Simplicity 1695079 Technical information

Purdue University
Purdue e-Pubs
JTRP Technical Reports
Joint Transportation Research Program
2009
Validation of NCAT Structural Test Track
Experiment Using INDOT APT Facility
Eyal Levenberg
Purdue University
Rebecca S. McDaniel
Purdue University
Jan Olek
Purdue University
Recommended Citation
Levenberg, E., R. S. McDaniel, and J. Olek. Validation of NCAT Structural Test Track Experiment
Using INDOT APT Facility. Publication FHWA/IN/JTRP-2008/26. Joint Transportation Research
Program, Indiana Department of Transportation and Purdue University, West Lafayette, Indiana,
2009. doi: 10.5703/1288284314311.
This document has been made available through Purdue e-Pubs, a service of the Purdue University Libraries. Please contact epubs@purdue.edu for
additional information.
FHWA/IN/JTRP-2008/26
Final Report
VALIDATION OF NCAT STRUCTURAL TEST
TRACK EXPERIMENT USING INDOT APT
FACILITY
Eyal Levenberg
Rebecca S. McDaniel
September 2009
Final Report
FHWA/IN/JTRP-2008/26
VALIDATION OF NCAT STRUCTURAL TEST TRACK EQUIPMENT
USING INDOT APT FACILITY
By
Eyal Levenberg
Postdoctoral Researcher
Rebecca S. McDaniel
Technical Director
and
Jan Olek
Professor of Civil Engineering and Director
North Central Superpave Center
School of Civil Engineering
Purdue University
Joint Transportation Research Program
Project No. C-36-31R
File No. 2-11-18
SPR-2813
Conducted in Cooperation with the
Indiana Department of Transportation and the
Federal Highway Administration
U.S. Department of Transportation
The contents of this report reflect the views of the authors who are responsible for the
facts and accuracy of the data presented herein. The contents do not necessarily reflect
the official views or policies of the Indiana Department of Transportation and Federal
Highway Administration. This report does not constitute a standard, specification, or
regulation.
Purdue University
West Lafayette, Indiana
September 2009
TECHNICAL REPORT STANDARD TITLE PAGE
1. Report No.
2. Government Accession No.
3. Recipient's Catalog No.
FHWA/IN/JTRP-2008/26
4. Title and Subtitle
5.
Validation of NCAT Structural Test Track Experiment Using INDOT APT
Facility
Report Date
September 2009
6. Performing Organization Code
8. Performing Organization Report No.
7. Author(s)
Eyal Levenberg, Rebecca S. McDaniel, and Jan Olek
FHWA/IN/JTRP-2008/26
9. Performing Organization Name and Address
10. Work Unit No.
Joint Transportation Research Program
550 Stadium Mall Drive
Purdue University
West Lafayette, IN 47907-2051
11. Contract or Grant No.
SPR-2813
12. Sponsoring Agency Name and Address
13. Type of Report and Period Covered
Indiana Department of Transportation
State Office Building
100 North Senate Avenue
Indianapolis, IN 46204
Final Report
14. Sponsoring Agency Code
15. Supplementary Notes
Prepared in cooperation with the Indiana Department of Transportation and Federal Highway Administration.
16.
Abstract
The National Center for Asphalt Technology (NCAT) operates a full-scale test road for studying the response and
performance of asphalt pavements. During the 2003 - 2005 testing phase, NCAT instrumented eight of their test sections
with stress and strain gauges. Two of the test sections were later replicated, along with embedded instrumentation, for
subsequent testing in the accelerated pavement testing (APT) facility operated by the Indiana Department of Transportation.
The availability of similarly constructed and instrumented pavement systems loaded in different conditions offered a unique
opportunity to develop and test the forecastability of pavement models. Exploring this aspect is the topic of the present
work, in which an attempt is made to use the APT experiment in conjunction with laboratory test results, and forecast
resilient responses obtained at NCAT that were generated under completely different loading and environmental conditions.
The modeling and analysis methodologies are outlined in detail and the calculation results are compared with NCAT
measurements. Findings are discussed and recommendations for future research are given.
.
17. Key Words
18. Distribution Statement
asphalt pavements, accelerated pavement testing, embedded
instrumentation, resilient response, forecastability, inverse
analysis, elasticity, viscoelasticity, transverse isotropy.
No restrictions. This document is available to the public through the
National Technical Information Service, Springfield, VA 22161
19. Security Classif. (of this report)
Unclassified
Form DOT F 1700.7 (8-69)
20. Security Classif. (of this page)
Unclassified
21. No. of Pages
173
22. Price
ACKNOWLEDGMENTS
The principal author is thankful to various individuals for their assistance and
involvement during the completion of the report. First, thanks are due to the members of
the Study Advisory Committee (SAC): Tommy Nantung, Dave Andrewski, and Kurt
Sommer from the Indiana Department of Transportation; Gerald Huber from Heritage
Research; Lee Gallivan from the Federal Highway Administration (FHWA) Indiana
Division; Lloyd Bandy from the Asphalt Pavement Association of Indiana; John
Haddock from Purdue University; and Dudley Bonte from Rieth-Riley Construction.
A sincere appreciation goes to Robert “Buzz” Powell, manager of the NCAT
Test Track at Opelika, Alabama, and to David H. Timm from Auburn University for
their collaboration and cooperation in developing the instrumentation installation
procedures and for sharing the response data recorded at the Track.
Special thanks are owed to Tom Robertson and Calvin Reck from Purdue
University for their collaboration in all aspects of the APT experiment starting with
construction and instrumentation, and ending with loading and data collection.
I would like to take this opportunity and also thank Terhi Pellinen, the original
Principal Investigator (PI) of this project along with her research assistants Jose Llenín
and Greg Webster who jointly designed and overlooked the APT experiments and left
me with well organized notes and photographs to work with.
Last but not least, and on a more personal level, would like to thank the people
of the North Central Superpave Center (NCSC), especially: Jan Olek, Becky McDaniel,
and Ayesha Shah, for inviting me to join the NCSC ‘family’, for giving me the
opportunity to work on this project, and for placing their trust in me to see it through to
completion.
TABLE OF CONTENTS
ACKNOWLEDGMENTS........................................................................................ ii
LIST OF TABLES .................................................................................................. vi
LIST OF FIGURES ............................................................................................... viii
CHAPTER 1 - INTRODUCTION ........................................................................ 1-1
1.1 BACKGROUND AND MOTIVATION................................................................................ 1-1
1.2 PROBLEM STATEMENT, STUDY OBJECTIVES AND SCOPE ..................................... 1-2
1.3 WORK PLAN ........................................................................................................................ 1-3
1.4 STRUCTURE OF THE REPORT ......................................................................................... 1-4
CHAPTER 2 - THE NCAT EXPERIMENT ........................................................ 2-1
2.1 FACILITY DESCRIPTION ................................................................................................... 2-1
2.2 CONSTRUCTION OF TEST SECTIONS ............................................................................ 2-3
2.2.1 Pavement Structures ............................................................................................... 2-3
2.2.2 Subgrade ................................................................................................................. 2-4
2.2.3 Granular Base ......................................................................................................... 2-5
2.2.4 Hot Mix Asphalt ..................................................................................................... 2-6
2.3 MECHANICAL TESTING ........................................................................................2-9
2.3.1 Resilient Modulus of Subgrade and Base Materials............................................... 2-9
2.3.2 HMA Complex Modulus ...................................................................................... 2-12
2.3.3 Falling Weight Deflections .................................................................................. 2-18
2.4 EMBEDDED INSTRUMENTATION .....................................................................2-20
2.4.1 Environmental Monitoring ................................................................................... 2-20
2.4.2 Mechanical Responses ......................................................................................... 2-21
2.5 STRUCTURAL BEHAVIOR AT NCAT ................................................................2-24
2.5.1 Resilient Response ............................................................................................... 2-24
2.5.2 Cracking and Rutting Performance ...................................................................... 2-30
CHAPTER 3 - THE APT EXPERIMENT............................................................ 3-1
3.1 FACILITY DESCRIPTION ................................................................................................... 3-1
3.2 COMPOSITION OF TEST SECTIONS ................................................................................ 3-5
3.3 MECHANICAL TESTING AND PRELIMINARY ANALYSIS ......................................... 3-7
3.3.1 Resilient Modulus of Unbound Materials .............................................................. 3-7
3.3.2 HMA Complex Modulus ...................................................................................... 3-10
3.3.3 Falling Weight Deflections .................................................................................. 3-16
3.4 LOADING HISTORY AND ANALYSIS DATASET ........................................................ 3-17
3.4.1 Application of Load Passes .................................................................................. 3-17
3.4.2 Identification of Dataset for Structural Investigation ........................................... 3-21
3.5 STRUCTURAL BEHAVIOR .............................................................................................. 3-22
3.5.1 Instrumentation..................................................................................................... 3-22
3.5.2 Resilient Response ............................................................................................... 3-25
3.5.3 Rutting and Cracking Performance ...................................................................... 3-29
3.6 DIRECT COMPARISON WITH NCAT RESULTS ........................................................... 3-38
iv
CHAPTER 4 - BASIC MECHANISTIC ANALYSIS ......................................... 4-1
4.1 SCOPE AND APPROACH.................................................................................................... 4-1
4.2 LAYERED ELASTIC ISOTROPIC MODEL ....................................................................... 4-2
4.2.1 Theory and Computational Implementation........................................................... 4-2
4.2.2 Calibration to APT Conditions ............................................................................... 4-7
4.2.3 Interim Results and Discussion ............................................................................ 4-10
4.3 NCAT RESPONSE PREDICTION ..................................................................................... 4-13
4.3.1 Methodology ........................................................................................................ 4-13
4.3.2 Falling Weight Deflections .................................................................................. 4-17
4.3.3 Traffic Induced Stresses and Strains .................................................................... 4-19
4.4 APPRAISAL OF BASIC ANALYSIS................................................................................. 4-29
CHAPTER 5 - ADVANCED MECHANISTIC METHODS ............................... 5-1
5.1 LAYERED ELASTIC ANISOTROPIC MODEL ................................................................. 5-1
5.1.1 Theory and Computational Implementation........................................................... 5-1
5.1.2 Calibration to APT Conditions ............................................................................... 5-9
5.1.3 NCAT Response Prediction ................................................................................. 5-12
5.2 LAYERED VISCOELASTIC ISOTROPIC MODEL ......................................................... 5-18
5.2.1 Theory and Computational Implementation......................................................... 5-18
5.2.2 Calibration to APT Conditions ............................................................................. 5-25
5.2.3 NCAT Response Prediction ................................................................................. 5-30
5.3 APPRAISAL OF ADVANCED METHODS ...................................................................... 5-38
CHAPTER 6 - CONCLUSION ............................................................................ 6-1
6.1 SUMMARY AND FINDINGS .............................................................................................. 6-1
6.2 RECCOMENDATIONS AND IMPLEMENTATION .......................................................... 6-4
REFERENCES ...................................................................................................... 7-1
APPENDICES (DVD Available Upon Request) ................................................. 8-1
APPENDIX A: RAW EXPERIMENTAL DATA .......................................................... 8-1
APPENDIX B: COMPUTER PROGRAMS ................................................................... 8-2
APPENDIX C: REPORTS AND CORRESPONDANCE .............................................. 8-2
v
LIST OF TABLES
Table 2.1.1: Breakdown of axle properties for an ‘average’ NCAT truck (Priest and
Timm, 2006) ................................................................................................................. 2-3
Table 2.2.1: HMA design parameters for mixes 1 to 4 (Timm and Priest, 2006) ........................ 2-8
Table 2.3.1: Resilient modulus of subgrade soil with average compaction level of
96%. Raw test results from laboratory reports ........................................................... 2-10
Table 2.3.2: Resilient modulus of aggregate base with average compaction level of
93%. Raw test results from laboratory reports ........................................................... 2-11
Table 2.3.3: Average complex modulus test results for Mix 1 (Barde and Cardone,
2004) ........................................................................................................................... 2-14
Table 2.3.4: Average complex modulus test results for Mix 2 (Barde and Cardone,
2004) ........................................................................................................................... 2-15
Table 2.3.5: Average complex modulus test results for Mix 3 (Barde and Cardone,
2004) ........................................................................................................................... 2-16
Table 2.3.6: Average complex modulus test results for Mix 4 (Barde and Cardone,
2004) ........................................................................................................................... 2-17
Table 2.3.7: FWD deflections at N1 section (location according to NCAT database:
station 2 inside the wheel path) .................................................................................. 2-19
Table 2.3.8: FWD deflections at N2 section (location according to NCAT database:
station 2 inside the wheel path) .................................................................................. 2-19
Table 2.5.1: Temperature profile in Section N1 for analysis of resilient response data............. 2-24
Table 2.5.2: Tabulated progression of N1 and N2 rutting levels vs. number of applied
ESALs......................................................................................................................... 2-31
Table 3.3.1: Resilient modulus of unbound materials (calibrated equation 3.3.2
parameters) ................................................................................................................... 3-9
Table 3.3.2: Complex modulus analysis results for a reference temperature of 15.5ºC
based on the approach in Levenberg and Shah (2008) ............................................... 3-13
Table 3.3.3: Peak FWD deflections measured in the center of sections n1 and n2 .................... 3-17
Table 3.4.1: APT pass application log for sections n1 and n2 (original structure) .................... 3-18
Table 3.4.2: APT pass application log for Section n2 (rehabilitated structure) ......................... 3-19
Table 3.4.3: APT pass application log for Section n1 (rehabilitated structure) ......................... 3-20
Table 3.5.1: Location of APT instrumentation in Section n1 (relate to Figure 3.5.1)................ 3-25
Table 4.2.1: Backcalculated layer moduli for pass #5,000 and pass #80,000 ............................ 4-10
Table 4.3.1: Combined complex modulus properties for APT n1 / NCAT N1 (based
on equations 4.3.1 to 4.3.4) ........................................................................................ 4-16
Table 4.3.2: Matching errors between isotropic LET predictions and NCAT measured
responses. Errors are in percent after normalization using the corresponding
peak to peak response shown in brackets (in microstrains) ........................................ 4-23
vi
Table 5.1.1: Backcalculated anisotropic layer moduli for pass #5,000 ...................................... 5-10
Table 5.1.2: Adjusted anisotropic HMA moduli for FWD response prediction ........................ 5-13
Table 5.2.1: Backcalculated material properties for the layered viscoelastic model
during APT pass #5,000 ............................................................................................. 5-27
Table 5.2.2: Relative improvement in response predictions for the isotropic LVT
compared to the isotropic LET case given in Table 4.3.2 .......................................... 5-32
vii
LIST OF FIGURES
Figure 2.1.1: Schematic layout of the 46 test sections at NCAT (Phase II) experiment .............. 2-2
Figure 2.1.2: Photograph of a typical NCAT truck (Priest and Timm, 2006) .............................. 2-3
Figure 2.2.1: Structural layers for sections N1 to N8 (Priest and Timm, 2006)........................... 2-4
Figure 2.2.2: Final gradation of upper subgrade soil at N1 and N2 sections (Timm and
Priest, 2006) .................................................................................................................. 2-5
Figure 2.2.3: Gradation of base material at N1 and N2 sections (Timm and Priest,
2006) ............................................................................................................................. 2-6
Figure 2.2.4: Sub-layering of HMA in test sections N1 to N8 (Timm and Priest, 2006)............. 2-7
Figure 2.2.5: Design gradation of different mix types used in the NCAT ‘structural
study’ (Timm and Priest, 2006) .................................................................................... 2-8
Figure 2.4.1: Photograph and dimensions (in inches) of an asphalt strain gauge (Timm
et al. 2004) .................................................................................................................. 2-21
Figure 2.4.2: Photograph of Geokon Earth Pressure cell Model 3500 ....................................... 2-22
Figure 2.4.3: Sensor layout for section N1 (based on Timm et al., 2004).................................. 2-23
Figure 2.5.1: Vertical stresses (i.e. stress in Z) on top of the base course (upper chart)
and on top of the subgrade (lower chart) as a result of one truck pass.
Gauges positioned along the Y-axis in Figure 2.4.3 ................................................... 2-26
Figure 2.5.2: Horizontal strains in the loading direction (i.e., strain in Y) at the bottom
of the HMA course as a result of one truck pass. Gauges positioned along
the Y-axis in Figure 2.4.3 ........................................................................................... 2-27
Figure 2.5.3: Horizontal strains in the loading direction (i.e., strain in Y) at the bottom
of the HMA course as a result of one truck pass. Gauges offset by 24 in.
(610 mm) compared to the Y-axis in Figure 2.4.3...................................................... 2-28
Figure 2.5.4: Horizontal strains in the transverse direction (i.e., strain in X) at the
bottom of the HMA course as a result of one truck pass. Gauges positioned
along Y-axis in Figure 2.4.3 ....................................................................................... 2-29
Figure 2.5.5: Fatigued sections N1 (left photo) and N2 (right photo) ........................................ 2-30
Figure 2.5.6: Graphical progression of N1 and N2 rutting levels vs. number of applied
ESALs at the Track..................................................................................................... 2-32
Figure 3.1.1: Schematic floor plan of INDOT APT facility......................................................... 3-2
Figure 3.1.2: Picture of empty test pit and APT loading system .................................................. 3-3
Figure 3.2.1: Composition of APT test pavements n1 and n2...................................................... 3-5
Figure 3.3.1: Resilient modulus of unbound materials - a cross plot of calibrated
equation 3.3.2 values and test data ............................................................................. 3-10
Figure 3.3.2: Mix 1 dynamic modulus and phase angle master curves @ 15.5ºC ..................... 3-13
Figure 3.3.3: Mix 2 dynamic modulus and phase angle master curves @ 15.5ºC ..................... 3-14
Figure 3.3.4: Mix 3 dynamic modulus and phase angle master curves @ 15.5ºC ..................... 3-14
viii
Figure 3.3.5: Mix 4 dynamic modulus and phase angle master curves @ 15.5ºC ..................... 3-15
Figure 3.3.6: Superimposed dynamic modulus master curves @ 15.5ºC for mixes 1 to
4 .................................................................................................................................. 3-15
Figure 3.3.7: Superimposed phase angle master curves @ 15.5ºC for mixes 1 to 4 .................. 3-16
Figure 3.3.8: Peak FWD deflections measured in the center of sections n1 and n2 .................. 3-17
Figure 3.5.1: Plan of embedded instrumentation in APT lane 1 (Section n1) ............................ 3-24
Figure 3.5.2: Measured vertical stresses in Section n1 on top of the base and on top of
the subgrade during pass #5,000 (solid line) and pass #80,000 (dashed line) ............ 3-26
Figure 3.5.3 Measured horizontal strains at the bottom of the HMA in the direction of
loading during pass #5,000 (solid line) and pass #80,000 (dashed line) .................... 3-27
Figure 3.5.4: Measured horizontal strains at the bottom of the HMA in the transverse
direction to the loading during pass #5,000 (solid line) and pass #80,000
(dashed line) ............................................................................................................... 3-29
Figure 3.5.5: Rutting development in Section n1 at the central cross section during the
first 90,000 load passes (applied without wheel wander) ........................................... 3-31
Figure 3.5.6: Contour plot of Section n1 rutting after 100 passes .............................................. 3-32
Figure 3.5.7: Contour plot of Section n1 rutting after 500 passes .............................................. 3-32
Figure 3.5.8: Contour plot of Section n1 rutting after 1,000 passes ........................................... 3-33
Figure 3.5.9: Contour plot of Section n1 rutting after 5,000 passes ........................................... 3-33
Figure 3.5.10: Contour plot of Section n1 rutting after 10,000 passes ....................................... 3-34
Figure 3.5.11: Contour plot of Section n1 rutting after 20,000 passes ....................................... 3-34
Figure 3.5.12: Contour plot of Section n1 rutting after 30,000 passes ....................................... 3-35
Figure 3.5.13: Contour plot of Section n1 rutting after 40,000 passes ....................................... 3-35
Figure 3.5.14: Contour plot of Section n1 rutting after 50,000 passes ....................................... 3-36
Figure 3.5.15: Contour plot of Section n1 rutting after 60,000 passes ....................................... 3-36
Figure 3.5.16: Contour plot of Section n1 rutting after 70,000 passes ....................................... 3-37
Figure 3.5.17: Contour plot of Section n1 rutting after 80,000 passes ....................................... 3-37
Figure 3.5.18: Contour plot of Section n1 rutting after 90,000 passes ....................................... 3-38
Figure 4.2.1: User interface of the isotropic LET program ELLEA1 (see Appendix B) ............ 4-6
Figure 4.2.2: Resilient responses during APT pass #5,000. Both measured (solid
markers) and model generated (solid line) are shown ................................................ 4-11
Figure 4.2.3: Resilient responses during APT pass #80,000. Both measured (solid
markers) and model generated (solid line) are shown ................................................ 4-12
Figure 4.3.1: Combined HMA dynamic modulus and phase angle master curves for a
reference temperature of 15.5ºC (based on Table 4.3.1) ............................................ 4-15
Figure 4.3.2: Combined HMA time-temperature shifting for a reference temperature
of 15.5ºC (based on Table 4.3.1) ................................................................................ 4-17
Figure 4.3.3: Measured and projected peak FWD deflections at NCAT N1.............................. 4-18
ix
Figure 4.3.4: Layout of N1 gauge array (refer to Figure 2.4.3) and travel path
positioning of the center point of the rightmost truck tire (connecting
arrows) ........................................................................................................................ 4-21
Figure 4.3.5: Travel paths of center of rightmost truck wheels over the N1 gauge array
at NCAT for the different axles in Table 2.1.1 ........................................................... 4-22
Figure 4.3.6: Calculated and measured N1 responses - right side of steering axle (1S) ............ 4-24
Figure 4.3.7: Calculated and measured N1 responses - right side of drive axle (1D and
2D) .............................................................................................................................. 4-25
Figure 4.3.8: Calculated and measured N1 responses - right side of first trailer axle
(1T) ............................................................................................................................. 4-26
Figure 4.3.9: Calculated and measured N1 responses - right side of third trailer axle
(3T) ............................................................................................................................. 4-27
Figure 4.3.10: Calculated and measured N1 responses - right side of last trailer axle
(5T) ............................................................................................................................. 4-28
Figure 5.1.1: User interface of the anisotropic LET program ELLEA2...................................... 5-8
Figure 5.1.2: ELLEA2 display of property and algorithm restriction for the example
shown in Figure 5.1.1 ................................................................................................... 5-8
Figure 5.1.3: Comparison of measured resilient responses in the APT during pass
#5,000 with responses computed using the anisotropic layered model
(isotropic case is reproduced from Figure 4.2.2) ........................................................ 5-12
Figure 5.1.4: Comparison of measured peak FWD deflections at NCAT N1 with
projected peak deflections using anisotropic LET (isotropic case reproduced
from Figure 4.3.3) ....................................................................................................... 5-14
Figure 5.1.5: Comparison of anisotropic LET projections with measured N1 responses
- right side of steering axle (1S). Isotropic case reproduced from Figure
4.3.6 ............................................................................................................................ 5-16
Figure 5.1.6: Comparison of anisotropic LET projections with measured N1 responses
- right side of third trailer axle (3T). Isotropic case reproduced from Figure
4.3.9 ............................................................................................................................ 5-17
Figure 5.2.1: Scheme for simulating a moving load on a layered viscoelastic model .............. 5-22
Figure 5.2.2: Indicial admittance of a layered viscoelastic system (example). Strain
response due to a unit intensity ‘input’ of an APT half-axle passing along
four offset distances from the evaluation point .......................................................... 5-24
Figure 5.2.3: Comparison of backcalculated relaxation modulus with that
interconverted from complex modulus test results ..................................................... 5-28
Figure 5.2.4: Comparison of measured resilient responses in the APT during pass
#5,000 with responses computed using the isotropic viscoelastic layered
model .......................................................................................................................... 5-30
Figure 5.2.5: Comparison of isotropic LVT projections with measured N1 responses right side of steer axle (1S). Isotropic case reproduced from Figure 4.3.6 ................. 5-33
x
Figure 5.2.6: Comparison of isotropic LVT projections with measured N1 responses right side of drive axle (1D and 2D). Isotropic case reproduced from Figure
4.3.7 ............................................................................................................................ 5-34
Figure 5.2.7: Comparison of isotropic LVT projections with measured N1 responses right side of first trailer axle (1T). Isotropic case reproduced from Figure
4.3.8 ............................................................................................................................ 5-35
Figure 5.2.8: Comparison of isotropic LVT projections with measured N1 responses right side of third trailer axle (3T). Isotropic case reproduced from Figure
4.3.9 ............................................................................................................................ 5-36
Figure 5.2.9: Comparison of isotropic LVT projections with measured N1 responses right side of fifth (last) trailer axle (5T). Isotropic case reproduced from
Figure 4.3.10 ............................................................................................................... 5-37
xi
CHAPTER 1 - INTRODUCTION
1.1 BACKGROUND AND MOTIVATION
Despite years of systematic research the design and analysis of asphalt pavements still
includes dominant empirical components. This state of affairs may be ascribed to the
complexity of the problem: first, the mechanical behavior of pavement materials and
subgrades is not well understood and cannot always be sufficiently controlled; second,
material properties continuously change during the pavement service life due to natural
processes such as oxidation and age hardening; third, only rough estimates can be
provided for basic design inputs such as traffic loads and environmental conditions; and
finally, the notion of structural failure is not defined in a clear cut manner. For these
reasons available analysis methods are heavily based on past experience and as such
cannot be used reliably with non-traditional materials and components and cannot aid in
optimizing pavement designs.
Accelerated pavement testing (APT) facilities were built in an effort to address
some of the aforementioned limitations. These facilities offer controlled study
conditions in which pavements and subgrades of known materials can be loaded and
closely monitored. Traditionally, the design and use of these facilities was driven by
empirical approaches, resulting in studies in which the performance of different
pavements under similar loading conditions was monitored and compared. In these tests
significant effort was placed on accelerating rutting and cracking damage in an attempt
to capture in a relatively short period of time (of the order of months) equivalent field
experience that can only be gained over a period of many years. More recent APT
studies have been driven by pavement mechanics principles in an effort to
accommodate or aid in the development of more rational design methods. In either case
it is widely accepted that experimental results obtained in APT facilities are not directly
applicable to the field and that the sophisticated interpretation is required.
Pavetrack is a full scale test road located near the campus of Auburn University
in Alabama, operated and managed by the National Center for Asphalt Technology
(NCAT); it is a closed-loop facility that applies accelerated truck traffic to 46 adjoining
1-1
experimental sections paved with Hot Mix Asphalt (HMA). The 2003 - 2005 testing
phase at NCAT, also known as Phase II, included the construction, loading and
continuous monitoring of eight different instrumented pavement structures, referred to
as sections N1 to N8. The primary objective of this so-called ‘structural study’ was to
provide high quality data for validating the Mechanistic-Empirical Pavement Design
Guide (MEPDG) (ARA Inc., 2004). For this purpose, during the two year loading
period, performance data (e.g., cracking, rutting, roughness and skid resistance) and
response data (i.e., stresses, strains and deflections) were recorded within the structure
and subgrade along with prevailing environmental conditions (e.g., temperatures and
moisture levels).
In 2004, the Indiana Department of Transportation (INDOT) and Purdue
University engaged in a smaller-scale research project that is closely related to the
NCAT ‘structural study’ experiment. In this project, the two NCAT test sections N1 and
N2 were replicated in the INDOT APT facility along with embedded instrumentation.
These sections, referred to herein as n1 and n2, were loaded in the APT over a two year
period between 2004 and 2006.
Consequently, similar instrumented pavement structures were made available,
loaded in completely different conditions with a closely monitored environment. It is
the overall motivation of this study to try and establish a relation between the behavior
(i.e., both response and performance) of the pavements in the APT facility and their
corresponding behavior at the NCAT test track. By establishing such a relation, the
methodology used can be potentially applied with confidence to future APT studies as
means of forecasting field behavior of replicate pavement systems.
1.2 PROBLEM STATEMENT, STUDY OBJECTIVES AND SCOPE
By their very nature, the testing conditions in the APT facility are considerably more
uniform compared to field conditions. For example, in the current APT study use was
made of a single axle configuration, single axle weight and one loading speed; also, the
entire experiment was carried out under constant temperature conditions. Accordingly,
the observed pavement behavior in the APT is the result of these limited conditions
1-2
only. Therefore, the problem is how to interpret APT experimental results such that they
could be applied to different environmental and loading conditions.
The main objective of this study is to devise and validate an analysis scheme by
which experimental data collected in the APT experiment can be used to successfully
forecast the corresponding pavement behavior at the NCAT test track. The analysis
scheme is based on mechanistic principals in order to provide a rational basis for the
interpretation and allow the incorporation of different complexity levels depending on
the desired accuracy of the outcome. The work includes an underlying basic assumption
that the pavement structures in both cases were similar. It is beyond the scope of this
research to consider the case of different structures.
Focus is placed on the analysis of responses, i.e., stresses, strains and
deflections, and less on cracking and rutting performance. This is mainly because
accurate response prediction is the underlying key for reliable performance forecasting
and due to a scarcity of adequate performance data from both experiments. Moreover, it
is important to note that only resilient (recoverable) responses will be addressed. This is
mainly because the type of embedded instrumentation installed in both experiments was
only suited for monitoring dynamic (transient) pavement reactions and not for recording
permanent (irrecoverable) responses. In fact, this latter point is a known shortcoming of
all available strain and stress gauges commercially available at this time.
This study has also a secondary objective which is to summarize the work
performed and document the available experimental data. A clear description of what
was done, how it was done and what data was collected may encourage additional
studies using the existing records.
1.3 WORK PLAN
The initial work plan, described in a previous report by Llenín and Pellinen (2004),
consisted of tasks related to construction of the APT experiment and execution of the
accompanying laboratory tests (see Appendix C). At this time test data from both the
APT and NCAT experiments are available; therefore, the work plan outlined herein
1-3
includes only the tasks required to achieve the aforementioned study objectives. The
scope and purpose of each task are described hereafter.
Task 1 consists of careful study, systematic documentation and presentation of
pertinent test data from both the APT and NCAT experiments. The aim here is to
familiarize the reader with relevant details of the work done. It is mostly descriptive in
nature with limited pre-processing of the data. The raw test results are contained in
Appendix A.
Task 2 includes the identification of response data suitable for mechanistic
analysis; the aim here is to identify a subset of the available data that is most suitable
and sufficient for accomplishing the main study objective.
Task 3 contains the development and calibration of a mechanistic pavement
model based on APT data only. This task is the central element of the entire
methodology. As a basic case, the mechanistic ‘engine’ of the MEPDG is applied, i.e.,
layered elastic theory (LET) with isotropic material properties. Also considered are two
more advanced models, namely: LET with transversely isotropic material properties and
layered viscoelastic theory (LVT) with linear isotropic material properties. In each case
the numerical values of the model parameters are obtained from inverse analysis by
simulating the APT experiment and matching the measured responses.
In Task 4 the capabilities of the APT model are enhanced. For this purpose an
analysis scheme is developed by which the derived material properties obtained in Task
3 are adjusted in order for them to apply to other loading configurations, other loading
speeds, and different environmental conditions. These adjustments are based primarily
on the analysis of laboratory test data.
Finally, Task 5 deals with model validation using NCAT results. For this
purpose the loading and environment at NCAT are simulated and the forecastability of
the ‘enhanced’ model (Task 4) is assessed by comparison with NCAT measurements.
1.4 STRUCTURE OF THE REPORT
Chapters 2 and 3 address tasks 1 and 2. Chapter 2 is mainly narrative with minimal
interpretation; containing relevant information from the NCAT experiment such as
description of the facility loading conditions, composition and instrumentation of the
1-4
test sections and recorded field and laboratory behavior. Chapter 3 summarizes the APT
work; it includes a description of the loading history and environment prevailing during
the experiment, some preliminary analyses of available the test data and identification
of dataset most suitable for structural investigation. This chapter also presents the
recorded structural behavior for the selected dataset. Thereafter a direct comparison
with the NCAT results is provided to emphasize the need for more intricate and
fundamental analysis.
Tasks 3, 4 and 5 are addressed by Chapters 4 and 5; both contain the
development of mechanistic models for representing the pavement systems considered.
Chapter 4 deals only with isotropic LET while Chapter 5 deals anisotropic LET and
isotropic LVT. In these chapters it is shown how the necessary material properties are
calibrated using the APT experiment. Thereafter, they explain how to apply the models
to other loading and environmental conditions that were not included in the calibration.
Finally, selected responses at NCAT are predicted using the models and compared with
field measurements for validation purposes. The final chapter, Chapter 6, includes a
short summary of the entire report outlining the main findings. It also provides some
general recommendations and implementation suggestions for INDOT.
1-5
CHAPTER 2 - THE NCAT EXPERIMENT
This chapter summarizes the ‘structural study’ experiment conducted by NCAT
between the years 2003 and 2005 (phase II). The focus is on sections N1 and N2 which
were later replicated in the APT experiment. Reference to the original reports is
provided so the reader can trace the source and obtain additional data.
2.1 FACILITY DESCRIPTION
The NCAT test track is a 1.7 miles (2.8 km) oval shaped closed-loop asphalt road
located near Opelika, Alabama. The primary objective for building the track was to
provide a practical, engineering driven, research tool for validation of laboratory tests
and pavement design procedures under accelerated and controlled traffic conditions
(Brown et al., 2002). The general track layout is shown in Figure 2.1.1. As can be seen,
the facility allows for the simultaneous loading of 46 experimental sections; there are 26
sections on the tangents, each about 200 ft long (60.96 m). The curved potions of the
track host the remaining 20 sections. The track was constructed to have two lanes and
paved shoulders on each side of the roadway. Both lanes were designed and constructed
to have the same materials and thicknesses within a given test section. One lane was
used for traffic and pavement performance monitoring. The other lane provided a safety
lane in case of truck breakdowns and construction access for repairing existing sections.
Construction of the track was completed in 2000 and the first cycle of tests (i.e.,
Phase I) took place between the years 2000 and 2002. The second testing cycle (Phase
II), which is relevant to this report, took place between the years 2003 and 2005. Most
of the sections from the 2000 experiment were either left as-is for the 2003 experiment
or rehabilitated by shallow milling and inlaying. These sections were originally
involved in a comparative rutting study which was extended to the second cycle of
testing. Only eight new sections were rebuilt from the subgrade up for the 2003
experiment: sections N1 to N8 (Timm et al., 2004; 2006). These sections were devoted
to a ‘structural study’ focusing primarily upon the effects of HMA thickness and binder
type/grade on the dynamic pavement responses under truck loading. These pavements
were instrumented to monitor load induced horizontal strains in the bottom of the HMA,
2-1
vertical compressive stresses on top of the base and subgrade, moisture in the unbound
materials and temperature within the HMA. Additionally, these sections were
investigated by periodic deflection testing and monitored for structural distresses by
employing routine surface condition surveys. Reportedly this was done to allow for
later validation of the MEPDG. The focus herein is on sections N1 and N2, which were
the only sections from the ‘structural study’ replicated in the INDOT APT.
Figure 2.1.1: Schematic layout of the 46 test sections at NCAT (Phase II) experiment.
Traffic loadings at the track are applied using a designated fleet of tractor-trailer
trucks (triple trailer), each traveling at 45 mph or 792 in./s (72.4 km/h or 20.1 m/s).
Drivers are utilized to operate the trucks, and their operation consists of two 7.5 hour
shifts, five days a week. Typically, each truck completes about 26 laps in an hour. A
picture of one NCAT truck is provided in Figure 2.1.2. As can be seen, the truck has
several wheel assemblies: single-axle single-wheels (steer axle); tandem-axle dualwheels (drive axle); and single-axle dual-wheels (trailer axle). The ‘average’ weight
carried by the individual axles is shown in table 2.1.1. This is an ‘average’ weight
because each of the five trucks in the fleet has a slightly different load. The coefficient
of variation (COV) of the loads is also shown in the table with values in the range of
2-2
1.7% to 4.9% (average of 3.1%). Standard tires were used (Priest et al., 2005) identified
as 275/80R22.5 and inflated to 100 psi.
Figure 2.1.2: Photograph of a typical NCAT truck (Priest and Timm, 2006).
Table 2.1.1: Breakdown of axle properties for an ‘average’ NCAT truck (Priest and
Timm, 2006).
Axle-name
AxleNumber
Steer
Axle-type
SingleAxle
No. of
Wheels
Average
Axle-Load,
lb (kg)
COV for 5
Trucks
1S
Drive
1D
Trailer
2D
Tandem-Axle
1T
2T
3T
4T
5T
Single
-Axle
SingleAxle
Single
-Axle
Single
-Axle
Single
-Axle
2
4
4
4
4
4
4
4
10,680
20,320
20,290
21,010
20,760
21,310
20,550
20,613
(4,850)
(9,225)
(9,210)
(9,540)
(9,425)
(9,675)
(9,330)
(9,360)
3.9%
3.9%
4.9%
2.2%
2.5%
1.7%
3.6%
2.2%
2.2 CONSTRUCTION OF TEST SECTIONS
2.2.1 Pavement Structures
Pavement structures in the NCAT ‘structural study’ were designed using the 1993
AASHTO guide employing three different traffic levels. The final outcome is shown in
Figure 2.2.1. It may be seen that each structure includes 6 in. (152 mm) of unbound
granular base under HMA layers of varying thickness having different composition
and/or binder grade. Both N1 and N2 sections were designed with 5 in. (127 mm) of
HMA, sections N3 and N4 were designed with 9 in. (229 mm) of HMA, and sections
2-3
N5 to N8 were designed with 7 in. (178 mm) of HMA each. It is important to note that
these are design values and that the actual as-constructed thicknesses varied slightly.
Figure 2.2.1: Structural layers for sections N1 to N8 (Priest and Timm, 2006).
