring the area within the stress-strain hysteresis loop
energy reaches a plateau, representing a period where the
for each captured load pulse. This methodology is used by
amount of damage occurring to the sample is constant. Finally,
the IPC Global beam fatigue device used in the study by
in Region III, sample instability begins as the rate of change
NCAT, Asphalt Institute, and the University of Illinois. Alter-
of dissipated energy rapidly increases. A lower dissipated
natively, the dissipated energy can be calculated according
energy ratio (DER) plateau value implies that less damage
to Equation 14.
is occurring per cycle. Therefore, a sample with a low DER
plateau value would be expected to have a longer fatigue life wn = π × σ n × εn × sin δn (14)
than a sample with a high DER plateau value. Shen and
Carpenter (40) refined this technique and suggested that the where,
RDEC plateau value (PV) should be calculated at the number σn = maximum tensile stress in cycle n, in kPa,
of cycles that produced 50% of the initial sample stiffness (Nf). εn = maximum tensile strain in cycle n,
A PV of 8.57E-9 was proposed by Shen and Carpenter as indica- δn = 360 × f × s,
tive of a long life pavement (40). f = loading frequency, Hz, and
The RDEC analysis procedure is described in Appendix C, s = time lag in seconds between peak load and peak deflec-
Proposed Standard Practice for Extrapolating Long-Life Beam tion in seconds.
Fatigue Tests Using the Ratio of Dissipated Energy Change
(RDEC). RDEC is the ratio of dissipated energy change between Due to testing noise, as shown in Figure 4.8, the raw dissi-
two data points divided by the number of cycles between the pated energy data are not directly usable for calculating RDEC

0.006

0.004 III
RDEC

II
0.002

PV

0
10 510 1,010 1,510 2,010 2,510 3,010 3,510 4,010
Number of Load Cycles

Figure 4.8. Typical RDEC versus loading cycles plot and the indication of PV.
28

IDOT03 mix 3N704B f

⎛ 100 ⎞
DE vs. Loading cycles 800 microstrains 1− ⎜1+ ⎟
1.4 ⎝ a ⎠
RDECa = (15)
1.2 100
1.0
where,
DE, kPa

0.8
Exponential Slope k f = the slope from the regressed dissipated energy-loading
0.6
y = 3.4255x -0.164 cycle curve.
0.4 R² = 0.9512
0.2 In the RDEC approach, PV is defined as the RDEC value at
Nf50
0.0 the 50% stiffness reduction failure point (Nf50). Therefore,
0 1000 2000 3000 4000 5000
the PV value can be obtained using Equation 16.
Loading Cycles

Figure 4.9. DE versus LC chart for one IDOT mix ⎛ 100 ⎞

f

1− ⎜1+
⎝ Nf50 ⎟⎠
with fitted curve.
PV = (16)
100
and PV. A curve fitting procedure is recommended to obtain
the best fit equation for the dissipated energy-loading cycle Here, the PV value depends only on the f factor of the regressed
data. It is assumed that the regression equation of dissipated power law DE-LC curve and the defined failure point, Nf50.
energy-loading cycle relationship follows a power law rela- For long-life tests, where Nf50 was not known, the stiffness-
tionship, Axf (as indicated in Figure 4.9). The key for the loading cycle curve first needed to be extrapolated to deter-
curve fitting process is to obtain a slope (in the power law mine Nf50, resulting in the calculation being based on a double
relation plot) of the curve, f, which can best represent the extrapolation.
original curve. In general, there are two rules for evaluating Using this approach, Shen and Carpenter (40) demon-
the goodness of the fitted curve: (1) a high R-square value, strated a unique PV-Nf curve for all HMA mixes at normal
and (2) correct trend of the DE-LC curve. This is similar to strain/damage level testing, regardless of the testing condi-
the procedures described previously for fitting logarithmic tion, loading modes, and mixture types. The tests used for es-
or power models to the stiffness-loading cycle curves and is tablishing this relationship at all normal testing were carried
illustrated in Figure 4.10. to or beyond the failure point (i.e., the Nf50 values are known).
The average RDEC for an arbitrary 100 cycles at cycle a can Also, using the results from long-term fatigue testing, Shen
be simply calculated using the following equation: and Carpenter (40) demonstrated that the unique relation-

Predicted DE Sample 2 at 200 ms

0.100 Legend Number Denotes
0.095 Starting Cycle
0.090 R^2 = 0.90
y = 0.7162x-0.1384
Predicted DE, kPa

0.085 66910
0.080
4617890
0.075
8403190
0.070
0.065 Measured
0.060
0.055
0.050

Loading Cycles

Figure 4.10. Dissipated energy versus loading cycle for raw data and
various power models.
 29

ship between PV-Nf is also valid for low strain testing. In other The unique PV-loading cycle curve is also illustrated in
words, the PV-Nf relationship is unique for the whole loading Figure 4.11. As illustrated in Appendix C, an alternative
range including both normal and low strain/damage level. approach for extrapolating the fatigue life of a sample that
An error analysis was performed for two dissipated energy does not fail is to plot the sample’s RDEC versus loading
curve fitting routines using 19 Illinois DOT mixtures tested at cycle on a log-log plot and determine the intersection with
low strain levels, and the results are presented in Figure 4.11. the unique PV-Nf line.
All the samples were tested to extended load repetitions
(i.e., 5 to 30 million load repetitions). Hence, the Nf values
Comparison of Fatigue Life Methods
obtained from the stiffness reduction curve are reasonable,
and the comparison can be focused on the errors that could This extended discussion on fatigue life extrapolation
be involved due to fitting the DE-LC curve. Two PV-Nf lines techniques is provided as an introduction to future method-
are shown in Figure 4.11. One is with the PVs obtained from ologies for identifying the endurance limit. In this study,
highest R2 fitting of the dissipated energy-loading cycle curve; tests were conducted to a maximum of 50 million cycles in
the other is with the PVs from visual fitting that represents the order to confirm the existence of the endurance limit. As
dissipated energy-loading cycle curve’s best trend (especially noted previously, it takes approximately two months to com-
the extension trend). They are closer at relatively higher PV plete a single test at this high number of cycles. This extended
(shorter fatigue life). With the decrease of PV (increase in test time is not practical for routine determination of the
fatigue life), the PV calculated from the highest R2 fitting endurance limit. One alternative to determine the strain
gives a greater value compared to the values obtained from level that corresponds to the endurance limit for a given mix-
visual fitting. For low strain testing, the segment that gives ture would be to conduct beam fatigue tests at a low strain
higher R2 does not necessarily represent the real trend of the level to a more limited number of cycles (perhaps less than
curve, which could induce error. Therefore, for low strain 10 million or approximately 10 days) and extrapolate the
long fatigue testing, the “highest R2 fitting” rule is not best data. Thus, a model would be fit to the stiffness versus load-
suited. For most cases, the initial segment of the DE-LC curve ing cycle data and the number of cycles required to reach 50%
has to be eliminated and only the later segment that gives of the initial stiffness would be extrapolated. A significant
a good extension trend of the curve should be used, since deviation from a log-log plot of strain versus cycles to failure
it is more representative for the actual long-term fatigue would indicate the strain level corresponding to the endurance
performance. limit (this will be shown later in the section on Existence of

1.E-06

1.E-09

1.E-12
Endurance Limit
1.E-15

1.E-18 highest R^2 fitting

y = 0.0708x-1.024
PV at Nf50

1.E-21 R² = 0.9956
Visual fitting:
1.E-24
y = 0.3553x-1.101
1.E-27 R² = 0.9989

1.E-30 Unique PV-Nf curve:

PV=0.4428Nf^(-1.1102)
1.E-33

1.E-36

1.E-39
1.E+05 1.E+08 1.E+11 1.E+14 1.E+17 1.E+20 1.E+23 1.E+26 1.E+29 1.E+32 1.E+35
Nf at 50% stiffness reduction

PV-Nf from visual fitting PV-Nf from highest R^2 fitting

PV-Nf unique curve Power (PV-Nf from visual fitting)
Power (PV-Nf from highest R^2 fitting)

Figure 4.11. PV-Nf curve for IDOT03 mix at low strain with error bars indicated.
30

120

Predicted Percent of Actual Stiffness at 50 Million Cycles

110
100
90
80
70
60
50
40
30
20
10
0
0 10,000,000 20,000,000 30,000,000 40,000,000 50,000,000 60,000,000
Number of Cycles Used for Model

Exponential Model Logarithmic Model Power Model Weibull Function

Figure 4.12. Convergence of extrapolated stiffness for PG 67-22 Sample 4.

the Endurance Limit). The two main requirements for this logarithmic, power, and Weibull function. All of the initial
technique that need to be evaluated are (1) the appropriate cycles were included when fitting the models. Figures 4.12
form of the model and (2) the minimum number of cycles and 4.13 show the percentage of the actual measured stiff-
that need to be tested. ness at 50 million cycles for each of the models for PG 67-22
The samples tested at 100 ms for the PG 67-22 at optimum at optimum for Samples 4 and 13, respectively. The cycles
asphalt content were first used to evaluate the ability of the var- shown in Figures 4.12 and 4.13 represent the total number
ious models to predict the sample stiffness at 50 million cycles. of cycles (starting at the first cycle) used to fit the model. The
Four models were considered: exponential (AASHTO T321), stiffness at 50 million cycles extrapolated using that model

120
Predicted Percent of Actual Stiffness at 50 Million Cycles

110
100
90
80
70
60
50
40
30
20
10
0
0 10,000,000 20,000,000 30,000,000 40,000,000 50,000,000 60,000,000
Number of Cycles Used for Model

Exponential Model Logarithmic Model Power Model Weibull Function

Figure 4.13. Convergence of extrapolated stiffness for PG 67-22 Sample 13.

 31

is then shown as a percentage of the measured stiffness on loading cycles and the first 10 million loading cycles. A pre-
the y-axis. For example, if Sample 4 would have been tested vious study on the endurance limit by Peterson and Turner
to 10 million cycles and a logarithmic model fit to the data, the (38) extrapolated the fatigue life based on testing to 4 million
extrapolated stiffness at 50 million cycles would be 108.2% of cycles. Shen and Carpenter (40) extrapolated test results
the measured stiffness at 50 million cycles. based on tests conducted to greater than 8 million cycles.
Examination of Figures 4.12 and 4.13 show that the expo- Table 4.2 shows the fatigue life predictions for the five sam-
nential model consistently underestimates the stiffness at ples using six different extrapolation methods: exponential
50 million cycles and is slow to converge on the measured model, logarithmic model, power model, single-stage Weibull,
stiffness (testing would need to be conducted to a high num- three-stage Weibull, and RDEC.
ber of cycles to even approach the measured stiffness). This The RDEC procedure consistently overestimates the sam-
would suggest that the exponential model recommended by ples’ fatigue lives by three to seven orders of magnitude. For
AASHTO T321 is not a good choice for extrapolating fa- example, the actual fatigue life of Sample 2 of the PG 67-22 at
tigue data. optimum asphalt content is 2.60E+07 whereas the predicted
The predicted stiffness values using the logarithmic and fatigue life using the RDEC procedure for the first 10 million
power models are basically the same in Figures 4.12 and 4.13. cycles is 3.08E+11. The power model also consistently over-
Both converge to a reasonable predicted stiffness within 10 mil- estimates fatigue life by two to seven orders of magnitude.
lion cycles. However, when all of the loading cycles are used, For the remaining methods, with the exception of the expo-
both overestimate the stiffness at 50 million cycles and, con- nential model, the fatigue life of Sample 5 of the PG 67-22 at
sequently, would overestimate the fatigue life. The single-stage optimum asphalt content was overestimated by a larger degree
Weibull function converges quickly and provides the most than for the other samples. Sample 5 was tested at 170 ms. The
accurate results for Sample 4, but does a relatively poor job logarithmic model overestimated the fatigue life of Sample 5
for Sample 13. Recall that the Weibull function for Sample 13 by five orders of magnitude, but the remaining four samples
had the concave down shape in Figure 4.4. by one to two orders of magnitude. The three-stage Weibull
The accuracy of the stiffness prediction is not the only factor model overestimated fatigue life of Sample 5 by three or
which will affect the accuracy of the fatigue life extrapolation. four orders of magnitude based on the data from 4 million
The shape of the model will also have an effect. Logarith- and 10 million loading cycles, respectively. However, for the
mic and power models can produce very flat slopes at high remaining samples, the three-stage Weibull function over-
numbers of loading cycles that result in overestimation of estimated fatigue life by zero to two orders of magnitude.
the fatigue life (particularly if some of the initial cycles are not Sample 23 of the same mix was also tested at 170 ms. It was
eliminated to better match the slope of stiffness versus load- tested to 60 million cycles without reaching 50% of its ini-
ing cycles at a high number of loading cycles). tial stiffness.
Five samples tested in Phase I had fatigue lives between The exponential and single-stage Weibull function pro-
20 and 50 million cycles. These samples were used to evaluate duced the most accurate fatigue life predictions. As shown in
the accuracy of the extrapolation techniques. Predictions Figure 4.14, the exponential model consistently underesti-
were based on models developed using the first 4 million mates fatigue life. The fatigue life of points above the line of

Table 4.2. Comparison of fatigue life extrapolations.

Mix Sample Cycles Used Actual Exponential Logarithmic Power Single- Three- RDEC
for Stage Stage
Extrapolation Weibull Weibull
(Millions)
4 7.89E+06 2.54E+09 3.80E+10 3.31E+07 1.41E+08 6.78E+11
2 2.60E+07
10 1.87E+07 2.83E+09 3.43E+10 5.59E+07 1.89E+08 3.08E+11
PG 67-22 4 4.94E+06 4.84E+08 4.19E+09 1.44E+07 4.43E+07 5.53E+10
21 2.08E+07
Optimum 10 9.88E+06 3.17E+08 1.78E+09 1.91E+07 5.04E+07 1.92E+09
4 1.27E+07 2.13E+12 2.15E+14 3.87E+08 7.23E+11 1.66E+14
5 3.47E+07
10 2.88E+07 2.27E+12 2.15E+14 7.24E+08 1.92E+10 8.02E+13
PG 67-22 4 8.62E+06 5.57E+09 1.58E+11 3.11E+07 1.20E+09 2.26E+15
Optimum 4 3.90E+07
Plus 10 1.84E+07 4.59E+09 2.97E+10 4.97E+07 3.39E+09 1.06E+11
PG 76-22 4 7.69E+06 6.56E+08 7.16E+09 1.13E+07 3.28E+08 2.56E+12
Optimum 10 3.96E+07
Plus 10 1.73E+07 7.77E+08 7.41E+09 2.01E+07 4.54E+08 1.71E+10
32

Extrapolations from 4 Million Cycles 10 million cycles using the single-stage Weibull function, is
Measured Fatigue Life, Cycles

1.0E+08 7.75. This indicates that the two predictions are relatively close.
8.0E+07
In summary, the Weibull functions were selected for extrap-
olating fatigue tests that did not fail within 50 million cycles or
6.0E+07 when the test was interrupted prior to failure (as occurred with
4.0E+07 Sample 6 of the PG 67-22 mix at optimum asphalt content).
Sample 5 Weibull
For long-life fatigue tests, at strain levels slightly above the
2.0E+07 endurance limit, the single-stage Weibull function appears
0.0E+00 to provide the most accurate extrapolation of fatigue life.
0.0E+00 2.0E+07 4.0E+07 6.0E+07 8.0E+07 1.0E+08 The three-stage Weibull function, however, provides the best
Predicted Fatigue Life, Cycles fit to the stiffness versus loading cycle data. Fatigue extrapo-
Exponential Single-Stage Weibull lations from both methods are shown when discussing evi-
dence of the endurance limit. Based on the results from this
Figure 4.14. Comparison of exponential and single- section, an AASHTO Standard Practice for Predicting the En-
stage Weibull extrapolations with measured
durance Limit of Hot Mix Asphalt (HMA) for Long-Life Pave-
fatigue lives.
ment Design was developed and is presented in Appendix A.
The draft format includes extrapolation techniques using both
the single- and three-stage Weibull functions.
equality is underestimated and below the line of equality is
overestimated. The error is greater for larger extrapolations
(e.g., testing to 10 million cycles for a sample with a fatigue Existence of the Endurance Limit
life of 40 million cycles). The extrapolations for the single-stage Samples tested below the fatigue endurance limit are ex-
Weibull model are distributed around the line of equality. As pected to have an essentially infinite fatigue life. As noted
noted previously, the fatigue life for Sample 5 is significantly previously, testing was only conducted to 50 million cycles.
overestimated. However, the single-stage Weibull function Therefore, the failure point of these samples needed to be
appears to give the most reasonable extrapolation of fatigue extrapolated. Two techniques were used to extrapolate the
test results. stiffness versus loading cycle data, the single- and three-stage
When looking at the accuracy of fatigue predictions, it Weibull functions. Additionally, the data were analyzed using
should be considered that strain versus fatigue life data typ- the RDEC procedures to determine the plateau value. The fol-
ically is looked at on a log-log plot. The log of 26 million, lowing sections present the results from the testing and dis-
the measured fatigue life for Sample 2, is 7.41, while the log cuss each of the analyses. Tables 4.3 through 4.6 present the
of 56 million, the fatigue life estimated based on the first data collected in Phase I of the study.

Table 4.3. Granite 19.0 mm NMAS mix with PG 67-22 at optimum asphalt.

Beam Air Initial Micro- Cycles Extrapolated Cycles to 50% PV Cycles to 50% Average
ID Voids, Flexural Strain Tested Initial Stiffness Initial Cycles
% Stiffness, Single-Stage Three-Stage Stiffness to Failure
MPa Weibull Weibull
18 6.6 5,175 800 6,000 NA NA 3.66E-5 6,000
3 6.8 4,686 800 7,130 NA NA 2.06E-5 7,130 6,377
7 7.4 4,522 800 6,000 NA NA 2.63E-5 6,000
10 6.8 5,153 400 246,220 NA NA 6.25E-7 246,220
46 7.0 5,239 400 57,000 NA NA 2.24E-7 267,8081 252,136
1 7.0 5,868 400 242,380 NA NA 3.17E-7 242,380
2 6.6 5,175 200 26,029,000 NA NA 5.33E-94 26,029,000
6 7.2 6,435 200 12,930,000 NA NA 6.19E-94 14,537,1862 20,445,922
21 7.4 6,240 200 20,771,580 NA NA 6.35E-94 20,771,580
5 6.7 4,519 170 34,724,500 NA NA 2.30E-94 34,724,500
69,362,250
23 6.8 5,645 170 60,000,000 1.04E+08 9.16E+07 5.37E-104 1.04E+083
4 6.7 6,602 100 50,000,000 5.49E+09 5.52E+09 9.25E-154 5.49E+093
2.90E+09
13 7.4 5,059 100 50,000,000 3.00E+08 1.04E+11 6.37E-164 3.00E+083
Notes:
1
Failure extrapolated. Testing suspended at 58% of initial stiffness at 57,000 due to computer problem.
2
Software froze, apparently due to error writing to network drive. Sample stiffness 3,439 MPa, at 53.4% of initial stiffness. Result extrapolated using
linear regression of latter cycles.
3
Results extrapolated using single-stage Weibull model.
4
Less than 8.57E-9 proposed by Shen and Carpenter (40) as indicative of long-life pavement.
 33

Table 4.4. Granite 19.0 mm NMAS mix with PG 76-22 at optimum asphalt.

Beam Air Initial Micro- Cycles Extrapolated Cycles to 50% PV Cycles to 50% Average
ID Voids, Flexural Strain Tested Initial Stiffness Initial Cycles
% Stiffness, Single-Stage Three-Stage Stiffness to Failure
MPa Weibull Weibull
4 7.2 3,025 800 42,240 NA NA 4.02E-06 42,240
26,160
7 6.7 5,445 800 10,080 NA NA 1.58E-05 10,080
2 7.1 4,191 400 3,609,470 NA NA 1.66E-08 3,609,470
8 6.6 4,976 400 591,770 NA NA 2.87E-07 591,770 1,664,400
13 7.2 3,675 400 791,960 NA NA 2.15E-07 791,960
1 7.3 4637 250 14,837,450 NA NA 3.30E-08 14,837,450
NA
11 6.8 4148 250 50,000,000 2.91E+09 1.31E+15 2.22E-182 2.91E+091
5 6.7 4,460 200 50,000,000 2.75E+09 1.53E+11 0.00E+002 2.75E+091
2.72E+09
3 7.1 4,062 200 50,000,000 2.68E+09 2.61E+21 0.00E+002 2.68E+091
Notes:
1
Results extrapolated using single-stage Weibull model.
2
Less than 8.57E-9 proposed by Shen and Carpenter (40) as indicative of long-life pavement.

Table 4.5. Granite 19.0 mm NMAS mix with PG 67-22 at optimum plus asphalt.

Beam Air Initial Micro- Cycles Extrapolated Cycles to 50% PV Cycles to 50% Average
ID Voids, Flexural Strain Tested Initial Stiffness Initial Cycles
% Stiffness, Single-Stage Three-Stage Stiffness to Failure
MPa Weibull Weibull
Cox & Sons fixture in Interlaken Load Frame, except as noted
8 3.0 5,054 800 15,464 NA NA 15,464
24,982
14 3.2 5,306 800 34,500 NA NA 34,500
10 3.2 5,896 400 468,343 NA NA 468,343
403,232
15 3.3 6,698 400 338,121 NA NA 338,121
9 3.4 6,094 200 10,000,000 24,944,621 1.14E+08 24,944,6211
42 3.5 6,923 200 38,985,510 NA NA 38,985,510 62,310,044
13 3.8 6,219 200 50,000,000 1.23E+08 9.95E+07 1.23E+081
IPC Global fatigue device
6 4.7 6,862 800 5,570 NA NA 4.17E-05 5,570
5,400
3 4.1 7,472 800 5,230 NA NA 3.99E-05 5,230
7 5.1 7,675 400 131,390 NA NA 1.49E-06 131,390
94,615
4 4.9 7,653 400 57,840 NA NA 6.26E-06 57,840
2 4.7 7,512 200 3,584,740 NA NA 1.58E-07 3,584,740 3,584,740
6a 3.3 8,605 100 15,350,090 5.81E+08 NA4 NA 5.81E+081 5.81E+08
Notes:
1
Results extrapolated using single-stage Weibull model.
2
Testing conducted by Rutgers University on an IPC Global fatigue device.
3
Tested on Asphalt Institute IPC Global fatigue device.
4
No solution.

Table 4.6. Granite 19.0 mm NMAS mix with PG 76-22 at optimum plus asphalt.

Beam Air Initial Micro- Cycles Extrapolated Cycles to 50% PV Cycles to 50% Average
ID Voids, Flexural Strain Tested Initial Stiffness Initial Cycles
% Stiffness, Single-Stage Three-Stage Stiffness to Failure
MPa Weibull Weibull
8 3.7 3,520 800 252,450 NA NA 2.61E-07 252,4501
5 3.0 5,451 800 32,520 NA NA 3.12E-06 32,520 48,050
14 3.3 5,764 800 63,580 NA NA 1.09E-06 63,580
1 2.8 5,532 400 2,860,000 NA NA 1.17E-08 2,860,000
6,257,500
4 3.0 5,532 400 9,655,000 NA NA 2.05E-093 9,655,000
10 3.6 4,308 300 39,624,000 NA NA 1.71E-093 39,624,000
2 3.5 5,427 300 8,811,8104 4.88E+7 5.63E+09 1.84E-133 4.88E+72
7.57E+07
12 3.5 4,105 300 20,080,7504 1.47E+8 2.46E+10 7.33E-173 1.47E+82
11 4.0 5,162 300 50,000,000 6.75E+7 4.50E+09 1.25E-133 6.75E+72
13 3.1 6,841 200 50,000,000 5.96E+9 6.79E+11 2.22E-183 5.96E+92
1.85E+10
9 3.1 5,609 200 50,000,000 3.10E+10 1.58E+16 0.00E+003 3.10E+102
Notes:
1
Not included in average.
2
Results extrapolated using single-stage Weibull model.
3
Less than 8.57E-9 proposed by Shen and Carpenter (40) as indicative of long-life pavement.
4
Sample did not fail, extrapolated using single-stage Weibull model.
34

PG 67-22 Mix at Optimum Asphalt Content Figure 4.17 shows a log-log plot of cycles to failure versus
strain. For samples that did not fail within 50 million cycles,
The results for the PG 67-22 mix tested at optimum asphalt
extrapolations are shown using single-stage Weibull, three-
content were presented in Table 4.3. The PG 67-22 mix at opti-
stage Weibull, and RDEC. The data from 800 through 170 ms
mum asphalt content was tested by NCAT. The extrapolations
were used to fit the regression line. A fatigue life of 60 million
shown in Table 4.3 are based on the single- and three-stage
cycles was assigned to Sample 23, tested at 170 ms (the actual
Weibull functions, as well as the RDEC. Sample 23, tested at
number of cycles tested). Ninety-five percent confidence
170 ms, produced Nf of 1.04E+08 and 9.16E+07 using the
limits are shown for the regression line and 95% prediction
single- and three-stage Weibull functions, respectively. Sam-
intervals are shown from 170 through 50 ms. Although the
ples 4 and 13, tested at 100 ms produced extrapolated Nf
three-stage Weibull extrapolation of Nf for Samples 23 and 4
using the single-stage Weibull function of 5.49E+09 and
fall on the upper prediction limit, the fatigue life estimate at
3.00E+08, respectively. Although both of these numbers rep-
100 ms for Sample 13 indicates a deviation from the log-log
resent extraordinarily long fatigue lives, Table 4.3 indicates
that Sample 13 would be expected to have a longer fatigue life. regression line, which in turn indicates the existence of an
The single-stage Weibull function fits for Samples 4 and 13 endurance limit between 100 and 170 ms.
was previously presented in Figure 4.4. As shown in Figure 4.4, Recall that 170 ms was selected to produce a beam fatigue
the slope of the data decreases above approximately 1 million life of 50 million cycles, or approximately 500 million load
loading cycles, indicating less damage. The best fit line for the repetitions in the field. Based on Sample 23’s deviation
Weibull function for Sample 13 has a steeper slope, resulting from the prediction limits in Figure 4.17, this strain level
in the prediction of a shorter fatigue life. appears to be close to the endurance limit, but slightly high.
Tsai et al. (57) developed a three-stage Weibull function to Nf was substituted in the regression as the predictor for strain
more accurately model the changes in slope observed in the level and the regression re-run, resulting in Equation 17, as
data. As discussed previously, three regions can be observed follows:
with the Weibull function. In the third region, damage can
ε = 103.54 × ( N )
−0.170
either increase rapidly—leading to failure—or decrease as (17)
observed for Sample 13. The three-stage Weibull results for
Sample 13 are shown in Figures 4.15 and 4.16. The three- Ninety-five percent prediction limit, in terms of strain,
stage Weibull function resulted in Nf predictions of 5.52E+09 was calculated for N = 50 million cycles. The lower predic-
and 1.04E+11 for Samples 4 and 13, respectively. tion limit for Equation 17 at 50 million cycles was 151 ms.

-2
Ln (-Ln(Stiffness Ratio))

-4

-6
Raw Data
Fit
-8

-10

-12
0 2 4 6 8 10 12 14 16 18 20
Ln (Loading Cycles)

Figure 4.15. Three-stage Weibull function for Sample 13, PG 67-22 at optimum.
 35

6,000

5,000

Stiffness, MPa 4,000

3,000
Raw Data
Fit
2,000

1,000

0
0 10,000,000 20,000,000 30,000,000 40,000,000 50,000,000 60,000,000
Loading Cycles

Figure 4.16. Three-stage Weibull fit for Sample 13, PG 67-22 at optimum.

By using the lower prediction limit, the strain level resulting pavements. Although all three strain levels appear to provide
in 50 million cycles should be below the endurance limit for a long fatigue life, 170 ms appears to be at, or slightly above,
the PG 67-22 mix at optimum asphalt content. the endurance limit based on the other analyses. The recom-
The RDEC plateau values were calculated for each of the mended plateau value may not define the endurance limit,
PG 67-22 at optimum samples. The results are shown in but rather a long fatigue life.
Table 4.3. The samples tested at 200, 170, and 100 ms produce Figure 4.18 shows the relationship between cycles to failure
plateau values lower than the critical value, 8.57E-9 recom- and plateau value. The relationship for low strain tests devel-
mended by Shen and Carpenter (40) as indicative of long-life oped by Shen and Carpenter (40) based on testing 602 beams

1000

R² = 0.99 Sam ple 23

Sa mp le 4

100
Micro-Strain

Sa mp le 13
10

1
1 100 10000 1000000 100000000 1E+10 1E+12 1E+14
Cycles to Failure (50% Stiffness)

Measured Single-Stage Weibull Function

3-Stage Weibull Function RDEC
Confidence Limits Prediction Limits

Figure 4.17. Cycles to failure versus strain for PG 67-22 at optimum.

36

1
0.1
0.01
y = 0.2937x-1.05
0.001
0.0001 R2 = 0.999
1E-05

Plateau Value
1E-06
1E-07
1E-08
1E-09
1E-10
1E-11
1E-12
1E-13 y = 0.2871x -1.0793
1E-14
1E-15
1 100 10000 1000000 10000000 1E+10 1E+12 1E+14
0
Cycles to 50% Initial Stiffness

NCAT NCHRP 9-38 Shen and Carpenter

Power (NCAT NCHRP 9-38) Power (Shen and Carpenter)

Figure 4.18. Cycles to failure versus plateau value.

is shown for comparison. The relationships are very similar. Weibull function. Extrapolations were also conducted using
This would support Shen and Carpenter’s proposal that there the three-stage Weibull function and RDEC. Sample 2 was
is one relationship between cycles to failure and plateau value evaluated as a potential outlier using the repeatability data
for all mixes, regardless of the manner of testing. developed in Phase II. The acceptable difference between two
results is estimated to be 0.69 (on a log basis), while the differ-
ence between Sample 2 and Sample 13 is 0.66 on a log basis.
PG 76-22 at Optimum Asphalt Content
This indicates that Sample 2 is within acceptable variation.
The results for the PG 76-22 mix tested at optimum asphalt Sample 2 increased the variability of the data, producing an
content were presented in Table 4.4. The PG 76-22 mix at R = 0.92 and resulting in larger prediction and confidence
2

optimum asphalt content was tested by NCAT. The extrap- intervals (Figure 4.19). However, both points at 200 ms and
olations shown in Table 4.4 are based on the single-stage one point at 250 ms indicate a deviation from the log-log plot

1000
Sample 11 Sample 5

R² = 0.92
100
Micro-Strain

Sample 3

10

1
1.E+00 1.E+02 1.E+04 1.E+06 1.E+08 1.E+10 1.E+12 1.E+14 1.E+16 1.E+18 1.E+20 1.E+22

Cycles to Failure (50% Stiffness)

Measured Three-Stage Weibull Single-Stage Weibull

95% Confidence Interval 95% Prediction Interval

Figure 4.19. Cycles to failure versus strain for PG 76-22 at optimum.