2.2.2 Subgrade
The upper subgrade for all eight test sections (N1 to N8) was processed to a depth of 30
in. (762 mm) from the pavement surface (Timm and Priest, 2006). The material was
then compacted in layers using vibratory pad-foot rollers (e.g., Dynapac CA15PD). It
should be noted that large cobbles, which were originally present in the material, broke
down under rolling. The final outcome was classified as an A-4(0) soil; the resulting
gradation curve after rolling operations is presented in Figure 2.2.2. As can be seen, 100
percent of the material is smaller than the 1.5 in. sieve (38.1 mm opening), and more
than 45% passes the #200 sieve (0.075 mm opening). Below the upper subgrade the
material originating from the 2000 track construction remained untouched.
When the upper subgrade soil was compacted in the lab using modified Proctor
effort, the resulting maximum dry unit weight was 119.6 pcf (1918 kg/m³) with a
corresponding moisture content of 8.6% (i.e., laboratory optimum). The average as-built
2-4
moisture contents for sections N1 and N2 were 10% and 11% respectively. The average
in-place wet unit weight was 132.0 pcf (2116 kg/m³) for both sections, and the
corresponding dry unit weights were 120.0 and 118.9 pcf (1924 and 1906 kg/m³). From
this information it may be concluded that the relative in-place degree of compaction,
based on dry densities, was 100.3% and 99.4% for sections N1 and N2 respectively.
Figure 2.2.2: Final gradation of upper subgrade soil at N1 and N2 sections (Timm and
Priest, 2006).
2.2.3 Granular Base
The granular base consisted of a well graded crushed granite material, compacted in a
single 6 in. (152 mm) layer. The gradation of this material is shown in Figure 2.2.3. As
can be seen, 100 percent of the material is smaller than the 1.5 in. sieve (38.1 mm
opening). Also note that less than 10% of the material is passes the #200 sieve (0.075
mm opening). Loose base samples were recompacted in the lab using modified Proctor
effort, resulting in a maximum dry unit weight of 137.9 pcf (2211 kg/m³) and a
corresponding moisture content of 9.2% (i.e., laboratory optimum). The average field
2-5
dry density values were 138.0 (2213 kg/m³) for both sections. With respect to dry
densities, the relative compaction degree for the base layer in both sections was 100.1%.
The as-built moisture content varied slightly: 6.4% in section N1 and 6.6% in section
N2.
Figure 2.2.3: Gradation of base material at N1 and N2 sections (Timm and Priest,
2006).
2.2.4 Hot Mix Asphalt
Figure 2.2.4 presents the sub-layering of the HMA in the different test sections in the
‘structural study’ (Timm and Priest, 2006). Sections N1 and N2 were paved in three lifts
to a total HMA thickness of 5 in. (127 mm). The bottom and intermediate lifts are each
2 in. (50.8 mm) thick while the top lift is 1.0 in. (25.4 mm) thick.
2-6
Figure 2.2.4: Sub-layering of HMA in test sections N1 to N8 (Timm and Priest, 2006).
Table 2.2.1 presents the individual mixture design parameters for mixes 1 to 4
which were paved in the N1 and N2 sections. It may be seen that the surface mixes 1
and 3 differ by the type of binder and corresponding preparation temperatures. SBS
modified PG 76-22 was used for Mix 1, and unmodified PG 67-22 was used for Mix 3.
Both mixes (see Figure 2.2.5) were designed with a ‘wearing’ (dense) gradation having
a nominal maximum aggregate size (NMAS) of 9.5 mm (3/8 in.) and a compactive
effort of 80 gyrations. At the optimum binder content (i.e., 6.13% effective), the
samples had 4.3% air voids and 17.9% VMA. Similar to the surface mixes, the
intermediate and bottom mixes (mixes 2 and 4) differed by the type of binder. SBS
modified PG 76-22 was used for Mix 2, and unmodified PG 67-22 was used for Mix 4.
Both mixes (see Figure 2.2.5) were designed with a ‘Base’ (also dense) gradation
having a NMAS of 19 mm (3/4 in.) and a compactive effort of 80 gyrations. At
optimum binder content (4.27% effective), the samples had 4.3% air voids and 14.5%
VMA.
2-7
Table 2.2.1: HMA design parameters for mixes 1 to 4 (Timm and Priest, 2006).
Property
Units
Mix 1 Mix 2 Mix 3 Mix 4
Binder Grade
76-22
67-22
Compactive Effort,
gyrations
80
Mixing Temperature
ºF (ºC)
345 (174)
325 (163)
Effective Binder Content
percent
6.13
4.27
6.13
4.27
Dust to Binder Ratio
0.88
1.10
0.88
1.10
pcf
147.8 153.6 147.8 153.6
Bulk Unit Weight of Compacted Pills
(kg/m³) (2370) (2463) (2370) (2463)
Air Void Content
percent
4.3
Voids in Mineral Aggregate
percent
17.9
14.5
17.9
14.5
Figure 2.2.5: Design gradation of different mix types used in the NCAT ‘structural
study’ (Timm and Priest, 2006).
Surveys of the as-built thicknesses were conducted during construction. These
focused on the pavement areas that included instrumentation. The average thicknesses
of individual lifts were as follows. For section N1, the bottom lift was 2.2 in. (56 mm)
thick, the intermediate was 2.1 in. (53 mm) thick, and the top lift was 0.6 in. (15 mm)
thick. For section N2, the bottom lift was 1.8 in. (46 mm) thick, the intermediate lift
was 2.0 in. (51 mm), and the top lift was 1.1 in. (28 mm) thick. Therefore, the average
total HMA thickness over the instrumented areas was 4.9 in. (124 mm) in both sections.
Cores taken after construction confirmed these results. It should be noted that the HMA
2-8
thickness in Section N1 was not constant at 5.0 in. (127 mm), but varied between 6.7 to
4.5 in. (170 to 114 mm).
The as-built air void contents were also surveyed during construction. On
average, all three lifts in Section N1 were compacted to an air void content of 7.0%.
However, in Section N2 only the top lift was compacted to 7.0% voids while the two
bottom lifts were compacted to an air void content of 6.0%. The asphalt content also
varied slightly relative to the design values. Additional details regarding the HMA
construction can be found in Powell and Brown (2004).
2.3 MECHANICAL TESTING
2.3.1 Resilient Modulus of Subgrade and Base Materials
The resilient modulus of the upper subgrade soil was tested according to the AASHTO
T307 protocol. Tests were done either on recompacted material or on undisturbed
specimens obtained from the field after subgrade construction. The test conditions
included three levels of moisture content (denoted by ω ): 7.2%, 9.7% and 20.1%; three
levels of applied confining pressure: 2, 4 and 6 psi (13.8, 27.6 and 41.4 kPa); and five
levels of applied cyclic axial stress: 2, 4, 6, 8 and 10 psi (13.8, 27.6, 41.4 and 68.9 kPa).
The average dry density of the tested samples was 114.4 pcf (1834 kg/m³) which
represents a relative compaction degree of about 96% (recall that the compaction degree
in the field was about 100%). The raw test results are presented in Table 2.3.1. These
were obtained directly from the laboratory reports.
Timm and Priest (2006) provided three regression equations, each fitting the test
results for a single level of moisture content; the mathematical expression they used was
as follows:
M R = b1 ⋅ ( Sc )b2 ( S3 )b3 ...................................................................................... (2.3.1)
where M R is the resilient modulus (in psi), b1 , b2 and b3 are regression constants, S c
is the peak applied uniaxial cyclic stress (in psi), and S 3 is the applied confining
pressure (in psi).
2-9
Table 2.3.1: Resilient modulus of subgrade soil with average compaction level of 96%.
Raw test results from laboratory reports.
Confining
Pressure,
psi (kPa)
Peak
Cyclic
Stress
psi
(kPa)
Resilient
Modulus,
psi (MPa)
ω=7.2%
6.0
(41.4)
4.0
(27.6)
2.0
(13.8)
2.1
(14.5)
4.0
(27.6)
5.8
(40.0)
7.7
(53.1)
9.6
(66.2)
2.1
(14.5)
3.9
(26.9)
5.7
(39.3)
7.6
(52.4)
9.5
(65.5)
1.9
(13.1)
3.8
(26.2)
5.6
(38.6)
7.5
(51.7)
9.5
(65.5)
10,876
(75.0)
11,738
(80.9)
11,730
(80.9)
12,260
(84.5)
12,837
(88.5)
10,311
(71.1)
10,222
(70.5)
10,252
(70.7)
10,465
(72.2)
10,965
(75.6)
7,871
(54.3)
7,783
(53.7)
7,920
(54.6)
8,216
(56.6)
8,776
(60.5)
Peak
Cyclic
Stress,
psi
(kPa)
Resilient
Modulus,
psi (MPa)
ω =9.7%
2.1
(14.5)
4.0
(27.6)
5.9
(40.7)
7.7
(53.1)
9.6
(66.2)
2.0
(13.8)
3.9
(26.9)
5.9
(40.7)
7.8
(53.8)
9.7
(66.9)
2.0
(13.8)
4.0
(27.6)
6.0
(41.4)
7.9
(54.5)
9.7
(66.9)
9,970
(68.7)
11,177
(77.1)
11,478
(79.1)
11,922
(82.2)
12,384
(85.4)
7,310
(50.4)
7,762
(53.5)
8,487
(58.5)
9,225
(63.6)
9,987
(68.9)
4,339
(29.9)
4,701
(32.4)
5,507
(38.0)
6,285
(43.3)
7,084
(48.8)
Peak
Cyclic
Stress,
psi
(kPa)
Resilient
Modulus,
psi (MPa)
ω=20.1%
2.2
(15.2)
4.1
(28.3)
5.9
(40.7)
7.7
(53.1)
9.6
(66.2)
2.1
(14.5)
4.0
(27.6)
5.9
(40.7)
7.7
(53.1)
9.6
(66.2)
2.0
(13.8)
4.0
(27.6)
5.9
(40.7)
7.8
(53.8)
9.7
(66.9)
10,034
(69.2)
10,653
(73.4)
10,568
(72.9)
10,511
(72.5)
10,477
(72.2)
8,502
(58.6)
8,614
(59.4)
8,736
(60.2)
8,854
(61.0)
9,101
(62.7)
5,032
(34.7)
5,451
(37.6)
6,119
(42.2)
6,641
(45.8)
7,149
(49.3)
The resilient modulus of the base material was also tested according to
AASHTO T307 procedure. Tests were done on specimens recompacted to a dry density
of 128.6 pcf (2062 kg/m³). This value represents a relatively low compaction degree of
2-10
about 93% (recall that 100% compaction was achieved in the field). The test conditions
included two levels of moisture content ( ω ): 5.3% and 9.8%; three levels of applied
confining pressure: 2, 4 and 6 psi (13.8, 27.6 and 41.4 kPa); and five levels of applied
cyclic axial stress between 2 and 10 psi (13.8 to 68.9 kPa). The raw test results are
given in Table 2.3.2.
Table 2.3.2: Resilient modulus of aggregate base with average compaction level of
93%. Raw test results from laboratory reports.
Confining
Pressure, psi
(kPa)
6.0
(41.4)
4.0
(27.6)
2.0
(13.8)
Peak Cyclic
Stress, psi
(kPa)
Resilient
Modulus, psi
(MPa)
Peak Cyclic
Stress, psi
(kPa)
ω=5.3%
2.2 (15.2)
4.1 (28.3)
5.9 (40.7)
7.8 (53.8)
9.7 (66.9)
2.1 (14.5)
3.9 (26.9)
5.8 (40.0)
7.7 (53.1)
9.7 (66.9)
2.0 (13.8)
3.8 (26.2)
5.7 (39.3)
7.7 (53.1)
9.7 (66.9)
9,062 (62.5)
9,236 (63.7)
9,080 (62.6)
9,017 (62.2)
9,138 (63.0)
7,615 (52.5)
7,091 (48.9)
6,986 (48.2)
7,164 (49.4)
7,481 (51.6)
5,194 (35.8)
4,843 (33.4)
4,983 (34.4)
5,339 (36.8)
5,731 (39.5)
Resilient
Modulus, psi
(MPa)
ω =9.8%
2.1 (14.5)
4.0 (27.6)
5.8 (40.0)
7.6 (52.4)
9.5 (65.5)
2.1 (14.5)
4.0 (27.6)
5.9 (40.7)
7.7 (53.1)
9.5 (65.5)
2.0 (13.8)
3.9 (26.9)
5.9 (40.7)
7.7 (53.1)
9.5 (65.5)
8,698 (60.0)
8,950 (61.7)
8,785 (60.6)
8,751 (60.3)
8,864 (61.1)
6,830 (47.1)
6,322 (43.6)
6,366 (43.9)
6,642 (45.8)
6,945 (47.9)
4,280 (29.5)
3,966 (27.3)
4,233 (29.2)
4,643 (32.0)
5,003 (34.5)
Another set of tests was done on the base material with a relative compaction
degree of 97.5% and a moisture content of 5.5%. These tests were done under applied
confining pressure levels ranging from 3 to 20 psi (20.7 to 137.9 kPa), and cyclic stress
levels in the range of 3 to 40 psi (20.7 to 275.8 kPa). Relating to equation 2.1, the
obtained coefficients were: b1 = 5677 , b2 = 0 and b3 = 0.4711 .
2-11
2.3.2 HMA Complex Modulus
Complex modulus testing of the NCAT asphalt mixtures was conducted under the
direction of Dr. Terhi Pellinen at Purdue University. A detailed description of the work
is presented in a report by Barde and Cardone (2004) which can be found in Appendix
C. The materials were sampled in loose state from the delivery trucks during
construction and then shipped to Purdue University. Specimens were prepared and
tested in accordance with the NCHRP 1-37A protocol (NCHRP, 2002).
Initially, mixtures were heated to compaction temperature in the oven; a
temperature of 155 °C was selected for mixes 1 and 3, and a temperature of 145 °C was
selected for mixes 2 and 4. Next, specimens were prepared using a gyratory compactor.
For each mixture type, four specimens were prepared for testing (i.e., a total of 16) by
targeting an air void content of 7%. The first two specimens, denoted by ‘a’ and ‘b’ in
the report, were compacted with assumed correction factors (i.e., ratio between sample
bulk specific gravity that is measured using saturated surface-day method and bulk
specific gravity that is calculated based on sample height in the gyratory mold). The
next two specimens, denoted by ‘c’ and ‘d’ in the report, were prepared with calibrated
correction factors after studying the resulting densities of the first two.
The gyratory compacted pills were further cored to produce cylindrical
specimens 4 in. (102 mm) in diameter and 6 in. (152 mm) in height. The final air void
content for the 16 samples varied in the range of 6.8 to 8.2%. Complex modulus tests
were done on each of the 16 samples in stress-controlled uniaxial compression mode
with a Haversine load pattern. Six different frequencies were employed: 25, 10, 5, 1, 0.5
and 0.1 Hz; applied in descending order. The frequency sweep was executed five times
for five different temperature levels: -10, 4.4, 21.1, 37.8 and 54.4 °C; applied in
ascending order.
A dummy specimen was used to verify that the target temperature level was
reached. The testing was carried out using a UTM 25 with a 25 kN load cell. Stress
levels were adjusted in order to maintain the values of both permanent and recoverable
strain within specified protocol limits. In an effort to minimize edge effects, latex sheets
covered with silicon grease were inserted between loading system and specimen at both
2-12
ends. Specimens were instrumented with three linear variable displacement
transformers (LVDTs), each spanning 100 mm, for measurement of axial (vertical)
deformations; the LVDTs were mounted to the periphery of the specimen 120° apart. It
should be noted that lateral strains were not monitored.
The complex modulus test results are shown in Tables 2.3.3 to 2.3.6. Each table
presents the dynamic moduli values and the corresponding phase angles. It was found,
separately for each mix, that the initial differences in voids did not affect the test results
in a statistically significant manner. Therefore, each value in these tables constitutes the
average result of four separate tests.
2-13
Table 2.3.3: Average complex modulus test results for Mix 1 (Barde and Cardone,
2004).
Test Temperature Test Frequency Dynamic Modulus Phase Angle
ºC (ºF)
-10.0
(+14)
+4.4
(+40)
+21.1
(+70)
+37.8
(+100)
+54.4
(+130)
[Hz]
MPa (ksi)
[degrees]
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
23,057 (3,344)
21,808 (3,163)
20,585 (2,986)
18,129 (2,629)
17,017 (2,468)
14,364 (2,083)
13,792 (2,000)
12,365 (1,793)
11,418 (1,656)
9,391 (1,362)
8,563 (1,242)
6,722 (975)
7,099 (1,030)
5,983 (868)
5,052 (733)
3,529 (512)
3,009 (436)
2,061 (299)
2,622 (380)
2,139 (310)
1,694 (246)
1,020 (148)
862 (125)
589 (85)
1,054 (153)
855 (124)
685 (99)
436 (63)
382 (55)
286 (41)
6.1
9.2
9.7
10.9
11.5
13.2
10.1
11.2
13.6
15.3
16.3
18.9
16.0
17.0
23.1
26.5
27.4
29.3
27.5
23.5
25.3
33.6
32.3
30.2
22.2
15.1
16.6
21.1
22.8
26.9
2-14
Table 2.3.4: Average complex modulus test results for Mix 2 (Barde and Cardone,
2004).
Test Temperature Test Frequency Dynamic Modulus Phase Angle
ºC (ºF)
-10.0
(+14)
+4.4
(+40)
+21.1
(+70)
+37.8
(+100)
+54.4
(+130)
[Hz]
MPa (ksi)
[degrees]
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25,440 (3,690)
24,150 (3,503)
22,781 (3,304)
20,048 (2,908)
18,791 (2,725)
15,703 (2,278)
16,581 (2,405)
14,813 (2,148)
13,710 (1,988)
11,191 (1,623)
10,168 (1,475)
7,951 (1,153)
8,575 (1,244)
7,321 (1,062)
6,302 (914)
4,503 (653)
3,898 (565)
2,722 (395)
4,238 (615)
3,498 (507)
2,792 (405)
1,752 (254)
1,481 (215)
1,044 (151)
1,504 (218)
1,200 (174)
985 (143)
655 (95)
578 (84)
450 (65)
6.6
9.7
10.5
12.1
12.8
15.1
10.9
12.2
13.9
16.0
17.2
20.4
15.3
15.3
22.6
25.0
26.1
28.6
26.2
27.3
24.9
32.5
32.1
31.2
26.2
22.7
21.5
25.4
26.4
27.4
2-15
Table 2.3.5: Average complex modulus test results for Mix 3 (Barde and Cardone,
2004).
Test Temperature Test Frequency Dynamic Modulus Phase Angle
ºC (ºF)
-10.0
(+14)
+4.4
(+40)
+21.1
(+70)
+37.8
(+100)
+54.4
(+130)
[Hz]
MPa (ksi)
[degrees]
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
22,559 (3,272)
21,386 (3,102)
20,300 (2,944)
17,801 (2,582)
16,667 (2,417)
13,991 (2,029)
11,884 (1,724)
10,786 (1,564)
9,879 (1,433)
7,858 (1,140)
7,111 (1,031)
5,457 (791)
7,113 (1,032)
5,928 (860)
5,087 (738)
3,504 (508)
2,970 (431)
1,992 (289)
2,634 (382)
2,038 (296)
1,624 (236)
1,008 (146)
851 (123)
604 (88)
673 (98)
534 (77)
448 (65)
313 (45)
276 (40)
216 (31)
4.9
7.8
8.9
10.5
11.3
13.5
11.2
13.1
15.6
17.6
18.9
21.8
16.7
19.1
21.9
27.3
28.0
30.1
24.3
20.5
23.7
28.7
28.6
27.9
31.2
28.9
29.7
26.5
25.4
24.1
2-16
Table 2.3.6: Average complex modulus test results for Mix 4 (Barde and Cardone,
2004).
Test Temperature Test Frequency Dynamic Modulus Phase Angle
ºC (ºF)
-10.0
(+14)
+4.4
(+40)
+21.1
(+70)
+37.8
(+100)
+54.4
(+130)
[Hz]
MPa (ksi)
[degrees]
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
24,940 (3,617)
23,592 (3,422)
22,678 (3,289)
19,975 (2,897)
18,783 (2,724)
15,847 (2,298)
16,045 (2,327)
14,393 (2,088)
13,241 (1,920)
10,641 (1,543)
9,622 (1,396)
7,394 (1,072)
9,036 (1,311)
7,341 (1,065)
6,419 (931)
4,436 (643)
3,772 (547)
2,612 (379)
3,958 (574)
3,072 (446)
2,480 (360)
1,608 (233)
1,363 (198)
962 (140)
1,485 (215)
1,157 (168)
931 (135)
619 (90)
548 (79)
435 (63)
5.8
7.9
9.1
10.6
11.5
13.6
10.5
13.0
14.8
17.4
18.7
22.3
16.9
22.0
23.0
26.7
28.2
30.9
23.4
23.9
28.1
30.1
29.8
29.4
24.0
16.5
17.2
20.7
21.6
24.6
2-17
2.3.3 Falling Weight Deflections
Falling weight deflectometer (FWD) testing was conducted more or less on a monthly
basis at NCAT. These tests were conducted at identical locations within each test
section. The locations themselves were randomly selected at the beginning of the entire
two year experiment. The FWD was a Dynatest 8000 model equipped with seven
sensors spaced at 12 in. (304.8 mm) intervals starting from the center of the load plate.
The load plate had a radius of 300 mm (5.91 in.) and was equipped with a split
configuration to ensure better contact with the pavement surface. For each location, two
drops of about 9,000 lb (4,086 kg) load were executed.
Between November 3, 2003, and February 7, 2005, 135 FWD tests were
conducted on sections N1 and N2 (i.e., 270 drops). The testing locations were later
cored to obtain the as-build pavement thicknesses. The data provided by Table 2.3.7
consists only of those deflections (average of two drops) conducted over a point located
inside the wheel path for which the HMA thickness was verified via coring to be
exactly 5.0 in. (127 mm). It should be noted that the first few deflection bowls
characterize the N1 pavement in its pristine state, soon after construction and before it
had incurred significant damage. Similarly, Table 2.3.8 includes the results for section
N2. Both tables show the HMA temperature at the time of testing measured at a depth
of 50.8 mm (2 in.) from the surface. Also, for preparing both tables the measured
deflections were linearly normalized to a peak load of exactly 9,000 lb (4,086 kg) in
magnitude.
2-18
Table 2.3.7: FWD deflections at N1 section (location according to NCAT database:
station 2 inside the wheel path).
Date
D0
[μm]
D1
[μm]
D2
[μm]
D3
[μm]
D4
[μm]
D5
[μm]
D6
[μm]
Temp.
ºC (ºF)
Nov. 3, 2003
565.7
353.0
153.4
76.9
40.9
32.1
20.7
32.9 (91.2)
Dec. 15, 2003
451.2
332.5
178.4
93.9
47.6
31.2
24.2
16.7 (62.1)
Jan. 26, 2004
338.1
263.2
160.4
99.2
60.9
38.0
29.5
10.1 (50.2)
Feb. 23, 2004
354.0
268.1
160.3
92.1
52.7
30.9
22.5
12.6 (54.7)
Mar. 22, 2004
620.2
416.7
200.7
99.6
55.3
36.0
26.3
26.8 (80.2)
June 14, 2004
771.7
511.9
224.4
101.6
53.3
34.6
26.7
32.3 (90.1)
Sep. 20, 2004
841.3
545.0
237.4
94.8
44.1
36.6
31.6
33.9 (93.0)
Nov. 1, 2004
723.8
490.1
219.8
93.6
45.9
32.0
23.3
29.2 (84.6)
Dec. 6, 2004
866.5
593.6
279.4
109.8
50.0
34.8
28.3
16.2 (61.2)
Feb. 7, 2005
782.8
486.8
190.4
79.3
44.0
30.2
24.4
21.3 (70.3)
Table 2.3.8: FWD deflections at N2 section (location according to NCAT database:
station 2 inside the wheel path).
Date
D0
[μm]
D1
[μm]
D2
[μm]
D3
[μm]
D4
[μm]
D5
[μm]
D6
[μm]
Temp.
ºC (ºF)
Dec. 15, 2003
311.8
230.9
125.6
72.2
43.9
29.7
24.0
15.8 (60.4)
Jan. 26, 2004
328.0
255.1
156.0
93.2
51.4
32.8
22.9
10.2 (50.4)
Feb. 23, 2004
299.1
219.5
126.4
75.1
48.0
30.9
21.9
12.5 (54.5)
Mar. 22, 2004
489.8
327.8
159.7
82.8
50.7
34.7
27.6
25.8 (78.4)
June 14, 2004
632.5
418.9
177.9
83.9
48.7
34.4
26.6
32.8 (91.0)
Sep. 20, 2004
610.6
384.8
157.9
71.6
35.0
33.0
28.7
31.4 (88.5)
Nov. 1, 2004
605.2
377.9
158.7
71.5
42.1
30.7
24.7
28.5 (83.3)
Dec. 6, 2004
570.8
381.1
182.7
85.1
45.0
33.1
28.2
16.0 (60.8)
Jan. 10, 2005
572.4
364.5
173.5
84.5
45.8
31.9
26.8
20.2 (68.4)
Feb. 7, 2005
506.0
314.1
142.2
66.2
35.7
25.0
22.3
19.8 (67.6)
2-19
2.4 EMBEDDED INSTRUMENTATION
2.4.1 Environmental Monitoring
Instrumentation devoted to monitoring environmental changes included moisture probes
and temperature gauges. Campbell Scientific moisture probes (model CS615) were
installed at NCAT during the phase I experiment (Freeman et al., 2001). This type of
gauge was also selected to be installed in the ‘structural study’ (Timm et al., 2004).
These probes indicate changes in volumetric moisture content (i.e., volume of water per
unit bulk volume of soil) by detecting changes in the dielectric constant of the
surrounding material. The dielectric constant of soils is a composition of the dielectric
constants of its individual constituents. Solid soil particles like sand and clay have
dielectric constants in the range of 2 to 4. Water, however, has a much higher dielectric
constant of about 80. Thus, increases in the moisture content of soil can be identified by
measured increases in the soil’s dielectric constant. The CS615 probes, sometimes
referred to as water content reflectometers, use time-domain measurement methods.
They consist of two parallel stainless steel rods, spaced 2 in. (51 mm) apart, connected
to a printed circuit board which is encapsulated in epoxy. Each rod is 12 in. (305 mm)
long and 1/8 in. (3.2 mm) in diameter. The circuit board transmits electrical waves that
travel along the rods with travel times that depend primarily on the dielectric constant of
the surrounding material. When such probes are calibrated for a specific soil, their
accuracy is typically ±2% moisture by volume. The readings from these probes were
not considered reliable enough for the ‘structural study’ (personal communication, D.
H. Timm, 2007).
Temperatures were monitored using passive elements based on thermistor
technology (Model 108 temperature probes manufactured by Campbell Scientific, Inc.).
A thermistor is essentially a resistor whose resistance varies according to temperature.
For each test section in the ‘structural study,’ four thermistors were bundled together to
provide temperature information near the surface and at the following depths: 2, 4 and
10 in. (51, 102 and 254 mm). These probes were installed after paving had been
completed given their survival temperature range is -50°C to +100°C (-58°F to 212°F).
Over the range of -3°C to 90°C (26.6°F to 194°F) the measurement accuracy of these
2-20
probes is ±0.3°C (±0.54°F). Unlike the moisture probes, the temperature probes do not
need to be calibrated for the particular environment in which they are to be used. In
Freeman et al. (2001) they were tested and found repeatable and accurate.
2.4.2 Mechanical Responses
Mechanical responses at the ‘structural study’ were measured with strain gauges and
pressure cells. Both device types are considered suitable for measuring dynamic
responses only because they experience drift over time (for various reasons) that
precludes their use in monitoring permanent changes. In each section an array of 12
stain gauges was attached to the bottom of the HMA course at a depth of 5 in. (i.e.,
z=127 mm). The asphalt strain gauges were manufactured by Construction
Technologies Laboratories (CTL Group) Model ASG-152. A picture of one such a
gauge with corresponding dimensions (in inches) is shown in Figure 2.4.1. It may be
seen that the gauge is made of two ‘T’ shaped metal elements interconnected by a 2 in.
long (51 mm) measuring sensor.
Figure 2.4.1: Photograph and dimensions (in inches) of an asphalt strain gauge (Timm
et al. 2004).
Two pressures cells were embedded within the pavement system: one on top of
the base course (i.e., z=5 in. or z=127 mm) and another on top of the subgrade (z=11 in.
or z=280 mm). Both pressure cells were manufactured by Geokon (3500 circular
model). These devices (see Figure 2.4.2) are constructed from two slightly convex
stainless steel plates welded together around their periphery and separated by a narrow
2-21
gap filled with de-aired hydraulic fluid. When external pressure is applied to the plates,
the two plates are squeezed together causing a corresponding increase of fluid pressure
inside the cell. High pressure stainless steel tubing connects the pressure cell to a semiconductor pressure transducer which converts the increased pressure of the compressed
fluid into an electrical signal. This signal is transmitted through a signal cable to the
readout location.
D=230 mm
t=6 mm
Figure 2.4.2: Photograph of Geokon Earth Pressure cell Model 3500.
The instrumentation layout shown in Figure 2.4.3 refers to Section N1. An
essentially identical layout was assembled in Section N2. The center point of this array
was positioned along the right wheel path, about 8 ft (2.44 m) away from the centerline.
The Y-axis in the figure points in the direction of truck travel while the X-axis points in
the transverse direction. The array of 12 asphalt strain gauges can be seen in the figure,
spaced evenly 24 in. (610 mm) apart in both X and Y directions.
Strain gauges BLL, BLC, BLR, ALL, ALC, and ALR are all measuring strains
in Y. The rest of the strain gauges, namely BTL, BTC, BTR, ATL, ATC and ATR, are
measuring strains in X. It should be noted that five of the strain gauges, ALL, ATL,
BLL, ATR and BTR, did not survive the construction process. The two pressure cells
are also shown in the figure as BBC and ASC. The BBC gauge was located on top of
2-22
the base course and under the HMA course (z=5 in. or z=127 mm). It was designated to
capture vertical stresses (i.e., stress in Z) at this interface. This gauge had a 36.3 psi
(0.25 MPa) range. The ASC gauge was designated to measure the vertical stresses at the
interface between the subgrade and the granular base (z=11 in. or z=280 mm). This
gauge had a 14.5 psi (0.1 MPa) range given that it is placed at a greater depth.
Y-axis [ft]
8
ASC
6
4
ALL
ALR
ALC
2
ATL
-4
ATR
ATC
-2
BTL
BLL
2
4
BTR
BTC
-2
BLC
X-axis [ft]
BLR
-4
-6
BBC
Figure 2.4.3: Sensor layout for section N1 (based on Timm et al., 2004).
2-23
2.5 STRUCTURAL BEHAVIOR AT NCAT
2.5.1 Resilient Response
The resilient responses presented hereafter were obtained from Section N1 while in its
pristine state. The data was collected soon after construction, on November 7, 2003,
during which a single NCAT truck was doing multiple laps. The corresponding raw data
file provided by NCAT for the purpose of this study contained three such laps (see
Appendix A). Herein, the responses due to the first of the three passes is presented and
discussed.
Each time the truck approached the gauge array, the data acquisition system was
switched on and the gauge readings were recorded for a period of about 2 seconds.
Between truck passes, for a period of about 2 minutes, the data acquisition system was
switched off in order to save storage space. The pavement temperatures in different
depths during this experiment are reported in Table 2.5.1. The values in the table are
hourly averages, based on minute-by-minute readings, corresponding to the hour in
which the data files were created (personal communication, D. H. Timm, 2007).
Table 2.5.1: Temperature profile in Section N1 for analysis of resilient response data.
Depth, in. (mm)
Temperature, ºF (ºC)
0
88.5 (31.4)
2 (51)
80.9 (27.2)
4 (102)
77.1 (25.1)
10 (254)
73.7 (23.2)
The following figures present the measured N1 responses to one truck pass as
recorded by the gauge array layout shown in Figure 2.4.3. In all figures the abscissa
represents test time and ranges between 0 and 2.2 seconds (the zero is arbitrary). In each
case circular markers denote NCAT data; the solid line passing between the data points
was created using cubic spline interpolation as means of increasing the ‘measurement’
density (this increased measurement density will be used later on in the report). It is
possible to identify and relate graphically the data in the charts to the specific axle. This
is done using the terminology in Table 2.4.1 where ‘1S’ refers to the steering axle, ‘1D’
2-24
and ‘2D’ refer to the drive axle and ‘1T’ to ‘5T’ refer to the trailer axles. When
reviewing the charts it should be borne in mind that the location of each axle relative to
the gauges was not measured. In fact, the tractor unit and trailers did not follow a
straight line and did not move along or parallel to the Y-axis in Figure 2.4.3. Mainly for
this reason, axles having similar weight recorded different peak responses and
sometimes even responses of opposite sign (compare response due to axles 1T, 2T, 3T
and 4T with the response due to axle 5T in Figure 2.5.4).
Figure 2.51 includes two charts, both presenting measured vertical stresses (i.e.,
stress in Z). The upper chart shows readings from gauge placed on top of the aggregate
base (i.e., gauge BBC) and the lower chart shows readings from the gauge placed on top
of the subgrade (i.e., gauge ASC). The ordinate in both charts ranges between 0 and 18
psi. In both cases it is easy to identify the individual axles. As expected, peak
magnitudes are generally lower in the bottom chart. Also it can be seen that, excluding
the drive axle (dual-tandem), the pressure cells recover fully, i.e., return to a zero
reading, in between individual axle passes. Also interesting to note is that the speed of
the truck can be calculated using the figures. For example, the peak stress due to the
steer axle occurs at t=0.395 seconds according to gauge ASC and at t=0.213 seconds in
gauge BBC. From Figure 2.4.3 it can be seen that these gauges are spaced 12 feet apart.
Therefore, the NCAT truck covered a distance of 12 feet in 0.182 seconds. The
calculated truck speed is 65.93 foot per second or 44.95 mph which is extremely close
to the target speed of 45 mph.
Figure 2.5.2 includes two charts, both presenting strains measured at the bottom
of the HMA course in the direction loading (i.e., strain in Y) by gauges ALC and BLC.
Both gauges are located on the Y-axis in Figure 2.4.3. Similarly, Figure 2.5.3 presents
the strains in Y measured at the bottom of the HMA course by gauges ALR and BLR.
Recall that the location of these gauges is offset by 24 in. (610 mm) compared to the Yaxis in Figure 2.4.3. The ordinate in both cases ranges from -400 microstrains to +200
microstrains. As can be seen, strain reversal is induced to the HMA course as the axle
wheels travel over the pavement. As the load is approaching the gauges measure
compressive strains. Tensile strains are induced as the load gets closer to the gauge. The
strains go back into compression as the load is receding.
2-25
18.0
BBC
Stress in Z (z=5", x=0") [psi]
16.0
1D
14.0
2D
1T
2T
12.0
3T
10.0
4T
8.0
6.0
1S
5T
4.0
2.0
Time [s]
0.0
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
2.20
18.0
ASC
Stress in Z (z=11", x=0") [psi]
16.0
14.0
12.0
2D
10.0
1T
2T
1D
4T
3T
8.0
5T
6.0
1S
4.0
2.0
Time [s]
0.0
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
2.20
Figure 2.5.1: Vertical stresses (i.e. stress in Z) on top of the base course (upper chart)
and on top of the subgrade (lower chart) as a result of one truck pass. Gauges positioned
along the Y-axis in Figure 2.4.3.
2-26
200
BLC
150
Strain in Y (z=5", x=0") [μstrain]
100
50
0
-50
-100
-150
5T
1S
-200
4T
2D
-250
1D
3T
1T
-300
2T
-350
Time [s]
-400
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
2.20
200
ALC
150
Strain in Y (z=5", x=0") [μstrain]
100
50
0
-50
-100
1S
-150
5T
-200
3T
4T
-250
2D
-300
1T
1D
2T
-350
Time [s]
-400
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
2.20
Figure 2.5.2: Horizontal strains in the loading direction (i.e., strain in Y) at the bottom
of the HMA course as a result of one truck pass. Gauges positioned along the Y-axis in
Figure 2.4.3.
2-27
200
BLR
Strain in Y (z=5", x=+24") [μstrain]
150
100
50
0
1S
-50
1T
2D
-100
1D
2T
-150
-200
3T
4T
-250
-300
5T
-350
Time [s]
-400
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
2.20
200
ALR
Strain in Y (z=5", x=+24") [μstrain]
150
100
50
1S
0
-50
1D
-100
2D
1T
2T
-150
3T
-200
4T
-250
-300
5T
-350
Time [s]
-400
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
2.20
Figure 2.5.3: Horizontal strains in the loading direction (i.e., strain in Y) at the bottom
of the HMA course as a result of one truck pass. Gauges offset by 24 in. (610 mm)
compared to the Y-axis in Figure 2.4.3.
The final two charts in Figure 2.5.4 present horizontal strains at the bottom of
the HMA course (i.e., z=5 in. or z=127 mm) as measured by gauges ATC and BTC in
Figure 2.4.3 (i.e., strain in X). As can be seen, tensile (negative) strains of up to 400
2-28
microstrains in magnitude were induced in the transverse direction by all axles except
for the last one (axle 5T). As mentioned earlier, the reason for this change in sign is
related to the location of the load relative to the gauge.
150
BTC
100
5T
Strain in X (z=5", x=0") [μstrain]
50
0
-50
1S
-100
4T
-150
-200
-250
3T
-300
1D
2D
1T
2T
-350
-400
Time [s]
-450
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
2.20
150
ATC
100
5T
Strain in X (z=5", x=0") [μstrain]
50
0
-50
-100
1S
4T
-150
-200
-250
3T
-300
1T
-350
1D
2D
2T
-400
Time [s]
-450
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
2.20
Figure 2.5.4: Horizontal strains in the transverse direction (i.e., strain in X) at the
bottom of the HMA course as a result of one truck pass. Gauges positioned along Yaxis in Figure 2.4.3.
2-29
2.5.2 Cracking and Rutting Performance
Sections N1 and N2 experienced cracking in a very similar manner (Priest and Timm,
2006; Timm et al., 2006). Both failed in fatigue within two months of each other (see
Figure 2.5.5. Section N1 (modified HMA) failed prior to section N2 (unmodified HMA)
after six months of traffic. First, small transverse cracks appeared in the wheel path.