 37

N f = 2.738 × 105 exp0.077 VFA ( ε 0 ) (S0′′)−2.720

of cycles to failure versus strain, indicative of the endurance −3.624
(18)
limit. Similar to the PG 67-22 results at optimum, the regres-
sion was reversed to solve for the strain level that would produce
where,
50 million and 100 million cycles. The lower 95% prediction
Nf = fatigue life,
interval indicated a strain level of 146 ms to produce 50 mil-
VFA = voids filled with asphalt, percent,
lion cycles to failure. This strain level is slightly lower than
ε0 = initial strain, and
that determined for the PG 67-22 at optimum even though
S″0 = initial loss stiffness, psi.
the endurance limit appears to be at a higher value (between
200 and 250 ms). This is due to the increased variability in
Table 4.7 shows the predicted fatigue lives using Equa-
the testing.
tion 18 and the actual and predicted percent difference between
One sample at 250 ms and both samples at 200 ms pro-
the two sets of beams. The SHRP A-404 surrogate fatigue
duced plateau values less than the critical value indicated by
equation fairly accurately predicts the percentage reduction
Shen and Carpenter to be indicative of long fatigue life.
in fatigue life at 800 ms (estimated 70% versus actual 78%).
It underestimated the reduction at the lower strain levels. All
PG 67-22 at Optimum Plus Asphalt Content of the fatigue lives predicted using the SHRP A-404 surrogate
Testing for the PG 67-22 mix at optimum plus asphalt con- model are considerably lower than the measured values. Sen-
tent was initially conducted on the Asphalt Institute’s Inter- sitivity analyses indicated that most of the effect was due to the
laken servo-hydraulic frame using a Cox & Sons fixture. Due increased initial stiffness. For the PG 76-22 mixes, initial stiff-
to problems with the Interlaken system, low-strain beams ness increased with the lower air voids determined for the
were later tested on IPC Global beam fatigue devices operated optimum plus samples. Comparisons between the asphalt
by both the Asphalt Institute and Rutgers University. Due to content and gradation of randomly selected beams indicated
concerns about possible differences caused by the various no significant differences. There were also no differences in
machines, it was decided to retest the cells using the Asphalt the environmental chamber temperature of measured stiff-
Institute’s IPC Global beam fatigue device. The initial results ness of a plastic test beam to support the change in initial stiff-
from the three machines and the retests using the IPC Global ness. Therefore, it is felt that the difference in initial stiffness
machine were presented in Table 4.5. must be attributed to the different binders used to produce
The fatigue lives for the retests are significantly shorter the beams.
than for the original mix. A number of factors appear to con- Figure 4.20 shows a log-log plot of cycles to failure versus
tribute to this difference. The initial stiffness for the original strain. Sample 1 from the first set at 200 ms and Sample 6a
set of beams averaged 6,027 MPa; the initial stiffness of the from the second set at 100 ms show deviations indicative of
replacement beams averaged 7,435 MPa. The beams were the endurance limit from their respective best-fit lines. Based
prepared at air voids contents outside of the tolerance for the on the first data set, the 95% lower prediction interval for a
optimum plus target (3.3 ± 0.5%) and used NCAT’s lab stock fatigue life of 50 million cycles is 158 ms, which is essentially
PG 67-22 binder instead of the dual graded PG 64/67-22 the same as that determined for the PG 67-22 at optimum
binder, which was used at the NCAT Test Track. (151 ms).
The SHRP A-404 surrogate fatigue model (Equation 18)
(18) was used to assess whether the differences in initial stiff-
PG 76-22 at Optimum Plus Asphalt Content
ness and voids filled with asphalt (VFA) would tend to cause
the degree of observed difference in the measured fatigue The results for the PG 76-22 mix tested at optimum plus
lives. VFA would be lower due to the higher air voids. asphalt content were presented in Table 4.6. The PG 76-22

Table 4.7. Fatigue life predictions based on SHRP A-404

surrogate model.
Binder Micro- Initial VFA Predicted Predicted Measured
Strain Stiffness, Nf Percent Percent
psi Reduction Reduction
First Set 800 751,307 76 1,659
70 78
Second Set 800 1,039,502 72 504
First Set 400 913,317 76 12,027
57 77
Second Set 400 1,111,587 72 5,180
First Set 200 929,996 76 141,164
52 94
Second Set 200 1,089,540 72 67,442
38

1000
R² = 0.96

R² = 0.98
10 0

Micro-Strain
Sample 1, First Set
Sample 6a, Second Set

10

1
1

10

10

10

10

1E
0

00

00

00

+1
0

00

00

0
0

00
0
Cycles to Failure (50% Stiffness)

First Set Measured Three-Stage Weibull Single-Stage Weibull

Figure 4.20. Cycles to failure versus strain for PG 67-22 at optimum plus.

mix at optimum asphalt content was tested by NCAT. The Summary of Phase I Observations
extrapolations shown in Table 4.6 are based on the single- Regarding Endurance Limit
stage Weibull function. Extrapolations also were conducted
using the three-stage Weibull function and RDEC. Figure 4.21 Clear indications of the endurance limit were shown for
shows the log-log plot of cycles to failure versus strain. Although three of four mixes (not PG 67-22 at optimum plus). The strain
a deviation from the regression line is first indicated at 300 ms, level corresponding to the endurance limit appears to be mix
particularly for the three-stage Weibull and RDEC extrapola- dependent. Visually, the endurance limit appears to be more
tions, it is not clear until 200 ms. The 95% lower confidence sensitive to binder properties than to asphalt content/air void
limit for the endurance limit using the methodology described content. An endurance limit (predicted value, not lower pre-
in Appendix A is 200 ms. This represents an increase as com- diction interval) of approximately 170 ms was determined for
pared to both the PG 76-22 at optimum and PG 67-22 at opti- the PG 67-22 mix at optimum asphalt content. The endurance
mum plus mixes. limit for the PG 76-22 mixture appears to be on the order of

1000

R² = 0.93
100
Micro-Strain

10

1
1 100 10000 1000000 100000000 1E+10 1E+12 1E+14 1E+16
Cycles to Failure (50% Stiffness)

Measured Three-Stage Weibull Single-Stage Weibull RDEC

Figure 4.21. Cycles to failure versus strain for PG 76-22 at optimum plus.
 39

Table 4.8. Granite 19.0 mm NMAS mix with PG 58-28 at optimum asphalt.
Sample Air Initial Micro- Cycles Extrapolation Cycles to 50% Cycles to Average
Voids, Stiffness, Strain Tested Stiffness 50% Cycles to
% MPa Single-Stage Three-Stage Stiffness Failure
Weibull Weibull
4 6.8 3,216 800 12,730 NA NA 12,730 10,097
8 7.2 3,014 800 8,730 NA NA 8,730
9 7.0 2,974 800 8,830 NA NA 8,830
2 6.9 3,372 400 166,290 NA NA 166,290 183,793
6 7.4 3,424 400 148,090 NA NA 148,090
7 7.4 3,424 400 237,000 NA NA 237,000
5 7 4,217 76 12,000,000 1.01E+09 1.11E+08 1.01E+091 1.03E+09
15 7 4,332 76 12,000,000 1.57E+09 5.30E+09 1.57E+091
16 7 4,706 76 12,000,000 5.04E+08 4.39E+09 5.04E+081
Note:
1
Extrapolated using single-stage Weibull model.

220 ms, and approximately 300 ms for the PG 76-22 at opti- Phase II Testing to Investigate
mum plus. The plateau value criteria determined by Shen and Additional Binder Grades
Carpenter (40) appears to be indicative of very long fatigue
life, but not necessarily the endurance limit. Testing was conducted in Phase II to evaluate additional
The Weibull function appears to be the best technique for binder grades. The 19.0 mm NMAS granite test track mix was
extrapolation of low strain stiffness results. One rapid technique replicated using a true grade PG 64-22 and PG 58-28. Three
for determining the endurance limit may be to test three repli- beams were tested at 800 ms and three beams were tested at
cates at each of two strain levels, nominally 800 and 400 ms 400 ms for each mixture. The fatigue testing was conducted
and then fit a log-log relationship between cycles to failure and by NCAT. The endurance limit was estimated using the one-
strain. Testing additional samples at normal strain levels would sided 95% lower prediction interval for a strain level corre-
potentially reduce the estimate of the standard deviation sponding to 50 million cycles according to the methodology
and therefore increase the confidence in the prediction. The described in Appendix A. The lower 95% prediction inter-
predicted endurance limit for the PG 76-22 mix at optimum vals were 82 and 75 ms, respectively, for the PG 58-28 and
asphalt content would most likely benefit from testing addi- PG 64-22 mixtures. Confirmation tests for the endurance
tional samples. The lower 95% prediction limit for a fatigue limit of the mixture using the PG 58-28 binder were carried
life of 50 million cycles appears to be reasonably close to the out at 76 ms due to an error in the t-value used in determin-
endurance limit. This technique was originally presented in ing the lower prediction limit. The tests should have been
the Proposed Standard Practice for Predicting the Endurance carried out at 82 ms. The test data is presented in Tables 4.8
Limit of Hot Mix Asphalt (HMA) for Long-Life Pavements and 4.9 and shown graphically in Figures 4.22 and 4.23.
in Appendix A. This technique was used for the testing con- For the PG 58-28 mixture, the single-stage Weibull extrap-
ducted in Phase II. Appendix A was later modified in an effort olations for the samples tested at 76 ms indicate fatigue lives
to obtain a more accurate estimate of the endurance limit. that are longer than that predicted from the log-log regression,

Table 4.9. Granite 19.0 mm NMAS mix with PG 64-22 at optimum asphalt.
Sample Air Initial Micro- Cycles Extrapolation Cycles to 50% Cycles to 50% Average
Voids, Stiffness, Strain Tested Stiffness Stiffness Cycles to
% MPa Single-Stage Three-Stage Failure
Weibull Weibull
5 7.5 3,635 800 5,580 NA NA 5,580 5,377
6 7.5 3,736 800 5,060 NA NA 5,060
8 7.5 4,234 800 5,490 NA NA 5,490
2 7.3 4,666 400 98,120 NA NA 98,120 95,377
4 7.5 4,449 400 111,250 NA NA 111,250
9 7.4 4,227 400 76,760 NA NA 76,760
3 6.6 5,190 75 12,000,000 8.18E+10 1.08E+09 8.18E+101 3.95E+10
7 6.7 4,667 75 12,000,000 3.64E+10 3.18E+08 3.64E+101
10 6.7 5,989 75 12,000,000 2.82E+08 2.82E+08 2.82E+081
Note:
1
Extrapolated using single-stage Weibull model.
40

10000

1000

Micro-Strain
100
R² = 0.98

10

1
1 100 10000 1000000 100000000 1E+10
Cycles to Failure (50% Stiffness)

Measured Three-Stage Weibull Single-Stage Weibull

Confidence Limits Prediction Limits

Figure 4.22. Cycles to failure versus strain for PG 58-28 at optimum

asphalt content.

but within the prediction limits for the extrapolation. Two of prediction limits for the log-log regression line. This is a clear
the three-stage Weibull extrapolations exceed the prediction indication of the endurance limit.
limits, but Sample 16 indicates a shorter fatigue life. In general, the predicted endurance limits for the PG 58-28
The single- and three-stage Weibull fatigue life extrapola- and PG 64-22 binders were lower than what might have been
tions for the PG 64-22 mixture samples exceeded the fatigue expected based on the Phase I testing. Historically, softer
lives estimated from the log-log plot of cycles to failure ver- binders are believed to perform better in constant strain tests
sus strain. The extrapolated fatigue lives also exceeded the (see Table 2.1).

10000

1000
Micro-Strain

R² = 0.99

100

10

1
1.0E+00 1.0E+01 1.0E+02 1.0E+03 1.0E+04 1.0E+05 1.0E+06 1.0E+07 1.0E+08 1.0E+09 1.0E+10 1.0E+11
Loading Cycles
Measured Three-Stage Weibull Single-Stage Weibull Confidence Limit
Series5 Prediction Limit Series7 Power (Measured)

Figure 4.23. Cycles to failure versus strain for PG 64-22 at optimum asphalt content.
 41

Estimate of Precision optimum, PG 67-22 at optimum plus, and PG 76-22 at opti-

of Beam Fatigue Tests mum asphalt content mixes. The strain levels used for the
beams tested to evaluate the low strain extrapolations were
In Phase II, a small-scale round-robin was conducted to 130 and 220 ms, respectively, for the PG 67-22 and PG 76-22
develop precision estimates for the beam fatigue test and the mixes at optimum asphalt content. The differences arose
extrapolation procedures. ASTM C802 recommends a mini- since not all labs had reported their data when the confirma-
mum of 10 labs for a round-robin study. Due to the testing tion strain levels were selected.
time commitments involved with beam fatigue testing, only The round-robin data were analyzed according to ASTM
five laboratories, three represented by the research team, and E691. When analyzing round-robin data, it is desirable to have
two volunteers, participated. The volunteer labs only tested a constant standard deviation or constant coefficient of varia-
one mix at three strain levels. The experimental matrix for the
tion at the various test (strain) levels. To achieve this, the pre-
mini round-robin is shown in Table 4.10.
cision estimates are based on the log base 10 of the actual test
Variability is expected to consist of four components:
results. This is reasonable since fatigue transfer functions uti-
materials, sample preparation, beam testing, and analysis. To
lized in pavement design are logarithmic.
minimize materials variability, NCAT prepared and batched
Three potential outliers were identified based on the h and k
all of the aggregates. SEM Materials and the University of
statistics utilized in ASTM E691. Only one outlier was removed
California participated on a voluntary basis.
from the data set, Lab 2’s results for the PG 67-22 at optimum
The directions provided for preparation of the test samples
asphalt content samples tested at 120 ms and extrapolated using
are presented in Appendix F. Each lab mixed and compacted
the logarithmic model.
their own beams for testing. The aggregate batches were ran-
domized prior to shipping. In some cases, however, additional The precision estimates, without the outlier, are presented
batches were required in order to obtain beams with the appro- in Tables 4.11 and 4.12 for the normal strain and extrapo-
priate air void levels. The samples were short-term oven aged lated data, respectively. The complete data are presented in
for 4 h at 275°F (135°C) according to AASHTO R30 prior to Appendix D. An examination of Table 4.11 suggests that the
compaction. The range of equipment used to compact the repeatability and reproducibility standard deviations for the
beams included linear kneading compactors, vibratory com- normal strain tests are consistent for the different mixes and
pactors, and rolling wheel compactors. strain levels. Based on the data in Table 4.11, the log of two
Testing was conducted in accordance with the draft prac- properly conducted tests by the same operator should not dif-
tice for the determination of the endurance limit, which has fer by more than 0.69 with 95% confidence. Similarly, the log
since been modified (Appendix A). Fatigue lives were also of two properly conducted normal strain tests by two differ-
extrapolated using the logarithmic and RDEC described in ent laboratories should not differ by more than 0.89.
Appendix C. Each lab first tested three beams each at 800 and For the extrapolation methods shown in Table 4.12, the
400 ms. Extrapolations were conducted using each lab’s data single-stage Weibull function produces the least variable
to estimate the endurance limit. The average estimated en- results, followed by the logarithmic model. The RDEC pro-
durance limit for all of the participating labs was determined. cedure produces the most variable results. The use of the
Using a single strain level for the endurance limit allowed re- unique plateau value versus cycles to failure line proposed
peatability and reproducibility calculations to be performed in Appendix C appeared to create erroneously low fatigue
on the fatigue life extrapolations. Each lab then tested three lives with one of the data sets and exceptionally long lives
beams at the average estimated endurance limit to 12 million with one of the other lab’s data. Based on the potential for
cycles. Low strain beams were not tested for the PG 67-22 mix overestimating fatigue life and the associated variability in
at optimum plus. The average estimated endurance limits calculation, the RDEC model is not recommended for extrap-
were 151, 175, and 188 ms, respectively, for the PG 67-22 at olating endurance limit data.

Table 4.10. Mini round-robin testing matrix.

Lab/Mix PG 67-22 at PG 67-22 at PG 76-22 at
Optimum Optimum Optimum
Plus
NCAT X X X
Asphalt Institute X X X
University of Illinois X X X
SEM Materials X
University of California X
42

Table 4.11. Summary of normal strain round-robin results.

Code No. Log of Standard Repeatability Reproducibility Between- Within-Lab Between-
of Average Deviation Standard Standard Lab Coefficient Lab
Labs of All between Deviation (Sr) Deviation (SR) Standard of Coefficient
Labs Log of Deviation Variation, of
Cell of Log of % Variation,
Averages Lab %
(Sx) Means
(SL)
PG 67-22 Opt. at 800 ms 7 3.876 0.220 0.249 0.300 0.167 6.4 7.7
PG 67-22 Opt. at 400 ms 7 5.370 0.365 0.240 0.414 0.338 4.5 7.7
PG 76-22 Opt. at 800 ms 3 3.932 0.127 0.261 0.261 0.000 6.6 6.6
PG 76-22 Opt. at 400 ms 3 5.624 0.299 0.295 0.384 0.246 5.2 6.8
PG 67-22 Opt.+ at 800 ms 3 4.203 0.107 0.207 0.207 0.000 4.9 4.9
PG 67-22 Opt.+ at 400 ms 3 5.717 0.134 0.243 0.243 0.000 4.2 4.2
Pooled 0.234 0.248 0.318 0.164 5.4 6.8

Indirect Tensile Strength as Samples of the PG 67-22 mixtures at optimum and opti-
a Surrogate for Endurance mum plus asphalt content and PG 76-22 mixtures at optimum
Limit Determination and optimum plus asphalt content were compacted in the
Superpave Gyratory Compactor (SGC) for IDT testing to
Indirect Tensile Strength (IDT)—or, more correctly, the the same air void levels used for the beam fatigue tests. The
tensile strain at failure from the IDT test—was examined samples were tested in the IDT at 25°C. Testing was conducted
as a surrogate for beam fatigue tests to identify the fatigue according to AASHTO T322. Tensile strain was calculated as
endurance limit. The Asphalt-Aggregate Mixture Analysis described by Kim and Wen (60). Strain was calculated at 98%
System (AAMAS) used a mixture’s resilient modulus and ten- of the peak stress, and the results are shown in Figure 4.24
sile strain at failure from the IDT test to assess fatigue resist- versus the predicted and 95% lower confidence limits for the
ance (59). In the AAMAS system, testing was conducted at the endurance limit determined from the beam fatigue tests.
following three temperatures: 5°C, 25°C, and 40°C. The mea- From Figure 4.24, it is apparent the predicted and 95% lower
sured resilient modulus and tensile strain at failure were com- confident interval for the endurance limit are approximately
pared to the properties of a “standard” mix, the dense-graded 5 and 3%, respectively, of the indirect tensile failure strain.
mix used at the AASHO Road Test. Von Quintus (personal However, although the indirect tensile test appears to be sen-
communication) suggested that long-life pavements be designed sitive to the two different binders, it does not appear to be
with tensile strains at the bottom of the asphalt layer that sensitive to binder content. This procedure appears to have
were < 1% of the tensile strain at failure. Maupin and Freeman some potential to predict the magnitude of the endurance
(9) demonstrated satisfactory correlations between constant limit. Additional work is necessary with a broader range of
strain fatigue-life curves and indirect tensile test results. materials.

Table 4.12. Summary of round-robin extrapolations at the estimated endurance limit.

Code Number Log of Standard Repeatability Reproducibility Between- Within- Between-
of Labs Average Deviation Standard Standard Lab Lab Lab
of All between Deviation Deviation (SR) Standard Coefficient Coefficient
Labs Log of Cell (Sr) Deviation of of
Averages of Log of Variation, Variation,
(Sx) Lab % %
Means
(SL)
Logarithmic
PG 67-22 Opt. at 130 ms 5 10.609 1.642 1.322 1.965 1.454 12.5 18.5
PG 76-22 Opt. at 220 ms 3 8.209 0.441 1.207 1.207 0.000 14.7 14.7
Pooled 1.192 1.279 1.681 0.909 13.3 17.1
Weibull
PG 67-22 Opt. at 130 ms 5 8.794 0.915 0.586 1.032 0.850 6.7 11.7
PG 76-22 Opt. at 220 ms 3 7.563 0.309 0.719 0.719 0.000 9.5 9.5
Pooled 0.688 0.636 0.915 0.531 7.7 10.9
RDEC
PG 67-22 Opt. at 130 ms 4 10.275 3.421 2.341 3.919 3.143 22.8 38.1
PG 76-22 Opt. at 220 ms 3 9.607 0.967 2.151 2.151 0.000 22.4 22.4
Pooled 2.369 2.259 3.161 1.796 22.6 31.4
 43

350

300
y = 0.047x

Endurance Limit, Micro-Strain

R² = 0.650
250

200

150

100
y = 0.031x
R² = 0.400
50

0
0 1000 2000 3000 4000 5000 6000
Tensile Strain at Failure, Micro-Strain
Predicted Endurance Limit 95% Lower Confidence Endurance Limit
Linear (Predicted Endurance Limit) Linear (95% Lower Confidence Endurance Limit)

Figure 4.24. IDT tensile strain at failure versus beam fatigue endurance limit.
44

CHAPTER 5

Uniaxial Tension Results and Analyses

The testing performed at the University of New Hampshire Dynamic Modulus and Phase Angle
(UNH) includes complex modulus, monotonic, and fatigue Master Curves
tests in uniaxial tension. Analysis of the data is done using
Uniaxial complex modulus tests were performed on at least
viscoelastic and continuum damage mechanics principles to
three specimens from each mixture. The testing was done at the
identify the fatigue endurance limit of the PG 67-22 opti-
following five temperatures: −10⬚C, 0⬚C, 10⬚C, 20⬚C, and 30⬚C
mum, PG 67-22 optimum plus, PG 76-22 optimum, and
and at eight frequencies: 0.1, 0.2, 0.5, 1, 2, 5, 10, and 20 Hz.
PG 76-22 optimum plus mixtures. Frequency sweeps at var-
The dynamic modulus and phase angle values at each
ious temperatures are run to measure the complex dynamic
frequency and temperature were calculated from stresses
modulus of each mix. The dynamic modulus and phase angle and LVDT measured strains. The individual dynamic mod-
master curves are then constructed from these data and the ulus curves were shifted along the frequency axis to construct
relaxation master curve is obtained through linear viscoelas- the dynamic modulus master curve using the time-temperature
tic conversion. The damage characteristic curve of each mix superposition principle. The reference temperature was selected
is obtained by running uniaxial monotonic tests to failure or as 20⬚C. The resulting master curve for dynamic modulus (|E∗|)
by running constant amplitude fatigue tests to failure. The is expressed by the following sigmoidal equation:
characteristic damage curve is used to predict the number of
cycles to failure at different strain amplitudes to determine b
log Eⴱ = a + (19)
the fatigue endurance limit of the mixture. 1 + exp ( −c − d log fr )
Fatigue tests with increasing strain amplitude are run in uni-
axial tension to directly identify the fatigue endurance of the where,
mixtures. The testing and analysis procedures are described in a, b, c, and d are positive regression coefficients.
the following sections, starting with the dynamic modulus
tests, followed by monotonic tests, fatigue tests, and finally the The term fr represents the reduced frequency, given by
prediction of endurance limit.
log fr = log f + log aT (20)

Test Specimens where,

aT is the shift factor which is a function of temperature and
A summary of the air void content and testing performed f is the actual frequency at which the individual curves are
on each specimen at UNH is shown in Table 5.1. All test spec- obtained.
imens were fabricated at NCAT using an SGC and shipped to
UNH, where they were cut and cored to 75-mm diameter, The solver function in Microsoft Excel is used to simul-
150-mm tall specimens. Steel plates were glued to the ends of taneously determine the sigmoidal fit coefficients and the
the uniaxial specimens using plastic epoxy glue in a gluing jig shift factors for the individual specimens using error min-
designed to align the specimen vertically. Four LVDTs spaced imization. An overall mixture master curve is determined
90⬚ apart around the circumference of the specimen were by fitting the sigmoidal function in Equation 19 to all of the
attached to the surface of the specimen using a 100-mm gage individual specimen master curves. Figures 5.1 through 5.4
length. show the individual and overall dynamic modulus master
 45

Table 5.1. Summary of specimens and tests conducted.

Specimen Air |E*| Monotonic

Mix Fatigue Test Comments
No. Voids % Test Test
(After
Constant Increasing
Core and
Amplitude Amplitude
Cut)
1 8.6 Broke
during |E*|
testing
2 7.0 X
3 7.3 Broke
during |E*|
testing
4 7.3 X
5 7.5 X
6 6.8 X
7 7.7 X
8 7.2 X
PG 67-22,
Optimum 9 7.7 X
10 7.5 X
11 7.7 X X
12 7.0 Broke
during |E*|
testing
13 7.7 X X
14 6.9 X
15 6.9 X
16 6.7 X
17 6.7 X
18 6.6 X
1 6.5 Broke
during |E*|
testing
PG 76-22, 2 6.9 X X
Optimum 3 6.1 X X
4 7.0 X X
5 6.2 X X
1 2.2 X X
2 1.5 X X
3 2.2 X X Monotonic
test data not
usable
4 1.3 X X
5 2.1 X X
PG 67-22, 6 3.6 X X
Optimum+ 7 3.6 Broke
during |E*|
testing
8 2.7 X X
9 3.0 Broke
during |E*|
testing
10 3.0 X X
1 1.5 X X
2 1.3 X
3 1.9 X X
4 1.2 X X
5 1.0 Broke
PG 76-22, during |E*|
Optimum+ testing
6 3.1 X X
7 3.2 X X
8 2.7 X X
9 3.7 X X
10 2.8 X X
46

50000
Master Curve at 20°C

Dynamic Modulus, MPa

40000
Individual Specimens

30000

20000

10000

0
1.E-03 1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
Reduced Frequency, Hz

Figure 5.1. Dynamic modulus master curves for PG 67-22 optimum.

50000
Individual Specimens
Dynamic Modulus, MPa

40000 Master Curve at 20°C

30000

20000

10000

0
1.E-03 1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
Reduced Frequency, Hz

Figure 5.2. Dynamic modulus master curves for PG 76-22 optimum.

50000
Master Curve at 20°C

40000 Individual Specimens

Dynamic Modulus, MPa

30000

20000

10000

0
1.E-03 1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
Reduced Frequency, Hz

Figure 5.3. Dynamic modulus master curves for PG 67-22 optimum plus.
 47

50000
Mix Master Curve at 20°C

Dynamic Modulus, MPa

40000
Individual Specimens

30000

20000

10000

0
1.E-03 1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
Reduced Frequency, Hz

Figure 5.4. Dynamic modulus master curves for PG 76-22 optimum plus.

50000
67-22, Opt
76-22, Opt
Dynamic Modulus, MPa

40000
67-22, Opt+
30000 76-22, Opt+

20000

10000

0
1.E-03 1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
Reduced Frequency, Hz

Figure 5.5. Dynamic modulus master curves for all mixes tested.

curves obtained for the four mixtures. Figure 5.5 shows the curve for each specimen. The phase angle master curves are
overall dynamic modulus curves (obtained from fit) for all fit with a sigmoidal function of the form in Equation 19. Fig-
four mixtures together. ures 5.6 through 5.9 show the individual and overall phase
The shift factors determined from the dynamic modulus angle master curves for individual mixtures. Figure 5.10 shows
master curves were then used to shift the individual phase angle the overall phase angle master curves for all four mixtures
curves at each temperature to obtain the phase angle master together.

60
Combined Mix

Individual Specimens
Phase Angle, Degree

40

20

0
1.E-03 1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
Reduced Frequency, Hz

Figure 5.6. Phase angle master curves for PG 67-22 optimum.

48

60
Master Curve at 20°C

Phase Angle, Degree

Individual Specimens
40

20

0
1.E-03 1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
Reduced Frequency, Hz

Figure 5.7. Phase angle master curves for PG 76-22 optimum.

60
Master Curve at 20°C
Phase Angle, Degree

Individual Specimens
40

20

0
1.E-03 1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
Reduced Frequency, Hz

Figure 5.8. Phase angle master curves for PG 67-22 optimum plus.

60
Individual Specimens
Phase Angle, Degree

Master Curve
40

20

0
1.E-03 1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
Reduced Frequency, Hz

Figure 5.9. Phase angle master curves for PG 76-22 optimum plus.
 49

60
67-22, Opt
76-22, Opt

Phase Angle, Degree

67-22, Opt+
40
76-22, Opt+

20

0
1.E-03 1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
Reduced Frequency, Hz

Figure 5.10. Phase angle master curves for all mixes tested.

Damage Characteristic Curve men data are fit using both a generalized power model and an
exponential model (60, 61), given in Equations 21 and 22,
The characteristic curve for a mixture describes the rela- respectively.
tionship between C, which is normalized pseudo stiffness
C = 1 − C11 ( S )
C12
calculated through viscoelastic theory, and S, which is the dam- (21)
age parameter calculated from continuum damage mechanics
C = e − kS (22)
theory. Practically, C can be thought of as the material’s
integrity at any point in time and S as the level of damage over where,
time. As the amount of damage in the material increases, the C11, C12, and k are regression coefficients (hereafter referred
material integrity decreases. A mixture’s characteristic curve to as damage curve coefficients) obtained by ordinary
can be constructed from monotonic or cyclic tests. The steps least squares technique.
involved in the construction of the characteristic curves are
described in Appendix E. The generalized power law does a better job fitting the actual
data obtained from the tests. Figures 5.15(a) and (b) show the
characteristic curve fits for all mixtures obtained using the gen-
Monotonic Testing
eralized power model and exponential model, respectively. The
Figures 5.11 through 5.14 show the characteristic curves behavior of the PG 76-22 optimum plus mixture appears to
obtained from on-specimen LVDTs for each individual mix- be significantly different from the other mixtures; the curve for
ture tested under monotonic loading. The individual speci- this mixture only represents one specimen, so more testing

1
Individual Specimens

Generalized power model

0.75
Exponential model

0.5
C

0.25

0
0 0.5 1 1.5 2 2.5 3
S

Figure 5.11. Characteristic curves using on-specimen LVDTs for different

strain rates for PG 67-22 optimum.
50

1
Individual Specimens

Generalized power model

0.75
Exponential model

0.5
C

0.25

0
0 0.5 1 1.5 2 2.5 3
S

Figure 5.12. Characteristic curves using on-specimen LVDTs for PG 76-22 optimum.

1
Individual Specimens

Generalized power model

0.75
Exponential model

0.5
C

0.25

0
0 0.5 1 1.5 2 2.5 3
S

Figure 5.13. Characteristic curves using on-specimen LVDTs for PG 67-22

optimum plus.

1
Individual Specimens

Generalized power model

0.75
Exponential model

0.5
C

0.25

0
0 0.5 1 1.5 2 2.5 3
S

Figure 5.14. Characteristic curves using on-specimen LVDTs for PG 76-22

optimum plus.
 51

1
67-22, opt
76-22, opt
0.75 67-22, opt+
76-22, opt+

C1
0.5

0.25

0
0 0.5 1 1.5 2 2.5 3
S1
(a) Generalized power model

1
67-22, opt
76-22, opt
0.75 67-22, opt+
76-22, opt+

0.5
C

0.25

0
0 0.5 1 1.5 2 2.5 3
S
(b) Exponential model

Figure 5.15. Overall characteristic curves using on-specimen LVDTs for all
mixtures tested under monotonic loading.

must be done to determine if this is representative behavior or versus S curves will be different for the various strain rates.
a problem with the one test specimen. Figures 5.16(a) and (b) show the C versus S curves at differ-
Initially, specimens failed near the end plates during mono- ent strain rates measured using the on-specimen and plate-
tonic tests, which is likely due to the presence of high air void to-plate LVDTs, respectively. It can be seen from these figures
content at the ends of the specimens. The specimens were that the curves for the various strain rates fall on top of each
shortened by cutting more material from the top and bottom. other except for the plate-to-plate curve for Specimen 10.
This was successful in producing failure in the middle of the There are some issues related to the strain data obtained from
specimens. In addition to making shorter specimens, LVDTs plate-to-plate LVDTs for this specimen. Thus, the data for the
were located plate to plate along with the on-specimen LVDTs remaining specimens indicates that the effect of viscoplastic-
during the monotonic tests (two on-specimen and two plate- ity in the monotonic test results is not significant.
to-plate LVDTs). The advantage of using the plate-to-plate
LVDTs is that the material response can be captured even if Fatigue Testing
the specimens fail outside the on-specimen LVDT gage length.
Monotonic tests performed at various strain rates can be A total of five specimens of PG 67-22 optimum were tested
used to investigate whether there is any significant effect of under cyclic fatigue loading using four on-specimen LVDTs.
plasticity in the mixture behavior. If there is a significant The testing was performed at 20°C by controlling the cross-
portion of viscoplastic strain in the material response, the C head displacements. Haversine waveforms were applied to the
52

0.75

0.5
C

0.25

0
0 0.5 1 1.5 2 2.5
S
rate=0.135/min (Spec #5) rate=0.135/min (Spec #6) 0.135/min (Spec #7)
rate=0.02/min (Spec #9) rate=0.1/min (Spec #8) rate=0.0675/min (Spec #10)
(a) On-specimen LVDTs

0.75

0.5
C

0.25

0
0 0.5 1 1.5 2 2.5 3
S
rate=0.135/min (Spec #6) rate=0.135/min (Spec #7) rate=0.02/min (Spec #9)
rate=0.1/min (Spec #8) rate=0.0675/min (Spec #10)
(b) Plate to Plate LVDTs

Figure 5.16. C versus S curves at different strain rates for PG 67-22

optimum specimens.

specimen until failure occurred. Table 5.2 summarizes the fa- The fatigue tests are neither controlled strain nor con-
tigue testing. Specimen 17 was the first specimen tested. The trolled stress tests, rather a mixed mode of loading occurs be-
research team anticipated that the specimen would fail within cause the crosshead rate is controlled, but the on-specimen
10,000 cycles, so the test was halted after about 26,000 cycles strains are used for analysis. Figures 5.17(a) and (b) show typ-
to determine what was happening. Due to machine compli- ical stress and strain history, respectively, recorded during a
ance, the actual strains measured by the LVDTs were roughly constant amplitude fatigue test. The load continuously de-
one-quarter of the applied crosshead strain, so the strain am- creases due to the development of damage, and decreases dra-
plitude was increased for subsequent tests. matically at failure. The mean strain increases continuously

Table 5.2. Summary of fatigue tests performed on PG 67-22

optimum specimens.
Specimen ID Average On- Crosshead Nf Comments
Specimen LVDT Strain, ms
Strain, ms
Specimen 14 242 1050 17,480 Failed in the middle
Specimen 15 254 1017 10,728 Failed in the middle
Specimen 16 n/a n/a n/a Failed at first cycle
Specimen 17 134 1300 >26,687 Did not fail
Specimen 18 223 1017 1731 Failed near top end plate
 53

(a) Stress history (b) Strain history

Figure 5.17. Typical stress/strain history for constant amplitude uniaxial fatigue test.