Then the cracks progressed to the edge of the wheel path and often curled in the
direction of traffic. Later, the individual transverse cracks became interconnected into a
classical alligator pattern. Pumping of the fines from the unbound aggregate base
through the cracks was also observed in the individual transverse cracks as well as the
alligator cracked areas. The progression of fatigue failure was fairly rapid once the first
cracks appeared and especially once pumping began. Subsequently, the responses
quickly over ranged the embedded instrumentation. More detailed crack mapping can
be found in Priest and Timm (2006).
Figure 2.5.5: Fatigued sections N1 (left photo) and N2 (right photo).
With respect to rutting, sections N1 and N2 did not rut much by the time they failed in
fatigue. The progression of average rut depth is shown in Table 2.5.2 (personal
communication, B. Powell, 2007); this data is also presented graphically in Figure 2.5.6.
As can be seen, after the application of about 4.5 million equivalent single axle loads
(ESALs), the final average rut depth was only about 8 mm.
2-30
Table 2.5.2: Tabulated progression of N1 and N2 rutting levels vs. number of applied
ESALs.
#
Cumulative
ESALs
Measurement
Date
Rutting in
N1 mm, (in.)
Rutting in
N2, mm (in.)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
0
10,670
40,068
218,357
643,101
727,772
1,055,293
1,115,778
1,179,894
1,259,225
1,608,513
1,695,079
1,787,799
2,233,571
2,412,440
2,501,549
2,595,810
2,694,521
2,863,433
2,970,955
3,060,410
3,166,906
3,450,232
3,559,500
3,741,860
3,921,345
4,031,786
4,122,945
4,219,483
4,300,286
4,382,730
4,553,790
4,671,075
9/12/03
10/27/03
11/10/03
12/8/03
1/12/04
1/19/04
2/16/04
2/23/04
3/1/04
3/8/04
4/5/04
4/12/04
4/19/04
5/24/04
6/7/04
6/14/04
6/21/04
6/28/04
7/12/04
7/19/04
7/26/04
8/2/04
8/23/04
8/30/04
9/13/04
9/26/04
10/4/04
10/11/04
10/18/04
10/25/04
11/1/04
11/15/04
11/29/04
0.00 (0.00)
0.21 (0.01)
0.27 (0.01)
0.27 (0.01)
0.65 (0.03)
0.33 (0.01)
0.65 (0.03)
0.68 (0.03)
0.68 (0.03)
0.84 (0.03)
1.05 (0.04)
1.30 (0.05)
1.33 (0.05)
2.07 (0.08)
2.78 (0.11)
2.84 (0.11)
2.98 (0.12)
3.07 (0.12)
3.43 (0.14)
3.86 (0.15)
4.10 (0.16)
4.28 (0.17)
4.32 (0.17)
4.14 (0.16)
4.91 (0.19)
5.02 (0.20)
5.36 (0.21)
5.84 (0.23)
5.68 (0.22)
6.70 (0.26)
8.04 (0.32)
7.97 (0.31)
7.64 (0.30)
0.00 (0.00)
0.30 (0.01)
0.35 (0.01)
0.42 (0.02)
0.71 (0.03)
0.43 (0.02)
0.85 (0.03)
0.85 (0.03)
0.84 (0.03)
1.34 (0.05)
2.03 (0.08)
1.51 (0.06)
1.51 (0.06)
2.93 (0.12)
3.54 (0.14)
3.49 (0.14)
3.78 (0.15)
3.92 (0.15)
4.31 (0.17)
4.87 (0.19)
5.08 (0.20)
5.11 (0.20)
5.21 (0.21)
5.59 (0.22)
6.05 (0.24)
6.18 (0.24)
6.44 (0.25)
6.11 (0.24)
6.95 (0.27)
7.21 (0.28)
7.52 (0.30)
7.99 (0.31)
8.32 (0.33)
2-31
9.0
N1
N2
8.0
Average Rutting [mm]
7.0
6.0
5.0
4.0
3.0
2.0
1.0
ESALs
0.0
-
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
Figure 2.5.6: Graphical progression of N1 and N2 rutting levels vs. number of applied
ESALs at the Track.
2-32
CHAPTER 3 - THE APT EXPERIMENT
This chapter summarizes the APT work conducted jointly by INDOT and Purdue
University through the North Central Superpave Center between the years 2004 and
2006. It describes the loading history and environment prevailing during the APT
experiment and some preliminary analyses of pertinent test data. Here, a subset of the
available data is identified as suitable and sufficient for carrying out the main study
objective. Finally, to emphasize the need for fundamental analysis, as performed in the
following chapters, a direct comparison with the NCAT results is provided.
3.1 FACILITY DESCRIPTION
The INDOT APT facility was fabricated in the early 1990’s; it was designed and
constructed by Purdue University under a joint venture between INDOT and the School
of Civil Engineering (White et al., 1990; Galal et al., 1998). The facility is housed in a
2,000 sq. ft (~186 sq. m) hangar located near INDOT’s Office of Research and
Development in West Lafayette. The hangar is divided into three major areas, a testing
area, a utility area and an operator control area. Figure 3.1.1 shows a schematic floor
plan of the facility, in which the three different areas can be seen.
The testing area consists of a test pit embedded in a concrete floor. It is 6 ft deep
(1.83 m) and shaped as a square with 20 ft (6.1 m) long sides. Prototype pavements for
testing are constructed inside this pit. Typically, different structures are constructed in
the pit, each 20 ft (6.1 m) long. The width of each section depends on the experimental
configuration. The n1 and n2 sections addressed herein were each 10 ft (3.05 m) wide;
accordingly, as can be seen in Figure 3.1.1 the test pit is split into two lanes (1 and 2).
The NCAT N1 section was replicated in Lane 1 and the N2 section was replicated in
Lane 2. Additionally, since the pit is surrounded by concrete from all sides and bottom,
water can be introduced into the subgrade. Overhanging the test pit from the ceiling is a
radiant heating system. The purpose of this system is to stabilize and control the air
temperature during testing; it is capable of heating the air in the test area up to 140ºF
(60ºC), and maintaining constant pavement surface temperatures to within ±2ºF (if all
doors remain closed). Additionally, the facility has some cooling capabilities using an
3-1
air-conditioning unit. In its maximum capacity, the air temperature inside the test area
can be reduced relative to warmer outdoor conditions to 60ºF (15.5ºC). In this study the
temperature in the testing area was set to 60ºF (15.5ºC). A humidity detector is
positioned close to the test pit; in principle, it is possible to increase the humidity in the
test area by intentionally ponding water.
Figure 3.1.1: Schematic floor plan of INDOT APT facility.
The APT loading system is mounted on a large steel frame with beams spanning
across and bridging the test pit. The frame itself is fixed on each side to steel rails
embedded in the concrete floor. The fixture can be loosened to allow positioning of the
frame above any test-lane of choice. The loading system is designed to produce a
downward force of up to 20,000 lb (9,080 kg). This force is produced by four
3-2
interconnected pneumatic cylinders. A precision air pressure gauge is used to adjust and
control the magnitude of the force throughout the test. The downward force is applied to
the pavement surface through a wheel assembly. Two tire assembly types are available:
dual/conventional and single wide-base. The tire inflation pressures in each case are
adjustable, up to a maximum of 120 psi (0.84 MPa). Either assembly is mounted on a
carriage capable of traversing the test pit by traveling on the steel beams. The carriage is
cable driven by a motor and a control drive. This motor was designed to accelerate the
carriage (with wheel assembly) from a static startup position to a speed of 5 mph (8
km/h or 2.235 m/s) within the first 5 ft (1.52 m) of the test pit. The carriage speed is
then maintained constant, at 5 mph (8 km/h), for the next 12 ft (3.66 m). Finally, the
speed is reduced back to zero within the last 3 ft (0.91 m) of the test pit. For this study
passes were applied via the dual-wheel assembly loaded to 15,000 lb (6,810 kg) with
tires inflated to 100 psi (0.70 MPa). Figure 3.1.2 shows a picture of the APT loading
system (the test pit is empty in this picture).
Figure 3.1.2: Picture of empty test pit and APT loading system.
3-3
Depending on the desired mode of application, the wheel assembly can be raised
from the ground by reversing the action of the four pneumatic cylinders and returned to
the startup position for another loading cycle; this will result in a unidirectional mode of
loading. Alternatively, the pavement can also be loaded while the carriage travels back
to the startup position; this will result in a bidirectional mode of loading. Moreover,
trafficking in the APT can be applied repeatedly in the same wheel path or with wander.
In the latter case, the wheel path is randomly selected by a computer to within ±5 in
(±125 mm) from the centerline. The transverse movement for producing wander is
governed by an electric servo-motor also attached to the steel frame. Under this study
APT passes were applied in unidirectional mode only. As will be discussed later (see
Section 3.4), except for the initial part of the experiment passes were applied with
wander.
The utility room houses a boiler, transformer, heating controls and a water
circulating system. The latter can be used to introduce hot water into pipes embedded in
one of the pavement layers. This feature is mostly used when testing rigid pavements or
composite pavements (i.e., HMA overlaying concrete) to heat up the concrete slabs. The
operator control room houses computers and an interface to fully interact with and
control the APT operation. Currently this room is equipped with three personal
computers networked together; one computer is used for operational control of the APT;
the second computer is used for data collection and reduction; the third computer is set
up to monitor the APT functions. The control room also houses two scanners for
collecting data from sensors embedded in the pavement structure. At this time the
system used for the data acquisition is the Vishay Measurements Group System 6100.
Each scanner unit accepts up to 20 input cards and has the ability to scan at a very high
sampling rate, up to 10,000 samples per second per channel.
In this study a gauge array similar to the NCAT study was installed (see Section
3.5). As the APT carriage traversed the pavement, gauges were sampled at a rate of 100
scans per second. However, not all load passes were captured; usually every tenth cycle
was recorded. Also no data was collected in between load passes during which the
pavement recovered as the APT carriage was lifted in the air and returned to the startup
position. This shutdown of the data recording was done in order to save storage space. It
3-4
should be noted that whenever wander was applied, the lateral carriage position was not
recorded. Hence, the exact location of the load relative to the embedded gauge array is
known only for the initial part of the experiment in which passes were applied without
wander.
3.2 COMPOSITION OF TEST SECTIONS
The pavement structures (and subgrade) of sections n1 and n2 in lanes 1 and 2
respectively (see Figure 3.1.1) are shown in Figure 3.2.1. As can be seen, similar to the
N1 and N2 NCAT sections both are comprised of 5 in. (127 mm) of hot mix asphalt
(HMA) overlaying 6 in. (152 mm) of crushed granite aggregate base course, placed on
top of an A-4(0) soil serving as subgrade. Materials used to construct the sections were
sampled at NCAT and hauled to Indiana for placement in the APT. The properties of
each of the pavement components are discussed in more detail in the following
subsections. Llenín and Pellinen (2004) and Llenín et al. (2006) provide details on the
overall project planning and pavement construction process (see also Appendix C).
HMA
5 in.
Section n1
Section n2
Mix 1: NMAS 9.5 mm; PG 76-22; AVC 7.5%
Mix 3: NMAS 9.5 mm; PG 67-22; AVC 7.3%
Mix 2: NMAS 19 mm; PG 76-22; AVC 10.2%
Mix 4: NMAS 19 mm; PG 67-22; AVC 9.7%
Mix 2: NMAS 19 mm; PG 76-22; AVC 9.1%
Mix 4: NMAS 19 mm; PG 67-22; AVC 9.4%
6 in.
Aggregate
Base
Crushed granite material; Average dry unit
weight: 133.5 pcf = 97%
19 in.
Upper Subgrade
A-4(0) soil; Average dry unit weight:
111.5 pcf = Average compaction degree:
93%
Lower Subgrade: 90% compaction
Figure 3.2.1: Composition of APT test pavements n1 and n2.
3-5
The subgrade in the APT was compacted in lifts, each up to 6 in. (152 mm)
thick, using vibratory plate equipment. The average compaction densities for each layer
are given in Llenin and Pellinen (2004). Referring to the top 19 in. (483 mm) of the
subgrade, the average as-built wet density was 126.9 and 128.0 pcf (2035 and 2052
kg/m³) for sections n1 and n2 respectively. The corresponding average moisture
contents were 14.7% and 14.3%. Hence, the following average dry densities were
obtained: 110.6 and 112.0 pcf (1773 and 1796 kg/m³). Since the material was similar to
the NCAT subgrade with a maximum laboratory dry density (Proctor modified) of
119.6 pcf (1918 kg/m³), then the relative compaction degree achieved in the APT was
92.5% and 93.6% for sections n1 and n2 respectively (an average of 93% is shown in
Figure 3.2.1). The lower portion of the subgrade was compacted to an average wet
density of 123.8 pcf (1985 kg/m³) and water content of 14.7%. The corresponding dry
density was therefore 107.9 pcf (1730 kg/m³) which represents a compaction degree of
90% (as shown in Figure 3.2.1).
The aggregate base was compacted in a single 6 in. (152.4 mm) lift. The average
as-built wet density for this layer was 139.6 and 136.7 pcf (2238 and 2192 kg/m³) for
sections n1 and n2 respectively. The corresponding average moisture contents were
3.5% and 3.6%. Hence, the following average dry densities were obtained: 134.9 and
131.9 pcf (2163 and 2115 kg/m³). Since the material was similar to the NCAT base,
having a maximum laboratory dry density (Proctor modified) of 137.9 pcf (2211
kg/m³), then the relative compaction degree achieved in the APT was 97.8% and 95.6%
for sections n1 and n2 respectively (an average of 97% is shown in Figure 3.2.1).
The HMA mixes in the APT had similar composition to the corresponding
NCAT mixes. The mixes were produced by a local contractor in Indiana using NCAT
aggregates. The HMA course in each lane was made of two mixes and constructed in
three lifts. With reference to Figure 3.2.1, the surface lifts (mixes 1 and 3), were 1.0 in.
(25.4 mm) thick while the intermediate and bottom lifts (mixes 2 and 4), were 4.0 in.
(101.6 mm) thick constructed in two 2 in. (50.8 mm) lifts. The average as-constructed
air void content for the three lifts in Section n1 was as follows (top to bottom): 7.5%,
10.2% and 9.1%. The average as-constructed air void content for the three lifts in
3-6
Section n2 was: 7.3%, 9.7% and 9.4%. Recall from Chapter 2 (Subsection 2.2.4) that at
NCAT the corresponding average void contents were about 6 to 7% for all lifts.
3.3 MECHANICAL TESTING AND PRELIMINARY ANALYSIS
3.3.1 Resilient Modulus of Unbound Materials
The accepted mathematical expression for representing resilient modulus test results,
unlike equation 2.3.1, is (Uzan, 1985; 1992; Witczak and Uzan, 1988):
⎛θ
M R = (k1 ⋅ Pa ) ⋅ ⎜⎜
⎝ Pa
k2
k3
⎞
⎞ ⎛ τ oct
⎟⎟ ⋅ ⎜⎜
+ 1⎟⎟ ............................................................... (3.3.1)
⎠
⎠ ⎝ Pa
where Pa is atmospheric pressure, θ = σ 1 + σ 2 + σ 3 is the bulk stress, τ oct is the
octahedral
shear
stress
defined
by
the
expression
2
9 ⋅ τ oct
= (σ 1 − σ 2 ) 2 + (σ 1 − σ 3 ) 2 + (σ 2 − σ 3 ) 2 . Peak applied (total) principal stresses in
the triaxial apparatus are dented as σ 1 , σ 2 and σ 3 ; these are related to the AASHTO
T307 terminology as follows: σ 1 = S 3 + S c + S s and σ 2 = σ 3 = S 3 in which S 3 in the
confining pressure, S c is the peak cyclic stress and S s is the uniaxial static or seating
load. S s is relatively small, with values of about 0.1 ⋅ S c . The three model parameters:
k1 , k 2 and k 3 (unitless) represent the specific material within the range of applied
stresses and at given moisture ( ω ) and density levels.
In typical pavement applications, compacted unbound materials are unsaturated
with the moisture phase in tension (i.e., negative pore pressures). In this case it was
found advantageous to modify equation 3.3.1 such that it includes the influence of
suction on the modulus. Assuming that increased suction has similar influence as
increased confining stresses, the resulting modified equation is (see also Lytton et al.,
1993; Andrei et al., 2004):
k2
k3
⎞
⎛ θ + k6 ⎞ ⎛ τ oct
⎟⎟ ⋅ ⎜⎜
M R = (k1 ⋅ Pa ) ⋅ ⎜⎜
+ 1⎟⎟ ........................................................... (3.3.2)
⎠
⎝ Pa ⎠ ⎝ Pa
3-7
in which the additional (positive) parameter k 6 , having units of stress, represents
suction effects. If we further assume that all moisture sensitivity is lumped into k 6 , i.e.,
k 6 = k 6 (ω ) , then the remaining three parameters in equation 3.3.2 (namely: k1 , k 2 and
k 3 ) are independent of moisture content (all four parameters remain density dependent).
The parameter k 6 should attain a value of zero whenever there is no moisture present in
the material and also when the moisture levels are high and the suction has negligible
effect on the modulus. Between these two extremes k 6 will arrive at some maximum
(positive) value.
Equation 3.3.2 was applied to analyze the resilient modulus test data for the
subgrade and base materials presented in Chapter 2. For this purpose, the numerical
values of the parameters were manipulated by a nonlinear optimization algorithm until a
best fit was achieved between the model projections and the test results. The goodness
of fit was defined based on absolute relative errors. Three data sets were analyzed,
namely: subgrade soil compacted to 96% (see Table 2.3.1), aggregate base compacted
to 93% (Table 2.3.2) and aggregate base compacted to 97.5%. In the latter case, a
synthetic set of test data was generated using equation 2.3.1 with three levels of
confining pressure (5, 10 and 15 psi or equivalently 34.5, 68.9 and 103.4 kPa) and five
levels of cyclic stress (5, 10, 15, 20 and 25 psi or equivalently 34.5, 68.9, 103.4, 137.9
and 172.4 kPa).
The resulting values of the equation 3.3.2 parameters, in each of the three cases,
are summarized in Table 3.3.1 (with Pa = 14.5 psi). As can be seen, k1 is positive and
equals about 500 in all three cases; k 2 is also positive and ranges between 0.8 and 1.0;
k 3 is negative and equals about -0.75 for the high density cases (i.e., 96% subgrade and
97.5% base). For the low density base k 3 has doubled in value. The suction component
k 6 was found to equal about 5 psi (34.5 kPa) for the subgrade soil at 7.2% moisture; at
moisture levels higher than 9.7% it was found negligible. For the base material at about
5.4% water content k 6 is seen to increase in value with density, from 2 psi at 93%
3-8
compaction to 7.2 psi at 97.5% compaction; at a moisture level of 9.8% k 6 was found
negligible (in the low density case).
The as-constructed subgrade moisture content at NCAT was about 10.5% (see
Chapter 2, Subsection 2.2.2) and about 14.5% in the APT (see Section 3.3). Based on
the above results the suction component in the subgrade should be negligible. The asconstructed base moisture content was 6.5% at NCAT (see Chapter 2, Subsection 2.2.3)
and about 3.5% in the APT (see Section 3.3). Hence, the suction component in the base
can be estimated at around 7 psi (48.3 kPa); this is equivalent to a weight of 86 in. (2.2
m) of base material with a total (wet) density of 140 pcf (2245 kg/m³).
Table 3.3.1: Resilient modulus of unbound materials (calibrated equation 3.3.2
parameters).
Material
Subgrade Soil
Aggregate Base
Compaction, %
k1
k2
k3
k6, psi
96.0
554
0.827
-0.770
4.8 ( ω = 7.2% )
0.0 ( ω = 9.7% )
0.0 ( ω = 20.1% )
93.0
477
0.999
-1.580
2.0 ( ω = 5.3% )
0.0 ( ω = 9.8% )
97.5
527
0.833
-0.715
7.2 ( ω = 5.5% )
The entire set of test data and model forecasts are cross plotted in Figure 3.3.1
(log-log scale). The goodness of fit may be graphically assessed from this figure. As can
be seen all the data points fall very close to the equality line (oblique dashed line). It
may also be seen that the resilient modulus of the subgrade at 96% compaction (square
markers) ranges between 4,500 and 15,000 psi (31 and 103 MPa). The aggregate base at
93% (triangular markers) has modulus values in the range of 4,000 to 10,000 psi (28 to
69 MPa). The resilient modulus of the 97.5% base (circular markers) has a range of
13,000 to 21,000 psi (90 to 145 MPa). It should be noted that, for the tested conditions,
considerable overlap is seen in the modulus ranges of the subgrade and base. This
means that they may exhibit comparable stiffness as part of the pavement system.
3-9
100,000
Subgrade 96%
Resilient Modulus - Model [psi]
Base 93%
Base 97.5%
10,000
Resilient Modulus - Test Data [psi]
1,000
1,000
10,000
100,000
Figure 3.3.1: Resilient modulus of unbound materials - a cross plot of calibrated
equation 3.3.2 values and test data.
3.3.2 HMA Complex Modulus
In this subsection the complex modulus test results presented in Chapter 2 (Tables 2.3.3
to 2.3.6) are analyzed. In general terms the interpretation involves horizontal shifting,
along the frequency axis, of the measured dynamic modulus and phase angle data
obtained at different temperatures. This is done with respect to a pre-selected reference
temperature until two separate but continuous curves are attained. The first is the socalled ‘dynamic modulus master curve’ and the second ‘phase angle master curve’. The
analysis performed herein follows the approach recommended in Levenberg and Shah
(2008). This method is slightly different than the common/usual methods because use is
made of both dynamic modulus and phase angle data, simultaneously, to obtain the
master curves. This approach was chosen herein because it was developed specifically
for asphalt mixtures. A short theoretical background and description of the approach is
provided hereafter.
When a time varying uniaxial stress σ (t ) is applied to a linear viscoelastic solid
at a given test temperature T0 , in the form: σ (t ) = σ 0 ⋅ exp(i ⋅ ω ⋅ t ) with i 2 = −1 , σ 0 as
3-10
the stress amplitude (constant) and ω as the angular frequency (units of radians per
second), the resulting steady state strain response is also sinusoidal. The quotient of
stress and strain in the frequency domain is may be represented by a complex number:
E* = E * ⋅ (cos φ + i ⋅ sin φ ) = E1 + i ⋅ E 2 .......................................................... (3.3.3)
in which E * is the material’s complex modulus, E * is the dynamic modulus and φ
denotes the phase lag by which the strain lags behind the applied stress. These
quantities, although not shown explicitly, are functions of both ω and T0 .
The components of the complex modulus, E1 and E 2 in equation 3.3.3, can be
expressed using one fundamental viscoelastic function known as the relaxation
spectrum h and an additional material constant known as the equilibrium modulus E∞ :
∞
E1 (ω , T0 ) = E ∞ + ∫ h(τ , T0 ) ⋅
0
∞
E 2 (ω , T0 ) = ∫ h(τ , T0 ) ⋅
0
ω 2 ⋅τ 2
⋅ d (ln τ ) ............................................. (3.3.4)
1 + ω 2 ⋅τ 2
ω ⋅τ
⋅ d (ln τ ) ..................................................... (3.3.5)
1 + ω 2 ⋅τ 2
As can be seen, the relaxation spectrum has units of stress and is a function of time τ
and temperature T0 , i.e., h = h(τ , T0 ) . The equilibrium modulus E ∞ is temperature
independent, defined as: E ∞ = lim ω →0 E1 (ω ) = lim t →∞ E (t ) in which E (t ) is the
viscoelastic relaxation modulus (units of stress).
Equations 3.3.4 and 3.3.5 are appropriate for a given constant test or reference
temperature, T0 . The assumption of thermo-rheological simplicity (Schwarzl and
Staverman, 1952) states that these relations can remain applicable for a different
(constant) temperature, T , simply by replacing physical time, τ , with reduced (or
pseudo) time, τ r , defined as: τ r = τ / aT , where aT = aT (T , T0 ) is the so-called timetemperature shift factor which is a unitless function of temperature only. Due to the
reciprocal nature of time and frequency, the reduced angular frequency, ω r , is simply
obtained by ω r = ω ⋅ aT with ω as the applied angular frequency. For a certain class of
3-11
polymers (Plazek, 1996), and for a limited range of temperatures, aT tend to follow the
Williams-Landel-Ferry equation (Williams et al., 1955):
log(aT ) =
− c1 ⋅ (T − T0 )
................................................................................. (3.3.6)
c 2 + (T − T0 )
where c1 and c2 are both positive constants ( c1 is unitless and c2 has units of
temperature). This equation was found applicable to HMA mixtures (e.g., Di Benedetto
et al., 2007).
As suggested in Levenberg and Shah (2008), a mathematical expression for the
relaxation spectrum h(τ , T0 ) of the following form is assumed:
(
)
h(τ ) = a1 ⋅ exp − a 2 ⋅ [ln(τ ) − ln(a3 )]2 ............................................................ (3.3.7)
where a1 , a2 and a3 are all temperature dependent positive constants. The variable a1
has units of modulus, a2 is unitless, and a3 has units of time (similar to τ ). This
equation, along with equations 3.3.4, 3.3.5 and 3.3.6 were used to fit the complex
modulus test data given in Chapter 2 by Tables 2.3.5 to 2.3.8 (separate analysis for each
case). In this process six parameters needed evaluation, namely: a1 , a2 , a3 (equation
3.3.7), E∞ (equation 3.3.4), c1 and c2 (equation 3.3.6). Their numerical value obtained
simultaneously using a nonlinear minimization algorithm where the goodness of fit was
defined based on relative errors. The chosen reference temperature was 15.5ºC. A
summary of their derived values, for each of the four mix types, is given in Table 3.3.2.
The test data and corresponding master curves are shown in Figures 3.3.2 to
3.3.5. In these figures the dynamic modulus is depicted on the left ordinate and the
phase angle on the right ordinate. The abscissa represents reduced frequency f r defined
by the expression: ω r = 2 ⋅ π ⋅ f r . In Figures 3.3.6 and 3.3.7 the master curves are
superimposed for graphical comparison. The four dynamic modulus muster curves are
all plotted in Figure 3.3.6, and the four phase angle master curves are plotted in Figure
3.3.7. It may be seen that mixes 1 and 3, and separately mixes 2 and 4, have very
similar master curves for all practical purposes.
3-12
Table 3.3.2: Complex modulus analysis results for a reference temperature of 15.5ºC
based on the approach in Levenberg and Shah (2008).
Relaxation Modulus Parameters
(equation 3.3.7)
Mix
Equilibrium
Modulus
Time-Temperature
Shifting Parameters
(equation 3.3.6)
a1 , MPa
(ksi)
a 2 ⋅10 3
a3 ⋅10 5 , s
E∞ , MPa
(ksi)
c1
c2 , ºC (ºF)
1
1,983
(287.6)
9.84
3.91
172
(24.9)
26.8
215.7
(420.3)
2
2,421
(351.1)
10.63
8.81
159
(23.1)
35.6
338.2
(640.8)
3
1,822
(264.3)
12.10
17.79
91
(13.2)
44.1
376.4
(709.5)
4
2,223
(322.4)
11.55
16.38
272
(39.5)
34.5
311.9
(593.4)
100,000
45
15.5 ºC
Dynamic Modulus (data)
Phase Angle (data)
Viscoelastic Model
40
Dynamic Modulus [MPa]
30
25
1,000
20
15
100
Phase Angle [degrees]
35
10,000
10
5
Reduced Frequency [Hz]
10
1E-08
1E-06
1E-04
1E-02
1E+00
1E+02
1E+04
1E+06
0
1E+08
Figure 3.3.2: Mix 1 dynamic modulus and phase angle master curves @ 15.5ºC.
3-13
100,000
45
15.5 ºC
Dynamic Modulus (data)
Phase Angle (data)
Viscoelastic Model
40
Dynamic Modulus [MPa]
30
25
1,000
20
15
100
Phase Angle [degrees]
35
10,000
10
5
Reduced Frequency [Hz]
10
1E-08
1E-06
1E-04
1E-02
1E+00
1E+02
1E+04
1E+06
0
1E+08
Figure 3.3.3: Mix 2 dynamic modulus and phase angle master curves @ 15.5ºC.
100,000
45
15.5 ºC
Dynamic Modulus (data)
Phase Angle (data)
Viscoelastic Model
40
Dynamic Modulus [MPa]
30
25
1,000
20
15
100
Phase Angle [degrees]
35
10,000
10
5
Reduced Frequency [Hz]
10
1E-08
1E-06
1E-04
1E-02
1E+00
1E+02
1E+04
1E+06
0
1E+08
Figure 3.3.4: Mix 3 dynamic modulus and phase angle master curves @ 15.5ºC.
3-14
100,000
45
15.5 ºC
Dynamic Modulus (data)
Phase Angle (data)
Viscoelastic Model
40
Dynamic Modulus [MPa]
30
25
1,000
20
15
100
Phase Angle [degrees]
35
10,000
10
5
Reduced Frequency [Hz]
10
1E-08
1E-06
1E-04
1E-02
1E+00
1E+02
1E+04
1E+06
0
1E+08
Figure 3.3.5: Mix 4 dynamic modulus and phase angle master curves @ 15.5ºC.
Dynamic Modulus [MPa]
100000
Mix 1
Mix 2
Mix 3
Mix 4
10000
1000
Reduced Frequency [Hz]
100
1.00E-07
1.00E-05
1.00E-03
1.00E-01
1.00E+01
1.00E+03
1.00E+05
1.00E+07
Figure 3.3.6: Superimposed dynamic modulus master curves @ 15.5ºC for mixes 1 to 4.
3-15
30
Mix 1
Mix 2
Mix 3
Mix 4
Phase Angle [degrees]
25
20
15
10
5
Reduced Frequency [Hz]
0
1.00E-07
1.00E-05
1.00E-03
1.00E-01
1.00E+01
1.00E+03
1.00E+05
1.00E+07
Figure 3.3.7: Superimposed phase angle master curves @ 15.5ºC for mixes 1 to 4.
3.3.3 Falling Weight Deflections
FWD testing was conducted in the APT on June 14, 2004, before passes were applied
(embedded instrumentation was not activated during the test). The FWD loading plate
was 11.81 in. (300 mm) in diameter. A set of nine geophones was used, located at the
following offset distances from the center of the plate: 0, 8, 12, 18, 24, 36, 48, 60 and
72 in. (0, 0.20, 0.30, 0.46, 0.61, 0.91, 1.22, 1.52 and 1.83 m). Testing was performed in
centers of lane 1 and lane 2. Six drop sequences were applied in each lane, with each
drop consisting of three load levels: 65, 85 and 105 psi (0.448, 0.586 and 0.724 MPa).
The pavement surface temperature during the test was 86.5ºF (30.3ºC). The average
peak deflections measured at each load level are shown in Table 3.3.3. These results are
plotted in Figure 3.3.8 in which solid lines represent n1 deflections and dashed lines
represent n2 deflections; the three different marker types represent the three load levels.
As can be seen, when comparing the response between the n1 and n2 sections, the
deflection basins are very similar for offset distances greater than 0.46 m (18 in.).
Closer to the loading plate, the deflections in lane 2 are slightly but consistently larger.
3-16
Table 3.3.3: Peak FWD deflections measured in the center of sections n1 and n2.
Lane
n1
n2
0
Load Level
MPa, (psi)
D0
[μm]
D1
[μm]
D2
[μm]
D3
[μm]
D4
D5
D6
[μm] [μm] [μm]
0.448 (65)
501.9 360.0 268.0 106.8
47.7
28.1
18.8
0.586 (85)
661.3 482.6 362.7 149.2
67.6
39.0
26.4
0.724 (105)
826.3 610.5 459.4 194.9
88.4
51.4
34.6
0.448 (65)
576.4 413.7 298.7 117.3
50.6
29.2
19.5
0.586 (85)
750.1 547.1 395.1 158.2
70.3
39.6
26.3
0.724 (105)
941.4 697.2 501.3 202.4
91.5
51.5
34.0
200
400
600
800
1000
1200
1400
1600
1800
2000
10.0
Deflection [microns]
Offset [mm]
100.0
Section n1
0.448 Mpa
0.586 Mpa
Section n2
0.724 Mpa
1000.0
Figure 3.3.8: Peak FWD deflections measured in the center of sections n1 and n2.
3.4 LOADING HISTORY AND ANALYSIS DATASET
3.4.1 Application of Load Passes
Passes in the APT were applied via a dual wheel assembly loaded to 15,000 lb (6810
kg). Each wheel was equipped with a Goodyear radial Unisteel tire model G159A
(designation 11R22.5) inflated to 100 psi (0.70 MPa). All passes were applied in
unidirectional mode, some without wander but most with wander (indicated by ‘w’ in
the following tables).
3-17
Loading of Section n1 began on July 19, 2004. About 90,000 passes were
applied by August 11, 2004, all without wander. The last 2,500 cycles where applied
overnight, during which a bond failure occurred between the surface and intermediate
HMA lifts. At the onset of failure, the surface lift in the wheel path area was sheared-off
in the direction of loading, exposing the intermediate HMA lift. Subsequently the
loading of lane 1 was stopped, and the APT loading frame was switched to lane 2.
Section n2 incurred about 2,500 passes beginning September 13, 2004, before it became
apparent that another bond failure was rapidly developing, again at the interface
between the surface and intermediate HMA lifts. Subsequently, it was decided not to
wait for complete shear failure, but to mill and repave the surface HMA lift in both
sections. This sequence of events is summarized in Table 3.4.1.
Table 3.4.1: APT pass application log for sections n1 and n2 (original structure).
Date
Section Cumulative Passes
Days
19-Jul-04
0
0
22-Jul-04
5,000
3
23-Jul-04
10,000
4
25-Jul-04
20,000
6
27-Jul-04
30,000
8
29-Jul-04
40,000
10
n1
30-Jul-04
50,000
11
2-Aug-04
60,000
14
6-Aug-04
70,000
18
9-Aug-04
80,000
21
11-Aug-04
90,000
23
Loading stopped due to bond failure between top and intermediate
HMA layers.
13-Sep-04
n2
2,500
1
Signs of bond failure between top and intermediate HMA layers;
surface asphalt layer on both sections was milled and repaved.
Resurfacing of the surface HMA lift took place between September 13 and
September 28, 2004 (no construction details are available). Reloading of section n2
began on the latter date and continued until February 1, 2005, at which point 187,500
passes had been applied. The first 40,000 passes were applied without wander; wheel
3-18
wander was employed for the reminder of the test. Trafficking of section n2 was
discontinued because another bond failure was seen to take place, this time at the
interface between the intermediate and bottom lifts. This sequence of events is shown in
Table 3.4.2.
Table 3.4.2: APT pass application log for Section n2 (rehabilitated structure).
Date
Section Cumulative Passes
Days
28-Sep-04
0
0
6-Oct-04
20,000
8
8-Oct-04
25,000
10
12-Oct-04
30,000
14
15-Oct-04
40,000
17
18-Oct-04
50,000 (w)
20
21-Oct-04
60,000 (w)
23
25-Oct-04
70,000 (w)
27
n2
27-Oct-04
90,000 (w)
29
15-Nov-04
130,000 (w)
48
22-Nov-04
140,000 (w)
55
6-Dec-04
160,000 (w)
69
15-Dec-04
174,000 (w)
78
21-Dec-04
180,000 (w)
84
1-Feb-05
187,500 (w)
126
Testing stopped due to bond failure between the intermediate and
bottom HMA layers.
Water was introduced to the subgrade pit beginning June 27, 2005 (details can
be found in Appendix C). This was accomplished by localized removal of the HMA in
lane 2 (by means of saw cutting and coring), and inundating the openings. On July 14,
2005, a dynamic cone penetration test indicated that the shear strength of the subgrade
had reduced considerably compared to the initial conditions. On August 2, 2005, APT
loadings of the rehabilitated Section n1 were renewed. Between August 2, 2005, and
April 4, 2006, an additional 250,000 passes, all with wander, were applied to the
section. Trafficking was discontinued because of time constraints. Very little cracking
was reportedly seen on the pavement surface by that time. Both structural sections and
the upper part of the subgrade were removed and discarded, along with the embedded
3-19
instrumentation, so that another research project could be installed in the APT. This
sequence of events is shown in Table 3.4.3.
Table 3.4.3: APT pass application log for Section n1 (rehabilitated structure).
Date
Section Cumulative Passes
Days
th
Water introduced to subgrade beginning June 27 2005
2-Aug-05
1,000 (w)
0
15-Aug-05
3,000 (w)
13
16-Aug-05
4,000 (w)
14
17-Aug-05
5,000 (w)
15
18-Aug-05
8,000 (w)
16
22-Aug-05
10,000 (w)
20
31-Aug-05
11,000 (w)
29
1-Sep-05
13,000 (w)
30
6-Sep-05
17,500 (w)
35
7-Sep-05
20,000 (w)
36
26-Sep-05
25,000 (w)
55
3-Oct-05
35,000 (w)
62
24-Oct-05
35,000 (w)
83
1-Nov-05
50,000 (w)
91
2-Nov-05
55,000 (w)
92
7-Nov-05
60,000 (w)
97
14-Nov-05
70,000 (w)
104
N1
15-Nov-05
75,000 (w)
105
22-Nov-05
80,000 (w)
112
28-Nov-05
85,000 (w)
118
29-Nov-05
90,000 (w)
119
5-Dec-05
95,000 (w)
125
6-Dec-05
100,000 (w)
126
12-Dec-05
105,000 (w)
132
13-Dec-05
110,000 (w)
133
19-Dec-05
120,000 (w)
139
3-Jan-06
125,000 (w)
154
9-Jan-06
130,000 (w)
160
16-Jan-06
140,000 (w)
167
23-Jan-06
150,000 (w)
174
24-Jan-06
160,000 (w)
175
6-Feb-06
170,000 (w)
188
13-Feb-06
180,000 (w)
195
27-Feb-06
200,000 (w)
209
4-Apr-06
250,100 (w)
245
Testing stopped; both sections removed and replaced.