2
Specimen #15

Specimen #14
1.5
Monotonic

1
C

0.5

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
S

Figure 5.18. Comparison monotonic and cyclic curves for PG 67-22

optimum.

during the tests until failure, at which point the mean strain fSp
ε 02α = (23)
p ( 0.125 IC11C12 ) N ( E ⴱ )
drops. α 2α

Figure 5.18 shows the characteristic curves constructed

from both the fatigue and monotonic tests for the PG 67-22
2α f [ e − αkS − 1]
optimum mixture. Previous research has shown better agree- ε 02α = (24)
αI α ( −k ) N ( Eⴱ )
1+α 2α
ment between samples and between monotonic and cyclic
tests (32).
where,
ε0 is the strain level required to sustain N number of load
Evaluation of Endurance Limit repetitions,
Prediction from Characteristic S is the damage parameter value at failure (measured from
Damage Curves the damage characteristic curve for the mixture at the
point where C = 0.3, identified in previous research [32]),
Once dynamic modulus, initial stiffness, testing frequency, I is the initial pseudo stiffness,
and damage curve coefficients are known, the strain level |E∗| is the dynamic modulus at testing frequency (f ), and
required to sustain any number of design load repetitions can α is the material constant, p = 1 + (1 − C12)α.
be predicted. Equations 23 and 24 can be used to find required
strain level with generalized power law and exponential mod- Plots of strain level (ε0) versus design load repetitions (N)
els, respectively (60, 61). obtained for individual asphalt mixtures are presented in
54

1.0E-02
Exponential model

Generalized power model

1.0E-03

Strain
1.0E-04

1.0E-05
1.0E+00 1.0E+02 1.0E+04 1.0E+06 1.0E+08
Number of repetitions

Figure 5.19. Plot of strain level versus load repetitions for PG 67-22 optimum.

Figures 5.19 through 5.22. A comparison of the relations sented graphically in Figure 5.24. The values obtained using
of all mixtures together is presented Figures 5.23(a) and the exponential model are much lower than those obtained
(b) for the generalized power law and exponential models, from the generalized power law model. There is more confi-
respectively. dence in the values from the generalized power law because
The strain levels required to sustain 50 million cycles of this function fits the C-versus-S data for these mixtures bet-
repetitions for all mixtures are shown in Table 5.3 and pre- ter than the exponential function.

1.0E-02
Exponential model

Generalized power model

1.0E-03
Strain

1.0E-04

1.0E-05
1.0E+00 1.0E+02 1.0E+04 1.0E+06 1.0E+08
Number of repetitions

Figure 5.20. Plot of strain level versus load repetitions for PG 76-22 optimum.

1.0E-02
Exponential model

Generalized power model

1.0E-03
Strain

1.0E-04

1.0E-05
1.0E+00 1.0E+02 1.0E+04 1.0E+06 1.0E+08
Number of repetitions

Figure 5.21. Plot of strain level versus load repetitions for PG 67-22 at
optimum plus.
 55

1.0E-02
Exponential model

Generalized power model

1.0E-03
Strain

1.0E-04

1.0E-05
1.0E+00 1.0E+02 1.0E+04 1.0E+06 1.0E+08
Number of repetitions

Figure 5.22. Plot of strain level versus load repetitions for PG 76-22 at
optimum plus.

1.0E-03
Strain

1.0E-04

1.0E-05
1.0E+00 1.0E+02 1.0E+04 1.0E+06 1.0E+08
Number of repetitions
67-22 @ Opt. 76-22 @ Opt. 67-22 @ Opt. + 76-22 @ Opt. +
(a) Generalized power models

1.0E-02
Strain

1.0E-03

1.0E-04
1.0E+00 1.0E+02 1.0E+04 1.0E+06 1.0E+08
Number of repetitions
67-22 @ Opt. 76-22 @ Opt 67-22 @ Opt. + 76-22 @ Opt+
(b) Negative exponential models

Figure 5.23. Plot of strain level versus load repetitions LVDTs for all mixtures.
56

Table 5.3. Computed critical micro-strain to sustain of the material. The hysteresis loop collapses when the pseudo
50 million cycles. strain is plotted versus the stress, as seen in Figure 5.27. If dam-
age occurs in the material, a hysteresis loop will appear in the
Damage Asphalt Mixture Grade
Characteristic 67-22 76-22 67-22 76-22 stress-pseudo strain plot.
Curve Form Optimum Optimum Optimum+ Optimum+ The analysis of the increasing strain amplitude fatigue tests
Exponential
model
96 70 64 47 involves calculating the pseudo strain and then plotting
Generalized pseudo strain versus stress at each strain amplitude to deter-
261 197 194 164
power model mine the strain level at which loops begin to appear. The pres-
ence of a loop in the stress-pseudo strain plot indicates that
damage is occurring in the specimen. Figures 5.28 and 5.29
Increasing Strain Amplitude Test show the stress versus pseudo strain plots for PG 67-22 op-
This test consists of applying blocks of haversine loading to timum plus and PG 76-22 optimum plus specimens, respec-
a uniaxial test specimen. Initially, a relatively low strain am- tively. For the PG 67-22 optimum plus specimen, it is clear
plitude that is thought to be below the fatigue endurance limit that there is no loop at the lowest strain amplitude, a loop
is applied. Typically, this is close to the same amplitude at appears to just be forming at the middle strain amplitude and
which dynamic modulus tests are performed. Approximately a loop is definitely apparent at the highest strain amplitude.
10,000 cycles are applied at this amplitude to allow the speci- These two figures indicate that the fatigue endurance limit
men to reach steady-state response. The applied strain ampli- (strain level below which no damage is occurring) for this
tude is then increased and 10,000 more cycles are applied. This specimen is around 150 ms. For the PG 76-22 optimum plus
procedure is continued until the specimen fails. Figures 5.25(a) specimen shown in Figure 5.29, the loop appears in the sec-
and (b) show a typical load and strain history, respectively, ond load level, which is 245 ms. Hence, the fatigue endurance
recorded during an increasing amplitude fatigue test. It should limit is somewhere between 93 and 245 ms. Table 5.4 sum-
be noted that, for this project, the crosshead displacement marizes the results of the increasing amplitude fatigue tests
was controlled while the on-specimen LVDT measurements for all specimens tested. It should be noted that only three
were used for strain analysis. Machine compliance made it mixtures were subject to this test method.
difficult to precisely control the strain amplitude in the spec- The third column in Table 5.4 shows the bounds of the
imen at the different levels. strain level at which loop formation was observed. For sev-
eral specimens, the first level tested resulted in loop forma-
tion, so only an upper bound is reported. The specimen
Concept of Pseudo Strain
stiffness at 50 cycles and air void content are also shown in
The use of pseudo strain instead of engineering strain in Table 5.4. From this information, it is evident that there are
constitutive analysis removes the hysteretic effect of vis- two groups of specimens for the 67-22 optimum plus and
coelasticity. For example, a plot of stress versus strain data 76-22 optimum plus mixtures. The 67-22 optimum plus Spec-
obtained from a dynamic modulus test produces a hysteresis imens 1 and 2, and the 76-22 optimum plus Specimens 1
loop, as shown in Figure 5.26. The load levels applied during a and 3, have lower air void contents and correspondingly higher
dynamic modulus test are low enough not to induce damage, stiffnesses and strain range at which loop formation occurs
so the hysteresis loop is purely due to the viscoelastic response than the other specimens. Using engineering judgment, the

0.0003
Exponential Model
Generalized Power Model
0.0002
Strain

0.0001

0.0000
67-22 @ 76-22 @ 67-22 @ 76-22 Optimum+
Optimum Optimum Optimum+
Asphalt concrete mix

Figure 5.24. Comparison of critical strain to sustain 50 million

cycles for all mixtures.
 57

Stress, MPa
Load

Pseudo Strain, m/m

Time
(a) Load History Figure 5.27. Stress versus pseudo strain plot for
several cycles of loading.
LVDT reading 4

Stress, MPa

Time
(b) Strain History Pseudo Strain

Figure 5.25. Typical stress/strain history for increasing Figure 5.28. Stress versus pseudo strain plots for
amplitude uniaxial fatigue test. PG 67-22 optimum plus.
Stress, MPa

Stress (MPa)

Strain, m/m

Figure 5.26. Stress versus strain plot for

Pseudo strain
several cycles of loading.
Figure 5.29. Stress versus pseudo strain plots for
PG 76-22 optimum plus.
58

Table 5.4. Summary of increasing amplitude fatigue test data.

Loop Estimated
Specimen Stiffness @ 50th Air Voids,
Mixture Formation Endurance
Number Cycle (MPa) %
Strain Range Limit
15 <1811 8955 6.9
67-22 Opt 17 107<LF<121 11594 6.7 ~120
18 <2291 7927 6.6
1 150<LF<262 14756 2.2
~250
2 247<LF<345 17698 1.5
67-22 Opt+ 6 67<LF<82 7747 3.6
8 94<LF<147 9752 2.7 ~150
10 73<LF<155 10653 3.0
1 <2361 8997 1.5
~230
3 <2371 8896 1.9
6 59<LF<125 11250 3.1
76-22 Opt+ 7 102<LF<225 7905 3.2
8 110<LF<123 9859 2.7 ~115
9 96<LF<240 6636 3.7
10 <1311 8868 2.8

Note:
1
Indicates loop formed at lowest strain level tested.

Table 5.5. Overall fatigue endurance limit summary.

Estimated Fatigue Endurance Limit (ms)
Test Method PG 67-22 PG 67-22 PG 76-22 PG 76-22
Optimum Optimum+ Optimum Optimum+
Generalized
C versus S 261 194 197 164
power law
Prediction
Exponential 96 64 70 47
Increasing Low air void 250 230
120
Amplitude Not tested
High air void 150 115
Fatigue Test

estimated endurance limit values for the different mixtures The increasing amplitude fatigue test is promising and
are shown in the last column. requires some continued research to refine the method. One
of the main challenges in this test is controlling the strain
amplitude measured on the specimen. It is difficult to do
Summary of Endurance Limit Values
this by controlling the crosshead during the test because of
A summary of the fatigue endurance limit values deter- machine compliance. The amount of compliance changes
mined from the different test methods is shown in Table with different loading levels and is difficult to predict. There
5.5. The estimated endurance limits for the different groups are several ways to mitigate this problem. The ideal solution
of specimens based on air void content are shown in the is to control the test using the on-specimen LVDTs, however,
table. The C versus S prediction method does not follow ex- great care must be taken in running tests on closed-loop sys-
pected trends with respect to increasing asphalt content. tems in this way. Another alternative is to run load-controlled
The increasing amplitude fatigue test shows an increase in tests and determine the appropriate load levels by using the
endurance limit with an increase in asphalt content for the measured dynamic modulus and target strain amplitude.
PG 67-22 mixtures. However, there is not much difference A draft Proposed Standard Practice for Predicting the
between the estimated endurance limits for the different as- Endurance Limit of Hot Mix Asphalt (HMA) by Pseudo Strain
phalt grades. Approach is presented in Appendix B.
 59

CHAPTER 6

Examination of LTPP Database for Indications

of an Endurance Limit

Introduction 2. Design procedures that use the equivalent temperature

concept but the axle load distribution for each axle type—
Flexible pavements have traditionally been designed to These procedures also use the cumulative damage con-
limit load-related cracking. The more traffic, the thicker the cept to determine the amount of fracture damage for each
HMA layer to limit the load-related cracks to some design structure. The PerRoad Program would fall within this
limit. As noted, however, industry has been proposing the category (63).
use of an endurance limit as a mixture property for HMA 3. Design procedures that calculate and use pavement tem-
layers. The endurance limit is defined as the tensile strain peratures at specific depths over some time interval, gener-
below which no fracture or fatigue damage occurs and applies ally less than a month—These procedures typically use the
to cracks that initiate at the bottom of the HMA layer. Almost incremental damage concept to determine the amount of
all design and analysis procedures that use the endurance fracture damage within specific time intervals and at spe-
limit concept assume that one value applies to all HMA mix- cific depths within the pavement structure. The MEPDG
tures and temperatures. Values that have been used vary from would fall within this category (64).
65 to 120 ms.
This section of the report has three objectives: discuss the The equivalent temperature concept simply defines one
incorporation of the endurance limit design premise into temperature for which the annual or seasonal damage equals
mechanistic-empirical based pavement design procedures, the cumulative damage determined at monthly or more fre-
confirm the reality and values suggested for the endurance quent intervals. The equivalent temperature is used to estimate
limit, and recommend field studies to support use of this con- the dynamic modulus for calculating the tensile strain at the
cept in the MEPDG software. bottom of the HMA layer on an annual or seasonal basis.
All M-E based design procedures, regardless of the group,
use Miner’s hypothesis to calculate fracture damage, and
Including the Endurance
assume that wheel-load-related alligator cracks initiate at the
Limit Design Premise into
bottom of the HMA layer and propagate to the surface with
Mechanistic-Empirical-Based
continued truck loadings, with the exception of the MEPDG.
Pavement Design Procedures
In addition, all M-E based design procedures use the max-
All mechanistic-empirical pavement design procedures can imum tensile strain at the bottom of the HMA layer as the
be grouped into three types relative to wheel-load-induced pavement response parameter for calculating fracture damage
cracking. These are as follows: and predicting the amount of alligator cracks. Those design
procedures apply the endurance limit design premise in one
1. Design procedures that use the equivalent axle load and of three methods, which are summarized as follows:
equivalent temperature concepts—The equivalent tem-
perature is determined based on an annual or monthly 1. The introduction of the endurance limit design premise
basis. These procedures typically use the cumulative dam- into those design procedures that use the equivalent tem-
age concept to determine the amount of fracture damage perature and equivalent axle load concepts is straight for-
over the design period for each structure. The DAMA Pro- ward. Stated simply, the maximum tensile strain is cal-
gram would fall within this category (62). culated at the equivalent temperature and axle load and
60

Standard Mix; Conv. Low Modulus, Conv. High Modulus, Conv.

Standard Mix, Full-Depth Endurance Limit

180

Tensile Strain, Bottom of HMA,

160
140
120

micro-strains
100
80
60
40
20
0
7 8 9 10 11 12 13 14 15 16 17 18 19
HMA Thickness, inches

Figure 6.1. Tensile strains calculated for an 18-kip single-axle load

for the equivalent annual temperature for different HMA mixtures.

compared to the endurance limit. The HMA layer thickness with this method is that the higher loads result in signifi-
is simply determined for which the maximum tensile strain cantly higher damage indices; an increase in axle load will
equals or is less than the endurance limit. Figure 6.1 illus- result in an increase in damage to a power of about four.
trates the use of the endurance limit within this method. Thus, the probability of cracking is much higher than the
2. The introduction of the endurance limit into those design probability of a specific tensile strain being exceeded.
procedures that use the equivalent temperature concept 3. Those design procedures that use the incremental dam-
but use the actual axle load distribution is also fairly age concept establish a threshold value for the tensile
straight forward. The maximum tensile strain is calculated strain, below which the fracture damage is assumed to
at the equivalent temperature for each axle load within the be zero. In other words, the procedure simply ignores
axle load distribution. The axle load distribution for each calculated tensile strains that are equal to or less than the
axle type is used to determine the probability of the tensile value set as the endurance limit for determining the incre-
strain exceeding the endurance limit. The designer then mental damage within a specific time period and depth.
considers that probability of exceeding that critical value Successive runs have been made with the MEPDG to deter-
in designing an HMA layer for which no fatigue damage mine the difference in calculated fracture damage with and
would accumulate over time. Figure 6.2 illustrates the use without using the endurance limit as an HMA mixture
of the endurance limit within this method. One concern property. Figures 6.3 and 6.4 illustrate the increasing

Standard Mix, Conv. Low Modulus, Conv.

High Modulus, Conv. Standard Mix, Full-Depth
105
Endurance Limit (65 micro-
Probability of Exceeding

100
95
strains), %

90
85
80
75
70
7 8 9 10 11 12 13 14 15 16 17 18 19
HMA Thickness, inches

Figure 6.2. Probability of exceeding the endurance limit for

different HMA mixtures using typical axle load distributions and
seasonal temperatures.
 61

Summer, 250 ksi E=450 ksi E=650 ksi Winter, 900 ksi Endurance Limit

160

Tensile Strain, Bottom of HMA

140

Layer, micro-strains
120
100
80
60
40
20
0
6 16 26 36
Single Axle Load, kips

Figure 6.3. Increasing tensile strains for varying single-axle loads

for different seasons or dynamic modulus within those seasons
(HMA thickness equals 15 in.).

maximum tensile strains for varying single-axle loads for ing in the wheel paths is assumed to initiate at the surface and
different dynamic modulus values and HMA thicknesses, propagate downward. The MEPDG assumes that both types
respectively. of cracking are caused by load-induced tensile strains. That
hypothesis, however, has yet to be confirmed.
Version 0.9 of the MEPDG did not include the endurance As noted above, the new MEPDG uses an incremental dam-
limit design premise in the recalibration process of the design age index. Fracture damage is computed on a grid basis with
methodology or software. In other words, Version 0.9 assumes depth for each month within the analysis or design period.
that any tensile strain in the HMA layer induces some fac- Temperatures are computed with the Integrated Climatic
ture damage. Two types of load-related cracking are pre- Model at specific depth intervals for each hour of the day.
dicted for designing flexible pavements in accordance with These temperatures are then grouped into five averages at
the MEPDG—alligator cracking and longitudinal cracking in each depth interval for each month. The fatigue cracking
the wheel path. Alligator cracking, the more common crack- (alligator cracking) equation is used to calculate the amount
ing distress used in design, is assumed to initiate at the bottom of fracture damage for each depth interval and month. The
of the HMA layer. These cracks propagate to the surface with monthly damage indices are then summed over time to predict
continued fracture damage accumulation. Longitudinal crack- the area of fatigue cracking at each depth interval.

8-kip Load 14-kip Load 18-kip Load

22-kip Load 28-kip Load 34-kip Load
Endurance Limit
300
Maximum Tensile Strain,

250
micro-strain

200

150

100

50

0
6 8 10 12 14 16 18 20
HMA Thickness, inches

Figure 6.4. Increasing maximum tensile strains for varying single-

axle loads for different HMA thicknesses (HMA dynamic modulus
equals 450 ksi; equivalent annual modulus).
62

All design methods that accept the endurance limit design Survival curves are typically based on age but can also be
premise assume that the endurance limit is independent of based on traffic loadings or the probability of exceeding a spe-
the mixture and temperature. That endurance limit is the cific level of distress. The age or condition at failure must be
tensile strain below which no fracture damage occurs. If an based on a clearly defined condition. Mathematical models
endurance limit value was an input into the MEPDG or are best fitted to the points in the survival curves to predict the
used within other M-E based design methods, the question probability of survival or failure as a function of age, thick-
becomes, what value should be used as the endurance limit? ness, cumulative traffic, or some other pavement feature. The
The purpose of this section is to use the LTPP database to try general form of these models for use in life cycle cost analysis
and answer three questions related to the endurance limit is as follows (67, 68):
design premise, as follows:
a
Probability of Failure = bⴱ( Age −c )
+d (25)
1. Do field observations of alligator cracking support the exis- 1+ e
tence of an endurance limit as an HMA mixture property?
2. If the field observations support the endurance limit the- a
Probability of Failure = bⴱ( ESAL −c )
+d (26)
ory or hypothesis, what is the tensile strain below which no 1+ e
more alligator cracking has been exhibited?
3. Is the endurance limit independent of mixture type and where,
dynamic modulus? Failure = Existing pavement is overlaid or reconstructed,
or a specified level of distress has been exceeded;
Age = Number of years since construction (new pave-
Defining the Endurance Limit— ment or overlay);
A Survivability Analysis ESAL = Cumulative equivalent single 18-kip axle loads
since construction (new pavement or overlay),
A survivability analysis was used to try to answer the above
millions; and
questions using the LTPP database. The survivability analysis
a, b, c, d = Regression coefficients determined from the
completed within this project is an expansion of work com-
analysis.
pleted using the LTPP database in the mid-1990s. This section
of the report describes the use of survival curves in determin- The probability of survival is 1 minus the probability of fail-
ing the thickness or level of tensile strain at which only lim- ure. Optimization is typically used to determine the regression
ited cracking has occurred over long periods of time. coefficients that best fit the survival points. A survival analysis
also can be completed using a specific level of distress and
Development and Application pavement response value. In other words, the survival curves
of Survival Curves can be used to define the probability that a specific area of
alligator cracking will be less than some specified amount for
Survival or probability of failure analyses have been used for different HMA thicknesses or tensile strains at the bottom of
decades in actuarial sciences. They have also been used in the the HMA layer.
pavement industry to determine the expected service life of It is important to note that survival curves for pavements
pavement structures for use in life cycle cost analysis, and to are necessarily based on previously built designs, materials,
compare the mean and standard deviation of the expected construction, and maintenance. The data used to develop the
service life for different design features and site factors in eval- survival rates or probability of failure represent typical con-
uating the adequacy of the design procedure (65, 66). Survival struction, materials, mixture designs, and thicknesses that
curves are uniquely useful because every point on the curve have been built by agencies within the past time period rep-
represents the probability that a given pavement section will resented by the data. These can be defined as “benchmark”
be rehabilitated or exceed a specific level of distress. survival curves.
Survival analysis is a statistical method for determining the The reliability of a pavement depends on the length of time
distribution of lives or “Life Expectancy,” as well as the occur- it has been in service and design features and site factors that
rence of a specific distress for a subset of pavements. Since not are not properly accounted for in a thickness design proce-
all of the pavements included in the analysis have reached the dure. Thus, the distribution of the time to failure of a pave-
end of their service life or a specific level of distress, mean ment type or thickness level is of fundamental importance in
values can not be used. The age or amount of alligator crack- reliability studies. A method used to characterize this distribu-
ing and probability of occurrence are computed considering tion is the failure rate, or rate of occurrence, for a specific level
all sections in the subset using statistical techniques. of distress. The failure rate can be defined as follows.
 63

If f(t) is the probability density of the time to failure of a 0.45

given pavement type and thickness, that is, the probability 0.4 Premature
Failure Part
that the pavement will fail between times t and t + Δt is given 0.35

Failure Rate
by f(t)* Δt, then the probability that the pavement will fail on 0.3
0.25
the interval from 0 to t is given by
0.2
t 0.15
F ( t ) = ∫ f ( t ) dx
Wear Out
(27) 0.1 Failure Part
0
Chance Failure Part
0.05
0
The reliability function, expressing the probability that it 0 10 20 30
survives to time t, is given by Age of Pavement, years

Figure 6.5. Typical failure rate relationship for

R (t ) = 1 − F (t ) (28) pavement structures.

Thus, the probability that the pavement will fail in the inter-
val from t to t + Δt is F(t + Δt) − F(t), and the conditional first part, and when failure is a result of multi-distresses
probability of failure in this interval, given that the pavement as related to a combination of parameters over time (for
survived to time t, is expressed by example, exponential growth increases in traffic, past the
design period from which thickness was determined).
F ( t + Δt ) − F ( t )
(29)
R (t ) The failure rate can be determined by organizing the per-
formance data in terms of the distribution of pavement age
Dividing by Δt, one can obtain the average rate of failure in exceeding a critical level (failure) versus the distribution of
the interval from t to t + Δt, given that the pavement survived age for those pavements exhibiting a value lower than the
to time t by critical value. Figure 6.6 shows a typical probability of failure
relationship from actual data included in the LTPP database
F ( t + Δt ) − F ( t ) ⎡ 1 ⎤ for roughness measured on flexible pavements in the general
⎢ R (t ) ⎥ (30) pavement studies (GPS-1 and GPS-2) and special pavement
Δt ⎣ ⎦
studies (SPS-1) experiments. GPS-1 sections consist of HMA
For small Δt, one can get the failure rate, which is on granular base. GPS-2 sections consist of HMA on bitumi-
nous, hydraulic cement, lime, fly ash, or other pozzolan bound
f (t ) f (t ) or stabilized base. SPS-1 sections are part of a strategic study
Z (t ) = = (31) of structural factors for flexible pavements.
R (t ) 1 − F (t )
Given the above definition of each part of the probabil-
ity of failure relationship with time, the failure rate can be
The failure rate is expressed in terms of the distribution of defined as
failure times. A typical failure rate curve is composed of three
parts or can be grouped into three areas, as shown in Figure 6.5 ⎡ − t Z (t )dx ⎤
and defined as follows: ⎢ ∫ ⎥
f ( t ) = Z ( t ) ⎣e 0 ⎦ (32)
1. The first part is characterized by a decreasing failure rate
with time and is representative of the time period during Assuming that the failure rate is constant within the second
which early failure or premature failures occur. This area or part, and replacing Z(t) with α, the distribution of failure times
time typically represents pavements that were inadequately is an exponential distribution as shown below.
designed or built, using inferior materials.
2. The second part is characterized by a constant failure rate. f ( t ) = α [ e − αt ] (33)
A constant failure rate represents the time period when
chance failures occur, or the failure occurs at random with Many survival curves or, conversely, the probability of fail-
pavement age. In some survival methods, this area is referred ure, are based on the above relationships and assumptions.
to as the useful life of a pavement. Unfortunately, the failure rate within the second part is not
3. The third part is characterized by an increasing failure rate usually constant, and the failure rates for the first and third
with time. This area or time represents the reverse of the parts are not inversely proportional to one another. For these
64

120

100

Probability of Exceeding
Roughness Level, %
Low Roughness, >1.5 m/km
80

Moderate Roughness, >2.0 m/km

60

40 Excessive Roughness, >2.5 m/km

20

0
0 10 20 30 40
Age, years

Figure 6.6. Illustration of results from a survivability analysis or probability of

exceeding a critical roughness magnitude.

cases, which are typical for pavements, the failure rate can be result in rehabilitation of the roadway (69). The test sections
estimated by the following relationship: used in the survival analysis were from the GPS-1 and GPS-2
experiments. Figure 6.7 shows the distribution of pavement
Z ( t ) = αβ ( t )
β−1
(34) age for the GPS-1 and GPS-2 test sections (LTPP database
version 13.1/NT3.1 released in January 2002), and Figure 6.8
Thus, shows the number of test sections with different areas of alli-
gator (fatigue) cracking. As shown in Figure 6.8, many of the
f ( t ) = αβ ( t )
β−1
⎡⎣e − αt β ⎤⎦ (35) LTPP test sections have no alligator cracking.
Figures 6.9 and 6.10 show the survival curves from the
This density function is termed the Weibull distribution, LTPP data for different levels of alligator cracking that would
and is typically used in failure analyses. cause some type of rehabilitation activities. As shown, the
average life (50% probability) to crack initiation and a low
cracking amount (less than 10% of wheel-path area) is 19
LTPP Database to Establish the Initial
and 23 years, respectively.
Survival Curve
A similar survivability analysis was completed by Von
A survivability analysis was completed by Von Quintus Quintus in 1995 for a subset of the test sections included in
et al. for the Asphalt Pavement Alliance to determine the the GPS-1 and GPS-2 experiments. The test sections were
expected age for an amount of fatigue cracking that would randomly selected from the LTPP program for the thicker

50

40

30
No. of Test
Sections
20

10

0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33
= 0 9 27 38 17 29 35 46 39 23 25 27 16 16 17 7 1
Age, years

Figure 6.7. Frequency histogram of pavement age at the time

of the distress survey for the LTPP GPS test sections included in
the Asphalt Pavement Alliance study (69).
 65

180
160
140
120
No. of Test 100
Sections 80
60
40
20
0
50- 100- 150- 200- 300- 400-
0 0-10 13-50
100 150 200 300 400 500
= 177 65 45 27 16 15 14 9 1
Area of Fatigue Cracking, sq. m.

Figure 6.8. Histogram of the number of test sections with

different levels of alligator cracking (69).

120
Probability of Cracking Level, %

100
Crack Initiations
80
Low Cracking, >10%
60
Moderate Cracking, >20%

40
Excessive Cracking, >50%

20

0
0 10 20 30 40
Age, years

Figure 6.9. Graphical illustration of the probability of failure or

exceeding a specified area of alligator cracking (69).

60
Area of Fatigue Cracking, %

50
50% Probability of
40 Occurrence
25% Probability of
30
Occurrence
20 75% Probability of
Occurrence
10

0
0 10 20 30 40
Age, years

Figure 6.10. Probability of occurrence for alligator cracking (69).

66

HMA layers. This survivability analysis was completed to try ulus and tensile strain at failure, where both properties are
to estimate a value for the endurance limit based on alligator determined at 77°F (25°C). Points below the line in Figure 6.12
cracking observations within the LTPP program, rather than are assumed to have inferior fatigue properties, and those
just using values estimated from limited laboratory testing above the line exceed the fatigue strength/life of the standard
programs. This survival analysis was a desk-top study that has mixture. Laboratory tests and field observations of alliga-
yet to be formally documented. The LTPP data were used to tor cracking have been used to check the validity of this rela-
determine the probability of occurrence of different amounts tionship over time. More alligator cracking has been observed
of alligator cracking for different HMA thicknesses and ten- where the tensile strain at failure is less than that value from
sile strains. Figure 6.12 for a specific HMA modulus value based on the
The EVERSTRESS Program was used to calculate the maxi- equivalent temperature concept.
mum tensile strain at the bottom of the HMA layer for each test Von Quintus used this relationship to estimate or define
section using the equivalent annual temperature and equivalent the endurance limit for different HMA mixtures as 1% of the
(18-kip) single-axle load concepts. The HMA modulus value tensile strain at failure measured in accordance with the test
used in the calculation of tensile strain was determined using protocol from the AAMAS study (59). That definition has yet
the Witczak equation (70) based on volumetric data and to be confirmed and validated.
physical properties of the HMA for the equivalent annual
temperature. The modulus values for the other pavement
Updated Survivability Analysis
and soil layers were based on resilient modulus testing per-
Using LTPP Data
formed in the laboratory. Figure 6.11 shows the survival
curve from that limited study. A magnitude of 2% cracking The survivability analysis completed for this project included
was used in this initial survival analysis because of the mea- the same test sections from the 1995 study, plus additional
surement error in alligator cracking with time. A small mea- test sections within the GPS-1, GPS-2, and SPS-1 experiments.
surement error could result in significant changes to this A subset of the LTPP test sections was used, which was ran-
definition of the endurance limit. In summary, the endurance domly selected to cover all environmental regions, soil types,
limit was determined to be 65 ms at a 95% confidence level and HMA thicknesses. The additional test sections used in the
for an 18-kip single-axle load applied to the pavement at the updated survivability analysis were from the GPS-1, GPS-2,
equivalent annual temperature for each LTPP site included and SPS-1 experiments—LTPP database version VR 2004.06,
in the analysis. release 18.0 (2004). Figure 6.13 shows the distribution of
HMA thickness for all test sections included in the updated
survivability analysis.
Preliminary Definition of the Endurance
Figure 6.14 shows the distribution of pavement age for the
Limit as an HMA Mixture Property
test sections with more than 10 in. of HMA that were used to
The AAMAS project sponsored by NCHRP recommended update the survival curve (Figure 6.11). As shown, the age of
use of the indirect tensile strength and modulus tests to esti- 45% of the thicker test sections included in the updated study
mate the fatigue strength/life of specific HMA mixtures (59). is greater than 15 years. Many of these additional test sec-
Figure 6.12 illustrates that relationship between HMA mod- tions were from the SPS-1 experiment that was excluded from

Fatigue Cracking <2%

100.00
95.00
90.00
85.00
Survival, %

80.00
75.00
70.00
65.00
60.00
55.00
50.00
40 60 80 100 120 140 160 180 200 220 240 260 280 300
Tensile Strain, micro-strains

Figure 6.11. Survival curve for flexible pavements developed from data
included in the LTPP GPS-1 and GPS-2 experiments (undocumented
study, 1995).
 67

100

Tensile Strain at Failure, mils/in.

10

1
10 100 1000 10000
Total Resilient Modulus, ksi

Note: ms and mils/in. are the same unit.

Figure 6.12. Relationship between modulus and tensile strain at

failure to estimate the fatigue strength of HMA Mixtures at 77ⴗF
(25ⴗC) (59).

110
Number of sections in each thickness layer

100

90

80

70

60

50

40

30

20

10

0
<2 2.1-4 4.1-6 6.1-8 8.1-10 10.1-12 12.1-15 15.1-20 > 20
Number 94 65 88 101 104 92 71 88 25
Thickness Layer of HMA Layers

Figure 6.13. Distribution of HMA thickness for the test sections used in the
updated survivability analysis.

HMA Thickness > 10 inches

70
Number of Test Sections

60
50
40
30
20
10
0
<11 11 12 13 14 15 16 17 18 19 20+
Age, years

Figure 6.14. Distribution of age for those test sections with HMA layer
thicknesses in excess of 10 in.
68

the initial study to estimate the endurance limit. The reason there are an appreciable number of test sections with thicker
that the SPS-1 test sections were excluded from the study in HMA layers (15 in. or more) that have levels of fatigue crack-
1995 is that most of the projects within the SPS-1 experiment ing exceeding 5%.
were relatively new at that time. Figure 6.17 compares the maximum tensile strain calcu-
The other important parameter in the survivability analy- lated for each section and HMA thickness. The modulus of
sis is the truck traffic applied to each of these test sections. the HMA layer was determined using the equivalent temper-
Without significant truck traffic, defining the endurance limit ature concept for an 18-kip ESAL, as described previously for
from field observations has limited meaning. Figure 6.15 the original survivability analysis. As shown and expected, the
shows the distribution of the cumulative number of 18-kip tensile strains decrease with increasing HMA layer thickness.
ESALs for the test sections included in the updated surviv- Figure 6.18 compares the maximum tensile strain at the bottom
ability study that have HMA thickness in excess of 10 in. The of the HMA layer and the amount of fatigue cracking observed
cumulative truck traffic for these thicker test sections is con- on the LTPP test sections from the most recent distress survey
sidered moderate traffic with most test sections having less included in the LTPP database. As shown and expected, the
than 15 million cumulative 18-kip ESALs. test sections with the lower tensile strains have less fatigue or
In summary, the test sections with the thicker HMA lay- alligator cracking.
ers are not new pavements (Figure 6.14), but do have truck
traffic levels that are lower than what would be considered
Maximum Tensile-Strain-Based Definition
heavy truck traffic (Figure 6.15). This level of truck traffic is
a concern to the definition established for the endurance An updated survival curve to the one presented in Fig-
limit. Much higher levels of truck traffic are needed to vali- ure 6.11 was developed for the additional alligator cracking
date the endurance limit design premise with field observa- data and LTPP test sections. Figure 6.19 shows the results
tions and data. from the survival analysis for a range of fatigue cracking levels.
The results from the updated survival analysis are signifi-
cantly different from the 1995 desk-top study. In fact, the
HMA Thickness-Based Definition
updated survival curve for the 1% and 2% alligator crack-
The asphalt industry has proposed some maximum HMA ing levels would indicate that there is no endurance limit for
thicknesses that are believed to be resistant to alligator crack- these sections. The relationships shown in Figure 6.19 for
ing. The LTPP database was used to determine the level of the 1% and 2% cracking levels have a peak survival rate sig-
HMA thickness at which none or little alligator cracking has nificantly less than 100% and then begin to decrease with
been observed on HMA pavement surfaces. Figure 6.16 com- lower tensile strain values.
pares the amount of fatigue cracking (percent of wheel-path Some of the GPS test sections that were without alligator
area) from the most recent distress survey and HMA thick- cracks in 1995 now have some alligator cracks recorded in the
ness. As shown and expected, the test sections with thinner LTPP database for the test sections with the thickest HMA
HMA layers generally have more fatigue cracking. However, layers. Possible reasons for the significant difference in results

HMA Thickness > 10 inches

120
Number of Test Sections

100

80
60

40

20

0
<5 5-10.0 10-15.0 15-20 20-25 25-30 >30
Range of Cumulative Axles, millions of ESALs

Figure 6.15. Distribution of cumulative equivalent single-axle loads

for the test sections used in the survivability analysis with HMA layer
thicknesses in excess of 10 in.
 69

Series1 Log. (Series1)

100

Area Fatigue Cracking, %

90
80
70
60
50
40
30
20
10
0
0 5 10 15 20 25 30
HMA Thickness, inches

Figure 6.16. Comparison of area fatigue cracking (area alligator

cracking based on a percent of wheel-path area) and HMA layer
thickness.

Series1 Power (Series1)

1000
Tensile Strain, Micro-Strain

100

10
0 5 10 15 20 25 30
HMA Thickness, inches

Figure 6.17. Comparison of the maximum tensile strain at the

bottom of the HMA layer and HMA thickness.

Fatigue Cracking Log. (Fatigue Cracking)

100
90
Fatigue Cracking, %

80
70
60
50
40
30
20
10
0
10 100 1000 10000
Tensile Strain, micro-strains

Figure 6.18. Comparison of the area fatigue cracking and

maximum tensile strain computed at the bottom of the
HMA layer.
70

Fatigue <1% Fatigue <2% Fatigue <4% Fatigue <8%

100.00
95.00
90.00
85.00

Survival, %
80.00
75.00
70.00
65.00
60.00
55.00
50.00
0 50 100 150 200 250 300 350 400
Tensile Strain, micro-strains

Figure 6.19. Survival curves based on the maximum tensile strain

at the bottom of the HMA layers of flexible pavements included in
the LTPP program.

or the survival curve from the one developed in 1995 are sum- It is expected that this is not the reason for the difference
marized as follows: in survival curves.
• There was a change in the LTPP definition of longitudinal
• The SPS-1 projects were added to the updated analysis. It cracking in the wheel path. The change in definition defi-
is expected that including the SPS-1 projects did not cause nitely could have affected the updated survival curve. Many
this difference in findings, unless the fatigue cracking ini- of the previously measured longitudinal cracks that are as-
tiated from some other design-site feature that would have sumed to have initiated at the surface are now recorded as
a higher probable occurrence within the SPS-1 test sec- alligator cracking and are assumed to have initiated at the
tions, as compared to the GPS sections. In addition, the bottom of the HMA layer. The cracking maps and video dis-
study completed for the Asphalt Pavement Alliance con- tress data logs can be reviewed to segregate longitudinal
cluded that there was a possibility that the GPS test sections cracks with crack deterioration along the edges from tradi-
selected by the individual agencies for the LTPP program tional alligator cracks. This evaluation process is time con-
are biased towards the better performing pavements. The suming. Figure 6.20 graphically presents the change in per-
SPS-1 projects were built during the LTPP program and centage of survival sections as a function for varying alligator
would not be biased toward better performing pavements. cracking levels for different tensile strains at the bottom of

Tensile Strain=50 mils/in. Tensile Strain=75 mils/in.

Tensile Strain=100 mils/in. Tensile Strain=150 mils/in.

100
95
90
85
Survival, %

80
75
70
65
60
55
50
0 1 2 3 4 5 6 7 8 9
Alligator Cracking, % of Wheel Path

Figure 6.20. Percent survival of test sections for different levels of

alligator cracking and tensile strain at the bottom of the HMA layer.
 71

the HMA layer. As shown in Figure 6.20, the 150 ms (mils/ • An additional reason or explanation for the difference in
in.) curve deviates from the other relationship. Errors in results is that there is no endurance limit for HMA mixtures.
measuring small amounts of alligator cracking as well as a
change in the definition for alligator cracking could have In summary, it is still believed that the endurance limit is
caused this anomaly. This indicates that other types of crack- an HMA mixture property. Based on the results from the
ing may be included as fatigue cracks for pavements with updated survival analysis, however, forensic investigations
strain levels of 100 ms or less at the bottom of the asphalt of the test sections with the thicker HMA layers are needed
layer calculated using an equivalent annual temperature to confirm the location of crack initiation and other assump-
and 18-kip axle load. To determine the cause of the anom- tions used noted above in the survivability analysis.
aly requires that forensic investigations be completed on
these much thicker HMA sections to determine the cause of Affect of Polymer Modification
the recorded alligator cracking. on Field Performance
• The location where alligator cracks recorded in the LTPP
database initiated are assumed. As noted above, alligator Von Quintus et al. (71) conducted a study to quantify the
cracks are assumed to initiate at the bottom of the HMA effects of polymer modification on pavement performance.
layer and propagate to the surface. The validity of this Sites were selected from the LTPP database, NCAT Test Track,
assumption would have an effect on the survival curve. FHWA Accelerated Loading Facility (ALF), and a number
In addition, the maximum tensile strain at the bottom of of Canadian provinces and U.S. states with good records of
the HMA layer was based on the assumption of full-bond performance and material properties. For each polymer mod-
between all HMA lifts. If partial bond exists between two ified section, a control mix or two to three unmodified sites
lifts near the surface, load-related cracks can initiate at that were selected for comparison. The unmodified sections were
location and propagate downward as well as upward. A termed companion sites.
full forensic investigation will be needed to determine the The performance of the polymer modified asphalt (PMA)
location of where these cracks, recorded in the LTPP data, test sections and their companion sections was compared
initiated and the mechanism (debonding between adjacent using a normalization technique that is based on computing
HMA lifts) that resulted in those cracks. damage indices for each test section. This normalization tech-
• The initial and updated survivability analysis was performed nique uses M-E models to reduce the effect from confounding
assuming that stripping or moisture damage is not present factors between projects. The M-E performance prediction
within the HMA layer. Stripping and moisture damage were models were calibrated to local conditions using performance
adequately identified during the initial sampling and coring data from companion sections (without any additive or mod-
program for the GPS test sections. For the SPS-1 projects, ifier in the HMA mixture). This local calibration procedure
stripping or moisture damage may have occurred on some was used to estimate the true effect of PMA because of the vari-
of the projects and resulted in premature alligator cracking ation and errors associated with the models selected for use.
for the thicker sections. This possible cause for the difference Figure 6.21 presents a comparison of the fatigue cracking
in findings can be resolved with forensic investigations. predicted using the locally calibrated fatigue cracking equation

Companion Sites PMA Sites Line of Equality

100
Measured Fatigue Cracking, %

90
80
70
60
50
40
30
20
10
0
0 20 40 60 80 100
Predicted Fatigue Cracking, %

Figure 6.21. Comparison between the predicted and measured

fatigue cracking for neat HMA and PMA mixtures (71).
72

Trend Line for Companion Sites Companion Sites PMA Sites

100

Measured Fatigue Cracking, %

90
80
70
60
50
40
30
20
10
0
0.01 0.1 1 10
Locally Calibrated Fracture Damage Index

Figure 6.22. Comparison between the fracture damage index

and measured fatigue cracking for neat HMA mixtures and
PMA mixtures (71).

and that actually measured for the sites. The data for the constructed with PMA can withstand a larger percentage
unmodified sections fall along the line of equality, whereas of their maximum load repetitions for a given level of
the data for the polymer modified section indicate that the cracking.
actual cracking is less than the predicted cracking. Fig- Both Figure 6.21 and 6.22 support the findings in Chapter 4
ure 6.22 shows the damage index (DI) or ratio applied loads that indicate the polymer modified PG 76-22 should have a
to allowable load before failure occurs versus measured higher endurance limit. This also indicates that the endurance
fatigue cracking. Figure 6.22 also indicates that pavements limit is mixture specific.
 73

CHAPTER 7

Sensitivity of Pavement Thickness

to the Endurance Limit

Summary of Predicted Estimate of Shift Factors

Endurance Limits between Laboratory Tests
and Field Performance
Estimates of the endurance limit were obtained from beam
fatigue tests, uniaxial tension tests, and analysis of field per- In Chapter 3, calculations were provided to show that the
formance. Table 7.1 summarizes the endurance limit predic- maximum expected load repetitions in a 40-year period is
tions from the beam fatigue and uniaxial tension tests. An approximately 500 million. Then, using a shift factor of 10
analysis of LTPP data, presented in Chapter 6, indicated an between laboratory and field performance, a limit of 50 mil-
endurance limit of 65 ms. There appears to be good corre- lion cycles was determined for the laboratory beam fatigue
lation between the beam fatigue estimates of the endurance testing. The selection of a shift factor of 10 was based on rec-
limit determined in Phase I and those determined during ommendations from SHRP (46). The structural sections
the mini round-robin analysis. Based on the predicted val- from the 2003 NCAT Test Track provide an opportunity to
ues determined from testing at NCAT and the Asphalt Insti- verify the shift factors. Three methodologies were used to as-
tute, stiffer high-temperature binder grades and optimum sess the shift factors for the 2003 NCAT Test Track structural
plus asphalt contents produce higher endurance limit values. sections: measured strains from in-place instrumentation,
The 95% lower prediction limit samples follow the same gen- PerRoad, and the MEPDG.
eral trend. The prediction of the endurance limit from the The NCAT Test Track was initially constructed in 2000. All
round-robin data shows small differences based on binder of the sections of the 2000 track consisted of 19 in. of HMA
grade and asphalt content. (15 in. of which were the same dense-graded mixes for all of
The rankings from the uniaxial tension tests are the reverse the sections and the remaining 4 in. that were the experi-
of those determined from the beam fatigue tests. For the mental mixes that varied between sections), 5 in. of perme-
uniaxial tension tests, the PG 76-22 mixes generally result in able asphalt-treated drainage layer, 6 in. of crushed aggregate
lower estimates of the endurance limit. This appears to be the base, and 12 in. of improved (AASHTO A-2) subgrade.
opposite of field experience with polymer modified binders, The 2003 NCAT Test Track cycle included eight struc-
illustrated in Figures 6.21 and 6.22. There are a number of tural sections. Three pavement sections were designed using
potential reasons for the differences observed between beam the 1993 AASHTO Pavement Design Guide to carry approx-
fatigue and uniaxial tension testing. There were very few uni- imately 1⁄3, 2⁄3, and the full 10 million ESAL loading (for that
axial replicates tested, especially compared with the number cycle). The design reliabilities are summarized in Table 7.2.
of beams tested. In some cases, the trends are based on results The input parameters are summarized in Von Quintus (72).
from only one or two samples. Early in the uniaxial testing, The eight structural sections were selected to evaluate pave-
there were a number of problems with end failures. Speci- ment sections designed for varying levels of traffic, polymer
men fabrication was altered to reduce this problem. Finally, modified and unmodified or neat asphalt binders, stone
the modes of loading uniaxial and beam fatigue samples are matrix asphalt (SMA), and rich bottom layer. All eight sec-
different. To date, the basic research has not been completed tions were placed on 6 in. of granular base. The section lay-
to understand the differences between the modes of loading out is shown in Figure 7.1. The thin, medium, and thick pave-
or how the stress state changes with damage. The state of stress ment sections were constructed using both a neat PG 67-22
in pavements is not completely understood either. More effort binder and a styrene-butadiene-styrene (SBS) polymer mod-
is needed prior to considering this approach. ified PG 76-22. The 19.0-mm NMAS base course was used for
74

Table 7.1. Summary of estimates of the endurance limit (ms).

Mix Beam Fatigue Beam Fatigue Round-Robin Uniaxial C versus S Uniaxial Increasing
Amplitude
Predicted 95% Predicted1 95% Average Power Exponential Average Average
Lower Lower 95% Model Model for Loop Lowest for
Confidence Confidence Lower Formation3 Loop
2
Limit Limit Confidence Formation3
Limit2
PG 58-22 107 82 NA NA NA NA NA NA NA
PG 64-22 89 75 NA NA NA NA NA NA NA
PG 67-22 172 151 182 130 103 261 96 195 133
PG 67-22 Opt. + 184 158 176 141 121 194 64 223 150
PG 76-22 220 146 195 148 126 197 70 NA NA
PG 76-22 Opt. + 303 200 NA NA NA 164 47 189 124
Notes:
1
Calculated using the pooled data from the round-robin.
2
Average of the 95% lower confidence limit calculated by each individual lab.
3
Averages calculated based on individual specimen data not presented in Chapter 5.

Table 7.2. 2003 NCAT Test Track structural section design

reliabilities (72).

Traffic HMA, In. Granular Fill1, In. Structural Reliability

Base, In. Number at 10
Million
ESALs
Full 9 6 15 6.2 92%
2/3 7 6 17 5.4 68%
1/3 5 6 19 4.6 30%
Note:
1
Fill was placed from the top of the original improved subgrade to the bottom of the granular base
to maintain the surface elevation of the structural sections with respect to the 2000 track.

the fatigue testing in this study with the same binders. For the 9 vehicle). A typical triple trailer is shown in Figure 7.2. For
seventh section, the 1-in. thick wearing course of the medium the triple trailer, the average steer axle weight was 10,680
thickness design was replaced with SMA. SMA was also used lbs, the average weight of the tandem drive axles was 40,610
as the wearing surface of the eighth section; in addition, the lbs, and the average weight of the single-axle trailer ranged
bottom 2 in. of the 19.0-mm base were replaced with a rich from 20,550 to 21,010 lbs. For the five-axle single trailer,
bottom layer containing an additional 0.5% asphalt (medium the steer axle weighed 11,550 lbs, the tandem drive axles
total thickness). weighed 33,850 lbs, and the tandem rear axles weighed
For the 2003 track, loading was provided by five triple- 32,900 lbs (73). Hot tire inflation pressures were typically
trailer trucks and one five-axle single trailer (FHWA Class 105 psi.

N1 N2 N3 N4 N5 N6 N7 N8
0 Aggregate Base
1 1 1 1 1 1 1 1
2 PG 67-22, 19.0 mm Opt+
4 4 4
4 6 6 6 PG 67-22, 19.0 mm NMAS
Thickness, in.

8 8
6 2
PG 76-22 , 19.0 mm NMAS
8 6 6
PG 76-22 SMA
10 6 6 6 6
12 6 6 PG 67-22, 9.5 mm NMAS

14 PG 76-22, 9.5 mm NMAS

16

Figure 7.1. Layout of 2003 NCAT Test Track structural sections.

 75

tions and the two loading configurations (triple trailer and

five-axle single trailer). The regression model used a power
relationship (Equation 36).

ε t = β1T β2 (36)

where,
εt = horizontal tensile strain (ms),
Figure 7.2. Typical triple-trailer weight distribution T = mid-depth HMA temperature, °F, and
for 2003 NCAT Test Track (73). β1 and β2 = regression constants.

The regression constants are summarized in Tables 7.4

Six of the eight structural sections developed some degree and 7.5. For Section N8, the data exhibited three different
of fatigue cracking during the 2003 test track cycle (all but the performance periods. Prior to April 20, 2004, the strain re-
9-in. thick HMA sections N3 and N4). Three of the sections sponse appears to indicate that the layers were bonded. From
failed, with failure defined as fatigue cracking exceeding 20% April 20 to September 1, 2004, the strain response shifts, but
of the total lane area (73). The MEPDG assumes cracking of maintains a power model relationship with temperature.
50% of the total lane area when damage equals 100% (74). After September 1, 2004, increasing strains, even with low
The cracking data at failure and failure dates for Sections N1, mid-depth HMA temperatures, appear to indicate that cracks
N2, and N8 are shown in Table 7.3. The evolution, monitor- are propagating through the upper layers. Two separate mod-
ing, and calculation of the crack areas are documented by els are shown in Table 7.4 with coefficients for traffic before
Priest and Timm (73). The degree of cracking in Section N8, and after the point when the bond appears to have failed.
with the rich bottom layer, was unexpected, particularly when A generic relationship for both the triple trailer and five-
compared with the performance of Section N7. Willis and axle single trailer was developed for the sections as a function
Timm (75) conducted a forensic evaluation that indicated of thickness (73). The generic equation (37) was used to cal-
slippage or debonding between the rich bottom layer and the culate the strains resulting from the five-axle single trailer for
overlying base layer. Cracking apparently began in the over- Section N1. Insufficient data were collected before N1 cracked
lying layer. to develop a specific model for Section N1. The surveyed
HMA thickness of Section N1 was 4.8 in. A thickness of 5 in.
was used for Section N8 after September 1, 2004 to represent
Shift Factors Based on Measured Strain
debonding between the base and rich bottom layer.
The eight structural sections from the 2003 NCAT Test Track
were instrumented to measure in situ strain in the asphalt, com- ε t = 2.1228T 1.190 − 26.448t (37)
pressive stresses in the unbound layers, and moisture and tem-
perature as a function of depth in the pavement structure (72). where,
Temperature and moisture data were collected on an hourly εt = horizontal tensile strain (ms),
basis throughout the two-year loading cycle. High-speed data T = mid-depth HMA temperature, °F, and
(2,000 samples per second) from the asphalt strain gauges t = HMA thickness (in.).
and unbound layer pressure cells were collected at least once
a month for at least three truck passes for each section and The low-speed data acquisition system collected the average
weekly after cracking was initially observed. The layout of the mid-depth HMA temperature for each section on an hourly
instrumentation for each section is shown in Figure 7.3. basis. The numbers of laps by each triple trailer and five-axle
Based on the high-speed data acquisition, Priest and Timm single trailer truck were recorded for each hour. Both Priest
(73) developed strain prediction models for each of the sec- (76) and Willis (77) developed spreadsheets to calculate the

Table 7.3. 2003 NCAT Test Track structural section failure

data (73).
Section Failure Date Cracking, % of Cracking, % of
Total Lane Wheel Path
N1 6/14/2004 20.2 58.3
N2 7/19/2004 19.5 56.3
N8 8/15/2005 18.5 53.5
76

Figure 7.3. Typical gauge array for 2003 NCAT Test Track structural sections (72).

Table 7.4. Triple-trailer regression analyses for

strain-temperature relationship (73 [not N8]).

Section 1 2 R2
N11 4.0439 1.066 0.76
N2 0.0005 3.081 0.88
N3 0.0508 1.899 0.91
N4 0.0211 2.086 0.82
N5 0.0109 2.291 0.88
N6 0.0132 2.293 0.81
N7 0.0022 2.652 0.71
N8 (prior 4/20/2004) 0.1487 1.556 0.90
N8 (4/20/2004 and after) 0.0926 1.824 0.81
1
Limited data were available.

Table 7.5. Five-axle single-trailer regression analyses for

strain-temperature relationship (73 [not N8]).

Section 1 2 R2
N1 Insufficient data to perform regression
N2 3.922E-05 3.579 0.871
N3 5.501E-03 2.332 0.773
N4 1.304E-03 2.632 0.773
N5 1.440E-04 3.185 0.887
N6 1.852E-02 2.155 0.881
N7 8.310E-04 2.796 0.821
N8 (prior to 1.170E-04 3.157 0.850
9/1/2004)
 77

Table 7.6. Fatigue transfer function coefficients for 2003 NCAT

Test Track 19.0-mm NMAS base mixes.

Mix k1 k2 R2
PG 67-22 Optimum 7.189E-15 5.782 0.99
PG 67-22 Optimum+ 4.420E-09 4.107 0.98
PG 76-22 Optimum 4.663E-12 5.052 0.92

number of axle repetitions and respective strain for each hour ni = number of load applications at condition i, and
of the 2003 NCAT Test Track loading cycles. Transfer func- Nfi = allowable load applications to failure for condition i
tions were developed from the laboratory beam fatigue tests (in this case, determined using the laboratory fatigue
in the form of Equation 38, as follows: transfer function shown in Equation 38 and Table 7.6).

k2
⎛ 1⎞ Based on the observed level of cracking, Section N6 was
N f = β1 × k1 ⎜ ⎟ (38)
⎝ εt ⎠ assigned a cumulative damage factor of 0.7 after the applica-
tion of 10 million ESALs at the end of the 2003 NCAT Test
where, Track cycle (73). Willis (77) calculated strain for each hour
Nf = number of load repetitions (cycles) to failure, based on the loading conditions using Equation 36 and the
εt = tensile strain at the bottom of the HMA layer, coefficients in Tables 7.4 and 7.5 based on the recorded mid-
k1 and k2 = regression constants, and depth HMA temperature. The spreadsheets were modified to
β1 = shift factor between laboratory and field per- calculate Ni for each section, hour, and loading condition
formance (initially set at 1.0). using Equation 38 and the coefficients in Table 7.6. The num-
ber of load repetitions of a given truck in each hour was divided
The regression constants for the 2003 NCAT Test Track by their respective Ni to determine the incremental damage.
19.0-mm NMAS base mixes, based on the beam fatigue tests The cumulative damage, D, was determined by summing all
conducted as part of this study are shown in Table 7.6. the incremental damage according to Equation 39 until the
The incremental damage was calculated for each hour of date when failure occurred as identified in Table 7.3. For Sec-
loading by truck type (triple trailer or five-axle single trailer) tion N6, the damage was summed until the end of the 2003
using Miner’s Hypothesis, shown in Equation 39, where fail- loading cycle (application of 10 million ESALs). The shift fac-
ure is approached when the cumulative damage approaches tor for each section, β1, in Equation 38 was solved for using
1.0 (78). The failure criterion was defined as fatigue cracking Microsoft Excel’s Solver Function such that the cumulative
equal to 20% of the total lane area. damage, D, = 1.0 on the failure date. For Section N6, the shift
factor was determined for D = 0.7 at the end of the loading
n n
ni cycles. The shift factors were determined with and without
D = ∑ Di = ∑ (39)
i =1 i =1 N f i the inclusion of an endurance limit. The endurance limit
represented by the lower limit of the 95% confidence inter-
where, val was used in the calculations. The calculated shift factors
D = cumulative damage, between laboratory and field performance are summarized
Di = incremental damage for condition (in this case, repeti- in Table 7.7. Based on the data in Table 7.7, the use of an
tions of a given axle’s load at a given HMA temperature endurance limit does not affect the shift factor for pavement
over a given hour), sections likely to fail in fatigue.

Table 7.7. Calculated shift factors between laboratory and

field performance based on measured strains.
Section 19.0-mm NMAS Base Shift Factor Shift Factor with
Mix Endurance Limit
N1 PG 76-22 Optimum 4.24 4.24
N2 PG 67-22 Optimum 75.77 75.77
N6 PG 67-22 Optimum 38.00 38.00
N8 PG 67-22 Optimum+1 8.33 8.33
Note:
1
Until debonding occurred, then PG 67-22 optimum.
78

Shift Factors Based on Calculated Strains ko = at-rest earth pressure coefficient,

Using M-E Design Programs τoct = octahedral shear stress = 1⁄3((σ1 − σ2)2 + (σ1−
σ3)2 + (σ2-σ3)2)1/2, and
In Chapter 6, three classes of M-E pavement design programs
k1, k2, k3 = regression coefficients.
were described, as follows:

1. Those that use equivalent axle loads and equivalent tem- A design resilient modulus was calculated using Equation
peratures; 40 for the pavement structures shown in Figure 7.1. Addi-
2. Those that use equivalent temperatures in the form of tional values were calculated for a 12-in. thick HMA section
seasons and axle load distributions for each axle type; and for use in sensitivity analyses to be described later in this
3. Those that use detailed temperature, load, and incremen- chapter. The design values were calculated using an iterative
tal damage calculations. process as described in the MEPDG (80). The wheel-load
stresses were calculated for a 20,000-lb, single-axle load using
PerRoad is an example of the second type and the MEPDG WESLEA for Windows. Dynamic modulus values for the PG
is an example of the third type of pavement design program. 67-22 and PG 76-22 mixes were calculated at a mean annual
Both programs were used to estimate shift factors between temperature of 74.3°F and 10 Hz frequency. The design resilient
predicted and field performance. modulus values are summarized in Table 7.9. All of the design
The inputs used in the M-E design programs are described values matched the calculated values within 5%. The lower Mr
below. Where possible, the same inputs were used in both Per- values for the granular base were unexpected; however, they
Road and the MEPDG. Resilient modulus (Mr) values were cal- were consistent with previous laboratory testing conducted by
culated for both the subgrade and the granular base. Labora- the Alabama Department of Transportation (81) and back-
tory triaxial resilient modulus tests were performed on samples calculated values from falling-weight deflectometer tests (79).
of both unbound materials according to the NCHRP 1-28A Also, the subgrade consisted of a very angular material with
protocol by Burns, Cooley, Dennis, Inc. The multivariable,
good compactibility and strength.
non-linear stress sensitivity model (Equation 40) recom-
Dynamic modulus testing was conducted on lab-compacted,
mended by the MEPDG was fit to the test data (79, 80). The
field mixed material sampled from the 2003 NCAT Test Track
model coefficients are shown in Table 7.8 for the subgrade
by Purdue University. For the MEPDG, the dynamic modu-
and unbound granular base materials.
lus (E∗) results were used as Level 1 inputs. For PerRoad, a sig-
k2 k3
⎛ θ⎞ ⎡⎛ τ ⎞ ⎤ moidal function, in the form recommended by the MEPDG
M r = k1 pa × ⎜ ⎟ × ⎢⎜ oct ⎟ + 1⎥ (440)
⎝ pa ⎠ ⎣⎝ pa ⎠ ⎦ (82), was fit to the experimental data. The E∗ data are pre-
sented in Table 7.10 for the PG 67-22 and PG 76-22 19.0-mm
where,
NMAS base mixtures. The E∗ master curves for the PG 67-22
Mr = resilient modulus,
pa = atmospheric pressure (14.696 psi), and PG 76-22 mixtures are shown graphically in Figure 7.4.
θ = bulk stress = σ1 + σ2 + σ3 = σ1 + 2σx,y, Observation of the data in Table 7.10 and Figure 7.4 suggests
σ1 = major principal stress = σz + po, that the E∗ values for the PG 67-22 and PG 76-22 mixtures are
σ2 = intermediate principal stress = σ3 for Mr approximately equal.
test on cylindrical specimen, The temperature data from the 2003 cycle of the NCAT
σ3 = minor principal stress/confining pressure Test Track were used for the PerRoad analysis. Figure 7.5
= σx,y + ko (po), shows a frequency distribution of the air temperature data.
σz = vertical stress from wheel load(s) calcu- The temperature corresponding to the 10th, 30th, 50th, 70th,
lated using layered-elastic theory, and 90th percentiles of the frequency distribution were used
σx,y = horizontal stress from wheel load(s) calcu- to identify five “seasons” for the analysis. The air tempera-
lated using layered-elastic theory, tures were converted to pavement temperatures at 1⁄3 of the
po = at-rest vertical pressure from overburden HMA depth using Equation 41 (7). Dynamic modulus values
of paving layers above unbound layer or were then calculated at 10 Hz for five temperatures determined
subgrade, from a frequency distribution of the pavement temperatures

Table 7.8. Unbound materials stress-sensitivity model

coefficients (79).