3-20
3.4.2 Identification of Dataset for Structural Investigation
In general terms, this research aims at devising a method for applying APT results to
field conditions. As put forward in Chapter 1: (i) the scope is limited to the case of
duplicate pavement systems; (ii) the work plan consists of calibrating a mechanistic
model to APT conditions and extending it using laboratory data; and (iii) the extended
model is to be validated using NCAT results. Accordingly, it is argued that Section n1
dataset, collected in the APT between July and August 2004 (see Table 3.4.1), is best
suited for carrying out structural investigation and achieving the main study objective.
First and foremost, there is maximum similarity between this section and its replicate at
NCAT, especially in the initial part of the experiment when both pavements were in
their pristine state. This similarity was severely ‘damaged’ during the second round of
n1 testing that took place between August 2005 and April 2006 (see Table 3.4.3),
mainly because of the ‘artificial’ subgrade weakening and also because n1 became a
rehabilitated structure: the surface HMA was replaced after incurring about 90,000 load
passes and then the structure was allowed to rest/heal without traffic for almost a year
before more loads were applied. Second, in this dataset APT passes were applied
without wander, which means that the exact carriage position relative to the embedded
gauges is known and available. This information is critical for model calibration.
Finally, the construction operations are well documented and supplemented by
laboratory tests.
This state of affairs is not the case for Section n2. First, the original surface
HMA lift was replaced but the construction data and properties of the new mix are not
available. Second, loading of n2 between September 2004 and February 2005 included
wheel wander for which the exact carriage position was recorded only in the loading
direction but not laterally. When the exact position of the loading is unknown, the
approach followed herein fails because the APT model cannot be calibrated using
inverse analysis. Finally, the main difference between sections n1 and n2 (or
equivalently between N1 and N2) is the binder type. However, the master curves for
n2/N2 mixes 3 and 4 were very similar to the n1/N1 master curves for mixes 1 and 2
(respectively). Hence, the resilient response of the two sections should also be very
similar. In this connection, see also the FWD results in Figure 3.3.8. Moreover, as will
3-21
be shown in the following section, large differences were recorded between gauges
installed in Section n1 that were expected to measure identical responses. These
differences are assumed to originate from structural heterogeneity and slight
dissimilarity in gauge installation conditions. Consequently, any inherent dissimilarity
in the response to load of the two sections is masked by these differences.
In summary, and based on the aforementioned sequence of events, the
experimental dataset obtained during loading of Section n1 in the APT between July
and August 2004 is selected for pursual of the study objectives. Although in Chapter 4
this dataset will be analyzed twice, for the pavement in its initial condition and also
after 80,000 passes, the forecasting of NCAT response will focus on the initial loading
phases during which the two experiments were most closely linked.
3.5 STRUCTURAL BEHAVIOR
3.5.1 Instrumentation
Instrumentation types and placement techniques used in the APT study were similar to
these in the NCAT study (see Chapter 2, section 2.4). Pressure cells (Model 3500
manufactured by Geokon) were used as vertical stress gauges; CTL Group gauges
(Model ASG-152) were used for measuring horizontal strains at the bottom of the
HMA. These gauges were checked for functionality before and after embedment. The
achieved level of survivability was 100%. For further details refer to the reports
included in Appendix C.
From a mechanical point of view, the introduction of a gauge in a pavement
system produces changes to the stress and strain fields which influence the response of
the pavement in the vicinity of the gauge and hence influence the recorded values. This
disturbance was mostly investigated for the case of pressure cells (e.g., Tory and
Sparrow, 1967; Brown, 1977; Tabatabaee and Sebaaly, 1990). Numerous factors have
been identified that affect the measurements of pressure cells, including the ratio of cell
thickness to diameter, the ratio of medium stiffness to cell stiffness, cell size, and field
placement effects (Weiler and Kulhawy, 1982; Dunnicliff, 1988).
3-22
Early synthetic work by Taylor (1945) and Monfore (1950) have shown that
measurement errors can be reduced by minimizing the thickness to diameter ratio of the
cell and making it as incompressible as possible. Based on experimental work, Peattie
and Sparrow (1954) have shown that if these criteria are fulfilled then the measurement
error relative to the ‘true’ stress level (in percent) equals 0.6 times the thickness to
diameter ratio. Geokon earth pressure cells are classified as hydraulic type gauges; by
design they have a thickness to diameter ratio of 0.026 (=6/230 see Figure 2.4.2) and are
relatively incompressible. According to Peattie and Sparrow (1954), they are expected
to record pressures that are higher than the ‘true’ stress levels by about 1.6%. This is a
relatively low error level for a geotechnical application.
More recently calibration chambers have been proposed as means for pressure
cell calibration (Theroux et al., 2001; Labuz and Theroux, 2005). For pavement
applications a more feasible method of calibration would be to apply a known load to
the pavement at increasing distances from the gauge and calculate the resulting
‘volume’ of stresses. In theory this ‘volume’ should equal the applied load. This in situ
type of calibration is best done with a single tire since the shape of the stress trace is
symmetric and thus only a few measurement points need to be considered. In the
current study dual-tires were used which resulted in an asymmetric stress trace. Also,
the location of the tires relative to the gauge array was only measured in the
longitudinal direction and not laterally. Hence it was not possible to perform such
calibration herein. Similar reasoning precludes in situ calibration of the pressure cells
for the NCAT study. With respect to the strain gauges, analysis of near field strain
disturbance and resulting measurement errors could not be found in the literature.
Subsequently, the stress and strain gauge readings were used as-is without applying any
correction or calibration factors. In addition, it should be mentioned that both types of
gauges experience drift over time due to temperature sensitivity (see Tesarik et al.,
2006) and other reasons; this makes them suitable for capturing dynamic responses only
and not for monitoring permanent changes.
A plan showing the embedded instrumentation aimed at capturing mechanical
responses in test section n1 is provided in Figure 3.5.1. The loading centerline is
denoted in the figure by the Y-axis and the transverse direction by the X-axis. The
3-23
loading direction was from left to right along the Y-axis as indicated by the arrow. The
entire gauge array is seen to be located in an eight foot (2.44 m) long strip, 2 ft (0.61 m)
wide, in the central part of the test section. The first and last 6 ft (1.83 m) of the test
section were not instrumented because the loading speed in these zones is not constant,
with the carriage either accelerating or decelerating. In the central strip the loads are
applied at a constant speed, which, in this study, was always 5 mph (~2.2 m/s). With
respect to the X and Y axes in Figure 3.5.1, Table 3.5.1 lists the location of each gauge.
Figure 3.5.1: Plan of embedded instrumentation in APT lane 1 (Section n1).
As can be seen in Figure 3.5.1 (and Table 3.5.1), the pavement system was
instrumented with a total of 12 gauges consisting of four pressure gauges and eight
strain gauges. The two pressure gauges (#1178 and #1185) were measuring vertical
stresses on top of the base course or bottom of the HMA. These were installed at a
depth of 5 in. (127 mm) from the surface along the centerline of the loading path (i.e.,
Y-axis). Two additional pressure gauges (#1179 and #1184) were measuring vertical
stresses on top of the subgrade or bottom of the base course. These were installed at a
depth of 11.0 in. (279.4 mm) from the surface, also along the centerline.
All eight strain gauges were attached to the bottom of the HMA, i.e., at a depth
of 5 in. (127 mm) from the surface. Gauges G-1, G-2, G-3 and G-4 were located along
3-24
the centerline of the loading path. Strain gauges G-5, G-6, G-7 and G-8 were located
along a parallel line positioned two feet (0.61 m) from the loading path. Gauges G-2, G4, G-5 and G-7 were measuring horizontal strains in the loading direction (i.e., strain in
Y) while gauges G-1, G-3, G-6 and G-8 were measuring horizontal strains in the
transverse direction (i.e., strain in X).
Table 3.5.1: Location of APT instrumentation in Section n1 (relate to Figure 3.5.1).
Location in X,
in. (m)
Pressure
Cell
#1178
# 1179
#1184
#1185
G-1
G-2
G-3
G-4
G-5
G-6
G-7
G-8
Gauge Type
0.0
Strain Gauge
Gauge ID
24 (6.1)
Location in Y,
in. (m)
Depth in Z,
in. (mm)
72 (1.83)
96 (2.44)
144 (3.66)
168 (4.27)
84 (2.13)
108 (2.74)
132 (3.35)
156 (3.96)
84 (2.13)
108 (2.74)
132 (3.35)
156 (3.96)
5 (127)
11 (279)
5 (127)
3.5.2 Resilient Response
The resilient response data presented and discussed hereafter was obtained from Section
n1 during testing that took place between July and August 2004 (see Table 3.4.1 and
discussion in Subsection 3.4.2). Recall that the loading was stopped after 90,000 passes
due to bond failure that occurred at the interface between the surface and intermediate
HMA lifts. Figures 3.5.2, 3.5.3 and 3.5.4 show the resilient strains and stresses
measured during APT passes 5,000 and 80,000 vs. the APT carriage location which
corresponds to the Y-axis in Figure 3.5.1. In each of these figures, a solid line
represents pass #5,000 and a dashed line represents pass #80,000. As is customary for
geomaterials, a positive sign indicates compression and a negative sign indicates tension
of either stress or strain.
3-25
Figure 3.5.2 presents the vertical stresses measured on top of the subgrade and
on top of the base course by the four pressure gauges. As can be seen, the resulting
curves are bell-shaped and nearly symmetric. For loading pass #5,000, peak vertical
stresses on top of the base course were 30 and 35 psi (0.21 and 0.24 MPa). On top of the
subgrade, the measured peak stresses were 16 and 20 psi (0.11 and 0.14 MPa). In theory
the readings of each gauge pair should be identical. Furthermore, it may be seen that
peak vertical stresses during pass #80,000 are slightly higher compared to pass #5,000.
The difference is more significant in both absolute and relative terms for the gauges
located on top of the subgrade compared to those located on top of the base course. It
should be noted that the stress peaks occur slightly after the APT carriage had passed
over the gauges and moved further along by about 2 to 5 in. (51 to 127 mm).
40
Pass #5,000
Pass #80,000
Loading Direction
36
Vertical Stress [psi]
32
28
24
20
16
#1184, top of
subgrade
12
#1178,
top of base
8
#1185,
top of base
#1179, top of
subgrade
4
0
0
40
80
120
160
200
240
APT Carriage Location [in.]
Figure 3.5.2: Measured vertical stresses in Section n1 on top of the base and on top of
the subgrade during pass #5,000 (solid line) and pass #80,000 (dashed line).
Figure 3.5.3 presents the measured horizontal strains at the bottom of the HMA
course in the direction of loading. Four gauge readings are shown, two of which were
located along the loading centerline (G-2 and G-4), and two were located along a
parallel line (G-5 and G-7) that is offset by two feet (0.61 m); see Figure 3.5.1 and
Table 3.5.1. In all four cases it can be seen that as the load approaches a gauge, the
3-26
bottom of the HMA goes into compression. Then, the strain direction is reversed and
the gauges go into tension. The point of maximum tension occurs when the APT
carriage has passed the gauge positions along the Y-axis by about 1 to 3 in. (25 to 76
mm). Finally, when the load is receding (APT carriage moves further along), the tensile
strains are reversed and compression is induced once more at the bottom of the HMA.
This pattern is more pronounced for the gauges aligned along the centerline (G-2 and G4).
150
Loading Direction
Bottom of HMA Strain (in Y) [microstrains]
100
50
0
G-5,
2ft Off-Center
-50
G-4,
Centerline
-100
-150
G-7,
2ft Off-Center
G-2,
Centerline
-200
-250
-300
-350
Pass #5,000
Pass #80,000
-400
APT Carriage Location [in.]
-450
0
40
80
120
160
200
240
Figure 3.5.3 Measured horizontal strains at the bottom of the HMA in the direction of
loading during pass #5,000 (solid line) and pass #80,000 (dashed line).
It can be graphically seen that the approaching branch of the strain response is
different from the receding branch, resulting in a non-symmetrical time history curve.
The two most noticeable differences are: (i) peak compressive strain is usually higher in
the approaching branch compared to the receding branch; and (ii) the spacing along the
Y-axis between the tension and compression strain peaks is larger in the receding
branch compared to the approaching curve.
Referring to the approaching branch of pass #5,000 for gauges G-2 and G-4
(both centerline gauges), peak strains were 84 and 119 microstrains in compression and
3-27
431 and 314 microstrains in tension (respectively). For gauges G-5 and G-7 (both offcenter gauges) the peak compressive stains were 2 and 15 microstrains while the peak
tensile strains were 50 and 100 microstrains (respectively). In theory, the readings from
each gauge pair should be identical. When comparing the response between pass #5,000
and pass #80,000 the most noticeable difference is seen in the peak tensile strain
magnitudes for gauges G-2 and G-4 (the centerline gauges). For the G-2 strain gauge,
peak strain in tension during pass #80,000 is 380 microstrains (compared to 431
microstrains during pass #5,000). For G-4 gauge the peak strain in tension during pass
#80,000 is 163 microstrains (compared to 314 microstrains during pass #5,000).
Figure 3.5.4 presents the measured horizontal strains at the bottom of the HMA
course in the transverse direction relative to the loading centerline. In this case the
centerline gauges behave differently compared to the off-center gauges. Referring to
pass #5,000 data, it can be seen that gauges G-1 and G-3 (centerline gauges) go into
tension as the load is approaching, with peak strains of 108 and 154 microstrains
respectively. Contrary to the previous two figures, these peaks occur 4 to 6 in. (102 to
152 mm) before the APT carriage reaches the gauge. As the APT carriage passes the
gauges and moves further along, the strain direction is reversed until a small level of
compression is induced. This compressive strain slowly recovers during the time period
(not shown in the figure) in which the APT load is lifted from pavement and moved
back to the startup position. Gauges G-6 and G-8 (off-center gauges) go into
compression as the APT carriage approaches, with peak strains of 103 and 152
microstrains respectively. These peaks, however, occur 3 to 5 in. (76 to 127 mm) after
the load had passed each gauge. The receding branch of the response shows that the
strain direction is reversed until a small level of tension is induced in the gauges.
The response of the gauges, as seen in Figure 3.5.4, is very confusing. First, the
G-3 gauge shows two peaks instead of one as seen in the rest of the gauges. Next, when
comparing the response between pass #5,000 and pass #80,000 the trends are not
uniform: (i) it seems that the response of the G-3 gauge has shifted (delayed), as if the
gauge was physically moved a few inches along the Y-axis during the experiment; and
(ii) the peak strains decreased during the test for gauges G-1, G-3 and G-6 but not for
3-28
the G-8 gauge. At this point we do not have a good explanation for these obscure
behaviors.
200
Bottom of HMA Strain (in X) [microstrains]
Loading Direction
150
G-8,
2ft Off-Center
100
G-6,
2ft Off-Center
50
0
-50
G-1,
Centerline
G-3,
Centerline
-100
Pass #5,000
Pass #80,000
-150
APT Carriage Location [in.]
-200
0
40
80
120
160
200
240
Figure 3.5.4: Measured horizontal strains at the bottom of the HMA in the transverse
direction to the loading during pass #5,000 (solid line) and pass #80,000 (dashed line).
3.5.3 Rutting and Cracking Performance
Periodically, the APT carriage was halted for mapping of surface cracks and so that
rutting measurements could take place. No cracking information is available from the
APT study as cracks did not appear on the surface throughout the experiment.
Reportedly, a very few hairline cracks did appear at the very end of the experiment in
Section n1 (see Table 3.4.3). Their location and orientation, however, was not mapped
before digging out the materials to make room for the following APT study.
The device for obtaining transverse profiles consists of a vertical rod with a
small wheel attached to its tip. The rod is moved manually along a guided straight line
across the pavement, with its rolling wheel in continuous contact with the surface.
During this process both vertical and horizontal movements are collected providing
about 680 data pairs. Horizontal position is measured with a cable-based transducer
while the vertical position is measured using an LVDT. The entire device is attached to
3-29
the overhanging APT beams that span the test pit thus providing a fixed reference for
the measurements throughout the experiment. In an effort to assess the accuracy of the
profiler, a nominally flat concrete surface was measured repeatedly along the same line.
The standard deviation of readings was found to be 0.015 in. (0.38 mm). Referring to a
single point on the pavement, the maximum difference between two individual profile
readings taken at different times was found to be 0.069 in. (1.75 mm). On an average
this difference was 0.048 in. (1.22 mm). More details on the profiler can be found in
Huang (1995) and in Galal and White (1999). Table 3.4.1 shows when rutting profiles
were taken in Section n1. As can be seen, the dates in the table correspond to the
following cumulative number of APT passes (without wander): 0, 5k, 10k, 20k, 30k,
40k, 50k, 60k, 70k, 80k and 90k. In each case nine cross sections were determined,
spaced 2 ft (0.61 m) apart. Corresponding to the Y-axis in Figure 3.5.1, their locations
were: 24, 48, 72, 120, 144, 168, 192, 216, and 240 in. (0.61, 1.22, 1.83, 2.44, 3.05, 3.66,
4.27, 4.88 and 5.49 m). The complete profile data set can be found in Appendix A.
For illustration purposes, the profiles measured at the central cross section with
Y=120 in. (2.44 m) are shown in Figure 3.5.5. This figure shows 12 profiles relative to
the initial profile. The magnitude of surface depression is shown on the ordinate on the
left. The dashed horizontal line at zero rutting represents the pavement immediately
before testing. The abscissa indicates the offset in the X direction relative to the loading
centerline (refer to Figure 3.5.1). The results for passes 100, 500 and 1000 were
obtained using interpolation assuming that the rutting increased linearly when APT
passes #1 to #5000 are depicted on a logarithmic scale.
As can be seen in Figure 3.5.5, during the application of the first 60,000 passes
the vertical surface displacement is seen to continuously increase. In absolute terms, a
maximum surface depression of 0.45 in. (11.4 mm) was reached directly under the tires.
With additional APT passes, this maximum is slightly reduced while surface heaving
takes place outside the wheel path. First on the right side at about pass #70,000 and then
also on the left side during pass # 90,000.
3-30
0.45
0.40
100 Passes
1000 Passes
10000 Passes
50000 Passes
70000 Passes
0.35
0.30
0.25
Surface Rutting [in.]
0.20
500 Passes
5000 Passes
20000 Passes
60000 Passes
90000 Passes
0.15
0.10
0.05
0.00
-0.05
-0.10
-0.15
-0.20
-0.25
-0.30
-0.35
-0.40
-0.45
Offset in X from Loading Center Line [in.]
-0.50
-30
-25
-20
-15
-10
-5
0
5
10
15
20
25
30
Figure 3.5.5: Rutting development in Section n1 at the central cross section during the
first 90,000 load passes (applied without wheel wander).
Figures 3.5.6 to 3.5.18 contain contour charts (13 charts in total) that show the
development of rutting in the APT. Each figure is basically a plan view of lane 1 with X
and Y axes as defined in Figure 3.5.1. The different colors (or shades) in the charts
represent different rutting depths (legend is identical in all figures). In preparing these
figures, use was made of all available profiles measured during the entire experiment;
each figure represents a different pass level, shown in a box on the lower right corner;
starting with pass #100 (Figure 3.5.6) and ending at pass #90,000 (Figure 17). The
outcome helps visualize how rutting progressed under load. One immediate observation
from these charts is that rutting was consistently deeper in the first half of the test lane
(i.e., Y<144 in.). This indicates structural heterogeneity that may, at least partially,
explain the dissimilarity in readings from gauge pairs that should, in theory, be
measuring identical response (see discussion in previous subsection).
3-31
-0.65--0.50
-0.50--0.35
-0.35--0.20
-0.20--0.05
-0.05-0.10
0.10-0.25
0.25-0.40
0.40-0.55
-30
-25
-20
-10
-5
0
5
10
15
Lateral Offset in X
-15
20
Pass #100
24
30
36
42
48
54
60
66
72
78
84
90
25
30
96 102 108 114 120 126 132 138 144 150 156 162 168 174 180 186 192 198 204 210 216
Location in Y [in.]
Figure 3.5.6: Contour plot of Section n1 rutting after 100 passes.
-0.65--0.50
-0.50--0.35
-0.35--0.20
-0.20--0.05
-0.05-0.10
0.10-0.25
0.25-0.40
0.40-0.55
-30
-25
-20
-10
-5
0
5
10
15
Lateral Offset in X
-15
20
Pass #500
24
30
36
42
48
54
60
66
72
78
84
90
25
30
96 102 108 114 120 126 132 138 144 150 156 162 168 174 180 186 192 198 204 210 216
Location in Y [in.]
Figure 3.5.7: Contour plot of Section n1 rutting after 500 passes.
3-32
-0.65--0.50
-0.50--0.35
-0.35--0.20
-0.20--0.05
-0.05-0.10
0.10-0.25
0.25-0.40
0.40-0.55
-30
-25
-20
-10
-5
0
5
10
15
Lateral Offset in X
-15
20
Pass #1,000
24
30
36
42
48
54
60
66
72
78
84
90
25
30
96 102 108 114 120 126 132 138 144 150 156 162 168 174 180 186 192 198 204 210 216
Location in Y [in.]
Figure 3.5.8: Contour plot of Section n1 rutting after 1,000 passes.
-0.65--0.50
-0.50--0.35
-0.35--0.20
-0.20--0.05
-0.05-0.10
0.10-0.25
0.25-0.40
0.40-0.55
-30
-25
-20
-10
-5
0
5
10
15
Lateral Offset in X
-15
20
Pass #5,000
24
30
36
42
48
54
60
66
72
78
84
90
25
30
96 102 108 114 120 126 132 138 144 150 156 162 168 174 180 186 192 198 204 210 216
Location in Y [in.]
Figure 3.5.9: Contour plot of Section n1 rutting after 5,000 passes.
3-33
-0.65--0.50
-0.50--0.35
-0.35--0.20
-0.20--0.05
-0.05-0.10
0.10-0.25
0.25-0.40
0.40-0.55
-30
-25
-20
-10
-5
0
5
10
15
Lateral Offset in X
-15
20
Pass #10,000
24
30
36
42
48
54
60
66
72
78
84
90
25
30
96 102 108 114 120 126 132 138 144 150 156 162 168 174 180 186 192 198 204 210 216
Location in Y [in.]
Figure 3.5.10: Contour plot of Section n1 rutting after 10,000 passes.
-0.65--0.50
-0.50--0.35
-0.35--0.20
-0.20--0.05
-0.05-0.10
0.10-0.25
0.25-0.40
0.40-0.55
-30
-25
-20
-10
-5
0
5
10
15
Lateral Offset in X
-15
20
Pass #20,000
24
30
36
42
48
54
60
66
72
78
84
90
25
30
96 102 108 114 120 126 132 138 144 150 156 162 168 174 180 186 192 198 204 210 216
Location in Y [in.]
Figure 3.5.11: Contour plot of Section n1 rutting after 20,000 passes.
3-34
-0.65--0.50
-0.50--0.35
-0.35--0.20
-0.20--0.05
-0.05-0.10
0.10-0.25
0.25-0.40
0.40-0.55
-30
-25
-20
-10
-5
0
5
10
15
Lateral Offset in X
-15
20
Pass #30,000
24
30
36
42
48
54
60
66
72
78
84
90
25
30
96 102 108 114 120 126 132 138 144 150 156 162 168 174 180 186 192 198 204 210 216
Location in Y [in.]
Figure 3.5.12: Contour plot of Section n1 rutting after 30,000 passes.
-0.65--0.50
-0.50--0.35
-0.35--0.20
-0.20--0.05
-0.05-0.10
0.10-0.25
0.25-0.40
0.40-0.55
-30
-25
-20
-10
-5
0
5
10
15
Lateral Offset in X
-15
20
Pass #40,000
24
30
36
42
48
54
60
66
72
78
84
90
25
30
96 102 108 114 120 126 132 138 144 150 156 162 168 174 180 186 192 198 204 210 216
Location in Y [in.]
Figure 3.5.13: Contour plot of Section n1 rutting after 40,000 passes.
3-35
-0.65--0.50
-0.50--0.35
-0.35--0.20
-0.20--0.05
-0.05-0.10
0.10-0.25
0.25-0.40
0.40-0.55
-30
-25
-20
-10
-5
0
5
10
15
Lateral Offset in X
-15
20
Pass #50,000
24
30
36
42
48
54
60
66
72
78
84
90
25
30
96 102 108 114 120 126 132 138 144 150 156 162 168 174 180 186 192 198 204 210 216
Location in Y [in.]
Figure 3.5.14: Contour plot of Section n1 rutting after 50,000 passes.
-0.65--0.50
-0.50--0.35
-0.35--0.20
-0.20--0.05
-0.05-0.10
0.10-0.25
0.25-0.40
0.40-0.55
-30
-25
-20
-10
-5
0
5
10
15
Lateral Offset in X
-15
20
Pass #60,000
24
30
36
42
48
54
60
66
72
78
84
90
25
30
96 102 108 114 120 126 132 138 144 150 156 162 168 174 180 186 192 198 204 210 216
Location in Y [in.]
Figure 3.5.15: Contour plot of Section n1 rutting after 60,000 passes.
3-36
-0.65--0.50
-0.50--0.35
-0.35--0.20
-0.20--0.05
-0.05-0.10
0.10-0.25
0.25-0.40
0.40-0.55
-30
-25
-20
-10
-5
0
5
10
15
Lateral Offset in X
-15
20
Pass #70,000
24
30
36
42
48
54
60
66
72
78
84
90
25
30
96 102 108 114 120 126 132 138 144 150 156 162 168 174 180 186 192 198 204 210 216
Location in Y [in.]
Figure 3.5.16: Contour plot of Section n1 rutting after 70,000 passes.
-0.65--0.50
-0.50--0.35
-0.35--0.20
-0.20--0.05
-0.05-0.10
0.10-0.25
0.25-0.40
0.40-0.55
-30
-25
-20
-10
-5
0
5
10
15
Lateral Offset in X
-15
20
Pass #80,000
24
30
36
42
48
54
60
66
72
78
84
90
25
30
96 102 108 114 120 126 132 138 144 150 156 162 168 174 180 186 192 198 204 210 216
Location in Y [in.]
Figure 3.5.17: Contour plot of Section n1 rutting after 80,000 passes.
3-37
-0.65--0.50
-0.50--0.35
-0.35--0.20
-0.20--0.05
-0.05-0.10
0.10-0.25
0.25-0.40
0.40-0.55
-30
-25
-20
-10
-5
0
5
10
15
Lateral Offset in X
-15
20
Pass #90,000
24
30
36
42
48
54
60
66
72
78
84
90
25
30
96 102 108 114 120 126 132 138 144 150 156 162 168 174 180 186 192 198 204 210 216
Location in Y [in.]
Figure 3.5.18: Contour plot of Section n1 rutting after 90,000 passes.
3.6 DIRECT COMPARISON WITH NCAT RESULTS
Assuming similar pavement systems were built at NCAT and in the APT, there should
be a way to relate the two experiments. The simplest and most direct method of analysis
is to compare the observed behavior by contrasting (separately) the observed
performance and response.
In the NCAT study the N1 (and N2) pavement failed predominantly in fatigue
mode (see Figure 2.5.5) while very little rutting was observed. The maximum recorded
rut depth (Figure 2.5.6) was 0.31 in. (8 mm). In contrast, in the APT study very little (if
any) cracking occurred, and the n1 pavement experienced under the first 90,000 passes
alone (see Table 3.4.1) maximum rutting of about 0.65 in. (11.5 mm) and heaving of
similar magnitude (see Figure 3.5.18).
NCAT responses to one truck pass were shown in Figures 2.5.1 to 2.5.4. Each
figure includes strain or stress pulses recorded by one gauge. As can be seen, one pass
induced eight pulses with different peak magnitudes even for nominally identical axle
loads. Also, the pavement was allowed to rest for 45 seconds between truck passes
(assuming three running trucks) and for about two full days during weekends. APT
3-38
responses were shown in Figures 3.5.2 to 3.5.4. In contrast to the NCAT case, each
APT pass resulted in one stress (or strain) pulse; also, the pavement was allowed to rest
for 8 seconds between passes. It should be noted that the APT was operated on
weekends also. Graphically, the response traces from the two experiments are very
much different and any attempt to directly compare them is futile given that they
embody many dissimilarities:
(i) Loading Speed: 45 mph at NCAT vs. 5 mph in the APT. The loading speed
influences the duration of stress and stress pulses and affects the HMA stiffness;
(ii) Axle Configuration: NCAT had three axle types, single, dual and dual
tandem vs. one axle type in the APT, dual;
(iii) Axle Load: NCAT axles were loaded to 20,000 pounds except for the steer
axle which was loaded to 10,000 pounds vs. one load level in the APT of 15,000
pounds;
(iv) Location of Loads relative to the Measuring Gauge: NCAT loads were
applied near the gauge array but the exact location is unknown while in the APT study
the load could be positioned accurately;
(v) HMA Temperature: NCAT temperature changed with the environment while
it was constant in the APT at 15.5ºC. The prevailing temperature influences the stiffness
of the HMA and hence the resulting responses.
Based on the above discussion it may be concluded that a direct relation
between the two experiments cannot be established as both performance and response
are distinct. Consequently a more fundamental approach is needed to link the observed
behavior. This is pursued in the following chapters using mechanistic principals
assuming the pavement systems in the two experiments have similar material properties.
Mechanistic treatment can account for each of the aforementioned dissimilarities. As
put forward and explained in Chapter 1 (Section 1.2) the analysis will focus on resilient
responses.
3-39
CHAPTER 4 - BASIC MECHANISTIC ANALYSIS
This chapter addresses the primary study objective by establishing a relation between
the NCAT and APT experiments. The scope and approach are discussed in Section 4.1.
Section 4.2 contains the development of a basic mechanistic model for the pavement
systems considered; it is based on layered elastic theory (LET) with the necessary
material properties obtained via inverse analysis of APT results. Section 4.3 deals with
NCAT response prediction; the calculation methodology is first explained and then
applied to forecast selected responses.
4.1 SCOPE AND APPROACH
In Section 3.6 it was argued that a fundamental approach is needed to link the NCAT
and APT experiments. This is pursued hereafter focusing on load related resilient
responses using isotropic LET. The APT experiment is first analyzed. The n1 pavement
system is modeled using four isotropic layers (see Subsection 4.2.1) comprised of
HMA, base and subgrade on top of a semi-infinite concrete medium. The latter
represents the concrete floor present at the bottom of the APT pit. The unknown
material properties (layer moduli) are obtained through backcalculation by matching the
gauge readings collected during one pass of the APT carriage (see Subsection 4.2.2).
Due to the temperature and rate sensitivity of the HMA, the resulting properties
represent the environment and loading configuration in the APT only. This ‘inverse
analysis’ is performed twice, for the pavement in the initial phases of the experiment,
after 5,000 passes, and also later in the experiment after 80,000 passes. Pavement
properties in both cases (i.e., layer moduli) are presented and compared (Subsection
4.2.3), showing that the structure experienced permanent property changes under the
repetitive APT passes.
Next, the calibrated APT model is extended to apply to other loading
configurations and environments. Assuming similarly constructed pavement systems,
the methodology consists of changing the HMA modulus to reflect different loading
speeds and temperatures. This is done after additional analysis of laboratory complex
modulus results (see details in Subsection 4.3.1). The unbound material properties,
4-1
although stress-state sensitive, are assumed unaffected by the changes in the HMA
stiffness. This may be justified, at least as a first order approximation, considering that
there are preexisting effective confining stresses in these materials (of unknown
magnitudes) originating from the construction process; these include vertical stresses
due to self weight, locked-in horizontal stresses from the compaction process, and
confining stresses due to negative pore pressures (see Subsection 3.3.1).
Finally, the loading and environment at NCAT are simulated and the extended
model is used in a forward calculation mode to forecast NCAT responses. Three types
of resilient responses are investigated (subsections 4.3.2 and 4.3.3): (i) peak surface
deflections observed during FWD testing (Subsection 4.3.2); (ii) vertical stresses on top
of the subgrade and aggregate base caused by an NCAT truck; and (iii) horizontal
strains at the bottom of the HMA induced by an NCAT truck. Owing to the permanent
property changes occurring in the APT experiment, the analysis focuses on the
pavement in the initial stages of the experiment. The concrete floor in the APT is left in
place during the forward analysis because deeper into the NCAT subgrade a rigid
bedrock material is expected (see Subsection 2.2.2).
4.2 LAYERED ELASTIC ISOTROPIC MODEL
4.2.1 Theory and Computational Implementation
Since its introduction by Burmister (1943; 1945), isotropic LET has been used by
engineers and researchers for representing the load induced resilient responses of
asphalt pavement systems. At this time, isotropic LET serves as the main ‘engine’ for
the MEPDG through the JULEA computer code (Uzan, 1976). According to the theory,
pavement materials are assumed to be linear elastic, homogeneous, isotropic and
weightless, characterized by an elastic (Young’s) modulus, E , and a Poisson’s ratio, ν .
Using a cylindrical coordinate system ( r , θ , z ), and assuming an axially
symmetric deformation field, the constitutive law is:
εr =
1
ν
⋅ σ r + ⋅ (σ θ + σ z ) ............................................................................ (4.2.1a)
E
E
4-2
εθ =
1
ν
⋅ σ θ + ⋅ (σ r + σ z ) ............................................................................ (4.2.1b)
E
E
εz =
ν
1
⋅ σ z + ⋅ (σ θ + σ r ) ............................................................................ (4.2.1c)
E
E
⎛ 1 +ν ⎞
⎟ ⋅ τ rz ........................................................................................... (4.2.1d)
⎝ E ⎠
ε rz = ⎜
ε zθ = ε rθ = 0 ................................................................................................. (4.2.1e)
A scalar ‘stress function’ φ (r , z ) that satisfies ∇ 4φ = 0 can be used to derive stresses
and displacements (Love, 1923):
σr =
∂ ⎛
∂ 2φ ⎞
⎜⎜ν ⋅ ∇ 2φ − 2 ⎟⎟ .............................................................................. (4.2.2a)
∂z ⎝
∂r ⎠
σθ =
1 ∂φ ⎞
∂ ⎛
2
⎜ν ⋅ ∇ φ − ⋅ ⎟ ........................................................................... (4.2.2b)
r ∂r ⎠
∂z ⎝
σz =
∂ ⎛
∂ 2φ ⎞
⎜⎜ (2 − ν ) ⋅ ∇ 2φ − 2 ⎟⎟ ...................................................................... (4.2.2c)
∂z ⎝
∂z ⎠
τ rz =
∂ ⎛
∂ 2φ ⎞
⎜⎜ (1 −ν ) ⋅ ∇ 2φ − 2 ⎟⎟ ........................................................................ (4.2.2d)
∂r ⎝
∂z ⎠
w=
1 +ν
E
u=−
⎛
∂ 2φ 1 ∂φ ⎞
⋅ ⎜⎜ (1 − 2 ⋅ν ) ⋅ ∇ 2φ + 2 + ⋅ ⎟⎟ .................................................... (4.2.2e)
r ∂r ⎠
∂r
⎝
1 +ν
E
⎛ ∂ 2φ ⎞
⎟⎟ ........................................................................................ (4.2.2f)
⋅ ⎜⎜
⎝ ∂r∂z ⎠
where u and w denote the r and z components of the displacement (respectively) and
∇ 2 is the Laplace operator ∇ 2 = ∂ 2 / ∂r 2 + (1 / r ) ⋅ ∂ / ∂r + ∂ 2 / ∂z 2 .
Consider a semi-infinite linear elastic isotropic and homogeneous medium made
of n − 1 parallel layers lying over a half-space. Each layer is identified by a subscript i
with material properties Ei and ν i . The layers are numbered serially, with the layer at
the top being layer 1 and the half-space, layer n . The origin of the cylindrical
4-3
coordinate system is placed at the surface of the first layer with the z -axis drawn into
the medium and the r -axis parallel to the layers. The depth to the individual interfaces,
measured from the surface, is denoted by z i ( i = 1, 2, .. n − 1 ). Hence, z1 is the thickness
of layer 1, z 2 is the combined thickness of layers 1 and 2, and so on. The combined
thickness of the n − 1 layers is denoted by H (i.e., H = z n −1 ).
Following Huang (2004), a ‘stress function’ that complies with all of the above
requirements is:
− m ⋅ ( λi − λ )
⎞
− Bi ⋅ e − m⋅( λ − λi−1 )
⎛ H 3 ⋅ J 0 (m ⋅ ρ ) ⎞ ⎛⎜ Ai ⋅ e
⎟ .... (4.2.3)
⎟
⋅
2
⎟ ⎜
− m ⋅ ( λi − λ )
m
− Di ⋅ m ⋅ λ ⋅ e − m⋅( λ − λi−1 ) ⎟⎠
⎝
⎠ ⎝ + Ci ⋅ m ⋅ λ ⋅ e
φi ( ρ , λ ) = ⎜⎜
in which ρ = r / H , λ = z / H , λi = z i / H and m is a unitless parameter; Ai , Bi , Ci
and Di are all unitless functions of m ; J k denotes a Bessel function of the first kind of
order k ; and the subscript i refers to the layer number. Substitution of this equation
into equations 4.2.2 yields the response of interest in a given layer i due to a vertical
non-dimensional surface load of the form m ⋅ J 0 (m ⋅ ρ ) . The value of the functions
Ai (m) , Bi (m) , Ci (m) and Di (m) cannot be expressed analytically; they must be
determined, for any given value of m , by solving a set of linear equations. This set of
equations transpires from the boundary and continuity conditions of the problem as
follows:
(σ z* )1 = m ⋅ J 0 (m ⋅ ρ )
(τ rz* )1 = 0
for λ = 0 ............................................................ (4.2.4a)
for λ = 0 ............................................................................... (4.2.4b)
(σ z* ) i = (σ z* ) i +1
for λ = λi .................................................................... (4.2.4c)
(τ rz* ) i = (τ rz* ) i +1
for λ = λi .................................................................... (4.2.4d)
( w* ) i = ( w* ) i +1
for λ = λi .................................................................... (4.2.4e)
(u * ) i = (u * ) i +1
for λ = λi ...................................................................... (4.2.4f)
(R* ) n = 0
for λ → ∞ ........................................................................... (4.2.4g)
4-4
in which the asterisk is used to indicate that the response is due to vertical surface
loading m ⋅ J 0 (m ⋅ ρ ) as can also be seen in equations 4.2.4a-b. Equations 4.2.4c-f
express the continuity of stresses and displacements inside the structure at the layer
interfaces; full bonding is suggested by equation 4.2.4f. Equation 4.2.4g means that all
response types (denoted using R ) must vanish for the n th layer and at infinite depth
(i.e., lim z →∞ R = 0 ). Finally, the response due to a uniform load q distributed over a
circular area of radius a is obtained by performing the integration:
R = (q ⋅ α ) ⋅
∞
R*
∫ m ⋅ J1 (m ⋅α ) ⋅ dm .................................................................... (4.2.5)
m =0
in which α = a / H and R is the stress or displacement of interest. Strains are thereafter
obtained using the constitutive relations (i.e., equations 4.2.1a-d).
For the purpose of this study, the entire aforementioned derivation was
programmed into an Excel worksheet (see program ELLEA1 in Appendix B). This was
done for the case of five layers and considering two separate loaded areas. The
combined effect of the two independent loads is calculated using superposition after
converting the axially symmetric results in each case to a Cartesian coordinate system.
The integration in equation 4.2.5 was carried out numerically between the first 200
zeros of the Bessel functions involved. The Gauss integration scheme was used for this
purpose whereby the first interval was integrated using a 30-point Gaussian formula, the
second interval was integrated using a 20-point formula, the third interval was
integrated using a ten-point formula and the remaining intervals were integrated using a
five-point formula. In order to speed the computational time, the number of matrix
inversions required for solving equations 4.2.4 was limited to 96, corresponding to 96
predetermined values of the integration variable m in the range of 0 to 50,000. A cubic
spline interpolation scheme was used to derive intermediate results within this range.
Furthermore, in order to improve the convergence of the integration, especially for
points residing close to the surface, one step of Richardson extrapolation was employed
(Sugihara, 1987).
4-5
The program’s user interface is shown in Figure 4.2.1. As can be seen (from top
to bottom), the input of material properties and layer thicknesses is done in the topmost
table. For each layer three attributes are required: Young’s modulus, Poisson’s ratio,
and thickness. In the example shown in Figure 4.2.1 only four layers are considered
because identical material properties are assigned to layers 3 and 4 (recall that all layers
are fully bonded). Next, the required loading information is defined for each of the two
loads, consisting of a vertical stress magnitude, loading radius, and the location of
application. In the example shown the loaded areas differ in their stress magnitude (105
vs. 55) and radius (4 vs. 7), and their locations of application are specified by the X and
Y coordinates. Finally, the coordinates of the evaluation point within the structure are
required. In the example they are x=2, y=2 and z=2; these values are relative to the
selected loading coordinates.
Figure 4.2.1: User interface of the isotropic LET program ELLEA1 (see Appendix B).
Based on the chosen depth for the evaluation point z, the program identifies
automatically the layer in question (layer #1 in the example). The resulting stresses,
strains and vertical deflection are shown in the bottom table. It should be noted that
4-6
there is no requirement to press a ‘run’ button to execute the code; in fact, any change
of value in one of the input tables will be automatically reflected in real time in the
results table. This feature is what makes this program extremely easy to use and
appealing for further analyses compared to any other available LET code. Note also that
no units are specified as the computations are done in dimensionless form (see Equation
4.2.3); the user must be consistent with his choice. In the example, units of psi are used
for moduli and stresses; inches are used for thicknesses, radii and coordinates.
4.2.2 Calibration to APT Conditions
It is well recognized that pavement materials do not comply with isotropic LET
assumptions. The resilient response of HMA mixtures is known to be anisotropic and
nonlinear viscoelastic (Shields et al., 1998; Levenberg, 2006; Uzan and Levenberg,
2007). The resilient response of unbound layers is nonlinear elastic and stress-state
sensitive (Uzan, 1985; 1992) and also anisotropic (e.g., Tutumluer and Thompson,
1997). As argued in Section 4.1 use of isotropic LET may be considered appropriate, at
least as a first order approximation, given that as-constructed pavement layers are not
stress-free even without external loads. These result in built-in stresses which diminish
somewhat the inconsistency with actual material behavior. Nevertheless, a systematic
error is introduced into the analysis when isotropic LET is applied. Minimizing this
error can be accomplished by deriving the free model parameters (i.e., elastic moduli)
through a process of inverse analysis (or backcalculation) using the time history of
embedded gauge readings.
Following this approach, subsequent stresses, strains and deflections calculated
with the calibrated model will resemble measured responses even though the model
assumptions are fundamentally incorrect and over-simplified. In this connection it
should be noted that LET cannot inherently simulate certain features that were seen in
the experiment (see Subsection 3.5.2). One example refers to the offset observed
between peak responses and load location which in LET must coincide. Another
example is the non-symmetric response relative to the load location, i.e., the differences
between approaching and receding curves as recorded by the gauges which in LET is
always symmetric (see also Elseifi et al., 2006; Al-Qadi, 2007).
4-7
For performing the backcalculation, the n1 pavement system was represented
using four layers. The three HMA lifts were combined into one (top) layer, 5 in. (127
mm) thick with an assumed Poisson’s ratio of ν 1 = 0.30 . The second layer from the top
represented the crushed aggregate base course, with a thickness of 6 in. (152.4 mm) and
ν 2 = 0.35 (assumed). Because no instrumentation was embedded in the subgrade (only
on top), there was no available data to support its sub-layering. Hence, the upper and
lower subgrade layers were combined into one layer (third layer from the top) having a
total thickness of 61 in. (1.55 m) and ν 3 = 0.40 (assumed). The fourth and final layer,
with semi-infinite thickness, represented the concrete floor of the test pit. The elastic
properties of this layer were fixed to the following values: E 4 = 4,000,000 psi (27,580
MPa) and ν 4 = 0.20 . The dual-wheel loading was represented by two circular areas,
each 8 in. (203 mm) in diameter, transferring uniform vertical stresses of 150 psi (1.03
MPa) to the pavement surface. The spacing between the centers of the loads was taken
as 13.5 in. (343 mm). For simulating the moving APT carriage, the quasi-static
approach was applied in which dynamic (inertial) effects are disregarded. This
assumption seemed reasonable because of the relatively slow loading speeds in the
APT.
Generated model responses were compared to measured responses and a
nonlinear optimization algorithm (Fylstra et al., 1998) was applied to manipulate the
material properties until a best fit was achieved. This process was repeated twice to
separately analyze the structure during pass #5,000 and during pass #80,000. Due to the
non-symmetric strain response of the pavement, only data from the approaching branch
were used for the comparison. Subsequently, 25 data points were pre-selected from each
time history, corresponding to 25 different APT carriage positions relative to the gauge
location with denser spacing closer to the gauge. These ‘offset’ distances ranged
between 70 in. (1.78 m), for which readings were negligible, and zero, in which the
APT carriage was exactly in line with the gauge along the Y-axis (see Figure 3.5.1).
Regardless of the number of data points used for the comparison between model
and experiment there were only three moduli that needed to be backcalculated (for a
given pass level), namely the HMA modulus ( E1 ), the aggregate base modulus ( E2 ),
4-8
and the subgrade modulus ( E3 ). In order to derive their numerical values, an objective
(scalar) function describing the agreement between the model and test data was
formulated. First, for each gauge separately, out of the total twelve gauges available, an
error term was defined as follows:
ERRg =
1
⋅
N
∑ [R
N
n =1
APT
n
− Rnmodel
]
2
.................................................................... (4.2.6)
in which N is the number of data points used for the comparison for the g th gauge
(i.e., N = 25 ). R APT represents the measured APT response of either stress or strain and
R model is the corresponding isotropic LET response. Note that ERR g has the same units
as R APT (or equivalently R model ) and is always positive. Next, these individual errors
were combined to formulate a unitless global error term, defined as follows:
Global _ Error =
⎤
1 G ⎡ ERRg
⋅∑⎢
− 1⎥ ..................................................... (4.2.7)
G g =1 ⎢⎣ min( ERRg ) ⎥⎦
where G is the total number of gauges considered in the analysis (i.e., G = 12 ), and
min( ERRg ) represents the lowest achievable error between the model and the test data
for the g th gauge. The numerical value of min( ERRg ) was obtained by employing an
over-fitting technique; i.e., the layer moduli were first manipulated using the
optimization algorithm in an effort to separately minimize each of the individual errors
(equation 4.2.6).
Note that min( ERRg ) is always greater than zero; even if the model were
perfect, all test data contain some random noise. However, the global error term can, in
principal, equal zero. This situation occurs mathematically when all individual errors
are minimal. Therefore, equation 4.2.7 serves as a weighted average of the individual
errors, making sure that neither of the gauge readings is underweighted or overweighted
in the backcalculation process compared to the others. In order to enable a direct
comparison between the global error for pass #5,000 and pass #80,000, values of
4-9
min( ERRg ) obtained for pass #5,000 were also used for the backcalculation of pass
#80,000.
4.2.3 Interim Results and Discussion
Table 4.2.1 presents the backcalculated layer moduli for pass #5,000 and pass #80,000.
The global error term (equation 4.2.7) was 4.89% for pass #5,000 and 6.27% for pass
#80,000. In both cases it can be seen that the stiffness of the pavement structure is
decreasing from top to bottom. During pass #5,000 the HMA is 14.6 times stiffer than
the underlying aggregate base. The aggregate base is seen to be twice as stiff as the
subgrade. By comparing these results with pass #80,000, it is clear that during the APT
experiment the individual layer moduli increased: (i) the HMA experienced a slight
stiffness increase of about 8.5%; (ii) the stiffness of the base increased significantly by
about 54%; and (iii) the subgrade increased in stiffness by about 16.5%. Subsequently,
the relative stiffness within the structure also changed, with the HMA ending up 10.3
times stiffer than the underlying base, and the base becoming 2.6 times stiffer than the
subgrade. In lieu of direct test data, these changes are believed to be the result of further
densification under the APT carriage passes, especially of the unbound materials.
Table 4.2.1: Backcalculated layer moduli for pass #5,000 and pass #80,000.
#
Layer
Thickness,
in. (mm)
Poisson’s
Ratio
Pass #5,000
Pass #80,000
Backcalculated Moduli, psi (MPa)
1
HMA
5 (127)
0.30
350,000 (2,412)
380,000 (2,618)
2
Base
6 (152)
0.35
24,000 (165)
37,000 (255)
3
Subgrade
61 (1,549)
0.40
12,000 (83)
14,000 (96)
4
Concrete
Semiinfinite
0.20
4,000,000 (27,580)
Figures 4.2.2 and 4.2.3 show both the measured and calibrated model responses
for pass #5,000 and #80,000 (respectively) vs. offset distance from the gauge. Each
figure contains six charts. The two topmost charts show horizontal strains in X (left)
4-10
and in Y (right) for gauges located along the loading centerline. The charts in the
middle of the figure show horizontal strains in X (left) and in Y (right) for gauges
positioned outside the loading path. The bottom charts show vertical stresses as
measured by pressure cells located on top of the base (left) and on top of the subgrade
(right). In each chart the measured gauge data is represented by solid markers. Because
the pavement was instrumented with pairs of gauges measuring the same response, two
160
0
80
-40
Computed Response
-60
G1_measured
-80
G3_measured
-100
-120
-140
0
-80
-160
Computed Response
G2_measured
G4_measured
-240
-320
-400
-480
-160
Offset [in.]
Offset [in.]
-180
-60
-50
-40
-30
-20
-10
-560
0
-70
-60
-50
-40
-30
-20
-10
0
180
150
G6_measured
120
G8_measured
90
60
30
0
40
Strain in X (z=5", x=24") [mstrains]
Computed Response
20
0
-20
Computed Response
-40
G5_measured
-60
G7_measured
-80
-100
Offset [in.]
Offset [in.]
-30
-70
-60
-50
-40
-30
-20
-10
-120
0
-70
-60
-50
-40
-30
-20
-10
0
21
40
35
1178_measured
30
1185_measured
25
20
15
10
5
0
-5
Offset [in.]
18
Computed Response
1179_measured
1184_measured
Stress in Z (z=5", x=0) [psi]
Computed Response
15
12
9
6
3
0
Offset [in.]
-10
-70
-60
-50
-40
-30
-20
-10
0
Stress in Z (z=11", x=0) [psi]
-70
Strain in Y (z=5", x=24") [μstrains]
-20
Strain in X (z=5", x=0) [μstrains]
20
Strain in Y (z=5", x=0) [μstrains]
types of markers are used. The calibrated model responses are shown using a solid line.
-3
-70
-60
-50
-40
-30
-20
-10
0
Figure 4.2.2: Resilient responses during APT pass #5,000. Both measured (solid
markers) and model generated (solid line) are shown.
These figures provide some intuition and information on several experimental
and modeling aspects. First, the large difference between measured responses of the
gauge pairs is demonstrated. Graphically, these differences seem to be smaller for the
4-11
stress measurements compared to the strain readings. During pass #5,000 the maximum
relative difference in the peak stress readings with respect to the average reading at the
peak is 10.6% (for pressure gauges 1178 and 1185). Similarly, the maximum relative
difference in the peak strain readings with respect to their average at the peak is 35.0%
(for strain gauges G5 and G7). During pass #80,000 the corresponding differences are
10.8% (again, pressure gauges 1178 and 1185) and 44.7% (strain gauges G6 and G8).
These differences are believed to represent both structural heterogeneity (see also
160
0
80
-40
-60
Computed Response
-80
G1_measured
-100
G3_measured
-120
-140
-160
Offset [in.]
0
-80
Computed Response
-160
G2_measured
-240
G4_measured
-320
-400
-480
Offset [in.]
-560
-180
-50
-40
-30
-20
-10
-70
0
Computed Response
-50
-40
-30
-20
-10
0
180
40
150
20
G6_measured
120
G8_measured
-60
90
60
30
0
Offset [in.]
0
-20
Computed Response
-40
G5_measured
-60
G7_measured
-80
-100
Offset [in.]
-30
-70
-60
-50
-40
-30
-20
-10
-120
-70
0
-60
-50
-40
-30
-20
-10
0
21
40
35
Computed Response
30
1179_measured
25
20
15
10
5
0
-5
Offset [in.]
Stress in Z (z=5", x=0) [psi]
Computed Response
1178_measured
1185_measured
18
15
1184_measured
12
9
6
3
0
Offset [in.]
-3
-10
-70
-60
-50
-40
-30
-20
-10
0
Stress in Z (z=11", x=0) [psi]
-60
Strain in X (z=5", x=24") [μstrains]
-70
Strain in Y (z=5", x=24") [μstrains]
-20
Strain in X (x=5", x=0) [μstrains]
20
Strain in Y (z=5", x=0) [μstrains]
Figures 3.5.6 to 3.5.18) and slight dissimilarity in gauge installation conditions.
-70
-60
-50
-40
-30
-20
-10
0
Figure 4.2.3: Resilient responses during APT pass #80,000. Both measured (solid
markers) and model generated (solid line) are shown.
Next, the goodness of fit of the calibrated model can be visualized. It may be
graphically seen that the isotropic LET captures relatively well the horizontal strains in
4-12
X and Y directions for the off-center gauges (G5 to G8). For the centerline gauges (G1
to G4), the horizontal strains are captured relatively well only in the direction of loading
(G2 and G4). The fit is not very good for the strains in the transverse direction (G1 and
G3). The vertical stress peaks on top of the base (1178 and 1185) and on top of the
subgrade (1179 and 1184) are underpredicted by the model. The above findings,
however, should not be expected to hold in general. In other cases, the stress
dependence of the unbound layers, and perhaps even anisotropy, may impair the
theory’s reproducibility. Also, the ability to successfully use LET is likely to weaken
when the pavement structure is comprised of thicker HMA layers. In this case, the
HMA’s time dependence will be more dominant and the non-symmetry in the strain and
stress response within the structure will be more pronounced.
4.3 NCAT RESPONSE PREDICTION
4.3.1 Methodology
The calibrated APT model (Table 4.2.1) cannot be used directly to forecast NCAT
responses. The main differences that need to be taken into account are HMA
temperature, axle configuration, axle weight and loading speed. It is straightforward to
apply the layered model with different axle configurations and different axle loads.
Because the quasi-static approach is applied for simulating the moving load, speed is
not an issue from a computational standpoint. However, the HMA properties
themselves are sensitive to the loading speed and temperature. This can be accounted
for exogenously by changing the HMA modulus in the NCAT simulation; the
methodology is described hereafter.
As a first step, the complex modulus test data from the individual HMA mixes
are combined into one dataset representative of one HMA layer that is 5 in. (127 mm)
thick. The following equations are suggested for this purpose:
3
E1com
⎡h ⋅3 E + h ⋅3 E ⎤
1
1,1
2
1, 2
⎥ ...................................................................... (4.3.1)
=⎢
h1 + h2
⎥
⎢
⎦
⎣
4-13
3
E2com
⎡h ⋅3 E + h ⋅3 E ⎤
1
2 ,1
2
2, 2
⎥ ...................................................................... (4.3.2)
=⎢
h1 + h2
⎥
⎢
⎦
⎣
*
Ecom
=
(E ) + (E )
com 2
1
com 2
2
.............................................................................. (4.3.3)
⎛ E2com ⎞
⎟ ....................................................................................... (4.3.4)
com ⎟
⎝ E1 ⎠
φcom = arctan⎜⎜
in which h1 = 1.0 in. (25.4 mm) is the lift thickness of Mix 1, h2 = 4.0 in. (102 mm) is
the lift thickness of Mix 2 (refer to Figure 3.2.1), E1,i and E2,i ( i = 1, 2 ) are the
components of the complex modulus for Mix i at a given test temperature and
*
frequency (refer to Table 2.3.5, Table 2.3.6 and equation 3.3.3), and Ecom
and φcom are
the combined dynamic modulus and phase angle at a given test temperature and
frequency.
The results from these computations are shown in Table 4.3.1 from which
master curves were constructed for a reference temperature of 15.5ºC (60ºF) using the
approach developed by Levenberg and Shah (2008). Referring to equations 3.3.4, 3.3.5,
3.3.6 and 3.3.7 (Subsection 3.3.2), the derived parameters for the combined properties
were: a1 = 2,328 MPa (337.6 ksi), a 2 = 1.03 ⋅ 10 −2 , a3 = 6.889 ⋅ 10 −5 s, E ∞ = 164 MPa
(23.8 ksi), c1 = 33.0 , and c 2 = 302.1 ºC (575.8ºF). The resulting dynamic modulus and
phase angle master curves (vs. reduced frequency f r ) are plotted in Figure 4.3.1. The
corresponding time-temperature shift factor ( aT ) vs. physical temperature is plotted in
Figure 4.3.2.
Next step, referring to Table 4.2.1, recall that the backcalculated modulus of the
HMA layers was 350,000 psi (2,412 MPa) in the initial part of the experiment. This
value is suitable for a temperature of 15.5ºC (60ºF) and a loading speed of 5 mph (~2.2
m/s). Using Figures 4.3.1 and 4.3.2, it may be seen that this stiffness level is paired with
a reduced frequency of 0.0232 Hz and a time-temperature shift factor of 1.0. Using the
same figures a new HMA modulus can be computed for any given loading speed and
temperature by adjusting f r and aT relative to the APT conditions. For example, at
4-14
NCAT the trucks are traveling at 45 mph (20.1 m/s); this speed is 9.0 ( = 45 / 5 ) times
higher than in the APT. Now, if the temperature was 15.5ºC (60ºF) then aT = 1.0 and
the reduced frequency would become 0.2088 Hz ( = 0.0232 ⋅ 9.0 ⋅ 1.0 ); the corresponding
modulus (using Figure 4.3.1) is therefore 4,028 MPa (584,100 psi). If, on the other
hand, the HMA temperature at NCAT was 30.0ºC (86ºF) instead of 15.5ºC (60ºF), then
aT = 0.0308
and
the
reduced
frequency
would
become
0.0064
Hz
( = 0.0232 ⋅ 9.0 ⋅ 0.0308 ); the corresponding modulus in this case is 1,685 MPa (244,300
psi).
In summary, using the combined dynamic modulus master curve with the newly
computed (reduced) frequency, one can adjust the HMA modulus to adequately
represent different conditions. This methodology is applied in the next subsections to
forward calculate responses of interest at NCAT. Thereafter, the forecast is compared
with measured values to assess the scheme.
100,000
45
15.5 ºC
Dynamic Modulus (data)
Phase Angle (data)
Viscoelastic Model
40
Dynamic Modulus [MPa]
30
25
1,000
20
15
100
Phase Angle [degrees]
35
10,000
10
5
Reduced Frequency [Hz]
10
1E-08
1E-06
1E-04
1E-02
1E+00
1E+02
1E+04
1E+06
0
1E+08
Figure 4.3.1: Combined HMA dynamic modulus and phase angle master curves for a
reference temperature of 15.5ºC (based on Table 4.3.1).
4-15
Table 4.3.1: Combined complex modulus properties for APT n1 / NCAT N1 (based on
equations 4.3.1 to 4.3.4).
Test Temperature Test Frequency Dynamic Modulus Phase Angle
ºC (ºF)
-10.0
(+14)
+4.4
(+40)
+21.1
(+70)
+37.8
(+100)
+54.4
(+130)
[Hz]
MPa (ksi)
[degrees]
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
25
10
5
1
0.5
0.1
24,951 (3619)
23,669 (3433)
22,329 (3239)
19,653 (2850)
18,426 (2672)
15,427 (2237)
15,995 (2320)
14,299 (2074)
13,229 (1919)
10,814 (1568)
9,832 (1426)
7,693 (1116)
8,265 (1199)
7,038 (1021)
6,037 (876)
4,295 (623)
3,708 (538)
2,580 (374)
3,873 (562)
3,188 (462)
2,543 (369)
1,584 (230)
1,339 (194)
939 (136)
1,405 (204)
1,122 (163)
918 (133)
606 (88)
534 (77)
413 (60)
6.5
9.6
10.3
11.8
12.5
14.7
10.7
12.0
13.9
15.9
17.0
20.1
15.4
15.6
22.7
25.2
26.3
28.7
26.4
26.6
25.0
32.7
32.1
31.0
25.4
21.2
20.6
24.6
25.8
27.3
4-16
1E+05
Time-Temperature Shift Factor [-]
-10.0ºC
1E+03
+4.4ºC
1E+01
aT=1.0
+21.1ºC
1E-01
+37.8ºC
1E-03
+54.4ºC
15.5ºC
0
Temperature [ C]
1E-05
-20.0
-10.0
0.0
10.0
20.0
30.0
40.0
50.0
60.0
Figure 4.3.2: Combined HMA time-temperature shifting for a reference temperature of
15.5ºC (based on Table 4.3.1).
4.3.2 Falling Weight Deflections
In this subsection an attempt is made to use the layered model, calibrated against APT
data (Subsection 4.2.2) and adjusted using laboratory tests (Subsection 4.3.1), to
forecast peak FWD deflections measured at NCAT (see Table 2.3.9). Given that the
diameter of the FWD plate is 11.8 in. (300 mm) and assuming a load pulse of 0.04
seconds in duration, the ‘speed’ of the loading in the FWD test is 295 in./s (=
11.8/0.04). This is 3.35 times faster than the APT speed (= 295/88). Referring to NCAT
section N1 before it incurred significant damage, the first three deflection tests in Table
2.3.9 are considered. The corresponding HMA temperatures were (respectively): 32.9ºC
(91.2ºF) on November 3, 2003, 16.7ºC (62.1ºF) on December 15, 2003, and 10.1ºC
(50.2ºF) on January 26, 2004. The appropriate time-temperature shift factors from
Figure 4.3.2 are therefore: aT
Nov . 03
= 0.016 , aT
Dec. 03
= 0.704 and aT
Jan. 04
= 3.987 .
Using these values the adjusted reduced frequencies become:
fr
Nov . 03
= 0.0232 ⋅ 3.35 ⋅ 0.016 = 0.001Hz .................................................... (4.3.5a)
4-17
fr
Dec. 03
fr
Jan. 04
= 0.0232 ⋅ 3.35 ⋅ 0.704 = 0.055 Hz .................................................... (4.3.5b)
= 0.0232 ⋅ 3.35 ⋅ 3.987 = 0.310 Hz .................................................... (4.3.5c)
in which 0.0232 is the reduced frequency derived from inverse analysis of APT
conditions (see Subsection 4.3.1). Using Figure 4.3.1 the corresponding HMA moduli
are: EHMA Nov. 03 = 1,000 MPa (145,000 psi), EHMA Dec. 03 = 2,900 MPa (420,500 psi) and
EHMA Jan. 04 = 4,600 MPa (667,000 psi).
Figure 4.3.3 shows the peak measured FWD deflections at NCAT and also the
corresponding computed deflections. The latter are based on the isotropic LET with
fixed moduli for the base and subgrade, but with different HMA moduli based on the
discussion above.
0
100
FWD Deflection [microns]
th
January 26 ,
2004
200
Isotropic LET
Nov. 2003 Data
Dec. 2003 Data
Jan. 2004 Data
300
400
th
December 15 ,
2003
500
rd
600
November 3 ,
2003
Distance from Center of FWD Plate [mm]
700
0
200
400
600
800
1000
1200
1400
1600
1800
2000
Figure 4.3.3: Measured and projected peak FWD deflections at NCAT N1.
The agreement between the test data and model projections can be graphically
evaluated from the figure. It may be seen that the trend of the computations matches the
trend in the deflection data. Quantitatively, the average absolute difference between test
and model for the three data sets is 32.4 microns; the average absolute relative error is
4-18
44.1%. If the isotropic LET model was directly calibrated using the NCAT deflections,
allowing only for the HMA modulus to differ in each case, the errors would have been
12.8 microns and 22.2% respectively. Therefore, although the trend in the computations
follows the trend in the data, the forecasting errors are 2.5 to 2.0 times higher compared
to the calibrated case.
4.3.3 Traffic Induced Stresses and Strains
In this subsection an attempt is made to forecast resilient stresses and strains at NCAT
caused by one truck pass. As indicated in Section 4.1 the analysis is focused on the
initial phase of the experiment before the HMA had incurred any visible fatigue
cracking. Assuming no interaction between the different truck axles (see Table 2.1.1),
the separate responses due to each half-axle are computed using isotropic LET. Model
predictions are then compared with the responses shown in Figures 2.5.1 to 2.5.4 as
measured by the gauge array in Figure 2.4.3.
First, the average temperature in the HMA is determined so that the timetemperature shift factor can be obtained from Figure 4.3.2. Using the three topmost data
points in Table 2.5.1, the temperature profile in the HMA is expressed as a second
degree polynomial: T ( z ) = 0.475 ⋅ z 2 − 4.75 ⋅ z + 88.5 in which z represents depth from
the surface in inches and T has units of ºF. By substitution of the appropriate depths it
can be verified that the expression reproduces the temperatures in Table 2.5.1. The
average HMA temperature is thereafter calculated by integrating T (z ) with respect to z
over the total HMA thickness, between 0 and 5 in. (0 to 127 mm), and dividing the
outcome by 5. Consequently, the average HMA temperature was found to be 80.6ºF or
27.0ºC which pairs with aT = 0.062 in Figure 4.3.2. Next, the HMA modulus is
determined from Figure 4.3.1; the reduced frequency is simply calculated as follows:
f r = 0.0232 ⋅ 9.0 ⋅ 0.062 = 0.013 Hz in which 0.0232 Hz is the reduced frequency that
represents the APT loading speed and temperature (Subsection 4.3.1), 9.0 accounts for
the difference in loading speed between APT and NCAT (Subsection 4.3.1) and 0.062 is
the time-temperature shifting (above). The resulting HMA modulus (see Figure 4.3.1) is
2,000 MPa (290,000 psi). All other layer properties are given in Table 4.2.1.
4-19
In the forward computations, the radius of contact area for each of the tires was
always taken as 4.0 in. (101.6 mm). The corresponding stress intensity was calculated
using the axle weights in Table 2.1.1. For the dual axles, center to center tire spacing
was taken as 13.5 in. (343 mm); for the dual tandem axles, axle spacing was taken as 50
in. (1.27 m). The moving NCAT truck was simulated by applying the array of tire loads
at different locations relative to the gauges. Calculations were performed for every
0.001 seconds during which the axles traveled forward 0.792 in. (20.1 mm) based on a
45 mph (792 in./s) speed. It is important to recall (see Subsection 2.5.1) that the wheel
positions relative to the gauges were not measured in the NCAT experiment. These
positions are necessary for performing the LET computations. In an effort to resolve
this issue it is assumed hereafter that peak gauge readings were attained when the load
was in line with the corresponding gauge along the Y-axis in Figure 2.4.3. This
assumption helps position the moving axles in the longitudinal direction (i.e., direction
of travel). It is further assumed that the truck wheels were moving in a straight line, not
necessarily parallel to road centerline while passing over the gauge array.
In Figure 4.3.4 the gauge array layout from Figure 2.4.3 is reproduced, showing
only the functioning gauges that survived the construction process. Also shown in the
figure, using connecting arrows, is the travel path of the center of the rightmost truck
tire. This line is located at an unknown transverse distance from the BBC gauge,
denoted in the figure as X 0 and an unknown distance from the ASC gauge, denoted as
X 1 . The determination of X 0 and X 1 is subsequently done, separately for each half-
axle considered, such that model predictions best conform with measured responses of
these two gauges. Hence, the matching between model and experiment for gauges BBC
and ASC should not be considered as pure prediction given that it was consciously
minimized to position the axles.
4-20
Y-axis [ft]
X1
ASC
6
S=45 mph
=792 in./s
4
ALR
ALC
2
ATC
X-axis [ft]
-2
-4
BTC
2
4
BTL
-2
BLC
BLR
-4
X0
BBC
-6
Figure 4.3.4: Layout of N1 gauge array (refer to Figure 2.4.3) and travel path
positioning of the center point of the rightmost truck tire (connecting arrows).
The resulting numerical values of X 0 and X 1 are shown in Figure 4.3.5 for the
different axles. Note that in this figure (and unlike Figure 4.3.4), the horizontal and
vertical scales are different with the horizontal scale stretched to better illustrate the
findings. The actual axle travel paths are not expected to be identical given that the
‘train’ of trailers has flexibility to move and ‘worm around’ in the transverse direction.
Accordingly, as can be seen in the figure, the computed travel paths are similar but not
identical, lying within a few inches from each other. Also noteworthy is the tendency of
4-21
the axles following the drive axle to drift to the right side relative to the Y-axis. This is
realistic considering the fact that the NCAT trucks traversed the N1 Section after
completing a left turn on the East curve (see Figure 2.1.1).
8
ASC
Y-axis [ft]
6
4
ALC
ALR
2
ATC
X-axis [ft]
0
-2.0
-1.6
BTL
-1.2
-0.8
-0.4
0.0
BTC
-2
BLC
0.4
0.8
1.2
1.6
2.0
S
BLR
D
1T
2T
-4
BBC
5T
3T
4T
-6
Figure 4.3.5: Travel paths of center of rightmost truck wheels over the N1 gauge array
at NCAT for the different axles in Table 2.1.1.
The following Figures 4.3.6 to 4.3.10 graphically contrast the computational
model and the measured resilient responses at NCAT (vs. time). Each figure separately
presents the stresses and strains due to a different half-axle. Referring to Table 2.1.1,
these are respectively: steering wheel (1S), drive axle (1D and 2D), first trailer axle
(1T), third trailer axle (3T), and last (fifth) trailer axle (5T). Each figure is comprised of
nine charts, depicting the calculated and measured response of the individual gauges
shown in Figure 4.3.5. The abscissa represents time in seconds, matching the timeline in
Figures 2.5.1 to 2.5.4. The ordinate depicts either vertical stress (in psi) or horizontal
strain (in microstrains) depending on the gauge considered (note that the scale changes
from case to case). In addition, each figure also includes a picture of the NCAT truck
with an arrow identifying the half-axle considered.
As a general observation, these figures show that the model predictions capture
relatively well the magnitudes as well as the trends in the measured responses. Similar
4-22
to the APT case (refer to Subsection 4.2.3), better matching is usually achieved for the
strains in the travel (longitudinal) direction (i.e., strains in Y) compared to the strains in
the transverse direction (i.e., strains in X). The forecastability of the vertical stresses
cannot be assessed because these were used to allocate the loads.
Quantitatively, the matching errors in Figures 4.3.6 to 4.3.10, between the
isotropic LET predictions and NCAT measured responses, are summarized in Table
4.3.2. These errors were computed using an expression similar to equation 4.2.6 using
about 200 points of comparison spanning the timeframe shown in each of the charts.
The reported error values should have all units of microstrains. However, because the
strain magnitudes in the charts were different in each case a low error level would not
necessarily mean better match. For this reason the individual errors were further
normalized by the peak to peak magnitude of the corresponding measured response; the
latter are shown in the table inside the brackets (units of microstrains). Subsequently,
the resulting errors in the table are dimensionless (reported in percent), and as can be
seen, range between 0.205% and 2.540%. The lowest error refers to matching the
response of the ALC gauge due to the fifth trailer axle (5T) - see Figure 4.3.10. The
highest error refers to mismatching the response of the ATC gauge due to the first
trailer axle (1T) - see Figure 4.3.8.
Table 4.3.2: Matching errors between isotropic LET predictions and NCAT measured
responses. Errors are in percent after normalization using the corresponding peak to
peak response shown in brackets (in microstrains).
Gauge
NCAT truck axle designation from Table 2.1.1
1S
1D+2D
1T
3T
5T
BLC
0.251 (220)
0.307 (450)
0.541 (410)
0.538 (370)
0.250 (210)
BLR
0.346 (100)
0.653 (260)
1.030 (200)
0.656 (360)
0.357 (510)
BTC
0.553 (80)
1.336 (360)
2.377 (300)
1.774 (220)
1.030 (115)
BTL
0.540 (40)
0.863 (100)
1.220 (100)
1.376 (90)
0.858 (80)
ATC
0.726 (90)
1.411 (350)
2.540 (280)
1.936 (200)
0.899 (150)
ALC
0.371 (200)
0.270 (500)
0.438 (450)
0.668 (320)
0.205 (220)
ALR
0.498 (80)
0.659 (200)
1.033 (180)
0.929 (280)
0.282 (500)
4-23
100
8
Stress in Z (z=5", x=0") [psi]
Strain in Y (z=5", x=0") [μstrain]
BBC_S Data
Model
7
6
5
4
3
2
1
0
50
0
-50
-100
-150
Time [s]
-1
0.000
0.050
0.100
0.150
0.200
0.250
0.300
0.350
-200
0.000
0.400
20
0
-20
-40
BLR_S Data
Model
-60
-80
0.000
0.050
0.100
Time [s]
0.150
0.200
0.250
0.300
0.350
0.150
0.200
0.250
0.300
0.350
0.200
0.250
0.300
0.350
0.400
0.250
0.300
0.350
0.400
0.450
0.250
0.300
0.350
0.400
0.450
0.400
35
30
BTL_S Data
Model
25
20
15
10
5
0
-5
-10
0.050
Time [s]
0.100
0.150
0.200
0.250
0.300
0.350
0.400
-30
-40
-50
-60
-70
BTC_S Data
Model
0.100
0.150
Time [s]
0.450
10
0
-10
-20
-30
-40
-50
-60
-70
-80
ATC_S Data
Model
0.150
0.200
Time [s]
0.500
0
-50
-100
-150
ALC_S Data
Model
0.150
0.200
Time [s]
0.250
0.300
0.350
0.400
0.450
0.500
Strain in Y (z=5", x=+24") [μstrain]
40
50
-250
0.100
-20
-90
0.100
0.450
100
-200
0
-10
20
Strain in X (z=5", x=0") [μstrain]
Strain in X (z=5", x=-24") [μstrain]
0.100
10
-80
0.050
0.400
40
Strain in Y (z=5", x=0") [μstrain]
0.050
Time [s]
20
Strain in X (z=5", x=0") [μstrain]
Strain in Y (z=5", x=+24") [μstrain]
40
BLC_S Data
Model
30
20
10
0
-10
-20
-30
-40
-50
-60
0.100
ALR_S Data
Model
0.150
0.200
Time [s]
0.500
Stress in Z (z=11", x=0") [psi]
8
7
ASC_S Data
Model
6
5
4
3
2
1
0
Time [s]
-1
0.150
0.200
0.250
0.300
0.350
0.400
0.450
0.500
0.550
Figure 4.3.6: Calculated and measured N1 responses - right side of steering axle (1S).
4-24
18
200
Stress in Z (z=5", x=0") [psi]
14
Strain in Y (z=5", x=0") [μstrain]
BBC_D Data
Model
16
12
10
8
6
4
2
0
-2
0.350
0.400
0.450
0.500
0.550
0.600
0.650
-200
-300
BLR_D Data
Model
100
50
0
-50
-100
-150
0.400
0.450
0.500
0.550
0.600
0.650
0.700
40
20
0
-20
Strain in X (z=5", x=0") [μstrain]
BTL_D Data
Model
60
0.550
0.600
0.650
0.700
0.550
0.600
0.650
0.700
0.750
0.550
0.600
0.650
0.700
0.750
0.550
0.600
0.650
0.700
0.750
0.750
-100
-150
-200
-250
-300
-350
BTC_D Data
Model
0.450
0.500
Time [s]
0.800
0
-50
-100
-150
-200
-250
-300
-350
Time [s]
0.450
0.500
0.550
0.600
0.650
0.700
0.750
-400
0.400
0.800
ATC_D Data
Model
0.450
0.500
Time [s]
0.800
100
100
0
-100
-200
ALC_D Data
Model
0.450
0.500
Time [s]
0.550
0.600
0.650
0.700
0.750
0.800
0.850
0.800
Strain in Y (z=5", x=+24") [μstrain]
200
-400
0.400
Time [s]
0.500
50
80
-300
0.450
0
-400
0.400
0.750
100
-40
0.400
0.400
-50
Time [s]
-200
0.350
BLC_D Data
Model
50
Strain in X (z=5", x=0") [μstrain]
Strain in Y (z=5", x=+24") [μstrain]
-100
-400
0.350
0.700
150
Strain in X (z=5", x=-24") [μstrain]
0
Time [s]
-4
0.300
Strain in Y (z=5", x=0") [μstrain]
100
50
0
-50
-100
-150
0.400
ALR_D Data
Model
0.450
0.500
Time [s]
0.800
18
Stress in Z (z=11", x=0") [psi]
16
14
ASC_D Data
Model
12
10
8
6
4
2
0
-2
-4
0.500
Time [s]
0.550
0.600
0.650
0.700
0.750
0.900
Figure 4.3.7: Calculated and measured N1 responses - right side of drive axle (1D and
2D).