Material k1 k2 k3 R2
Granite Base 716.28 0.8468 -0.4632 0.93
Subgrade 1878.97 0.4067 -0.7897 0.42
 79

Table 7.9. Design resilient modulus values for 2003 NCAT

Test Track.
Section Binder Grade Total HMA Granular Base Subgrade Mr,
Thickness, in. Mr, psi psi
N1 PG 76-22 5 17,500 26,000
N2 PG 67-22 5 17,500 26,000
N3 PG 67-22 9 12,000 18,000
N4 PG 76-22 9 12,000 18,000
N5 PG 76-22 7 14,000 21,000
N6 PG 67-22 7 14,000 21,000
Perpetual PG 67-22 12 9,000 14,000
Perpetual PG 76-22 12 9,000 14,000

experienced by the structural sections of the 2003 NCAT Test Z = depth below surface at the upper third point of
Track. The modulus values used for determining shift factors the layer, in.
for the test track sections and subsequent sensitivity analyses
are presented in Table 7.11. For PerRoad, the load spectra were determined from the
2003 NCAT Test Track records. The average daily truck traffic
⎛ 1 ⎞ 34 in the design lane was 1,237 trucks. The triple trailers made up
MMPT = MMAT ⎜ 1 + − +6
⎝ Z + 4 ⎟⎠ Z + 4
(41)
88.5% of the truck traffic and the five-axle, single trailer com-
prised the remaining 11.5%. This results in 8,091 axle groups
where, (or repetitions of an axle configuration) per day. The axle
MMPT = mean monthly pavement temperature, °F, configurations had the following distribution: 15.3% steer axles
MMAT = mean monthly air temperature, °F, and (10,000–12,000 lbs); 67.7% single axles (20,000–22,000 lbs); and

Table 7.10. Dynamic modulus data of 2003 NCAT Test Track base
mixtures (after 81).
Temperature, Frequency, PG 76-22 Mixture PG 67-22 Mixture
°F Hz Avg. E*, Std. COV, Avg. E*, Std. COV,
psi Dev. % psi Dev. %
E*, psi E*, psi
0.1 2,277,563 301,583 13.2% 2,298,485 192,823 8.4%
0.5 2,725,367 337,116 12.4% 2,724,243 255,820 9.4%
1 2,907,715 357,532 12.3% 2,897,164 276,114 9.5%
14
5 3,304,176 410,690 12.4% 3,289,201 315,239 9.6%
10 3,502,587 421,351 12.0% 3,421,801 320,689 9.4%
25 3,689,759 450,429 12.2% 3,617,276 355,717 9.8%
0.1 1,153,195 203,310 17.6% 1,072,481 131,115 12.2%
0.5 1,474,779 268,533 18.2% 1,395,625 157,677 11.3%
1 1,623,189 300,838 18.5% 1,543,418 168,029 10.9%
40
5 1,988,503 376,758 18.9% 1,920,335 188,822 9.8%
10 2,148,479 421,027 19.6% 2,087,564 198,544 9.5%
25 2,404,942 474,083 19.7% 2,327,093 210,715 9.1%
0.1 394,829 44,604 11.3% 378,875 67,659 17.9%
0.5 565,357 71,916 12.7% 547,046 83,215 15.2%
1 653,141 84,367 12.9% 643,423 88,274 13.7%
70
5 913,955 125,840 13.8% 930,961 106,425 11.4%
10 1,061,821 150,939 14.2% 1,064,722 116,677 11.0%
25 1,243,734 184,064 14.8% 1,310,597 148,638 11.3%
0.1 151,383 15,348 10.1% 139,490 29,780 21.3%
0.5 214,801 16,759 7.8% 197,650 38,716 19.6%
1 254,106 17,495 6.9% 233,293 42,084 18.0%
100
5 404,945 28,959 7.2% 359,657 57,351 15.9%
10 507,342 34,227 6.7% 445,592 64,055 14.4%
25 614,706 47,519 7.7% 574,095 105,440 18.4%
0.1 65,267 3,879 5.9% 63,055 6,364 10.1%
0.5 83,832 5,467 6.5% 79,553 7,576 9.5%
1 95,036 6,564 6.9% 89,815 9,182 10.2%
130
5 142,898 4,857 3.4% 135,103 19,120 14.2%
10 174,009 4,556 2.6% 167,845 25,931 15.4%
25 218,137 16,623 7.6% 215,345 40,140 18.6%
80

10000

1000
|E*|, ksi

100

PG 67-22
PG 76-22

10
-6.00 -4.00 -2.00 0.00 2.00 4.00
Log Reduced Frequency, Hz

Figure 7.4. Master curves for PG 67-22 and PG 76-22 19.0-mm NMAS base
mixtures at optimum asphalt content.

7000 120%

6000 100%

5000
80%
Frequency

4000
60%
3000
Frequency
40%
2000 Cumulative %

1000 20%

0 0%
10
20
30
40
50
60
70
80
90

0
0
e
10
11

or
M

Air Temperature, ˚F

Figure 7.5. Frequency distribution of air temperature data for the 2003
NCAT Test Track.

Table 7.11. Summary of PerRoad E* inputs.

Sections N1 and N2, 5 in. thick Sections N3 and N4, 9 in. thick Sections N5 and N6, 7 in. thick Perpetual Section
E*, psi E*, psi E*, psi E*, psi
20th Percentile
Length, Air
Season Weeks Temperature, °F MMPT, °F PG 67-22 PG 76-22 MMPT, °F PG 67-22 PG 76-22 MMPT, °F PG 67-22 PG 76-22 MMPT, °F PG 67-22 PG 76-22
Summer 9 80.6 94.8 158,401 168,421 93.3 166,643 177,089 94.0 162,895 173,149 89.6 187,724 199,180
Fall 12 72.2 84.9 219,579 232,367 83.7 229,303 242,454 84.2 224,897 237,886 80.8 252,532 266,478
Winter 10 37.1 43.6 770,325 798,342 43.5 772,034 800,051 43.6 771,269 799,286 44.0 765,289 793,302
Spring 11 53.2 62.6 501,586 526,939 61.9 509,352 534,864 62.2 505,866 531,307 60.9 522,602 548,371
Spring 2 10 65.5 77.1 286,911 301,854 76.0 297,443 312,652 76.5 292,685 307,775 73.8 320,829 336,568
 81

Table 7.12. Shift factors from PerRoad analyses.

Section Shift Factor Based on All Shift Factor Neglecting
Seasons Summer Season
N1 13.4 6.7
N2 45.0 19.2
N6 17.6 NA
NA = Not applicable because Section N6 went through two summer seasons.

17.0% tandem axles (20.6% of which were 32,000–34,000 lbs transfer functions developed as part of this research). The
and 79.4% of which were 40,000–42,000 lbs). For the MEPDG, climate data were determined from the Enhanced Integrated
the triple-trailer trucks were modeled as FHWA Class 13 vehi- Climatic Model using the test track’s coordinates: latitude
cles, neglecting the steer axle, and the five-axle single trailer 32.36, longitude −85.18, elevation 630 ft, and depth to water
truck was modeled as a Class 9 vehicle. table (on north tangent) of 12.5 ft. Three weather stations’
Shift factors were calculated using PerRoad for three sec- data were used in the analysis: Columbus, GA; Troy, AL; and
tions: N1, N2, and N6. It was not felt that Section N8, with Montgomery, AL. The subgrade and granular base were mod-
debonding occurring part way through the loading cycle, eled as Level 2 inputs. The Mr values are summarized in Table
could be modeled with PerRoad. PerRoad runs a 5,000-cycle 7.9. The remaining granular base and subgrade inputs were
Monte Carlo simulation. The default variability was used for taken from Taylor and Timm (79).
the modulus and thickness of the HMA and granular layers. Level 1 inputs were used for the HMA. The entire HMA
Monte Carlo runs are distributed over the five seasons based layer was modeled as the base material for simplicity (recall
on the number of weeks for each season and based on the dis- the surface layer is only 1-in. thick). The dynamic modulus
tribution of axle loads. A resulting strain at the bottom of results are summarized in Table 7.10. The binder proper-
the HMA layer is output for each iteration of the Monte Carlo ties are summarized in Table 7.13. The volumetric proper-
simulation. The strain was converted to damage per axle ties for specific sections were taken from Taylor and Timm
repetition using Equations 38 and 39 and the coefficients in (79). For the sensitivity analyses, described in the next
Table 7.6 corresponding to the mixture used in a particular section, the average volumetric properties were used, as
section. The average damage per axle repetition was calcu- follows:
lated and multiplied by the total number of axle repetitions
when the section failed or, in the case of Section N6, reached • Air voids = 6%,
the end of the 2003 loading cycle. A shift factor was determined • Volume of effective binder = 10.5%, and
for the transfer function for a given section using Microsoft • Unit weight = 150.5 pcf.
Excel Solver by minimizing the difference squared between
the observed and calculated total damage. The shift factors Unless otherwise stated, the MEPDG default values were
are summarized in Table 7.12. Because Sections N1 and N2 used for uncommon items, such as thermal properties.
did not go through a summer before they failed, the summer Unlike PerRoad, the MEPDG does not output its raw
season data were removed and a new average damage and layered elastic calculations. Therefore, a shift factor could not
fatigue transfer function shift factor calculated. This shift be calculated directly in the same manner as was done with
factor is lower since higher strains (and hence more damage) PerRoad. The predicted cracking as a function of time is shown
were observed during the summer season. in Figure 7.6 for Section N1 and in Figure 7.7 for Sections N2
An estimate of predicted cracking for the 2003 Test Track and N6. The MEPDG provides the predicted (50% reliability)
Sections was determined using the MEPDG (Version 1.0) bottom-up fatigue cracking, termed maximum cracking, and
and the nationally calibrated fatigue model (not the fatigue the predicted cracking at some level of reliability, in this case,

Table 7.13. 2003 NCAT Test Track binder properties for MEPDG.
Test PG 67-22 PG 76-22
Temperature, G*, Pa Delta, ° G*, Pa Delta, °
°F
136.4 13.610 73.2 -- --
147.2 6.125 76.7 6.597 65.9
158.0 2.832 80.0 3.683 67.3
168.8 -- -- 2.057 69.1
82

Bottom-Up Cracking - Alligator

100
Maximum Cracking
90
Bottom-Up Reliability
80
Maximum Cracking Limit
Alligator Cracking (%)

70

60

50

40

30

20
20.2% Cracking Observed , 20.1% Predicted, Section N1
10

0
0 6 12 18 24
Pavement Age (month)

Figure 7.6. MEPDG predicted cracking for Section N1 (PG 76-22), 2003 NCAT Test Track.

90%, termed bottom-up reliability in Figures 7.6 and 7.7 (and ing (19.5%) exceeded the maximum cracking (14.4%), but
in the MEPDG output). The MEPDG recommends reliabilities was again less than the 90% reliability cracking (32.5%).
between 85% and 97% for urban interstate-type pavements In the MEPDG, damage is related to predicted cracking
and between 80 and 95% for rural interstate-type pave- according to Equation 42 (84). Note that the minus sign
ments (83). Observation of Figure 7.6 suggests that the max- between C1 and C2 in the equation in El-Basyouny and Witczak
imum cracking (20.1%) closely approximates the observed (74) is incorrect, and should be a plus sign as shown below.
cracking (20.2%) for Section N1 at the 2003 NCAT Test Track.
However, the 90% reliability bottom-up cracking is signif- ⎛ 6, 000 ⎞ ⎛ 1⎞
FC = ⎜ ⎟⎠ × ⎜⎝ ⎟⎠ (42)
icantly higher (38.2%). For Section N2, the observed crack- ⎝ 1+ e C1 + C 2 × Log D 60

Bottom-Up Cracking - Alligator

100
N2 Maximum Cracking
90 N2 Bottom-Up Reliability
Maximum Cracking Limit
80
N6 Maximum Cracking
Alligator Cracking (%)

70 N6 Bottom-Up Reliability

60

50
19.5% Cracking Observed, 14.4% Predicted Section N2
40

30

20

10

0
0 6 12 18 24
Pavement Age (month)

Figure 7.7. MEPDG predicted cracking for Sections N2 and N6 (PG 67-22), 2003 NCAT Test Track.
 83

where, In Chapter 3, a practical definition of a long-life or per-

FC = fatigue cracking (% of lane area), petual pavement is one able to withstand 500 million axle
C1 = 2 × C2, repetitions in a 40-year period without failing. A shift fac-
C2 = –2.40874 – 39.748 x (1+hac)–2.85609 tor of 10 was assumed, resulting in a laboratory equivalent of
hac = thickness of the HMA (in.), 50 million repetitions. Based on the analyses in this section, a
D = cumulative theoretical fatigue damage, %. shift factor of 10 appears reasonable. Varying the shift factor
when determining the endurance limit is not recommended.
Solving Equation 42 for a 5-in.-thick pavement, a cumula-
tive damage of 0.294 (29.4%) corresponds to 20% cracking of
Sensitivity of Mechanistic-Empirical
the lane area. Considering the predicted cracking described
Pavement Design Methods
above and back-calculating the corresponding cumulative
to the Endurance Limit
damage, the shift factor would be approximately 1.0 for Sec-
tion N1, and approximately 0.73 for Section N2. Four sensitivity analyses were conducted to assess the impact
of the endurance limit on pavement design. The first analysis
compared pavement design thicknesses using conventional
Summary of Observed Shift Factors from
and perpetual design procedures. The second analysis looked
2003 NCAT Test Track Structural Sections
at the sensitivity of perpetual designs to the measured value of
The observed shift factors based on the performance of the the endurance limit. The traffic and materials from the 2003
2003 NCAT Test Track structural sections are summarized in NCAT Test Track were used in the first two analyses. Since
Table 7.14. Based on the measured strains and the PerRoad the NCAT Test Track used a limited range of axle weights, the
analyses, the shift factor for the PG 67-22 mix at optimum third and fourth analyses were performed using the materials
asphalt content exceeds the assumed shift factor of 10.0. The from the 2003 NCAT Test Track but the MEPDG’s default
shift factors for the PG 76-22 mix at optimum asphalt con- truck traffic classification No. 1 for principal arterials. The
tent, represented by Section N1, from both the measured third and fourth analyses repeated the first two analyses with
strain and PerRoad analyses, are less than 10.0. The beam a distribution of axle types and weights that are representa-
fatigue results for the PG 76-22 mix were variable, with some tive of typical traffic on a principal arterial.
long fatigue lives observed at relatively high strain levels (see
Table 4.4). This same variability resulted in a reduced 95%
NCAT Test Track Traffic
prediction limit of the endurance limit. The shift factor for
Section N8, the PG 67-22 mix at optimum plus binder content, The 2003 NCAT Test Track pavement section was designed
is slightly less than 10.0; however, this analysis was based on using three methodologies: 1993 AASHTO procedure, MEPDG
several assumptions regarding the measured strains once slip- Version 1.0, and PerRoad Version 3.3. The 1993 AASHTO
page between the layers occurred. design was conducted using a change in pavement service-
The fatigue equations developed from the laboratory testing ability index (PSI) = 1.2, design reliability = 95%, and an
were not used in the MEPDG, rather the NCHRP 1-37A cal- overall standard deviation = 0.45 for 200 million ESALs, the
ibrated fatigue models (64) were used. Based on these analyses, expected traffic over a 40-year period. The structural number
the MEPDG fatigue model reasonably predicts the observed was determined from the AASHTO design equation (42) nu-
cracking. However, design is based on the predicted cracking merically using the bisection method. Two design subgrade
at some level of reliability being less than 25% area cracking. Mr values were used in the analysis, 5,500 psi used in the orig-
The 90% reliability cracking in Sections N1 and N2 exceeded inal design of the 2000 Track (72), and 14,000 psi, the value
the observed cracking by 91% and 68%, respectively. from Table 7.9 used in the perpetual designs. Layer coefficients

Table 7.14. Summary of observed shift factors.

Section Measured Strain PerRoad MEPDG
N1 4.2 6.7 1.0
N2 75.8 19.2 0.73
N6 38.0 17.6 *
N8 8.3 NA NA
Notes:
*Since the MEPDG bases failure on 20% cracking and this did not occur in Section N6, shift factor
could not be calculated.
NA = Not applicable; no attempt was made to model debonding as part of this research.
84

of 0.14 and 0.44 were assigned to the granular base and The results are summarized in Figure 7.8. All three MEPDG
HMA, respectively. design thicknesses are less than that determined from the 1993
AASHTO Pavement Design Guide. Although the 20-year and
log10W18 = Z R × So + 9.36 × log10 ( SN + 1) − 20 perpetual MEPDG designs are the same thickness, the impli-
cations are significantly different. In the first case, at 90%
⎡ ΔPSI ⎤ reliability, bottom-up cracking over 20% of the lane area
log10 ⎢
+ ⎣ 4.2 − 1.5 ⎥⎦ + 2.32 × log M − 8.07 (43) would be expected after 20 years; in the second case, no crack-
10 R
1094
0.40 + ing would be expected after 40 years. This is further illustrated
(SN + 1) 5.19
in Table 7.15 where pavement thickness was iterated in the
MEPDG without specifying an endurance limit and the result-
where, ing damage determined. Thicknesses of 39 in. and 35 in. were
W18 = the number of expected 18-kip ESALs in the design required for the PG 67-22 and PG 76-22 mixes at optimum
lane over the design life 200 × 106, asphalt content to achieve predicted cracking performance
ZR = the normal deviate associated with the chosen level similar to that achieved when an endurance limit was con-
of reliability, 90% = −1.28, sidered (for maximum cracking of 0% to be predicted at the
S0 = materials standard deviation, 0.45, end of 40 years). Some damage was predicted in all of the cases
SN = structural number, tested, including 1.45% bottom-up cracking at 90% reliabil-
ΔPSI = initial minus terminal serviceability, 1.2, and ity, which was predicted to occur in the first month of service.
MR = effective soils resilient modulus, psi. It was expected that the PerRoad perpetual thickness would
be less than that determined with the MEPDG. This expecta-
MEPDG analyses were performed using the inputs described tion was based on the differences in the manner in which both
previously. The three scenarios examined included 20-year and programs handle pavement temperatures. For PerRoad, up
40-year designs with 90% reliability of bottom-up cracking of to five seasons can be specified, with corresponding moduli
less than 25% of the total lane area, and a 40-year perpetual for each season. Typically, this would be based on grouping
analysis where the pavement thickness was selected to provide average monthly temperature data. In this analysis, actual
maximum damage and cracking = 0% at the end of 40 years. temperature data from the 2003 NCAT Test Track cycle were
The PerRoad analyses were performed using the inputs grouped and used in the analyses. This most likely resulted in
described previously. Both the MEPDG and PerRoad perpet- higher temperatures being selected for the warmer season,
ual analyses used the respective one-sided 95% lower predic- which results in correspondingly lower design moduli. Dur-
tion limits of the endurance limit for the PG 67-22 (151 ms) ing PerRoad’s Monte Carlo simulations, modulus is allowed to
and PG 76-22 (146 ms) mixes at optimum asphalt content vary within a season based on a log-normal distribution. The
determined in Phase I of this study. The design thickness was default coefficient of variation used in this study is 30%. For
selected such that approximately 95% of the load applications the MEPDG, temperatures are predicted using the Enhanced
were less than the endurance limit (85). Integrated Climatic Model on an hourly basis. They are then

20
18 18
18 PG 67-22
16 16
16 14.5 14.5 PG 76-22
HMA Thickness, in.

14 13
12 11 11 11
10 10
10
8
6
4
2
0
AASHTO 93, AASHTO 93, M-E PDG, 20 M-E PDG, 40 M-E PDG PerRoad
Mr = 5,500 psi Mr = 14,000 yr. yr. Perpetual Perpetual
psi

Figure 7.8. Comparison of pavement thicknesses from empirical, M-E, and

perpetual design methodologies.
 85

Table 7.15. Damage as a function of pavement thickness for the

MEPDG with no endurance limit.

PG 67-22 at Optimum PG 76-22 at Optimum

90% 90%
Reliability, Reliability,
HMA Maximum Maximum Bottom- Maximum Maximum Bottom-
Thickness, Damage, Cracking, Up Damage, Cracking, Up
In. % % Cracking % % Cracking
10 11.5 9.07 27.18
11 14.5 11.4 29.51 6.57 5.28 23.29*
12 8.92 7.21 25.31
13 5.67 4.61 22.45*
15 2.47 1.99 6.52
20 0.442 0.34 1.79 0.195 0.14 1.59
22 0.109 0.08 1.53
23 0.0838 0.06 1.51
25 0.112 0.08 1.53 0.0501 0.03 1.48
27 0.031 0.02 1.47
30 0.0352 0.02 1.47 0.0159 0.01 1.46
33 0.0191 0.01 1.46
35 0.0132 0.01 1.46 0.00601 0* 1.45
37 0.00927 0.01 1.45
38 0.00782 0.01 1.45
39 0.00664 0* 1.45
40 0.00564 0 1.45 0.00258 0 1.45
41 0.00481 0 1.45
42 0.00413 0 1.45
45 0.00264 0 1.45
Note:
*Indicates minimum thickness with cracking less than 25% of total lane area at the end of the design life.

collected into five “bins” on a monthly basis for determination The second set of analyses examined the sensitivity of the
of layer moduli. This would be expected to result in higher MEPDG and PerRoad to the measured endurance limit using
temperatures occurring at some points during the year and, the NCAT Test Track traffic. Pavement design simulations
hence, lower moduli and higher strains. were conducted using both the PG 67-22 and PG 76-22 mixes
A second difference that may have affected the MEPDG at optimum asphalt content, the previously described pave-
versus the PerRoad results is the way that the layers were sub- ment design parameter, and three levels of the endurance
divided for calculation purposes. For the PerRoad analysis, limit: 70 ms, 100 ms, and the measured endurance limits (151
the HMA was treated as a single layer. The pavement temper- and 146 ms, respectively). The results, illustrated graphically
ature was calculated according to Equation 44 (86). in Figures 7.9 and 7.10 for the PG 67-22 and PG 76-22 mixes,
respectively, indicate that the perpetual pavement design
MMPT = 1.05 × MMAT + 5 (44) thickness is extremely sensitive to the measured endurance

where,
MMPT = mean monthly pavement temperature, °F and
30
MMAT = mean monthly air temperature, °F.
HMA Thickness, in.

25 25
20 20
Equation 44 is representative of the average temperature of 19
16.5
the HMA layer for pavement ≥ 10 in. thick. By comparison, the 15 15
10 11
10-in.-thick MEPDG section was subdivided (automatically)
into seven layers, the top 0.5 in., the next 0.5 in., three 1.0-in. 5
M-E PDG Per Road
sublayers, a 4.0-in. sublayer and a bottom 2.0-in. sublayer. 0
Pavement temperatures and corresponding moduli were cal- 50 100 150 200
Endurance Limit, micro-strains
culated for each of these layers. The net result of this is that the
temperature of the bottom 2-in. layer tends to be lower, result- Figure 7.9. Sensitivity of pavement thickness to
ing in a higher layer moduli, and, therefore, lower strains. endurance limit for PG 67-22 mixes.
86

30 the previously described pavement design parameters, and

HMA Thickness, in.

25 24 three levels of the endurance limit: 70 ms, 100 ms, and the
20 20 measured (151 ms) endurance limit.
15 16 16 The MEPDG’s Level 3 default truck traffic classification
13
10 10
No. 1 and associated axle weight distributions for principal
5
arterials were used for the load spectra. The MEPDG pro-
M-E PDG Per Road duces a file that records the accumulated ESALs on a monthly
0
50 100 150 200 basis throughout the design life of the project. For the NCAT
Endurance Limit, micro-strains Test Track traffic, the MEPDG calculated 171,514,458 ESALs
at the end of 40 years, assuming no growth. The average annual
Figure 7.10. Sensitivity of pavement thickness to
endurance limit for PG 76-22 mixes. daily truck traffic (AADTT) was adjusted using the Level 3
default truck traffic classification No. 1 to produce a similar
number of ESALs after 40 years (171,561,129). An AADTT of
21,833 with a 50% directional split, two lanes in each direc-
limit. The use of polymer modified PG 76-22 has a more sub-
tion, and 90% of the trucks in the design lane were used for
stantial impact on pavement thickness with the MEPDG, as
the calculations.
compared to PerRoad with the difference in thickness ranging
Traffic can be defined in PerRoad in two manners: FHWA
between 1.0 and 3.0 in., depending on the endurance limit.
vehicle class using default axle weight distributions or axle
Larger differences were observed with lower endurance limits.
weight distribution by type (single, tandem, etc.). The MEPDG
default truck traffic classification No. 1 was converted to the
Typical Principal Arterial
format used by PerRoad. The axle load configuration con-
Truck Traffic Classification
sisted of 9,824 axle groups per day in the design lane, of which
The third and fourth set of analyses examined the sensitiv- 45.2% were single axles, 54.3% were tandem axles, and 0.5%
ity of the MEPDG and PerRoad to the measured endurance were tridem axles. The load spectra for the three axle types are
limit using a normal load spectra that might be expected on shown in Figure 7.11.
a principal arterial. Pavement design simulations were con- Figure 7.12 shows the sensitivity of the MEPDG and Per-
ducted using the PG 67-22 mix at optimum asphalt content, Road to the measured endurance limit. Both the MEPDG and

30%

25%

20%
Frequency, %

15%

10%

5%

0%
0

10

20

30

40

50

60

70

80

90

10

11
,

,0

,0

,0

,0

,0

,0

,0

0,

0,
00

00
00

00

00

00

00

00

00

00

0
0

00
0

Axle Weight, lbs

Single Tandem Tridem

Figure 7.11. Axle weight distribution used in PerRoad.

 87

40

35 35

30

HMA Thickness, in.

27
25

20 20
19

15 16
12.5
10

5
ME-PDG Per Road
0
50 100 150 200
Endurance Limit, micro-strains

Figure 7.12. Sensitivity of pavement thickness to the endurance

limit for typical axle distribution.

PerRoad are sensitive to changes in the measured endurance endurance limit. For the MEPDG, the thickness that resulted
limit. For the MEPDG, a change in the endurance limit of in at least 90% reliability against both bottom-up and top-
50 ms results in a change in pavement thickness of approxi- down fatigue cracking was determined. Figure 7.13 shows
mately 7 to 8 in. For PerRoad, a change in the endurance limit a comparison of the conventional and perpetual design
of 50 ms results in a change in pavement thickness of approx- thicknesses.
imately 4 in. This sensitivity highlights the need to measure The PerRoad perpetual design using the measured endur-
the endurance limit as accurately as possible. ance limit of 151 ms is slightly less thick than the 40-year
Finally, design thicknesses were determined using the empirical design using the 1993 AASHTO Design Guide. The
MEPDG and 1993 AASHTO Pavement Design Guide using MEPDG thicknesses considering only fatigue cracking are sim-
design lives of both 20 and 40 years without considering an ilar to the PerRoad thickness. However, at an 11-in. pavement

20 19

18

16
14
14 13
HMA Thickness, in.