4-25
18
200
Stress in Z (z=5", x=0") [psi]
14
Strain in Y (z=5", x=0") [μstrain]
BBC_1T Data
Model
16
12
10
8
100
0
-100
6
4
-200
2
0
-300
-2
Time [s]
-4
0.700
0.725
0.750
0.775
0.800
0.825
0.850
0.875
0.900
-400
0.750
Strain in X (z=5", x=0") [μstrain]
Strain in Y (z=5", x=+24") [μstrain]
50
0
-50
-100
-200
0.750
BLR_1T Data
Model
0.775
0.800
0.825
0.850
0.875
0.900
0.925
0.950
Strain in X (z=5", x=0") [μstrain]
Strain in X (z=5", x=-24") [μstrain]
BTL_1T Data
Model
40
20
0
0.775
0.800
0.825
0.850
0.875
0.900
0.925
0.875
0.900
0.925
0.825
0.850
0.875
0.900
0.925
0.875
0.900
0.925
0.950
0.975
0.900
0.925
0.950
0.975
1.000
0.950
-100
-150
-200
-250
-300
-350
BTC_1T Data
Model
-400
0.750
0.775
0.800
Time [s]
0.950
0
-50
-100
-150
-200
-250
-300
-350
-400
0.800
0.950
150
ATC_1T Data
Model
0.825
0.850
Time [s]
1.000
100
100
50
0
-50
-100
-150
-200
-250
-350
ALC_1T Data
Model
-400
0.825
0.850
0.875
Time [s]
0.900
0.925
0.950
0.975
1.000
1.025
Strain in Y (z=5", x=+24") [μstrain]
Strain in Y (z=5", x=0") [μstrain]
0.850
0
Time [s]
-300
0.825
50
60
-20
0.750
0.800
-50
Time [s]
100
80
0.775
Time [s]
50
100
-150
BLC_1T Data
Model
50
0
-50
-100
-150
0.825
ALR_1T Data
Model
0.850
0.875
Time [s]
1.025
12
Stress in Z (z=11", x=0") [psi]
10
ASC_1T Data
Model
8
6
4
2
0
Time [s]
-2
0.900
0.925
0.950
0.975
1.000
1.025
1.050
1.075
1.100
Figure 4.3.8: Calculated and measured N1 responses - right side of first trailer axle
(1T).
4-26
18
150
Stress in Z (z=5", x=0") [psi]
14
Strain in Y (z=5", x=0") [μstrain]
BBC_3T Data
Model
16
12
10
8
100
50
0
-50
-100
6
-150
4
-200
2
-250
0
-2
Time [s]
-4
1.150
1.175
1.200
1.225
1.250
1.275
1.300
1.325
1.350
-300
-350
1.200
100
50
0
-50
-100
-150
-200
-250
-300
BLR_3T Data
Model
-350
1.200
1.225
1.250
1.275
1.300
1.325
1.350
1.375
1.400
1.275
1.300
1.325
1.350
1.375
1.325
1.350
1.375
1.400
1.425
BTL_3T Data
Model
60
40
20
0
-20
0
-100
-150
-200
-250
-300
-350
BTC_3T Data
Model
-400
1.250
1.275
1.300
Time [s]
1.275
1.300
1.325
1.350
1.375
1.400
1.425
0
-100
ATC_3T Data
Model
-150
-200
-250
Time [s]
-300
1.300
1.450
1.325
1.350
1.375
1.400
1.425
1.450
1.475
1.375
1.400
1.425
1.450
1.475
1.500
150
100
50
0
-50
-100
-150
-200
-250
-350
ALC_3T Data
Model
-400
1.300
1.325
1.350
Time [s]
1.375
1.400
1.425
1.450
1.475
1.500
Strain in Y (z=5", x=+24") [μstrain]
Strain in Y (z=5", x=0") [μstrain]
150
-300
1.450
-50
Time [s]
-40
1.250
1.400
50
Strain in X (z=5", x=0") [μstrain]
Strain in X (z=5", x=-24") [μstrain]
1.250
-50
Time [s]
100
80
1.225
Time [s]
50
Strain in X (z=5", x=0") [μstrain]
Strain in Y (z=5", x=+24") [μstrain]
150
BLC_3T Data
Model
100
50
0
-50
-100
-150
-200
-250
ALR_3T Data
Model
-300
1.300
1.325
1.350
Time [s]
1.500
12
Stress in Z (z=11", x=0") [psi]
10
ASC_3T Data
Model
8
6
4
2
0
Time [s]
-2
1.350
1.375
1.400
1.425
1.450
1.475
1.500
1.525
1.550
Figure 4.3.9: Calculated and measured N1 responses - right side of third trailer axle
(3T).
4-27
8
100
Stress in Z (z=5", x=0") [psi]
6
Strain in Y (z=5", x=0") [μstrain]
BBC_5T Data
Model
7
5
4
3
2
0
-50
-150
Time [s]
-2
1.600
1.650
1.700
1.750
1.800
1.850
1.900
1.950
2.000
-200
1.600
100
0
-100
-200
-300
-400
-500
1.600
BLC_5T Data
Model
1.650
1.700
Time [s]
1.750
1.800
1.850
1.900
BLR_5T Data
Model
1.650
1.750
1.800
1.850
60
40
20
0
-20
1.900
1.950
Time [s]
-40
1.650
2.000
1.700
1.750
1.800
1.850
1.900
1.950
40
30
20
10
0
-10
Strain in X (z=5", x=0") [μstrain]
BTL_5T Data
Model
50
1.700
1.750
1.800
1.850
1.900
1.950
2.000
100
80
60
40
20
0
-20
Time [s]
-40
1.700
2.050
1.750
1.800
1.850
1.900
1.950
2.000
2.050
2.100
200
50
0
-50
-100
ALC_5T Data
Model
1.750
1.800
Time [s]
1.850
1.900
1.950
2.000
2.050
2.100
Strain in Y (z=5", x=+24") [μstrain]
100
Strain in Y (z=5", x=0") [μstrain]
2.050
ATC_5T Data
Model
120
Time [s]
-200
1.700
2.000
140
60
-150
2.000
BTC_5T Data
Model
80
Time [s]
1.700
70
-20
1.650
1.950
100
Strain in X (z=5", x=0") [μstrain]
200
Strain in Y (z=5", x=+24") [μstrain]
0
-100
1
-1
Strain in X (z=5", x=-24") [μstrain]
50
ALR_5T Data
Model
100
0
-100
-200
-300
Time [s]
-400
1.700
1.750
1.800
1.850
1.900
1.950
2.000
2.050
2.100
Stress in Z (z=11", x=0") [psi]
10
ASC_5T Data
Model
8
6
4
2
0
Time [s]
-2
1.800
1.850
1.900
1.950
2.000
2.050
2.100
2.150
2.200
Figure 4.3.10: Calculated and measured N1 responses - right side of last trailer axle
(5T).
4-28
4.4 APPRAISAL OF BASIC ANALYSIS
This chapter tackled the primary study objective, dealing with: (i) development of a
basic mechanistic pavement model based on isotropic LET; (ii) calibration of the
layered model using APT data through a process of inverse analysis; (iii) extension of
the model capabilities to apply to other loading configurations, other loading speeds,
and different environmental conditions; (iv) application of the extended model to
simulate the loading and environment at NCAT; and (v) assessment of forecastability.
Analysis of the APT structure was focused on the pavement in the initial phase
of the experiment, after 5,000 load applications, and also after 80,000 load applications.
The pavement was modeled using isotropic LET with material properties derived
through a process of backcalculation. The analysis was performed twice, separately for
each experimental stage, using the time history of the all gauge readings collected
during one pass of the APT carriage. Overall, the calibrated model captured relatively
well the trends in the test data (see Subsection 4.2.3).
Contrasting the backcalculated moduli in the two test stages revealed that the
pavement components increased in stiffness during the experiment. Most significant
was the increase in base stiffness, which was about 50% stiffer after 80,000 load
applications compared to the initial conditions, followed by about 16% increase in
subgrade modulus. Further densification under load is believed to be the main cause for
these changes.
The calibrated APT model was next extended to apply to other experimental
conditions by adjusting the stiffness of the HMA to reflect changes in loading speed and
temperature compared to the APT experiment. For this purpose the complex modulus
data (see Subsection 3.3.2) were further analyzed; resulting in a new set of master
curves and time-temperature shifting which were based on the combined properties of
the different HMA lifts. Subsequently, changes in HMA temperature were related to
modulus adjustments using time-temperature shifting; changes in loading speed were
also related to modulus adjustments by manipulating the effective frequency level.
Focusing on the initial phase of the experiment, the extended APT model was
then used to forecast load generated responses at NCAT, consisting of: peak FWD
4-29
deflections and stresses and strains induced by a moving truck. Considering all
simplifying assumptions made in the aforementioned scheme and the relatively few free
parameters used to represent the pavement system, the extended model performed
relatively well in projecting resilient responses at NCAT.
4-30
CHAPTER 5 - ADVANCED MECHANISTIC METHODS
In this chapter two more advanced models compared with the preceding chapter are
developed with the aim of establishing a superior link between the APT and NCAT
experiments. The first involves LET with anisotropic material properties and the second
involves layered viscoelastic theory (LVT) with isotropic material properties. The
underlying theories and computational implementations are discussed in subsections
5.1.1 and 5.2.1. In subsections 5.1.2 and 5.2.2 the models are calibrated using the APT
experiment; in subsections 5.1.3 and 5.2.3 they are further extended using laboratory
data and then applied to forecast selected NCAT responses. A short summary and
discussion of findings is provided in Section 5.3.
5.1 LAYERED ELASTIC ANISOTROPIC MODEL
5.1.1 Theory and Computational Implementation
This subsection contains the derivation of stresses and displacements for the case of
transversely isotropic multilayered elastic half-space with vertical loads applied at the
surface spread evenly over a circle. In terms of a Cartesian coordinate system ( x, y , z ),
with x − y as the plane of material isotropy, the constitutive law is:
ε x = a11 ⋅ σ x + a12 ⋅ σ y + a13 ⋅ σ z .................................................................. (5.1.1a)
ε y = a12 ⋅ σ x + a11 ⋅ σ y + a13 ⋅ σ z ................................................................... (5.1.1b)
ε z = a13 ⋅ σ x + a13 ⋅ σ y + a33 ⋅ σ z ................................................................... (5.1.1c)
ε yz = (a 44 / 2) ⋅ τ yz ......................................................................................... (5.1.1d)
ε xz = (a 44 / 2) ⋅ τ xz ......................................................................................... (5.1.1e)
ε xy = (a11 − a12 ) ⋅ τ xy ...................................................................................... (5.1.1f)
in which the six ε ’s are components of the strain tensor, the σ ’s and the τ ’s are
components of the stress tensor, and the a ’s are the material properties (i.e., elastic
constants).
5-1
In terms of a cylindrical coordinate system ( r , θ , z ), with z as the axis of
material symmetry, and assuming an axially symmetric deformation field (i.e.,
ε zθ = ε rθ = 0 ), the constitutive law becomes:
ε r = a11 ⋅ σ r + a12 ⋅ σ θ + a13 ⋅ σ z ................................................................... (5.1.2a)
ε θ = a12 ⋅ σ r + a11 ⋅ σ θ + a13 ⋅ σ z ................................................................... (5.1.2b)
ε z = a13 ⋅ σ r + a13 ⋅ σ θ + a33 ⋅ σ z ................................................................... (5.1.2c)
ε rz = (a 44 / 2) ⋅ τ rz .......................................................................................... (5.1.2d)
in which a11 = 1 / E x , a33 = 1 / E z , a12 = −ν xy / E x , a13 = −ν zx / E z and a44 = 1 / Gxz .
Hence, five elastic constants are included, namely: two Young’s moduli E x (= E y ) and
E z ; two Poisson’s ratios ν xy (= ν yx ) and ν zx (= ν xz ⋅ E z / E x ) ; and one shear modulus
G xz (= G yz ) . The condition that the strain energy must be positive imposes the following
property restrictions (PRs) on the values of the elastic constants (e.g., Poulus and Davis,
1974): (PR1) E x , E z , G xz > 0 ; (PR2) 1 −ν xy − 2 ⋅ν xz ⋅ν zx > 0 ; and (PR3) 1 − ν xy > 0 .
Following Lekhnitskii (1963) and Singh (1986), the stresses ( σ r , σ θ , σ z ,τ rz ) and
displacements ( u, w in the r, z directions respectively) can be derived from a stress
function φ (r , z ) as follows:
σr = −
∂ ⎛ ∂φ 2 b ∂φ
∂φ 2
⎜⎜ 2 + ⋅
+a⋅ 2
r ∂r
∂z ⎝ ∂r
∂z
⎞
⎟⎟ ............................................................. (5.1.3a)
⎠
∂ ⎛ ∂φ 2 1 ∂φ
∂φ 2
⎜
σθ = − ⎜b ⋅ 2 + ⋅
+a⋅ 2
r ∂r
∂z ⎝ ∂r
∂z
σz =
τ rz
∂ ⎛ ∂φ 2 c ∂φ
∂φ 2
⎜⎜ c ⋅ 2 + ⋅
+d⋅ 2
r ∂r
∂z ⎝ ∂r
∂z
∂ ⎛ ∂φ 2 1 ∂φ
∂φ 2
⎜
= ⎜ 2 + ⋅
+a⋅ 2
r ∂r
∂r ⎝ ∂r
∂z
⎞
⎟⎟ ......................................................... (5.1.3b)
⎠
⎞
⎟⎟ ............................................................ (5.1.3c)
⎠
⎞
⎟⎟ ................................................................ (5.1.3d)
⎠
5-2
u = (a12 − a11 ) ⋅ (1 − b) ⋅
∂ 2φ
......................................................................... (5.1.3e)
∂r∂z
w = (2 ⋅ a13 ⋅ a − a33 ⋅ d ) ⋅
⎛ ∂φ 2 1 ∂φ ⎞
∂ 2φ
a
−
44 ⎜
⎜ ∂r 2 + r ⋅ ∂r ⎟⎟ ........................................ (5.1.3f)
∂z 2
⎠
⎝
in which the parameters a , b , c and d are functions of the elastic constants:
a=
a13 ⋅ (a11 − a12 )
....................................................................................... (5.1.4a)
a11 ⋅ a33 − a132
b=
a13 ⋅ (a13 + a 44 ) − a12 ⋅ a 33
....................................................................... (5.1.4b)
a11 ⋅ a33 − a132
c=
a13 ⋅ (a11 − a12 ) + a11 ⋅ a 44
....................................................................... (5.1.4c)
a11 ⋅ a 33 − a132
a112 − a122
d=
......................................................................................... (5.1.4d)
a11 ⋅ a33 − a132
and the stress function φ (r , z ) satisfies the ‘compatibility’ equation ∇α2 ∇ 2β φ = 0 in
which ∇ α2 = ∂ 2 / ∂r 2 + r −1 ⋅ ∂ / ∂r + α −2 ⋅ ∂ 2 / ∂z 2 and ∇ 2β is identical to ∇α2 except for
β in place of α . The terms α and β are also derived from the material properties, as
follows:
⎛ α 2 ⎞ a + c ± (a + c) 2 − 4 ⋅ d
⎜ 2⎟=
...................................................................... (5.1.5)
⎜β ⎟
2⋅d
⎝ ⎠
The above algorithm imposes additional restrictions (named algorithm
restrictions or ARs) on the material properties: (AR1) a11 ⋅ a33 − a132 ≠ 0 ; (AR2)
(a + c) 2 − 4 ⋅ d > 0 ; (AR3) α 2 > 0 ; and (AR4) β 2 > 0 . In addition, α and β must be
distinct or the following derivation becomes singular; for this reason the isotropic case,
in which α = β = 1 , can only be approached but not directly computed.
Consider a semi-infinite medium made of n − 1 parallel layers lying over a halfspace. Each layer is identified by a subscript i with material properties ( E z ) i , ( E x ) i ,
5-3
(ν zx ) i , (ν xy ) i and (G xz ) i . Layers are numbered serially, the layer at the top being layer
1 and the half-space, layer n . Similar to the isotropic case (see Subsection 4.2.1), the
origin of the cylindrical coordinate system is placed at the surface of the first layer with
the z -axis pointing into the medium and the r -axis parallel to the layers. As before, the
depth to the individual interfaces (measured from the surface) is denoted by z i
( i = 1, 2, .. n − 1 ). Hence, z1 is the thickness of layer 1, z 2 is the combined thickness of
layers 1 and 2, and so on. The combined thickness of the n − 1 layers is denoted by H
(i.e., H = zn −1 ).
Inspired by Huang (2004) and Lekhnitskii (1963), a stress function that complies
with all of the above requirements is offered:
− m ⋅α ⋅( λ − λ )
− m ⋅α ⋅( λ − λi −1 )
⎞
⎛ H 3 ⋅ J 0 (m ⋅ ρ ) ⎞ ⎛⎜ Ai ⋅ e i i + Bi ⋅ e i
⎟ ................. (5.1.6)
⎜
⎟
⋅
φi ( ρ , λ ) = ⎜
2
⎟ ⎜
− m ⋅ β i ⋅( λi − λ )
− m ⋅ β i ⋅( λ − λi −1 ) ⎟
m
+
⋅
+
⋅
C
e
D
e
⎝
⎠ ⎝ i
i
⎠
in which ρ = r / H , λ = z / H , λi = z i / H and m is a unitless parameter; Ai , Bi , Ci
and Di are all unitless functions of m ; J k denotes a Bessel function of the first kind of
order k ; and the subscript i refers to the layer number. Substitution of this equation
into equations 5.1.3a-f yields the responses of interest in a given layer i due to a
vertical non-dimensional surface load of the form m ⋅ J 0 (m ⋅ ρ ) and not due to a
uniform load distributed over a circular area (this fact is indicated by an asterisk). The
resulting expressions are presented in what follows for completeness of the derivation.
(
)
⎛ α i ⋅ Lαi ⋅ Ai ⋅ e − m⋅α i ⋅( λi −λ ) − Bi ⋅ e − m⋅α i ⋅( λ −λi−1 )
(σ ) = ⎜
⎜ + β ⋅ Lβ ⋅ C ⋅ e − m⋅β i ⋅( λi −λ ) − D ⋅ e − m⋅βi ⋅( λ −λi−1 )
i
i
⎝ i i
*
r i
(
(
)
⎛ α i ⋅ Ai ⋅ e − m⋅α i ⋅( λi −λ ) − Bi ⋅ e − m⋅α i ⋅( λ −λi−1 )
+⎜
⎜ + β ⋅ C ⋅ e − m⋅β i ⋅( λi −λ ) − D ⋅ e − m⋅βi ⋅( λ −λi−1 )
i
i
⎝ i
(
(
)
(
(
)
)
⎞ (bi − 1) ⋅ J1 (m ⋅ ρ )
⎟⋅
⎟
ρ
⎠
............ (5.1.7a)
)
⎛ Qiα ⋅ Ai ⋅ e − m⋅α i ⋅( λi −λ ) − Bi ⋅ e − m⋅α i ⋅( λ −λi−1 )
(σ θ* ) i = ⎜
⎜ + Q β ⋅ C ⋅ e − m⋅β i ⋅( λi −λ ) − D ⋅ e − m⋅βi ⋅( λ −λi−1 )
i
i
i
⎝
− m⋅α i ⋅( λi − λ )
α
⎛ Si ⋅ Ai ⋅ e
− Bi ⋅ e − m⋅α i ⋅( λ −λi−1 )
⎜
+
⎜ + S α ⋅ C ⋅ e − m⋅βi ⋅( λi −λ ) − D ⋅ e − m⋅β i ⋅( λ −λi−1 )
i
i
⎝ i
(
⎞
⎟ ⋅ m ⋅ J 0 (m ⋅ ρ )
⎟
⎠
⎞
⎟ ⋅ m ⋅ J 0 (m ⋅ ρ )
⎟
⎠
................... (5.1.7b)
⎞ J1 (m ⋅ ρ )
⎟⋅
⎟
ρ
⎠
)
)
5-4
(
)
⎛ K iα ⋅ Ai ⋅ e − m⋅α i ⋅( λi −λ ) − Bi ⋅ e − m⋅α i ⋅( λ −λi−1 )
(σ z* )i = m ⋅ J 0 (m ⋅ ρ ) ⋅ ⎜
⎜ + K β ⋅ C ⋅ e −m⋅βi ⋅( λi −λ ) − D ⋅ e − m⋅βi ⋅( λ −λi−1 )
i
i
i
⎝
(
(
)
⎛ Lαi ⋅ Ai ⋅ e − m⋅α i ⋅( λi −λ ) + Bi ⋅ e − m⋅α i ⋅( λ −λi−1 )
(τ rz* ) i = m ⋅ J1 (m ⋅ ρ ) ⋅ ⎜ β
⎜ + L ⋅ C ⋅ e − m⋅β i ⋅( λi −λ ) + D ⋅ e −m⋅βi ⋅( λ −λi−1 )
i
i
⎝ i
(
(
⎞
⎟ ................... (5.1.7c)
⎟
⎠
)
⎞
⎟ ..................... (5.1.7d)
⎟
⎠
)
)
⎛ α i ⋅ Ai ⋅ e − m⋅α i ⋅( λi −λ ) − Bi ⋅ e − m⋅α i ⋅( λ −λi−1 )
(u * ) i = H ⋅ Fi ⋅ J1 (m ⋅ ρ ) ⋅ ⎜
⎜ + β ⋅ C ⋅ e −m⋅βi ⋅( λi −λ ) − D ⋅ e − m⋅βi ⋅( λ −λi−1 )
i
i
⎝ i
(
(
)
⎛ Giα ⋅ Ai ⋅ e − m⋅α i ⋅( λi −λ ) + Bi ⋅ e − m⋅α i ⋅( λ −λi−1 )
( w* ) i = H ⋅ J 0 (m ⋅ ρ ) ⋅ ⎜
⎜ + G β ⋅ C ⋅ e − m⋅βi ⋅( λi −λ ) + D ⋅ e − m⋅β i ⋅( λ −λi−1 )
i
i
i
⎝
(
⎞
⎟ ................ (5.1.7e)
⎟
⎠
)
⎞
⎟ .................... (5.1.7f)
⎟
⎠
)
in which the following settings were used in order to save space:
Lαi = 1 − ai ⋅ α i2 .............................................................................................. (5.1.8a)
Lβi = 1 − ai ⋅ β i2 .............................................................................................. (5.1.8b)
Qiα = α i ⋅ (bi − ai ⋅ α i2 ) ................................................................................... (5.1.8c)
Qiβ = β i ⋅ (bi − ai ⋅ β i2 ) ................................................................................... (5.1.8d)
Siα = α i ⋅ (1 − bi ) ............................................................................................ (5.1.8e)
Siβ = β i ⋅ (1 − bi ) ............................................................................................. (5.1.8f)
K iα = d i ⋅ α i3 − ci ⋅ α i .................................................................................... (5.1.8g)
K iβ = di ⋅ βi3 − ci ⋅ βi ...................................................................................... (5.1.8h)
Fi = ((a12 )i − (a11 )i ) ⋅ (bi − 1) ........................................................................... (5.1.8i)
Giα = (a44 )i − α i2 ⋅ ((a33 )i ⋅ di − 2 ⋅ (a13 )i ⋅ ai ) .................................................... (5.1.8j)
Giβ = (a44 )i − β i2 ⋅ ((a33 )i ⋅ di − 2 ⋅ (a13 )i ⋅ ai ) ................................................... (5.1.8k)
The value of each of the functions Ai (m) , Bi (m) , Ci (m) and Di (m) in
equations 5.1.7a-f is determined, for any given m , by solving a set of linear equations
that represent the boundary conditions. For the lowest layer, where i = n and λ → ∞ ,
5-5
the stresses and displacements must vanish; this leads to An = C n = 0 . For a vertical
load of the form m ⋅ J 0 (m ⋅ ρ ) applied to the surface of layer 1 (i.e., i = 1 , λ = 0 ), and
in the absence of shearing forces, we obtain the two conditions: (σ z* )1 = m ⋅ J 0 (m ⋅ ρ )
and (τ rz* )1 = 0 . Using equations 5.1.7c and 5.1.7d these conditions can be written
explicitly as follows:
(
)
(
)
K1α ⋅ A1 ⋅ e− m⋅α1 ⋅λ1 − B1 + K1β ⋅ C1 ⋅ e− m⋅ β1 ⋅λ1 − D1 = 1 ........................................ (5.1.9a)
(
)
(
)
Lα1 ⋅ A1 ⋅ e− m⋅α1 ⋅λ1 + B1 + L1β ⋅ C1 ⋅ e− m⋅ β1 ⋅λ1 + D1 = 0 .......................................... (5.1.9b)
Continuity of vertical and shear stresses at the interface between layers i and
i + 1 when λ = λi , is represented by the equations (σ z* ) i = (σ z* ) i +1 and (τ rz* ) i = (τ rz* ) i +1
or more explicitly:
(
)
⎛ K iα ⋅ Ai − Bi ⋅ e − m⋅α i ⋅( λi − λi−1 )
⎜
⎜ + K β ⋅ C − D ⋅ e − m⋅ β i ⋅( λi − λi−1 )
i
i
i
⎝
(
(
)
⎛ Lαi ⋅ Ai + Bi ⋅ e − m⋅α i ⋅( λi − λi−1 )
⎜
⎜ + Lβ ⋅ C + D ⋅ e − m⋅ β i ⋅( λi − λi−1 )
i
i
⎝ i
(
(
)
⎞ ⎛ K iα+1 ⋅ Ai +1 ⋅ e − m⋅α i+1 ⋅( λi+1 − λi ) − Bi +1
⎟=⎜
⎟ ⎜ + K β ⋅ C ⋅ e − m⋅ β i+1 ⋅( λi+1 − λi ) − D
i +1
i +1
i +1
⎠ ⎝
)
(
(
)
⎞ ⎛ Lαi +1 ⋅ Ai +1 ⋅ e − m⋅α i+1 ⋅( λi+1 − λi ) + Bi +1
⎟=⎜
⎟ ⎜ + Lβ ⋅ C ⋅ e − m⋅ β i+1 ⋅( λi+1 − λi ) + D
i +1
⎠ ⎝ i +1 i +1
)
(
⎞
⎟ ....... (5.1.9c)
⎟
⎠
)
⎞
⎟ ......... (5.1.9d)
⎟
⎠
)
Continuity of vertical and radial displacements at the interface between layers i and
i + 1 when λ = λi , is represented by the equations ( w* ) i = ( w* ) i +1 and (u * ) i = (u * ) i +1 or
explicitly:
(
)
⎧⎪Giα ⋅ Ai + Bi ⋅ e − m⋅α i ⋅( λi − λi−1 )
⎨
⎪⎩+ Giβ ⋅ Ci + Di ⋅ e − m⋅ β i ⋅( λi − λi−1 )
(
(
(
)
)
− m ⋅α ⋅ ( λ − λ )
Fi ⎧⎪α i ⋅ Ai − Bi ⋅ e i i i−1
⋅⎨
Fi +1 ⎪⎩+ βi ⋅ Ci − Di ⋅ e − m⋅ β i ⋅( λi − λi−1 )
(
)
⎫⎪ ⎧⎪Giα+1 ⋅ Ai +1 ⋅ e − m⋅α i+1 ⋅( λi+1 − λi ) + Bi +1
⎬=⎨
⎪⎭ ⎪⎩+ Giβ+1 ⋅ Ci +1 ⋅ e − m⋅ β i+1 ⋅( λi+1 − λi ) + Di +1
(
(
⎫⎪
⎬ ....... (5.1.9e)
⎪⎭
)
)
⎫⎪ ⎧⎪α i +1 ⋅ Ai +1 ⋅ e − m⋅α i +1 ⋅( λi +1 − λi ) − Bi +1
⎬=⎨
⎪⎭ ⎪⎩+ βi +1 ⋅ Ci +1 ⋅ e − m⋅ β i+1 ⋅( λi+1 − λi ) − Di +1
)
(
⎫⎪
⎬ .. (5.1.9f)
⎪⎭
)
Note that equation 5.1.9f represents full bonding at the interface between layers i and
i + 1 . Finally, the response of interest R generated by a uniform load q distributed over
a circular area of radius a is obtained by performing the integration:
5-6
R = (q ⋅ α ) ⋅
∞
R*
∫ m ⋅ J1 (m ⋅ α ) ⋅ dm ................................................................. (5.1.10)
m=0
in which α = a / H and R* is any stress or displacement of interest from equations
5.1.7a-f. Thereafter, the strains are calculated from equations 5.1.2a-d.
The aforementioned derivation was programmed into an Excel worksheet (see
program ELLEA2 in Appendix B). Similar to the isotropic program ELLEA1 (see
Chapter 4 and also Appendix B), the case of fully bonded five layers and two
independent loaded areas was considered. The integration method for equation 5.1.10
was also similar to that used for equation 4.2.5. The program was verified against three
cases: (i) closed form solution for a transversely-isotropic half-space acted upon by a
concentrated force applied at the surface (Lekhnitskii, 1963); (ii) tabulated solution for
a transversely-isotropic half-space due to a distributed load over a circular area (Poulos
and Davis, 1974); and (iii) numerical solution for isotropic LET (De Jong et al., 1973)
when the transversely-isotropic material properties approach the isotropic case.
The program’s user interface is shown in Figure 5.1.1. As can be seen, the
structural information is located in the topmost table. For each layer six attributes are
needed, namely the thickness and the five elastic constants. The loading information is
located in the following table; two loads are considered, each requiring four input
values: the stress intensity q , the radius a , and two relative coordinates in X and in Y
identifying the location of application. The computational results are shown in the
bottom table. Negative values for stresses or strains (or both) in the X, Y and Z
directions indicate tension while positive values indicate compression; positive
displacement refers to a downward deflection while a negative displacement refers to
heaving. Similar to ELLEA1, the calculations are done in real-time which means that
any change in the input parameters is automatically reflected in the results. It should be
noted that the program also calculates and displays (separately for each layer), property
restrictions 2 and 3 (i.e., PR2 and PR3) and algorithm restrictions 2, 3 and 4 (i.e., AR2,
AR3 and AR4). For the example in Figure 5.1.1 these are shown separately in Figure
5.1.2. In ELLEA2 these restrictions are displayed to the right hand side of the topmost
input table.
5-7
Figure 5.1.1: User interface of the anisotropic LET program ELLEA2.
Figure 5.1.2: ELLEA2 display of property and algorithm restriction for the example
shown in Figure 5.1.1.
5-8
5.1.2 Calibration to APT Conditions
The anisotropic LET properties were backcalculated using the APT study. This was
performed for the initial stages of the experiment, during pass number 5,000. Similar to
the isotropic analysis, a four layered structure was assumed with a semi-infinite
(isotropic) concrete bottom. The properties of the HMA ( i = 1 ), aggregate base ( i = 2 )
and subgrade ( i = 3 ) were obtained by matching model generated responses with
measured responses. Calculation steps used for this purpose were identical to those
outlined in Subsection 4.2.2 except for the number of unknowns. In the most general
case, since each anisotropic layer is characterized by five independent elastic properties,
there are 15 unknowns to be determined. In order to simplify matters and reduce the
number of free parameters to a more realistic level, only the vertical and horizontal
moduli were backcalculated for each layer, namely ( E z ) i and ( E x ) i .
The numerical values of the remaining elastic constants were a priori assumed.
The shear modulus Gxz in each layer was related to the other elastic properties
according to the following expression (Wolf, 1935; Barden, 1963; Christian, 1968):
Gxz =
Ex ⋅ Ez
............................................................................ (5.1.11)
Ez + Ex ⋅ (1 + 2 ⋅ν zx )
Also, the two Poisson’s ratios were assumed to be identical, i.e., (ν zx ) i = (ν xy ) i = ν i
with values predetermined similar to the isotropic case (see Table 4.2.1). This latter
assumption is also made in the Australian pavement design guide (Austroads, 2004).
Other assumptions are known to exist, most noteworthy of which (although not applied
herein) was developed by Graham and Houlsby (1983):
ν xy = ν zx ⋅
Gxz =
Ex
............................................................................................ (5.1.12a)
Ez
Ex ⋅ Ez
......................................................................................... (5.1.12b)
2 ⋅ (1 + ν xy )
5-9
in which it can be seen that the two Poisson’s ratios ν xy and ν zx are dissimilar. Note
that both equation 5.1.12b and equation 5.1.11 yield the isotropic shear modulus (as
expected) when the isotropic case is introduced, with E z = E x = E and ν xy = ν zx = ν .
Table 5.1.1 presents the calibrated (backcalculated) anisotropic elastic constants
for APT pass number 5,000. The global error term (equation 4.2.7) was 4.73% which is
only slightly lower compared to 4.89% in the isotropic case (note that similar values of
min(ERRg ) were used to make this comparison valid). Also listed in the table are the
Poisson’s ratios and the resulting shear moduli calculated according to equation 5.1.11.
Table 5.1.1: Backcalculated anisotropic layer moduli for pass #5,000.
#
Layer
Thickness,
in. (mm)
Poisson’s
Ratio
ν zx = ν xy
Backcalculated Moduli,
psi (MPa)
Shear
Modulus Gxz
(equation 5.1.11)
Ez
Ex
psi (MPa)
1
HMA
5 (127)
0.30
152,000 (1,050)
358,500 (2,470)
75,100 (520)
2
Base
6 (152)
0.35
61,300 (422)
10,850 (75)
8,340 (58)
3
Subgrade
61 (1,549)
0.40
11,950 (82)
9,250 (64)
3,865 (27)
4
Concrete
Semiinfinite
0.20
4,000,000 (27,580)
1,666,670
(11,490)
It may be seen that both the subgrade and aggregate base were found to be
stiffer in the vertical direction compared to the horizontal direction. Because
compaction processes produce preferred aggregate orientation (e.g., Oda et al., 1985;
Saadeh et al., 2002) and lock-in of horizontal stresses within the different layers (e.g.,
Uzan 1985; Duncan et al. 1991), this outcome is expected (at least conceptually). For
the HMA, however, the trend is reversed with greater stiffness in the horizontal
direction. Quantitatively, the ratio of E z / E x is 1.29, 5.65 and 0.42 for the subgrade,
base and HMA (respectively).
5-10
In the vertical ( z ) direction, the HMA is only 2.5 times stiffer than the
underlying aggregate base; this ratio seems relatively low. The base itself is about 5.0
times stiffer than the subgrade; a ratio that is relatively high considering the thinness of
the layer and structure. In the horizontal ( x − y ) direction, the HMA was found to be 33
times stiffer than the aggregate base; the value of E x seems too low for the base (only
1.2 stiffer than the subgrade), but reasonable for the HMA.
By comparison with the isotropic case for pass number 5,000 (see Table 4.2.1) it
may be seen that the isotropic analysis gives moduli values that more or less range
between E z and E x (separately for each layer); the stiffness ratios, however, do not
match.
Figure 5.1.3 graphically contrasts the measured and computed responses in the
APT (refer to Figure 3.5.1). Gauge readings are shown using solid markers (two types)
and the anisotropic model responses are shown using solid lines. The isotropic case,
reproduced from Figure 4.2.2, is also included (using dashed lines) for comparison. As
can be seen, the anisotropic model offers only a slight advantage in capturing the
measured responses compared to the isotropic case. Both models underestimate the
peak vertical stresses on top of the base (gauges 1178 and 1187) and the peak horizontal
strains in the X direction measured between the dual-wheel assembly (gauges G1 and
G3). The other three responses are relatively well reproduced.
5-11
Computed Response
Isotropic Case
G1_measured
G3_measured
-60
-80
-100
-120
-140
-160
Offset [in.]
0
-80
-160
Computed Response
Isotropic Case
G2_measured
G4_measured
-240
-320
-400
-480
Offset [in.]
-180
-70
-60
-50
-40
-30
-20
-10
-560
0
-70
-60
-50
-40
-30
-20
-10
0
180
150
120
90
60
30
0
40
Strain in X (z=5", x=24") [μstrains]
Computed Response
Isotropic Case
G6_measured
G8_measured
Offset [in.]
20
0
-20
Computed Response
Isotropic Case
G5_measured
G7_measured
-40
-60
-80
-100
Offset [in.]
-30
-70
-60
-50
-40
-30
-20
-10
-120
0
-70
-60
-50
-40
-30
-20
-10
0
40
Computed Response
Isotropic Case
1178_measured
1185_measured
21
Computed Response
Isotropic Case
1179_measured
1184_measured
25
20
15
10
5
0
-5
Offset [in.]
Stress in Z (z=5", x=0) [psi]
35
30
18
15
12
9
6
3
0
Offset [in.]
-10
-70
-60
-50
-40
-30
-20
-10
0
Strain in Y (z=5", x=0) [μstrains]
-40
80
Strain in Y (z=5", x=24") [μstrains]
-20
160
Stress in Z (z=11", x=0) [psi]
0
Strain in X (z=5", x=0) [μstrains]
20
-3
-70
-60
-50
-40
-30
-20
-10
0
Figure 5.1.3: Comparison of measured resilient responses in the APT during pass
#5,000 with responses computed using the anisotropic layered model (isotropic case is
reproduced from Figure 4.2.2).
5.1.3 NCAT Response Prediction
In this subsection the anisotropic layered model is applied to forecast resilient responses
measured at NCAT. Guided by the methodology developed in Chapter 4, the
computations are done with the calibrated properties shown in Table 5.1.1; the HMA
moduli are adjusted to account for the differences in loading speed and temperature
(relative to the APT experiment). For the anisotropic HMA case, there is an additional
assumption that the ratio E z / E x is unaffected by loading speed and temperature (i.e.,
remains equal to 0.42).
5-12
Referring first to peak FWD deflections, the first three basins obtained on Nov.
3, 2003, Dec. 15, 2003, and on Jan. 26, 2004 are considered (Table 2.3.9). In the
isotropic analysis of these same deflections (see Subsection 4.3.2), the adjusted HMA
moduli were: E HMA
psi) and E HMA
Nov . 03
Jan. 04
= 1,000 MPa (145,000 psi), E HMA
Dec. 03
= 2,900 MPa (420,500
= 4,600 MPa (667,000 psi). These stiffness values reflected
(respectively): 41.4%, 120.1% and 190.6% of the APT backcalculated modulus of 2,412
MPa (350,000 psi; see Table 4.2.1). These percentages were applied to the anisotropic
HMA moduli in Table 5.1.1; the resulting stiffnesses are listed in Table 5.1.2.
Table 5.1.2: Adjusted anisotropic HMA moduli for FWD response prediction.
FWD Test Date
Vertical Modulus E z ,
Horizontal Modulus E x ,
(refer to Table 2.3.9)
psi (MPa)
psi (MPa)
Nov. 3, 2003
63,000 (435)
148,420 (1,023)
Dec. 15, 2003
182,550 (1,260)
430,560 (2,970)
Jan. 26, 2004
289,700 (2,000)
683,300 (4,711)
Figure 5.1.4 shows the peak measured FWD deflections at NCAT (solid
markers) and also the corresponding computed/projected deflections using the
anisotropic LET (solid lines); the isotropic case from Figure 4.3.3 is reproduced here for
comparison (dashed lines). It may graphically be seen that the trend in the computations
follows the trend in the data. Quantitatively, the average absolute difference between
test data and the anisotropic model for the three dates considered is 33.6 microns (vs.
32.4 microns in the isotropic case); the average absolute relative error is 40.1% (vs.
44.1% in the isotropic case). If the anisotropic model was directly calibrated using the
NCAT deflections, allowing only for the HMA modulus to differ in each case (while
maintaining the ratio: E z / E x = 0.42 ), the aforementioned errors would have been 12.0
microns and 20.1% respectively (vs. 12.8 microns and 22.2% in the isotropic analysis).
Hence, the forecasting errors of the anisotropic model are 2.8 to 2.0 times higher
compared to the calibrated case. This outcome was also obtained in the isotropic case.
5-13
Based on these findings it may be concluded that the anisotropic analysis FWD
deflections offers little advantage over an isotropic analysis.
0.0
100.0
FWD Deflection [microns]
th
January 26 ,
2004
200.0
Anisotropic LET
Isotropic Case
Nov. 2003 Data
Dec. 2003 Data
Jan. 2004 Data
300.0
400.0
th
December 15 ,
2003
500.0
rd
600.0
November 3 ,
2003
Distance from Center of FWD Plate [mm]
700.0
0.0
200.0
400.0
600.0
800.0
1000.0
1200.0
1400.0
1600.0
1800.0
2000.0
Figure 5.1.4: Comparison of measured peak FWD deflections at NCAT N1 with
projected peak deflections using anisotropic LET (isotropic case reproduced from
Figure 4.3.3).
Referring next to the projection of truck traffic responses, recall that in the
isotropic case the HMA modulus was adjusted from 350,000 psi (2,414 MPa) in the
calibrated APT model to 290,000 psi (2,000 MPa) for NCAT conditions (Subsection
4.3.3). This 17.1% reduction was calculated based on the combined dynamic modulus
master curve (Figure 4.3.1) and on the combined time-temperature shifting (Figure
4.3.2) to reflect both truck speeds and HMA temperatures. For the anisotropic case,
assuming a constant E z / E x (=0.42), the new HMA moduli for NCAT speed and
temperature are simply those in Table 5.1.1 but each reduced by 17.1%. As before, all
other layer properties remain unchanged for the forward calculations.
In the following two figures, the computed and measured NCAT responses are
graphically contrasted. Given that the location of the NCAT trucks was not measured,
assumptions were made regarding their location relative to the gauge array. First, the
travel paths of the truck axles shown in Figure 4.3.5 were reused for performing the
5-14
forward calculations in the anisotropic case. Recall that these were obtained by the
requirement that the model matches the measured vertical stress responses (hence the
stress responses should not used to assess the modeling and predictive scheme). Second,
in order to position the axles longitudinally, the calculated and measured peaks were
made (forced) to coincide with each other.
With reference to Figure 2.1.2, Figure 5.1.5 is devoted to the responses caused
by the steer axle (1S), and Figure 5.1.6 is devoted to the responses caused by the third
trailer axle (3T). As performed in Subsection 4.3.3, each figure is comprised of nine
charts, depicting the calculated (solid lines) and measured (circular markers) response
of the individual gauges shown in Figure 4.3.5. The abscissa represents time in seconds,
matching the timeline in Figures 2.5.1 to 2.5.4. The ordinate depicts either vertical
stress (in psi) or horizontal strain (in microstrains) depending on the gauge considered.
Note that the scale changes from chart to chart. Each figure also includes a picture of
the NCAT truck with an arrow identifying the axle being considered. The isotropic
predictions, seen in Figures 4.3.6 and 4.3.9, are reproduced here for graphical reference
using dashed lines.
As a general observation, it may be seen in these figures that the solid and
dashed lines essentially coincide, indicating that the anisotropic model offers little
advantage over the isotropic case. Consequently, for the pavement system herein
considered, it seems that the added complexity involved in the anisotropic analysis
(although conceptually appealing) did not prove worthy. It very well may be that this
outcome transpired from imposing a time-independent model on the test data.
Henceforth, in the following section, time-dependence is introduced into the (isotropic)
model by treating the HMA, and therefore the entire system response, as viscoelastic.
5-15
8
100
BBC_S Data
Model
Isotropic Case
6
5
Strain in Y (z=5", x=0") [μstrain]
Stress in Z (z=5", x=0") [psi]
7
4
3
2
1
0
50
0
-50
-100
-150
Time [s]
-1
0.000
0.050
0.100
0.150
0.200
0.250
0.300
0.350
-200
0.000
0.400
0
-20
-40
-60
BLR_S Data
Model
Isotropic Case
0.050
0.100
0.150
Time [s]
0.200
0.250
0.300
0.350
0.400
0.150
Time [s]
0.200
0.250
0.300
0.350
0.250
0.300
0.350
0.400
0.300
0.350
0.400
0.450
0.300
0.350
0.400
0.450
0.400
30
25
BTL_S Data
Model
Isotropic Case
20
15
10
5
0
-5
-10
0.050
0.100
0.150
0.200
0.250
0.300
0.350
0.400
-50
-70
BTC_S Data
Model
Isotropic Case
-80
0.050
0.100
-60
0.150
0.200
Time [s]
0.450
0
-10
-20
-30
-40
-50
-60
-70
ATC_S Data
Model
Isotropic Case
0.150
0.200
0.250
Time [s]
0.500
Strain in Y (z=5", x=+24") [μstrain]
40
0
-50
-100
-250
0.100
-40
-90
0.100
0.450
50
-200
-30
-80
Time [s]
100
-150
-20
10
Strain in X (z=5", x=0") [μstrain]
35
0
-10
20
40
Strain in X (z=5", x=-24") [μstrain]
0.100
10
20
-80
0.000
Strain in Y (z=5", x=0") [μstrain]
0.050
20
Strain in X (z=5", x=0") [μstrain]
Strain in Y (z=5", x=+24") [μstrain]
40
BLC_S Data
Model
Isotropic Case
ALC_S Data
Model
Isotropic Case
0.150
0.200
0.250
Time [s]
0.300
0.350
0.400
0.450
0.450
0.500
0.500
30
20
10
0
-10
-20
-30
-40
-50
-60
0.100
ALR_S Data
Model
Isotropic Case
0.150
0.200
0.250
Time [s]
0.500
9
Stress in Z (z=11", x=0") [psi]
8
7
6
ASC_S Data
Model
Isotropic Case
5
4
3
2
1
0
Time [s]
-1
0.150
0.200
0.250
0.300
0.350
0.400
0.550
Figure 5.1.5: Comparison of anisotropic LET projections with measured N1 responses right side of steering axle (1S). Isotropic case reproduced from Figure 4.3.6.
5-16
18
150
BBC_3T Data
Model
Isotropic Case
14
12
100
Strain in Y (z=5", x=0") [μstrain]
Stress in Z (z=5", x=0") [psi]
16
10
8
6
4
2
0
-2
1.175
1.200
1.225
1.250
1.275
1.300
1.325
Strain in X (z=5", x=0") [μstrain]
Strain in Y (z=5", x=+24") [μstrain]
50
0
-50
-100
-150
-300
-350
1.200
BLR_3T Data
Model
Isotropic Case
1.225
1.250
1.275
1.300
1.325
1.350
1.375
-250
BLC_3T Data
Model
Isotropic Case
Time [s]
1.225
1.250
1.275
1.300
1.325
1.350
1.375
1.350
1.375
1.400
1.425
0
-100
-150
-200
-250
-300
-350
BTC_3T Data
Model
Isotropic Case
Time [s]
1.275
1.300
1.325
BTL_3T Data
Model
Isotropic Case
20
0
-20
0
-50
-100
ATC_3T Data
Model
Isotropic Case
-150
-200
-250
Time [s]
Time [s]
1.300
1.325
1.350
1.375
1.400
1.425
-300
1.300
1.450
150
100
100
Strain in Y (z=5", x=+24") [μstrain]
Strain in Y (z=5", x=0") [μstrain]
1.275
150
50
0
-50
-100
-150
-200
-250
-300
-350
-400
1.300
1.450
50
40
-40
1.250
1.400
-50
-400
1.250
1.400
Strain in X (z=5", x=0") [μstrain]
Strain in X (z=5", x=-24") [μstrain]
60
-200
Time [s]
100
80
-150
50
100
-250
-100
-350
1.200
1.350
150
-200
0
-50
-300
Time [s]
-4
1.150
50
ALC_3T Data
Model
Isotropic Case
1.325
1.350
1.375
1.425
1.450
1.475
1.500
1.525
1.500
1.350
1.375
1.400
1.425
1.450
1.475
1.400
1.425
1.450
1.475
1.500
50
0
-50
-100
-150
-200
-250
Time [s]
1.400
1.325
-300
1.300
ALR_3T Data
Model
Isotropic Case
1.325
1.350
1.375
Time [s]
1.500
20
Stress in Z (z=11", x=0") [psi]
18
16
14
ASC_3T Data
Model
Isotropic Case
12
10
8
6
4
2
0
-2
1.350
Time [s]
1.375
1.400
1.425
1.450
1.475
1.550
Figure 5.1.6: Comparison of anisotropic LET projections with measured N1 responses right side of third trailer axle (3T). Isotropic case reproduced from Figure 4.3.9.
5-17
5.2 LAYERED VISCOELASTIC ISOTROPIC MODEL
5.2.1 Theory and Computational Implementation
The time-dependent response R ve (t ) of a linear viscoelastic (non-aging) system under
isothermal conditions to a given time-dependent input I (t ) is fully characterized by a
function RHve (t ) named indicial admittance (von Karman and Biot, 1940). This function
represents the response of the system to a unit input applied as a step function in time,
i.e., RHve (t ) = R ve (t ) for the case where I (t ) = H (t ) in which H (t ) is the Heaviside unit
step function. In essence RHve (t ) is a function of the problem geometry, boundary
conditions and the material properties; it also depends on the preselected input (e.g.,
force or displacement on the boundary), and on the type of response of interest (e.g.,
stress, strain or displacement at a point). Whatever the case may be, once the unit
response function RHve (t ) is available, the response to an arbitrary time-varying input
can be calculated using Boltzmann’s superposition integral (e.g., Schapery, 1974):
t
R ve (t ) =
R
∫
τ
ve
H
(t − τ ) ⋅ dI (τ ) ............................................................................ (5.2.1)
=0
in which t is physical time and τ is a time-like integration variable.
In many cases (e.g., Pipkin, 1972; Lockett, 1972) the analysis of a linear
viscoelastic system can converted to a mathematically equivalent linear elasticity
problem. In general terms, this ‘elastic-viscoelastic’ correspondence principal states that
a Laplace transformed viscoelastic response to a given input can be obtained from the
corresponding response in an associated elastic problem having the same geometry, but
with the input replaced by its Laplace transform and all elastic material parameters
replaced by the s-multiplied Laplace transform of the corresponding viscoelastic
material functions. Subsequently, the time-domain (viscoelastic) response is obtained
from Laplace transform inversion of the associated elastic solution (Lakes, 1998).
Accordingly, let R e (t ) represent the elastic response in the associated elastic
problem corresponding to R ve (t ) in equation 5.2.1. Then, the elastic response can be
expressed as follows:
5-18
R e (t ) = RHe (c1e , c 2e , ... ) ⋅ I (t ) ............................................................................ (5.2.2)
in which I (t ) is the input and RHe is the indicial admittance of the associated elastic
system. The function RHe is seen to depend on a set of elastic time-independent material
constants denoted here as c1e , c2e , etc, corresponding to a set of viscoelastic (timedependent) material properties: c1ve , c2ve , etc. Although not shown explicitly, RHe also
depends on the geometry of the problem, boundary conditions and the type of response
being calculated.
The indicial admittance of the viscoelastic system RHve (t ) can be obtained by
first setting the input I (t ) in equation 5.2.2 to equal a unit input H (t ) and then
applying the elastic-viscoelastic correspondence principal. The resulting expression is:
RHve ( s ) = RHe (s ⋅ c1ve ( s ), s ⋅ c2ve ( s ), ...) ⋅ (1 / s ) ........................................................ (5.2.3)
where the macron (or overbar) denotes a Laplace transformed time-function with s
being the transform variable. Furthermore, it can be seen that the elastic constants were
replaced by the s-multiplied Laplace transforms of the corresponding viscoelastic
material properties. The term (1 / s ) on the right hand side is the Laplace transform of
the unit load H (t ) . By defining the Carson transform as the s-multiplied Laplace
transform (denoted by an overtilde), equation 5.2.3 can be rearranged as follows:
(
)
~
RHve ( s ) = RHe c~1ve ( s ), c~2ve ( s ), ... .......................................................................... (5.2.4)
At this point, the correspondence principal requires performing a Laplace
~
transform inversion of RHve ( s ) to obtain RHve (t ) . This last step is most often impossible
to perform using exact methods because either the elastic solution in known only
numerically or its analytic form cannot be inverted; the use of realistic relaxation and
creep functions adds further complexity to analytical inversion methods. In view of
these complications, Schapery (1962) had proposed two transform inversion methods
for which only the numerical values of the elastic solutions are required. The first is the
so-called ‘collocation method’ in which the elastic solution is first collocated by a finite
Dirichlet (or Prony) series and later inverted analytically. In principle, increasing the
5-19
length of the series increases the accuracy of the method, but at the cost of additional
computation time. The second inversion method is known as the ‘direct method’. It is
much easier and faster to apply, and although it offers little means of reducing error,
several studies have found it adequate for practical (engineering) purposes (Schapery,
1965; Hufferd and Lai, 1978). The ‘direct’ method was selected for use herein, and is
briefly described in what follows.
The Carson transform of f (t ) is defined as:
~
f (s) = s ⋅
+∞
∫ f (t ) ⋅ e
− st
⋅ dt ................................................................................. (5.2.5)
t =0
Using the change of variable w = log(s ⋅ t ) , equation 5.2.5 becomes:
~
f ( s) =
+∞
⎛ 10 w ⎞
∫−∞ f ⎜⎜⎝ s ⎟⎟⎠ ⋅ g (w) ⋅ dw .......................................................................... (5.2.6)
In which g ( w) = 10 w ⋅ e −10 ⋅ ln(10) . The key point of the ‘direct inversion’ method lies
w
in the properties of the function g (w) which is quasi-null everywhere except for
− 2 ≤ w ≤ 1 and also satisfies the relation:
+∞
∫ g (w) ⋅ dw = 1 ............................................................................................... (5.2.7)
−∞
Therefore, g (w) may be replaced in equation 5.2.6 by a Dirac delta function of the
form: δ ( w − w0 ) . This approximation leads to the direct inversion formula:
~
f ( s ) ≈ f ( β / s ) ............................................................................................ (5.2.8a)
or inversely
~
f (t ) ≈ f ( β / t ) ............................................................................................. (5.2.8b)
in which β = 10 w0 . The free parameter β (or equivalently w0 ) needs to be determined
to optimize the approximation accuracy. In the case where f (t ) is a power law in time,
β can be selected to yield the exact inversion. Schapery (1962; 1965; 1974) proposed
using β ≈ 0.5 .
5-20
Without assuming the numerical value of β , equation 5.2.8b is first applied to
equation 5.2.4, giving:
~
R Hve (t ) ≈ R Hve (β / t ) = R He (c~1ve ( β / t ), c~2ve ( β / t ), ...) ............................................ (5.2.9)
After applying equation 5.2.8b again, this time to the right hand side of equation 5.2.9,
the parameter β cancels out, resulting in:
RHve (t ) ≈ RHe (c1ve (t ), c2ve (t ), ...) .......................................................................... (5.2.10)
The meaning of equation 5.2.10 is that the indicial admittance of the viscoelastic system
RHve (t ) can be approximated at any given time t using the unit response of the
associated elastic problem R He with viscoelastic material functions replacing the
corresponding elastic constants and evaluated at time t . This procedure is also referred
to as the ‘quasi-elastic’ approximation (Schapery, 1962; 1965; 1974).
It is important to point out that, due to the assumption of linearity, equation 5.2.1
along with the entire aforementioned derivation can be generalized to compute the
response due to several α inputs (α = 1, 2, ...) :
R ve (t ) = ∑
α
t
∫τ R
ve
Hα
(t − τ ) ⋅ dI α (τ ) ................................................................. (5.2.11)
=0
in which RHveα is the viscoelastic response R ve (t ) due to a unit input I α (t ) = H (t ) , with
all other inputs zero (i.e., the viscoelastic indicial admittance for an isolated α input).
Equation 5.2.11 was ultimately used here to calculate preselected responses
(e.g., horizontal strains, vertical stresses) due to a moving half-axle at a given speed and
under constant temperature conditions. For computational implementation of the above
derivation, the isotropic LET program ELLEA1 was applied for generating solutions to
the associated elastic problem. Only the HMA modulus was assumed to be timedependent, while the moduli of all other layers and the Poisson’s ratio of all layers
(including that of the HMA) were assumed to be time-independent. Axle movements
ware simulated by sequentially loading and unloading the pavement surface at different
points located along the line of travel. A total of 63 points were used in equation 5.2.11
5-21
(i.e., α = 1,2,...,63 ) to yield 63 unit response time functions RHveα (t ) at different offset
distances. These offsets ranged from -76 in. (1.93 m) to +76 in. (1.93 m) relative to the
evaluation point, with 31 points before (approaching) the evaluation point, 31 points
after the evaluation point, and one additional point exactly in line with the evaluation
point. Spacing of these points ranged between 4 in. (101.6 mm) to 1.0 in. (25.4 mm)
with denser spacing closer to the evaluation point.
A triangular loading shape was applied at each of the loading points. This was
done in such a way that the pavement system always carried the full load of the halfaxle. As a point on the pavement was loaded, the previous (adjacent) point was
unloaded. When the peak load was reached at a given loading point, the load was
completely removed from the previous loading point that same instant. This scheme is
demonstrated in Figure 5.2.1. The lower part of the figure shows one evaluation point
(crossed circle) and also six load application points (out of the 63) located along the Yaxis (refer also to Figure 3.5.1). The spacing of these points is denoted by Δyα , with the
subscript indicating that the spacing was not uniform.
Figure 5.2.1: Scheme for simulating a moving load on a layered viscoelastic model.
5-22
The upper part of Figure 5.2.1 shows the individual ‘input’ functions vs. time,
illustrating the triangular load-unload shape applied to each point. Given that the
loading points were not spaced uniformly apart, the time difference between adjacent
load peaks (denoted as Δt ) was varied such that the load would appear to be moving at
a constant speed of choice (denoted as V in the figure).
Given a set of time-independent material properties, and a loading point α ,
R Hveα (t ) in equation 5.2.11 was first evaluated using ELLEA1. For this purpose, the
response of interest due to a unit stress intensity was computed 51 times, each with a
different HMA modulus level ranging between an upper bound E0 and a lower bound
E∞ . The time t associated with each modulus level depended on the shape of the
relaxation modulus E (t ) , was which was represented using the expression:
E∞ ⋅ [1 + (t / τ D ) nD ]
.......................................................................... (5.2.12)
E (t ) =
(t / τ D ) nD + ( E∞ / E0 )
in which τ D (units of time) and n D (unitless) are constants that control the shape that
E (t ) takes in the ‘transition’ between the two extreme values E0 and E∞ (note that
lim t →0 E (t ) = E0 and also that lim t →∞ E (t ) = E∞ ).
The resulting elastic response, plotted vs. time, formed points on the R Hveα (t )
curves. These points were thereafter interpolated using piecewise linear functions (with
time in logarithmic scale) for generating RHveα (t ) values for any given α and for any t
of choice. As an example, Figure 5.2.2 shows four such curves, each representing the
strain response in the travel direction at the bottom of the asphalt layer due to the APT
dual-wheel loading with an applied stress level of 1 psi (i.e., unit intensity input). Each
curve corresponds to a different offset distance from the evaluation point (indicated in
the chart). The strain responses are denoted using solid lines with values depicted on the
left ordinate. Values of E (t ) are shown on the right ordinate, and time is depicted on
the abscissa. In generating these curves, layer properties were taken from Table 4.2.1
(pass number 5,000 data). The HMA modulus was varied between E0 = 40825 MPa
(5,920,000 psi) and E∞ = 164 MPa (23,800 psi), such that the associated times will be
5-23
evenly spaced on a logarithmic scale. Both extreme modulus values were obtained from
*
the combined dynamic modulus master curve in Figure 4.3.1 using E0 = lim f r → ∞ Ecom
*
and E ∞ = lim f r →0 E com
. The constants τ D and nD were arbitrarily determined as:
τ D = 35,000 seconds and nD = 0.240.
1.E+05
2.0
1.5
Offset=20.0 [in.]
0.5
Offset=10.0 [in.]
0.0
1.E+04
-0.5
-1.0
Offset=5.0 [in.]
-1.5
1.E+03
-2.0
-2.5
-3.0
-3.5
1.E+02
Offset=0.0 [in.]
-4.0
Strain
Modulus
-4.5
-5.0
-5.5
1.E-10
1.E-08
1.E-06
HMA Relaxation Modulus [MPa]
Strain in Y (x=0; y=5") [μstrain]
1.0
Time [s]
1.E-04
1.E-02
1.E+00
1.E+02
1.E+04
1.E+06
1.E+08
1.E+01
1.E+10
Figure 5.2.2: Indicial admittance of a layered viscoelastic system (example). Strain
response due to a unit intensity ‘input’ of an APT half-axle passing along four offset
distances from the evaluation point.
Subsequently, the integration in equation 5.2.11 was commenced; this was
performed numerically based on the expression:
⎞
⎛
R ve (t n ) = ∑ ⎜⎜ ∑ H (t n − τ m ) ⋅ R Hveα (t n − τ m ) ⋅ ΔI α (τ m ) ⎟⎟ ................................ (5.2.13)
α ⎝ m
⎠
in which R ve (t n ) is the viscoelastic response of interest evaluated at time t n , H (⋅) is
the unit Heaviside function, R Hveα (t n − τ m ) is the indicial admittance of the viscoelastic
system for a given α , evaluated at time t n − τ m , and ΔI α (τ m ) = I α (τ m ) − I α (τ m −1 ) . The
parameter n ranged between 1 and 23 resulting in 23 values of t n , chosen such that the
5-24
shape of the viscoelastic response was adequately captured within the offset range of
±70 in. (±1.78 m). Physically, the calculation points were spaced 2 to 5 in. (50.8 to 127
mm) apart; a cubic spline interpolation scheme was used to generate intermediate
responses. The parameter m ranged between 1 and 3,200 with values of τ m chosen
such that each triangular load-unload ‘input’ (see Figure 5.2.1) was divided into 100
time intervals: 50 during loading and 50 during unloading.
It should be noted that the time t associated with each relaxation modulus level
E (t ) depends also on the temperature considered in the analysis. Using the timetemperature superposition concept it was possible to include temperature dependence in
the aforementioned formulation by horizontal shifting of the E (t ) curve in Figure 5.2.2.
Mathematically, this was accomplished by multiplying τ D in equation 5.2.12 by the
time-temperature shift factor aT (see equation 3.3.6). Naturally, for analysis performed
under the reference temperature in which E (t ) was determined, aT equaled unity and
equation 5.2.12 was unaffected. For an analysis temperature that was different than the
reference temperature, aT departed from unity (see Figure 4.3.2), resulting in horizontal
shifting of the E (t ) curve. For example, when the analysis temperature was higher than
the reference temperature, aT became greater than unity, resulting in horizontal shifting
of E (t ) to the left. As expected, under such circumstances the material appears softer
because the transition from E0 to E ∞ becomes, in effect, faster.
5.2.2 Calibration to APT Conditions
The layered viscoelastic model was calibrated using the time history of the responses
measured in the APT during pass number 5,000. Similar to the previous analyses
(subsections 4.2.2 and 5.1.2) material properties were determined using a nonlinear
error minimization algorithm by matching computed and measured responses as best as
possible. The objective function had similar structure to that used previously (see
equations 4.2.6 and 4.2.7) so as to ensure that all gauge readings were equally weighted
in the backcalculation process. Ideally, the moving half-axle load in the APT should
have been applied many times over in the simulation, and the matching performed using
5-25
the responses computed in the last simulated pass. However, due to computational
power limitations, only one movement of the half-axle was simulated. Furthermore, it
should be noted that unlike the previous analyses (refer to Figures 4.2.2 and 5.1.3), the
matching of computed and measured responses was not limited to the approaching
branches and was also performed for the receding branches.
The pavement system was modeled as a four layered half-space. Only the top
layer, representing the HMA, with a thickness of 5 in. (127 mm) was treated as
viscoelastic while the remaining three layers were treated as time-independent (elastic).
A constant Poisson’s ratio of 0.30 was assumed for the HMA with a relaxation modulus
E (t ) that follows equation 5.2.12. The extreme values of E (t ) at t → 0 and t → ∞
were prefixed to: 5,920,000 psi (40,825 MPa) and 23,800 psi (164 MPa) respectively.
These values were obtained from the combined dynamic modulus master curve in
*
*
and E ∞ = lim f r →0 E com
. The second layer from
Figure 4.3.1 using E 0 = lim f r →∞ E com
the top represented the aggregate base with a thickness of 6 in. (152 mm), Poisson’s
ratio of 0.35 and modulus E 2 . The third layer from the top represented the subgrade
with a thickness of 61 in. (1,549 mm), Poisson’s ratio of 0.40 and modulus E3 . The
bottom (fourth) layer with semi-infinite thickness represented the concrete floor of the
test pit, having the following properties: ν 4 = 0.20 and E 4 = 4,000,000 psi (27,580
MPa). Consequently, four unknown parameters were determined by the inverse
analysis, namely: the two remaining HMA properties τ D and n D (equation 5.2.12), the
base modulus E 2 , and the subgrade modulus E3 .
Table 5.2.1 presents the calibrated material properties of the layered viscoelastic
model corresponding to APT pass number 5,000. The global error term (equation 4.2.7)
was 1.32% which is much lower than, but not directly comparable to, the timeindependent cases, mainly because matching was performed for both the approaching
and receding branches of the responses. As can be seen in the table, the base modulus
was found to be lower than the subgrade modulus, which contradicts the findings from
the time-independent analyses. With reference to the resilient modulus tests (see Figure
3.3.1), a subgrade modulus of 25,915 psi (180 MPa) seems too high (i.e., exceeding the
5-26
resilient modulus range of test results), and a base modulus of 6,820 psi (47 MPa)
appears too low for a material compacted to 97% (which is the compaction degree in
the APT experiment). In comparison with the isotropic LET analysis (Table 4.2.1), the
subgrade here is about 2.2 times stiffer; the base modulus here is merely 28% of that
backcalculated in the time-independent isotropic case.
Table 5.2.1: Backcalculated material properties for the layered viscoelastic model
during APT pass #5,000.
Modulus, psi (MPa)
#
Layer
Thickness,
in. (mm)
Poisson’s
Ratio
1
HMA
5 (127)
0.30
E0 = 5,920,000 (40,825); E∞ = 23,800 (164);
τ D = 21.8 s; nD = 0.532
2
Base
6 (152)
0.35
6,820 (47)
3
Subgrade
61 (1,549)
0.40
25,915 (180)
4
Concrete
Semiinfinite
0.20
4,000,000 (27,580)
+ equation 5.2.12 parameters
As for the HMA, Figure 5.2.3 superimposes the backcalculated relaxation
modulus E (t ) , i.e., equation 5.2.12 and parameters from Table 5.2.1, with the
relaxation modulus interconverted from the combined dynamic modulus and phase
angle master curves in Figure 4.3.1. The interconversion from the frequency domain to
the time was performed using the following equation (Levenberg and Shah, 2008):
∞
E (t ) = E∞ +
∫ h(τ ) ⋅ e
τ
−t / τ
⋅ d (ln τ ) .................................................................. (5.2.14)
=0
in which h(τ ) is the relaxation spectrum given in equation 3.3.7 with parameters listed
in Subsection 4.3.1 (calibrated to results in the frequency domain). As can be seen in the
figure, although both curves were derived for a reference temperature of 15.5ºC, and
although the extreme values E0 and E ∞ were forced to coincide, the transitions from
E0 to E ∞ are completely different (faster in the backcalculated case).
5-27
100,000
Relaxation Modulus [MPa]
15.5 ºC
10,000
From Complex Modulus Tests
1,000
From APT Inverse Analysis
100
Reduced Time [s]
10
1.E-10
1.E-08
1.E-06
1.E-04
1.E-02
1.E+00
1.E+02
1.E+04
1.E+06
1.E+08
1.E+10
Figure 5.2.3: Comparison of backcalculated relaxation modulus with that interconverted
from complex modulus test results.
In the context of inverse analysis, the base and subgrade moduli values, together
with the viscoelastic HMA properties, provided the best overall match with the
embedded gauge readings; as such they are expected to ensure good forecastability. The
apparent unrealistic base and subgrade moduli values, and the differences observed
between the relaxation modulus curves in Figure 5.2.3, can be mainly caused by: (i) use
of linear elastic isotropic properties to represent the base and subgrade layers and
isotropic properties to represent the HMA course; and (ii) simulating just a single load
pass to match a response measured after many consecutive load passes. In future studies
the first cause can be addressed by using anisotropic elastic (linear or nonlinear)
properties to represent the base and subgrade materials and by treating the HMA as
anisotropic (viscoelastic); the second cause can be addressed in future studies by
simulating more load passes over the viscoelastic system while performing the inverse
analysis.
In addition, it should be noted that the dissimilarity in the relaxation modulus
curves in Figure 5.2.3 can also reflect differences in strain levels, given that complex
modulus tests were performed under very small strains (~100 microstrains) while the
5-28
backcalculated viscoelastic properties are associated with strain levels that are about
four times higher. Furthermore, the dissimilarity can originate from the differences in
aggregate structure between an HMA prepared in the laboratory versus an HMA
prepared using full-scale construction equipment.
Figure 5.2.4 shows the measured and calibrated model responses for APT pass
number 5,000. Six charts are included, each showing a different response vs. offset
distance from the gauge. With reference to Figure 3.5.1, the two topmost charts show
horizontal strains in X (left) and in Y (right) for gauges located along the loading
centerline (between the dual tires). The charts in the middle of the figure show
horizontal strains in X (left) and in Y (right) for gauges positioned outside the loading
path. The bottom charts show vertical stresses as measured by pressure cells located on
top of the base (left) and on top of the subgrade (right). In each chart the measured
gauge data is represented by two types of solid markers and the calibrated model
responses are shown using solid lines.
From the figure it may be graphically seen that the isotropic LVT captures
relatively well both the shape and magnitude of all measured responses except for the
vertical stresses on top of the base (gauges 1178 and 1185) which are under predicted. It
should be noted that the non-symmetry is the responses were also captured very well;
the LVT was able to simulate the differences between the approaching and receding
response branches and the delay in the peaks (occurring slightly after the load had
passed a gauge).
5-29
150
20
100
Strain in Y (z=5"; x=0) [μstrain]
Strain in X (z=5"; x=0) [μstrain]
40
0
-20
-40
-60
-80
-100
-120
Viscoelastic Model
-140
G1_measured
-160
G3_measured
Offset [in.]
-180
0
-50
-100
-150
-200
-250
-300
Viscoelastic Model
-350
G2_measured
-400
G4_measured
-60
-50
-40
-30
-20
-10
0
10
20
30
40
50
60
70
-70
-60
-50
-40
-30
-20
-10
0
10
20
30
40
-10
0
10
20
30
40
50
60
70
40
160
Strain in Y (z=5"; x=24") [μstrain]
Viscoelastic Model
140
G6_measured
120
G8_measured
100
80
60
40
20
0
20
0
-20
-40
-60
Viscoelastic Model
-80
G5_measured
-100
G7_measured
Offset [in.]
Offset [in.]
-120
-20
-70
-60
-50
-40
-30
-20
-10
0
10
20
30
40
50
60
-70
70
40
-60
-50
-40
-30
-20
50
60
70
25
Viscoelastic Model
35
Viscoelastic Model
1178_measured
30
Stress in Z (z=11"; x=0) [psi]
Stress in Z (z=5"; x=0) [psi]
Offset [in.]
-450
-70
Strain in X (z=5"; x=24") [μstrain]
50
1185_measured
25
20
15
10
5
0
-5
1179_measured
20
1184_measured
15
10
5
0
Offset [in.]
-10
Offset [in.]
-5
-70
-60
-50
-40
-30
-20
-10
0
10
20
30
40
50
60
70
-70
-60
-50
-40
-30
-20
-10
0
10
20
30
40
50
60
70
Figure 5.2.4: Comparison of measured resilient responses in the APT during pass
#5,000 with responses computed using the isotropic viscoelastic layered model.
5.2.3 NCAT Response Prediction
The calibrated APT model (Table 5.2.1) was used to forecast the NCAT responses due
to a moving truck. No attempt was made to forecast peak FWD deflections as was done
using the time-independent models, mainly because the time history of the FWD
loading is unavailable. Therefore, the forecasting here is aimed at projecting the
collected responses shown in Figures 2.5.1 to 2.5.4 as measured by the gauge array in
Figure 2.4.3. The fact that layered viscoelastic theory is employed means that all main
differences between the APT experiment and NCAT (i.e., axle configuration, axle
weight, loading speed and HMA temperature) are endogenously and naturally taken into
account.
5-30
As before, the elastic properties of the base and subgrade (and rigid bottom)
were assumed to be identical in both experiments, unaffected by the different loading
conditions. The appropriate HMA temperature at NCAT was determined in Subsection
4.3.3 to be 80.6ºF (27.0ºC); based on Figure 4.3.2 this temperature level is associated
with a time temperature shift factor of 0.062 (i.e., aT = 0.062). Using a loading speed
( V in Figure 5.2.1) that is nine times higher than the APT, responses were generated by
the calibrated viscoelastic model for the different axle configurations and loads given by
Table 2.1.1. The travel paths of the truck axles shown in Figure 4.3.5 were reused for
performing the forward calculations here. Recall that these were obtained by forcing the
measured and calculated stresses to match in the isotropic LET case (see subsections
4.3.3 and 5.1.3). Hence, the modeling capabilities should only be judged based on the
strain responses. Additionally, in order to position the axles longitudinally (as their
location was not measured in NCAT), the calculated and measured peaks were made
(forced) to coincide with each other.
Figures 5.2.5 to 5.2.9 graphically contrast the computational model and the
measured resilient responses at NCAT (vs. time). Each figure separately presents the
stresses and strains due to a different half-axle. Referring to Figure 2.1.2 and Table
2.1.1, these are respectively: steering wheel (1S), drive axle (1D and 2D), first trailer
axle (1T), third trailer axle (3T), and fifth (last) trailer axle (5T). Each figure is
comprised of nine charts, individually depicting the measured response (circular
markers) and calculated response of the viscoelastic model (solid lines) for the gauges
shown in Figure 4.3.5. The isotropic LET case is also shown for graphical comparison
(dashed lines), reproduced from Figures 4.3.6 to 4.3.10. The abscissa represents time in
seconds, matching the timeline in Figures 2.5.1 to 2.5.4. The ordinate depicts either
vertical stress (in psi) or horizontal strain (in microstrains) depending on the gauge
considered (note that the scale changes from case to case). In addition, each figure also
includes a picture of the NCAT truck with an arrow identifying the half-axle
considered.
As a general observation, these figures show that the model predictions capture
relatively well the magnitudes as well as the trends in the measured responses.
5-31
Quantitatively, the matching errors in these figures were almost consistently lower than
those in Figures 4.3.6 to 4.3.10. This is shown in Table 5.2.2, which lists the
improvement in predictive power (in percent) of the viscoelastic model over the
isotropic elastic case (as given in Table 4.3.2). As can be seen, the improvement ranged
between 0.9% to 71.7% with an overall average of 27.8%; only a single negative
(worsening) case was obtained (1S axle, ATC gauge).
Table 5.2.2: Relative improvement in response predictions for the isotropic LVT
compared to the isotropic LET case given in Table 4.3.2.
Gauge
NCAT truck axle designation from Table 2.1.1
1S
1D+2D
1T
3T
5T
BLC
39.9%
47.6%
49.3%
46.3%
20.5%
BLR
21.2%
9.9%
7.5%
19.8%
21.2%
BTC
64.6%
3.9%
6.8%
42.7%
49.0%
BTL
26.4%
71.7%
67.5%
27.2%
0.9%
ATC
-28.0%
2.6%
5.0%
41.9%
51.4%
ALC
29.0%
42.3%
52.0%
32.7%
41.0%
ALR
15.5%
11.8%
5.0%
7.8%
18.3%
5-32
Figure 5.2.5: Comparison of isotropic LVT projections with measured N1 responses right side of steer axle (1S). Isotropic case reproduced from Figure 4.3.6.