12.5
12
12 11

10

0
AASHTO 93, AASHTO 93, M-E PDG, M-E PDG, M-E PDG PerRoad
20-yr. 40-yr. 20-yr. 40-yr. Perpetual Perpetual

Figure 7.13. Comparison of pavement thicknesses from empirical, M-E,

and perpetual design methodologies for typical axle distribution.
88

thickness, the 20-year MEPDG design fails the reliability 20% of the lane area to be cracked at the end of the design
criteria for terminal international roughness index (IRI), life whereas the perpetual pavements would not be expected
total pavement rutting, and HMA rutting. The HMA thick- to have any cracking. This significantly changes the required
ness must be increased to 14 in. to produce an acceptable maintenance and rehabilitation requirements in a life-cycle
reliability for terminal IRI. The HMA thickness must be cost analysis.
increased to 30 in. to produce an acceptable reliability against
total pavement rutting. The reliability for HMA rutting can not
Considerations for Incorporating
be achieved for this level of traffic in the NCAT Test Track’s
the Endurance Limit into
climate using a PG 67-22 binder. The perpetual thickness
M-E Design Procedures
determined using the MEPDG is significantly thicker than
that determined using PerRoad when considering a typical In the preceding section, sensitivity analyses were presented
axle distribution. demonstrating the affect of incorporating the endurance limit
Although the pavement thicknesses appear similar between into two M-E programs, the MEPDG and PerRoad. Certainly
conventional empirical or mechanistic designs and the per- the predicted performance from the MEPDG in terms of
petual pavement thickness determined using PerRoad, again bottom-up cracking was improved compared to a conven-
the implications are very different. Based on the conven- tional 20-year design. Using the experimentally determined
tional MEPDG results, the 11-in. HMA pavement would be endurance limits from this study, there was no increase in the
expected to have maximum cracking of 4.8% of the lane design thickness determined using the MEPDG for a 20-year
area after 20 years. The 90% reliability for bottom-up crack- or perpetual design. The thicknesses determined, 11.0- and
ing is 22.72% of the lane area. Similarly, for the 12-in. pavement 10.0-in., respectively, for the PG 67-22 and PG 76-22 mixes
after 40 years, maximum bottom-up cracking was predicted at optimum asphalt content are consistent with Nunn’s (10)
at 5.9% with a 90% reliability of 23.99% of the lane area. Based recommendations that long-life pavements should range
on the conventional analysis, the pavement will have failed between 7.9 and 15.4 in. Further, Section N3 and N4 of the
and be in need of reconstruction, whereas the perpetual 2003 NCAT Test Track have now gone through two test track
analysis suggests that at a similar thickness there should be loading cycles without any observed fatigue cracking (77). This
no bottom-up fatigue cracking after 40 years. This difference indicates that pavement thicknesses close to those designed
would have a significant effect on life-cycle cost analysis. as perpetual pavements (N3 and N4 are 9 in. thick) with
the MEPDG are performing well after a fairly high number
(20 million ESALs) of load applications. However, the sen-
Summary of Sensitivity Analyses
sitivity of the required pavement thickness to the measured
The design thickness for a perpetual pavement is very endurance limit also has been demonstrated, as well as
sensitive to the measured endurance limit using both the the apparent sensitivity to temperature as evidenced by the
MEPDG and PerRoad. Considering a typical traffic stream, increased pavement thicknesses determined using PerRoad.
a 50-ms change in the endurance limit resulted in a 7- to 8-in. Therefore, consideration should be given as to whether the
or 4-in. change in HMA thickness, respectively, with the endurance limit is really best represented by a single value,
MEPDG and PerRoad. This sensitivity highlights the need for determined at a single temperature.
accurate determination of the endurance limit. To improve One hypothesis is that the fatigue endurance limit is driven,
accuracy, the number of strain levels used to predict the in part, by the ability of asphalt mixtures to heal. Healing
endurance limit in Appendix A was increased from two to occurs more readily at higher temperatures. Therefore, a
three with three replicates at each level from that used in mixture’s fatigue capacity or endurance limit may be higher at
Phase II of the study. Additional samples should reduce the higher temperatures. Testing was only conducted at a single
standard error of the log-log regression and result in a smaller temperature, 20°C, as part of this study. Tsai et al. (87) tested
t-value when calculating the lower, one-sided prediction limit mixes at three temperatures, 10°C, 20°C, and 30°C, as part of
(endurance limit). a reflective-cracking study. A total of six samples was tested
Again considering a normal traffic stream, the PerRoad at each temperature, three each at two strain levels. The sam-
perpetual design thickness was slightly less than that deter- ples were compacted to 6% air voids (target). Five binders
mined using the 1993 AASHTO Pavement Design Guide, and were tested: AR-4000, type G asphalt rubber (RAC-G), and
approximately the same as that required to satisfy the fatigue three modified binders termed MB4, MB15, and MAC16.
requirements of a 20- or 40-year MEPDG design (not consid- The endurance limit was predicted from published data
ering the endurance limit). However, the predicted conditions included in Tsai et al. (87) using the procedure described in
of the pavement at the end of the design life are significantly Appendix A. The results are shown in Table 7.16. Varia-
different. At 90% reliability, the MEPDG would predict over tions in the predicted endurance limit were observed both
 89

Table 7.16. Predicted endurance limit as a function of

test temperature.
Predicted Lower 95%
Test Temperature,
Binder Endurance Limit, Prediction Limit,
°C
ms ms
10.2 101 39
AR-4000 19.9 52 12
30.4 105 80
10.3 130 91
RAC-G 20.2 190 106
30.0 183 124
10.0 176 127
MAC15 20.0 255 178
30.3 461 100
9.9 377 284
MB4 20.5 394 230
29.8 555 247
10.1 348 261
MB15 20.0 215 135
30.4 171 82

with changes in test temperature and binder. Three of the Based on measured strains from the NCAT Test Track from
binders generally followed the expected trend of increasing sections that have not experienced fatigue cracking, Willis (77),
endurance limit with increasing test temperature. In two of proposed designing perpetual pavements based on a cumula-
the cases, the predicted endurance followed the expected tive frequency distribution of allowable strains. A similar con-
trend, while the 95% lower prediction limit was more vari- cept was initially proposed by Priest (76). The proposed upper
able, due to variability in the beam fatigue test results. For bound for a cumulative frequency distribution of endurance
the RAC-G, the 95% lower prediction limit showed a trend limit strain is shown in Table 7.17 for Sections N3 and N4 of
of increasing endurance limit with increasing temperature. the NCAT Test Track. A cumulative frequency distribution
The MB15 binder indicated decreasing endurance limit with defines the percentage of observed data below a given value.
increasing temperature. The AR-4000 binder indicated its Based on Table 7.17, 50% of the in-service strain values should
lowest endurance limit at 20°C. be less than 181 ms to prevent fatigue cracking. It should be

Table 7.17. Cumulative distribution of strain criteria for

long-life pavements (77).
Percentile Upper Bound Fatigue Maximum Fatigue Ratio
Limit
99 394 2.83
95 346 2.45
90 310 2.18
85 282 1.98
80 263 1.85
75 247 1.74
70 232 1.63
65 218 1.53
60 205 1.44
55 193 1.35
50 181 1.27
45 168
40 155
35 143
30 132
25 122
20 112
15 101
10 90
5 72
1 49
90

noted that the mean annual air temperature measured during the 55th percentile. It should be reiterated that these distri-
the 2003 NCAT Test Track cycle was 65.5°F, which corre- butions are based on two sections for which bottom-up fatigue
sponds to a pavement temperature (based on Equation 44) of cracking has not been observed after the application of approx-
73.8°F. This is greater than the 20°C (68.8°F) test temperature imately 20 million ESALs. It is possible that fatigue cracking
used for the beam fatigue tests. Hence, the fact that the 50% could occur with additional loading.
strain values are greater than the endurance limits measured The last concept that needs to be considered in long-life
for this study is not unexpected. Table 7.17 also presents strain pavement design is how different rates of loading may be
ratios, which are ratios of the upper bound for the allowable accommodated. In addition to designing against damage
strain at a given percentile to the measured endurance limits from expected axle loads, frequency of application needs to
(151 ms and 146 ms for the PG 67-22 and PG 76-22 mixes, be considered. Low volume roads may, in some cases, experi-
respectively) determined as part of this study. This offers an ence the same distribution of axle loads over time, but differ-
opportunity to adjust the distribution based on measured ent frequencies of application. Infrequent load applications
material properties. may offer more time for healing to occur and hence less accu-
Both concepts were developed based on observations mulated damage. If load frequency is not considered, all per-
that the cumulative frequency distribution of measured petual pavements designed for the same distribution of axles
strains of sections that did and did not crack differed above on the same subgrade will have the same thickness.
 91

CHAPTER 8

Conclusions and Recommendations

Conclusions Additional evaluation is needed to reconcile the difference

between beam and uniaxial fatigue results, which produced
The following conclusions may be drawn based on the data trends for neat versus polymer modified binders that were
presented to date: the opposite of those determined with beam fatigue testing.
• Field observations, from the data discussed in Chapter 6
• A practical definition of the endurance limit or long-life and from the NCAT Test Track, support the importance
pavement would be a pavement able to withstand 500 mil- of good construction in addition to pavement thickness
lion design load repetitions in a 40-year period. design and materials selection.
• Several techniques can be used to evaluate beam fatigue data • Shift factors between laboratory and field performance,
near the endurance limit. These include logarithmic extrap- based on fatigue transfer functions developed as part of this
olation of the loading cycle versus stiffness curve, single- research, ranged from 4.2 to 75.8.
or three-stage Weibull model using the stiffness ratio, and • Pavement thicknesses for a perpetual design determined
ratio of dissipated energy change (RDEC). The single-stage using the endurance limits measured as part of this study
Weibull model produced fairly accurate extrapolations that were consistent with thicknesses observed in previous
appear to be conservative. Extrapolations performed using studies of in-service pavements.
the single-stage Weibull model resulted in the lowest vari- • The MEPDG and PerRoad perpetual design methodologies
ability in the mini round-robin. Therefore, the single-stage are sensitive to changes in the endurance limit.
Weibull model is recommended for extrapolating low strain • Considering a typical principal arterial traffic stream, the
fatigue tests to confirm the existence of the endurance limit. thickness of a perpetual pavement designed using PerRoad
In certain cases, the three-stage Weibull model may pro- was similar to that determined using the 1993 AASHTO
vide a better fit to the experimental data. Procedures for Pavement Design Guide or MEPDG (without the endur-
both methods are supplied in Appendix A. ance limit) for a 20- or 40-year design life. The thickness
• The data support the existence of an endurance limit for of a perpetual pavement designed with the MEPDG was
each of the six mixes tested. The 95% lower prediction limit approximately 50% thicker. The predicted condition of
varied from 75 to 200 ms. the pavement at the end of 20 to 40 years was signifi-
• All of the estimated endurance limits were above 70 ms. An cantly different, with no cracking expected in the perpet-
analysis of LTPP data indicated an endurance limit of 65 ms. ual pavements versus over 20% of lane area cracking at
• On a log basis for normal strain fatigue tests, the repeata- 90% reliability (based on the MEPDG). Damage would
bility (within-lab) standard deviation was determined to be also be expected based on a change in serviceability index
0.248 and the reproducibility (between-lab) standard devi- of 1.2 for the 1993 AASHTO design procedure.
ation was determined to be 0.318. This results in within-
and between-lab coefficients of variation of 5.4% and 6.8%,
Recommendations
respectively.
• Uniaxial tension testing provides a promising technique. Recommendations from this study address the following
Results from the uniaxial tension test can be determined five areas: (1) investigation of the endurance limit as a mixture
more quickly than beam fatigue tests, but the data are more property, (2) additional research and development to further
complicated to analyze. There are difficulties controlling the development of the uniaxial tension test, (3) field testing
the strain that is actually applied to the sample being tested. to investigate cracking observed in thicker LTPP sections that
92

appear to refute the existence of the endurance limit, (4) in- ments in the LTPP database that exhibit cracking, to deter-
corporation of the endurance limit into pavement design, mine the cause of that cracking.
and (5) cataloging endurance limit values. For pavements designed using equivalent annual or equiv-
Only a single gradation and two aggregate sources blended alent seasonal temperatures, the use of a single value for the
in a single mix were tested in this study. Further, only a sin- endurance limit appears to be reasonable. However, field data
gle form of binder modification was evaluated. Table 2.1 presented from the NCAT Test Track indicate that pavements
presented the affect of a range of factors on fatigue life (1). can withstand a cumulative distribution of strains (which
Of these, binder stiffness and air void content were expected includes strain levels that exceed the mixtures’ endurance limit,
to have a larger affect than aggregate type and gradation. determined at a single temperature, as described in this study)
Differences were observed in the predicted endurance limit and still exhibit perpetual behavior (77). Further, there is evi-
based on binder stiffness. The affect of binder content and dence that the endurance limit, determined from beam fatigue
in-place air voids was mixed. Few affects were observed for tests, varies as a function of temperature. Thus, future efforts
the PG 67-22, but a more pronounced affect was observed for to incorporate the endurance limit into the MEPDG should
the polymer modified PG 76-22. Additional evaluations should consider a distribution of acceptable strains or endurance lim-
be conducted with a wider range of mixtures, binder types, its that vary as a function of temperature.
and modified binders. Agencies should use Appendix A, Proposed AASHTO
The samples tested in this study were short-term samples, Practice for Beam Fatigue Testing, to determine catalogs of
oven aged for four hours at 275°F (135°C) according to endurance limit values for typical mixes, binder grades, and
AASHTO R30. No long-term aging was evaluated. Nunn (10) binder sources. Only a single aggregate source and gradation
indicates that the stiffness of thick pavements increases with were tested in this study. Therefore, it is difficult to assess
time. This should reduce the strain at the bottom of the as- the affect of those parameters on the endurance limit. Based
phalt layer. However, the endurance limit or strain “capacity” on the literature review, these seem to be secondary factors.
of the mix may decrease with increased oxidative aging and The endurance limit catalog should concentrate on mixes
physical hardening. The affect of long-term aging on the en- used at the bottom of the HMA layer where cracking initi-
durance limit should be investigated. ates. An experimental plan should include the binder grades
Samples for the uniaxial tension test can be prepared on and types of modifiers the agency typically uses in the bottom-
a gyratory compactor. Uniaxial tension testing is less time- lift, laboratory compaction efforts (e.g., Ndesign) if historical
consuming than beam fatigue testing. Therefore, this test experience indicates these different levels result in different
method deserves additional development. Basic research optimum asphalt content, and major aggregate types. If the
should be conducted to better understand the stress states in agency is considering the use of a rich-bottom layer, with or
uniaxial tension samples. Better techniques should be devel- without polymer modified binder or a high-modulus base,
oped to control the strain experienced by the test sample. they should also be included. Laboratory samples should be
Forensic investigations should be conducted on a stratified compacted to a density representative of the level typically
random sample of the thickest GPS-1, GPS-2, and SPS-1 pave- achieved in the field.
 93

References

1. Epps, J. A. and C. L. Monismith. “Fatigue of Asphalt Concrete 14. Newcomb, D. APA 101: Perpetual Pavements: A Synthesis. Asphalt
Mixtures-Summary of Existing Information.” ASTM STP 508: Pavement Alliance, Lanham, MD, 2002.
Fatigue of Compacted Bituminous Aggregate Mixtures. American So- 15. Harvey, J. T. and B. W. Tsai. “Effects of Asphalt Content and Air Void
ciety for Testing and Materials, Coshocken, PA, 1972, pp. 19–45. Content on Mix Fatigue and Stiffness.” In Transportation Research
2. Tangella, S. C. S. R., J. Craus, J. A. Deacon, and C. L. Monismith. Record No. 1543, Transportation Research Board, Washington, DC,
“Summary Report on Fatigue Response of Asphalt Mixtures.” 1996, pp. 38–45.
TM-UCB-A-003A-89-3, Strategic Highway Research Program, 16. Monismith, C. L. “Some Applications of Theory in the Design of
Project A-003-A, University of California, Berkley, CA, 1990. Asphalt Pavements.” Fifth Annual Nevada Streets and Highway
3. Tayebali, A. A., J. A. Deacon, J. S. Coplantz, J. T. Harvey, and C. L. Conference, University of Nevada, 1970.
Monismith. “Fatigue Response of Asphalt-Aggregate Mixes: Part I 17. Goodrich, J. L. “Asphalt and Polymer Modified Asphalt Properties
Test Method Selection.” SHRP A-404 Strategic Highway Research Related to the Performance of Asphalt Concrete Mixes.” In Proceed-
Program, National Research Council, Washington, DC, 1994. ings of Association of Asphalt Paving Technologists Technical Sessions,
4. Callister, W. D. Jr. Materials Science and Engineering: An Introduction. Vol. 57, Williamsburg, VA, 1988, pp. 116–175.
John Wiley & Sons, New York, 1985. 18. Tayebali, A. A., J. A. Deacon, J. S. Coplantz, F. N. Finn. and C. L.
5. Barret, C. R., W. D. Nix, and A. S. Tetelman. The Principles of Engi- Monismith. “Fatigue Response of Asphalt-Aggregate Mixes: Part II
neering Materials. Prentice Hall, Englewood Cliffs, NJ, 1973. Extended Test Program.” SHRP A-404 Strategic Highway Research
6. Baladi, G. Y. and M. B. Snyder. “Highway Pavements.” Pavement Program, National Research Council, Washington, D.C. 1994.
Design, Publication No. FHWA HI-90-027, Vol. II, Part III, National 19. Bahia, H. U., D. I. Hanson, M. Zeng, H. Zhai, M. A. Khatri, and
Highway Institute, Washington, DC, 1992. R. M. Anderson. “Characterization of Modified Asphalt Binders
7. Huang, Y. H. Pavement Analysis and Design. Prentice Hall, Upper
in Superpave Mix Design.” NCHRP Report 459, Transportation
Saddle River, NJ, 1993.
Research Board, National Research Council, Washington, DC,
8. Monismith, C. L. and D. B. McLean. “Structural Design Consid-
2001, p. 95.
erations.” In Proceedings of the Association of Asphalt Paving Tech-
20. Monismith, C. L., F. Long, and J. T. Harvey. “California’s Interstate-
nologists, Vol. 41, 1972.
710 Rehabilitation: Mix and Structural Section Designs, Construc-
9. Maupin, G. W. Jr. and J. R. Freeman Jr. “Simple Procedure for
tion Specifications.” Journal of the Association of Asphalt Paving
Fatigue Characterization of Bituminous Concrete.” Final Report
Technologists, Vol. 70, Clearwater Beach, FL, 2001, pp. 762–799.
No. FHWA-RD-76-102, Federal Highway Administration, Wash-
ington, DC, 1976. 21. Lee, H. J., J. Y. Choi, Y. Zhao, and Y. R. Kim. “Laboratory Evaluation
10. Nunn, M. “Long-life Flexible Roads.” In Proceedings of the 8th Inter- of the Effects of Aggregate Gradation and Binder Type on Perfor-
national Conference on Asphalt Pavements, Vol. 1, University of mance of Asphalt Mixtures.” In Proceedings of the Ninth International
Washington, Seattle, WA, August 1997, pp. 3–16. Conference on Asphalt Pavements, Copenhagen, Denmark. 2002.
11. Nishizawa, T., S. Shimeno, and M. Sekiguchi. “Fatigue Analysis 22. Asphalt Pavement Alliance. “Determination of Threshold Strain
of Asphalt Pavements with Thick Asphalt Mixture Layer.” In Level for the Fatigue Endurance Limit in Asphalt Mixtures: Phase I.”
Proceedings of the 8th International Conference on Asphalt Pave- Asphalt Pavement Alliance, 2003.
ments, Vol. 2, University of Washington, Seattle, WA, August 1997, 23. Von Quintus, H. “Quantifications of the Effects of Polymer-Modified
pp. 969–976. Asphalt for Reducing Pavement Distress.” ER-215, Asphalt Institute,
12. Wu, Z., Z. Q. Siddique, and A. J. Gisi. “Kansas Turnpike—An Lexington, KY 2004.
Example of Long Lasting Asphalt Pavement.” Proceedings Inter- 24. Harvey, J., C. Monismith, R. Hornjeff, M. Berjarano, B. W. Tsai,
national Symposium on Design and Construction of Long Lasting and V. Kannekanti. “Long-Life AC Pavements: A Discussion of
Asphalt Pavements, National Center for Asphalt Technology, Auburn, Design and Construction Experience Based on California Expe-
AL, 2004, pp. 857–876. rience.” In Proceedings of International Symposium on Design and
13. Mahoney, J. P. “Study of Long-Lasting Pavements in Washington Construction of Long Lasting Asphalt Pavements, National Center
State.” In Transportation Research Circular 503: Perpetual Bitumi- for Asphalt Technology, Auburn, AL, 2004, pp. 285–333.
nous Pavements, Transportation Research Board, Washington, DC, 25. Harvey, J. T. and B. W. Tsai. “Effects of Asphalt Content and Air Void
2001, pp. 88–95. Content on Mix Fatigue and Stiffness.” In Transportation Research
94

Record No. 1543, Transportation Research Board, Washington, DC, Record No. 1929, Transportation Research Board, Washington, DC,
1996, pp. 38–45. 2005, pp. 165–173.
26. Anderson, R. M. and R. Bentsen. “Influence of Voids in Mineral 41. El-Basyouny, M. M. and M. Witczak. “Development of the Fatigue
Aggregate (VMA) on the Mechanical Properties of Coarse and Cracking Models for the 2002 Design Guide.” Presented at 84th
Fine Asphalt Mixtures.” Journal of the Association of Asphalt Paving Annual Meeting of The Transportation Research Board, Transporta-
Technologists, Vol. 70, Clearwater Beach, FL, 2001, pp. 1–37. tion Research Board, Washington, DC, 2005.
27. Maupin, G. W. Jr. “Additional Asphalt to Increase the Durability of 42. American Society for Testing and Materials. “Standard Practice for
Virginia’s Superpave Surface Mixes.” VTRC 03-R15, Virginia Trans- Statistical Analysis of Linear or Linearized Stress-Life (S-N) and
portation Research Council, Charlottesville, VA, 2003. Strain-Life (ε-N) Fatigue Data.” E739-91, In ASTM Standards on Pre-
28. Corte, J. F. “Development and Uses of Hard-Grade Asphalt and cision and Bias for Various Applications, Conshohocken, PA, 1992.
of High-Modulus Asphalt Mixtures in France.” In Transportation 43. Harvey, J. T., J. A. Deacon, A. A. Tayebali, and R. B. Leahy. “A
Research Circular 503: Perpetual Bituminous Pavements, Transporta- Reliability-Based Mix Design and Analysis System for Mitigating
tion Research Board, Washington, DC, 2001, pp. 12–31. Fatigues Distress.” In Proceedings of the 8th International Confer-
29. Ferne, B. W. and M. Nunn. “The European Approach to Long ence on Asphalt Pavements, Vol. 1. University of Washington,
Lasting Asphalt Pavements—A State of the Art Review by ELLPAG.” Seattle, WA, August 1997, pp. 301–323.
Proceedings International Symposium on Design and Construction of 44. Savard, Y., M. Boutonnet, C. Mauduit, and N. Pouliot. “Compar-
Long Lasting Asphalt Pavements, National Center for Asphalt Tech- ison of Pavement Design Methods in France and Quebec.” In Pro-
nology, Auburn, AL, 2004, pp. 87–101. ceedings of International Symposium on Design and Construction of
30. Tayebali, A. A., J. A. Deacon, J. S. Coplantz, and C. L. Monismith. Long Lasting Asphalt Pavements, National Center for Asphalt
“Modeling Fatigue Response of Asphalt-Aggregate Mixtures.” Jour- Technology, Auburn, AL, 2004, pp. 153–198.
nal of Association of Asphalt Paving Technologists, Vol. 62, Austin, 45. Deacon, J., J. Coplantz, F. Finn, and C. Monismith. “Temperature
TX, 1993, pp. 385–421. Considerations in Asphalt-Aggregate Mixture Analysis and Design.”
31. Lee, H. J., J. Y. Choi, Y. Zhao, and Y. R. Kim. “Laboratory Evaluation In Transportation Research Record No. 1454, Transportation Research
of the Effects of Aggregate Gradation and Binder Type on Perfor- Board, Washington, DC, 1994, pp. 97–112.
mance of Asphalt Mixtures.” In Proceedings of the Ninth International 46. Leahy, R. B., R. G. Hicks, C. L. Monismith, and F. N. Finn. “Frame-
Conference on Asphalt Pavements, Copenhagen, Denmark, 2002. work for Performance—Based Approach to Mix Design and Analy-
sis.” In Proceedings of the Association of Asphalt Paving Technologists,
32. Daniel, J. S. and Y. R. Kim, “Development of a Simplified Fatigue
Vol. 64, Association of Asphalt Paving Technologists, Minneapolis,
Test and Analysis Procedure Using a Viscoelastic, Continuum
MN, 1995, pp. 431–473.
Damage Model.” Journal of the Association of Asphalt Paving Tech-
47. Pierce, L. M. and J. P. Mahoney. “Asphalt Concrete Overlay Design
nologists, Vol. 71, 2002, pp. 619–645.
Case Studies.” In Transportation Research Record No. 1543, Trans-
33. Chehab, G. R., Y. R. Kim, R. A. Schapery, M. W. Witczak, and
portation Research Board, Washington, DC, 1996, pp. 3–9.
R. Bonaquist. “Time-Temperature Superposition Principle for
48. Highway Capacity Manual. Transportation Research Board, National
Asphalt Concrete Mixtures with Growing Damage in Tension
Research Council, Washington, DC, 2000.
State.” Journal of Association of Asphalt Paving Technologists, Vol. 71,
49. Timm, D. H. and A. L. Priest. “Dynamic Pavement Response Data
2002, pp. 559–593.
Collection and Processing at the NCAT Test Track.” NCAT Report
34. Ghuzlan, K. and S. Carpenter. “Energy-Derived, Damage-Based
04-03, National Center for Asphalt Technology, Auburn University,
Failure Criterion for Fatigue Testing.” In Transportation Research Auburn, AL, June 2004.
Record No. 1723, Transportation Research Board, Washington, DC, 50. University of Washington Civil Engineering Department. “Trucks
2000, pp. 141–149. and Buses.” http://training.ce.washington.edu/WSDOT/Modules/
35. Rowe, G. M. “Performance of Asphalt Mixtures in the Trapezoidal 04_design_parameters/trucks_buses.htm Accessed August 18, 2006.
Fatigue Test.” Journal of the Association of Asphalt Paving Technol- 51. Deacon, J. A., A. A. Tayebali, J. S. Coplantz, F. N. Finn, and C. L.
ogists, Vol. 62, Austin, TX, 1993, pp. 344–384. Monismith. “Fatigue Response of Asphalt-Aggregate Mixes: Part III
36. Carpenter, S. H., K. A. Ghuzlan, and S. Shen. “A Fatigue Endurance Mix Design and Analysis.” SHRP A-404, Strategic Highway Research
Limit for Highway and Airport Pavements.” In Transportation Program, National Research Council, Washington, DC, 1994.
Research Record No. 1832, Transportation Research Board, Wash- 52. American Association of State Highway and Transportation Officials.
ington, DC, 2003, pp. 131–138. “Standard Specifications for Transportation Materials and Methods
37. Thompson, M. R. and S. H. Carpenter. “Design Principles for of Sampling and Testing: Part 2B: Tests,” 25th ed., Washington, DC,
Long Lasting HMA Pavements.” In Proceedings of International 2005.
Symposium on Design and Construction of Long Lasting Asphalt 53. Rowe, G. M. and M. G. Bouldin. “Improved Techniques to Evalu-
Pavements. National Center for Asphalt Technology, Auburn, AL, ate the Fatigue Resistance of Asphaltic Mixes.” In 2nd Eurasphalt
2004, pp. 365–384. and Eurobitume Congress Barcelona 2000–Proceedings 0081,UK.
38. Peterson, R. L. and P. Turner. “Determination of Threshold Strain 54. Personal communication with S. Shen, January 2005.
Level for Infinite Fatigue Life in Asphalt Mixtures.” Interim Report 55. Tsai, B. W., J. T. Harvey, and C. L. Monismith. “High Temperature
for Asphalt Pavement Alliance, Asphalt Institute, Lexington, KY, Fatigue and Fatigue Damage Process of Aggregate-Asphalt Mixes.”
2002. Journal of the Association of Asphalt Paving Technologists, Vol. 71,
39. Peterson, R. L., P. Turner, M. Anderson, and M. Buncher. “Deter- 2002, pp. 345–385.
mination of Threshold Strain Level for Fatigue Endurance Limit in 56. Tsai, B.W. “High Temperature Fatigue and Fatigue Damage Process
Asphalt Mixtures.” In Proceedings of International Symposium on of Aggregate-Asphalt Mixes.” Report to California Department of
Design and Construction of Long Lasting Asphalt Pavements, National Transportation, Pavement Research Center, Institute of Transporta-
Center for Asphalt Technology, Auburn, AL, 2004, pp. 385–410. tion Studies, University of California, Berkley, CA,
40. Shen, S. and S. H. Carpenter. “Application of Dissipated Energy Con- 57. Tsai, B. W., J. T. Harvey, and C. L. Monismith. “Using the Three-
cept in Fatigue Endurance Limit Testing.” In Transportation Research Stage Weibull Equation and Tree-Based Model to Characterize the
 95

Mix Fatigue Damage Process.” In Transportation Research Record tress.” Final Report No. 5504-2/2, Applied Research Associates,
No. 1929, Transportation Research Board, Washington, DC, 2005, Inc., prepared for the Asphalt Institute, October 2004.
pp. 227–237. 72. Timm, D. H., A. L. Priest, and T. V. McEwen. “Design and Instru-
58. Tsai, B. W., V. N. Kannekanti, and J. T. Harvey. “Application of mentation of the Structural Pavement Experiment at the NCAT
Genetic Algorithm in Asphalt Pavement Design.” In Transporta- Test Track.” NCAT Report 04-01, Auburn, AL, April 2004.
tion Research Record No. 1891, Transportation Research Board, 73. Priest, A. L. and D. H. Timm. “Methodology and Calibration of
Washington, DC, 2004, pp. 112–120. Fatigue Transfer Functions for Mechanistic-Empirical Flexible
59. Von Quintus, Harold L., J. A. Scherocman, C. S. Hughes, and Pavement Design.” NCAT Report 06-03, Auburn, AL, December
T. W. Kennedy. “Asphalt-Aggregate Mixture Analysis System: 2006.
AAMAS.” NCHRP Report No. 338, National Cooperative Highway 74. El-Basyouny, M. M. and M. Witczak. “Calibration of Alligator
Research Program, National Research Council, Washington, DC, Fatigue Cracking Model for 2002 Design Guide.” In Transporta-
March 1991. tion Research Record No. 1919, Transportation Research Board,
60. Kim, Y. R. and H. Wen. “Fracture Energy from Indirect Tension National Academies, Washington, DC, 2005, pp. 77–86.
Testing.” Journal of the Association of Asphalt Paving Technologists, 75. Willis, J. R. and D. H. Timm. “Forensic Investigation of a Rich-
Vol. 71, Colorado Springs, CO, 2002, pp. 779–793. Bottom Pavement.” NCAT Report 06-04, Auburn, AL, December
61. Christensen, D.W. and R.F. Bonaquist. “Practical Application of 2006.
Continuum Damage Theory to Fatigue Phenomena in Asphalt 76. Priest, A. L. “Calibration of Fatigue Transfer Functions for
Concrete Mixtures.” Journal of the Association of Asphalt Paving Mechanistic-Empirical Pavement Design.” Master’s Thesis. Auburn
Technologists, Vol. 74, 2005, pp. 963–1002. University, Auburn, AL, 2005.
62. Computer Program DAMA (CP-1/1991 Revision). “Pavement 77. Willis, J. R. “Field-Based Strain Thresholds for Flexible Perpetual
Structural Analysis Using Multi-Layered Elastic Theory.” Asphalt Pavement Design.” Doctoral Dissertation. Auburn University,
Institute, Lexington, KY, 1991. Auburn, AL, 2008.
63. Timm, D. H. Computer Program PerRoad. Version 3.3, Auburn 78. Miner, M. A. “Estimation of Fatigue Life with Particular Emphasis
University, February 2008. on Cumulative Damage.” Metal Fatigue, Sines and Waisman, eds.,
64. Applied Research Associates and Arizona State University. “Com- McGraw Hill, Inc., New York, 1959, pp. 278–89.
puter Program Mechanistic Empirical Pavement Design Guide 79. Taylor, A. J. and D. H. Timm. “Mechanistic Characterization of
Version 1.003.” May 2007. Resilient Moduli for Unbound Pavement Layer Materials.” NCAT
65. Gharaibeh, N. G. and M. I. Darter. “Longevity of Highway Pave- Report 09-06, Auburn, AL, 2009.
ments In Illinois—2000 Update.” Final Report FHWA-IL-UI-283, 80. ARA, Inc. Guide for Mechanistic-Empirical Design of New and Re-
Illinois Department of Transportation, Springfield, Illinois, 2002. habilitated Pavement Structures. Final Report, Part 2: Design In-
66. Gharaibeh, N. G. and M. I. Darter. “Probabilistic Analysis of High- puts, “Chapter 1: Subgrade/Foundation Design Inputs,” NCHRP,
way Pavement Life for Illinois.” In Transportation Research Record Transportation Research Board, Washington, DC, 2004.
No. 1823, Transportation Research Board, Washington, DC, 2003, 81. Timm, D. H. and A. L. Priest. “Material Properties of the 2003
pp. 111–120. NCAT Test Track Structural Study.” NCAT Report 06-01, Auburn,
67. Smith, K. L., N. G. Gharaibeh, M. I. Darter, H. L. Von Quintus, AL, 2006.
B. Killingsworth, R. Barton, and K. Kobia, “Review of Life Cycle 82. ARA, Inc. “Guide for Mechanistic-Empirical Design of New and
Costing Analysis Procedures” (in Ontario), Final Report prepared Rehabilitated Pavement Structures.” Final Report, Part 2 Design
for the Ministry of Transportation of Ontario, Toronto, Ontario, Inputs, Chapter 2: Subgrade/Foundation Design Inputs, NCHRP,
Canada, 1998. Transportation Research Board, Washington, DC, 2004.
68. Bradbury, A., T. Kazmierowski, K. L. Smith, and H. L. Von Quin- 83. ARA, Inc. “Guide for Mechanistic-Empirical Design of New and
tus. “Life Cycle Costing of Freeway Pavements in Ontario.” Paper Rehabilitated Pavement Structures.” Final Report, Part 1 Intro-
presented at the 79th Annual Meeting of the Transportation Re- duction, Chapter 2: Background, Scope, and Overview.
search Board, Washington, DC, 2000. 84. ARA, Inc., “Guide for Mechanistic-Empirical Design of New and
69. Von Quintus, H. L., J. Mallela, and J. Jiang. “Expected Service Life Rehabilitated Pavement Structures.” Final Report, Appendix II-1:
and Performance Characteristics of HMA Pavements in LTPP.” Calibration of Fatigue Cracking Models for Flexible Pavements.
Final Report 5672-2/1, Applied Research Associates, Inc., prepared 85. Newcomb, D. E. “A Guide to PerRoad 2.4 for PerRoad Perpetual
for the Asphalt Pavement Alliance, June 2004. Pavement Software.” Asphalt Pavement Alliance, 2004.
70. Fonseca, O. A. and M. W. Witczak. “A Predictive Methodology 86. Yoder, E. J. and M. W. Witczak. Principles of Pavement Design,
for the Dynamic Modulus of In-Place Aged Asphalt Mixtures.” 2nd Ed., John Wiley & Sons, New York, 1975.
Journal of the Association of Asphalt Paving Technologists, Vol. 65, 87. Tsai, B. W., D. Jones, J. T. Harvey, and C. L. Monismith. “Reflective
1996, pp. 532–567. Cracking Study: First-Level Report on Laboratory Fatigue Testing.”
71. Von Quintus, H. L., J. Mallela, and J. Jiang. “Quantification of the Research Report UCPRC-RR-2006-08, Institute of Transportation
Effects of Polymer-Modified Asphalt for Reducing Pavement Dis- Studies, University of California, Davis, 2008.
 97

APPENDIX A

Proposed Standard Practice for

Predicting the Endurance Limit of Hot Mix
Asphalt (HMA) for Long-Life Pavement Design
AASHTO Designation: PP XX-XX

1. Scope
1.1 This practice describes methodology for predicting the endurance limit for hot mix
asphalt for long-life pavement design.

1.2 This standard may involve hazardous materials, operations, and equipment. This standard
does not purport to address all of the safety problems associated with its use. It is the
responsibility of the user of this procedure to establish appropriate safety and health
practices and to determine the applicability of regulatory limitations prior to its use.

2. Referenced Documents
2.1 AASHTO Standards
• T 321, Determining the Fatigue Life of Compacted Hot-Mix Asphalt (HMA) Sub-
jected to Repeated Flexural Bending.
• R 30, Mixture Conditioning of Hot Mix Asphalt (HMA)

2.2 Other Publications

• NCHRP 9-38, “Endurance Limit of Hot Mix Asphalt Mixtures to Prevent Fatigue
Cracking in Flexible Pavements,” Draft Final Report.

3. Terminology
3.1 Endurance limit – the strain level, at a given temperature, below which no bottom-up
fatigue damage occurs in the HMA.

3.2 Long-Life Pavement Design – a pavement designed to last a minimum of forty years
without bottom-up fatigue failure, or need for structural strengthening.

3.3 Normal strain levels – strain levels where failure (50 percent of initial stiffness) occurs
in less than 12 million cycles. For tests conducted at 20°C, strain levels of 300 micro-
strain or greater generally meet this requirement.
98

3.4 Low Strain levels – strain levels where failure (50 percent of initial stiffness) does not
occur by 12 million cycles. The failure point of low strain tests generally needs to be
extrapolated by one of the methods described in this document.

4. Summary of Practice
4.1 This practice describes the analysis needed to determine the endurance limit for hot-
mix asphalt concrete mixtures. It involves collecting beam fatigue test data at specified
strain rates, predicting the endurance limit based on a log-log extrapolation, and then
running tests at the predicted strain level to confirm endurance limit behavior. Since
the tests conducted at the predicted strain level should not fail, the failure point is
extrapolated from the test data by use of one of several different techniques to confirm
the endurance limit of the asphalt mixture.

5. Significance and Use

5.1 The endurance limit can be used during pavement design to determine a pavement
thickness which will prevent bottom-up fatigue cracking.

6. Apparatus
6.1 Specimen Fabrication Equipment – Equipment for fabricating beam fatigue test speci-
mens as described in AASHTO T 321, Determining the Fatigue Life of Compacted Hot-
Mix Asphalt (HMA) Subjected to Repeated Flexural Bending.

6.2 Beam Fatigue Test System – Equipment for testing beam fatigue samples as described in
AASHTO T 321, Determining the Fatigue Life of Compacted Hot-Mix Asphalt (HMA)
Subjected to Repeated Flexural Bending.

6.3 Analysis Software – Data is collected during the test using a data acquisition system
described in section 6.2. Data analysis can be conducted using a spreadsheet program,
or a variety of statistical packages.

7. Hazards
7.1 This practice and associated standards involve handling of hot asphalt binder, aggregates
and asphalt mixtures. It also includes the use of sawing and coring machinery and servo-
hydraulic or pneumatic testing equipment. Use standard safety precautions, equipment,
and clothing when handling hot materials and operating machinery.

8. Standardization
8.1 Items associated with this practice that require calibration are included in the documents
referenced in Section 2. Refer to the pertinent section of the referenced documents for
information concerning calibration.
 99

9. Beam Fatigue Test Data

9.1 Test Specimen Fabrication
9.1.1 Prepare at least twelve test specimens to the target air void content in accordance
with AASHTO T 321. Samples should be short-term oven aged according to AASHTO
R 30 for four hours at 135°C prior to compaction. The target air void content should
be representative of that expected to be obtained in the field. A target air void content
of 7 percent was used for mixes produced at optimum asphalt content in the NCHRP
9-38 research. A reduced air void content would be expected for optimum plus or
so-called rich-bottom type mixes.
Note 1 – A reasonable air void tolerance for test specimen fabrication is ± 0.5 %.
Note 2 – The following are estimated based on a limited round robin. The coefficient
of variation of the log (base 10) of the fatigue life of a properly conducted beam fatigue
tests at normal strain levels is 5.4 and 6.8 percent, respectively, for within- and between-
lab variability. The difference between the logs of the fatigue lives (log sample 1 – log
sample 2) of two properly conducted test should not exceed 0.69 for a single operator
or 0.89 between two labs.

9.2 Testing Conditions

9.2.1 Separate the test specimens into four groups with approximately equal air voids. Deter-
mine the number of cycles to failure (50 percent of initial stiffness) of each specimen in
one group using the beam fatigue test system at a strain level of 700 micro-strain. Deter-
mine the number of cycles to failure in a second group at a strain level of 500 micro-
strain. Determine the number of cycles to failure in the third group at a strain level of
300 micro-strain. The test results at the normal strain level should not be extrapolated;
all nine sample should be tested to failure.
Note 3 – The original research, including round robin testing, was based on testing
three samples each at 800 and 400 micro strain. Additional samples were added to
improve the estimation of the endurance limit due to the sensitivity of pavement thick-
ness to the endurance limit.

9.3 Beam Fatigue Data Summary to Predict Endurance Limit

9.3.1 Transform the data by taking the log (base 10) of both the micro-strain levels and cycles
to failure. Research has shown that the log-log transformation of the data from tests
conducted at normal strain levels (above the endurance limit) produce a straight line.
Perform a simple linear regression on the transformed data using fatigue life as a pre-
dictor for micro-strain level. Using the regression coefficients determined from the
regression analysis, determine the micro-strain level corresponding to a fatigue life of
log 50,000,000 cycles = 7.69897. Designate this value yo. Calculate the one-sided lower
95% prediction interval according to Equation 1.

1 ( xo − x )
2

Lower Prediction Limit = ŷo − t α s 1 + + (1)

n Sxx

where:
tα = value of t distribution for n − 2 degrees of freedom = 1.89458 for n = 9 with
α = 0.05%,
100

s = estimate of standard deviation from the regression analysis, also referred to in

Microsoft Excel as the standard error,
n = number of samples = 9,
n
Sxx = ∑ ( x i − x ) (Note: log of fatigue lives),
2

i =1
x0 = log 50,000,000 = 7.69897,
_
x = average of the fatigue life results determined in 9.2.1.

Note 4 – A simple spreadsheet has been developed to perform the calculations described
above.

9.3.2 Conduct the beam fatigue test at the strain level corresponding to the 95% one-sided
lower prediction limit for a fatigue life of 50,000,000 cycles to 10,000,000 cycles. Research
has shown that tests conducted to a minimum of 10,000,000 cycles can extrapolate to
estimate long-life fatigue lives.

10. Data Analysis to Extrapolate Long-life

Fatigue Tests
Beam fatigue tests conducted at low strain levels are unlikely to fail in a reasonable
number of cycles. This is particularly true for tests conducted at the 95% one-sided
lower prediction limit for the endurance limit determined in section 9.3.1. In order to
confirm the existence of the endurance limit, the test data needs to be extrapolated to
predict a failure point. This section provides three procedures for extrapolating the
beam fatigue failure point for low-strain tests.

10.1 Data Extrapolation Using the Single-Stage Weibull Survivor Function

10.1.1 General Form. The general form of the Weibull Survivor function is shown as Equa-
tion 2:

⎡ ⎛ t −δ ⎞γ ⎤
R ( t ) = exp ⎢ − ⎜ ⎟ ⎥ (2)
⎣ ⎝ θ−δ⎠ ⎦

where:
R(t) = the reliability at time t where t might be time or another life parameter such as
loading cycles,
γ = the slope,
δ = the minimum life, and
θ = the characteristic life.

10.1.2 Simplified Form. A specialized case of the Weibull function assumes the mini-
mum life, δ, equals zero. Therefore, the hazard function would equal 1/γ, which
simplifies Equation 2 into Equation 3. Since the beam fatigue loading cycles are
applied at a constant frequency of 10 Hz, the loading cycles, n, can be substituted
for time, t.

S (n ) = exp ( − λ × n γ ) (3)
 101

where:
S(t) = probability of survival until time t,
n = number of loading cycles,
λ = scale parameter (intercept),
γ = shape parameter (slope).

10.1.3 Final Form. The stiffness ratio (SR) is used to characterize fatigue damage. The stiff-
ness ratio is the stiffness measured at cycle n, divided by the initial stiffness, determined
at the 50th cycle. Tsai reports that at a given cycle n, the beam being tested has a prob-
ability of survival past cycle n equal to the stiffness ratio times 100 percent. Thus, SR(n)
can be substituted for S(t). Equation 4 allows the scale and shape parameters for labo-
ratory beam fatigue data to be determined by linear regression.

ln ( − ln ( SR )) = ln ( λ ) + γ × ln (n ) (4)

10.1.4 Plot the left-hand side of Equation 4 versus the natural logarithm of the number of
cycles, n so a straight line regression can be determined. If the measured stiffness at a
given number of cycles is greater than the initial stiffness, e.g. SR > 1.0, ln(-ln(SR)) can-
not be computed. Eliminate these data from the regression analysis. Using the shape
(slope) and scale (intercept) parameters determined from the simple linear regression,
solve Equation 4 for n which produces an SR of 0.5. This can be readily done using a
solver function or by trial and error in a spreadsheet. This value of n is the extrapolated
fatigue life for 50 percent initial stiffness.

10.1.5 The endurance limit is confirmed by the divergence of the test data below the regres-
sion line at a high number of cycles. Figure 1 illustrates this divergence.

PG 64-22 at Optimum

0 Upper divergence, indicating endurance limit

-1

-2
Ln(-Ln(Stiffness Ratio))

-3

-4
y = 0.401x -8.214
-5 R² = 0.722
-6

-7

-8

-9

-10
0 2 4 6 8 10 12 14 16 18 20
Ln (Cycles)

Sample 13 Linear (Sample 13)

Figure 1. Weibull functions.

102

10.2 Data Extrapolation Using Three-Stage Weibull Survivor Function

10.2.1 To improve upon the accuracy of the single-stage Weibull function, Tsai et al. (2)
developed a methodology for fitting a three-stage Weibull curve. Tsai et al. (2) theo-
rized that a plot of loading cycles versus stiffness ratio could be divided into three
stages: initial heating and temperature equilibrium, crack initiation, and crack propa-
gation. In the case of low strain tests (below the endurance limit), the third stage does
not appear to represent crack propagation, but rather concave down stage with a reduced
rate of damage.
In most cases, low strain fatigue tests can be most accurately extrapolated using the single-
stage Weibull function. However, in some cases, there are three distinct slopes to the
transformed data, in which case the three-stage Weibull function may be used to pro-
vide a better estimate of the fatigue life.

10.2.2 First, plot the data as described in Section 10.1.4. Visually examine the data to deter-
mine three stages, determined by groups of data exhibiting distinct slopes. Assign a data
series to each group of data, representing a single slope. Perform a linear regression on
the transformed data for each stage. The regression coefficients become seed values for
either a compiled Fortran program or Microsoft Excel spreadsheet solution. An exam-
ple is shown in Figure 2.

0
0 2 4 6 8 10 12 14 16 18 20

y = 0.0583x -2.554
R² = 0.5874
-1

y = 0.1605x -3.949
R² = 0.9193
-2

Series 1
Stage 1
Stage 2
-3 Stage 3
Linear (Stage 1)
y = 0.5158x -6.51 Linear (Stage 2)
R² = 0.6636
Linear (Stage 3)

-4

-5

-6

Figure 2. Estimation of three Weibull stages by inspection.

 103

10.2.3 A Weibull function is fit to each of these stages as shown in Equation 5:

SR 1 = e ( )
− α1 ×nβ1
for 0 ≤ n < n1 (5)

SR 2 = e ( )
− α 2 ×(n− γ 1 )β2
for n1 ≤ n < n2

SR 3 = e ( )
− α3 ×(n− γ 2 )β3
for n2 ≤ n < n3

10.2.4 Coefficients α1, α2, α3, β1, β2, β3, n1, and n2 are illustrated in Figure 3. Using a series of
mathematical manipulations (2), n1, γ1, n2, and γ2 can be calculated sequentially as follows:
1
⎡ α 2 ⎛ β 2 ⎞ β2 ⎤ β2 −β1
n1 = ⎢ × ⎜ ⎟ ⎥ (6)
⎢⎣ α1 ⎝ β1 ⎠ ⎥⎦

⎡ β ⎤
γ 1 = ⎢1 − 2 ⎥ × n1 (7)
⎣ β1 ⎦

1
⎡ α ⎛ β ⎞ β3 ⎤ β 2 − β3
n2 = γ 1 + ⎢ 3 × ⎜ 3 ⎟ ⎥ (8)
⎢⎣ α 2 ⎝ β 2 ⎠ ⎥⎦

PG 76-22 Optimum Plus Sample 9 at 200 ms

0

y = 0.0667x -2.6913
R² = 0.7617
-1
y = 0.1769x -4.1281
R² = 0.9075 β3
1
Ln( -Ln(Stiffness Ratio))

-2

y = 0.5427x -6.6613 β2
R² = 0.547
1
-3 Stage 1
Stage 2
β1 Intercept = Ln α3
Stage 3
-4
1 Intercept = Ln α2

Intercept = Ln α1 LN (n1) Ln (n2)

-5

Stage 1 Stage 2 Stage 3

Initial Heating
-6
0 2 4 6 8 10 12 14 16 18 20
Ln (Loading Cycles)

Figure 3. Three-stage Weibull curve definitions.

104

⎛ β ⎞ β
γ 2 = ⎜ 1 − 3 ⎟ × n2 + 3 × γ 1 (9)
⎝ β2 ⎠ β2

10.2.4 The trial values for α1, α2, α3, β1, β2, β3 and he test data are entered into a FORTRAN
program, N3stage.exe, developed by Tsai or into a spreadsheet developed as part of
NCHRP 9-38. N3stage.exe provides a robust solution, but can take up to 45 minutes
to run. The spreadsheet uses Microsoft Excel’s Solver Function and takes less than a
moment, but in a few cases does not identify a solution. The N3stage.exe program can
be obtained from: BWTsai@Berkeley.edu

10.2.5 Using the results from the NStage3.exe program or NCHRP 9-38 spreadsheet, the pre-
dicted fatigue life can be calculated according to Equation 10.

⎛ ⎛ α3 ⎞ ⎞
⎜ ( ln( − ln( 0.5))− ln⎜⎝ ⎟⎟
Nf = e ⎝ β3 ⎠ ⎠ + γ2 (10)

11. Report
11.1 For each sample, report the following:
11.1.1 Sample air voids

11.1.2 Test Temperature

11.1.3 Initial flexural stiffness (measured at 50 cycles)

11.1.4 Normal Strain Tests

11.1.4.1 Micro-strain level
11.1.4.2 Number of cycles (measured) to 50 percent of initial stiffness

11.1.5 Low Strain Tests

11.1.5.1 Method of Extrapolation
11.1.5.2 Equation used for extrapolation and R2 value for equation
11.1.5.3 Extrapolated fatigue life Nf for 50 percent of initial stiffness

11.1.6 Endurance Limit

11.1.6.1 Log-Log plot of data with cycles to failure on x-axis and micro-strain on y-axis.
11.1.6.2 Best-fit regression line/equation of normal-strain data should be included on graph.
11.1.6.3 Predicted endurance limit from normal-strain regression with Nf = 50,000,000 cycles
11.1.6.4 Lower one-sided 95 percent confidence interval for micro-strain level corresponding
to Nf = 50,000,000 cycles, termed predicted endurance limit.
 105

12. Keywords
12.1 Beam fatigue, endurance limit, long-life pavement

13. References
13.1 Tsai, B-W, “High Temperature Fatigue and Fatigue Damage Process of Aggregate-
Asphalt Mixes.” Report to California Department of Transportation, Pavement
Research Center, Institute of Transportation Studies, University of California at
Berkeley.

13.2 Tsai, B.-W., J. T. Harvey, and C. L. Monismith. “Using the Three-Stage Weibull Equa-
tion and Tree-Based Model to Characterize the Mix Fatigue Damage Process.” In
Transportation Research Record No. 1929, Transportation Research Board, Washing-
ton, DC, 2005, Pp 227-237.
106

APPENDIX B

Proposed Standard Practice for

Predicting the Endurance Limit of Hot Mix
Asphalt (HMA) by Pseudo Strain Approach
AASHTO Designation: PP XX-XX

1. Scope
1.1 This practice describes methodology for predicting the endurance limit for hot mix
asphalt for long-life pavement design by pseudo strain approach.

1.2 This standard may involve hazardous materials, operations, and equipment. This standard
does not purport to address all of the safety problems associated with its use. It is the
responsibility of the user of this procedure to establish appropriate safety and health practices
and to determine the applicability of regulatory limitations prior to its use.

2. Referenced Documents
2.1 AASHTO Standards
• PP XX-XX, Preparation of Cylindrical Performance Test Specimens Using the Super-
pave Gyratory Compactor
• PP XX-XX, Determining the Dynamic Modulus and Flow Number for Hot Mix
Asphalt (HMA) Using the Simple Performance Test System
• PP XX-XX, Developing Dynamic Modulus Master Curves for Hot-Mix Asphalt
Concrete Using the Simple Performance Test System

2.2 Other Publications

• Equipment Specification for the Simple Performance Test System, Version 2.0,
Prepared for National Cooperative Highway Research Program (NCHRP), March
26, 2004.

3. Terminology
3.1 Dynamic Modulus – |Eⴱ|, the absolute value of the complex modulus calculated by
dividing the peak-to-peak stress by the peak-to-peak strain for a material subjected to
a sinusoidal loading.

3.2 Phase Angle – δ, the angle in degrees between a sinusoidally applied stress and the
resulting strain in a controlled-stress test.
 107

3.3 Endurance limit – the strain level, at a given temperature, below which no fatigue dam-
age occurs in the HMA.

4. Summary of Practice
4.1 This practice describes the analysis needed to determine the endurance limit for hot-
mix asphalt concrete mixtures by pseudo strain approach. It involves testing
continuous cyclic fatigue test data of cylindrical asphalt concrete.

5. Significance and Use

5.1 The endurance limit can be used during pavement design to determine a pavement
thickness which will prevent fatigue cracking.

6. Apparatus
Specimen Fabrication Equipment – Equipment for fabricating cylindrical test specimens
as described in AASHTO PP XX-XX, Preparation of Cylindrical Performance Test
Specimens Using the Superpave Gyratory Compactor.

6.1 Specimen Fabrication Equipment – Equipment for fabricating cylindrical test speci-
mens as described in PP XX-XX, Preparation of Cylindrical Performance Test Speci-
mens Using the Superpave Gyratory Compactor.

6.2 Dynamic Modulus Test System – A dynamic test system meeting the requirements of
Equipment Specification for the Simple Performance Test System, Version 2.0

6.3 Conditioning Chamber – An environmental chamber for conditioning the test spec-
imens to the desired testing temperature. The environmental chamber shall be capa-
ble of controlling the temperature of the specimen over a temperature range from
4 to 60°C (39 to 140°F) to an accuracy of ± 0.5°C (1°F). The chamber shall be large
enough to accommodate the number of specimens to be tested plus a dummy spec-
imen with a temperature sensor mounted at the center for temperature verification.

6.4 Analysis Software – Software capable of handling numerical approaches like numeri-
cal integration and nonlinear optimization. Data analysis can be conducted using a
spreadsheet program, or a variety of scientific computation packages like MATLAB,
Lab View.

7. Hazards
7.1 This practice and associated standards involve handling of hot asphalt binder,
aggregates and asphalt mixtures. It also includes the use of sawing and coring
machinery and servo-hydraulic or pneumatic testing equipment. Use standard safety
precautions, equipment, and clothing when handling hot materials and operating
machinery.
108

8. Standardization
8.1 Items associated with this practice that require calibration are included in the
documents referenced in Section 2. Refer to the pertinent section of the referenced
documents for information concerning calibration.

9. Test Specimen
9.1 Compaction – Prepare at least three test specimens to the target air void content and
aging condition in accordance with AASHTO PP XX-XX, Preparation of Cylindri-
cal Performance Test Specimens Using the Superpave Gyratory Compactor (SGC).
The target air void content should be representative of that expected to be obtained
in the field.

Note 1 – A reasonable air void tolerance for test specimen fabrication is ± 0.5 %.

Note 2 – The coefficient of variation for properly conducted fatigue tests has not yet
been determined

9.2 Preparation – Core the cylinders from gyratory compacted specimens and cut the ends
of cylinders thus obtained to ensure uniform air void content. Check the cylinder for
its geometry and axis alignment.

Note 3 – A reasonable diameter tolerance for test specimen is ± 1 mm.

9.3 Gluing – Clean the cored sample using compressed air and modified alcohol so that
the dust particles over the surface are totally removed. Apply glue on the ends of the
cleaned specimen. Place end plate over the specimen end and press against one
another by hand. Repeat same procedure for another end also. Mount the specimen
along with end plates on gluing jig and make adjustments such that cylinder axis is
aligned with that of end plates. Place dead weight on gluing jig so that endplates are
pressed against specimen. Remove the excess glue (if any) and leave the specimen for
sufficient curing.

Note 4 – Process of gluing end plates to test specimen shall be completed well within
initial setting time of glue.

10. Testing
10.1 Dynamic Modulus
10.1.1 Conduct dynamic modulus test in tension over a range of temperature and frequen-
cies using three specimens in accordance with AASHTO PP XX-XX, Determining the
Dynamic Modulus and Flow Number for Hot Mix Asphalt (HMA) Using the Simple
Performance Test System.

Note 5 – Applied loads should be within linear elastic limits of asphalt concrete (this
can be guaranteed when observed strain is less than 75 microns).

Note 6 – As flow number is not used in test practice described in this document, flow
number part of AASHTO PP XX-XX guidelines should be ignored.
 109

(a) Load History (b) Strain History

Figure 1. Typical stress/strain history for increasing amplitude uniaxial fatigue test.

10.2 Increasing Strain Amplitude Test

10.2.1 Conduct increasing strain amplitude fatigue test on other three specimens until fail-
ure starting from low strain amplitude at constant temperature (usually 20 C). A good
starting strain amplitude value is that at which dynamic modulus tests are conducted.
Around 10,000 cycles are applied at this amplitude to get steady state response.
Applied strain amplitude is then increased and 10,000 more cycles are applied. This
process shall be repeated at increasing strain amplitude level successively until speci-
men fails.

Note 7 – Mounting of LVDT’s and load cells shall be similar to that of dynamic mod-
ulus determination (refer 10.1).

Note 8 – Temperature at which fatigue test is conducted shall be henceforth referred as

reference temperature.

Typical plots of load and strain histories for an increasing amplitude fatigue test are
given in Figure 1a and 1b respectively.

11. Calculations
11.1 Dynamic Modulus Mastercurve Construction
11.1.1 Using data previously obtained dynamic modulus data (refer 10.1), construct dynamic
modulus master curves at reference temperature for individual specimens. General
form of the dynamic modulus master curve is given in Equation 1.

log E ⴱ = δ +
( Max − δ ) (1)
⎧ ΔEa ⎡⎛ 1 ⎞ ⎛ 1 ⎞ ⎤ ⎫
β+ γ ⎨ log ω + ⎜ ⎟ −⎜ ⎟ ⎬
1+ e ⎩ 19.14714 ⎣⎢⎝ T ⎠ ⎝ Tr ⎠ ⎦⎥ ⎭
110

where:
|Eⴱ| = dynamic modulus, MPa
ωr = reduced frequency, Hz
Max = limiting maximum modulus, MPa
Tr = reference temperature, °K
T = test temperature, °K
ΔEa = activation energy (treated as a fitting parameter)
δ, β, and γ = fitting parameters.

11.2 Relaxation Modulus Prediction

11.2.1 Using dynamic modulus (|Eⴱ|) and phase angle (ωr) data (refer 10.1), obtain relaxation
modulus for each specimen using following relations.

E ′ ( ω r ) = E ⴱ( ω r ) cos ( φ ( ω r ))
1 0..08
E (t r ) = E ′ (ω r ), ω r =
λ′ tr
⎛ nπ ⎞
λ ′ = Γ (1 − n ) cos ⎜ ⎟
⎝ 2⎠
d log E ′ ( ω )
n=
d log ω

where:
|E′(ωr)| = storage modulus,
tr = reduced time,
Γ = gamma function,
n = slope of log(E′(ω)) versus log(ω) obtained at each point of reduced frequency.

11.2.2 Fit Prony series for relaxation modulus values obtained previously (refer 11.2.1). Gen-
eral expression for relaxation modulus as Prony series is given by Equation 2.

⎛ t ⎞
E ( t ) = E∞ + ∑ En exp ⎜ − ⎟ (2)
n ⎝ ρn ⎠

where:
E∞ = Relaxation modulus as t→∞,
En = Prony series coefficients,
ρn = Relaxation time.

11.3 Calculation of Pseudo Strain

11.3.1 Using time history of strain (using data obtained in 10.2) and predicted relaxation
modulus (refer 11.2), calculate pseudo strains for entire strain history for each speci-
men. Equation 2 shall be used for calculating pseudostrains at reference temperature.

1 t dε
ε R (t ) = E ( t − τ ) dτ
E R ∫0
(3)
dτ
 111

Figure 2. Cross plot of Stress vs. Pseudo strain.

where:
ER = reference modulus (usually chosen as unity), MPa
E(t) = relaxation modulus, MPa
ε = measured strain

11.3.2 Crossplot stress vs. pseudo strain and check for formation of hysteretic loop. This shall
be started from lowest level of strain amplitude. Note down the strain level at which
hysteresis loop appears for first time. Also calculate average value of strain level at
which hysteresis loop forms. Typical plot comparing stress vs. pseudo strain before and
after damage is presented in Figure 2.
Note 6 – The loop is observed in cross plot of stress vs. pseudo strain if the damage has
occurred.

12. Report
12.1 For each specimen, report the following
12.1.1 Sample air voids
12.2 For each mix, report the following
12.2.1 Reference Temperature
12.2.2 Dynamic modulus master curve coefficients
12.2.3 Strain level at which hysteresis loop appears for first time in cross plot of stress vs.
pseudo strain
12.2.4 Average strain level for loop formation

13. Keywords
13.1 Fatigue, Viscoelastic Continuum damage, Endurance limit
112

APPENDIX C

Proposed Standard Practice for

Extrapolating Long-Life Beam Fatigue Tests Using
the Ratio of Dissipated Energy Change (RDEC)
AASHTO Designation: PP XX-XX

1. Scope
1.1 This practice describes methodology for extrapolating long-life Beam Fatigue Tests
Using the RDEC

1.2 This standard may involve hazardous materials, operations, and equipment. This standard
does not purport to address all of the safety problems associated with its use. It is the
responsibility of the user of this procedure to establish appropriate safety and health practices
and to determine the applicability of regulatory limitations prior to its use.

2. Referenced Documents
2.1 AASHTO Standards
• T 321, Determining the Fatigue Life of Compacted Hot-Mix Asphalt (HMA) Sub-
jected to Repeated Flexural Bending.

2.2 Other Publications

• NCHRP 9-38, “Endurance Limit of Hot Mix Asphalt Mixtures to Prevent Fatigue
Cracking in Flexible Pavements,” Draft Final Report.

3. Terminology
3.1 Normal strain levels – strain levels where failure (50 percent of initial stiffness) occurs
in less than 12 million cycles. For tests conducted at 20 °C, strain levels of 300 micro-
strain or greater generally meet this requirement.

3.2 Low Strain levels – strain levels where failure (50 percent of initial stiffness) does not
occur by 12 million cycles. The failure point of low strain tests generally needs to be
extrapolated by one of the methods described in this document.
 113

4. Summary of Practice
4.1 This practice describes the analysis needed to extrapolate the failure point of long-life
beam fatigue tests that are not tested to failure (50 percent reduction in initial stiffness).

5. Significance and Use

5.1 The extrapolation procedure can be used to estimate the failure point of fatigue tests
which do not fail in a reasonable amount of loading cycles (<12,000,000).

6. Apparatus
6.1 Specimen Fabrication Equipment – Equipment for fabricating beam fatigue test speci-
mens as described in AASHTO T 321, Determining the Fatigue Life of Compacted Hot-
Mix Asphalt (HMA) Subjected to Repeated Flexural Bending.

6.2 Beam Fatigue Test System – Equipment for testing beam fatigue samples as described
in AASHTO T 321, Determining the Fatigue Life of Compacted Hot-Mix Asphalt
(HMA) Subjected to Repeated Flexural Bending.

6.3 Analysis Software – Data is collected during the test using a data acquisition system
described in section 6.2. Data analysis can be conducted using a spreadsheet program
or variety of statistical packages.

7. Hazards
7.1 This practice and associated standards involve handling of hot asphalt binder,
aggregates and asphalt mixtures. It also includes the use of sawing and coring machinery
and servo-hydraulic or pneumatic testing equipment. Use standard safety precautions,
equipment, and clothing when handling hot materials and operating machinery.

8. Standardization
8.1 Items associated with this practice that require calibration are included in the documents
referenced in Section 2. Refer to the pertinent section of the referenced documents for
information concerning calibration.

9. Beam Fatigue Test Data

9.1 Test Specimen Fabrication
9.1.1 Prepare test specimens to the target air void content and aging condition in accordance
with AASHTO T 321. The target air void content should be representative of that
expected to be obtained in the field. A target air void content of 7 percent was used for
mixes produced at optimum asphalt content in the NCHRP 9-38 research. A reduced air
void content would be expected for optimum plus or so-called rich-bottom type mixes.
Note 1 – A reasonable air void tolerance for test specimen fabrication is ± 0.5 %.
114

9.2 Testing Conditions

9.2.1 Samples tested to a minimum of 10 million cycles, which have not reached 50 percent
of initial stiffness, may be extrapolated to determine a failure point as described in the
following section.

10. Data Analysis to Extrapolate Long-life Fatigue Test

Using RDEC
Beam fatigue tests conducted at low strain levels are unlikely to fail in a reasonable
number of cycles.

10.1 Ratio of Dissipated Energy Change (RDEC)

10.1.1 PV calculation for normal strain testing
10.1.1.1 Determine the number of loading cycles, Nf , to failure from testing.
10.1.1.2 Obtain a dissipated energy (kPa) vs. loading cycle (DE-LC) relationship. Obtain a best fit
equation for the DE-LC data, using a power law relationship. From the best fit equation,
record the slope, f, of the curve, that can best represent the original curve. Figure 1 illus-
trates a DE-LC fitted curve.
10.1.1.3 Calculate RDEC. By definition, RDEC is the ratio of dissipated energy change between
two loading cycles by the number between the cycles, that is, the average ratio of dissi-
pated energy change per loading cycles, as seen in Equation 1.

DE a − DEb
RDEC a = (1)
DE a ⴱ (b − a )

where,
RDECa = the average ratio of dissipated energy change at cycle a, comparing to next
cycle b;
a, b = load cycle a and b, respectively. The typical cycle count between cycle a and
b for RDEC calculation is 100, i.e., b − a = 100;
DEa, DEb = the dissipated energy (kPa) produced in load cycle a, and b, respectively.

IDOT03 mix 3N704B

DE vs. Loading cycles 800 microstrain
1.4
DE (kPa)

1.2
1
0.8
Slope f = - 0.1638
0.6
y = 3.4255x-0.1638
0.4 R2 = 0.9512
0.2
Nf50
0
0 1000 2000 3000 4000 5000
Loading cycles

Figure 1. DE vs. LC chart with fitted curve.