5-33
Figure 5.2.6: Comparison of isotropic LVT projections with measured N1 responses right side of drive axle (1D and 2D). Isotropic case reproduced from Figure 4.3.7.
5-34
Figure 5.2.7: Comparison of isotropic LVT projections with measured N1 responses right side of first trailer axle (1T). Isotropic case reproduced from Figure 4.3.8.
5-35
Figure 5.2.8: Comparison of isotropic LVT projections with measured N1 responses right side of third trailer axle (3T). Isotropic case reproduced from Figure 4.3.9.
5-36
Figure 5.2.9: Comparison of isotropic LVT projections with measured N1 responses right side of fifth (last) trailer axle (5T). Isotropic case reproduced from Figure 4.3.10.
5-37
5.3 APPRAISAL OF ADVANCED METHODS
As an extension to Chapter 4, two more advanced pavement models were employed
here to address the primary study objective of linking the APT and NCAT experiments,
namely: anisotropic layered elasticity, and isotropic layered viscoelasticity. The
analyses were performed for the pavements in the very early stages of the experiment,
focusing on resilient responses. First, the mathematical derivation of the models was
presented in detail. From then on, the models were calibrated using APT response data
and later enhanced using laboratory results to apply to other conditions not included in
the calibration. Thereafter, the loading and environment at NCAT were simulated and
the advanced models applied to forecast measured resilient responses consisting of peak
FWD deflections, and stresses and strains induced by a moving truck. The calculated
results were graphically and quantitatively compared with the measurements.
Referring first to the anisotropic LET, the model requires five elastic constants
for characterizing each layer. For calibration purposes, only two constants were
manipulated, namely the elastic moduli in the vertical and horizontal directions; the
numerical values of the remaining parameters were assumed. It was found that both the
subgrade and base were stiffer in the vertical direction (which can be expected) while
the HMA was found to be stiffer in the horizontal direction. While the calibration error
was only slightly lower compared to the isotropic case, questionable stiffness ratios and
levels were produced. After applying the model to forecast NCAT responses it was
found, both graphically and quantitatively, that the anisotropic treatment yielded only
slight improvements over the isotropic case. In conclusion, although the application of
anisotropic behavior is conceptually appealing, for the pavement system herein
considered, the added complexity involved in the anisotropic analysis did not prove
worthy.
Referring next to the isotropic LVT, the subgrade and aggregate base were each
characterized as time-independent using an elastic modulus and a Poisson’s ratio. The
HMA was characterized as viscoelastic using a relaxation modulus and a constant
Poisson’s ratio. A mathematical expression consisting of four parameters was used to
represent the relaxation modulus. Two out of the four parameters were directly
5-38
determined from complex modulus test results. The calibration to APT conditions was
performed after assuming the numerical values of the Poisson’s ratios and manipulating
the remaining material parameters. Hence, in effect, the viscoelastic analysis included
only one additional free material parameter compared to the basic isotropic layered
elastic analysis. The calibration produced seemingly unrealistic unbound material
properties with a very high subgrade stiffness and very low base stiffness, much lower
than the subgrade. Also, the backcalculated HMA relaxation modulus did not coincide
with that obtained from complex modulus tests. The probable reasons for these findings
were discussed in the text along with possible corrective measures (yet to be
implemented). Nevertheless, when forecastability is considered, the LVT proved
superior to the other theories and as such should be preferred in any future attempt to
apply APT results to other conditions.
5-39
CHAPTER 6 - CONCLUSION
This chapter offers a short summary of the entire report and highlights the main
findings/results (Section 6.1). In Section 6.2, general recommendations are suggested
including future research ideas, followed by specific advice on how INDOT should
implement the study results.
6.1 SUMMARY AND FINDINGS
The main objective of this study was to devise and validate an analysis scheme by
which experimental data collected in INDOT’s APT facility could be used to
successfully forecast the corresponding pavement behavior at the NCAT test track.
More details related to the overall objective and approach can be found in Chapter 1.
Only one pavement system was addressed, consisting of a relatively thin
structure composed of 5 in. (127 mm) of HMA and 6 in. (152 mm) of aggregate base
overlaying an untreated silty soil serving as subgrade. In both the APT and NCAT
studies this pavement system was built with embedded instrumentation, in order to
measure environmental changes (temperatures, moisture content) and load induced
transient responses (vertical stresses, horizontal strains). The pavement was loaded
during the Phase II experiment at NCAT between the years 2003 and 2005 while the
APT study took place between the years 2004 and 2006. In association with these
studies, laboratory tests were also performed on the individual pavement constituents.
The main study objective was pursued in this report according to the following
steps: (i) development of mechanistic models to represent the pavement system; (ii)
analysis of laboratory test results; (iii) calibration of the necessary material properties
from the APT experiment by means of inverse analysis; (iv) enhancement of the
modeling capabilities to apply to other loading and environmental conditions not
included in the calibration using laboratory test results; (v) simulation of the loading
and environmental conditions at NCAT and forward calculation of load induced
responses due to an FWD and a moving truck; and (vi) comparison of measured and
calculated responses to assess the forecastability of the proposed scheme.
6-1
Chapter 2 included relevant information from the NCAT study. Chapter 3
summarized the APT work and laid the groundwork for commencing the mechanistic
analyses. These two chapters revealed that:
(i)
Even though nominally identical pavement systems were constructed in the APT
and at NCAT, the differences in loading and environmental conditions produced
completely distinct responses and dissimilar cracking and rutting performances.
Hence, it was concluded that a direct (empirical) relation between the two experiments
cannot be established and that it is unavoidable to apply a more fundamental/rational,
mechanistic based approach;
(ii)
Neither the APT nor NCAT studies were designed and carried out with pure
mechanistic interpretation in mind. Partially for this reason, a large part of the
collected data in the APT could not be utilized. Similarly, only limited means for
validating the proposed scheme were offered by the NCAT data;
(iii) The instrumentation used in both experiments to observe and record mechanical
responses was suited for monitoring dynamic transient responses but not for
monitoring permanent responses. This fact confined the subsequent analyses to
focusing on resilient (recoverable) responses only;
(iv) In the APT study, large differences in response were observed in pairs of gauges
that were expected to record identical readings. These differences were assumed to be
the result of structural heterogeneity (also manifested in the rutting results) and slight
dissimilarities in gauge installation conditions. Because of these differences the
advantage of using more sophisticated models to represent the pavement system may
not be evident/noticeable;
(v)
Due to the controlled conditions in the APT facility it was possible to observe
that the resilient responses exhibited permanent changes during the experiment,
presumably due to load generated permanent changes in the material properties.
Similar findings could not be made for the NCAT study because of the changing
environmental conditions and because the axle loadings were not applied exactly at
the same location (and the location of application relative to the gauges was not
6-2
recorded). Mainly for this reason, subsequent analyses were focused on the initial
stages of both experiments.
Both Chapters 4 and 5 contained the development of mechanistic models for
representing the pavement system. These were followed by calibration procedures and
utilization of laboratory test results to enhance the applicability of these models to other
loading and environmental conditions. Then, the models were used in forward
calculation mode to forecast load induced resilient responses at NCAT. Finally, the
calculations were compared with the measurements to assess their forecastability.
In Chapter 4 the pavement was modeled as an isotropic layered elastic halfspace (see program ELLEA1 in Appendix B). Four layers were used with properties
calibrated under APT conditions using backcalculation by manipulating the elastic
constants until model-generated responses matched as closely as possible the time
history of all embedded gauge readings collected during one APT pass. This inverse
analysis procedure is a key point in the proposed methodology, both for the basic model
and for the more advanced models, as it minimizes any systematic errors associated
with the modeling simplifications. After calibration, the model was extended to apply to
other loading and environmental conditions using complex modulus test results. This
was accomplished by varying the stiffness of the HMA to reflect changes in
temperature and loading speed relative to the APT conditions while maintaining all
other material properties. The extended APT model was then used to forecast load
induced responses at NCAT, consisting of peak FWD deflections, and stresses and
strains resulting from a moving truck. Considering the relative simplicity of the above
scheme and the small number of free parameters used to represent the pavement system,
the isotropic layered elastic model performed relatively well in projecting resilient
responses at NCAT.
In Chapter 5 the pavement was represented using two more advanced models:
anisotropic layered elasticity (see program ELLEA2 in Appendix B), and isotropic
layered viscoelasticity. Again (and separately for each case), the models were first
calibrated using APT response data by means of inverse analysis and later enhanced
using laboratory results to apply to other conditions not included in the calibration.
6-3
Thereafter, the loading and environment at NCAT were simulated and the advanced
models applied to forecast measured resilient responses. Referring first to the
anisotropic model, it was found that only mild improvements over the isotropic elastic
case were offered and hence concluded that the added complexity involved in the
anisotropic analysis did not prove worthy. As for the viscoelastic model, it was found
that although computationally more demanding, a relatively simple calibration
procedure could be followed, involving laboratory test results and one additional free
parameter over the isotropic elastic case. The resulting calculations generated superior
forecastability compared to the other two theories suggesting that this model should be
preferred in the future. It is important to note that the related mathematical derivation
was based on the elastic-viscoelastic correspondence principle, considerably
simplifying the computational implementation, allowing readily available elastic
programs to be utilized.
6.2 RECCOMENDATIONS AND IMPLEMENTATION
As previously stated, the main objective of this study was to try and link INDOT’s APT
results with those obtained at NCAT for nominally similar pavement systems. In the
present work this objective was pursued by means of a mechanistic approach given that
establishing a direct comparison was deemed unachievable. However, since both
experiments were not a priori designed and carried out with a pure mechanistic
interpretation in mind, the collected results limited the analysis efforts to dealing with
resilient responses only, disregarding performance, i.e., permanent deformations
(rutting) and load induced cracking. Henceforth, as a general recommendation, it is
suggested that the traditional (empirical) approach of using APT studies as ‘rut testers’
or ‘pavement comparators’ be completely abandoned for future APT studies in favor of
more
modern
(advanced)
uses.
The
benefits
gained
from
employing
empirical/oversimplified approaches in the past are mostly exhausted, and if meaningful
improvements are to be made in the pavement design field, future efforts should be
placed on developing a more rational framework.
Subsequently, it is proposed that future APT research studies should aim at
developing a mechanistic scheme for applying APT results to field conditions for
6-4
similar constructions. Once such a methodology is available, huge financial benefits can
be gained, for example by using the facility as a learning tool to improve pavement
design methods or to promote the incorporation of new nontraditional materials. The
present work offered a mechanistic approach to account for resilient (recoverable)
responses. However, the proposed method was developed and validated using only one
pavement type. As a first future step, this work should be reapplied, validated, and
refined (if found necessary) using other pavement systems1. At a later stage, the
analysis process should be extended to account for permanent (irrecoverable)
deformations2 and to account for cracking of the HMA layers. It is further proposed that
the research approach continue along the same lines as done here, i.e., building two
nominally identical instrumented pavement systems, one in the INDOT APT and
another in the field. The APT pavement should be thoroughly investigated to generate
predictions for field conditions; these are then assessed/validated by comparison with
field measurements.
It is important to emphasize that laboratory work, and advancement of our basic
understanding of how the individual materials behave, plays an essential role here in
that it is equally as important as the structural experiments for achieving the
aforementioned objective. Only through the availability of both high quality structural
data and high quality laboratory test results may it be possible to link construction
processes, material properties and pavement behavior.
In what follows, and based on the study findings, a list of specific
recommendations is offered for implementation in future studies involving INDOT’s
APT. The recommendations are separated into several topics: (i) embedded
instrumentation; (ii) APT testing program; (iii) structural behavior; (iv) data acquisition;
(v) laboratory work; and (vi) field project testing. Topics associated with new
1
At the time of writing this report, a new research work was undertaken by INDOT and Purdue
University (via the NCSC), involving the replication of a section of US31 in the APT facility. The
pavement structure is composed of 17 in. (432 mm) HMA over a granular subgrade of which the top 16
in. (406 mm) were stabilized with cement.
2
At the time of writing this report, a new research proposal was prepared, targeting the rutting behavior
of pavements constructed with low void mixtures and involving the NCAT phase III test cycle. Naturally,
this study would require addressing permanent deformations.
6-5
construction of APT test sections (e.g., organization, scheduling, gauge installations,
etc) are covered in Llenín and Pellinen (2004).
(i)
Embedded Instrumentation. Before being installed, all gauge types should be
checked for functionality, investigated for temperature sensitivity, and calibration
factors validated. There is a need to devise ways to calibrate the gauges after
installation in the pavement system given that their presence influences the free field
behavior. Consider calibrating the pressure cells (used for vertical stresses) using
single tire loading and comparing the ‘volume’ of stresses to the total applied load.
There is a need to find more dependable and accurate methods to measure permanent
stress and strain changes, and also moisture content changes. In addition to the gauge
types installed in this study, it may be beneficial to add multi depth deflectometers. An
attempt should be made to measure horizontal stresses in the different pavement layers
(especially in the unbound layers) originating from the construction processes and
later on due to trafficking. Additional stress and strain gauges should also be
embedded deeper in the subgrade for better characterization and support for advanced
modeling. A way to measure or estimate suction levels in the unbound materials
should be sought.
(ii)
APT Testing Program. Each experiment should be performed under at least
three different temperature levels within the available range, e.g., 60ºF (15.5ºC), 77ºF
(25.0ºC), and 95ºF (35ºC). When switching from one temperature to another no
loading should be applied until sufficient time (of about a week) has elapsed for the
conditions to stabilize while monitoring and recording instrumentation readings; doing
this will help to quantify the sensitivity of the instrumentation to temperature changes
which could potentially be used later to adjust the raw readings. At any given
temperature, APT passes should initially be applied using the super-single tire and
later using the dual-tire assembly. For each temperature, the loadings should be
executed at four different speeds within the available range, e.g., 0.05 mph (0.08
km/h), 0.50 mph (0.80 km/h), 1.0 mph (1.61 km/h), and 5.0 mph (8.0 km/h). Also, at
any given speed data should be collected at different load levels to study nonlinear
response with respect to load level, e.g., 5 kips (2,270 kg); 7.5 kips (3,400 kg); 15.0
kips (6,800 kg) and 20 kips (9,070 kg). Modification of the APT should be considered
6-6
so that two new loading modes can be applied to the surface of the pavement:
horizontal (shear) and turning (torsion); these will allow the study of intersection
conditions. An investigation to study the effects of various tire pressures should also
be targeted. In this connection, a way to measure the actual loading area and
distribution of stresses under the APT tires should be explored.
(iii) Structural Behavior. Completed structures in the APT should be tested using the
Dynamic Cone Penetrometer, but also using more advanced geotechnical equipment
such as the pressuremeter (cavity expansion). FWD testing should periodically be
performed while recording the entire time history of the loading and deflections. At
the same time, the responses captured by the embedded gauge array should also be
recorded. Vertical surface deflection during APT testing and also during FWD testing
ought to be monitored with external LVDTs. In addition, stand-alone accelerometers
and geophones should be attached to the pavement surface to supplement the LVDTs.
During all experiment types rutting measurements should be collected; more profiles
ought to be collected at the initial stages of the experiment where most of the rutting is
accumulated. The first profile measurement, taken before any passes are applied,
should be repeated several times for each cross section since these are used as
reference for all other measurements and hence need to be determined at a higher
accuracy level. A method should be found for measuring surface profiles such that the
loading system does not have to stop; this will facilitate the study of permanent
deformation development. A systematic way to detect and record cracks as they
appear (and when they appear) on the surface should be found; in this connection, a
method should be developed to ascertain whether the observed cracks are so-called
‘bottom up’ or ‘top down’. Friction testing in the APT should be routinely performed,
e.g., using the Circular Texture Meter (ASTM E-2157) and the Dynamic Friction
Tester (ASTM E-1911). This type of data may be used to investigate the effects of
temperature, wheel wander, tire type, and loading intensity on frictional attributes.
With additional friction data from the corresponding field study, a scheme may be
developed for using APT experiments to forecast friction performance.
(iv) Data Logging. All types of available data should continuously be recorded from
the very instant of gauge installation, throughout all testing modes, until the test
6-7
sections are removed. Data sampling rates should be varied based on the current
‘action’; e.g., use 100 scans per second for APT passes (including when the wheels are
lifted of the ground and returned to the startup location), use 5,000 scans per second
for FWD tests, and record data every 5 to 10 minutes when monitoring environmental
changes. The carriage position relative to the gauge array ought to be recorded at all
time; this information is very important for performing inverse analysis. In the present
study the APT carriage position was only recorded in the longitudinal direction. In
future studies, whenever wander is applied, lateral positions of the carriage should also
be recorded.
(v)
Laboratory Work. Standard tests such as soil classification, laboratory
compaction, HMA complex modulus, and resilient modulus for the unbound materials
should be performed. An attempt should be made to employ a large range of
conditions in these tests, e.g., execute the resilient modulus tests at different
compaction and moisture levels. In addition, capabilities to perform more advanced
tests should be developed; e.g., creep and recovery for the HMA in uniaxial (either
tension or compression) or isotropic conditions. In this connection, radial strain
measurements in all mechanical tests should be included. Whenever possible, test
specimens should be fabricated from samples cut or cored from the as-built pavement.
(vi) Field project testing. As discussed in Topic (i) above, field project should be
instrumented during construction by embedding gauges similar to the APT study or by
retrofitting an existing pavement with gauges (e.g., temperature probes and multi
depth deflectometers). Response data in the field may be collected only at certain
times during which the lane will be closed to traffic and a truck of known weight and
speed will drive over the gauge array. It is very important that the applied loading be
accurately positioned in space ( x , y , z ) and time ( t ) relative to the gauges. For
longitudinal positioning, a triggering device is suggested that can sense reflective
targets mounted on the truck; for transverse positioning it may be sufficient to mark
the pavement with lines spaced 1 to 2 in. (25 to 51 mm) apart and take high definition
video of the moving truck. Other tests should be conducted to investigate the overall
structural behavior, as recommended in Topic (iii) above.
6-8
REFERENCES
Al-Qadi, I. L. (2007), “True Viscoelastic Analyses of Pavement Structures to Validate
& Enhance Current Modulus Selection from Load Pulse Duration in NCHRP 1-37a
Design Guide,” obtained through private communication from the FHWA, TFHRC.
Andrei, D., Witczak, M. W., Schwartz, C. W., and Uzan, J. (2004), “Harmonized
Resilient Modulus Test Method for Unbound Pavement Materials,” Transportation
Research Record 1874, Journal of the Transportation Research Board, pp. 29-37.
ARA Inc. (2004), “Guide for Mechanistic-Empirical Design of New and Rehabilitated
Pavement Structures,” Applied Research Associates Inc., ERES Consultants Division,
National Cooperative Highway Research Program (NCHRP) Project 1-37A, Final
Report, Transportation Research Board, Washington, DC.
Austroads (2004), “Pavement Design - A Guide to the Structural Design of Road
Pavements,” Austroads Publication No. AP-G17/04, Sydney, Australia.
Barde, V. and Cardone, F. (2004), “Dynamic Modulus Testing of NCAT Mixtures,” CE
597 Course Report, submitted to Dr. Terhi Pellinen, School of Civil Engineering,
Purdue University.
Barden, L. (1963), “Stresses and Displacements in a Cross-anisotropic Soils,”
Géotechnique, Vol. 13 (3), pp. 198-210.
Brown, E. R., Cooley, L. A., Hanson, D., Lynn, C., Powell, B., Prowell, B., and
Watson, D. (2002), “NCAT Test Track Design, Construction, and Performance,”
National Center for Asphalt Technology, Report No. 02-12.
Brown, S. F. (1977), “State-of-the-Art Report on Field Instrumentation for Pavement
Experiments,” Transportation Research Record 640, Journal of the Transportation
Research Board, pp. 13-28.
Burmister, D. M. (1943), “The Theory of Stresses and Displacements in Layered
Systems and Application to the Design of Airport Runways,” Proceedings of the
Highway Research Board, Vol. 23, Washington, D.C., pp. 126-148.
Burmister, D. M. (1945), “The General Theory of Stresses and Displacements in
Layered Systems,” Journal of Applied Physics, Vol. 16, pp. 89-94 (Part I); pp. 126-127
(Part II); pp. 296-302 (Part III).
Christian, J. T. (1968), “Undrained Stress Distribution by Numerical Methods,” ASCE
Journal of the Soil Mechanics and Foundation Engineering Division, Vol. 94 (6), pp.
1335-1345.
7-1
De Jong, D. L., Peatz, M. G. F., and Korswagen, A. R. (1973), “Computer Program
Bisar Layered Systems Under Normal and Tangential Loads,” Konin Klijke ShellLaboratorium, Amsterdam, External Report AMSR.0006.73.
Di Benedetto, H., Delaporte, B., and Sauzeat, C. (2007), “Three-Dimensional Linear
Behavior of Bituminous Materials: Experiments and Modeling, ” International Journal
of Geomechanics, Vol. 7(2), pp 149-157.
Duncan, J. M., Williams, G. W., Sehn, A. L., and Seed, R. B. (1991), “Estimation of
Earth Pressures due to Compaction,” Journal of Geotechnical Engineering, Vol. 117
(12), pp. 1833-1847.
Dunnicliff, J. (1988), “Geotechnical Instrumentation
Performance,” John Wiley and Sons, Inc., New York.
for
Monitoring
Field
Elseifi, M. A., Al-Qadi, I. L., and Yoo, P. J. (2006), “Viscoelastic Modeling and Field
Validation of Flexible Pavements,” Journal of Engineering Mechanics, Vol. 132 (2), pp.
172-178.
Freeman, R. B., Tommy, C. H, McEwen, T., and Powell, R. B. (2001), “Instrumentation
at the National Center for Asphalt Technology Test Track,” U.S. Army Corps of
Engineers, Engineer Research and Development Center, Report No. ERDC TR-01-9.
Fylstra, D., Lasdon, L., Watson, J., and Waren, A. (1998), “Design and Use of the
Microsoft Excel Solver,” Interfaces, Vol. 28(5), pp. 29-55.
Galal, K. A. and White, T. D. (1999), “INDOT-APT Test Facility Experience,” Paper
CS8-4, presented at the International Conference on Accelerated Pavement Testing,
Reno, Nevada, October 18-20.
Galal, K. A., White T. D., and Reck, C. (1998), “Accelerated Pavement Testing
Facility,” INDOT - Division of Research report, West Lafayette, Indiana.
Graham, J., and Houlsby, G. T. (1983), “Anisotropic Elasticity of a Natural Clay,”
Géotechnique, Vol. 33 (2), pp. 165-180.
Huang, H. (1995), “Analysis of Accelerated Pavement Tests and Finite Element
Modeling of Rutting Phenomenon,” Ph.D. dissertation, Purdue university.
Huang, Y. H. (2004), “Pavement Analysis and Design,” Second Edition, Pearson
Prentice Hall, p. 775.
Hufferd, W. L., and Lai, J. S. (1978), “Analysis of N-layered Viscoelastic Pavement
Systems,” Report FHWA-RD-78-82, Federal Highway Administration, p. 220.
Labuz, J. F., and Theroux, B. (2005), “Laboratory Calibration of Earth Pressure Cells,”
Geotechnical Testing Journal, Vol. 28 (2).
7-2
Lakes, R. S. (1998), Viscoelastic Solids, CRC Press (Boca Raton, FL), p. 476.
Lekhnitskii, S. G. (1963), “Theory of Elasticity of an Anisotropic Elastic Body, HoldenDay, San Francisco, p. 404.
Levenberg, E. (2006), “Constitutive Modeling of Asphalt-Aggregate Mixes with
Damage and Healing,” Ph.D. Dissertation, Technion - Israel Institute of Technology.
Levenberg, E., and Shah, A. (2008), “Interpretation of Complex Modulus Test Results
for Asphalt-Aggregate Mixes,” ASTM Journal of Testing and Evaluation, Vol. 36 (4).
Llenín, J. A., and Pellinen, T. K. (2004), “Validation of NCAT Structural Test Track
Experiment using INDOT APT Facility,” Interim Draft Final Report, Joint
Transportation Research Program, SPR 2813 Project, Purdue University.
Llenín, J. A., Pellinen, T. K., and Abraham, D. M. (2006), “Construction Management
of a Small-Scale Accelerated Pavement Testing Facility,” Journal of Performance of
Constructed Facilities, Vol. 20 (3), pp. 229-236.
Locket, F. J. (1972), Nonlinear Viscoelastic Solids, Academic Press Inc. (London), p.
195.
Love, A. E. H. (1923), “Treatise on the Mathematical Theory of Elasticity”, Cambridge
University Press, UK.
Lytton, R. L., Uzan, J., Fernando, E. G., Roque, R., Hiltunen, D., and Stoffels, S. M.
(1993), “Development and Validation of Performance Prediction Models and
Specifications for Asphalt Binders and Paving Mixes,” Report SHRP A-357, Strategic
Highway Research Program, National Research Council, Washington, D.C.
Monfore, G. E. (1950), “An Analysis of the Stress Distribution in and near Stress
Gauges Embedded in Elastic Soils,” Structural Laboratory Report No. SP 26, U.S.
Bureau of Reclamation, Denver, CO.
NCHRP (2002), “Standard Test Method for Dynamic Modulus of Asphalt Concrete
Mixtures”, Provisional Test Method DM-1, Project 1-37A, Arizona State University.
Oda, M., Nemat-Nasser, S., and Konish, J. (1985), “Stress-induced Anisotropy in
Granular Masses,” Journal of Soils and Foundation, Vol. 25 (3), pp. 85-97.
Peattie, K. R., and Sparrow, R. W. (1954), “The Fundamental Action of Earth Pressure
Cells,” Journal of the Mechanics and Physics of Solid, Vol. 2, pp. 141-155.
Pipkin, A. C. (1972), Lectures on Viscoelasticity Theory, Applied Mathematical
Sciences Vol. 7, Springer-Verlag Inc. (New York), pp. 180.
Plazek, D. J. (1996), “Oh, Thermorheological Simplicity, Wherefore Art Thou?,” 1995
Bingham Medal Address, Journal of Rheology, Vol. 40 (6), pp. 987-1014.
7-3
Poulos, H. G., and Davis, E. H. (1974), “Elastic Solutions for Soil and Rock Mechanics,
Center for Geotechnical Research,” John Wiley & Sons, Inc. (reprinted and corrected
1991).
Powell, R. B., and Brown, E. R. (2004), “Construction of the 2003 NCAT Pavement
Test Track,” National Center for Asphalt Technology, Draft Report.
Priest, A. L., and Timm, D. H. (2006), “Methodology and Calibration of Fatigue
Transfer Functions for Mechanistic-Empirical Flexible Pavement Design,” National
Center for Asphalt Technology, Report 06-03.
Priest, A. L., Timm, D. H., and Barrett, W. E. (2005), “Mechanistic Comparison of
Wide-base Single vs. Standard Dual Tire Configurations,” National Center for Asphalt
Technology, Report 05-03.
Saadeh, S., Tashman, L., Masad, E., and Mogawer, W. (2002), “Spatial and Directional
Distribution of Aggregates in Asphalt Mixes,” ASTM Journal of Testing and
Evaluation, Vol. 30 (6).
Schapery, R. A. (1962), “Approximate Methods of Transform Inversion for Viscoelastic
Stress Analysis,” Proceedings of the 4th U.S. National Congress on Applied Mechanics,
Vol. 2, ASME, pp. 1075-1085.
Schapery, R. A. (1965), “A Method of Viscoelastic Stress Analysis using Elastic
Solutions,” Journal of the Franklin Institute, Vol. 279 (4), pp. 268-289.
Schapery, R. A. (1974), “Viscoelastic Behavior and Analysis of Composite Materials, ”
Mechanics of Composite Materials, Vol. 2, pp. 85-168, edited by G. P. Sendeckyj,
Academic Press (New York).
Schwarzl, F., and Staverman, A. J. (1952), “Time-Temperature Dependence of Linear
Viscoelastic Behavior,” Journal of Applied Physics, Vol. 23 (8), pp. 838-843.
Shields, D. H., Zeng, M., and Kwok, R. (1998), “Nonlinear Viscoelastic Behavior of
Asphalt Concrete in Stress Relaxation,” Journal of the Association of Asphalt Pavement
Technologists, Vol. 67.
Singh, S. J. (1986), “Static Deformation of a Transversely Isotropic Multilayered Halfspace by Surface Loads,” Physics of the Earth and Planetary Interiors, Vol. 42 (4), pp.
263-273.
Sugihara, M. (1987), “Methods of Numerical Integration of Oscillatory Functions by
the DE-formula with Richardson Extrapolation,” Journal of Computational and Applied
Mathematics, Vol. 17, pp. 47-68.
Tabatabaee, N., and Sebaaly, P. (1990), “State-of-the-Art Pavement Instrumentation,”
Transportation Research Record 1260, Journal of the Transportation Research Board,
pp. 246-255.
7-4
Taylor, D. W., 1945, “Review of Pressure Distribution Theories, Earth Pressure Cell
Investigations, and Pressure Distribution Data,” Contract Report W22-053 eng-185,
U.S. Army Engineer Waterways Experiment Station, Vicksburg, MS.
Tesarik, D. R., Seymour, J. B., Williams, T. J., Martin, L. A., and Jones, F. M. (2006),
“Temperature Corrections to Earth Pressure Cells Embedded in Cemented Backfill,”
Report of Investigations No. 9665, U.S. Department of Health and Human Services,
Centers for Disease Control and Prevention, National Institute for Occupational Safety
and Health, Spokane Research Laboratory, Spokane, WA.
Theroux, B., Labuz, J. F., and Dai, S. (2001), “Field Installation of an Earth Pressure
Cell,” Transportation Research Record 1772, Journal of the Transportation Research
Board, pp. 12-19.
Timm, D. H., and Priest, A. L. (2006), “Material Properties of the 2003 NCAT Test
Track Structural Study,” National Center for Asphalt Technology, Report No. 06-01.
Timm, D. H., Priest, A. L., and McEwen, T. V. (2004), “Design and Instrumentation of
the Structural Pavement Experiment at the NCAT Test Track,” National Center for
Asphalt Technology, Report No. 04-01.
Timm, D. H., West, R. C., Priest, A. L., Powell, B., Selvaraj, I., Zhang, J., and Brown,
E. R. (2006), “Phase II NCAT Test Track Results,” National Center for Asphalt
Technology, Report 06-05.
Tory, A. C., and Sparrow, R. W. (1967), “The Influence of Diaphragm Flexibility on
the Performance of an Earth Pressure Cell,” Journal of Scientific Instruments, Vol. 44,
pp. 781-785.
Tutumluer, E., and Thompson, M. R. (1997), “Anisotropic Modeling of Granular Bases
in Flexible Pavements,” Transportation Research Record 1557, Journal of the
Transportation Research Board, pp. 18-26.
Uzan, J. (1976), “The Influence of the Interface Condition on the Stress Distribution in
a Layered System,” Transportation Research Record 616, Journal of the Transportation
Research Board, pp. 71-73.
Uzan, J. (1985), “Characterization of Granular Material,” Transportation Research
Record 1022, Journal of the Transportation Research Board, pp. 52-59.
Uzan J. (1992), “Resilient Characterization of Pavement Materials,” International
Journal of Numerical and Analytical Methods in Geomechanics, Vol. 16, pp. 453-459.
Uzan, J., and Levenberg, E. (2007), “Advanced Testing and Characterization of Asphalt
Concrete Materials in Tension,” International Journal of Geomechanics, Vol. 7 (2), pp.
158-165.
7-5
von Karman, T., and Biot, M. A. (1940), Mathematical Methods in Engineering: An
Introduction to the Mathematical Treatment of Engineering Problems, McGraw-Hill
Inc., New York, p. 505.
Weiler, W. A., and Kulhawy, F. H. (1982), “Factors Affecting Stress Cell
Measurements in Soil,” ASCE Journal of the Geotechnical and Foundation Division,
Vol. 108 (GT12), pp. 1529-1548.
White, T. D., Albers, J. M., and Haddock, J. E. (1990), “An Accelerated Testing System
to Determine Percent Crushed Aggregate Requirements in Bituminous Mixtures - Final
Report,” Joint Transportation Research Program, FHWA/IN/JTRP-90-8, Purdue
University.
Williams, M. L., Landel, R. F., and Ferry, J. D. (1955), “The Temperature Dependence
of Relaxation Mechanisms in Amorphous Polymers and Other Glass-forming Liquids,”
Journal of the American Chemical Society, Vol. 77, pp. 3701-3707.
Witczak, M. W., and Uzan, J. (1988), “The Universal Airport Design System, Report I
of IV: Granular Material Characterization,” Department of Civil Engineering,
University of Maryland, College Park.
Wolf, K. (1935), “Distribution of Stress in a Half-plane and a Half-space of Anisotropic
Material, “ ZAMM - Journal of Applied Mathematics and Mechanics, Vol. 15 (5), pp.
249-254.
7-6
APPENDICES (DVD UPON REQUEST)
‘Appendix_A’, ‘Appendix_B’ and ‘Appendix_C’ are available on a DVD from the
JTRP Office upon request by e-mail jtrp@ecn.purdue.edu or call 765-494-9310. Each
folder contains the associated appendix material. A short description of each is hereafter
provided.
APPENDIX A: RAW EXPERIMENTAL DATA
The ‘Appendix_A’ folder on the DVD contains the raw experimental data collected
during this entire study. Three subfolders are included, named: ‘A1_APT Experiment’,
‘A2_Common Data’, and ‘A3_NCAT Experiment’.
The A1 Subfolder contains information related to the APT study. It consists of
the following six folders: (i) A11_ RTR Files. These are text files that include
information recorded by the PC running the APT carriage. An explanation on how to
read these RTR files can be found in the PDF document placed in the same folder; (ii)
A12_Moisture Data. These files contain the recorded moisture data. A viewer program
is needed (also provided); (iii) A13_StrainSmart Data. These files contain the stress,
strain and temperature data. The write protected program ‘StrainSmart’ is required to
read these files (not included); (iv) A14_Pictures. Includes the pictures and video clips
taken during the APT experiment; (v) A15_Rutting Profiles. All rutting measurements
done in the APT are included, sorted according to measurement date. Indexes 1 and 2 in
the rutting file names refer to sections n1 and n2 before rehabilitation (respectively)
while indexes 3 and 4 refer to test sections n1 and n2 after rehabilitation (respectively);
and (vi) A16_FWD. Includes FWD files that consist of peak deflection data (and
corresponding HMA temperatures.
The A2 Subfolder contains information that is common to both the APT and
NCAT experiments. This subfolder consists of the following three folders, having self
explanatory names: (i) A21_Beam Fatigue; (ii) A22_Complex Modulus; and (iii)
A23_Presentations.
The A3 Subfolder contains information related to the NCAT study. It consists of
nine folders (again, self explanatory names): (i) A31_Pictures; (ii) A32_DCPT&CBR;
8-1
(iii) A33_FWD; (iv) A34_Reports&Presentations; (v) A35_HMA; (vi) A36_Resilient
Modulus; (vii) A37_Response&Performance; (viii) A38_Spray Applications; and (ix)
A39_Design&Spec.
APPENDIX B: COMPUTER PROGRAMS
The ‘Appendix_B’ folder on the DVD contains two structural analysis programs
developed under this study: (i) ‘ELLEA1’ which is based on isotropic LET (see Chapter
4); and (ii) ‘ELLEA2’ which is based on anisotropic LET (see Chapter 5). These
programs are built into Excel workbooks for real-time computations and ease of use.
Two versions are included for each program: (i) an Excel 2003 version identified by a
file name extension .xls; and (ii) an Excel 2007 version identified by a file name
extension .xlsx.
In order for these programs to run correctly it is important to enable the Analysis
ToolPak, an Add-In normally included in Excel, which sometimes needs manual
installation (a one time event). It is important to note that both structural analysis
programs do not contain any VBA code so that any Macro related virus warning can be
ignored. If you experience any trouble or have any additional questions or requests,
contact me through this Email address: Eyal.Levenberg@yahoo.com.
APPENDIX C: REPORTS AND CORRESPONDANCE
The ‘Appendix_C’ folder on the DVD contains previous reports and correspondence
related to the APT project (PDF format). These are chronologically ordered, each
placed in a separate folder. This report is also included so that it can be reviewed or
reproduced with colors if desired. Note that some of the PDF files are scanned versions
of the printed originals; as such they are relatively large in size.
The included titles are as follows: (i) C1_Research Proposal (Pellinen and Galal,
May 2003); (ii) C2_ Interim Instrumentation Plan Report (Llenín, April 2004); (iii)
C3_Interim Materials Shipping Plan Report (Llenín, April 2004); (iv) C4_APT
Construction Specifications (Llenín and Pellinen, May_2004); (v) C5_Instrumentation
Installation Procedure Report (Llenín, May 2004); (vi) C6_APT First Construction
Cycle Report (Llenín and Pellinen, August 2004); (vii) C7_Feasibility Analysis (Llenín
8-2
and Pellinen, November 2004); (viii) C8_Dynamic Modulus Testing of NCAT Mixes
(Barde and Cardone, December 2004); (ix) C9_Interim Draft Final Report (Llenín and
Pellinen, December 2004); (x) C10_APT Instrumentation and Loading Experiments
(Pellinen and Webster, January 2005); (xi) C11_Preparation of Beams for SPR-2813
(Pellinen, Webster and Brower, February 2005); (xii) C12_Advanced Analysis of Beam
Fatigue Test Results of NCAT mixes (Agrawal, August 2005); (xiii) C13_Laboratory
Fatige Testing of HMA (Webster and Pellinen, May 2005); (xiv) C14_Quarterly
Progress Report 2005 (Pellinen and Nantung, December 2005); and (xv) C15_This
(Final) Report (Levenberg, 2008).
8-3