 115

10.1.1.4 The average RDEC for an arbitrary 100 cycles at cycle ‘a’ can be simply calculated using
Equation 2.
f
⎛ 100 ⎞
1− ⎜1+ ⎟
⎝ a ⎠
RDEC a = (2)
100

where,
f = the slope from the regressed DE-LC curve

10.1.1.5 Calculate the plateau value (PV). The PV is defined as the RDEC value at the number
of cycles equal to the failure point (Nf50). Failure is defined as a 50 percent reduction in
initial stiffness, with the initial stiffness being determined at the 50th loading cycle. The
PV is determined using Equation 3.
f
⎛ 100 ⎞
1− ⎜1+
⎝ Nf50 ⎟⎠
PV = (3)
100

where,
f = the slope from the regressed DE-LC curve (kPa/cycle)
Nf50 = 50% stiffness reduction failure point

10.1.2 PV calculation for low strain testing

10.1.2.1 Plot the DE-LC data and fit the DE-LC curve using the power law relationship. Obtain
the f factor of the curve.

10.1.2.2 To achieve the best curve fit, it is recommended to eliminate the initial segment of the
DE-LC curve, but use the later part to ensure the fitted curve visually best represents
the curve’s outspread trend. The fitted segment should not be less than 1⁄4 of the total
testing length to avoid being misleading.

10.1.2.3 Calculate RDEC at each loading cycle using Equation 2, where f is given by the fitted
DE-LC curve. Plot the RDEC-LC curve (log-log).

10.1.2.4 Plot the unique PV-Nf curve as shown by Equation 4 on the same chart

PV = 0.4428 Nf50−1.1102 (4)

10.1.2.5 Extend the RDEC-LC curve until it crosses the unique PV-Nf curve. The intersection
point of these two curves produces: y = PV, x = Nf50

10.1.2.6 For |f| < 0.25 (which is the case for most fitted DE-LC curves from fatigue testing),
calculate PV and Nf50 using equations 5 and 6, respectively.
Note: Figure 2 illustrates the fatigue life prediction using the RDEC approach at low
strain testing.
f
⎛ 100 ⎞
1− ⎜1+
⎝ Nf50 ⎟⎠ f
PV = ≈− (5)
100 Nf50
116

0.01

0.001

0.0001
800 microstrain
1E-05
500 microstrain
1E-06

RDEC, log
300 microstrain
1E-07 Nf
PV-Nf Unique Line
Nf
1E-08
Nf
1E-09
FEL Tests 3N904A(Nf, PV)
300 microstrain
1E-10 5N90P2A(Nf,PV)
100 microstrain
1E-11
Nf Nf
1E-12
1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08 1.E+09 1.E+10 1.E+11 1.E+12
loading cycles, log

5N90P3B-800ms 5N90P4A-500ms 5N90P4B-300ms

5N90P2A-100ms 3N904A-300 ms Power (5N90P2A-100ms)
Power (3N904A-300 ms)

Figure 2. Fatigue life prediction using RDEC approach.

−9.0744
⎛ −f ⎞
Nf50 = ⎜
⎝ 0.4428 ⎟⎠
(6)

11. Report
11.1 For each sample, report the following:
11.1.1 Sample air voids
11.1.2 Test Temperature
11.1.3 Initial flexural stiffness (measured at 50 cycles)
11.1.4 Method of Extrapolation
11.1.5.1 Equation used for extrapolation and R2 value for equation
11.1.5.2 Extrapolated fatigue life Nf for 50 percent of initial stiffness

12. Keywords
12.1 Beam fatigue, long-life

13. References
13.1 Shen, S. and S. H. Carpenter. “Application of Dissipated Energy Concept in Fatigue
Endurance Limit Testing” In Transportation Research Record 1929, Transportation
Research Board, Washington, DC, 2005, Pp 165-173.
 117

APPENDIX D

NCHRP 9-38 Beam Fatigue

Fatigue Life

LAB MATERIAL
A (PG 67-22 Opt. 800 ms)
1 2 3 Avg. Std. Dev. Variance h k
1 3.13 3.66 3.72 3.50 0.324 0.105 -1.69 1.30
2 3.82 3.68 3.97 3.82 0.146 0.021 -0.25 0.58
3 3.89 3.99 3.42 3.77 0.305 0.093 -0.50 1.22
4 4.26 3.93 4.26 4.15 0.191 0.037 1.24 0.77
5 4.09 3.95 4.06 4.03 0.073 0.005 0.72 0.29
6 3.78 3.85 3.78 3.80 0.043 0.002 -0.33 0.17
7 4.39 4.18 3.59 4.06 0.415 0.172 0.82 1.66

Average of all Labs 3.876

Std. Dev. Between Cell Averages (Sx) 0.220
Repeatability Standard Deviation (Sr) 0.249
Reproducibility Standard Deviation (SR) 0.300
Between Lab Standard Deviation of Lab Means(SL) 0.167
Pooled within lab variance (SA2) 0.062 (Components of variance) W/L
2
Between Lab Variance of Lab Means(SL ) 0.028 (Components of variance) B/L
W/L Variance 0.062
B/L Variance 0.090
h critical = 2.05 k critical = 2.03

LAB MATERIAL
B (PG 67-22 Opt. 400 ms)
1 2 3 Avg. Std. Dev. Variance h k
1 5.49 5.60 5.43 5.51 0.087 0.008 0.38 0.36
2 4.68 4.94 5.06 4.89 0.196 0.038 -1.30 0.81
3 5.50 5.64 5.58 5.58 0.070 0.005 0.56 0.29
4 5.99 5.76 6.10 5.95 0.174 0.030 1.60 0.72
5 5.35 4.59 4.95 4.97 0.381 0.145 -1.11 1.59
6 5.39 5.43 5.38 5.40 0.023 0.001 0.09 0.10
7 5.57 4.81 5.49 5.29 0.421 0.177 -0.22 1.75

Average of all Labs 5.370

Std. Dev. Between Cell Averages (Sx) 0.365
Repeatability Standard Deviation (Sr) 0.240
Reproducibility Standard Deviation (SR) 0.414
Between Lab Standard Deviation of Lab Means(SL) 0.338
Pooled within lab variance (SA2) 0.058
2
Between Lab Variance of Lab Means(SL ) 0.114
W/L Variance 0.058
B/L Variance 0.172
h critical = 2.05 k critical = 2.03
118

LAB MATERIAL
C (PG 67-22 120 ms Logarithmic)
1 2 3 Avg. Std. Dev. Variance h k
1 9.92 7.60 7.24 8.25 1.453 2.110 -1.44 1.10
2 21.92 10.55 21.62 1.64 2.07
3 10.13 10.80 11.63 10.85 0.751 0.564 0.15 0.57
4 12.88 10.90 12.25 12.01 1.012 1.024 0.86 0.77
5 9.52 13.15 11.29 11.32 1.814 3.290 0.43 1.37

Average of all Labs 10.609

Std. Dev. Between Cell Averages (Sx) 1.642
Repeatability Standard Deviation (Sr) 1.322
Reproducibility Standard Deviation (SR) 1.965
Between Lab Standard Deviation of Lab Means(SL) 1.454
Pooled within lab variance (SA2) 1.747
2
Between Lab Variance of Lab Means(SL ) 2.114
W/L Variance 1.747
B/L Variance 3.861
h critical = 1.74 k critical = 1.92

Lab 2's data excluded as outlier

LAB MATERIAL
C (PG 67-22 120 ms Weibull )
1 2 3 Avg. Std. Dev. Variance h k
1 8.04 7.68 7.34 7.69 0.350 0.123 -1.21 0.60
2 10.49 9.31 10.22 10.01 0.620 0.385 1.33 1.06
3 7.84 8.25 8.40 8.16 0.292 0.085 -0.69 0.50
4 9.36 8.63 9.91 9.30 0.640 0.409 0.55 1.09
5 7.83 9.26 9.33 8.81 0.844 0.713 0.02 1.44

Average of all Labs 8.794

Std. Dev. Between Cell Averages (Sx) 0.915
Repeatability Standard Deviation (Sr) 0.586
Reproducibility Standard Deviation (SR) 1.032
Between Lab Standard Deviation of Lab Means(SL) 0.850
Pooled within lab variance (SA2) 0.343
2
Between Lab Variance of Lab Means(SL ) 0.723
W/L Variance 0.343
B/L Variance 1.066
h critical = 1.74 k critical = 1.92
 119

LAB MATERIAL
C (PG 67-22 120 ms RDEC)
1 2 3 Avg. Std. Dev. Variance h k
1 9.76 3.82 2.18 5.26 3.988 15.908 -1.47 1.70
2 13.73 9.81 13.43 12.32 2.177 4.740 0.60 0.93
3 11.57 13.29 12.85 12.57 0.892 0.795 0.67 0.38
4
5 10.65 11.74 10.47 10.95 0.690 0.476 0.20 0.29

Average of all Labs 10.275

Std. Dev. Between Cell Averages (Sx) 3.421
Repeatability Standard Deviation (Sr) 2.341
Reproducibility Standard Deviation (SR) 3.919
Between Lab Standard Deviation of Lab Means(SL) 3.143
Pooled within lab variance (SA2) 5.480
2
Between Lab Variance of Lab Means(SL ) 9.875
W/L Variance 5.480
B/L Variance 15.355
h critical = 1.49 k critical = 1.82

RDEC could not be calculated with dissipated energy data provided by Lab 4

LAB MATERIAL
A (PG 76-22 800 ms)
1 2 3 Avg. Std. Dev. Variance h k
1 3.66 3.77 4.30 3.91 0.343 0.117 -0.17 1.31
2 4.30 3.87 4.04 4.07 0.218 0.048 1.07 0.84
3 3.87 3.60 3.98 3.82 0.197 0.039 -0.90 0.76

Average of all Labs 3.932

Std. Dev. Between Cell Averages (Sx) 0.127
Repeatability Standard Deviation (Sr) 0.261
Reproducibility Standard Deviation (SR) 0.261
Between Lab Standard Deviation of Lab Means(SL) 0.000
Pooled within lab variance (SA2) 0.068 (Components of variance) W/L
2
Between Lab Variance of Lab Means(SL ) 0.000 (Components of variance) B/L
W/L Variance 0.068
B/L Variance 0.068
h critical = 1.15 k critical = 1.67
120

LAB MATERIAL
B (PG 76-22 400 ms)
1 2 3 Avg. Std. Dev. Variance h k
1 5.88 5.50 5.85 5.74 0.207 0.043 0.40 0.70
2 5.57 4.91 5.37 5.28 0.336 0.113 -1.14 1.14
3 5.51 6.16 5.86 5.84 0.324 0.105 0.74 1.10

Average of all Labs 5.624

Std. Dev. Between Cell Averages (Sx) 0.299
Repeatability Standard Deviation (Sr) 0.295
Reproducibility Standard Deviation (SR) 0.384
Between Lab Standard Deviation of Lab Means(SL) 0.246
Pooled within lab variance (SA2) 0.087 (Components of variance) W/L
2
Between Lab Variance of Lab Means(SL ) 0.060 (Components of variance) B/L
W/L Variance 0.087
B/L Variance 0.147
h critical = 1.15 k critical = 1.67

LAB MATERIAL
C (PG 67-22 220 ms Logarithmic)
1 2 3 Avg. Std. Dev. Variance h k
1 7.64 9.29 9.00 8.64 0.883 0.779 0.98 0.73
2 9.39 7.06 8.22 1.644 2.702 0.03 1.36
3 6.67 8.23 8.38 7.76 0.944 0.891 -1.02 0.78

Average of all Labs 8.209

Std. Dev. Between Cell Averages (Sx) 0.441
Repeatability Standard Deviation (Sr) 1.207
Reproducibility Standard Deviation (SR) 1.207
Between Lab Standard Deviation of Lab Means(SL) 0.000
Pooled within lab variance (SA2) 1.458 (Components of variance) W/L
2
Between Lab Variance of Lab Means(SL ) 0.000 (Components of variance) B/L
W/L Variance 1.458
B/L Variance 1.458
h critical = 1.15 k critical = 1.67

LAB MATERIAL
C (PG 67-22 220 ms Weibull)
1 2 3 Avg. Std. Dev. Variance h k
1 7.42 7.49 7.45 7.45 0.034 0.001 -0.36 0.05
2 8.76 7.06 7.91 1.204 1.451 1.13 1.68
3 6.99 7.37 7.61 7.32 0.314 0.099 -0.77 0.44

Average of all Labs 7.563

Std. Dev. Between Cell Averages (Sx) 0.309
Repeatability Standard Deviation (Sr) 0.719
Reproducibility Standard Deviation (SR) 0.719
Between Lab Standard Deviation of Lab Means(SL) 0.000
Pooled within lab variance (SA2) 0.517 (Components of variance) W/L
2
Between Lab Variance of Lab Means(SL ) 0.000 (Components of variance) B/L
W/L Variance 0.517
B/L Variance 0.517
h critical = 1.15 k critical = 1.67
 121

LAB MATERIAL
C (PG 67-22 220 ms RDEC)
1 2 3 Avg. Std. Dev. Variance h k
1 6.86 11.59 11.29 9.91 2.647 7.007 0.32 1.23
2 9.99 7.06 8.53 2.071 4.290 -1.12 0.96
3 8.54 11.47 11.14 10.39 1.606 2.578 0.80 0.75

Average of all Labs 9.607

Std. Dev. Between Cell Averages (Sx) 0.967
Repeatability Standard Deviation (Sr) 2.151
Reproducibility Standard Deviation (SR) 2.151
Between Lab Standard Deviation of Lab Means(SL) 0.000
Pooled within lab variance (SA2) 4.625 (Components of variance) W/L
2
Between Lab Variance of Lab Means(SL ) 0.000 (Components of variance) B/L
W/L Variance 4.625
B/L Variance 4.625
h critical = 1.15 k critical = 1.67

LAB MATERIAL
A (PG 67-22 Opt.+ 800 ms)
1 2 3 Avg. Std. Dev. Variance h k
1 3.88 4.42 4.12 4.14 0.268 0.072 -0.57 1.30
2 4.31 4.55 4.12 4.33 0.216 0.047 1.15 1.05
3 4.16 4.04 4.23 4.14 0.097 0.009 -0.58 0.47

Average of all Labs 4.203

Std. Dev. Between Cell Averages (Sx) 0.107
Repeatability Standard Deviation (Sr) 0.207
Reproducibility Standard Deviation (SR) 0.207
Between Lab Standard Deviation of Lab Means(SL) 0.000
Pooled within lab variance (SA2) 0.043 (Components of variance) W/L
2
Between Lab Variance of Lab Means(SL ) 0.000 (Components of variance) B/L
W/L Variance 0.043
B/L Variance 0.043
h critical = 1.15 k critical = 1.67

LAB MATERIAL
B (PG 67-22 Opt.+ 400 ms)
1 2 3 Avg. Std. Dev. Variance h k
1 5.60 5.97 6.04 5.87 0.239 0.057 1.14 0.98
2 5.94 5.34 5.59 5.62 0.303 0.092 -0.72 1.25
3 5.75 5.47 5.77 5.66 0.167 0.028 -0.42 0.69

Average of all Labs 5.717

Std. Dev. Between Cell Averages (Sx) 0.134
Repeatability Standard Deviation (Sr) 0.243
Reproducibility Standard Deviation (SR) 0.243
Between Lab Standard Deviation of Lab Means(SL) 0.000
Pooled within lab variance (SA2) 0.059
2
Between Lab Variance of Lab Means(SL ) 0.000
W/L Variance 0.059
B/L Variance 0.059
h critical = 1.15 k critical = 1.67
122

APPENDIX E

Construction of Characteristic Curve

Determination of Once the relaxation modulus is obtained from the storage

the Relaxation Modulus modulus, the Prony series coefficients in Equation 1 are
obtained through the following steps.
Construction of the damage characteristic curve requires First, Equation 1 is rewritten in matrix form as:
the calculation of pseudo strains, which in turn require the
relaxation modulus, E(t), of the mixture. The relaxation { A} = [ B ]{C}
modulus is difficult to measure directly in the laboratory and
is therefore determined from the dynamic modulus master A j = E ( t j ) − E∞
curve using linear viscoelastic theory. The relaxation modulus ⎛ tj ⎞
is expressed as the Prony series: B jk = exp ⎜ − ⎟ (4)
⎝ ρk ⎠
⎛ t ⎞ Cj = Ej
E ( t ) = E∞ + ∑ En exp ⎜ − ⎟ (1)
⎝ ρn ⎠
n
j = 1, 2, . . . , N
Where E∞ is the value of E(t) as t→∞, En are Prony series co- Where the relaxation times tj are chosen at decade intervals
efficients and ρn are relaxation times. To obtain the Prony along time axis and N is the number of data points used. The
series expression of relaxation modulus from dynamic mod- MATLAB optimization toolbox is used to obtain the solution
ulus, the first step is to determine the storage modulus master for Ej with the following constraint:
curve using:
1
E ′ ( ω r ) = E ∗ ( ω r ) cos ( ϕ ( ω r )) (2)
min
C 2
[ B ]{C} − { A} 22 , such that, {C} ≥ 0 (5)

Where |E′(ωr)| is the storage modulus, ϕ is the phase angle which provides positive values of Ej. The prony series repre-
and ωr is the reduced frequency in rad/sec. The relaxa- sentations of the relaxation modulus for the four mixtures are
tion modulus curve of each specimen is obtained from the shown in Figure E1.
storage modulus master curve by applying the following
relation: Monotonic Characteristic Curves
1 0.08 Pseudo Strain Calculation
E (t r ) = E ′ (ω r ), ω r =
λ′ tr Monotonic tests (constant crosshead) at 20°C are per-
⎛ nπ ⎞ formed on specimens at various crosshead strain rates. The
λ ′ = Γ (1 − n ) cos ⎜ ⎟ (3) pseudo strains are calculated using the strains measured from
⎝ 2⎠
the on-specimen LVDTs. Figure E2 shows the typical stress,
d log E ′ ( ω ) crosshead strain, and LVDT strain as a function of time for a
n=
d log ω monotonic test.
The pseudo strain is defined as:
Where tr is the reduced time, Γ is the gamma function and
n is the slope of log(E′(ω)) versus log(ω) curve which is 1 t dε
ε R (t ) = ∫ E ( t − τ ) dτ (6)
obtained at each point of reduced frequency. ER 0 dτ
 123

1.0E+05
67-22, Opt

Relaxation Modulus, MPa

76-22, Opt
67-22, Opt+
1.0E+03
76-22, Opt+

1.0E+01

1.0E-01
1.0E-09 1.0E-05 1.0E-01 1.0E+03 1.0E+07

Time, sec

Figure E1. Relaxation modulus curves for all four mixtures.

where ER is the reference modulus (which is chosen as unity), where σ(t) is the stress history. The damage parameter, S, is
E(t) is the relaxation modulus obtained from storage modulus obtained from the following equation:
and expressed as Prony series and ε is the on-specimen LVDT α
strain observed under monotonic tests. The above integration ⎛1 ⎞ 1+α 1
S1( t ) = ∑ ⎜ ( ε R , j ) (C j −1 − C j )⎟ ( t j − t j −1 )1+α
2
(9)
j ⎝ 2 ⎠
is evaluated numerically over the strain range up to the time
of failure. The strain history is discritized into N number of
small segments with time increment Δt and Equation 1 is 1
where I is the initial pseudo stiffness and α = 1 + , where m
substituted in Equation 6, resulting in the following form of m
numerical integration scheme: is the slope of the linear portion of the relaxation curve. The
damage characteristic curve is obtained by plotting the dam-
N +1
age parameter versus the pseudo stiffness.
ε R ( t ) = ∑ ci E (u ) t −t jj −1 , t 0 = 0, t N +1 = t
t −t

j =1
(7)
⎛ u⎞ Fatigue Characteristic Curves
E (u ) = E∞u − ∑ E j ρ j exp ⎜ − ⎟
j ⎝ ρj ⎠ The following steps are used to construct the characteristic
curve for the fatigue tests.
Calculation of Pseudo Stiffness (C)
and Damage Parameter (S)
Step 1: Calculate Average Strain
The pseudo stiffness, C, is defined as:
The strain history is decomposed into the mean strain and
σ (t ) cyclic strain components for analysis. The average strain is
C1( t ) = (8)
ε R (t ) calculated for each cycle from each LVDT. The individual

0.5 0.012

0.4 0.01
Stress, MPa

Strain, m/m

0.008
0.3
0.006
0.2
0.004
0.1 0.002

0 0
0 50 100 150 200
Time, Sec
Stress Crosshead Strain OSP Strain

Figure E2. Stress and strain histories for a typical

monotonic test.
124

3000

Average Strain (MS)

2000

1000

0
0 500 1000 1500 2000
Time, sec

Figure E3. Mean strain for a PG 67-22 at optimum specimen.

strains recorded by different LVDTs are then averaged to Step 4: Calculate Cyclic Pseudo Strain
determine the mean strain history for the specimen. Figure E3
The cyclic strain is determined by subtracting the mean
shows the mean strain during a constant amplitude fatigue
strain from total strain. The cyclic strain is then fit using the
test for a PG 67-22 optimum specimen.
following equation:

Step 2: Calculate Initial Pseudo Stiffness ε cy ( t ) = p + q cos ( ωt + ϕ ) (10)

Initial pseudo strain is needed to determine the initial
where, p, q, ω and ϕ are regression constants. The cyclic
pseudo stiffness, which is used to calculate the normalized
pseudo strains are calculated using:
pseudo stiffness for the entire fatigue test. Strains captured
during the first loading cycle are used to calculate the initial
pseudo strains in that cycle using Equation 6. The initial ε cyR ( t ) = q E ∗ cos ( ωt + ϕ ) (11)
pseudo stiffness is calculated as the slope of the initial linear
portion of the stress-vs-pseudo strain plot. Figure E4 demon- where, |E ⴱ| is the dynamic modulus of the mix at test temper-
strates the calculation of the initial pseudo stiffness for a par- ature and frequency.
ticular specimen.
Step 5: Calculate Maximum Pseudo Strain
Step 3: Calculate Mean Pseudo Strain in Each Cycle
The mean strains determined in Step 1 are used to calcu- The maximum pseudo strains for each cycle after the first
late the corresponding pseudo strains using the methodology cycle are calculated by adding the maximum cyclic pseudo
described for the monotonic tests. strain in each cycle to the corresponding mean pseudo

1.50

y = 1.1302x + 0.0302
2
R = 0.9966
1.00
Stress, MPa

0.50

0.00
0.00 0.20 0.40 0.60 0.80 1.00 1.20
Pseudostrain

Figure E4. Determination of initial pseudo stiffness.

 125

6.0

Max Pseudostrain
4.0

2.0

0.0
0 500 1000 1500 2000
Time, sec

Figure E5. Typical plot of maximum pseudo strain versus time.

α 1
strain value. Figure E5 shows the variation of the maximum 1+α t − t
⎡1
( ) (Ci−1 − Ci )⎤⎥ ⎛⎜⎝ i i−1 ⎞⎟⎠
N 1+α
S ( t ) = ∑ ⎢ I ε Rmax
2
pseudo strain over the course of a typical constant ampli- (15b)
tude fatigue test. i =1 ⎣ 2 ⎦ x

Equation 15a is used to calculate S during first loading cycle

Step 6: Calculate Pseudo Stiffness and Equation 15b is used during the rest of the loading
cycles until failure. The parameter x is the fraction of
The pseudo stiffness is defined as:
the total stress-vs-strain cycle during which damage can
σ grow. This is that portion of the loading curve where ten-
SR = (12)
εR sile stress occurs. To determine x, plots of the stress-vs-
strain curves are examined. From Figures E6 and E7, it can
For the first cycle, pseudo stiffness is calculated at each point be seen that an appropriate value for x is 4.0 (tensile stress
along the loading path. The secant pseudo stiffness is calcu- on the loading portion of the curve is approximately 1⁄4 of
lated for subsequent cycles to represent the change in slope of the whole loop).
the stress- pseudo strain loops:
σ max Step 9: C-vs-S Characteristic Curve
SR = (13)
ε max
R
The characteristic curve is constructed by cross-plotting C
where ε is the maximum pseudo strain in a cycle and σmax
R and S. An example is shown in Figure E8.
max
is the stress corresponding to ε Rmax.

Step 7: Calculate Normalized Pseudo Stiffness

The normalized pseudo stiffness, C, is calculated as:

SR
C= (14)
I

where, I is the initial pseudo stiffness, calculated in Step 2.

Step 8: Calculate Damage Parameter

The damage parameter, S, is calculated by using the follow-
ing equations:
α
Figure E6. Stress-vs-strain plots for specimen
N
⎡1 ⎤ 1+α 1
S ( t ) = ∑ ⎢ I ( ε R ) (Ci −1 − Ci ) ⎥ ( t i − t i −1 )1+α
2
(15a) No. 14.
i =1 ⎣ 2 ⎦
126

Figure E7. Stress-vs-strain plots for specimen

no. 15.

0.75
C1

0.5

0.25

0
0 0.5 1 1.5 2 2.5 3
S1

Figure E8. Example characteristic curve.

 127

APPENDIX F

NCHRP 9-38 Beam Fatigue Round Robin

The beam fatigue testing conducted to date indicates that 3. After mixing, the batch should be split to the size required
there is an endurance limit for hot mix asphalt (HMA). for your compaction device. We use the following formula
Techniques have been developed to identify the endurance for estimating the target sample weight for compaction:
limit using beam fatigue testing. One concern is that there is
no precision statement for AASHTO T321, the beam fatigue Target Weight = (Target Density − Correction Factor)
test. A full round robin is beyond the scope of this study. ÷ 100 × Gmm × Compacted Sample Volume
However, a mini-round robin should provide an indication
of the variability of beam fatigue testing and of the determi- where,
nation of the endurance limit. The round robin will encom- Target Density = 93%,
pass: sample preparation, beam fatigue testing, calculations Correction Factor = 2.5 accounts for surface voids and
to assess the endurance limit. Three techniques will be used the fact that the center of a com-
to analyze the beam fatigue results: single-stage Weibull pacted sample tends to be denser,
function, logarithmic extrapolation, and ratio of dissipated Gmm = 2.586,
energy. The following describes the sample preparation and Compacted Sample = length x width x height in cm3 =
testing. A second document will be sent at a later date, which Volume (for us) 7.78 × 39.8 × 11.25 =
will describe the data analysis. 3483 cm3.
The mixes included in the study and the labs testing each mix
are shown in Table 1. The mixes are the same mixes used previ- Please contact us if we did not supply a large enough sam-
ously in the NCHRP 9-38 study and are based on the lower lay- ple for your mold. If the initial sample does not produce 7 ±
ers of the structural sections of the 2003 NCAT Test Track. 0.5 percent air voids, this number may need to be adjusted.
Optimum asphalt content for both mixes is 4.5 percent by The target density for the optimum plus mix is 96.7 percent.
total weight of mix. The optimum plus asphalt content is 5.2
percent by total weight of mix. 4. The mix should then be aged for four hours at 275°F
(135 °C) according to AASHTO R30.
5. The sample should then be compacted using your in-
Directions for Preparation house procedure. Vary the compaction effort to achieve
of Samples
the predetermined volume. Normally this means com-
1. Each aggregate “batch” of material consists of two parts pacting to a specified height.
“A” and “B.” The aggregate batches were randomized 6. After the sample has cooled, it is a good idea to bulk the
before shipping to the individual labs. To make one com- sample according to AASHTO T166 before sawing the
plete batch, an “A” and “B” can should be dry mixed. The sample to the test dimensions. This will be used, if neces-
combined aggregate weight should be 8,776 grams. sary, to adjust the mass of future samples to achieve the
2. Therefore, 413.5 grams of binder should be mixed with correct air voids.
one batch for the optimum asphalt content samples and 7. Use a wet saw to cut the compacted beam to 380 ± 6 mm
481.4 grams of asphalt for the optimum plus samples. The in length, 63 ± 6 mm in width, and 50 ± 6 mm in height.
mixing temperatures are: 350°F for the PG 64-22 and 8. Determine the mass under water and SSD mass accord-
350°F for the PG 76-22. ing to AASHTO T166. Dry the sample in front of a fan to
128

Table 1. Testing matrix.

Lab/Mix PG 64-22 at PG 64-22 at PG 76-22 at
Optimum Optimum Optimum
Plus
NCAT X X X
Asphalt Institute X X X
University of Illinois X X X
VA Transportation Research Council X
SEM Materials X
University of California X

a constant mass to determine the dry mass. Calculate the You can then solve: Y = 2E + 19(x)−5.3087 for 50,000,000
sample density and air voids using a Gmm = 2.586. If the cycles.
air voids for the optimum asphalt content samples are
not 7 ± 0.5 percent or the air voids of the optimum plus 2 E + 19
Recall that x–5.3087=1/x5.3087, then x = 5.3087
samples are not 3.3 ± 0.5 percent, the sample density is 50, 000, 000
not in tolerance and a new sample must be made. Evalu- = 153 micro-strain
ate whether the sample mass should be adjusted. We nor-
mally adjusted by multiplying the dry mass in step 6 by You can also avoid the algebra and solve this using a
the desired density divided by the measured density. least-squares procedure and solver in Excel.
9. Is the sample is not going to be tested within 5 days, wrap
the sample in plastic wrap and store it in a freezer. 13. Test three beams at the strain value predicted to give
10. Condition the sample to 20.0 ± 0.5°C for two hours. 50,000,000 cycles, in this example 153 micro-strain. Ten-
Samples that have been frozen should be allowed to thaw tatively, we would like all of the labs to use the same strain
at room temperature for 16 hours prior to conditioning. level. Once you determine the strain level for 50,000,000
11. Test the sample according to AASHTO T321 cycles in Step 12, please contact Brian Prowell. These sets
a. Test three samples at 800 micro-strain of beams should be tested to a maximum of 12 million
b. Test three samples at 400 micro-strain cycles. Later instruction will describe the procedure to
12. Plot the results on a log-log graph and fit a regression extrapolate the data and confirm the identification of the
line. This would be a power model in excel (Figure 1). endurance limit.

100000000

10000000

1000000
Cycles to 50% Initial Stiffness

100000
-5.3087
y = 2E+19x
2
R = 0.9987
10000

1000

100

10

1
1 10 100 1000
micro-strain

Figure 1. Log-log plot of strain versus cycles to failure.

Abbreviations and acronyms used without definitions in TRB publications:
AAAE American Association of Airport Executives
AASHO American Association of State Highway Officials
AASHTO American Association of State Highway and Transportation Officials
ACI–NA Airports Council International–North America
ACRP Airport Cooperative Research Program
ADA Americans with Disabilities Act
APTA American Public Transportation Association
ASCE American Society of Civil Engineers
ASME American Society of Mechanical Engineers
ASTM American Society for Testing and Materials
ATA Air Transport Association
ATA American Trucking Associations
CTAA Community Transportation Association of America
CTBSSP Commercial Truck and Bus Safety Synthesis Program
DHS Department of Homeland Security
DOE Department of Energy
EPA Environmental Protection Agency
FAA Federal Aviation Administration
FHWA Federal Highway Administration
FMCSA Federal Motor Carrier Safety Administration
FRA Federal Railroad Administration
FTA Federal Transit Administration
HMCRP Hazardous Materials Cooperative Research Program
IEEE Institute of Electrical and Electronics Engineers
ISTEA Intermodal Surface Transportation Efficiency Act of 1991
ITE Institute of Transportation Engineers
NASA National Aeronautics and Space Administration
NASAO National Association of State Aviation Officials
NCFRP National Cooperative Freight Research Program
NCHRP National Cooperative Highway Research Program
NHTSA National Highway Traffic Safety Administration
NTSB National Transportation Safety Board
PHMSA Pipeline and Hazardous Materials Safety Administration
RITA Research and Innovative Technology Administration
SAE Society of Automotive Engineers
SAFETEA-LU Safe, Accountable, Flexible, Efficient Transportation Equity Act:
A Legacy for Users (2005)
TCRP Transit Cooperative Research Program
TEA-21 Transportation Equity Act for the 21st Century (1998)
TRB Transportation Research Board
TSA Transportation Security Administration
U.S.DOT United States Department of Transportation